xref: /linux/net/ipv6/datagram.c (revision 80d443e8876602be2c130f79c4de81e12e2a700d)
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 	addr_type = ipv6_addr_type(&usin->sin6_addr);
171 
172 	if (addr_type == IPV6_ADDR_ANY) {
173 		/*
174 		 *	connect to self
175 		 */
176 		usin->sin6_addr.s6_addr[15] = 0x01;
177 	}
178 
179 	daddr = &usin->sin6_addr;
180 
181 	if (addr_type == IPV6_ADDR_MAPPED) {
182 		struct sockaddr_in sin;
183 
184 		if (__ipv6_only_sock(sk)) {
185 			err = -ENETUNREACH;
186 			goto out;
187 		}
188 		sin.sin_family = AF_INET;
189 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
190 		sin.sin_port = usin->sin6_port;
191 
192 		err = __ip4_datagram_connect(sk,
193 					     (struct sockaddr *) &sin,
194 					     sizeof(sin));
195 
196 ipv4_connected:
197 		if (err)
198 			goto out;
199 
200 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
201 
202 		if (ipv6_addr_any(&np->saddr) ||
203 		    ipv6_mapped_addr_any(&np->saddr))
204 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
205 
206 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
207 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
208 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
209 					       &sk->sk_v6_rcv_saddr);
210 			if (sk->sk_prot->rehash)
211 				sk->sk_prot->rehash(sk);
212 		}
213 
214 		goto out;
215 	}
216 
217 	if (__ipv6_addr_needs_scope_id(addr_type)) {
218 		if (addr_len >= sizeof(struct sockaddr_in6) &&
219 		    usin->sin6_scope_id) {
220 			if (sk->sk_bound_dev_if &&
221 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
222 				err = -EINVAL;
223 				goto out;
224 			}
225 			sk->sk_bound_dev_if = usin->sin6_scope_id;
226 		}
227 
228 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
229 			sk->sk_bound_dev_if = np->mcast_oif;
230 
231 		/* Connect to link-local address requires an interface */
232 		if (!sk->sk_bound_dev_if) {
233 			err = -EINVAL;
234 			goto out;
235 		}
236 	}
237 
238 	sk->sk_v6_daddr = *daddr;
239 	np->flow_label = fl6_flowlabel;
240 
241 	inet->inet_dport = usin->sin6_port;
242 
243 	/*
244 	 *	Check for a route to destination an obtain the
245 	 *	destination cache for it.
246 	 */
247 
248 	err = ip6_datagram_dst_update(sk, true);
249 	if (err)
250 		goto out;
251 
252 	sk->sk_state = TCP_ESTABLISHED;
253 	sk_set_txhash(sk);
254 out:
255 	return err;
256 }
257 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
258 
259 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
260 {
261 	int res;
262 
263 	lock_sock(sk);
264 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
265 	release_sock(sk);
266 	return res;
267 }
268 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
269 
270 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
271 				 int addr_len)
272 {
273 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
274 	if (sin6->sin6_family != AF_INET6)
275 		return -EAFNOSUPPORT;
276 	return ip6_datagram_connect(sk, uaddr, addr_len);
277 }
278 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
279 
280 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
281 		     __be16 port, u32 info, u8 *payload)
282 {
283 	struct ipv6_pinfo *np  = inet6_sk(sk);
284 	struct icmp6hdr *icmph = icmp6_hdr(skb);
285 	struct sock_exterr_skb *serr;
286 
287 	if (!np->recverr)
288 		return;
289 
290 	skb = skb_clone(skb, GFP_ATOMIC);
291 	if (!skb)
292 		return;
293 
294 	skb->protocol = htons(ETH_P_IPV6);
295 
