xref: /openbmc/linux/net/ipv6/datagram.c (revision a06c488d)
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 && (addr_type&IPV6_ADDR_MULTICAST))
166 		fl6.flowi6_oif = np->mcast_oif;
167 
168 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
169 
170 	rcu_read_lock();
171 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
172 	final_p = fl6_update_dst(&fl6, opt, &final);
173 	rcu_read_unlock();
174 
175 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
176 	err = 0;
177 	if (IS_ERR(dst)) {
178 		err = PTR_ERR(dst);
179 		goto out;
180 	}
181 
182 	/* source address lookup done in ip6_dst_lookup */
183 
184 	if (ipv6_addr_any(&np->saddr))
185 		np->saddr = fl6.saddr;
186 
187 	if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
188 		sk->sk_v6_rcv_saddr = fl6.saddr;
189 		inet->inet_rcv_saddr = LOOPBACK4_IPV6;
190 		if (sk->sk_prot->rehash)
191 			sk->sk_prot->rehash(sk);
192 	}
193 
194 	ip6_dst_store(sk, dst,
195 		      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
196 		      &sk->sk_v6_daddr : NULL,
197 #ifdef CONFIG_IPV6_SUBTREES
198 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
199 		      &np->saddr :
200 #endif
201 		      NULL);
202 
203 	sk->sk_state = TCP_ESTABLISHED;
204 	sk_set_txhash(sk);
205 out:
206 	fl6_sock_release(flowlabel);
207 	return err;
208 }
209 
210 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
211 {
212 	int res;
213 
214 	lock_sock(sk);
215 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
216 	release_sock(sk);
217 	return res;
218 }
219 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
220 
221 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
222 				 int addr_len)
223 {
224 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
225 	if (sin6->sin6_family != AF_INET6)
226 		return -EAFNOSUPPORT;
227 	return ip6_datagram_connect(sk, uaddr, addr_len);
228 }
229 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
230 
231 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
232 		     __be16 port, u32 info, u8 *payload)
233 {
234 	struct ipv6_pinfo *np  = inet6_sk(sk);
235 	struct icmp6hdr *icmph = icmp6_hdr(skb);
236 	struct sock_exterr_skb *serr;
237 
238 	if (!np->recverr)
239 		return;
240 
241 	skb = skb_clone(skb, GFP_ATOMIC);
242 	if (!skb)
243 		return;
244 
245 	skb->protocol = htons(ETH_P_IPV6);
246 
247 	serr = SKB_EXT_ERR(skb);
248 	serr->ee.ee_errno = err;
249 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
250 	serr->ee.ee_type = icmph->icmp6_type;
251 	serr->ee.ee_code = icmph->icmp6_code;
252 	serr->ee.ee_pad = 0;
253 	serr->ee.ee_info = info;
254 	serr->ee.ee_data = 0;
255 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
256 				  skb_network_header(skb);
257 	serr->port = port;
258 
259 	__skb_pull(skb, payload - skb->data);
260 	skb_reset_transport_header(skb);
261 
262 	if (sock_queue_err_skb(sk, skb))
263 		kfree_skb(skb);
264 }
265 
266 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
267 {
268 	const struct ipv6_pinfo *np = inet6_sk(sk);
269 	struct sock_exterr_skb *serr;
270 	struct ipv6hdr *iph;
271 	struct sk_buff *skb;
272 
273 	if (!np->recverr)
274 		return;
275 
276 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
277 	if (!skb)
278 		return;
279 
280 	skb->protocol = htons(ETH_P_IPV6);
281 
282 	skb_put(skb, sizeof(struct ipv6hdr));
283 	skb_reset_network_header(skb);
284 	iph = ipv6_hdr(skb);
285 	iph->daddr = fl6->daddr;
286 
287 	serr = SKB_EXT_ERR(skb);
288 	serr->ee.ee_errno = err;
289 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
290 	serr->ee.ee_type = 0;
291 	serr->ee.ee_code = 0;
292 	serr->ee.ee_pad = 0;
293 	serr->ee.ee_info = info;
294 	serr->ee.ee_data = 0;
295 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
296 	serr->port = fl6->fl6_dport;
297 
298 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
299 	skb_reset_transport_header(skb);
300 
301 	if (sock_queue_err_skb(sk, skb))
302 		kfree_skb(skb);
303 }
304 
305 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
306 {
307 	struct ipv6_pinfo *np = inet6_sk(sk);
308 	struct ipv6hdr *iph;
309 	struct sk_buff *skb;
310 	struct ip6_mtuinfo *mtu_info;
311 
312 	if (!np->rxopt.bits.rxpmtu)
313 		return;
314 
315 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
316 	if (!skb)
317 		return;
318 
319 	skb_put(skb, sizeof(struct ipv6hdr));
320 	skb_reset_network_header(skb);
321 	iph = ipv6_hdr(skb);
322 	iph->daddr = fl6->daddr;
323 
324 	mtu_info = IP6CBMTU(skb);
325 
326 	mtu_info->ip6m_mtu = mtu;
327 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
328 	mtu_info->ip6m_addr.sin6_port = 0;
329 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
330 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
331 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
332 
333 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
334 	skb_reset_transport_header(skb);
335 
336 	skb = xchg(&np->rxpmtu, skb);
337 	kfree_skb(skb);
338 }
339 
340 /* For some errors we have valid addr_offset even with zero payload and
341  * zero port. Also, addr_offset should be supported if port is set.
