xref: /openbmc/linux/net/ipv6/datagram.c (revision 6c33a6f4)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	common UDP/RAW code
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  */
9 
10 #include <linux/capability.h>
11 #include <linux/errno.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/interrupt.h>
15 #include <linux/socket.h>
16 #include <linux/sockios.h>
17 #include <linux/in6.h>
18 #include <linux/ipv6.h>
19 #include <linux/route.h>
20 #include <linux/slab.h>
21 #include <linux/export.h>
22 
23 #include <net/ipv6.h>
24 #include <net/ndisc.h>
25 #include <net/addrconf.h>
26 #include <net/transp_v6.h>
27 #include <net/ip6_route.h>
28 #include <net/tcp_states.h>
29 #include <net/dsfield.h>
30 #include <net/sock_reuseport.h>
31 
32 #include <linux/errqueue.h>
33 #include <linux/uaccess.h>
34 
35 static bool ipv6_mapped_addr_any(const struct in6_addr *a)
36 {
37 	return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
38 }
39 
40 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk)
41 {
42 	struct inet_sock *inet = inet_sk(sk);
43 	struct ipv6_pinfo *np = inet6_sk(sk);
44 
45 	memset(fl6, 0, sizeof(*fl6));
46 	fl6->flowi6_proto = sk->sk_protocol;
47 	fl6->daddr = sk->sk_v6_daddr;
48 	fl6->saddr = np->saddr;
49 	fl6->flowi6_oif = sk->sk_bound_dev_if;
50 	fl6->flowi6_mark = sk->sk_mark;
51 	fl6->fl6_dport = inet->inet_dport;
52 	fl6->fl6_sport = inet->inet_sport;
53 	fl6->flowlabel = np->flow_label;
54 	fl6->flowi6_uid = sk->sk_uid;
55 
56 	if (!fl6->flowi6_oif)
57 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
58 
59 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr))
60 		fl6->flowi6_oif = np->mcast_oif;
61 
62 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
63 }
64 
65 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
66 {
67 	struct ip6_flowlabel *flowlabel = NULL;
68 	struct in6_addr *final_p, final;
69 	struct ipv6_txoptions *opt;
70 	struct dst_entry *dst;
71 	struct inet_sock *inet = inet_sk(sk);
72 	struct ipv6_pinfo *np = inet6_sk(sk);
73 	struct flowi6 fl6;
74 	int err = 0;
75 
76 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
77 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
78 		if (IS_ERR(flowlabel))
79 			return -EINVAL;
80 	}
81 	ip6_datagram_flow_key_init(&fl6, sk);
82 
83 	rcu_read_lock();
84 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
85 	final_p = fl6_update_dst(&fl6, opt, &final);
86 	rcu_read_unlock();
87 
88 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
89 	if (IS_ERR(dst)) {
90 		err = PTR_ERR(dst);
91 		goto out;
92 	}
93 
94 	if (fix_sk_saddr) {
95 		if (ipv6_addr_any(&np->saddr))
96 			np->saddr = fl6.saddr;
97 
98 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
99 			sk->sk_v6_rcv_saddr = fl6.saddr;
100 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
101 			if (sk->sk_prot->rehash)
102 				sk->sk_prot->rehash(sk);
103 		}
104 	}
105 
106 	ip6_sk_dst_store_flow(sk, dst, &fl6);
107 
108 out:
109 	fl6_sock_release(flowlabel);
110 	return err;
111 }
112 
113 void ip6_datagram_release_cb(struct sock *sk)
114 {
115 	struct dst_entry *dst;
116 
117 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
118 		return;
119 
120 	rcu_read_lock();
121 	dst = __sk_dst_get(sk);
122 	if (!dst || !dst->obsolete ||
123 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
124 		rcu_read_unlock();
125 		return;
126 	}
127 	rcu_read_unlock();
128 
129 	ip6_datagram_dst_update(sk, false);
130 }
131 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
132 
133 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
134 			   int addr_len)
135 {
136 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
137 	struct inet_sock	*inet = inet_sk(sk);
138 	struct ipv6_pinfo	*np = inet6_sk(sk);
139 	struct in6_addr		*daddr, old_daddr;
140 	__be32			fl6_flowlabel = 0;
141 	__be32			old_fl6_flowlabel;
142 	__be16			old_dport;
143 	int			addr_type;
144 	int			err;
145 
146 	if (usin->sin6_family == AF_INET) {
147 		if (__ipv6_only_sock(sk))
148 			return -EAFNOSUPPORT;
149 		err = __ip4_datagram_connect(sk, uaddr, addr_len);
150 		goto ipv4_connected;
