xref: /linux/net/ipv6/datagram.c (revision f474808acb3c4b30552d9c59b181244e0300d218)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	common UDP/RAW code
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  */
9 
10 #include <linux/capability.h>
11 #include <linux/errno.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/interrupt.h>
15 #include <linux/socket.h>
16 #include <linux/sockios.h>
17 #include <linux/in6.h>
18 #include <linux/ipv6.h>
19 #include <linux/route.h>
20 #include <linux/slab.h>
21 #include <linux/export.h>
22 
23 #include <net/ipv6.h>
24 #include <net/ndisc.h>
25 #include <net/addrconf.h>
26 #include <net/transp_v6.h>
27 #include <net/ip6_route.h>
28 #include <net/tcp_states.h>
29 #include <net/dsfield.h>
30 
31 #include <linux/errqueue.h>
32 #include <linux/uaccess.h>
33 
34 static bool ipv6_mapped_addr_any(const struct in6_addr *a)
35 {
36 	return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
37 }
38 
39 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk)
40 {
41 	struct inet_sock *inet = inet_sk(sk);
42 	struct ipv6_pinfo *np = inet6_sk(sk);
43 
44 	memset(fl6, 0, sizeof(*fl6));
45 	fl6->flowi6_proto = sk->sk_protocol;
46 	fl6->daddr = sk->sk_v6_daddr;
47 	fl6->saddr = np->saddr;
48 	fl6->flowi6_oif = sk->sk_bound_dev_if;
49 	fl6->flowi6_mark = sk->sk_mark;
50 	fl6->fl6_dport = inet->inet_dport;
51 	fl6->fl6_sport = inet->inet_sport;
52 	fl6->flowlabel = np->flow_label;
53 	fl6->flowi6_uid = sk->sk_uid;
54 
55 	if (!fl6->flowi6_oif)
56 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
57 
58 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr))
59 		fl6->flowi6_oif = np->mcast_oif;
60 
61 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
62 }
63 
64 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
65 {
66 	struct ip6_flowlabel *flowlabel = NULL;
67 	struct in6_addr *final_p, final;
68 	struct ipv6_txoptions *opt;
69 	struct dst_entry *dst;
70 	struct inet_sock *inet = inet_sk(sk);
71 	struct ipv6_pinfo *np = inet6_sk(sk);
72 	struct flowi6 fl6;
73 	int err = 0;
74 
75 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
76 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
77 		if (IS_ERR(flowlabel))
78 			return -EINVAL;
79 	}
80 	ip6_datagram_flow_key_init(&fl6, sk);
81 
82 	rcu_read_lock();
83 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
84 	final_p = fl6_update_dst(&fl6, opt, &final);
85 	rcu_read_unlock();
86 
87 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
88 	if (IS_ERR(dst)) {
89 		err = PTR_ERR(dst);
90 		goto out;
91 	}
92 
93 	if (fix_sk_saddr) {
94 		if (ipv6_addr_any(&np->saddr))
95 			np->saddr = fl6.saddr;
96 
97 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
98 			sk->sk_v6_rcv_saddr = fl6.saddr;
99 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
100 			if (sk->sk_prot->rehash)
101 				sk->sk_prot->rehash(sk);
102 		}
103 	}
104 
105 	ip6_sk_dst_store_flow(sk, dst, &fl6);
106 
107 out:
108 	fl6_sock_release(flowlabel);
109 	return err;
110 }
111 
112 void ip6_datagram_release_cb(struct sock *sk)
113 {
114 	struct dst_entry *dst;
115 
116 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
117 		return;
118 
119 	rcu_read_lock();
120 	dst = __sk_dst_get(sk);
121 	if (!dst || !dst->obsolete ||
122 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
123 		rcu_read_unlock();
