xref: /openbmc/linux/net/ipv6/datagram.c (revision 3e26a691)
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 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
44 {
45 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
46 	struct inet_sock	*inet = inet_sk(sk);
47 	struct ipv6_pinfo	*np = inet6_sk(sk);
48 	struct in6_addr	*daddr, *final_p, final;
49 	struct dst_entry	*dst;
50 	struct flowi6		fl6;
51 	struct ip6_flowlabel	*flowlabel = NULL;
52 	struct ipv6_txoptions	*opt;
53 	int			addr_type;
54 	int			err;
55 
56 	if (usin->sin6_family == AF_INET) {
57 		if (__ipv6_only_sock(sk))
58 			return -EAFNOSUPPORT;
59 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
60 		goto ipv4_connected;
61 	}
62 
63 	if (addr_len < SIN6_LEN_RFC2133)
64 		return -EINVAL;
65 
66 	if (usin->sin6_family != AF_INET6)
67 		return -EAFNOSUPPORT;
68 
69 	memset(&fl6, 0, sizeof(fl6));
70 	if (np->sndflow) {
71 		fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
72 		if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
73 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
74 			if (!flowlabel)
75 				return -EINVAL;
76 		}
77 	}
78 
79 	addr_type = ipv6_addr_type(&usin->sin6_addr);
80 
81 	if (addr_type == IPV6_ADDR_ANY) {
82 		/*
83 		 *	connect to self
84 		 */
85 		usin->sin6_addr.s6_addr[15] = 0x01;
86 	}
87 
88 	daddr = &usin->sin6_addr;
89 
90 	if (addr_type == IPV6_ADDR_MAPPED) {
91 		struct sockaddr_in sin;
92 
93 		if (__ipv6_only_sock(sk)) {
94 			err = -ENETUNREACH;
95 			goto out;
96 		}
97 		sin.sin_family = AF_INET;
98 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
99 		sin.sin_port = usin->sin6_port;
100 
101 		err = __ip4_datagram_connect(sk,
102 					     (struct sockaddr *) &sin,
103 					     sizeof(sin));
104 
105 ipv4_connected:
106 		if (err)
107 			goto out;
108 
109 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
110 
111 		if (ipv6_addr_any(&np->saddr) ||
112 		    ipv6_mapped_addr_any(&np->saddr))
113 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
114 
115 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
116 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
117 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
118 					       &sk->sk_v6_rcv_saddr);
119 			if (sk->sk_prot->rehash)
120 				sk->sk_prot->rehash(sk);
121 		}
122 
123 		goto out;
124 	}
125 
126 	if (__ipv6_addr_needs_scope_id(addr_type)) {
127 		if (addr_len >= sizeof(struct sockaddr_in6) &&
128 		    usin->sin6_scope_id) {
129 			if (sk->sk_bound_dev_if &&
130 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
131 				err = -EINVAL;
132 				goto out;
133 			}
134 			sk->sk_bound_dev_if = usin->sin6_scope_id;
135 		}
136 
137 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
138 			sk->sk_bound_dev_if = np->mcast_oif;
139 
140 		/* Connect to link-local address requires an interface */
141 		if (!sk->sk_bound_dev_if) {
142 			err = -EINVAL;
143 			goto out;
144 		}
145 	}
146 
147 	sk->sk_v6_daddr = *daddr;
148 	np->flow_label = fl6.flowlabel;
149 
150 	inet->inet_dport = usin->sin6_port;
151 
152 	/*
153 	 *	Check for a route to destination an obtain the
154 	 *	destination cache for it.