296 	serr = SKB_EXT_ERR(skb);
297 	serr->ee.ee_errno = err;
298 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
299 	serr->ee.ee_type = icmph->icmp6_type;
300 	serr->ee.ee_code = icmph->icmp6_code;
301 	serr->ee.ee_pad = 0;
302 	serr->ee.ee_info = info;
303 	serr->ee.ee_data = 0;
304 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
305 				  skb_network_header(skb);
306 	serr->port = port;
307 
308 	__skb_pull(skb, payload - skb->data);
309 	skb_reset_transport_header(skb);
310 
311 	if (sock_queue_err_skb(sk, skb))
312 		kfree_skb(skb);
313 }
314 
315 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
316 {
317 	const struct ipv6_pinfo *np = inet6_sk(sk);
318 	struct sock_exterr_skb *serr;
319 	struct ipv6hdr *iph;
320 	struct sk_buff *skb;
321 
322 	if (!np->recverr)
323 		return;
324 
325 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
326 	if (!skb)
327 		return;
328 
329 	skb->protocol = htons(ETH_P_IPV6);
330 
331 	skb_put(skb, sizeof(struct ipv6hdr));
332 	skb_reset_network_header(skb);
333 	iph = ipv6_hdr(skb);
334 	iph->daddr = fl6->daddr;
335 
336 	serr = SKB_EXT_ERR(skb);
337 	serr->ee.ee_errno = err;
338 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
339 	serr->ee.ee_type = 0;
340 	serr->ee.ee_code = 0;
341 	serr->ee.ee_pad = 0;
342 	serr->ee.ee_info = info;
343 	serr->ee.ee_data = 0;
344 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
345 	serr->port = fl6->fl6_dport;
346 
347 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
348 	skb_reset_transport_header(skb);
349 
350 	if (sock_queue_err_skb(sk, skb))
351 		kfree_skb(skb);
352 }
353 
354 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
355 {
356 	struct ipv6_pinfo *np = inet6_sk(sk);
357 	struct ipv6hdr *iph;
358 	struct sk_buff *skb;
359 	struct ip6_mtuinfo *mtu_info;
360 
361 	if (!np->rxopt.bits.rxpmtu)
362 		return;
363 
364 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
365 	if (!skb)
366 		return;
367 
368 	skb_put(skb, sizeof(struct ipv6hdr));
369 	skb_reset_network_header(skb);
370 	iph = ipv6_hdr(skb);
371 	iph->daddr = fl6->daddr;
372 
373 	mtu_info = IP6CBMTU(skb);
374 
375 	mtu_info->ip6m_mtu = mtu;
376 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
377 	mtu_info->ip6m_addr.sin6_port = 0;
378 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
379 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
380 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
381 
382 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
383 	skb_reset_transport_header(skb);
384 
385 	skb = xchg(&np->rxpmtu, skb);
386 	kfree_skb(skb);
387 }
388 
389 /* For some errors we have valid addr_offset even with zero payload and
390  * zero port. Also, addr_offset should be supported if port is set.
391  */
392 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
393 {
394 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
395 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
396 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
397 }
398 
399 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
400  *
401  * At one point, excluding local errors was a quick test to identify icmp/icmp6
402  * errors. This is no longer true, but the test remained, so the v6 stack,
403  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
404  *
405  * Timestamp code paths do not initialize the fields expected by cmsg:
406  * the PKTINFO fields in skb->cb[]. Fill those in here.
