xref: /linux/net/ipv6/datagram.c (revision 1fc31357ad194fb98691f3d122bcd47e59239e83)
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 	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_dst_store(sk, dst,
110 		      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
111 		      &sk->sk_v6_daddr : NULL,
112 #ifdef CONFIG_IPV6_SUBTREES
113 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
114 		      &np->saddr :
115 #endif
116 		      NULL);
117 
118 out:
119 	fl6_sock_release(flowlabel);
120 	return err;
121 }
122 
123 void ip6_datagram_release_cb(struct sock *sk)
124 {
125 	struct dst_entry *dst;
126 
127 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
128 		return;
129 
130 	rcu_read_lock();
131 	dst = __sk_dst_get(sk);
132 	if (!dst || !dst->obsolete ||
133 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
134 		rcu_read_unlock();
135 		return;
136 	}
137 	rcu_read_unlock();
138 
139 	ip6_datagram_dst_update(sk, false);
140 }
141 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
142 
143 static int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, 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 
257 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
258 {
259 	int res;
260 
261 	lock_sock(sk);
262 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
263 	release_sock(sk);
264 	return res;
265 }
266 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
267 
268 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
269 				 int addr_len)
270 {
271 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
272 	if (sin6->sin6_family != AF_INET6)
273 		return -EAFNOSUPPORT;
274 	return ip6_datagram_connect(sk, uaddr, addr_len);
275 }
276 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
277 
278 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
279 		     __be16 port, u32 info, u8 *payload)
280 {
281 	struct ipv6_pinfo *np  = inet6_sk(sk);
282 	struct icmp6hdr *icmph = icmp6_hdr(skb);
283 	struct sock_exterr_skb *serr;
284 
285 	if (!np->recverr)
286 		return;
287 
288 	skb = skb_clone(skb, GFP_ATOMIC);
289 	if (!skb)
290 		return;
291 
292 	skb->protocol = htons(ETH_P_IPV6);
293 
294 	serr = SKB_EXT_ERR(skb);
295 	serr->ee.ee_errno = err;
296 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
297 	serr->ee.ee_type = icmph->icmp6_type;
298 	serr->ee.ee_code = icmph->icmp6_code;
299 	serr->ee.ee_pad = 0;
300 	serr->ee.ee_info = info;
301 	serr->ee.ee_data = 0;
302 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
303 				  skb_network_header(skb);
304 	serr->port = port;
305 
306 	__skb_pull(skb, payload - skb->data);
307 	skb_reset_transport_header(skb);
308 
309 	if (sock_queue_err_skb(sk, skb))
310 		kfree_skb(skb);
311 }
312 
313 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
314 {
315 	const struct ipv6_pinfo *np = inet6_sk(sk);
316 	struct sock_exterr_skb *serr;
317 	struct ipv6hdr *iph;
318 	struct sk_buff *skb;
319 
320 	if (!np->recverr)
321 		return;
322 
323 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
324 	if (!skb)
325 		return;
326 
327 	skb->protocol = htons(ETH_P_IPV6);
328 
329 	skb_put(skb, sizeof(struct ipv6hdr));
330 	skb_reset_network_header(skb);
331 	iph = ipv6_hdr(skb);
332 	iph->daddr = fl6->daddr;
333 
334 	serr = SKB_EXT_ERR(skb);
335 	serr->ee.ee_errno = err;
336 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
337 	serr->ee.ee_type = 0;
338 	serr->ee.ee_code = 0;
339 	serr->ee.ee_pad = 0;
340 	serr->ee.ee_info = info;
341 	serr->ee.ee_data = 0;
342 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
343 	serr->port = fl6->fl6_dport;
344 
345 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
346 	skb_reset_transport_header(skb);
347 
348 	if (sock_queue_err_skb(sk, skb))
349 		kfree_skb(skb);
350 }
351 
352 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
353 {
354 	struct ipv6_pinfo *np = inet6_sk(sk);
355 	struct ipv6hdr *iph;
356 	struct sk_buff *skb;
357 	struct ip6_mtuinfo *mtu_info;
358 
359 	if (!np->rxopt.bits.rxpmtu)
360 		return;
361 
362 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
363 	if (!skb)
364 		return;
365 
366 	skb_put(skb, sizeof(struct ipv6hdr));
367 	skb_reset_network_header(skb);
368 	iph = ipv6_hdr(skb);
369 	iph->daddr = fl6->daddr;
370 
371 	mtu_info = IP6CBMTU(skb);
372 
373 	mtu_info->ip6m_mtu = mtu;
374 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
375 	mtu_info->ip6m_addr.sin6_port = 0;
376 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
377 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
378 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
379 
380 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
381 	skb_reset_transport_header(skb);
382 
383 	skb = xchg(&np->rxpmtu, skb);
384 	kfree_skb(skb);
385 }
386 
387 /* For some errors we have valid addr_offset even with zero payload and
388  * zero port. Also, addr_offset should be supported if port is set.