155 	 */
156 
157 	fl6.flowi6_proto = sk->sk_protocol;
158 	fl6.daddr = sk->sk_v6_daddr;
159 	fl6.saddr = np->saddr;
160 	fl6.flowi6_oif = sk->sk_bound_dev_if;
161 	fl6.flowi6_mark = sk->sk_mark;
162 	fl6.fl6_dport = inet->inet_dport;
163 	fl6.fl6_sport = inet->inet_sport;
164 
165 	if (!fl6.flowi6_oif)
166 		fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
167 
168 	if (!fl6.flowi6_oif && (addr_type&IPV6_ADDR_MULTICAST))
169 		fl6.flowi6_oif = np->mcast_oif;
170 
171 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
172 
173 	rcu_read_lock();
174 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
175 	final_p = fl6_update_dst(&fl6, opt, &final);
176 	rcu_read_unlock();
177 
178 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
179 	err = 0;
180 	if (IS_ERR(dst)) {
181 		err = PTR_ERR(dst);
182 		goto out;
183 	}
184 
185 	/* source address lookup done in ip6_dst_lookup */
186 
187 	if (ipv6_addr_any(&np->saddr))
188 		np->saddr = fl6.saddr;
189 
190 	if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
191 		sk->sk_v6_rcv_saddr = fl6.saddr;
192 		inet->inet_rcv_saddr = LOOPBACK4_IPV6;
193 		if (sk->sk_prot->rehash)
194 			sk->sk_prot->rehash(sk);
195 	}
196 
197 	ip6_dst_store(sk, dst,
198 		      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
199 		      &sk->sk_v6_daddr : NULL,
200 #ifdef CONFIG_IPV6_SUBTREES
201 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
202 		      &np->saddr :
203 #endif
204 		      NULL);
205 
206 	sk->sk_state = TCP_ESTABLISHED;
207 	sk_set_txhash(sk);
208 out:
209 	fl6_sock_release(flowlabel);
210 	return err;
211 }
212 
213 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
214 {
215 	int res;
216 
217 	lock_sock(sk);
218 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
219 	release_sock(sk);
220 	return res;
221 }
222 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
223 
224 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
225 				 int addr_len)
226 {
227 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
228 	if (sin6->sin6_family != AF_INET6)
229 		return -EAFNOSUPPORT;
230 	return ip6_datagram_connect(sk, uaddr, addr_len);
231 }
232 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
233 
234 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
235 		     __be16 port, u32 info, u8 *payload)
236 {
237 	struct ipv6_pinfo *np  = inet6_sk(sk);
238 	struct icmp6hdr *icmph = icmp6_hdr(skb);
239 	struct sock_exterr_skb *serr;
240 
241 	if (!np->recverr)
242 		return;
243 
244 	skb = skb_clone(skb, GFP_ATOMIC);
245 	if (!skb)
246 		return;
247 
248 	skb->protocol = htons(ETH_P_IPV6);
249 
250 	serr = SKB_EXT_ERR(skb);
251 	serr->ee.ee_errno = err;
252 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
253 	serr->ee.ee_type = icmph->icmp6_type;
254 	serr->ee.ee_code = icmph->icmp6_code;
255 	serr->ee.ee_pad = 0;
256 	serr->ee.ee_info = info;
257 	serr->ee.ee_data = 0;
258 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
259 				  skb_network_header(skb);
260 	serr->port = port;
261 
262 	__skb_pull(skb, payload - skb->data);
263 	skb_reset_transport_header(skb);
264 
265 	if (sock_queue_err_skb(sk, skb))
266 		kfree_skb(skb);
267 }
268 
269 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
270 {
271 	const struct ipv6_pinfo *np = inet6_sk(sk);
272 	struct sock_exterr_skb *serr;
273 	struct ipv6hdr *iph;
274 	struct sk_buff *skb;
275 
276 	if (!np->recverr)
277 		return;
278 
279 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
280 	if (!skb)
281 		return;
282 
283 	skb->protocol = htons(ETH_P_IPV6);
284 
285 	skb_put(skb, sizeof(struct ipv6hdr));
286 	skb_reset_network_header(skb);
287 	iph = ipv6_hdr(skb);
288 	iph->daddr = fl6->daddr;
289 
290 	serr = SKB_EXT_ERR(skb);
291 	serr->ee.ee_errno = err;
292 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
293 	serr->ee.ee_type = 0;
294 	serr->ee.ee_code = 0;
295 	serr->ee.ee_pad = 0;
296 	serr->ee.ee_info = info;
297 	serr->ee.ee_data = 0;
298 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
299 	serr->port = fl6->fl6_dport;
300 
301 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
302 	skb_reset_transport_header(skb);
303 
304 	if (sock_queue_err_skb(sk, skb))
305 		kfree_skb(skb);
306 }
307 
308 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
309 {
310 	struct ipv6_pinfo *np = inet6_sk(sk);
311 	struct ipv6hdr *iph;
312 	struct sk_buff *skb;
313 	struct ip6_mtuinfo *mtu_info;
314 
315 	if (!np->rxopt.bits.rxpmtu)
316 		return;
317 
318 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
319 	if (!skb)
320 		return;
321 
322 	skb_put(skb, sizeof(struct ipv6hdr));
323 	skb_reset_network_header(skb);
324 	iph = ipv6_hdr(skb);
325 	iph->daddr = fl6->daddr;
326 
327 	mtu_info = IP6CBMTU(skb);
328 
329 	mtu_info->ip6m_mtu = mtu;
330 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
331 	mtu_info->ip6m_addr.sin6_port = 0;
332 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
333 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
334 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
335 
336 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
337 	skb_reset_transport_header(skb);
338 
339 	skb = xchg(&np->rxpmtu, skb);
340 	kfree_skb(skb);
341 }
342 
343 /* For some errors we have valid addr_offset even with zero payload and
344  * zero port. Also, addr_offset should be supported if port is set.