407  */
408 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
409 				      struct sock_exterr_skb *serr)
410 {
411 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
412 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
413 		return true;
414 
415 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
416 		return false;
417 
418 	if (!skb->dev)
419 		return false;
420 
421 	if (skb->protocol == htons(ETH_P_IPV6))
422 		IP6CB(skb)->iif = skb->dev->ifindex;
423 	else
424 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
425 
426 	return true;
427 }
428 
429 /*
430  *	Handle MSG_ERRQUEUE
431  */
432 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
433 {
434 	struct ipv6_pinfo *np = inet6_sk(sk);
435 	struct sock_exterr_skb *serr;
436 	struct sk_buff *skb;
437 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
438 	struct {
439 		struct sock_extended_err ee;
440 		struct sockaddr_in6	 offender;
441 	} errhdr;
442 	int err;
443 	int copied;
444 
445 	err = -EAGAIN;
446 	skb = sock_dequeue_err_skb(sk);
447 	if (!skb)
448 		goto out;
449 
450 	copied = skb->len;
451 	if (copied > len) {
452 		msg->msg_flags |= MSG_TRUNC;
453 		copied = len;
454 	}
455 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
456 	if (unlikely(err)) {
457 		kfree_skb(skb);
458 		return err;
459 	}
460 	sock_recv_timestamp(msg, sk, skb);
461 
462 	serr = SKB_EXT_ERR(skb);
463 
464 	if (sin && ipv6_datagram_support_addr(serr)) {
465 		const unsigned char *nh = skb_network_header(skb);
466 		sin->sin6_family = AF_INET6;
467 		sin->sin6_flowinfo = 0;
468 		sin->sin6_port = serr->port;
469 		if (skb->protocol == htons(ETH_P_IPV6)) {
470 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
471 								  struct ipv6hdr, daddr);
472 			sin->sin6_addr = ip6h->daddr;
473 			if (np->sndflow)
474 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
475 			sin->sin6_scope_id =
476 				ipv6_iface_scope_id(&sin->sin6_addr,
477 						    IP6CB(skb)->iif);
478 		} else {
479 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
480 					       &sin->sin6_addr);
481 			sin->sin6_scope_id = 0;
482 		}
483 		*addr_len = sizeof(*sin);
484 	}
485 
486 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
487 	sin = &errhdr.offender;
488 	memset(sin, 0, sizeof(*sin));
489 
490 	if (ip6_datagram_support_cmsg(skb, serr)) {
491 		sin->sin6_family = AF_INET6;
492 		if (np->rxopt.all)
493 			ip6_datagram_recv_common_ctl(sk, msg, skb);
494 		if (skb->protocol == htons(ETH_P_IPV6)) {
495 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
496 			if (np->rxopt.all)
497 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
498 			sin->sin6_scope_id =
499 				ipv6_iface_scope_id(&sin->sin6_addr,
500 						    IP6CB(skb)->iif);
501 		} else {
502 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
503 					       &sin->sin6_addr);
504 			if (inet_sk(sk)->cmsg_flags)
505 				ip_cmsg_recv(msg, skb);
506 		}
507 	}
508 
509 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
510 
511 	/* Now we could try to dump offended packet options */
512 
513 	msg->msg_flags |= MSG_ERRQUEUE;
514 	err = copied;
515 
516 	consume_skb(skb);
517 out:
518 	return err;
519 }
520 EXPORT_SYMBOL_GPL(ipv6_recv_error);
521 
522 /*
523  *	Handle IPV6_RECVPATHMTU
524  */
525 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
526 		     int *addr_len)
527 {
528 	struct ipv6_pinfo *np = inet6_sk(sk);
529 	struct sk_buff *skb;
530 	struct ip6_mtuinfo mtu_info;
531 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
532 	int err;
533 	int copied;
534 
535 	err = -EAGAIN;
536 	skb = xchg(&np->rxpmtu, NULL);
537 	if (!skb)
538 		goto out;
539 
540 	copied = skb->len;
541 	if (copied > len) {
542 		msg->msg_flags |= MSG_TRUNC;
543 		copied = len;
544 	}
545 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
546 	if (err)
547 		goto out_free_skb;
548 
549 	sock_recv_timestamp(msg, sk, skb);
550 
551 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
552 
553 	if (sin) {
554 		sin->sin6_family = AF_INET6;
555 		sin->sin6_flowinfo = 0;
556 		sin->sin6_port = 0;
557 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
558 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
559 		*addr_len = sizeof(*sin);
560 	}
561 
562 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
563 
564 	err = copied;
565 
