xref: /openbmc/linux/net/ipv6/datagram.c (revision 3286f88f)
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 #include <linux/icmp.h>
23 
24 #include <net/ipv6.h>
25 #include <net/ndisc.h>
26 #include <net/addrconf.h>
27 #include <net/transp_v6.h>
28 #include <net/ip6_route.h>
29 #include <net/tcp_states.h>
30 #include <net/dsfield.h>
31 #include <net/sock_reuseport.h>
32 
33 #include <linux/errqueue.h>
34 #include <linux/uaccess.h>
35 
36 static bool ipv6_mapped_addr_any(const struct in6_addr *a)
37 {
38 	return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
39 }
40 
41 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk)
42 {
43 	struct inet_sock *inet = inet_sk(sk);
44 	struct ipv6_pinfo *np = inet6_sk(sk);
45 	int oif = sk->sk_bound_dev_if;
46 
47 	memset(fl6, 0, sizeof(*fl6));
48 	fl6->flowi6_proto = sk->sk_protocol;
49 	fl6->daddr = sk->sk_v6_daddr;
50 	fl6->saddr = np->saddr;
51 	fl6->flowi6_mark = sk->sk_mark;
52 	fl6->fl6_dport = inet->inet_dport;
53 	fl6->fl6_sport = inet->inet_sport;
54 	fl6->flowlabel = ip6_make_flowinfo(np->tclass, np->flow_label);
55 	fl6->flowi6_uid = sk->sk_uid;
56 
57 	if (!oif)
58 		oif = np->sticky_pktinfo.ipi6_ifindex;
59 
60 	if (!oif) {
61 		if (ipv6_addr_is_multicast(&fl6->daddr))
62 			oif = np->mcast_oif;
63 		else
64 			oif = np->ucast_oif;
65 	}
66 
67 	fl6->flowi6_oif = oif;
68 	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
69 }
70 
71 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
72 {
73 	struct ip6_flowlabel *flowlabel = NULL;
74 	struct in6_addr *final_p, final;
75 	struct ipv6_txoptions *opt;
76 	struct dst_entry *dst;
77 	struct inet_sock *inet = inet_sk(sk);
78 	struct ipv6_pinfo *np = inet6_sk(sk);
79 	struct flowi6 fl6;
80 	int err = 0;
81 
82 	if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
83 		flowlabel = fl6_sock_lookup(sk, np->flow_label);
84 		if (IS_ERR(flowlabel))
85 			return -EINVAL;
86 	}
87 	ip6_datagram_flow_key_init(&fl6, sk);
88 
89 	rcu_read_lock();
90 	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
91 	final_p = fl6_update_dst(&fl6, opt, &final);
92 	rcu_read_unlock();
93 
94 	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
95 	if (IS_ERR(dst)) {
96 		err = PTR_ERR(dst);
97 		goto out;
98 	}
99 
100 	if (fix_sk_saddr) {
101 		if (ipv6_addr_any(&np->saddr))
102 			np->saddr = fl6.saddr;
103 
104 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
105 			sk->sk_v6_rcv_saddr = fl6.saddr;
106 			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
107 			if (sk->sk_prot->rehash)
108 				sk->sk_prot->rehash(sk);
109 		}
110 	}
111 
112 	ip6_sk_dst_store_flow(sk, dst, &fl6);
113 
114 out:
115 	fl6_sock_release(flowlabel);
116 	return err;
117 }
118 
119 void ip6_datagram_release_cb(struct sock *sk)
120 {
121 	struct dst_entry *dst;
122 
123 	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
124 		return;
125 
126 	rcu_read_lock();
127 	dst = __sk_dst_get(sk);
128 	if (!dst || !dst->obsolete ||
129 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
130 		rcu_read_unlock();
131 		return;
132 	}
133 	rcu_read_unlock();
134 
135 	ip6_datagram_dst_update(sk, false);
136 }
137 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
138 
139 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
140 			   int addr_len)
141 {
142 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
143 	struct inet_sock	*inet = inet_sk(sk);
144 	struct ipv6_pinfo	*np = inet6_sk(sk);
145 	struct in6_addr		*daddr, old_daddr;
146 	__be32			fl6_flowlabel = 0;
147 	__be32			old_fl6_flowlabel;
148 	__be16			old_dport;
149 	int			addr_type;
150 	int			err;