151 	}
152 
153 	if (addr_len < SIN6_LEN_RFC2133)
154 		return -EINVAL;
155 
156 	if (usin->sin6_family != AF_INET6)
157 		return -EAFNOSUPPORT;
158 
159 	if (np->sndflow)
160 		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
161 
162 	if (ipv6_addr_any(&usin->sin6_addr)) {
163 		/*
164 		 *	connect to self
165 		 */
166 		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
167 			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
168 					       &usin->sin6_addr);
169 		else
170 			usin->sin6_addr = in6addr_loopback;
171 	}
172 
173 	addr_type = ipv6_addr_type(&usin->sin6_addr);
174 
175 	daddr = &usin->sin6_addr;
176 
177 	if (addr_type & IPV6_ADDR_MAPPED) {
178 		struct sockaddr_in sin;
179 
180 		if (__ipv6_only_sock(sk)) {
181 			err = -ENETUNREACH;
182 			goto out;
183 		}
184 		sin.sin_family = AF_INET;
185 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
186 		sin.sin_port = usin->sin6_port;
187 
188 		err = __ip4_datagram_connect(sk,
189 					     (struct sockaddr *) &sin,
190 					     sizeof(sin));
191 
192 ipv4_connected:
193 		if (err)
194 			goto out;
195 
196 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
197 
198 		if (ipv6_addr_any(&np->saddr) ||
199 		    ipv6_mapped_addr_any(&np->saddr))
200 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
201 
202 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
203 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
204 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
205 					       &sk->sk_v6_rcv_saddr);
206 			if (sk->sk_prot->rehash)
207 				sk->sk_prot->rehash(sk);
208 		}
209 
210 		goto out;
211 	}
212 
213 	if (__ipv6_addr_needs_scope_id(addr_type)) {
214 		if (addr_len >= sizeof(struct sockaddr_in6) &&
215 		    usin->sin6_scope_id) {
216 			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
217 				err = -EINVAL;
218 				goto out;
219 			}
220 			sk->sk_bound_dev_if = usin->sin6_scope_id;
221 		}
222 
223 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
224 			sk->sk_bound_dev_if = np->mcast_oif;
225 
226 		/* Connect to link-local address requires an interface */
227 		if (!sk->sk_bound_dev_if) {
228 			err = -EINVAL;
229 			goto out;
230 		}
231 	}
232 
233 	/* save the current peer information before updating it */
234 	old_daddr = sk->sk_v6_daddr;
235 	old_fl6_flowlabel = np->flow_label;
236 	old_dport = inet->inet_dport;
237 
238 	sk->sk_v6_daddr = *daddr;
239 	np->flow_label = fl6_flowlabel;
240 	inet->inet_dport = usin->sin6_port;
241 
242 	/*
243 	 *	Check for a route to destination an obtain the
244 	 *	destination cache for it.
245 	 */
246 
247 	err = ip6_datagram_dst_update(sk, true);
248 	if (err) {
249 		/* Restore the socket peer info, to keep it consistent with
250 		 * the old socket state
251 		 */
252 		sk->sk_v6_daddr = old_daddr;
253 		np->flow_label = old_fl6_flowlabel;
254 		inet->inet_dport = old_dport;
255 		goto out;
256 	}
257 
258 	reuseport_has_conns(sk, true);
259 	sk->sk_state = TCP_ESTABLISHED;
260 	sk_set_txhash(sk);
261 out:
262 	return err;
263 }
264 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
265 
266 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
267 {
268 	int res;
269 
270 	lock_sock(sk);
271 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
272 	release_sock(sk);
273 	return res;
274 }
275 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
276 
277 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
278 				 int addr_len)
279 {
280 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
281 	if (sin6->sin6_family != AF_INET6)
282 		return -EAFNOSUPPORT;
283 	return ip6_datagram_connect(sk, uaddr, addr_len);
284 }
285 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
286 
287 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
288 		     __be16 port, u32 info, u8 *payload)
289 {
290 	struct ipv6_pinfo *np  = inet6_sk(sk);
291 	struct icmp6hdr *icmph = icmp6_hdr(skb);
292 	struct sock_exterr_skb *serr;
293 
294 	if (!np->recverr)
295 		return;
296 
297 	skb = skb_clone(skb, GFP_ATOMIC);
298 	if (!skb)
299 		return;
300 
301 	skb->protocol = htons(ETH_P_IPV6);