566 out_free_skb:
567 	kfree_skb(skb);
568 out:
569 	return err;
570 }
571 
572 
573 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
574 				 struct sk_buff *skb)
575 {
576 	struct ipv6_pinfo *np = inet6_sk(sk);
577 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
578 
579 	if (np->rxopt.bits.rxinfo) {
580 		struct in6_pktinfo src_info;
581 
582 		if (is_ipv6) {
583 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
584 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
585 		} else {
586 			src_info.ipi6_ifindex =
587 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
588 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
589 					       &src_info.ipi6_addr);
590 		}
591 
592 		if (src_info.ipi6_ifindex >= 0)
593 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
594 				 sizeof(src_info), &src_info);
595 	}
596 }
597 
598 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
599 				    struct sk_buff *skb)
600 {
601 	struct ipv6_pinfo *np = inet6_sk(sk);
602 	struct inet6_skb_parm *opt = IP6CB(skb);
603 	unsigned char *nh = skb_network_header(skb);
604 
605 	if (np->rxopt.bits.rxhlim) {
606 		int hlim = ipv6_hdr(skb)->hop_limit;
607 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
608 	}
609 
610 	if (np->rxopt.bits.rxtclass) {
611 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
612 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
613 	}
614 
615 	if (np->rxopt.bits.rxflow) {
616 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
617 		if (flowinfo)
618 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
619 	}
620 
621 	/* HbH is allowed only once */
622 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
623 		u8 *ptr = nh + sizeof(struct ipv6hdr);
624 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
625 	}
626 
627 	if (opt->lastopt &&
628 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
629 		/*
630 		 * Silly enough, but we need to reparse in order to
631 		 * report extension headers (except for HbH)
632 		 * in order.
633 		 *
634 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
635 		 * (and WILL NOT be) defined because
636 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
637 		 */
638 		unsigned int off = sizeof(struct ipv6hdr);
639 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
640 
641 		while (off <= opt->lastopt) {
642 			unsigned int len;
643 			u8 *ptr = nh + off;
644 
645 			switch (nexthdr) {
646 			case IPPROTO_DSTOPTS:
647 				nexthdr = ptr[0];
648 				len = (ptr[1] + 1) << 3;
649 				if (np->rxopt.bits.dstopts)
650 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
651 				break;
652 			case IPPROTO_ROUTING:
653 				nexthdr = ptr[0];
654 				len = (ptr[1] + 1) << 3;
655 				if (np->rxopt.bits.srcrt)
656 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
657 				break;
658 			case IPPROTO_AH:
659 				nexthdr = ptr[0];
660 				len = (ptr[1] + 2) << 2;
661 				break;
662 			default:
663 				nexthdr = ptr[0];
664 				len = (ptr[1] + 1) << 3;
665 				break;
666 			}
667 
668 			off += len;
669 		}
670 	}
671 
672 	/* socket options in old style */
673 	if (np->rxopt.bits.rxoinfo) {
674 		struct in6_pktinfo src_info;
675 
676 		src_info.ipi6_ifindex = opt->iif;
677 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
678 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
679 	}
680 	if (np->rxopt.bits.rxohlim) {
681 		int hlim = ipv6_hdr(skb)->hop_limit;
682 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
683 	}
684 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
685 		u8 *ptr = nh + sizeof(struct ipv6hdr);
686 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
687 	}
688 	if (np->rxopt.bits.odstopts && opt->dst0) {
689 		u8 *ptr = nh + opt->dst0;
690 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
691 	}
692 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
693 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
694 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
695 	}
696 	if (np->rxopt.bits.odstopts && opt->dst1) {
697 		u8 *ptr = nh + opt->dst1;
698 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
699 	}
700 	if (np->rxopt.bits.rxorigdstaddr) {
701 		struct sockaddr_in6 sin6;
702 		__be16 *ports = (__be16 *) skb_transport_header(skb);
703 
704 		if (skb_transport_offset(skb) + 4 <= (int)skb->len) {