342  */
343 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
344 {
345 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
346 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
347 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
348 }
349 
350 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
351  *
352  * At one point, excluding local errors was a quick test to identify icmp/icmp6
353  * errors. This is no longer true, but the test remained, so the v6 stack,
354  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
355  *
356  * Timestamp code paths do not initialize the fields expected by cmsg:
357  * the PKTINFO fields in skb->cb[]. Fill those in here.
358  */
359 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
360 				      struct sock_exterr_skb *serr)
361 {
362 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
363 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
364 		return true;
365 
366 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
367 		return false;
368 
369 	if (!skb->dev)
370 		return false;
371 
372 	if (skb->protocol == htons(ETH_P_IPV6))
373 		IP6CB(skb)->iif = skb->dev->ifindex;
374 	else
375 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
376 
377 	return true;
378 }
379 
380 /*
381  *	Handle MSG_ERRQUEUE
382  */
383 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
384 {
385 	struct ipv6_pinfo *np = inet6_sk(sk);
386 	struct sock_exterr_skb *serr;
387 	struct sk_buff *skb;
388 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
389 	struct {
390 		struct sock_extended_err ee;
391 		struct sockaddr_in6	 offender;
392 	} errhdr;
393 	int err;
394 	int copied;
395 
396 	err = -EAGAIN;
397 	skb = sock_dequeue_err_skb(sk);
398 	if (!skb)
399 		goto out;
400 
401 	copied = skb->len;
402 	if (copied > len) {
403 		msg->msg_flags |= MSG_TRUNC;
404 		copied = len;
405 	}
406 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
407 	if (err)
408 		goto out_free_skb;
409 
410 	sock_recv_timestamp(msg, sk, skb);
411 
412 	serr = SKB_EXT_ERR(skb);
413 
414 	if (sin && ipv6_datagram_support_addr(serr)) {
415 		const unsigned char *nh = skb_network_header(skb);
416 		sin->sin6_family = AF_INET6;
417 		sin->sin6_flowinfo = 0;
418 		sin->sin6_port = serr->port;
419 		if (skb->protocol == htons(ETH_P_IPV6)) {
420 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
421 								  struct ipv6hdr, daddr);
422 			sin->sin6_addr = ip6h->daddr;
423 			if (np->sndflow)
424 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
425 			sin->sin6_scope_id =
426 				ipv6_iface_scope_id(&sin->sin6_addr,
427 						    IP6CB(skb)->iif);
428 		} else {
429 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
430 					       &sin->sin6_addr);
431 			sin->sin6_scope_id = 0;
432 		}
433 		*addr_len = sizeof(*sin);
434 	}
435 
436 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
437 	sin = &errhdr.offender;
438 	memset(sin, 0, sizeof(*sin));
439 
440 	if (ip6_datagram_support_cmsg(skb, serr)) {
441 		sin->sin6_family = AF_INET6;
442 		if (np->rxopt.all)
443 			ip6_datagram_recv_common_ctl(sk, msg, skb);
444 		if (skb->protocol == htons(ETH_P_IPV6)) {
445 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
446 			if (np->rxopt.all)
447 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
448 			sin->sin6_scope_id =
449 				ipv6_iface_scope_id(&sin->sin6_addr,
450 						    IP6CB(skb)->iif);
451 		} else {
452 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
453 					       &sin->sin6_addr);
454 			if (inet_sk(sk)->cmsg_flags)
455 				ip_cmsg_recv(msg, skb);
456 		}
457 	}
458 
459 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
460 
461 	/* Now we could try to dump offended packet options */
462 
463 	msg->msg_flags |= MSG_ERRQUEUE;
464 	err = copied;
465 
466 out_free_skb:
467 	kfree_skb(skb);
468 out:
469 	return err;
470 }
471 EXPORT_SYMBOL_GPL(ipv6_recv_error);
472 
473 /*
474  *	Handle IPV6_RECVPATHMTU
475  */
476 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
477 		     int *addr_len)
478 {
479 	struct ipv6_pinfo *np = inet6_sk(sk);
480 	struct sk_buff *skb;
481 	struct ip6_mtuinfo mtu_info;
482 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
483 	int err;
484 	int copied;
485 
486 	err = -EAGAIN;
487 	skb = xchg(&np->rxpmtu, NULL);
488 	if (!skb)
489 		goto out;
490 
491 	copied = skb->len;
492 	if (copied > len) {
493 		msg->msg_flags |= MSG_TRUNC;
494 		copied = len;
495 	}
496 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
497 	if (err)
498 		goto out_free_skb;
499 
500 	sock_recv_timestamp(msg, sk, skb);
501 
502 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
503 
504 	if (sin) {
505 		sin->sin6_family = AF_INET6;
506 		sin->sin6_flowinfo = 0;
507 		sin->sin6_port = 0;