389  */
390 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
391 {
392 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
393 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
394 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
395 }
396 
397 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
398  *
399  * At one point, excluding local errors was a quick test to identify icmp/icmp6
400  * errors. This is no longer true, but the test remained, so the v6 stack,
401  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
402  *
403  * Timestamp code paths do not initialize the fields expected by cmsg:
404  * the PKTINFO fields in skb->cb[]. Fill those in here.
405  */
406 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
407 				      struct sock_exterr_skb *serr)
408 {
409 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
410 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
411 		return true;
412 
413 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
414 		return false;
415 
416 	if (!skb->dev)
417 		return false;
418 
419 	if (skb->protocol == htons(ETH_P_IPV6))
420 		IP6CB(skb)->iif = skb->dev->ifindex;
421 	else
422 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
423 
424 	return true;
425 }
426 
427 /*
428  *	Handle MSG_ERRQUEUE
429  */
430 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
431 {
432 	struct ipv6_pinfo *np = inet6_sk(sk);
433 	struct sock_exterr_skb *serr;
434 	struct sk_buff *skb;
435 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
436 	struct {
437 		struct sock_extended_err ee;
438 		struct sockaddr_in6	 offender;
439 	} errhdr;
440 	int err;
441 	int copied;
442 
443 	err = -EAGAIN;
444 	skb = sock_dequeue_err_skb(sk);
445 	if (!skb)
446 		goto out;
447 
448 	copied = skb->len;
449 	if (copied > len) {
450 		msg->msg_flags |= MSG_TRUNC;
451 		copied = len;
452 	}
453 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
454 	if (unlikely(err)) {
455 		kfree_skb(skb);
456 		return err;
457 	}
458 	sock_recv_timestamp(msg, sk, skb);
459 
460 	serr = SKB_EXT_ERR(skb);
461 
462 	if (sin && ipv6_datagram_support_addr(serr)) {
463 		const unsigned char *nh = skb_network_header(skb);
464 		sin->sin6_family = AF_INET6;
465 		sin->sin6_flowinfo = 0;
466 		sin->sin6_port = serr->port;
467 		if (skb->protocol == htons(ETH_P_IPV6)) {
468 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
469 								  struct ipv6hdr, daddr);
470 			sin->sin6_addr = ip6h->daddr;
471 			if (np->sndflow)
472 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
473 			sin->sin6_scope_id =
474 				ipv6_iface_scope_id(&sin->sin6_addr,
475 						    IP6CB(skb)->iif);
476 		} else {
477 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
478 					       &sin->sin6_addr);
479 			sin->sin6_scope_id = 0;
480 		}
481 		*addr_len = sizeof(*sin);
482 	}
483 
484 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
485 	sin = &errhdr.offender;
486 	memset(sin, 0, sizeof(*sin));
487 
488 	if (ip6_datagram_support_cmsg(skb, serr)) {
489 		sin->sin6_family = AF_INET6;
490 		if (np->rxopt.all)
491 			ip6_datagram_recv_common_ctl(sk, msg, skb);
492 		if (skb->protocol == htons(ETH_P_IPV6)) {
493 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
494 			if (np->rxopt.all)
495 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
496 			sin->sin6_scope_id =
497 				ipv6_iface_scope_id(&sin->sin6_addr,
498 						    IP6CB(skb)->iif);
499 		} else {
500 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
501 					       &sin->sin6_addr);
502 			if (inet_sk(sk)->cmsg_flags)
503 				ip_cmsg_recv(msg, skb);
504 		}
505 	}
506 
507 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
508 
509 	/* Now we could try to dump offended packet options */
510 
511 	msg->msg_flags |= MSG_ERRQUEUE;
512 	err = copied;
513 
514 	consume_skb(skb);
515 out:
516 	return err;
517 }
518 EXPORT_SYMBOL_GPL(ipv6_recv_error);
519 
520 /*
521  *	Handle IPV6_RECVPATHMTU
522  */
523 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
524 		     int *addr_len)
525 {
526 	struct ipv6_pinfo *np = inet6_sk(sk);
527 	struct sk_buff *skb;
528 	struct ip6_mtuinfo mtu_info;
529 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
530 	int err;
531 	int copied;
532 
533 	err = -EAGAIN;
534 	skb = xchg(&np->rxpmtu, NULL);
535 	if (!skb)
536 		goto out;
537 
538 	copied = skb->len;
539 	if (copied > len) {
540 		msg->msg_flags |= MSG_TRUNC;
541 		copied = len;
542 	}
543 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
544 	if (err)
545 		goto out_free_skb;
546 
547 	sock_recv_timestamp(msg, sk, skb);
548 
549 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
550 
551 	if (sin) {
552 		sin->sin6_family = AF_INET6;
553 		sin->sin6_flowinfo = 0;
554 		sin->sin6_port = 0;
