xref: /openbmc/linux/net/ipv6/datagram.c (revision 32981ea5)
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 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 
58 	if (!fl6->flowi6_oif)
59 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
60 
61 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr))
62 		fl6->flowi6_oif = np->mcast_oif;
63 
64 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
65 }
66 
67 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
68 {
69 	struct ip6_flowlabel *flowlabel = NULL;
70 	struct in6_addr *final_p, final;
71 	struct ipv6_txoptions *opt;
72 	struct dst_entry *dst;
73 	struct inet_sock *inet = inet_sk(sk);
74 	struct ipv6_pinfo *np = inet6_sk(sk);
75 	struct flowi6 fl6;
76 	int err = 0;
77 
78 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
79 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
80 		if (!flowlabel)
81 			return -EINVAL;
82 	}
83 	ip6_datagram_flow_key_init(&fl6, sk);
84 
85 	rcu_read_lock();
86 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
87 	final_p = fl6_update_dst(&fl6, opt, &final);
88 	rcu_read_unlock();
89 
90 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
91 	if (IS_ERR(dst)) {
92 		err = PTR_ERR(dst);
93 		goto out;
94 	}
95 
96 	if (fix_sk_saddr) {
97 		if (ipv6_addr_any(&np->saddr))
98 			np->saddr = fl6.saddr;
99 
100 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
101 			sk->sk_v6_rcv_saddr = fl6.saddr;
102 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
103 			if (sk->sk_prot->rehash)
104 				sk->sk_prot->rehash(sk);
105 		}
106 	}
107 
108 	ip6_dst_store(sk, dst,
109 		      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
110 		      &sk->sk_v6_daddr : NULL,
111 #ifdef CONFIG_IPV6_SUBTREES
112 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
113 		      &np->saddr :
114 #endif
115 		      NULL);
116 
117 out:
118 	fl6_sock_release(flowlabel);
119 	return err;
120 }
121 
122 void ip6_datagram_release_cb(struct sock *sk)
123 {
124 	struct dst_entry *dst;
125 
126 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
127 		return;
128 
129 	rcu_read_lock();
130 	dst = __sk_dst_get(sk);
131 	if (!dst || !dst->obsolete ||
132 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
133 		rcu_read_unlock();
134 		return;
135 	}
136 	rcu_read_unlock();
137 
138 	ip6_datagram_dst_update(sk, false);
139 }
140 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
141 
142 static int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
143 {
144 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
145 	struct inet_sock	*inet = inet_sk(sk);
146 	struct ipv6_pinfo	*np = inet6_sk(sk);
147 	struct in6_addr		*daddr;
148 	int			addr_type;
149 	int			err;
150 	__be32			fl6_flowlabel = 0;
151 
152 	if (usin->sin6_family == AF_INET) {
153 		if (__ipv6_only_sock(sk))
154 			return -EAFNOSUPPORT;
155 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
156 		goto ipv4_connected;
157 	}
158 
159 	if (addr_len < SIN6_LEN_RFC2133)
160 		return -EINVAL;
161 
162 	if (usin->sin6_family != AF_INET6)
163 		return -EAFNOSUPPORT;
164 
165 	if (np->sndflow)
166 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
167 
168 	addr_type = ipv6_addr_type(&usin->sin6_addr);
169 
170 	if (addr_type == IPV6_ADDR_ANY) {
171 		/*
172 		 *	connect to self
173 		 */
174 		usin->sin6_addr.s6_addr[15] = 0x01;
175 	}
176 
177 	daddr = &usin->sin6_addr;
178 
179 	if (addr_type == IPV6_ADDR_MAPPED) {
180 		struct sockaddr_in sin;
181 
182 		if (__ipv6_only_sock(sk)) {
183 			err = -ENETUNREACH;
184 			goto out;
185 		}
186 		sin.sin_family = AF_INET;
187 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
188 		sin.sin_port = usin->sin6_port;
189 
190 		err = __ip4_datagram_connect(sk,
191 					     (struct sockaddr *) &sin,
192 					     sizeof(sin));
193 
194 ipv4_connected:
195 		if (err)
196 			goto out;
197 
198 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
199 
200 		if (ipv6_addr_any(&np->saddr) ||
201 		    ipv6_mapped_addr_any(&np->saddr))
202 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