508 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
509 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
510 		*addr_len = sizeof(*sin);
511 	}
512 
513 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
514 
515 	err = copied;
516 
517 out_free_skb:
518 	kfree_skb(skb);
519 out:
520 	return err;
521 }
522 
523 
524 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
525 				 struct sk_buff *skb)
526 {
527 	struct ipv6_pinfo *np = inet6_sk(sk);
528 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
529 
530 	if (np->rxopt.bits.rxinfo) {
531 		struct in6_pktinfo src_info;
532 
533 		if (is_ipv6) {
534 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
535 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
536 		} else {
537 			src_info.ipi6_ifindex =
538 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
539 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
540 					       &src_info.ipi6_addr);
541 		}
542 
543 		if (src_info.ipi6_ifindex >= 0)
544 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
545 				 sizeof(src_info), &src_info);
546 	}
547 }
548 
549 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
550 				    struct sk_buff *skb)
551 {
552 	struct ipv6_pinfo *np = inet6_sk(sk);
553 	struct inet6_skb_parm *opt = IP6CB(skb);
554 	unsigned char *nh = skb_network_header(skb);
555 
556 	if (np->rxopt.bits.rxhlim) {
557 		int hlim = ipv6_hdr(skb)->hop_limit;
558 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
559 	}
560 
561 	if (np->rxopt.bits.rxtclass) {
562 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
563 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
564 	}
565 
566 	if (np->rxopt.bits.rxflow) {
567 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
568 		if (flowinfo)
569 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
570 	}
571 
572 	/* HbH is allowed only once */
573 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
574 		u8 *ptr = nh + sizeof(struct ipv6hdr);
575 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
576 	}
577 
578 	if (opt->lastopt &&
579 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
580 		/*
581 		 * Silly enough, but we need to reparse in order to
582 		 * report extension headers (except for HbH)
583 		 * in order.
584 		 *
585 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
586 		 * (and WILL NOT be) defined because
587 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
588 		 */
589 		unsigned int off = sizeof(struct ipv6hdr);
590 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
591 
592 		while (off <= opt->lastopt) {
593 			unsigned int len;
594 			u8 *ptr = nh + off;
595 
596 			switch (nexthdr) {
597 			case IPPROTO_DSTOPTS:
598 				nexthdr = ptr[0];
599 				len = (ptr[1] + 1) << 3;
600 				if (np->rxopt.bits.dstopts)
601 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
602 				break;
603 			case IPPROTO_ROUTING:
604 				nexthdr = ptr[0];
605 				len = (ptr[1] + 1) << 3;
606 				if (np->rxopt.bits.srcrt)
607 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
608 				break;
609 			case IPPROTO_AH:
610 				nexthdr = ptr[0];
611 				len = (ptr[1] + 2) << 2;
612 				break;
613 			default:
614 				nexthdr = ptr[0];
615 				len = (ptr[1] + 1) << 3;
616 				break;
617 			}
618 
619 			off += len;
620 		}
621 	}
622 
623 	/* socket options in old style */
624 	if (np->rxopt.bits.rxoinfo) {
625 		struct in6_pktinfo src_info;
626 
627 		src_info.ipi6_ifindex = opt->iif;
628 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
629 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
630 	}
631 	if (np->rxopt.bits.rxohlim) {
632 		int hlim = ipv6_hdr(skb)->hop_limit;
633 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
634 	}
635 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
636 		u8 *ptr = nh + sizeof(struct ipv6hdr);
637 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
638 	}
639 	if (np->rxopt.bits.odstopts && opt->dst0) {
640 		u8 *ptr = nh + opt->dst0;
641 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
642 	}
643 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
644 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
645 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
646 	}
647 	if (np->rxopt.bits.odstopts && opt->dst1) {
648 		u8 *ptr = nh + opt->dst1;
649 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
650 	}
651 	if (np->rxopt.bits.rxorigdstaddr) {
652 		struct sockaddr_in6 sin6;
653 		__be16 *ports = (__be16 *) skb_transport_header(skb);
654 
655 		if (skb_transport_offset(skb) + 4 <= skb->len) {
656 			/* All current transport protocols have the port numbers in the
657 			 * first four bytes of the transport header and this function is
658 			 * written with this assumption in mind.