124 		return;
125 	}
126 	rcu_read_unlock();
127 
128 	ip6_datagram_dst_update(sk, false);
129 }
130 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
131 
132 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
133 			   int addr_len)
134 {
135 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
136 	struct inet_sock	*inet = inet_sk(sk);
137 	struct ipv6_pinfo	*np = inet6_sk(sk);
138 	struct in6_addr		*daddr, old_daddr;
139 	__be32			fl6_flowlabel = 0;
140 	__be32			old_fl6_flowlabel;
141 	__be16			old_dport;
142 	int			addr_type;
143 	int			err;
144 
145 	if (usin->sin6_family == AF_INET) {
146 		if (__ipv6_only_sock(sk))
147 			return -EAFNOSUPPORT;
148 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
149 		goto ipv4_connected;
150 	}
151 
152 	if (addr_len < SIN6_LEN_RFC2133)
153 		return -EINVAL;
154 
155 	if (usin->sin6_family != AF_INET6)
156 		return -EAFNOSUPPORT;
157 
158 	if (np->sndflow)
159 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
160 
161 	if (ipv6_addr_any(&usin->sin6_addr)) {
162 		/*
163 		 *	connect to self
164 		 */
165 		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
166 			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
167 					       &usin->sin6_addr);
168 		else
169 			usin->sin6_addr = in6addr_loopback;
170 	}
171 
172 	addr_type = ipv6_addr_type(&usin->sin6_addr);
173 
174 	daddr = &usin->sin6_addr;
175 
176 	if (addr_type & IPV6_ADDR_MAPPED) {
177 		struct sockaddr_in sin;
178 
179 		if (__ipv6_only_sock(sk)) {
180 			err = -ENETUNREACH;
181 			goto out;
182 		}
183 		sin.sin_family = AF_INET;
184 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
185 		sin.sin_port = usin->sin6_port;
186 
187 		err = __ip4_datagram_connect(sk,
188 					     (struct sockaddr *) &sin,
189 					     sizeof(sin));
190 
191 ipv4_connected:
192 		if (err)
193 			goto out;
194 
195 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
196 
197 		if (ipv6_addr_any(&np->saddr) ||
198 		    ipv6_mapped_addr_any(&np->saddr))
199 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
200 
201 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
202 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
203 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
204 					       &sk->sk_v6_rcv_saddr);
205 			if (sk->sk_prot->rehash)
206 				sk->sk_prot->rehash(sk);
207 		}
208 
209 		goto out;
210 	}
211 
212 	if (__ipv6_addr_needs_scope_id(addr_type)) {
213 		if (addr_len >= sizeof(struct sockaddr_in6) &&
214 		    usin->sin6_scope_id) {
215 			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
216 				err = -EINVAL;
217 				goto out;
218 			}
219 			sk->sk_bound_dev_if = usin->sin6_scope_id;
220 		}
221 
222 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
223 			sk->sk_bound_dev_if = np->mcast_oif;
224 
225 		/* Connect to link-local address requires an interface */
226 		if (!sk->sk_bound_dev_if) {
227 			err = -EINVAL;
228 			goto out;
229 		}
230 	}
231 
232 	/* save the current peer information before updating it */
233 	old_daddr = sk->sk_v6_daddr;
234 	old_fl6_flowlabel = np->flow_label;
235 	old_dport = inet->inet_dport;
236 
237 	sk->sk_v6_daddr = *daddr;
238 	np->flow_label = fl6_flowlabel;
239 	inet->inet_dport = usin->sin6_port;
240 
241 	/*
242 	 *	Check for a route to destination an obtain the
243 	 *	destination cache for it.