203 
204 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
205 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
206 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
207 					       &sk->sk_v6_rcv_saddr);
208 			if (sk->sk_prot->rehash)
209 				sk->sk_prot->rehash(sk);
210 		}
211 
212 		goto out;
213 	}
214 
215 	if (__ipv6_addr_needs_scope_id(addr_type)) {
216 		if (addr_len >= sizeof(struct sockaddr_in6) &&
217 		    usin->sin6_scope_id) {
218 			if (sk->sk_bound_dev_if &&
219 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
220 				err = -EINVAL;
221 				goto out;
222 			}
223 			sk->sk_bound_dev_if = usin->sin6_scope_id;
224 		}
225 
226 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
227 			sk->sk_bound_dev_if = np->mcast_oif;
228 
229 		/* Connect to link-local address requires an interface */
230 		if (!sk->sk_bound_dev_if) {
231 			err = -EINVAL;
232 			goto out;
233 		}
234 	}
235 
236 	sk->sk_v6_daddr = *daddr;
237 	np->flow_label = fl6_flowlabel;
238 
239 	inet->inet_dport = usin->sin6_port;
240 
241 	/*
242 	 *	Check for a route to destination an obtain the
243 	 *	destination cache for it.
244 	 */
245 
246 	err = ip6_datagram_dst_update(sk, true);
247 	if (err)
248 		goto out;
249 
250 	sk->sk_state = TCP_ESTABLISHED;
251 	sk_set_txhash(sk);
252 out:
253 	return err;
254 }
255 
256 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
257 {
258 	int res;
259 
260 	lock_sock(sk);
261 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
262 	release_sock(sk);
263 	return res;
264 }
265 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
266 
267 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
268 				 int addr_len)
269 {
270 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
271 	if (sin6->sin6_family != AF_INET6)
272 		return -EAFNOSUPPORT;
273 	return ip6_datagram_connect(sk, uaddr, addr_len);
274 }
275 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
276 
277 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
278 		     __be16 port, u32 info, u8 *payload)
279 {
280 	struct ipv6_pinfo *np  = inet6_sk(sk);
281 	struct icmp6hdr *icmph = icmp6_hdr(skb);
282 	struct sock_exterr_skb *serr;
283 
284 	if (!np->recverr)
285 		return;
286 
287 	skb = skb_clone(skb, GFP_ATOMIC);
288 	if (!skb)
289 		return;
290 
291 	skb->protocol = htons(ETH_P_IPV6);
292 
293 	serr = SKB_EXT_ERR(skb);
294 	serr->ee.ee_errno = err;
295 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
296 	serr->ee.ee_type = icmph->icmp6_type;
297 	serr->ee.ee_code = icmph->icmp6_code;
298 	serr->ee.ee_pad = 0;
299 	serr->ee.ee_info = info;
300 	serr->ee.ee_data = 0;
301 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
302 				  skb_network_header(skb);
303 	serr->port = port;
304 
305 	__skb_pull(skb, payload - skb->data);
306 	skb_reset_transport_header(skb);
307 
308 	if (sock_queue_err_skb(sk, skb))
309 		kfree_skb(skb);
310 }
311 
312 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
313 {
314 	const struct ipv6_pinfo *np = inet6_sk(sk);
315 	struct sock_exterr_skb *serr;
316 	struct ipv6hdr *iph;
317 	struct sk_buff *skb;
318 
319 	if (!np->recverr)
320 		return;
321 
322 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
323 	if (!skb)
324 		return;
325 
326 	skb->protocol = htons(ETH_P_IPV6);
327 
328 	skb_put(skb, sizeof(struct ipv6hdr));
329 	skb_reset_network_header(skb);
330 	iph = ipv6_hdr(skb);
331 	iph->daddr = fl6->daddr;
332 
333 	serr = SKB_EXT_ERR(skb);
334 	serr->ee.ee_errno = err;
335 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
336 	serr->ee.ee_type = 0;
337 	serr->ee.ee_code = 0;
338 	serr->ee.ee_pad = 0;
339 	serr->ee.ee_info = info;
340 	serr->ee.ee_data = 0;
341 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
342 	serr->port = fl6->fl6_dport;
343 
344 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
345 	skb_reset_transport_header(skb);
346 
347 	if (sock_queue_err_skb(sk, skb))
348 		kfree_skb(skb);
349 }
350 
351 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
352 {