555 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
556 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
557 		*addr_len = sizeof(*sin);
558 	}
559 
560 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
561 
562 	err = copied;
563 
564 out_free_skb:
565 	kfree_skb(skb);
566 out:
567 	return err;
568 }
569 
570 
571 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
572 				 struct sk_buff *skb)
573 {
574 	struct ipv6_pinfo *np = inet6_sk(sk);
575 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
576 
577 	if (np->rxopt.bits.rxinfo) {
578 		struct in6_pktinfo src_info;
579 
580 		if (is_ipv6) {
581 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
582 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
583 		} else {
584 			src_info.ipi6_ifindex =
585 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
586 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
587 					       &src_info.ipi6_addr);
588 		}
589 
590 		if (src_info.ipi6_ifindex >= 0)
591 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
592 				 sizeof(src_info), &src_info);
593 	}
594 }
595 
596 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
597 				    struct sk_buff *skb)
598 {
599 	struct ipv6_pinfo *np = inet6_sk(sk);
600 	struct inet6_skb_parm *opt = IP6CB(skb);
601 	unsigned char *nh = skb_network_header(skb);
602 
603 	if (np->rxopt.bits.rxhlim) {
604 		int hlim = ipv6_hdr(skb)->hop_limit;
605 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
606 	}
607 
608 	if (np->rxopt.bits.rxtclass) {
609 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
610 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
611 	}
612 
613 	if (np->rxopt.bits.rxflow) {
614 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
615 		if (flowinfo)
616 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
617 	}
618 
619 	/* HbH is allowed only once */
620 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
621 		u8 *ptr = nh + sizeof(struct ipv6hdr);
622 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
623 	}
624 
625 	if (opt->lastopt &&
626 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
627 		/*
628 		 * Silly enough, but we need to reparse in order to
629 		 * report extension headers (except for HbH)
630 		 * in order.
631 		 *
632 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
633 		 * (and WILL NOT be) defined because
634 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
635 		 */
636 		unsigned int off = sizeof(struct ipv6hdr);
637 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
638 
639 		while (off <= opt->lastopt) {
640 			unsigned int len;
641 			u8 *ptr = nh + off;
642 
643 			switch (nexthdr) {
644 			case IPPROTO_DSTOPTS:
645 				nexthdr = ptr[0];
646 				len = (ptr[1] + 1) << 3;
647 				if (np->rxopt.bits.dstopts)
648 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
649 				break;
650 			case IPPROTO_ROUTING:
651 				nexthdr = ptr[0];
652 				len = (ptr[1] + 1) << 3;
653 				if (np->rxopt.bits.srcrt)
654 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
655 				break;
656 			case IPPROTO_AH:
657 				nexthdr = ptr[0];
658 				len = (ptr[1] + 2) << 2;
659 				break;
660 			default:
661 				nexthdr = ptr[0];
662 				len = (ptr[1] + 1) << 3;
663 				break;
664 			}
665 
666 			off += len;
667 		}
668 	}
669 
670 	/* socket options in old style */
671 	if (np->rxopt.bits.rxoinfo) {
672 		struct in6_pktinfo src_info;
673 
674 		src_info.ipi6_ifindex = opt->iif;
675 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
676 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
677 	}
678 	if (np->rxopt.bits.rxohlim) {
679 		int hlim = ipv6_hdr(skb)->hop_limit;
680 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
681 	}
682 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
683 		u8 *ptr = nh + sizeof(struct ipv6hdr);
684 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
685 	}
686 	if (np->rxopt.bits.odstopts && opt->dst0) {
687 		u8 *ptr = nh + opt->dst0;
688 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
689 	}
690 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
691 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
692 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
693 	}
694 	if (np->rxopt.bits.odstopts && opt->dst1) {
695 		u8 *ptr = nh + opt->dst1;
696 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
697 	}
698 	if (np->rxopt.bits.rxorigdstaddr) {
699 		struct sockaddr_in6 sin6;
700 		__be16 *ports = (__be16 *) skb_transport_header(skb);
701 
702 		if (skb_transport_offset(skb) + 4 <= skb->len) {
703 			/* All current transport protocols have the port numbers in the
704 			 * first four bytes of the transport header and this function is
705 			 * written with this assumption in mind.