302 
303 	serr = SKB_EXT_ERR(skb);
304 	serr->ee.ee_errno = err;
305 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
306 	serr->ee.ee_type = icmph->icmp6_type;
307 	serr->ee.ee_code = icmph->icmp6_code;
308 	serr->ee.ee_pad = 0;
309 	serr->ee.ee_info = info;
310 	serr->ee.ee_data = 0;
311 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
312 				  skb_network_header(skb);
313 	serr->port = port;
314 
315 	__skb_pull(skb, payload - skb->data);
316 	skb_reset_transport_header(skb);
317 
318 	if (sock_queue_err_skb(sk, skb))
319 		kfree_skb(skb);
320 }
321 
322 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
323 {
324 	const struct ipv6_pinfo *np = inet6_sk(sk);
325 	struct sock_exterr_skb *serr;
326 	struct ipv6hdr *iph;
327 	struct sk_buff *skb;
328 
329 	if (!np->recverr)
330 		return;
331 
332 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
333 	if (!skb)
334 		return;
335 
336 	skb->protocol = htons(ETH_P_IPV6);
337 
338 	skb_put(skb, sizeof(struct ipv6hdr));
339 	skb_reset_network_header(skb);
340 	iph = ipv6_hdr(skb);
341 	iph->daddr = fl6->daddr;
342 	ip6_flow_hdr(iph, 0, 0);
343 
344 	serr = SKB_EXT_ERR(skb);
345 	serr->ee.ee_errno = err;
346 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
347 	serr->ee.ee_type = 0;
348 	serr->ee.ee_code = 0;
349 	serr->ee.ee_pad = 0;
350 	serr->ee.ee_info = info;
351 	serr->ee.ee_data = 0;
352 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
353 	serr->port = fl6->fl6_dport;
354 
355 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
356 	skb_reset_transport_header(skb);
357 
358 	if (sock_queue_err_skb(sk, skb))
359 		kfree_skb(skb);
360 }
361 
362 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
363 {
364 	struct ipv6_pinfo *np = inet6_sk(sk);
365 	struct ipv6hdr *iph;
366 	struct sk_buff *skb;
367 	struct ip6_mtuinfo *mtu_info;
368 
369 	if (!np->rxopt.bits.rxpmtu)
370 		return;
371 
372 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
373 	if (!skb)
374 		return;
375 
376 	skb_put(skb, sizeof(struct ipv6hdr));
377 	skb_reset_network_header(skb);
378 	iph = ipv6_hdr(skb);
379 	iph->daddr = fl6->daddr;
380 
381 	mtu_info = IP6CBMTU(skb);
382 
383 	mtu_info->ip6m_mtu = mtu;
384 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
385 	mtu_info->ip6m_addr.sin6_port = 0;
386 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
387 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
388 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
389 
390 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
391 	skb_reset_transport_header(skb);
392 
393 	skb = xchg(&np->rxpmtu, skb);
394 	kfree_skb(skb);
395 }
396 
397 /* For some errors we have valid addr_offset even with zero payload and
398  * zero port. Also, addr_offset should be supported if port is set.
399  */
400 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
401 {
402 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
403 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
404 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
405 }
406 
407 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
408  *
409  * At one point, excluding local errors was a quick test to identify icmp/icmp6
410  * errors. This is no longer true, but the test remained, so the v6 stack,
411  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
412  */
413 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
414 				      struct sock_exterr_skb *serr)
415 {
416 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
417 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
418 		return true;
419 
420 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
421 		return false;
422 
423 	if (!IP6CB(skb)->iif)
424 		return false;
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[2], *ports;
703 
704 		ports = skb_header_pointer(skb, skb_transport_offset(skb),
705 					   sizeof(_ports), &_ports);
706 		if (ports) {
707 			/* All current transport protocols have the port numbers in the
708 			 * first four bytes of the transport header and this function is
709 			 * written with this assumption in mind.