353 	struct ipv6_pinfo *np = inet6_sk(sk);
354 	struct ipv6hdr *iph;
355 	struct sk_buff *skb;
356 	struct ip6_mtuinfo *mtu_info;
357 
358 	if (!np->rxopt.bits.rxpmtu)
359 		return;
360 
361 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
362 	if (!skb)
363 		return;
364 
365 	skb_put(skb, sizeof(struct ipv6hdr));
366 	skb_reset_network_header(skb);
367 	iph = ipv6_hdr(skb);
368 	iph->daddr = fl6->daddr;
369 
370 	mtu_info = IP6CBMTU(skb);
371 
372 	mtu_info->ip6m_mtu = mtu;
373 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
374 	mtu_info->ip6m_addr.sin6_port = 0;
375 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
376 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
377 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
378 
379 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
380 	skb_reset_transport_header(skb);
381 
382 	skb = xchg(&np->rxpmtu, skb);
383 	kfree_skb(skb);
384 }
385 
386 /* For some errors we have valid addr_offset even with zero payload and
387  * zero port. Also, addr_offset should be supported if port is set.
388  */
389 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
390 {
391 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
392 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
393 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
394 }
395 
396 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
397  *
398  * At one point, excluding local errors was a quick test to identify icmp/icmp6
399  * errors. This is no longer true, but the test remained, so the v6 stack,
400  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
401  *
402  * Timestamp code paths do not initialize the fields expected by cmsg:
403  * the PKTINFO fields in skb->cb[]. Fill those in here.
404  */
405 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
406 				      struct sock_exterr_skb *serr)
407 {
408 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
409 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
410 		return true;
411 
412 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
413 		return false;
414 
415 	if (!skb->dev)
416 		return false;
417 
418 	if (skb->protocol == htons(ETH_P_IPV6))
419 		IP6CB(skb)->iif = skb->dev->ifindex;
420 	else
421 		PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex;
422 
423 	return true;
424 }
425 
426 /*
427  *	Handle MSG_ERRQUEUE
428  */
429 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
430 {
431 	struct ipv6_pinfo *np = inet6_sk(sk);
432 	struct sock_exterr_skb *serr;
433 	struct sk_buff *skb;
434 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
435 	struct {
436 		struct sock_extended_err ee;
437 		struct sockaddr_in6	 offender;
438 	} errhdr;
439 	int err;
440 	int copied;
441 
442 	err = -EAGAIN;
443 	skb = sock_dequeue_err_skb(sk);
444 	if (!skb)
445 		goto out;
446 
447 	copied = skb->len;
448 	if (copied > len) {
449 		msg->msg_flags |= MSG_TRUNC;
450 		copied = len;
451 	}
452 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
453 	if (unlikely(err)) {
454 		kfree_skb(skb);
455 		return err;
456 	}
457 	sock_recv_timestamp(msg, sk, skb);
458 
459 	serr = SKB_EXT_ERR(skb);
460 
461 	if (sin && ipv6_datagram_support_addr(serr)) {
462 		const unsigned char *nh = skb_network_header(skb);
463 		sin->sin6_family = AF_INET6;
464 		sin->sin6_flowinfo = 0;
465 		sin->sin6_port = serr->port;
466 		if (skb->protocol == htons(ETH_P_IPV6)) {
467 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
468 								  struct ipv6hdr, daddr);
469 			sin->sin6_addr = ip6h->daddr;
470 			if (np->sndflow)
471 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
472 			sin->sin6_scope_id =
473 				ipv6_iface_scope_id(&sin->sin6_addr,
474 						    IP6CB(skb)->iif);
475 		} else {
476 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
477 					       &sin->sin6_addr);
478 			sin->sin6_scope_id = 0;
479 		}
480 		*addr_len = sizeof(*sin);
481 	}
482 
483 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
484 	sin = &errhdr.offender;
485 	memset(sin, 0, sizeof(*sin));
486 