345  */
346 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
347 {
348 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
349 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
350 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
351 }
352 
353 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
354  *
355  * At one point, excluding local errors was a quick test to identify icmp/icmp6
356  * errors. This is no longer true, but the test remained, so the v6 stack,
357  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
358  *
359  * Timestamp code paths do not initialize the fields expected by cmsg:
360  * the PKTINFO fields in skb->cb[]. Fill those in here.
361  */
362 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
363 				      struct sock_exterr_skb *serr)
364 {
365 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
366 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
367 		return true;
368 
369 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
370 		return false;
371 
372 	if (!skb->dev)
373 		return false;
374 
375 	if (skb->protocol == htons(ETH_P_IPV6))
376 		IP6CB(skb)->iif = skb->dev->ifindex;
377 	else
378 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
379 
380 	return true;
381 }
382 
383 /*
384  *	Handle MSG_ERRQUEUE
385  */
386 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
387 {
388 	struct ipv6_pinfo *np = inet6_sk(sk);
389 	struct sock_exterr_skb *serr;
390 	struct sk_buff *skb;
391 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
392 	struct {
393 		struct sock_extended_err ee;
394 		struct sockaddr_in6	 offender;
395 	} errhdr;
396 	int err;
397 	int copied;
398 
399 	err = -EAGAIN;
400 	skb = sock_dequeue_err_skb(sk);
401 	if (!skb)
402 		goto out;
403 
404 	copied = skb->len;
405 	if (copied > len) {
406 		msg->msg_flags |= MSG_TRUNC;
407 		copied = len;
408 	}
409 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
410 	if (err)
411 		goto out_free_skb;
412 
413 	sock_recv_timestamp(msg, sk, skb);
414 
415 	serr = SKB_EXT_ERR(skb);
416 
417 	if (sin && ipv6_datagram_support_addr(serr)) {
418 		const unsigned char *nh = skb_network_header(skb);
419 		sin->sin6_family = AF_INET6;
420 		sin->sin6_flowinfo = 0;
421 		sin->sin6_port = serr->port;
422 		if (skb->protocol == htons(ETH_P_IPV6)) {
423 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
424 								  struct ipv6hdr, daddr);
425 			sin->sin6_addr = ip6h->daddr;
426 			if (np->sndflow)
427 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
428 			sin->sin6_scope_id =
429 				ipv6_iface_scope_id(&sin->sin6_addr,
430 						    IP6CB(skb)->iif);
431 		} else {
432 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
433 					       &sin->sin6_addr);
434 			sin->sin6_scope_id = 0;
435 		}
436 		*addr_len = sizeof(*sin);
437 	}
438 
439 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
440 	sin = &errhdr.offender;
441 	memset(sin, 0, sizeof(*sin));
442 
443 	if (ip6_datagram_support_cmsg(skb, serr)) {
444 		sin->sin6_family = AF_INET6;
445 		if (np->rxopt.all)
446 			ip6_datagram_recv_common_ctl(sk, msg, skb);
447 		if (skb->protocol == htons(ETH_P_IPV6)) {
448 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
449 			if (np->rxopt.all)
450 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
451 			sin->sin6_scope_id =
452 				ipv6_iface_scope_id(&sin->sin6_addr,
453 						    IP6CB(skb)->iif);
454 		} else {
455 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
456 					       &sin->sin6_addr);
457 			if (inet_sk(sk)->cmsg_flags)
458 				ip_cmsg_recv(msg, skb);
459 		}
460 	}
461 
462 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
463 
464 	/* Now we could try to dump offended packet options */
465 
466 	msg->msg_flags |= MSG_ERRQUEUE;
467 	err = copied;
468 
469 out_free_skb:
470 	kfree_skb(skb);
471 out:
472 	return err;
473 }
474 EXPORT_SYMBOL_GPL(ipv6_recv_error);
475 
476 /*
477  *	Handle IPV6_RECVPATHMTU
478  */
479 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
480 		     int *addr_len)
481 {
482 	struct ipv6_pinfo *np = inet6_sk(sk);
483 	struct sk_buff *skb;
484 	struct ip6_mtuinfo mtu_info;
485 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
486 	int err;
487 	int copied;
488 
489 	err = -EAGAIN;
490 	skb = xchg(&np->rxpmtu, NULL);
491 	if (!skb)
492 		goto out;
493 
494 	copied = skb->len;
495 	if (copied > len) {
496 		msg->msg_flags |= MSG_TRUNC;
497 		copied = len;
498 	}
499 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
500 	if (err)
501 		goto out_free_skb;
502 
503 	sock_recv_timestamp(msg, sk, skb);
504 
505 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
506 
507 	if (sin) {
508 		sin->sin6_family = AF_INET6;
509 		sin->sin6_flowinfo = 0;
510 		sin->sin6_port = 0;
