xref: /openbmc/linux/net/ipv6/datagram.c (revision 63dc02bd)
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 
26 #include <net/ipv6.h>
27 #include <net/ndisc.h>
28 #include <net/addrconf.h>
29 #include <net/transp_v6.h>
30 #include <net/ip6_route.h>
31 #include <net/tcp_states.h>
32 
33 #include <linux/errqueue.h>
34 #include <asm/uaccess.h>
35 
36 static inline int ipv6_mapped_addr_any(const struct in6_addr *a)
37 {
38 	return (ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0));
39 }
40 
41 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
42 {
43 	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
44 	struct inet_sock      	*inet = inet_sk(sk);
45 	struct ipv6_pinfo      	*np = inet6_sk(sk);
46 	struct in6_addr		*daddr, *final_p, final;
47 	struct dst_entry	*dst;
48 	struct flowi6		fl6;
49 	struct ip6_flowlabel	*flowlabel = NULL;
50 	struct ipv6_txoptions   *opt;
51 	int			addr_type;
52 	int			err;
53 
54 	if (usin->sin6_family == AF_INET) {
55 		if (__ipv6_only_sock(sk))
56 			return -EAFNOSUPPORT;
57 		err = ip4_datagram_connect(sk, uaddr, addr_len);
58 		goto ipv4_connected;
59 	}
60 
61 	if (addr_len < SIN6_LEN_RFC2133)
62 		return -EINVAL;
63 
64 	if (usin->sin6_family != AF_INET6)
65 		return -EAFNOSUPPORT;
66 
67 	memset(&fl6, 0, sizeof(fl6));
68 	if (np->sndflow) {
69 		fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
70 		if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
71 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
72 			if (flowlabel == NULL)
73 				return -EINVAL;
74 			usin->sin6_addr = flowlabel->dst;
75 		}
76 	}
77 
78 	addr_type = ipv6_addr_type(&usin->sin6_addr);
79 
80 	if (addr_type == IPV6_ADDR_ANY) {
81 		/*
82 		 *	connect to self
83 		 */
84 		usin->sin6_addr.s6_addr[15] = 0x01;
85 	}
86 
87 	daddr = &usin->sin6_addr;
88 
89 	if (addr_type == IPV6_ADDR_MAPPED) {
90 		struct sockaddr_in sin;
91 
92 		if (__ipv6_only_sock(sk)) {
93 			err = -ENETUNREACH;
94 			goto out;
95 		}
96 		sin.sin_family = AF_INET;
97 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
98 		sin.sin_port = usin->sin6_port;
99 
100 		err = ip4_datagram_connect(sk,
101 					   (struct sockaddr*) &sin,
102 					   sizeof(sin));
103 
104 ipv4_connected:
105 		if (err)
106 			goto out;
107 
108 		ipv6_addr_set_v4mapped(inet->inet_daddr, &np->daddr);
109 
110 		if (ipv6_addr_any(&np->saddr) ||
111 		    ipv6_mapped_addr_any(&np->saddr))
112 			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
113 
114 		if (ipv6_addr_any(&np->rcv_saddr) ||
115 		    ipv6_mapped_addr_any(&np->rcv_saddr)) {
116 			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
117 					       &np->rcv_saddr);
118 			if (sk->sk_prot->rehash)
119 				sk->sk_prot->rehash(sk);
120 		}
121 
122 		goto out;
123 	}
124 
125 	if (addr_type&IPV6_ADDR_LINKLOCAL) {
126 		if (addr_len >= sizeof(struct sockaddr_in6) &&
127 		    usin->sin6_scope_id) {
128 			if (sk->sk_bound_dev_if &&
129 			    sk->sk_bound_dev_if != usin->sin6_scope_id) {
130 				err = -EINVAL;
131 				goto out;
132 			}
133 			sk->sk_bound_dev_if = usin->sin6_scope_id;
134 		}
135 
136 		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
137 			sk->sk_bound_dev_if = np->mcast_oif;
138 
139 		/* Connect to link-local address requires an interface */
140 		if (!sk->sk_bound_dev_if) {
141 			err = -EINVAL;
142 			goto out;
143 		}
144 	}
145 
146 	np->daddr = *daddr;
147 	np->flow_label = fl6.flowlabel;
148 
149 	inet->inet_dport = usin->sin6_port;
150 
151 	/*
152 	 *	Check for a route to destination an obtain the
153 	 *	destination cache for it.