487 	if (ip6_datagram_support_cmsg(skb, serr)) {
488 		sin->sin6_family = AF_INET6;
489 		if (np->rxopt.all)
490 			ip6_datagram_recv_common_ctl(sk, msg, skb);
491 		if (skb->protocol == htons(ETH_P_IPV6)) {
492 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
493 			if (np->rxopt.all)
494 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
495 			sin->sin6_scope_id =
496 				ipv6_iface_scope_id(&sin->sin6_addr,
497 						    IP6CB(skb)->iif);
498 		} else {
499 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
500 					       &sin->sin6_addr);
501 			if (inet_sk(sk)->cmsg_flags)
502 				ip_cmsg_recv(msg, skb);
503 		}
504 	}
505 
506 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
507 
508 	/* Now we could try to dump offended packet options */
509 
510 	msg->msg_flags |= MSG_ERRQUEUE;
511 	err = copied;
512 
513 	consume_skb(skb);
514 out:
515 	return err;
516 }
517 EXPORT_SYMBOL_GPL(ipv6_recv_error);
518 
519 /*
520  *	Handle IPV6_RECVPATHMTU
521  */
522 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
523 		     int *addr_len)
524 {
525 	struct ipv6_pinfo *np = inet6_sk(sk);
526 	struct sk_buff *skb;
527 	struct ip6_mtuinfo mtu_info;
528 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
529 	int err;
530 	int copied;
531 
532 	err = -EAGAIN;
533 	skb = xchg(&np->rxpmtu, NULL);
534 	if (!skb)
535 		goto out;
536 
537 	copied = skb->len;
538 	if (copied > len) {
539 		msg->msg_flags |= MSG_TRUNC;
540 		copied = len;
541 	}
542 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
543 	if (err)
544 		goto out_free_skb;
545 
546 	sock_recv_timestamp(msg, sk, skb);
547 
548 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
549 
550 	if (sin) {
551 		sin->sin6_family = AF_INET6;
552 		sin->sin6_flowinfo = 0;
553 		sin->sin6_port = 0;
554 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
555 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
556 		*addr_len = sizeof(*sin);
557 	}
558 
559 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
560 
561 	err = copied;
562 
563 out_free_skb:
564 	kfree_skb(skb);
565 out:
566 	return err;
567 }
568 
569 
570 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
571 				 struct sk_buff *skb)
572 {
573 	struct ipv6_pinfo *np = inet6_sk(sk);
574 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
575 
576 	if (np->rxopt.bits.rxinfo) {
577 		struct in6_pktinfo src_info;
578 
579 		if (is_ipv6) {
580 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
581 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
582 		} else {
583 			src_info.ipi6_ifindex =
584 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
585 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
586 					       &src_info.ipi6_addr);
587 		}
588 
589 		if (src_info.ipi6_ifindex >= 0)
590 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
591 				 sizeof(src_info), &src_info);
592 	}
593 }
594 
595 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
596 				    struct sk_buff *skb)
597 {
598 	struct ipv6_pinfo *np = inet6_sk(sk);
599 	struct inet6_skb_parm *opt = IP6CB(skb);
600 	unsigned char *nh = skb_network_header(skb);
601 
602 	if (np->rxopt.bits.rxhlim) {
603 		int hlim = ipv6_hdr(skb)->hop_limit;
604 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
605 	}
606 
607 	if (np->rxopt.bits.rxtclass) {
608 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
609 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
610 	}
611 
612 	if (np->rxopt.bits.rxflow) {
613 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
614 		if (flowinfo)
615 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
616 	}
617 
618 	/* HbH is allowed only once */
619 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
620 		u8 *ptr = nh + sizeof(struct ipv6hdr);
621 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
622 	}
623 
624 	if (opt->lastopt &&
625 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
626 		/*
627 		 * Silly enough, but we need to reparse in order to
628 		 * report extension headers (except for HbH)
629 		 * in order.