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 	if (ipv6_addr_any(&usin->sin6_addr)) {
169 		/*
170 		 *	connect to self
171 		 */
172 		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
173 			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
174 					       &usin->sin6_addr);
175 		else
176 			usin->sin6_addr = in6addr_loopback;
177 	}
178 
179 	addr_type = ipv6_addr_type(&usin->sin6_addr);
180 
181 	daddr = &usin->sin6_addr;
182 
183 	if (addr_type & IPV6_ADDR_MAPPED) {
184 		struct sockaddr_in sin;
185 
186 		if (ipv6_only_sock(sk)) {
187 			err = -ENETUNREACH;
188 			goto out;
189 		}
190 		sin.sin_family = AF_INET;
191 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
192 		sin.sin_port = usin->sin6_port;
193 
194 		err = __ip4_datagram_connect(sk,
195 					     (struct sockaddr *) &sin,
196 					     sizeof(sin));
197 
198 ipv4_connected:
199 		if (err)
200 			goto out;
201 
202 		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
203 
204 		if (ipv6_addr_any(&np->saddr) ||
205 		    ipv6_mapped_addr_any(&np->saddr))
206 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
207 
208 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
209 		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
210 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
211 					       &sk->sk_v6_rcv_saddr);
212 			if (sk->sk_prot->rehash)
213 				sk->sk_prot->rehash(sk);
214 		}
215 
216 		goto out;
217 	}
218 
219 	if (__ipv6_addr_needs_scope_id(addr_type)) {
220 		if (addr_len >= sizeof(struct sockaddr_in6) &&
221 		    usin->sin6_scope_id) {
222 			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
223 				err = -EINVAL;
224 				goto out;
225 			}
226 			WRITE_ONCE(sk->sk_bound_dev_if, usin->sin6_scope_id);
227 		}
228 
229 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
230 			WRITE_ONCE(sk->sk_bound_dev_if, np->mcast_oif);
231 
232 		/* Connect to link-local address requires an interface */
233 		if (!sk->sk_bound_dev_if) {
234 			err = -EINVAL;
235 			goto out;
236 		}
237 	}
238 
239 	/* save the current peer information before updating it */
240 	old_daddr = sk->sk_v6_daddr;
241 	old_fl6_flowlabel = np->flow_label;
242 	old_dport = inet->inet_dport;
243 
244 	sk->sk_v6_daddr = *daddr;
245 	np->flow_label = fl6_flowlabel;
246 	inet->inet_dport = usin->sin6_port;
247 
248 	/*
249 	 *	Check for a route to destination an obtain the
250 	 *	destination cache for it.
251 	 */
252 
253 	err = ip6_datagram_dst_update(sk, true);
254 	if (err) {
255 		/* Restore the socket peer info, to keep it consistent with
256 		 * the old socket state
257 		 */
258 		sk->sk_v6_daddr = old_daddr;
259 		np->flow_label = old_fl6_flowlabel;
260 		inet->inet_dport = old_dport;
261 		goto out;
262 	}
263 
264 	reuseport_has_conns_set(sk);
265 	sk->sk_state = TCP_ESTABLISHED;
266 	sk_set_txhash(sk);
267 out:
268 	return err;
269 }
270 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
271 
272 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
273 {
274 	int res;
275 
276 	lock_sock(sk);
277 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
278 	release_sock(sk);
279 	return res;
280 }
281 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
282 
283 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
284 				 int addr_len)
285 {
286 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
287 	if (sin6->sin6_family != AF_INET6)
288 		return -EAFNOSUPPORT;
289 	return ip6_datagram_connect(sk, uaddr, addr_len);
290 }
291 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
292 
293 static void ipv6_icmp_error_rfc4884(const struct sk_buff *skb,