706 			 */
707 
708 			sin6.sin6_family = AF_INET6;
709 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
710 			sin6.sin6_port = ports[1];
711 			sin6.sin6_flowinfo = 0;
712 			sin6.sin6_scope_id =
713 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
714 						    opt->iif);
715 
716 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
717 		}
718 	}
719 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
720 		int val = opt->frag_max_size;
721 
722 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
723 	}
724 }
725 
726 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
727 			  struct sk_buff *skb)
728 {
729 	ip6_datagram_recv_common_ctl(sk, msg, skb);
730 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
731 }
732 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
733 
734 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
735 			  struct msghdr *msg, struct flowi6 *fl6,
736 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
737 {
738 	struct in6_pktinfo *src_info;
739 	struct cmsghdr *cmsg;
740 	struct ipv6_rt_hdr *rthdr;
741 	struct ipv6_opt_hdr *hdr;
742 	struct ipv6_txoptions *opt = ipc6->opt;
743 	int len;
744 	int err = 0;
745 
746 	for_each_cmsghdr(cmsg, msg) {
747 		int addr_type;
748 
749 		if (!CMSG_OK(msg, cmsg)) {
750 			err = -EINVAL;
751 			goto exit_f;
752 		}
753 
754 		if (cmsg->cmsg_level == SOL_SOCKET) {
755 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
756 			if (err)
757 				return err;
758 			continue;
759 		}
760 
761 		if (cmsg->cmsg_level != SOL_IPV6)
762 			continue;
763 
764 		switch (cmsg->cmsg_type) {
765 		case IPV6_PKTINFO:
766 		case IPV6_2292PKTINFO:
767 		    {
768 			struct net_device *dev = NULL;
769 
770 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
771 				err = -EINVAL;
772 				goto exit_f;
773 			}
774 
775 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
776 
777 			if (src_info->ipi6_ifindex) {
778 				if (fl6->flowi6_oif &&
779 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
780 					return -EINVAL;
781 				fl6->flowi6_oif = src_info->ipi6_ifindex;
782 			}
783 
784 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
785 
786 			rcu_read_lock();
787 			if (fl6->flowi6_oif) {
788 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
789 				if (!dev) {
790 					rcu_read_unlock();
791 					return -ENODEV;
792 				}
793 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
794 				rcu_read_unlock();
795 				return -EINVAL;
796 			}
797 
798 			if (addr_type != IPV6_ADDR_ANY) {
799 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
800 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
801 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
802 						   strict ? dev : NULL, 0) &&
803 				    !ipv6_chk_acast_addr_src(net, dev,
804 							     &src_info->ipi6_addr))
805 					err = -EINVAL;
806 				else
807 					fl6->saddr = src_info->ipi6_addr;
808 			}
809 
810 			rcu_read_unlock();
811 
812 			if (err)
813 				goto exit_f;
814 
815 			break;
816 		    }
817 
818 		case IPV6_FLOWINFO:
819 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
820 				err = -EINVAL;
821 				goto exit_f;
822 			}
823 
824 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
825 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
826 					err = -EINVAL;
827 					goto exit_f;
828 				}
829 			}
830 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
831 			break;
832 
833 		case IPV6_2292HOPOPTS:
834 		case IPV6_HOPOPTS:
835 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
836 				err = -EINVAL;
837 				goto exit_f;
838 			}
839 
840 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
841 			len = ((hdr->hdrlen + 1) << 3);
842 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
843 				err = -EINVAL;
844 				goto exit_f;
845 			}
846 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
847 				err = -EPERM;
848 				goto exit_f;
849 			}
850 			opt->opt_nflen += len;
851 			opt->hopopt = hdr;
852 			break;
853 
854 		case IPV6_2292DSTOPTS:
855 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
856 				err = -EINVAL;
857 				goto exit_f;
858 			}
859 
860 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
861 			len = ((hdr->hdrlen + 1) << 3);
862 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
863 				err = -EINVAL;
864 				goto exit_f;
865 			}
866 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