244 	 */
245 
246 	err = ip6_datagram_dst_update(sk, true);
247 	if (err) {
248 		/* Restore the socket peer info, to keep it consistent with
249 		 * the old socket state
250 		 */
251 		sk->sk_v6_daddr = old_daddr;
252 		np->flow_label = old_fl6_flowlabel;
253 		inet->inet_dport = old_dport;
254 		goto out;
255 	}
256 
257 	sk->sk_state = TCP_ESTABLISHED;
258 	sk_set_txhash(sk);
259 out:
260 	return err;
261 }
262 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
263 
264 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
265 {
266 	int res;
267 
268 	lock_sock(sk);
269 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
270 	release_sock(sk);
271 	return res;
272 }
273 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
274 
275 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
276 				 int addr_len)
277 {
278 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
279 	if (sin6->sin6_family != AF_INET6)
280 		return -EAFNOSUPPORT;
281 	return ip6_datagram_connect(sk, uaddr, addr_len);
282 }
283 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
284 
285 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
286 		     __be16 port, u32 info, u8 *payload)
287 {
288 	struct ipv6_pinfo *np  = inet6_sk(sk);
289 	struct icmp6hdr *icmph = icmp6_hdr(skb);
290 	struct sock_exterr_skb *serr;
291 
292 	if (!np->recverr)
293 		return;
294 
295 	skb = skb_clone(skb, GFP_ATOMIC);
296 	if (!skb)
297 		return;
298 
299 	skb->protocol = htons(ETH_P_IPV6);
300 
301 	serr = SKB_EXT_ERR(skb);
302 	serr->ee.ee_errno = err;
303 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
304 	serr->ee.ee_type = icmph->icmp6_type;
305 	serr->ee.ee_code = icmph->icmp6_code;
306 	serr->ee.ee_pad = 0;
307 	serr->ee.ee_info = info;
308 	serr->ee.ee_data = 0;
309 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
310 				  skb_network_header(skb);
311 	serr->port = port;
312 
313 	__skb_pull(skb, payload - skb->data);
314 	skb_reset_transport_header(skb);
315 
316 	if (sock_queue_err_skb(sk, skb))
317 		kfree_skb(skb);
318 }
319 
320 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
321 {
322 	const struct ipv6_pinfo *np = inet6_sk(sk);
323 	struct sock_exterr_skb *serr;
324 	struct ipv6hdr *iph;
325 	struct sk_buff *skb;
326 
327 	if (!np->recverr)
328 		return;
329 
330 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
331 	if (!skb)
332 		return;
333 
334 	skb->protocol = htons(ETH_P_IPV6);
335 
336 	skb_put(skb, sizeof(struct ipv6hdr));
337 	skb_reset_network_header(skb);
338 	iph = ipv6_hdr(skb);
339 	iph->daddr = fl6->daddr;
340 	ip6_flow_hdr(iph, 0, 0);
341 
342 	serr = SKB_EXT_ERR(skb);
343 	serr->ee.ee_errno = err;
344 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
345 	serr->ee.ee_type = 0;
346 	serr->ee.ee_code = 0;
347 	serr->ee.ee_pad = 0;
348 	serr->ee.ee_info = info;
349 	serr->ee.ee_data = 0;
350 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
351 	serr->port = fl6->fl6_dport;
352 
353 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
354 	skb_reset_transport_header(skb);
355 
356 	if (sock_queue_err_skb(sk, skb))
357 		kfree_skb(skb);
358 }
359 
360 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
361 {
362 	struct ipv6_pinfo *np = inet6_sk(sk);
363 	struct ipv6hdr *iph;
364 	struct sk_buff *skb;
365 	struct ip6_mtuinfo *mtu_info;
366 
367 	if (!np->rxopt.bits.rxpmtu)
368 		return;
369 
370 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
371 	if (!skb)
372 		return;
373 
374 	skb_put(skb, sizeof(struct ipv6hdr));
375 	skb_reset_network_header(skb);
376 	iph = ipv6_hdr(skb);
377 	iph->daddr = fl6->daddr;
378 
379 	mtu_info = IP6CBMTU(skb);
380 
381 	mtu_info->ip6m_mtu = mtu;
382 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
383 	mtu_info->ip6m_addr.sin6_port = 0;
384 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
385 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
386 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
387 
388 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
389 	skb_reset_transport_header(skb);
390 
391 	skb = xchg(&np->rxpmtu, skb);
392 	kfree_skb(skb);
393 }
394 
395 /* For some errors we have valid addr_offset even with zero payload and
396  * zero port. Also, addr_offset should be supported if port is set.
397  */
398 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
399 {
400 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
401 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
402 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
403 }
404 
405 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
406  *
407  * At one point, excluding local errors was a quick test to identify icmp/icmp6
408  * errors. This is no longer true, but the test remained, so the v6 stack,
409  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
410  */
411 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
412 				      struct sock_exterr_skb *serr)
413 {
414 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
415 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
416 		return true;
417 
418 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
419 		return false;
420 
421 	if (!IP6CB(skb)->iif)
422 		return false;
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[2], *ports;
701 
702 		ports = skb_header_pointer(skb, skb_transport_offset(skb),
703 					   sizeof(_ports), &_ports);
704 		if (ports) {
705 			/* All current transport protocols have the port numbers in the
706 			 * first four bytes of the transport header and this function is
707 			 * written with this assumption in mind.