154 	 */
155 
156 	fl6.flowi6_proto = sk->sk_protocol;
157 	fl6.daddr = np->daddr;
158 	fl6.saddr = np->saddr;
159 	fl6.flowi6_oif = sk->sk_bound_dev_if;
160 	fl6.flowi6_mark = sk->sk_mark;
161 	fl6.fl6_dport = inet->inet_dport;
162 	fl6.fl6_sport = inet->inet_sport;
163 
164 	if (!fl6.flowi6_oif && (addr_type&IPV6_ADDR_MULTICAST))
165 		fl6.flowi6_oif = np->mcast_oif;
166 
167 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
168 
169 	opt = flowlabel ? flowlabel->opt : np->opt;
170 	final_p = fl6_update_dst(&fl6, opt, &final);
171 
172 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
173 	err = 0;
174 	if (IS_ERR(dst)) {
175 		err = PTR_ERR(dst);
176 		goto out;
177 	}
178 
179 	/* source address lookup done in ip6_dst_lookup */
180 
181 	if (ipv6_addr_any(&np->saddr))
182 		np->saddr = fl6.saddr;
183 
184 	if (ipv6_addr_any(&np->rcv_saddr)) {
185 		np->rcv_saddr = fl6.saddr;
186 		inet->inet_rcv_saddr = LOOPBACK4_IPV6;
187 		if (sk->sk_prot->rehash)
188 			sk->sk_prot->rehash(sk);
189 	}
190 
191 	ip6_dst_store(sk, dst,
192 		      ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
193 		      &np->daddr : NULL,
194 #ifdef CONFIG_IPV6_SUBTREES
195 		      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
196 		      &np->saddr :
197 #endif
198 		      NULL);
199 
200 	sk->sk_state = TCP_ESTABLISHED;
201 out:
202 	fl6_sock_release(flowlabel);
203 	return err;
204 }
205 
206 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
207 		     __be16 port, u32 info, u8 *payload)
208 {
209 	struct ipv6_pinfo *np  = inet6_sk(sk);
210 	struct icmp6hdr *icmph = icmp6_hdr(skb);
211 	struct sock_exterr_skb *serr;
212 
213 	if (!np->recverr)
214 		return;
215 
216 	skb = skb_clone(skb, GFP_ATOMIC);
217 	if (!skb)
218 		return;
219 
220 	skb->protocol = htons(ETH_P_IPV6);
221 
222 	serr = SKB_EXT_ERR(skb);
223 	serr->ee.ee_errno = err;
224 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
225 	serr->ee.ee_type = icmph->icmp6_type;
226 	serr->ee.ee_code = icmph->icmp6_code;
227 	serr->ee.ee_pad = 0;
228 	serr->ee.ee_info = info;
229 	serr->ee.ee_data = 0;
230 	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
231 				  skb_network_header(skb);
232 	serr->port = port;
233 
234 	__skb_pull(skb, payload - skb->data);
235 	skb_reset_transport_header(skb);
236 
237 	if (sock_queue_err_skb(sk, skb))
238 		kfree_skb(skb);
239 }
240 
241 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
242 {
243 	struct ipv6_pinfo *np = inet6_sk(sk);
244 	struct sock_exterr_skb *serr;
245 	struct ipv6hdr *iph;
246 	struct sk_buff *skb;
247 
248 	if (!np->recverr)
249 		return;
250 
251 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
252 	if (!skb)
253 		return;
254 
255 	skb->protocol = htons(ETH_P_IPV6);
256 
257 	skb_put(skb, sizeof(struct ipv6hdr));
258 	skb_reset_network_header(skb);
259 	iph = ipv6_hdr(skb);
260 	iph->daddr = fl6->daddr;
261 
262 	serr = SKB_EXT_ERR(skb);
263 	serr->ee.ee_errno = err;
264 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
265 	serr->ee.ee_type = 0;
266 	serr->ee.ee_code = 0;
267 	serr->ee.ee_pad = 0;
268 	serr->ee.ee_info = info;
269 	serr->ee.ee_data = 0;
270 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
271 	serr->port = fl6->fl6_dport;
272 
273 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
274 	skb_reset_transport_header(skb);
275 
276 	if (sock_queue_err_skb(sk, skb))
277 		kfree_skb(skb);
278 }
279 
280 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
281 {
282 	struct ipv6_pinfo *np = inet6_sk(sk);
283 	struct ipv6hdr *iph;
284 	struct sk_buff *skb;
285 	struct ip6_mtuinfo *mtu_info;
286 
287 	if (!np->rxopt.bits.rxpmtu)
288 		return;
289 
290 	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
291 	if (!skb)
292 		return;
293 
294 	skb_put(skb, sizeof(struct ipv6hdr));
295 	skb_reset_network_header(skb);
296 	iph = ipv6_hdr(skb);
297 	iph->daddr = fl6->daddr;
298 
299 	mtu_info = IP6CBMTU(skb);
300 
301 	mtu_info->ip6m_mtu = mtu;
302 	mtu_info->ip6m_addr.sin6_family = AF_INET6;
303 	mtu_info->ip6m_addr.sin6_port = 0;
304 	mtu_info->ip6m_addr.sin6_flowinfo = 0;