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 
710 			sin6.sin6_family = AF_INET6;
711 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
712 			sin6.sin6_port = ports[1];
713 			sin6.sin6_flowinfo = 0;
714 			sin6.sin6_scope_id =
715 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
716 						    opt->iif);
717 
718 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
719 		}
720 	}
721 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
722 		int val = opt->frag_max_size;
723 
724 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
725 	}
726 }
727 
728 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
729 			  struct sk_buff *skb)
730 {
731 	ip6_datagram_recv_common_ctl(sk, msg, skb);
732 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
733 }
734 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
735 
736 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
737 			  struct msghdr *msg, struct flowi6 *fl6,
738 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
739 {
740 	struct in6_pktinfo *src_info;
741 	struct cmsghdr *cmsg;
742 	struct ipv6_rt_hdr *rthdr;
743 	struct ipv6_opt_hdr *hdr;
744 	struct ipv6_txoptions *opt = ipc6->opt;
745 	int len;
746 	int err = 0;
747 
748 	for_each_cmsghdr(cmsg, msg) {
749 		int addr_type;
750 
751 		if (!CMSG_OK(msg, cmsg)) {
752 			err = -EINVAL;
753 			goto exit_f;
754 		}
755 
756 		if (cmsg->cmsg_level == SOL_SOCKET) {
757 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
758 			if (err)
759 				return err;
760 			continue;
761 		}
762 
763 		if (cmsg->cmsg_level != SOL_IPV6)
764 			continue;
765 
766 		switch (cmsg->cmsg_type) {
767 		case IPV6_PKTINFO:
768 		case IPV6_2292PKTINFO:
769 		    {
770 			struct net_device *dev = NULL;
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 
779 			if (src_info->ipi6_ifindex) {
780 				if (fl6->flowi6_oif &&
781 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
782 					return -EINVAL;
783 				fl6->flowi6_oif = src_info->ipi6_ifindex;
784 			}
785 
786 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
787 
788 			rcu_read_lock();
789 			if (fl6->flowi6_oif) {
790 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
791 				if (!dev) {
792 					rcu_read_unlock();
793 					return -ENODEV;
794 				}
795 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
796 				rcu_read_unlock();
797 				return -EINVAL;
798 			}
799 
800 			if (addr_type != IPV6_ADDR_ANY) {
801 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
802 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
803 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
804 						   strict ? dev : NULL, 0) &&
805 				    !ipv6_chk_acast_addr_src(net, dev,
806 							     &src_info->ipi6_addr))
807 					err = -EINVAL;
808 				else
809 					fl6->saddr = src_info->ipi6_addr;
810 			}
811 
812 			rcu_read_unlock();
813 
814 			if (err)
815 				goto exit_f;
816 
817 			break;
818 		    }
819 
820 		case IPV6_FLOWINFO:
821 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
822 				err = -EINVAL;
823 				goto exit_f;
824 			}
825 
826 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
827 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
828 					err = -EINVAL;
829 					goto exit_f;
830 				}
831 			}
832 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
833 			break;
834 
835 		case IPV6_2292HOPOPTS:
836 		case IPV6_HOPOPTS:
837 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
838 				err = -EINVAL;
839 				goto exit_f;
840 			}
841 
842 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
843 			len = ((hdr->hdrlen + 1) << 3);
844 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
845 				err = -EINVAL;
846 				goto exit_f;
847 			}
848 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
849 				err = -EPERM;
850 				goto exit_f;
851 			}
852 			opt->opt_nflen += len;
853 			opt->hopopt = hdr;
854 			break;
855 
856 		case IPV6_2292DSTOPTS:
857 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
858 				err = -EINVAL;
859 				goto exit_f;
860 			}
861 
862 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
863 			len = ((hdr->hdrlen + 1) << 3);
864 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
865 				err = -EINVAL;
866 				goto exit_f;
867 			}
868 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
869 				err = -EPERM;
870 				goto exit_f;