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 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
721 		int val = opt->frag_max_size;
722 
723 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
724 	}
725 }
726 
727 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
728 			  struct sk_buff *skb)
729 {
730 	ip6_datagram_recv_common_ctl(sk, msg, skb);
731 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
732 }
733 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
734 
735 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
736 			  struct msghdr *msg, struct flowi6 *fl6,
737 			  struct ipcm6_cookie *ipc6)
738 {
739 	struct in6_pktinfo *src_info;
740 	struct cmsghdr *cmsg;
741 	struct ipv6_rt_hdr *rthdr;
742 	struct ipv6_opt_hdr *hdr;
743 	struct ipv6_txoptions *opt = ipc6->opt;
744 	int len;
745 	int err = 0;
746 
747 	for_each_cmsghdr(cmsg, msg) {
748 		int addr_type;
749 
750 		if (!CMSG_OK(msg, cmsg)) {
751 			err = -EINVAL;
752 			goto exit_f;
753 		}
754 
755 		if (cmsg->cmsg_level == SOL_SOCKET) {
756 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc6->sockc);
757 			if (err)
758 				return err;
759 			continue;
760 		}
761 
762 		if (cmsg->cmsg_level != SOL_IPV6)
763 			continue;
764 
765 		switch (cmsg->cmsg_type) {
766 		case IPV6_PKTINFO:
767 		case IPV6_2292PKTINFO:
768 		    {
769 			struct net_device *dev = NULL;
770 			int src_idx;
771 
772 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
773 				err = -EINVAL;
774 				goto exit_f;
775 			}
776 
777 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
778 			src_idx = src_info->ipi6_ifindex;
779 
780 			if (src_idx) {
781 				if (fl6->flowi6_oif &&
782 				    src_idx != fl6->flowi6_oif &&
783 				    (sk->sk_bound_dev_if != fl6->flowi6_oif ||
784 				     !sk_dev_equal_l3scope(sk, src_idx)))
785 					return -EINVAL;
786 				fl6->flowi6_oif = src_idx;
787 			}
788 
789 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
790 
791 			rcu_read_lock();
792 			if (fl6->flowi6_oif) {
793 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
794 				if (!dev) {
795 					rcu_read_unlock();
796 					return -ENODEV;
797 				}
798 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
799 				rcu_read_unlock();
800 				return -EINVAL;
801 			}
802 
803 			if (addr_type != IPV6_ADDR_ANY) {
804 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
805 				if (!ipv6_can_nonlocal_bind(net, inet_sk(sk)) &&
806 				    !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
807 							     dev, !strict, 0,
808 							     IFA_F_TENTATIVE) &&
809 				    !ipv6_chk_acast_addr_src(net, dev,
810 							     &src_info->ipi6_addr))
811 					err = -EINVAL;
812 				else
813 					fl6->saddr = src_info->ipi6_addr;
814 			}
815 
816 			rcu_read_unlock();
817 
818 			if (err)
819 				goto exit_f;
820 
821 			break;
822 		    }
823 
824 		case IPV6_FLOWINFO:
825 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
826 				err = -EINVAL;
827 				goto exit_f;
828 			}
829 
830 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
831 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
832 					err = -EINVAL;
833 					goto exit_f;
834 				}
835 			}
836 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
837 			break;
838 
839 		case IPV6_2292HOPOPTS:
840 		case IPV6_HOPOPTS:
841 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
842 				err = -EINVAL;
843 				goto exit_f;
844 			}
845 
846 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
847 			len = ((hdr->hdrlen + 1) << 3);
848 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
849 				err = -EINVAL;
850 				goto exit_f;
851 			}
852 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
853 				err = -EPERM;
854 				goto exit_f;
855 			}
856 			opt->opt_nflen += len;
857 			opt->hopopt = hdr;
858 			break;
859 
860 		case IPV6_2292DSTOPTS:
861 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
862 				err = -EINVAL;
863 				goto exit_f;
864 			}
865 
866 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
867 			len = ((hdr->hdrlen + 1) << 3);
868 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
869 				err = -EINVAL;
870 				goto exit_f;
871 			}
872 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
873 				err = -EPERM;
874 				goto exit_f;
875 			}
876 			if (opt->dst1opt) {