630 		 *
631 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
632 		 * (and WILL NOT be) defined because
633 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
634 		 */
635 		unsigned int off = sizeof(struct ipv6hdr);
636 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
637 
638 		while (off <= opt->lastopt) {
639 			unsigned int len;
640 			u8 *ptr = nh + off;
641 
642 			switch (nexthdr) {
643 			case IPPROTO_DSTOPTS:
644 				nexthdr = ptr[0];
645 				len = (ptr[1] + 1) << 3;
646 				if (np->rxopt.bits.dstopts)
647 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
648 				break;
649 			case IPPROTO_ROUTING:
650 				nexthdr = ptr[0];
651 				len = (ptr[1] + 1) << 3;
652 				if (np->rxopt.bits.srcrt)
653 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
654 				break;
655 			case IPPROTO_AH:
656 				nexthdr = ptr[0];
657 				len = (ptr[1] + 2) << 2;
658 				break;
659 			default:
660 				nexthdr = ptr[0];
661 				len = (ptr[1] + 1) << 3;
662 				break;
663 			}
664 
665 			off += len;
666 		}
667 	}
668 
669 	/* socket options in old style */
670 	if (np->rxopt.bits.rxoinfo) {
671 		struct in6_pktinfo src_info;
672 
673 		src_info.ipi6_ifindex = opt->iif;
674 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
675 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
676 	}
677 	if (np->rxopt.bits.rxohlim) {
678 		int hlim = ipv6_hdr(skb)->hop_limit;
679 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
680 	}
681 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
682 		u8 *ptr = nh + sizeof(struct ipv6hdr);
683 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
684 	}
685 	if (np->rxopt.bits.odstopts && opt->dst0) {
686 		u8 *ptr = nh + opt->dst0;
687 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
688 	}
689 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
690 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
691 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
692 	}
693 	if (np->rxopt.bits.odstopts && opt->dst1) {
694 		u8 *ptr = nh + opt->dst1;
695 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
696 	}
697 	if (np->rxopt.bits.rxorigdstaddr) {
698 		struct sockaddr_in6 sin6;
699 		__be16 *ports = (__be16 *) skb_transport_header(skb);
700 
701 		if (skb_transport_offset(skb) + 4 <= skb->len) {
702 			/* All current transport protocols have the port numbers in the
703 			 * first four bytes of the transport header and this function is
704 			 * written with this assumption in mind.