867 				err = -EPERM;
868 				goto exit_f;
869 			}
870 			if (opt->dst1opt) {
871 				err = -EINVAL;
872 				goto exit_f;
873 			}
874 			opt->opt_flen += len;
875 			opt->dst1opt = hdr;
876 			break;
877 
878 		case IPV6_DSTOPTS:
879 		case IPV6_RTHDRDSTOPTS:
880 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
881 				err = -EINVAL;
882 				goto exit_f;
883 			}
884 
885 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
886 			len = ((hdr->hdrlen + 1) << 3);
887 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
888 				err = -EINVAL;
889 				goto exit_f;
890 			}
891 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
892 				err = -EPERM;
893 				goto exit_f;
894 			}
895 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
896 				opt->opt_flen += len;
897 				opt->dst1opt = hdr;
898 			} else {
899 				opt->opt_nflen += len;
900 				opt->dst0opt = hdr;
901 			}
902 			break;
903 
904 		case IPV6_2292RTHDR:
905 		case IPV6_RTHDR:
906 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
907 				err = -EINVAL;
908 				goto exit_f;
909 			}
910 
911 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
912 
913 			switch (rthdr->type) {
914 #if IS_ENABLED(CONFIG_IPV6_MIP6)
915 			case IPV6_SRCRT_TYPE_2:
916 				if (rthdr->hdrlen != 2 ||
917 				    rthdr->segments_left != 1) {
918 					err = -EINVAL;
919 					goto exit_f;
920 				}
921 				break;
922 #endif
923 			default:
924 				err = -EINVAL;
925 				goto exit_f;
926 			}
927 
928 			len = ((rthdr->hdrlen + 1) << 3);
929 
930 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
931 				err = -EINVAL;
932 				goto exit_f;
933 			}
934 
935 			/* segments left must also match */
936 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
937 				err = -EINVAL;
938 				goto exit_f;
939 			}
940 
941 			opt->opt_nflen += len;
942 			opt->srcrt = rthdr;
943 
944 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
945 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
946 
947 				opt->opt_nflen += dsthdrlen;
948 				opt->dst0opt = opt->dst1opt;
949 				opt->dst1opt = NULL;
950 				opt->opt_flen -= dsthdrlen;
951 			}
952 
953 			break;
954 
955 		case IPV6_2292HOPLIMIT:
956 		case IPV6_HOPLIMIT:
957 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
958 				err = -EINVAL;
959 				goto exit_f;
960 			}
961 
962 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
963 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
964 				err = -EINVAL;
965 				goto exit_f;
966 			}
967 
968 			break;
969 
970 		case IPV6_TCLASS:
971 		    {
972 			int tc;
973 
974 			err = -EINVAL;
975 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
976 				goto exit_f;
977 
978 			tc = *(int *)CMSG_DATA(cmsg);
979 			if (tc < -1 || tc > 0xff)
980 				goto exit_f;
981 
982 			err = 0;
983 			ipc6->tclass = tc;
984 
985 			break;
986 		    }
987 
988 		case IPV6_DONTFRAG:
989 		    {
990 			int df;
991 
992 			err = -EINVAL;
993 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
994 				goto exit_f;
995 
996 			df = *(int *)CMSG_DATA(cmsg);
997 			if (df < 0 || df > 1)
998 				goto exit_f;
999 
1000 			err = 0;
1001 			ipc6->dontfrag = df;
1002 
1003 			break;
1004 		    }
1005 		default:
1006 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1007 					    cmsg->cmsg_type);
1008 			err = -EINVAL;
1009 			goto exit_f;
1010 		}
1011 	}
1012 
1013 exit_f:
1014 	return err;
1015 }
1016 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1017 
1018 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1019 			     __u16 srcp, __u16 destp, int bucket)
1020 {
1021 	const struct in6_addr *dest, *src;
1022 
1023 	dest  = &sp->sk_v6_daddr;
1024 	src   = &sp->sk_v6_rcv_saddr;
1025 	seq_printf(seq,
1026 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1027 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1028 		   bucket,
1029 		   src->s6_addr32[0], src->s6_addr32[1],
1030 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1031 		   dest->s6_addr32[0], dest->s6_addr32[1],
1032 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1033 		   sp->sk_state,
1034 		   sk_wmem_alloc_get(sp),
1035 		   sk_rmem_alloc_get(sp),
1036 		   0, 0L, 0,
1037 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1038 		   0,
1039 		   sock_i_ino(sp),
1040 		   atomic_read(&sp->sk_refcnt), sp,
1041 		   atomic_read(&sp->sk_drops));
1042 }
1043