511 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
512 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
513 		*addr_len = sizeof(*sin);
514 	}
515 
516 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
517 
518 	err = copied;
519 
520 out_free_skb:
521 	kfree_skb(skb);
522 out:
523 	return err;
524 }
525 
526 
527 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
528 				 struct sk_buff *skb)
529 {
530 	struct ipv6_pinfo *np = inet6_sk(sk);
531 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
532 
533 	if (np->rxopt.bits.rxinfo) {
534 		struct in6_pktinfo src_info;
535 
536 		if (is_ipv6) {
537 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
538 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
539 		} else {
540 			src_info.ipi6_ifindex =
541 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
542 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
543 					       &src_info.ipi6_addr);
544 		}
545 
546 		if (src_info.ipi6_ifindex >= 0)
547 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
548 				 sizeof(src_info), &src_info);
549 	}
550 }
551 
552 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
553 				    struct sk_buff *skb)
554 {
555 	struct ipv6_pinfo *np = inet6_sk(sk);
556 	struct inet6_skb_parm *opt = IP6CB(skb);
557 	unsigned char *nh = skb_network_header(skb);
558 
559 	if (np->rxopt.bits.rxhlim) {
560 		int hlim = ipv6_hdr(skb)->hop_limit;
561 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
562 	}
563 
564 	if (np->rxopt.bits.rxtclass) {
565 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
566 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
567 	}
568 
569 	if (np->rxopt.bits.rxflow) {
570 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
571 		if (flowinfo)
572 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
573 	}
574 
575 	/* HbH is allowed only once */
576 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
577 		u8 *ptr = nh + sizeof(struct ipv6hdr);
578 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
579 	}
580 
581 	if (opt->lastopt &&
582 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
583 		/*
584 		 * Silly enough, but we need to reparse in order to
585 		 * report extension headers (except for HbH)
586 		 * in order.
587 		 *
588 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
589 		 * (and WILL NOT be) defined because
590 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
591 		 */
592 		unsigned int off = sizeof(struct ipv6hdr);
593 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
594 
595 		while (off <= opt->lastopt) {
596 			unsigned int len;
597 			u8 *ptr = nh + off;
598 
599 			switch (nexthdr) {
600 			case IPPROTO_DSTOPTS:
601 				nexthdr = ptr[0];
602 				len = (ptr[1] + 1) << 3;
603 				if (np->rxopt.bits.dstopts)
604 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
605 				break;
606 			case IPPROTO_ROUTING:
607 				nexthdr = ptr[0];
608 				len = (ptr[1] + 1) << 3;
609 				if (np->rxopt.bits.srcrt)
610 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
611 				break;
612 			case IPPROTO_AH:
613 				nexthdr = ptr[0];
614 				len = (ptr[1] + 2) << 2;
615 				break;
616 			default:
617 				nexthdr = ptr[0];
618 				len = (ptr[1] + 1) << 3;
619 				break;
620 			}
621 
622 			off += len;
623 		}
624 	}
625 
626 	/* socket options in old style */
627 	if (np->rxopt.bits.rxoinfo) {
628 		struct in6_pktinfo src_info;
629 
630 		src_info.ipi6_ifindex = opt->iif;
631 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
632 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
633 	}
634 	if (np->rxopt.bits.rxohlim) {
635 		int hlim = ipv6_hdr(skb)->hop_limit;
636 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
637 	}
638 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
639 		u8 *ptr = nh + sizeof(struct ipv6hdr);
640 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
641 	}
642 	if (np->rxopt.bits.odstopts && opt->dst0) {
643 		u8 *ptr = nh + opt->dst0;
644 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
645 	}
646 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
647 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
648 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
649 	}
650 	if (np->rxopt.bits.odstopts && opt->dst1) {
651 		u8 *ptr = nh + opt->dst1;
652 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
653 	}
654 	if (np->rxopt.bits.rxorigdstaddr) {
655 		struct sockaddr_in6 sin6;
656 		__be16 *ports = (__be16 *) skb_transport_header(skb);
657 
658 		if (skb_transport_offset(skb) + 4 <= skb->len) {
659 			/* All current transport protocols have the port numbers in the
660 			 * first four bytes of the transport header and this function is
661 			 * written with this assumption in mind.