710 			 */
711 			sin6.sin6_family = AF_INET6;
712 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
713 			sin6.sin6_port = ports[1];
714 			sin6.sin6_flowinfo = 0;
715 			sin6.sin6_scope_id =
716 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
717 						    opt->iif);
718 
719 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
720 		}
721 	}
722 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
723 		int val = opt->frag_max_size;
724 
725 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
726 	}
727 }
728 
729 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
730 			  struct sk_buff *skb)
731 {
732 	ip6_datagram_recv_common_ctl(sk, msg, skb);
733 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
734 }
735 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
736 
737 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
738 			  struct msghdr *msg, struct flowi6 *fl6,
739 			  struct ipcm6_cookie *ipc6)
740 {
741 	struct in6_pktinfo *src_info;
742 	struct cmsghdr *cmsg;
743 	struct ipv6_rt_hdr *rthdr;
744 	struct ipv6_opt_hdr *hdr;
745 	struct ipv6_txoptions *opt = ipc6->opt;
746 	int len;
747 	int err = 0;
748 
749 	for_each_cmsghdr(cmsg, msg) {
750 		int addr_type;
751 
752 		if (!CMSG_OK(msg, cmsg)) {
753 			err = -EINVAL;
754 			goto exit_f;
755 		}
756 
757 		if (cmsg->cmsg_level == SOL_SOCKET) {
758 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc6->sockc);
759 			if (err)
760 				return err;
761 			continue;
762 		}
763 
764 		if (cmsg->cmsg_level != SOL_IPV6)
765 			continue;
766 
767 		switch (cmsg->cmsg_type) {
768 		case IPV6_PKTINFO:
769 		case IPV6_2292PKTINFO:
770 		    {
771 			struct net_device *dev = NULL;
772 			int src_idx;
773 
774 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
775 				err = -EINVAL;
776 				goto exit_f;
777 			}
778 
779 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
780 			src_idx = src_info->ipi6_ifindex;
781 
782 			if (src_idx) {
783 				if (fl6->flowi6_oif &&
784 				    src_idx != fl6->flowi6_oif &&
785 				    (sk->sk_bound_dev_if != fl6->flowi6_oif ||
786 				     !sk_dev_equal_l3scope(sk, src_idx)))
787 					return -EINVAL;
788 				fl6->flowi6_oif = src_idx;
789 			}
790 
791 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
792 
793 			rcu_read_lock();
794 			if (fl6->flowi6_oif) {
795 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
796 				if (!dev) {
797 					rcu_read_unlock();
798 					return -ENODEV;
799 				}
800 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
801 				rcu_read_unlock();
802 				return -EINVAL;
803 			}
804 
805 			if (addr_type != IPV6_ADDR_ANY) {
806 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
807 				if (!ipv6_can_nonlocal_bind(net, inet_sk(sk)) &&
808 				    !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
809 							     dev, !strict, 0,
810 							     IFA_F_TENTATIVE) &&
811 				    !ipv6_chk_acast_addr_src(net, dev,
812 							     &src_info->ipi6_addr))
813 					err = -EINVAL;
814 				else
815 					fl6->saddr = src_info->ipi6_addr;
816 			}
817 
818 			rcu_read_unlock();
819 
820 			if (err)
821 				goto exit_f;
822 
823 			break;
824 		    }
825 
826 		case IPV6_FLOWINFO:
827 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
828 				err = -EINVAL;
829 				goto exit_f;
830 			}
831 
832 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
833 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
834 					err = -EINVAL;
835 					goto exit_f;
836 				}
837 			}
838 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
839 			break;
840 
841 		case IPV6_2292HOPOPTS:
842 		case IPV6_HOPOPTS:
843 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
844 				err = -EINVAL;
845 				goto exit_f;
846 			}
847 
848 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
849 			len = ((hdr->hdrlen + 1) << 3);
850 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
851 				err = -EINVAL;
852 				goto exit_f;
853 			}
854 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
855 				err = -EPERM;
856 				goto exit_f;
857 			}
858 			opt->opt_nflen += len;
859 			opt->hopopt = hdr;
860 			break;
861 
862 		case IPV6_2292DSTOPTS:
863 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
864 				err = -EINVAL;
865 				goto exit_f;
866 			}
867 
868 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
869 			len = ((hdr->hdrlen + 1) << 3);
870 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
871 				err = -EINVAL;