305 	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
306 	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
307 
308 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
309 	skb_reset_transport_header(skb);
310 
311 	skb = xchg(&np->rxpmtu, skb);
312 	kfree_skb(skb);
313 }
314 
315 /*
316  *	Handle MSG_ERRQUEUE
317  */
318 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
319 {
320 	struct ipv6_pinfo *np = inet6_sk(sk);
321 	struct sock_exterr_skb *serr;
322 	struct sk_buff *skb, *skb2;
323 	struct sockaddr_in6 *sin;
324 	struct {
325 		struct sock_extended_err ee;
326 		struct sockaddr_in6	 offender;
327 	} errhdr;
328 	int err;
329 	int copied;
330 
331 	err = -EAGAIN;
332 	skb = skb_dequeue(&sk->sk_error_queue);
333 	if (skb == NULL)
334 		goto out;
335 
336 	copied = skb->len;
337 	if (copied > len) {
338 		msg->msg_flags |= MSG_TRUNC;
339 		copied = len;
340 	}
341 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
342 	if (err)
343 		goto out_free_skb;
344 
345 	sock_recv_timestamp(msg, sk, skb);
346 
347 	serr = SKB_EXT_ERR(skb);
348 
349 	sin = (struct sockaddr_in6 *)msg->msg_name;
350 	if (sin) {
351 		const unsigned char *nh = skb_network_header(skb);
352 		sin->sin6_family = AF_INET6;
353 		sin->sin6_flowinfo = 0;
354 		sin->sin6_port = serr->port;
355 		sin->sin6_scope_id = 0;
356 		if (skb->protocol == htons(ETH_P_IPV6)) {
357 			sin->sin6_addr =
358 				*(struct in6_addr *)(nh + serr->addr_offset);
359 			if (np->sndflow)
360 				sin->sin6_flowinfo =
361 					(*(__be32 *)(nh + serr->addr_offset - 24) &
362 					 IPV6_FLOWINFO_MASK);
363 			if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
364 				sin->sin6_scope_id = IP6CB(skb)->iif;
365 		} else {
366 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
367 					       &sin->sin6_addr);
368 		}
369 	}
370 
371 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
372 	sin = &errhdr.offender;
373 	sin->sin6_family = AF_UNSPEC;
374 	if (serr->ee.ee_origin != SO_EE_ORIGIN_LOCAL) {
375 		sin->sin6_family = AF_INET6;
376 		sin->sin6_flowinfo = 0;
377 		sin->sin6_scope_id = 0;
378 		if (skb->protocol == htons(ETH_P_IPV6)) {
379 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
380 			if (np->rxopt.all)
381 				datagram_recv_ctl(sk, msg, skb);
382 			if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
383 				sin->sin6_scope_id = IP6CB(skb)->iif;
384 		} else {
385 			struct inet_sock *inet = inet_sk(sk);
386 
387 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
388 					       &sin->sin6_addr);
389 			if (inet->cmsg_flags)
390 				ip_cmsg_recv(msg, skb);
391 		}
392 	}
393 
394 	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
395 
396 	/* Now we could try to dump offended packet options */
397 
398 	msg->msg_flags |= MSG_ERRQUEUE;
399 	err = copied;
400 
401 	/* Reset and regenerate socket error */
402 	spin_lock_bh(&sk->sk_error_queue.lock);
403 	sk->sk_err = 0;
404 	if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
405 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
406 		spin_unlock_bh(&sk->sk_error_queue.lock);
407 		sk->sk_error_report(sk);
408 	} else {
409 		spin_unlock_bh(&sk->sk_error_queue.lock);
410 	}
411 
412 out_free_skb:
413 	kfree_skb(skb);
414 out:
415 	return err;
416 }
417 
418 /*
419  *	Handle IPV6_RECVPATHMTU
420  */
421 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len)
422 {
423 	struct ipv6_pinfo *np = inet6_sk(sk);
424 	struct sk_buff *skb;
425 	struct sockaddr_in6 *sin;
426 	struct ip6_mtuinfo mtu_info;
427 	int err;
428 	int copied;
429 
430 	err = -EAGAIN;
431 	skb = xchg(&np->rxpmtu, NULL);
432 	if (skb == NULL)
433 		goto out;
434 
435 	copied = skb->len;
436 	if (copied > len) {
437 		msg->msg_flags |= MSG_TRUNC;
438 		copied = len;
439 	}
440 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
441 	if (err)
442 		goto out_free_skb;
443 
444 	sock_recv_timestamp(msg, sk, skb);
445 
446 	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
447 
448 	sin = (struct sockaddr_in6 *)msg->msg_name;
449 	if (sin) {
450 		sin->sin6_family = AF_INET6;
451 		sin->sin6_flowinfo = 0;
452 		sin->sin6_port = 0;