705 			 */
706 
707 			sin6.sin6_family = AF_INET6;
708 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
709 			sin6.sin6_port = ports[1];
710 			sin6.sin6_flowinfo = 0;
711 			sin6.sin6_scope_id =
712 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
713 						    opt->iif);
714 
715 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
716 		}
717 	}
718 }
719 
720 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
721 			  struct sk_buff *skb)
722 {
723 	ip6_datagram_recv_common_ctl(sk, msg, skb);
724 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
725 }
726 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
727 
728 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
729 			  struct msghdr *msg, struct flowi6 *fl6,
730 			  struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc)
731 {
732 	struct in6_pktinfo *src_info;
733 	struct cmsghdr *cmsg;
734 	struct ipv6_rt_hdr *rthdr;
735 	struct ipv6_opt_hdr *hdr;
736 	struct ipv6_txoptions *opt = ipc6->opt;
737 	int len;
738 	int err = 0;
739 
740 	for_each_cmsghdr(cmsg, msg) {
741 		int addr_type;
742 
743 		if (!CMSG_OK(msg, cmsg)) {
744 			err = -EINVAL;
745 			goto exit_f;
746 		}
747 
748 		if (cmsg->cmsg_level == SOL_SOCKET) {
749 			err = __sock_cmsg_send(sk, msg, cmsg, sockc);
750 			if (err)
751 				return err;
752 			continue;
753 		}
754 
755 		if (cmsg->cmsg_level != SOL_IPV6)
756 			continue;
757 
758 		switch (cmsg->cmsg_type) {
759 		case IPV6_PKTINFO:
760 		case IPV6_2292PKTINFO:
761 		    {
762 			struct net_device *dev = NULL;
763 
764 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
765 				err = -EINVAL;
766 				goto exit_f;
767 			}
768 
769 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
770 
771 			if (src_info->ipi6_ifindex) {
772 				if (fl6->flowi6_oif &&
773 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
774 					return -EINVAL;
775 				fl6->flowi6_oif = src_info->ipi6_ifindex;
776 			}
777 
778 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
779 
780 			rcu_read_lock();
781 			if (fl6->flowi6_oif) {
782 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
783 				if (!dev) {
784 					rcu_read_unlock();
785 					return -ENODEV;
786 				}
787 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
788 				rcu_read_unlock();
789 				return -EINVAL;
790 			}
791 
792 			if (addr_type != IPV6_ADDR_ANY) {
793 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
794 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
795 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
796 						   strict ? dev : NULL, 0) &&
797 				    !ipv6_chk_acast_addr_src(net, dev,
798 							     &src_info->ipi6_addr))
799 					err = -EINVAL;
800 				else
801 					fl6->saddr = src_info->ipi6_addr;
802 			}
803 
804 			rcu_read_unlock();
805 
806 			if (err)
807 				goto exit_f;
808 
809 			break;
810 		    }
811 
812 		case IPV6_FLOWINFO:
813 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
814 				err = -EINVAL;
815 				goto exit_f;
816 			}
817 
818 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
819 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
820 					err = -EINVAL;
821 					goto exit_f;
822 				}
823 			}
824 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
825 			break;
826 
827 		case IPV6_2292HOPOPTS:
828 		case IPV6_HOPOPTS:
829 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
830 				err = -EINVAL;
831 				goto exit_f;
832 			}
833 
834 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
835 			len = ((hdr->hdrlen + 1) << 3);
836 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
837 				err = -EINVAL;
838 				goto exit_f;
839 			}
840 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
841 				err = -EPERM;
842 				goto exit_f;
843 			}
844 			opt->opt_nflen += len;
845 			opt->hopopt = hdr;
846 			break;
847 
848 		case IPV6_2292DSTOPTS:
849 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
850 				err = -EINVAL;
851 				goto exit_f;
852 			}
853 
854 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
855 			len = ((hdr->hdrlen + 1) << 3);
856 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
857 				err = -EINVAL;
858 				goto exit_f;
859 			}
860 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
861 				err = -EPERM;
862 				goto exit_f;
863 			}