871 			}
872 			if (opt->dst1opt) {
873 				err = -EINVAL;
874 				goto exit_f;
875 			}
876 			opt->opt_flen += len;
877 			opt->dst1opt = hdr;
878 			break;
879 
880 		case IPV6_DSTOPTS:
881 		case IPV6_RTHDRDSTOPTS:
882 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
883 				err = -EINVAL;
884 				goto exit_f;
885 			}
886 
887 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
888 			len = ((hdr->hdrlen + 1) << 3);
889 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
890 				err = -EINVAL;
891 				goto exit_f;
892 			}
893 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
894 				err = -EPERM;
895 				goto exit_f;
896 			}
897 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
898 				opt->opt_flen += len;
899 				opt->dst1opt = hdr;
900 			} else {
901 				opt->opt_nflen += len;
902 				opt->dst0opt = hdr;
903 			}
904 			break;
905 
906 		case IPV6_2292RTHDR:
907 		case IPV6_RTHDR:
908 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
909 				err = -EINVAL;
910 				goto exit_f;
911 			}
912 
913 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
914 
915 			switch (rthdr->type) {
916 #if IS_ENABLED(CONFIG_IPV6_MIP6)
917 			case IPV6_SRCRT_TYPE_2:
918 				if (rthdr->hdrlen != 2 ||
919 				    rthdr->segments_left != 1) {
920 					err = -EINVAL;
921 					goto exit_f;
922 				}
923 				break;
924 #endif
925 			default:
926 				err = -EINVAL;
927 				goto exit_f;
928 			}
929 
930 			len = ((rthdr->hdrlen + 1) << 3);
931 
932 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
933 				err = -EINVAL;
934 				goto exit_f;
935 			}
936 
937 			/* segments left must also match */
938 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
939 				err = -EINVAL;
940 				goto exit_f;
941 			}
942 
943 			opt->opt_nflen += len;
944 			opt->srcrt = rthdr;
945 
946 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
947 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
948 
949 				opt->opt_nflen += dsthdrlen;
950 				opt->dst0opt = opt->dst1opt;
951 				opt->dst1opt = NULL;
952 				opt->opt_flen -= dsthdrlen;
953 			}
954 
955 			break;
956 
957 		case IPV6_2292HOPLIMIT:
958 		case IPV6_HOPLIMIT:
959 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
960 				err = -EINVAL;
961 				goto exit_f;
962 			}
963 
964 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
965 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
966 				err = -EINVAL;
967 				goto exit_f;
968 			}
969 
970 			break;
971 
972 		case IPV6_TCLASS:
973 		    {
974 			int tc;
975 
976 			err = -EINVAL;
977 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
978 				goto exit_f;
979 
980 			tc = *(int *)CMSG_DATA(cmsg);
981 			if (tc < -1 || tc > 0xff)
982 				goto exit_f;
983 
984 			err = 0;
985 			ipc6->tclass = tc;
986 
987 			break;
988 		    }
989 
990 		case IPV6_DONTFRAG:
991 		    {
992 			int df;
993 
994 			err = -EINVAL;
995 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
996 				goto exit_f;
997 
998 			df = *(int *)CMSG_DATA(cmsg);
999 			if (df < 0 || df > 1)
1000 				goto exit_f;
1001 
1002 			err = 0;
1003 			ipc6->dontfrag = df;
1004 
1005 			break;
1006 		    }
1007 		default:
1008 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1009 					    cmsg->cmsg_type);
1010 			err = -EINVAL;
1011 			goto exit_f;
1012 		}
1013 	}
1014 
1015 exit_f:
1016 	return err;
1017 }
1018 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1019 
1020 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1021 			     __u16 srcp, __u16 destp, int bucket)
1022 {
1023 	const struct in6_addr *dest, *src;
1024 
1025 	dest  = &sp->sk_v6_daddr;
1026 	src   = &sp->sk_v6_rcv_saddr;
1027 	seq_printf(seq,
1028 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1029 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1030 		   bucket,
1031 		   src->s6_addr32[0], src->s6_addr32[1],
1032 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1033 		   dest->s6_addr32[0], dest->s6_addr32[1],
1034 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1035 		   sp->sk_state,
1036 		   sk_wmem_alloc_get(sp),
1037 		   sk_rmem_alloc_get(sp),
1038 		   0, 0L, 0,
1039 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1040 		   0,
1041 		   sock_i_ino(sp),
1042 		   atomic_read(&sp->sk_refcnt), sp,
1043 		   atomic_read(&sp->sk_drops));
1044 }
1045