453 		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
454 		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
455 	}
456 
457 	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
458 
459 	err = copied;
460 
461 out_free_skb:
462 	kfree_skb(skb);
463 out:
464 	return err;
465 }
466 
467 
468 int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
469 {
470 	struct ipv6_pinfo *np = inet6_sk(sk);
471 	struct inet6_skb_parm *opt = IP6CB(skb);
472 	unsigned char *nh = skb_network_header(skb);
473 
474 	if (np->rxopt.bits.rxinfo) {
475 		struct in6_pktinfo src_info;
476 
477 		src_info.ipi6_ifindex = opt->iif;
478 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
479 		put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
480 	}
481 
482 	if (np->rxopt.bits.rxhlim) {
483 		int hlim = ipv6_hdr(skb)->hop_limit;
484 		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
485 	}
486 
487 	if (np->rxopt.bits.rxtclass) {
488 		int tclass = ipv6_tclass(ipv6_hdr(skb));
489 		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
490 	}
491 
492 	if (np->rxopt.bits.rxflow && (*(__be32 *)nh & IPV6_FLOWINFO_MASK)) {
493 		__be32 flowinfo = *(__be32 *)nh & IPV6_FLOWINFO_MASK;
494 		put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
495 	}
496 
497 	/* HbH is allowed only once */
498 	if (np->rxopt.bits.hopopts && opt->hop) {
499 		u8 *ptr = nh + opt->hop;
500 		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
501 	}
502 
503 	if (opt->lastopt &&
504 	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
505 		/*
506 		 * Silly enough, but we need to reparse in order to
507 		 * report extension headers (except for HbH)
508 		 * in order.
509 		 *
510 		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
511 		 * (and WILL NOT be) defined because
512 		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
513 		 */
514 		unsigned int off = sizeof(struct ipv6hdr);
515 		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
516 
517 		while (off <= opt->lastopt) {
518 			unsigned len;
519 			u8 *ptr = nh + off;
520 
521 			switch(nexthdr) {
522 			case IPPROTO_DSTOPTS:
523 				nexthdr = ptr[0];
524 				len = (ptr[1] + 1) << 3;
525 				if (np->rxopt.bits.dstopts)
526 					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
527 				break;
528 			case IPPROTO_ROUTING:
529 				nexthdr = ptr[0];
530 				len = (ptr[1] + 1) << 3;
531 				if (np->rxopt.bits.srcrt)
532 					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
533 				break;
534 			case IPPROTO_AH:
535 				nexthdr = ptr[0];
536 				len = (ptr[1] + 2) << 2;
537 				break;
538 			default:
539 				nexthdr = ptr[0];
540 				len = (ptr[1] + 1) << 3;
541 				break;
542 			}
543 
544 			off += len;
545 		}
546 	}
547 
548 	/* socket options in old style */
549 	if (np->rxopt.bits.rxoinfo) {
550 		struct in6_pktinfo src_info;
551 
552 		src_info.ipi6_ifindex = opt->iif;
553 		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
554 		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
555 	}
556 	if (np->rxopt.bits.rxohlim) {
557 		int hlim = ipv6_hdr(skb)->hop_limit;
558 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
559 	}
560 	if (np->rxopt.bits.ohopopts && opt->hop) {
561 		u8 *ptr = nh + opt->hop;
562 		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
563 	}
564 	if (np->rxopt.bits.odstopts && opt->dst0) {
565 		u8 *ptr = nh + opt->dst0;
566 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
567 	}
568 	if (np->rxopt.bits.osrcrt && opt->srcrt) {
569 		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
570 		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
571 	}
572 	if (np->rxopt.bits.odstopts && opt->dst1) {
573 		u8 *ptr = nh + opt->dst1;
574 		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
575 	}
576 	if (np->rxopt.bits.rxorigdstaddr) {
577 		struct sockaddr_in6 sin6;
578 		__be16 *ports = (__be16 *) skb_transport_header(skb);
579 
580 		if (skb_transport_offset(skb) + 4 <= skb->len) {
581 			/* All current transport protocols have the port numbers in the
582 			 * first four bytes of the transport header and this function is
583 			 * written with this assumption in mind.