294 				    struct sock_ee_data_rfc4884 *out)
295 {
296 	switch (icmp6_hdr(skb)->icmp6_type) {
297 	case ICMPV6_TIME_EXCEED:
298 	case ICMPV6_DEST_UNREACH:
299 		ip_icmp_error_rfc4884(skb, out, sizeof(struct icmp6hdr),
300 				      icmp6_hdr(skb)->icmp6_datagram_len * 8);
301 	}
302 }
303 
304 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
305 		     __be16 port, u32 info, u8 *payload)
306 {
307 	struct ipv6_pinfo *np  = inet6_sk(sk);
308 	struct icmp6hdr *icmph = icmp6_hdr(skb);
309 	struct sock_exterr_skb *serr;
310 
311 	if (!np->recverr)
312 		return;
313 
314 	skb = skb_clone(skb, GFP_ATOMIC);
315 	if (!skb)
316 		return;
317 
318 	skb->protocol = htons(ETH_P_IPV6);
319 
320 	serr = SKB_EXT_ERR(skb);
321 	serr->ee.ee_errno = err;
322 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
323 	serr->ee.ee_type = icmph->icmp6_type;
324 	serr->ee.ee_code = icmph->icmp6_code;
325 	serr->ee.ee_pad = 0;
326 	serr->ee.ee_info = info;
327 	serr->ee.ee_data = 0;
328 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
329 				  skb_network_header(skb);
330 	serr->port = port;
331 
332 	__skb_pull(skb, payload - skb->data);
333 
334 	if (inet6_sk(sk)->recverr_rfc4884)
335 		ipv6_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
336 
337 	skb_reset_transport_header(skb);
338 
339 	if (sock_queue_err_skb(sk, skb))
340 		kfree_skb(skb);
341 }
342 EXPORT_SYMBOL_GPL(ipv6_icmp_error);
343 
344 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
345 {
346 	const struct ipv6_pinfo *np = inet6_sk(sk);
347 	struct sock_exterr_skb *serr;
348 	struct ipv6hdr *iph;
349 	struct sk_buff *skb;
350 
351 	if (!np->recverr)
352 		return;
353 
354 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
355 	if (!skb)
356 		return;
357 
358 	skb->protocol = htons(ETH_P_IPV6);
359 
360 	skb_put(skb, sizeof(struct ipv6hdr));
361 	skb_reset_network_header(skb);
362 	iph = ipv6_hdr(skb);
363 	iph->daddr = fl6->daddr;
364 	ip6_flow_hdr(iph, 0, 0);
365 
366 	serr = SKB_EXT_ERR(skb);
367 	serr->ee.ee_errno = err;
368 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
369 	serr->ee.ee_type = 0;
370 	serr->ee.ee_code = 0;
371 	serr->ee.ee_pad = 0;
372 	serr->ee.ee_info = info;
373 	serr->ee.ee_data = 0;
374 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
375 	serr->port = fl6->fl6_dport;
376 
377 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
378 	skb_reset_transport_header(skb);
379 
380 	if (sock_queue_err_skb(sk, skb))
381 		kfree_skb(skb);
382 }
383 
384 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
385 {
386 	struct ipv6_pinfo *np = inet6_sk(sk);
387 	struct ipv6hdr *iph;
388 	struct sk_buff *skb;
389 	struct ip6_mtuinfo *mtu_info;
390 
391 	if (!np->rxopt.bits.rxpmtu)
392 		return;
393 
394 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
395 	if (!skb)
396 		return;
397 
398 	skb_put(skb, sizeof(struct ipv6hdr));
399 	skb_reset_network_header(skb);
400 	iph = ipv6_hdr(skb);
401 	iph->daddr = fl6->daddr;
402 
403 	mtu_info = IP6CBMTU(skb);
404 
405 	mtu_info->ip6m_mtu = mtu;
406 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
407 	mtu_info->ip6m_addr.sin6_port = 0;
408 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
409 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
410 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
411 
412 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
413 	skb_reset_transport_header(skb);
414 
415 	skb = xchg(&np->rxpmtu, skb);
416 	kfree_skb(skb);
417 }
418 
419 /* For some errors we have valid addr_offset even with zero payload and
420  * zero port. Also, addr_offset should be supported if port is set.