659 			 */
660 
661 			sin6.sin6_family = AF_INET6;
662 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
663 			sin6.sin6_port = ports[1];
664 			sin6.sin6_flowinfo = 0;
665 			sin6.sin6_scope_id =
666 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
667 						    opt->iif);
668 
669 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
670 		}
671 	}
672 }
673 
674 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
675 			  struct sk_buff *skb)
676 {
677 	ip6_datagram_recv_common_ctl(sk, msg, skb);
678 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
679 }
680 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
681 
682 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
683 			  struct msghdr *msg, struct flowi6 *fl6,
684 			  struct ipv6_txoptions *opt,
685 			  int *hlimit, int *tclass, int *dontfrag)
686 {
687 	struct in6_pktinfo *src_info;
688 	struct cmsghdr *cmsg;
689 	struct ipv6_rt_hdr *rthdr;
690 	struct ipv6_opt_hdr *hdr;
691 	int len;
692 	int err = 0;
693 
694 	for_each_cmsghdr(cmsg, msg) {
695 		int addr_type;
696 
697 		if (!CMSG_OK(msg, cmsg)) {
698 			err = -EINVAL;
699 			goto exit_f;
700 		}
701 
702 		if (cmsg->cmsg_level != SOL_IPV6)
703 			continue;
704 
705 		switch (cmsg->cmsg_type) {
706 		case IPV6_PKTINFO:
707 		case IPV6_2292PKTINFO:
708 		    {
709 			struct net_device *dev = NULL;
710 
711 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
712 				err = -EINVAL;
713 				goto exit_f;
714 			}
715 
716 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
717 
718 			if (src_info->ipi6_ifindex) {
719 				if (fl6->flowi6_oif &&
720 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
721 					return -EINVAL;
722 				fl6->flowi6_oif = src_info->ipi6_ifindex;
723 			}
724 
725 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
726 
727 			rcu_read_lock();
728 			if (fl6->flowi6_oif) {
729 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
730 				if (!dev) {
731 					rcu_read_unlock();
732 					return -ENODEV;
733 				}
734 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
735 				rcu_read_unlock();
736 				return -EINVAL;
737 			}
738 
739 			if (addr_type != IPV6_ADDR_ANY) {
740 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
741 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
742 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
743 						   strict ? dev : NULL, 0) &&
744 				    !ipv6_chk_acast_addr_src(net, dev,
745 							     &src_info->ipi6_addr))
746 					err = -EINVAL;
747 				else
748 					fl6->saddr = src_info->ipi6_addr;
749 			}
750 
751 			rcu_read_unlock();
752 
753 			if (err)
754 				goto exit_f;
755 
756 			break;
757 		    }
758 
759 		case IPV6_FLOWINFO:
760 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
761 				err = -EINVAL;
762 				goto exit_f;
763 			}
764 
765 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
766 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
767 					err = -EINVAL;
768 					goto exit_f;
769 				}
770 			}
771 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
772 			break;
773 
774 		case IPV6_2292HOPOPTS:
775 		case IPV6_HOPOPTS:
776 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
777 				err = -EINVAL;
778 				goto exit_f;
779 			}
780 
781 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
782 			len = ((hdr->hdrlen + 1) << 3);
783 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
784 				err = -EINVAL;
785 				goto exit_f;
786 			}
787 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
788 				err = -EPERM;
789 				goto exit_f;
790 			}
791 			opt->opt_nflen += len;
792 			opt->hopopt = hdr;
793 			break;
794 
795 		case IPV6_2292DSTOPTS:
796 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
797 				err = -EINVAL;
798 				goto exit_f;
799 			}
800 
801 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
802 			len = ((hdr->hdrlen + 1) << 3);
803 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
804 				err = -EINVAL;
805 				goto exit_f;
806 			}
807 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
808 				err = -EPERM;
809 				goto exit_f;
810 			}
811 			if (opt->dst1opt) {
812 				err = -EINVAL;
813 				goto exit_f;