864 			if (opt->dst1opt) {
865 				err = -EINVAL;
866 				goto exit_f;
867 			}
868 			opt->opt_flen += len;
869 			opt->dst1opt = hdr;
870 			break;
871 
872 		case IPV6_DSTOPTS:
873 		case IPV6_RTHDRDSTOPTS:
874 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
875 				err = -EINVAL;
876 				goto exit_f;
877 			}
878 
879 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
880 			len = ((hdr->hdrlen + 1) << 3);
881 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
882 				err = -EINVAL;
883 				goto exit_f;
884 			}
885 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
886 				err = -EPERM;
887 				goto exit_f;
888 			}
889 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
890 				opt->opt_flen += len;
891 				opt->dst1opt = hdr;
892 			} else {
893 				opt->opt_nflen += len;
894 				opt->dst0opt = hdr;
895 			}
896 			break;
897 
898 		case IPV6_2292RTHDR:
899 		case IPV6_RTHDR:
900 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
901 				err = -EINVAL;
902 				goto exit_f;
903 			}
904 
905 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
906 
907 			switch (rthdr->type) {
908 #if IS_ENABLED(CONFIG_IPV6_MIP6)
909 			case IPV6_SRCRT_TYPE_2:
910 				if (rthdr->hdrlen != 2 ||
911 				    rthdr->segments_left != 1) {
912 					err = -EINVAL;
913 					goto exit_f;
914 				}
915 				break;
916 #endif
917 			default:
918 				err = -EINVAL;
919 				goto exit_f;
920 			}
921 
922 			len = ((rthdr->hdrlen + 1) << 3);
923 
924 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
925 				err = -EINVAL;
926 				goto exit_f;
927 			}
928 
929 			/* segments left must also match */
930 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
931 				err = -EINVAL;
932 				goto exit_f;
933 			}
934 
935 			opt->opt_nflen += len;
936 			opt->srcrt = rthdr;
937 
938 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
939 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
940 
941 				opt->opt_nflen += dsthdrlen;
942 				opt->dst0opt = opt->dst1opt;
943 				opt->dst1opt = NULL;
944 				opt->opt_flen -= dsthdrlen;
945 			}
946 
947 			break;
948 
949 		case IPV6_2292HOPLIMIT:
950 		case IPV6_HOPLIMIT:
951 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
952 				err = -EINVAL;
953 				goto exit_f;
954 			}
955 
956 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
957 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
958 				err = -EINVAL;
959 				goto exit_f;
960 			}
961 
962 			break;
963 
964 		case IPV6_TCLASS:
965 		    {
966 			int tc;
967 
968 			err = -EINVAL;
969 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
970 				goto exit_f;
971 
972 			tc = *(int *)CMSG_DATA(cmsg);
973 			if (tc < -1 || tc > 0xff)
974 				goto exit_f;
975 
976 			err = 0;
977 			ipc6->tclass = tc;
978 
979 			break;
980 		    }
981 
982 		case IPV6_DONTFRAG:
983 		    {
984 			int df;
985 
986 			err = -EINVAL;
987 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
988 				goto exit_f;
989 
990 			df = *(int *)CMSG_DATA(cmsg);
991 			if (df < 0 || df > 1)
992 				goto exit_f;
993 
994 			err = 0;
995 			ipc6->dontfrag = df;
996 
997 			break;
998 		    }
999 		default:
1000 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1001 					    cmsg->cmsg_type);
1002 			err = -EINVAL;
1003 			goto exit_f;
1004 		}
1005 	}
1006 
1007 exit_f:
1008 	return err;
1009 }
1010 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1011 
1012 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1013 			     __u16 srcp, __u16 destp, int bucket)
1014 {
1015 	const struct in6_addr *dest, *src;
1016 
1017 	dest  = &sp->sk_v6_daddr;
1018 	src   = &sp->sk_v6_rcv_saddr;
1019 	seq_printf(seq,
1020 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1021 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1022 		   bucket,
1023 		   src->s6_addr32[0], src->s6_addr32[1],
1024 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1025 		   dest->s6_addr32[0], dest->s6_addr32[1],
1026 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1027 		   sp->sk_state,
1028 		   sk_wmem_alloc_get(sp),
1029 		   sk_rmem_alloc_get(sp),
1030 		   0, 0L, 0,
1031 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1032 		   0,
1033 		   sock_i_ino(sp),
1034 		   atomic_read(&sp->sk_refcnt), sp,
1035 		   atomic_read(&sp->sk_drops));
1036 }
1037