662 			 */
663 
664 			sin6.sin6_family = AF_INET6;
665 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
666 			sin6.sin6_port = ports[1];
667 			sin6.sin6_flowinfo = 0;
668 			sin6.sin6_scope_id =
669 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
670 						    opt->iif);
671 
672 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
673 		}
674 	}
675 }
676 
677 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
678 			  struct sk_buff *skb)
679 {
680 	ip6_datagram_recv_common_ctl(sk, msg, skb);
681 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
682 }
683 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
684 
685 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
686 			  struct msghdr *msg, struct flowi6 *fl6,
687 			  struct ipv6_txoptions *opt,
688 			  int *hlimit, int *tclass, int *dontfrag)
689 {
690 	struct in6_pktinfo *src_info;
691 	struct cmsghdr *cmsg;
692 	struct ipv6_rt_hdr *rthdr;
693 	struct ipv6_opt_hdr *hdr;
694 	int len;
695 	int err = 0;
696 
697 	for_each_cmsghdr(cmsg, msg) {
698 		int addr_type;
699 
700 		if (!CMSG_OK(msg, cmsg)) {
701 			err = -EINVAL;
702 			goto exit_f;
703 		}
704 
705 		if (cmsg->cmsg_level != SOL_IPV6)
706 			continue;
707 
708 		switch (cmsg->cmsg_type) {
709 		case IPV6_PKTINFO:
710 		case IPV6_2292PKTINFO:
711 		    {
712 			struct net_device *dev = NULL;
713 
714 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
715 				err = -EINVAL;
716 				goto exit_f;
717 			}
718 
719 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
720 
721 			if (src_info->ipi6_ifindex) {
722 				if (fl6->flowi6_oif &&
723 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
724 					return -EINVAL;
725 				fl6->flowi6_oif = src_info->ipi6_ifindex;
726 			}
727 
728 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
729 
730 			rcu_read_lock();
731 			if (fl6->flowi6_oif) {
732 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
733 				if (!dev) {
734 					rcu_read_unlock();
735 					return -ENODEV;
736 				}
737 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
738 				rcu_read_unlock();
739 				return -EINVAL;
740 			}
741 
742 			if (addr_type != IPV6_ADDR_ANY) {
743 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
744 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
745 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
746 						   strict ? dev : NULL, 0) &&
747 				    !ipv6_chk_acast_addr_src(net, dev,
748 							     &src_info->ipi6_addr))
749 					err = -EINVAL;
750 				else
751 					fl6->saddr = src_info->ipi6_addr;
752 			}
753 
754 			rcu_read_unlock();
755 
756 			if (err)
757 				goto exit_f;
758 
759 			break;
760 		    }
761 
762 		case IPV6_FLOWINFO:
763 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
764 				err = -EINVAL;
765 				goto exit_f;
766 			}
767 
768 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
769 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
770 					err = -EINVAL;
771 					goto exit_f;
772 				}
773 			}
774 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
775 			break;
776 
777 		case IPV6_2292HOPOPTS:
778 		case IPV6_HOPOPTS:
779 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
780 				err = -EINVAL;
781 				goto exit_f;
782 			}
783 
784 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
785 			len = ((hdr->hdrlen + 1) << 3);
786 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
787 				err = -EINVAL;
788 				goto exit_f;
789 			}
790 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
791 				err = -EPERM;
792 				goto exit_f;
793 			}
794 			opt->opt_nflen += len;
795 			opt->hopopt = hdr;
796 			break;
797 
798 		case IPV6_2292DSTOPTS:
799 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
800 				err = -EINVAL;
801 				goto exit_f;
802 			}
803 
804 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
805 			len = ((hdr->hdrlen + 1) << 3);
806 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
807 				err = -EINVAL;
808 				goto exit_f;
809 			}
810 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
811 				err = -EPERM;
812 				goto exit_f;
813 			}
814 			if (opt->dst1opt) {
815 				err = -EINVAL;
816 				goto exit_f;