421  */
422 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
423 {
424 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
425 	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
426 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
427 }
428 
429 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
430  *
431  * At one point, excluding local errors was a quick test to identify icmp/icmp6
432  * errors. This is no longer true, but the test remained, so the v6 stack,
433  * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
434  */
435 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
436 				      struct sock_exterr_skb *serr)
437 {
438 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
439 	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
440 		return true;
441 
442 	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
443 		return false;
444 
445 	if (!IP6CB(skb)->iif)
446 		return false;
447 
448 	return true;
449 }
450 
451 /*
452  *	Handle MSG_ERRQUEUE
453  */
454 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
455 {
456 	struct ipv6_pinfo *np = inet6_sk(sk);
457 	struct sock_exterr_skb *serr;
458 	struct sk_buff *skb;
459 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
460 	struct {
461 		struct sock_extended_err ee;
462 		struct sockaddr_in6	 offender;
463 	} errhdr;
464 	int err;
465 	int copied;
466 
467 	err = -EAGAIN;
468 	skb = sock_dequeue_err_skb(sk);
469 	if (!skb)
470 		goto out;
471 
472 	copied = skb->len;
473 	if (copied > len) {
474 		msg->msg_flags |= MSG_TRUNC;
475 		copied = len;
476 	}
477 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
478 	if (unlikely(err)) {
479 		kfree_skb(skb);
480 		return err;
481 	}
482 	sock_recv_timestamp(msg, sk, skb);
483 
484 	serr = SKB_EXT_ERR(skb);
485 
486 	if (sin && ipv6_datagram_support_addr(serr)) {
487 		const unsigned char *nh = skb_network_header(skb);
488 		sin->sin6_family = AF_INET6;
489 		sin->sin6_flowinfo = 0;
490 		sin->sin6_port = serr->port;
491 		if (skb->protocol == htons(ETH_P_IPV6)) {
492 			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
493 								  struct ipv6hdr, daddr);
494 			sin->sin6_addr = ip6h->daddr;
495 			if (np->sndflow)
496 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
497 			sin->sin6_scope_id =
498 				ipv6_iface_scope_id(&sin->sin6_addr,
499 						    IP6CB(skb)->iif);
500 		} else {
501 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
502 					       &sin->sin6_addr);
503 			sin->sin6_scope_id = 0;
504 		}
505 		*addr_len = sizeof(*sin);
506 	}
507 
508 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
509 	sin = &errhdr.offender;
510 	memset(sin, 0, sizeof(*sin));
511 
512 	if (ip6_datagram_support_cmsg(skb, serr)) {
513 		sin->sin6_family = AF_INET6;
514 		if (np->rxopt.all)
515 			ip6_datagram_recv_common_ctl(sk, msg, skb);
516 		if (skb->protocol == htons(ETH_P_IPV6)) {
517 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
518 			if (np->rxopt.all)
519 				ip6_datagram_recv_specific_ctl(sk, msg, skb);
520 			sin->sin6_scope_id =
521 				ipv6_iface_scope_id(&sin->sin6_addr,
522 						    IP6CB(skb)->iif);
523 		} else {
524 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
525 					       &sin->sin6_addr);
526 			if (inet_sk(sk)->cmsg_flags)
527 				ip_cmsg_recv(msg, skb);
528 		}
529 	}
530 
531 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
532 
533 	/* Now we could try to dump offended packet options */
534 
535 	msg->msg_flags |= MSG_ERRQUEUE;
536 	err = copied;
537 
538 	consume_skb(skb);
539 out:
540 	return err;
541 }
542 EXPORT_SYMBOL_GPL(ipv6_recv_error);
543 
544 /*
545  *	Handle IPV6_RECVPATHMTU
546  */
547 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
548 		     int *addr_len)
549 {
550 	struct ipv6_pinfo *np = inet6_sk(sk);
551 	struct sk_buff *skb;
552 	struct ip6_mtuinfo mtu_info;
553 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
554 	int err;
555 	int copied;
556 
557 	err = -EAGAIN;
558 	skb = xchg(&np->rxpmtu, NULL);
559 	if (!skb)
560 		goto out;
561 
562 	copied = skb->len;
563 	if (copied > len) {
564 		msg->msg_flags |= MSG_TRUNC;
565 		copied = len;
566 	}
567 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
568 	if (err)
569 		goto out_free_skb;
570 
571 	sock_recv_timestamp(msg, sk, skb);
572 
573 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
574 
575 	if (sin) {
576 		sin->sin6_family = AF_INET6;
577 		sin->sin6_flowinfo = 0;
578 		sin->sin6_port = 0;
579 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
580 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
581 		*addr_len = sizeof(*sin);
582 	}
