xref: /linux/net/ipv6/datagram.c (revision fd639726bf15fca8ee1a00dce8e0096d0ad9bd18)
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_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 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
144 			   int addr_len)
145 {
146 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
147 	struct inet_sock	*inet = inet_sk(sk);
148 	struct ipv6_pinfo	*np = inet6_sk(sk);
149 	struct in6_addr		*daddr;
150 	int			addr_type;
151 	int			err;
152 	__be32			fl6_flowlabel = 0;
153 
154 	if (usin->sin6_family == AF_INET) {
155 		if (__ipv6_only_sock(sk))
156 			return -EAFNOSUPPORT;
157 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
158 		goto ipv4_connected;
159 	}
160 
161 	if (addr_len < SIN6_LEN_RFC2133)
162 		return -EINVAL;
163 
164 	if (usin->sin6_family != AF_INET6)
165 		return -EAFNOSUPPORT;
166 
167 	if (np->sndflow)
168 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
169 
170 	if (ipv6_addr_any(&usin->sin6_addr)) {
171 		/*
172 		 *	connect to self
173 		 */
174 		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
175 			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
176 					       &usin->sin6_addr);
177 		else
178 			usin->sin6_addr = in6addr_loopback;
179 	}
180 
181 	addr_type = ipv6_addr_type(&usin->sin6_addr);
182 
183 	daddr = &usin->sin6_addr;
184 
185 	if (addr_type & IPV6_ADDR_MAPPED) {
186 		struct sockaddr_in sin;
187 
188 		if (__ipv6_only_sock(sk)) {
189 			err = -ENETUNREACH;
190 			goto out;
191 		}
192 		sin.sin_family = AF_INET;
193 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
194 		sin.sin_port = usin->sin6_port;
195 
196 		err = __ip4_datagram_connect(sk,
197 					     (struct sockaddr *) &sin,
198 					     sizeof(sin));
199 
200 ipv4_connected:
201 		if (err)
202 			goto out;
203 
204 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
205 
206 		if (ipv6_addr_any(&np->saddr) ||
207 		    ipv6_mapped_addr_any(&np->saddr))
208 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
209 
210 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
211 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
212 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
213 					       &sk->sk_v6_rcv_saddr);
214 			if (sk->sk_prot->rehash)
215 				sk->sk_prot->rehash(sk);
216 		}
217 
218 		goto out;
219 	}
220 
221 	if (__ipv6_addr_needs_scope_id(addr_type)) {
222 		if (addr_len >= sizeof(struct sockaddr_in6) &&
223 		    usin->sin6_scope_id) {
224 			if (sk->sk_bound_dev_if &&
225 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
226 				err = -EINVAL;
227 				goto out;
228 			}
229 			sk->sk_bound_dev_if = usin->sin6_scope_id;
230 		}
231 
232 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
233 			sk->sk_bound_dev_if = np->mcast_oif;
234 
235 		/* Connect to link-local address requires an interface */
236 		if (!sk->sk_bound_dev_if) {
237 			err = -EINVAL;
238 			goto out;
239 		}
240 	}
241 
242 	sk->sk_v6_daddr = *daddr;
243 	np->flow_label = fl6_flowlabel;
244 
245 	inet->inet_dport = usin->sin6_port;
246 
247 	/*
248 	 *	Check for a route to destination an obtain the
249 	 *	destination cache for it.