814 			}
815 			opt->opt_flen += len;
816 			opt->dst1opt = hdr;
817 			break;
818 
819 		case IPV6_DSTOPTS:
820 		case IPV6_RTHDRDSTOPTS:
821 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
822 				err = -EINVAL;
823 				goto exit_f;
824 			}
825 
826 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
827 			len = ((hdr->hdrlen + 1) << 3);
828 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
829 				err = -EINVAL;
830 				goto exit_f;
831 			}
832 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
833 				err = -EPERM;
834 				goto exit_f;
835 			}
836 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
837 				opt->opt_flen += len;
838 				opt->dst1opt = hdr;
839 			} else {
840 				opt->opt_nflen += len;
841 				opt->dst0opt = hdr;
842 			}
843 			break;
844 
845 		case IPV6_2292RTHDR:
846 		case IPV6_RTHDR:
847 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
848 				err = -EINVAL;
849 				goto exit_f;
850 			}
851 
852 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
853 
854 			switch (rthdr->type) {
855 #if IS_ENABLED(CONFIG_IPV6_MIP6)
856 			case IPV6_SRCRT_TYPE_2:
857 				if (rthdr->hdrlen != 2 ||
858 				    rthdr->segments_left != 1) {
859 					err = -EINVAL;
860 					goto exit_f;
861 				}
862 				break;
863 #endif
864 			default:
865 				err = -EINVAL;
866 				goto exit_f;
867 			}
868 
869 			len = ((rthdr->hdrlen + 1) << 3);
870 
871 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
872 				err = -EINVAL;
873 				goto exit_f;
874 			}
875 
876 			/* segments left must also match */
877 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
878 				err = -EINVAL;
879 				goto exit_f;
880 			}
881 
882 			opt->opt_nflen += len;
883 			opt->srcrt = rthdr;
884 
885 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
886 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
887 
888 				opt->opt_nflen += dsthdrlen;
889 				opt->dst0opt = opt->dst1opt;
890 				opt->dst1opt = NULL;
891 				opt->opt_flen -= dsthdrlen;
892 			}
893 
894 			break;
895 
896 		case IPV6_2292HOPLIMIT:
897 		case IPV6_HOPLIMIT:
898 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
899 				err = -EINVAL;
900 				goto exit_f;
901 			}
902 
903 			*hlimit = *(int *)CMSG_DATA(cmsg);
904 			if (*hlimit < -1 || *hlimit > 0xff) {
905 				err = -EINVAL;
906 				goto exit_f;
907 			}
908 
909 			break;
910 
911 		case IPV6_TCLASS:
912 		    {
913 			int tc;
914 
915 			err = -EINVAL;
916 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
917 				goto exit_f;
918 
919 			tc = *(int *)CMSG_DATA(cmsg);
920 			if (tc < -1 || tc > 0xff)
921 				goto exit_f;
922 
923 			err = 0;
924 			*tclass = tc;
925 
926 			break;
927 		    }
928 
929 		case IPV6_DONTFRAG:
930 		    {
931 			int df;
932 
933 			err = -EINVAL;
934 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
935 				goto exit_f;
936 
937 			df = *(int *)CMSG_DATA(cmsg);
938 			if (df < 0 || df > 1)
939 				goto exit_f;
940 
941 			err = 0;
942 			*dontfrag = df;
943 
944 			break;
945 		    }
946 		default:
947 			net_dbg_ratelimited("invalid cmsg type: %d\n",
948 					    cmsg->cmsg_type);
949 			err = -EINVAL;
950 			goto exit_f;
951 		}
952 	}
953 
954 exit_f:
955 	return err;
956 }
957 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
958 
959 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
960 			     __u16 srcp, __u16 destp, int bucket)
961 {
962 	const struct in6_addr *dest, *src;
963 
964 	dest  = &sp->sk_v6_daddr;
965 	src   = &sp->sk_v6_rcv_saddr;
966 	seq_printf(seq,
967 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
968 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
969 		   bucket,
970 		   src->s6_addr32[0], src->s6_addr32[1],
971 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
972 		   dest->s6_addr32[0], dest->s6_addr32[1],
973 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
974 		   sp->sk_state,
975 		   sk_wmem_alloc_get(sp),
976 		   sk_rmem_alloc_get(sp),
977 		   0, 0L, 0,
978 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
979 		   0,
980 		   sock_i_ino(sp),
981 		   atomic_read(&sp->sk_refcnt), sp,
982 		   atomic_read(&sp->sk_drops));
983 }
984