584 			 */
585 
586 			sin6.sin6_family = AF_INET6;
587 			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
588 			sin6.sin6_port = ports[1];
589 			sin6.sin6_flowinfo = 0;
590 			sin6.sin6_scope_id = 0;
591 
592 			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
593 		}
594 	}
595 	return 0;
596 }
597 
598 int datagram_send_ctl(struct net *net, struct sock *sk,
599 		      struct msghdr *msg, struct flowi6 *fl6,
600 		      struct ipv6_txoptions *opt,
601 		      int *hlimit, int *tclass, int *dontfrag)
602 {
603 	struct in6_pktinfo *src_info;
604 	struct cmsghdr *cmsg;
605 	struct ipv6_rt_hdr *rthdr;
606 	struct ipv6_opt_hdr *hdr;
607 	int len;
608 	int err = 0;
609 
610 	for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
611 		int addr_type;
612 
613 		if (!CMSG_OK(msg, cmsg)) {
614 			err = -EINVAL;
615 			goto exit_f;
616 		}
617 
618 		if (cmsg->cmsg_level != SOL_IPV6)
619 			continue;
620 
621 		switch (cmsg->cmsg_type) {
622 		case IPV6_PKTINFO:
623 		case IPV6_2292PKTINFO:
624 		    {
625 			struct net_device *dev = NULL;
626 
627 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
628 				err = -EINVAL;
629 				goto exit_f;
630 			}
631 
632 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
633 
634 			if (src_info->ipi6_ifindex) {
635 				if (fl6->flowi6_oif &&
636 				    src_info->ipi6_ifindex != fl6->flowi6_oif)
637 					return -EINVAL;
638 				fl6->flowi6_oif = src_info->ipi6_ifindex;
639 			}
640 
641 			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
642 
643 			rcu_read_lock();
644 			if (fl6->flowi6_oif) {
645 				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
646 				if (!dev) {
647 					rcu_read_unlock();
648 					return -ENODEV;
649 				}
650 			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
651 				rcu_read_unlock();
652 				return -EINVAL;
653 			}
654 
655 			if (addr_type != IPV6_ADDR_ANY) {
656 				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
657 				if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
658 				    !ipv6_chk_addr(net, &src_info->ipi6_addr,
659 						   strict ? dev : NULL, 0))
660 					err = -EINVAL;
661 				else
662 					fl6->saddr = src_info->ipi6_addr;
663 			}
664 
665 			rcu_read_unlock();
666 
667 			if (err)
668 				goto exit_f;
669 
670 			break;
671 		    }
672 
673 		case IPV6_FLOWINFO:
674 			if (cmsg->cmsg_len < CMSG_LEN(4)) {
675 				err = -EINVAL;
676 				goto exit_f;
677 			}
678 
679 			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
680 				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
681 					err = -EINVAL;
682 					goto exit_f;
683 				}
684 			}
685 			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
686 			break;
687 
688 		case IPV6_2292HOPOPTS:
689 		case IPV6_HOPOPTS:
690 			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
691 				err = -EINVAL;
692 				goto exit_f;
693 			}
694 
695 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
696 			len = ((hdr->hdrlen + 1) << 3);
697 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
698 				err = -EINVAL;
699 				goto exit_f;
700 			}
701 			if (!capable(CAP_NET_RAW)) {
702 				err = -EPERM;
703 				goto exit_f;
704 			}
705 			opt->opt_nflen += len;
706 			opt->hopopt = hdr;
707 			break;
708 
709 		case IPV6_2292DSTOPTS:
710 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
711 				err = -EINVAL;
712 				goto exit_f;
713 			}
714 
715 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
716 			len = ((hdr->hdrlen + 1) << 3);