250 	 */
251 
252 	err = ip6_datagram_dst_update(sk, true);
253 	if (err) {
254 		/* Reset daddr and dport so that udp_v6_early_demux()
255 		 * fails to find this socket
256 		 */
257 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
258 		inet->inet_dport = 0;
259 		goto out;
260 	}
261 
262 	sk->sk_state = TCP_ESTABLISHED;
263 	sk_set_txhash(sk);
264 out:
265 	return err;
266 }
267 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
268 
269 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
270 {
271 	int res;
272 
273 	lock_sock(sk);
274 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
275 	release_sock(sk);
276 	return res;
277 }
278 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
279 
280 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
281 				 int addr_len)
282 {
283 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
284 	if (sin6->sin6_family != AF_INET6)
285 		return -EAFNOSUPPORT;
286 	return ip6_datagram_connect(sk, uaddr, addr_len);
287 }
288 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
289 
290 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
291 		     __be16 port, u32 info, u8 *payload)
292 {
293 	struct ipv6_pinfo *np  = inet6_sk(sk);
294 	struct icmp6hdr *icmph = icmp6_hdr(skb);
295 	struct sock_exterr_skb *serr;
296 
297 	if (!np->recverr)
298 		return;
299 
300 	skb = skb_clone(skb, GFP_ATOMIC);
301 	if (!skb)
302 		return;
303 
304 	skb->protocol = htons(ETH_P_IPV6);
305 
306 	serr = SKB_EXT_ERR(skb);
307 	serr->ee.ee_errno = err;
308 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
309 	serr->ee.ee_type = icmph->icmp6_type;
310 	serr->ee.ee_code = icmph->icmp6_code;
311 	serr->ee.ee_pad = 0;
312 	serr->ee.ee_info = info;
313 	serr->ee.ee_data = 0;
314 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
315 				  skb_network_header(skb);
316 	serr->port = port;
317 
318 	__skb_pull(skb, payload - skb->data);
319 	skb_reset_transport_header(skb);
320 
321 	if (sock_queue_err_skb(sk, skb))
322 		kfree_skb(skb);
323 }
324 
325 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
326 {
327 	const struct ipv6_pinfo *np = inet6_sk(sk);
328 	struct sock_exterr_skb *serr;
329 	struct ipv6hdr *iph;
330 	struct sk_buff *skb;
331 
332 	if (!np->recverr)
333 		return;
334 
335 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
336 	if (!skb)
337 		return;
338 
339 	skb->protocol = htons(ETH_P_IPV6);
340 
341 	skb_put(skb, sizeof(struct ipv6hdr));
342 	skb_reset_network_header(skb);
343 	iph = ipv6_hdr(skb);
344 	iph->daddr = fl6->daddr;
345 
346 	serr = SKB_EXT_ERR(skb);
347 	serr->ee.ee_errno = err;
348 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
349 	serr->ee.ee_type = 0;
350 	serr->ee.ee_code = 0;
351 	serr->ee.ee_pad = 0;
352 	serr->ee.ee_info = info;
353 	serr->ee.ee_data = 0;
354 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
355 	serr->port = fl6->fl6_dport;
356 
357 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
358 	skb_reset_transport_header(skb);
359 
360 	if (sock_queue_err_skb(sk, skb))
361 		kfree_skb(skb);
362 }
363 
364 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
365 {
366 	struct ipv6_pinfo *np = inet6_sk(sk);
367 	struct ipv6hdr *iph;
368 	struct sk_buff *skb;
369 	struct ip6_mtuinfo *mtu_info;
370 
371 	if (!np->rxopt.bits.rxpmtu)
372 		return;
373 
374 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
375 	if (!skb)
376 		return;
377 
378 	skb_put(skb, sizeof(struct ipv6hdr));
379 	skb_reset_network_header(skb);
380 	iph = ipv6_hdr(skb);
381 	iph->daddr = fl6->daddr;
382 
383 	mtu_info = IP6CBMTU(skb);
384 
385 	mtu_info->ip6m_mtu = mtu;
386 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
387 	mtu_info->ip6m_addr.sin6_port = 0;
388 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
389 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
390 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
391 
392 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
393 	skb_reset_transport_header(skb);
394 
395 	skb = xchg(&np->rxpmtu, skb);
396 	kfree_skb(skb);
397 }
398 
399 /* For some errors we have valid addr_offset even with zero payload and
400  * zero port. Also, addr_offset should be supported if port is set.
401  */
402 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
403 {
404 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
405 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
406 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
407 }
408 
409 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
410  *
411  * At one point, excluding local errors was a quick test to identify icmp/icmp6
412  * errors. This is no longer true, but the test remained, so the v6 stack,
413  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
414  */
415 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
416 				      struct sock_exterr_skb *serr)
417 {
418 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
419 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
420 		return true;
421 
422 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
423 		return false;
424 
425 	if (!IP6CB(skb)->iif)
426 		return false;
427 
428 	return true;
429 }
430 
431 /*
432  *	Handle MSG_ERRQUEUE
433  */
434 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
435 {
436 	struct ipv6_pinfo *np = inet6_sk(sk);
437 	struct sock_exterr_skb *serr;
438 	struct sk_buff *skb;
439 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
440 	struct {