583 
584 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
585 
586 	err = copied;
587 
588 out_free_skb:
589 	kfree_skb(skb);
590 out:
591 	return err;
592 }
593 
594 
595 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
596 				 struct sk_buff *skb)
597 {
598 	struct ipv6_pinfo *np = inet6_sk(sk);
599 	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
600 
601 	if (np->rxopt.bits.rxinfo) {
602 		struct in6_pktinfo src_info;
603 
604 		if (is_ipv6) {
605 			src_info.ipi6_ifindex = IP6CB(skb)->iif;
606 			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
607 		} else {
608 			src_info.ipi6_ifindex =
609 				PKTINFO_SKB_CB(skb)->ipi_ifindex;
610 			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
611 					       &src_info.ipi6_addr);
612 		}
613 
614 		if (src_info.ipi6_ifindex >= 0)
615 			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
616 				 sizeof(src_info), &src_info);
617 	}
618 }
619 
620 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
621 				    struct sk_buff *skb)
622 {
623 	struct ipv6_pinfo *np = inet6_sk(sk);
624 	struct inet6_skb_parm *opt = IP6CB(skb);
625 	unsigned char *nh = skb_network_header(skb);
626 
627 	if (np->rxopt.bits.rxhlim) {
628 		int hlim = ipv6_hdr(skb)->hop_limit;
629 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
630 	}
631 
632 	if (np->rxopt.bits.rxtclass) {
633 		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
634 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
635 	}
636 
637 	if (np->rxopt.bits.rxflow) {
638 		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
639 		if (flowinfo)
640 			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
641 	}
642 
643 	/* HbH is allowed only once */
644 	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
645 		u8 *ptr = nh + sizeof(struct ipv6hdr);
646 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
647 	}
648 
649 	if (opt->lastopt &&
650 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
651 		/*
652 		 * Silly enough, but we need to reparse in order to
653 		 * report extension headers (except for HbH)
654 		 * in order.
655 		 *
656 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
657 		 * (and WILL NOT be) defined because
658 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
659 		 */
660 		unsigned int off = sizeof(struct ipv6hdr);
661 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
662 
663 		while (off <= opt->lastopt) {
664 			unsigned int len;
665 			u8 *ptr = nh + off;
666 
667 			switch (nexthdr) {
668 			case IPPROTO_DSTOPTS:
669 				nexthdr = ptr[0];
670 				len = (ptr[1] + 1) << 3;
671 				if (np->rxopt.bits.dstopts)
672 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
673 				break;
674 			case IPPROTO_ROUTING:
675 				nexthdr = ptr[0];
676 				len = (ptr[1] + 1) << 3;
677 				if (np->rxopt.bits.srcrt)
678 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
679 				break;
680 			case IPPROTO_AH:
681 				nexthdr = ptr[0];
682 				len = (ptr[1] + 2) << 2;
683 				break;
684 			default:
685 				nexthdr = ptr[0];
686 				len = (ptr[1] + 1) << 3;
687 				break;
688 			}
689 
690 			off += len;
691 		}
692 	}
693 
694 	/* socket options in old style */
695 	if (np->rxopt.bits.rxoinfo) {
696 		struct in6_pktinfo src_info;
697 
698 		src_info.ipi6_ifindex = opt->iif;
699 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
700 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
701 	}
702 	if (np->rxopt.bits.rxohlim) {
703 		int hlim = ipv6_hdr(skb)->hop_limit;
704 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
705 	}
706 	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
707 		u8 *ptr = nh + sizeof(struct ipv6hdr);
708 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
709 	}
710 	if (np->rxopt.bits.odstopts && opt->dst0) {
711 		u8 *ptr = nh + opt->dst0;
712 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
713 	}
714 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
715 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
716 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
717 	}
718 	if (np->rxopt.bits.odstopts && opt->dst1) {
719 		u8 *ptr = nh + opt->dst1;
720 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
721 	}
722 	if (np->rxopt.bits.rxorigdstaddr) {
723 		struct sockaddr_in6 sin6;
724 		__be16 _ports[2], *ports;
725 
726 		ports = skb_header_pointer(skb, skb_transport_offset(skb),
727 					   sizeof(_ports), &_ports);
728 		if (ports) {
729 			/* All current transport protocols have the port numbers in the
730 			 * first four bytes of the transport header and this function is
731 			 * written with this assumption in mind.