717 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
718 				err = -EINVAL;
719 				goto exit_f;
720 			}
721 			if (!capable(CAP_NET_RAW)) {
722 				err = -EPERM;
723 				goto exit_f;
724 			}
725 			if (opt->dst1opt) {
726 				err = -EINVAL;
727 				goto exit_f;
728 			}
729 			opt->opt_flen += len;
730 			opt->dst1opt = hdr;
731 			break;
732 
733 		case IPV6_DSTOPTS:
734 		case IPV6_RTHDRDSTOPTS:
735 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
736 				err = -EINVAL;
737 				goto exit_f;
738 			}
739 
740 			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
741 			len = ((hdr->hdrlen + 1) << 3);
742 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
743 				err = -EINVAL;
744 				goto exit_f;
745 			}
746 			if (!capable(CAP_NET_RAW)) {
747 				err = -EPERM;
748 				goto exit_f;
749 			}
750 			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
751 				opt->opt_flen += len;
752 				opt->dst1opt = hdr;
753 			} else {
754 				opt->opt_nflen += len;
755 				opt->dst0opt = hdr;
756 			}
757 			break;
758 
759 		case IPV6_2292RTHDR:
760 		case IPV6_RTHDR:
761 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
762 				err = -EINVAL;
763 				goto exit_f;
764 			}
765 
766 			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
767 
768 			switch (rthdr->type) {
769 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
770 			case IPV6_SRCRT_TYPE_2:
771 				if (rthdr->hdrlen != 2 ||
772 				    rthdr->segments_left != 1) {
773 					err = -EINVAL;
774 					goto exit_f;
775 				}
776 				break;
777 #endif
778 			default:
779 				err = -EINVAL;
780 				goto exit_f;
781 			}
782 
783 			len = ((rthdr->hdrlen + 1) << 3);
784 
785 			if (cmsg->cmsg_len < CMSG_LEN(len)) {
786 				err = -EINVAL;
787 				goto exit_f;
788 			}
789 
790 			/* segments left must also match */
791 			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
792 				err = -EINVAL;
793 				goto exit_f;
794 			}
795 
796 			opt->opt_nflen += len;
797 			opt->srcrt = rthdr;
798 
799 			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
800 				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
801 
802 				opt->opt_nflen += dsthdrlen;
803 				opt->dst0opt = opt->dst1opt;
804 				opt->dst1opt = NULL;
805 				opt->opt_flen -= dsthdrlen;
806 			}
807 
808 			break;
809 
810 		case IPV6_2292HOPLIMIT:
811 		case IPV6_HOPLIMIT:
812 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
813 				err = -EINVAL;
814 				goto exit_f;
815 			}
816 
817 			*hlimit = *(int *)CMSG_DATA(cmsg);
818 			if (*hlimit < -1 || *hlimit > 0xff) {
819 				err = -EINVAL;
820 				goto exit_f;
821 			}
822 
823 			break;
824 
825 		case IPV6_TCLASS:
826 		    {
827 			int tc;
828 
829 			err = -EINVAL;
830 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
831 				goto exit_f;
832 			}
833 
834 			tc = *(int *)CMSG_DATA(cmsg);
835 			if (tc < -1 || tc > 0xff)
836 				goto exit_f;
837 
838 			err = 0;
839 			*tclass = tc;
840 
841 			break;
842 		    }
843 
844 		case IPV6_DONTFRAG:
845 		    {
846 			int df;
847 
848 			err = -EINVAL;
849 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
850 				goto exit_f;
851 			}
852 
853 			df = *(int *)CMSG_DATA(cmsg);
854 			if (df < 0 || df > 1)
855 				goto exit_f;
856 
857 			err = 0;
858 			*dontfrag = df;
859 
860 			break;
861 		    }
862 		default:
863 			LIMIT_NETDEBUG(KERN_DEBUG "invalid cmsg type: %d\n",
864 				       cmsg->cmsg_type);
865 			err = -EINVAL;
866 			goto exit_f;
867 		}
868 	}
869 
870 exit_f:
871 	return err;
872 }
873