441 		struct sock_extended_err ee;
442 		struct sockaddr_in6	 offender;
443 	} errhdr;
444 	int err;
445 	int copied;
446 
447 	err = -EAGAIN;
448 	skb = sock_dequeue_err_skb(sk);
449 	if (!skb)
450 		goto out;
451 
452 	copied = skb->len;
453 	if (copied > len) {
454 		msg->msg_flags |= MSG_TRUNC;
455 		copied = len;
456 	}
457 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
458 	if (unlikely(err)) {
459 		kfree_skb(skb);
460 		return err;
461 	}
462 	sock_recv_timestamp(msg, sk, skb);
463 
464 	serr = SKB_EXT_ERR(skb);
465 
466 	if (sin && ipv6_datagram_support_addr(serr)) {
467 		const unsigned char *nh = skb_network_header(skb);
468 		sin->sin6_family = AF_INET6;
469 		sin->sin6_flowinfo = 0;
470 		sin->sin6_port = serr->port;
471 		if (skb->protocol == htons(ETH_P_IPV6)) {
472 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
473 								  struct ipv6hdr, daddr);
474 			sin->sin6_addr = ip6h->daddr;
475 			if (np->sndflow)
476 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
477 			sin->sin6_scope_id =
478 				ipv6_iface_scope_id(&sin->sin6_addr,
479 						    IP6CB(skb)->iif);
480 		} else {
481 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
482 					       &sin->sin6_addr);
483 			sin->sin6_scope_id = 0;
484 		}
485 		*addr_len = sizeof(*sin);
486 	}
487 
488 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
489 	sin = &errhdr.offender;
490 	memset(sin, 0, sizeof(*sin));
491 
492 	if (ip6_datagram_support_cmsg(skb, serr)) {
493 		sin->sin6_family = AF_INET6;
494 		if (np->rxopt.all)
495 			ip6_datagram_recv_common_ctl(sk, msg, skb);
496 		if (skb->protocol == htons(ETH_P_IPV6)) {
497 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
498 			if (np->rxopt.all)
499 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
500 			sin->sin6_scope_id =
501 				ipv6_iface_scope_id(&sin->sin6_addr,
502 						    IP6CB(skb)->iif);
503 		} else {
504 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
505 					       &sin->sin6_addr);
506 			if (inet_sk(sk)->cmsg_flags)
507 				ip_cmsg_recv(msg, skb);
508 		}
509 	}
510 
511 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
512 
513 	/* Now we could try to dump offended packet options */
514 
515 	msg->msg_flags |= MSG_ERRQUEUE;
516 	err = copied;
517 
518 	consume_skb(skb);
519 out:
520 	return err;
521 }
522 EXPORT_SYMBOL_GPL(ipv6_recv_error);
523 
524 /*
525  *	Handle IPV6_RECVPATHMTU
526  */
527 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
528 		     int *addr_len)
529 {
530 	struct ipv6_pinfo *np = inet6_sk(sk);
531 	struct sk_buff *skb;
532 	struct ip6_mtuinfo mtu_info;
533 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
534 	int err;
535 	int copied;
536 
537 	err = -EAGAIN;
538 	skb = xchg(&np->rxpmtu, NULL);
539 	if (!skb)
540 		goto out;
541 
542 	copied = skb->len;
543 	if (copied > len) {
544 		msg->msg_flags |= MSG_TRUNC;
545 		copied = len;
546 	}
547 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
548 	if (err)
549 		goto out_free_skb;
550 
551 	sock_recv_timestamp(msg, sk, skb);
552 
553 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
554 
555 	if (sin) {
556 		sin->sin6_family = AF_INET6;
557 		sin->sin6_flowinfo = 0;
558 		sin->sin6_port = 0;
559 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
560 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
561 		*addr_len = sizeof(*sin);
562 	}
563 
564 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
565 
566 	err = copied;
567 
568 out_free_skb:
569 	kfree_skb(skb);
570 out:
571 	return err;
572 }
573 
574 
575 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
576 				 struct sk_buff *skb)
577 {
578 	struct ipv6_pinfo *np = inet6_sk(sk);
579 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
580 
581 	if (np->rxopt.bits.rxinfo) {
582 		struct in6_pktinfo src_info;
583 
584 		if (is_ipv6) {
585 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
586 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
587 		} else {
588 			src_info.ipi6_ifindex =
589 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
590 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
591 					       &src_info.ipi6_addr);
592 		}
593 
594 		if (src_info.ipi6_ifindex >= 0)
595 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
596 				 sizeof(src_info), &src_info);
597 	}
598 }
599 
600 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
601 				    struct sk_buff *skb)
602 {
603 	struct ipv6_pinfo *np = inet6_sk(sk);
604 	struct inet6_skb_parm *opt = IP6CB(skb);
605 	unsigned char *nh = skb_network_header(skb);
606 
607 	if (np->rxopt.bits.rxhlim) {
608 		int hlim = ipv6_hdr(skb)->hop_limit;
609 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
610 	}
611 
612 	if (np->rxopt.bits.rxtclass) {
613 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
614 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
615 	}
616 
617 	if (np->rxopt.bits.rxflow) {
618 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
619 		if (flowinfo)
620 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
621 	}
622 
623 	/* HbH is allowed only once */
624 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
625 		u8 *ptr = nh + sizeof(struct ipv6hdr);
626 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
627 	}
628 
629 	if (opt->lastopt &&
630 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
631 		/*
632 		 * Silly enough, but we need to reparse in order to
633 		 * report extension headers (except for HbH)
634 		 * in order.