877 				err = -EINVAL;
878 				goto exit_f;
879 			}
880 			opt->opt_flen += len;
881 			opt->dst1opt = hdr;
882 			break;
883 
884 		case IPV6_DSTOPTS:
885 		case IPV6_RTHDRDSTOPTS:
886 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
887 				err = -EINVAL;
888 				goto exit_f;
889 			}
890 
891 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
892 			len = ((hdr->hdrlen + 1) << 3);
893 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
894 				err = -EINVAL;
895 				goto exit_f;
896 			}
897 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
898 				err = -EPERM;
899 				goto exit_f;
900 			}
901 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
902 				opt->opt_flen += len;
903 				opt->dst1opt = hdr;
904 			} else {
905 				opt->opt_nflen += len;
906 				opt->dst0opt = hdr;
907 			}
908 			break;
909 
910 		case IPV6_2292RTHDR:
911 		case IPV6_RTHDR:
912 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
913 				err = -EINVAL;
914 				goto exit_f;
915 			}
916 
917 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
918 
919 			switch (rthdr->type) {
920 #if IS_ENABLED(CONFIG_IPV6_MIP6)
921 			case IPV6_SRCRT_TYPE_2:
922 				if (rthdr->hdrlen != 2 ||
923 				    rthdr->segments_left != 1) {
924 					err = -EINVAL;
925 					goto exit_f;
926 				}
927 				break;
928 #endif
929 			default:
930 				err = -EINVAL;
931 				goto exit_f;
932 			}
933 
934 			len = ((rthdr->hdrlen + 1) << 3);
935 
936 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
937 				err = -EINVAL;
938 				goto exit_f;
939 			}
940 
941 			/* segments left must also match */
942 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
943 				err = -EINVAL;
944 				goto exit_f;
945 			}
946 
947 			opt->opt_nflen += len;
948 			opt->srcrt = rthdr;
949 
950 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
951 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
952 
953 				opt->opt_nflen += dsthdrlen;
954 				opt->dst0opt = opt->dst1opt;
955 				opt->dst1opt = NULL;
956 				opt->opt_flen -= dsthdrlen;
957 			}
958 
959 			break;
960 
961 		case IPV6_2292HOPLIMIT:
962 		case IPV6_HOPLIMIT:
963 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
964 				err = -EINVAL;
965 				goto exit_f;
966 			}
967 
968 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
969 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
970 				err = -EINVAL;
971 				goto exit_f;
972 			}
973 
974 			break;
975 
976 		case IPV6_TCLASS:
977 		    {
978 			int tc;
979 
980 			err = -EINVAL;
981 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
982 				goto exit_f;
983 
984 			tc = *(int *)CMSG_DATA(cmsg);
985 			if (tc < -1 || tc > 0xff)
986 				goto exit_f;
987 
988 			err = 0;
989 			ipc6->tclass = tc;
990 
991 			break;
992 		    }
993 
994 		case IPV6_DONTFRAG:
995 		    {
996 			int df;
997 
998 			err = -EINVAL;
999 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1000 				goto exit_f;
1001 
1002 			df = *(int *)CMSG_DATA(cmsg);
1003 			if (df < 0 || df > 1)
1004 				goto exit_f;
1005 
1006 			err = 0;
1007 			ipc6->dontfrag = df;
1008 
1009 			break;
1010 		    }
1011 		default:
1012 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1013 					    cmsg->cmsg_type);
1014 			err = -EINVAL;
1015 			goto exit_f;
1016 		}
1017 	}
1018 
1019 exit_f:
1020 	return err;
1021 }
1022 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1023 
1024 void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1025 			       __u16 srcp, __u16 destp, int rqueue, int bucket)
1026 {
1027 	const struct in6_addr *dest, *src;
1028 
1029 	dest  = &sp->sk_v6_daddr;
1030 	src   = &sp->sk_v6_rcv_saddr;
1031 	seq_printf(seq,
1032 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1033 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1034 		   bucket,
1035 		   src->s6_addr32[0], src->s6_addr32[1],
1036 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1037 		   dest->s6_addr32[0], dest->s6_addr32[1],
1038 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1039 		   sp->sk_state,
1040 		   sk_wmem_alloc_get(sp),
1041 		   rqueue,
1042 		   0, 0L, 0,
1043 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1044 		   0,
1045 		   sock_i_ino(sp),
1046 		   refcount_read(&sp->sk_refcnt), sp,
1047 		   atomic_read(&sp->sk_drops));
1048 }
1049