732 			 */
733 			sin6.sin6_family = AF_INET6;
734 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
735 			sin6.sin6_port = ports[1];
736 			sin6.sin6_flowinfo = 0;
737 			sin6.sin6_scope_id =
738 				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
739 						    opt->iif);
740 
741 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
742 		}
743 	}
744 	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
745 		int val = opt->frag_max_size;
746 
747 		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
748 	}
749 }
750 
751 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
752 			  struct sk_buff *skb)
753 {
754 	ip6_datagram_recv_common_ctl(sk, msg, skb);
755 	ip6_datagram_recv_specific_ctl(sk, msg, skb);
756 }
757 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
758 
759 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
760 			  struct msghdr *msg, struct flowi6 *fl6,
761 			  struct ipcm6_cookie *ipc6)
762 {
763 	struct in6_pktinfo *src_info;
764 	struct cmsghdr *cmsg;
765 	struct ipv6_rt_hdr *rthdr;
766 	struct ipv6_opt_hdr *hdr;
767 	struct ipv6_txoptions *opt = ipc6->opt;
768 	int len;
769 	int err = 0;
770 
771 	for_each_cmsghdr(cmsg, msg) {
772 		int addr_type;
773 
774 		if (!CMSG_OK(msg, cmsg)) {
775 			err = -EINVAL;
776 			goto exit_f;
777 		}
778 
779 		if (cmsg->cmsg_level == SOL_SOCKET) {
780 			err = __sock_cmsg_send(sk, cmsg, &ipc6->sockc);
781 			if (err)
782 				return err;
783 			continue;
784 		}
785 
786 		if (cmsg->cmsg_level != SOL_IPV6)
787 			continue;
788 
789 		switch (cmsg->cmsg_type) {
790 		case IPV6_PKTINFO:
791 		case IPV6_2292PKTINFO:
792 		    {
793 			struct net_device *dev = NULL;
794 			int src_idx;
795 
796 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
797 				err = -EINVAL;
798 				goto exit_f;
799 			}
800 
801 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
802 			src_idx = src_info->ipi6_ifindex;
803 
804 			if (src_idx) {
805 				if (fl6->flowi6_oif &&
806 				    src_idx != fl6->flowi6_oif &&
807 				    (READ_ONCE(sk->sk_bound_dev_if) != fl6->flowi6_oif ||
808 				     !sk_dev_equal_l3scope(sk, src_idx)))
809 					return -EINVAL;
810 				fl6->flowi6_oif = src_idx;
811 			}
812 
813 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
814 
815 			rcu_read_lock();
816 			if (fl6->flowi6_oif) {
817 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
818 				if (!dev) {
819 					rcu_read_unlock();
820 					return -ENODEV;
821 				}
822 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
823 				rcu_read_unlock();
824 				return -EINVAL;
825 			}
826 
827 			if (addr_type != IPV6_ADDR_ANY) {
828 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
829 				if (!ipv6_can_nonlocal_bind(net, inet_sk(sk)) &&
830 				    !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
831 							     dev, !strict, 0,
832 							     IFA_F_TENTATIVE) &&
833 				    !ipv6_chk_acast_addr_src(net, dev,
834 							     &src_info->ipi6_addr))
835 					err = -EINVAL;
836 				else
837 					fl6->saddr = src_info->ipi6_addr;
838 			}
839 
840 			rcu_read_unlock();
841 
842 			if (err)
843 				goto exit_f;
844 
845 			break;
846 		    }
847 
848 		case IPV6_FLOWINFO:
849 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
850 				err = -EINVAL;
851 				goto exit_f;
852 			}
853 
854 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
855 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
856 					err = -EINVAL;
857 					goto exit_f;
858 				}
859 			}
860 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
861 			break;
862 
863 		case IPV6_2292HOPOPTS:
864 		case IPV6_HOPOPTS:
865 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
866 				err = -EINVAL;
867 				goto exit_f;
868 			}
869 
870 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
871 			len = ((hdr->hdrlen + 1) << 3);
872 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
873 				err = -EINVAL;
874 				goto exit_f;
875 			}
876 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
877 				err = -EPERM;
878 				goto exit_f;
879 			}
880 			opt->opt_nflen += len;
881 			opt->hopopt = hdr;
882 			break;
883 
884 		case IPV6_2292DSTOPTS:
885 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
886 				err = -EINVAL;
887 				goto exit_f;
888 			}
889 
890 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
891 			len = ((hdr->hdrlen + 1) << 3);
892 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
893 				err = -EINVAL;
894 				goto exit_f;