635 		 *
636 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
637 		 * (and WILL NOT be) defined because
638 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
639 		 */
640 		unsigned int off = sizeof(struct ipv6hdr);
641 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
642 
643 		while (off <= opt->lastopt) {
644 			unsigned int len;
645 			u8 *ptr = nh + off;
646 
647 			switch (nexthdr) {
648 			case IPPROTO_DSTOPTS:
649 				nexthdr = ptr[0];
650 				len = (ptr[1] + 1) << 3;
651 				if (np->rxopt.bits.dstopts)
652 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
653 				break;
654 			case IPPROTO_ROUTING:
655 				nexthdr = ptr[0];
656 				len = (ptr[1] + 1) << 3;
657 				if (np->rxopt.bits.srcrt)
658 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
659 				break;
660 			case IPPROTO_AH:
661 				nexthdr = ptr[0];
662 				len = (ptr[1] + 2) << 2;
663 				break;
664 			default:
665 				nexthdr = ptr[0];
666 				len = (ptr[1] + 1) << 3;
667 				break;
668 			}
669 
670 			off += len;
671 		}
672 	}
673 
674 	/* socket options in old style */
675 	if (np->rxopt.bits.rxoinfo) {
676 		struct in6_pktinfo src_info;
677 
678 		src_info.ipi6_ifindex = opt->iif;
679 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
680 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
681 	}
682 	if (np->rxopt.bits.rxohlim) {
683 		int hlim = ipv6_hdr(skb)->hop_limit;
684 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
685 	}
686 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
687 		u8 *ptr = nh + sizeof(struct ipv6hdr);
688 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
689 	}
690 	if (np->rxopt.bits.odstopts && opt->dst0) {
691 		u8 *ptr = nh + opt->dst0;
692 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
693 	}
694 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
695 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
696 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
697 	}
698 	if (np->rxopt.bits.odstopts && opt->dst1) {
699 		u8 *ptr = nh + opt->dst1;
700 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
701 	}
702 	if (np->rxopt.bits.rxorigdstaddr) {
703 		struct sockaddr_in6 sin6;
704 		__be16 *ports = (__be16 *) skb_transport_header(skb);
705 
706 		if (skb_transport_offset(skb) + 4 <= (int)skb->len) {
707 			/* All current transport protocols have the port numbers in the
708 			 * first four bytes of the transport header and this function is
709 			 * written with this assumption in mind.
710 			 */
711 
712 			sin6.sin6_family = AF_INET6;
713 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
714 			sin6.sin6_port = ports[1];
715 			sin6.sin6_flowinfo = 0;
716 			sin6.sin6_scope_id =
717 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
718 						    opt->iif);
719 
720 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
721 		}
722 	}
723 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
724 		int val = opt->frag_max_size;
725 
726 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
727 	}
728 }
729 
730 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
731 			  struct sk_buff *skb)
732 {
733 	ip6_datagram_recv_common_ctl(sk, msg, skb);
734 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
735 }
736 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
737 
738 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
739 			  struct msghdr *msg, struct flowi6 *fl6,
740 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
741 {
742 	struct in6_pktinfo *src_info;
743 	struct cmsghdr *cmsg;
744 	struct ipv6_rt_hdr *rthdr;
745 	struct ipv6_opt_hdr *hdr;
746 	struct ipv6_txoptions *opt = ipc6->opt;
747 	int len;
748 	int err = 0;
749 
750 	for_each_cmsghdr(cmsg, msg) {
751 		int addr_type;
752 
753 		if (!CMSG_OK(msg, cmsg)) {
754 			err = -EINVAL;
755 			goto exit_f;
756 		}
757 
758 		if (cmsg->cmsg_level == SOL_SOCKET) {
759 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
760 			if (err)
761 				return err;
762 			continue;
763 		}
764 
765 		if (cmsg->cmsg_level != SOL_IPV6)
766 			continue;
767 
768 		switch (cmsg->cmsg_type) {
769 		case IPV6_PKTINFO:
770 		case IPV6_2292PKTINFO:
771 		    {
772 			struct net_device *dev = NULL;
773 
774 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
775 				err = -EINVAL;
776 				goto exit_f;
777 			}
778 
779 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
780 
781 			if (src_info->ipi6_ifindex) {
782 				if (fl6->flowi6_oif &&
783 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
784 					return -EINVAL;
785 				fl6->flowi6_oif = src_info->ipi6_ifindex;
786 			}
787 
788 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
789 
790 			rcu_read_lock();
791 			if (fl6->flowi6_oif) {
792 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
793 				if (!dev) {
794 					rcu_read_unlock();
795 					return -ENODEV;
796 				}
797 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
798 				rcu_read_unlock();
799 				return -EINVAL;
800 			}
801 
802 			if (addr_type != IPV6_ADDR_ANY) {
803 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
804 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
805 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
806 						   strict ? dev : NULL, 0) &&
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 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 		   sk_rmem_alloc_get(sp),
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