817 			}
818 			opt->opt_flen += len;
819 			opt->dst1opt = hdr;
820 			break;
821 
822 		case IPV6_DSTOPTS:
823 		case IPV6_RTHDRDSTOPTS:
824 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
825 				err = -EINVAL;
826 				goto exit_f;
827 			}
828 
829 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
830 			len = ((hdr->hdrlen + 1) << 3);
831 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
832 				err = -EINVAL;
833 				goto exit_f;
834 			}
835 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
836 				err = -EPERM;
837 				goto exit_f;
838 			}
839 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
840 				opt->opt_flen += len;
841 				opt->dst1opt = hdr;
842 			} else {
843 				opt->opt_nflen += len;
844 				opt->dst0opt = hdr;
845 			}
846 			break;
847 
848 		case IPV6_2292RTHDR:
849 		case IPV6_RTHDR:
850 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
851 				err = -EINVAL;
852 				goto exit_f;
853 			}
854 
855 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
856 
857 			switch (rthdr->type) {
858 #if IS_ENABLED(CONFIG_IPV6_MIP6)
859 			case IPV6_SRCRT_TYPE_2:
860 				if (rthdr->hdrlen != 2 ||
861 				    rthdr->segments_left != 1) {
862 					err = -EINVAL;
863 					goto exit_f;
864 				}
865 				break;
866 #endif
867 			default:
868 				err = -EINVAL;
869 				goto exit_f;
870 			}
871 
872 			len = ((rthdr->hdrlen + 1) << 3);
873 
874 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
875 				err = -EINVAL;
876 				goto exit_f;
877 			}
878 
879 			/* segments left must also match */
880 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
881 				err = -EINVAL;
882 				goto exit_f;
883 			}
884 
885 			opt->opt_nflen += len;
886 			opt->srcrt = rthdr;
887 
888 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
889 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
890 
891 				opt->opt_nflen += dsthdrlen;
892 				opt->dst0opt = opt->dst1opt;
893 				opt->dst1opt = NULL;
894 				opt->opt_flen -= dsthdrlen;
895 			}
896 
897 			break;
898 
899 		case IPV6_2292HOPLIMIT:
900 		case IPV6_HOPLIMIT:
901 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
902 				err = -EINVAL;
903 				goto exit_f;
904 			}
905 
906 			*hlimit = *(int *)CMSG_DATA(cmsg);
907 			if (*hlimit < -1 || *hlimit > 0xff) {
908 				err = -EINVAL;
909 				goto exit_f;
910 			}
911 
912 			break;
913 
914 		case IPV6_TCLASS:
915 		    {
916 			int tc;
917 
918 			err = -EINVAL;
919 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
920 				goto exit_f;
921 
922 			tc = *(int *)CMSG_DATA(cmsg);
923 			if (tc < -1 || tc > 0xff)
924 				goto exit_f;
925 
926 			err = 0;
927 			*tclass = tc;
928 
929 			break;
930 		    }
931 
932 		case IPV6_DONTFRAG:
933 		    {
934 			int df;
935 
936 			err = -EINVAL;
937 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
938 				goto exit_f;
939 
940 			df = *(int *)CMSG_DATA(cmsg);
941 			if (df < 0 || df > 1)
942 				goto exit_f;
943 
944 			err = 0;
945 			*dontfrag = df;
946 
947 			break;
948 		    }
949 		default:
950 			net_dbg_ratelimited("invalid cmsg type: %d\n",
951 					    cmsg->cmsg_type);
952 			err = -EINVAL;
953 			goto exit_f;
954 		}
955 	}
956 
957 exit_f:
958 	return err;
959 }
960 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
961 
962 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
963 			     __u16 srcp, __u16 destp, int bucket)
964 {
965 	const struct in6_addr *dest, *src;
966 
967 	dest  = &sp->sk_v6_daddr;
968 	src   = &sp->sk_v6_rcv_saddr;
969 	seq_printf(seq,
970 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
971 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
972 		   bucket,
973 		   src->s6_addr32[0], src->s6_addr32[1],
974 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
975 		   dest->s6_addr32[0], dest->s6_addr32[1],
976 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
977 		   sp->sk_state,
978 		   sk_wmem_alloc_get(sp),
979 		   sk_rmem_alloc_get(sp),
980 		   0, 0L, 0,
981 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
982 		   0,
983 		   sock_i_ino(sp),
984 		   atomic_read(&sp->sk_refcnt), sp,
985 		   atomic_read(&sp->sk_drops));
986 }
987