895 			}
896 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
897 				err = -EPERM;
898 				goto exit_f;
899 			}
900 			if (opt->dst1opt) {
901 				err = -EINVAL;
902 				goto exit_f;
903 			}
904 			opt->opt_flen += len;
905 			opt->dst1opt = hdr;
906 			break;
907 
908 		case IPV6_DSTOPTS:
909 		case IPV6_RTHDRDSTOPTS:
910 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
911 				err = -EINVAL;
912 				goto exit_f;
913 			}
914 
915 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
916 			len = ((hdr->hdrlen + 1) << 3);
917 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
918 				err = -EINVAL;
919 				goto exit_f;
920 			}
921 			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
922 				err = -EPERM;
923 				goto exit_f;
924 			}
925 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
926 				opt->opt_flen += len;
927 				opt->dst1opt = hdr;
928 			} else {
929 				opt->opt_nflen += len;
930 				opt->dst0opt = hdr;
931 			}
932 			break;
933 
934 		case IPV6_2292RTHDR:
935 		case IPV6_RTHDR:
936 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
937 				err = -EINVAL;
938 				goto exit_f;
939 			}
940 
941 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
942 
943 			switch (rthdr->type) {
944 #if IS_ENABLED(CONFIG_IPV6_MIP6)
945 			case IPV6_SRCRT_TYPE_2:
946 				if (rthdr->hdrlen != 2 ||
947 				    rthdr->segments_left != 1) {
948 					err = -EINVAL;
949 					goto exit_f;
950 				}
951 				break;
952 #endif
953 			default:
954 				err = -EINVAL;
955 				goto exit_f;
956 			}
957 
958 			len = ((rthdr->hdrlen + 1) << 3);
959 
960 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
961 				err = -EINVAL;
962 				goto exit_f;
963 			}
964 
965 			/* segments left must also match */
966 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
967 				err = -EINVAL;
968 				goto exit_f;
969 			}
970 
971 			opt->opt_nflen += len;
972 			opt->srcrt = rthdr;
973 
974 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
975 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
976 
977 				opt->opt_nflen += dsthdrlen;
978 				opt->dst0opt = opt->dst1opt;
979 				opt->dst1opt = NULL;
980 				opt->opt_flen -= dsthdrlen;
981 			}
982 
983 			break;
984 
985 		case IPV6_2292HOPLIMIT:
986 		case IPV6_HOPLIMIT:
987 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
988 				err = -EINVAL;
989 				goto exit_f;
990 			}
991 
992 			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
993 			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
994 				err = -EINVAL;
995 				goto exit_f;
996 			}
997 
998 			break;
999 
1000 		case IPV6_TCLASS:
1001 		    {
1002 			int tc;
1003 
1004 			err = -EINVAL;
1005 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1006 				goto exit_f;
1007 
1008 			tc = *(int *)CMSG_DATA(cmsg);
1009 			if (tc < -1 || tc > 0xff)
1010 				goto exit_f;
1011 
1012 			err = 0;
1013 			ipc6->tclass = tc;
1014 
1015 			break;
1016 		    }
1017 
1018 		case IPV6_DONTFRAG:
1019 		    {
1020 			int df;
1021 
1022 			err = -EINVAL;
1023 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1024 				goto exit_f;
1025 
1026 			df = *(int *)CMSG_DATA(cmsg);
1027 			if (df < 0 || df > 1)
1028 				goto exit_f;
1029 
1030 			err = 0;
1031 			ipc6->dontfrag = df;
1032 
1033 			break;
1034 		    }
1035 		default:
1036 			net_dbg_ratelimited("invalid cmsg type: %d\n",
1037 					    cmsg->cmsg_type);
1038 			err = -EINVAL;
1039 			goto exit_f;
1040 		}
1041 	}
1042 
1043 exit_f:
1044 	return err;
1045 }
1046 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1047 
1048 void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1049 			       __u16 srcp, __u16 destp, int rqueue, int bucket)
1050 {
1051 	const struct in6_addr *dest, *src;
1052 
1053 	dest  = &sp->sk_v6_daddr;
1054 	src   = &sp->sk_v6_rcv_saddr;
1055 	seq_printf(seq,
1056 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1057 		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1058 		   bucket,
1059 		   src->s6_addr32[0], src->s6_addr32[1],
1060 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1061 		   dest->s6_addr32[0], dest->s6_addr32[1],
1062 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1063 		   sp->sk_state,
1064 		   sk_wmem_alloc_get(sp),
1065 		   rqueue,
1066 		   0, 0L, 0,
1067 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1068 		   0,
1069 		   sock_i_ino(sp),
1070 		   refcount_read(&sp->sk_refcnt), sp,
1071 		   atomic_read(&sp->sk_drops));
1072 }
1073