872 				goto exit_f;
873 			}
874 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
875 				err = -EPERM;
876 				goto exit_f;
877 			}
878 			if (opt->dst1opt) {
879 				err = -EINVAL;
880 				goto exit_f;
881 			}
882 			opt->opt_flen += len;
883 			opt->dst1opt = hdr;
884 			break;
885 
886 		case IPV6_DSTOPTS:
887 		case IPV6_RTHDRDSTOPTS:
888 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
889 				err = -EINVAL;
890 				goto exit_f;
891 			}
892 
893 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
894 			len = ((hdr->hdrlen + 1) << 3);
895 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
896 				err = -EINVAL;
897 				goto exit_f;
898 			}
899 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
900 				err = -EPERM;
901 				goto exit_f;
902 			}
903 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
904 				opt->opt_flen += len;
905 				opt->dst1opt = hdr;
906 			} else {
907 				opt->opt_nflen += len;
908 				opt->dst0opt = hdr;
909 			}
910 			break;
911 
912 		case IPV6_2292RTHDR:
913 		case IPV6_RTHDR:
914 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
915 				err = -EINVAL;
916 				goto exit_f;
917 			}
918 
919 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
920 
921 			switch (rthdr->type) {
922 #if IS_ENABLED(CONFIG_IPV6_MIP6)
923 			case IPV6_SRCRT_TYPE_2:
924 				if (rthdr->hdrlen != 2 ||
925 				    rthdr->segments_left != 1) {
926 					err = -EINVAL;
927 					goto exit_f;
928 				}
929 				break;
930 #endif
931 			default:
932 				err = -EINVAL;
933 				goto exit_f;
934 			}
935 
936 			len = ((rthdr->hdrlen + 1) << 3);
937 
938 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
939 				err = -EINVAL;
940 				goto exit_f;
941 			}
942 
943 			/* segments left must also match */
944 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
945 				err = -EINVAL;
946 				goto exit_f;
947 			}
948 
949 			opt->opt_nflen += len;
950 			opt->srcrt = rthdr;
951 
952 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
953 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
954 
955 				opt->opt_nflen += dsthdrlen;
956 				opt->dst0opt = opt->dst1opt;
957 				opt->dst1opt = NULL;
958 				opt->opt_flen -= dsthdrlen;
959 			}
960 
961 			break;
962 
963 		case IPV6_2292HOPLIMIT:
964 		case IPV6_HOPLIMIT:
965 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
966 				err = -EINVAL;
967 				goto exit_f;
968 			}
969 
970 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
971 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
972 				err = -EINVAL;
973 				goto exit_f;
974 			}
975 
976 			break;
977 
978 		case IPV6_TCLASS:
979 		    {
980 			int tc;
981 
982 			err = -EINVAL;
983 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
984 				goto exit_f;
985 
986 			tc = *(int *)CMSG_DATA(cmsg);
987 			if (tc < -1 || tc > 0xff)
988 				goto exit_f;
989 
990 			err = 0;
991 			ipc6->tclass = tc;
992 
993 			break;
994 		    }
995 
996 		case IPV6_DONTFRAG:
997 		    {
998 			int df;
999 
1000 			err = -EINVAL;
1001 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1002 				goto exit_f;
1003 
1004 			df = *(int *)CMSG_DATA(cmsg);
1005 			if (df < 0 || df > 1)
1006 				goto exit_f;
1007 
1008 			err = 0;
1009 			ipc6->dontfrag = df;
1010 
1011 			break;
1012 		    }
1013 		default:
1014 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1015 					    cmsg->cmsg_type);
1016 			err = -EINVAL;
1017 			goto exit_f;
1018 		}
1019 	}
1020 
1021 exit_f:
1022 	return err;
1023 }
1024 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1025 
1026 void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1027 			       __u16 srcp, __u16 destp, int rqueue, int bucket)
1028 {
1029 	const struct in6_addr *dest, *src;
1030 
1031 	dest  = &sp->sk_v6_daddr;
1032 	src   = &sp->sk_v6_rcv_saddr;
1033 	seq_printf(seq,
1034 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1035 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1036 		   bucket,
1037 		   src->s6_addr32[0], src->s6_addr32[1],
1038 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1039 		   dest->s6_addr32[0], dest->s6_addr32[1],
1040 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1041 		   sp->sk_state,
1042 		   sk_wmem_alloc_get(sp),
1043 		   rqueue,
1044 		   0, 0L, 0,
1045 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1046 		   0,
1047 		   sock_i_ino(sp),
1048 		   refcount_read(&sp->sk_refcnt), sp,
1049 		   atomic_read(&sp->sk_drops));
1050 }
1051