xref: /linux/net/ipv6/exthdrs.c (revision f3a8b6645dc2e60d11f20c1c23afd964ff4e55ae)
1 /*
2  *	Extension Header handling for IPv6
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Andi Kleen		<ak@muc.de>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  *
10  *	This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  */
15 
16 /* Changes:
17  *	yoshfuji		: ensure not to overrun while parsing
18  *				  tlv options.
19  *	Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
20  *	YOSHIFUJI Hideaki @USAGI  Register inbound extension header
21  *				  handlers as inet6_protocol{}.
22  */
23 
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/netdevice.h>
30 #include <linux/in6.h>
31 #include <linux/icmpv6.h>
32 #include <linux/slab.h>
33 #include <linux/export.h>
34 
35 #include <net/dst.h>
36 #include <net/sock.h>
37 #include <net/snmp.h>
38 
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/rawv6.h>
43 #include <net/ndisc.h>
44 #include <net/ip6_route.h>
45 #include <net/addrconf.h>
46 #include <net/calipso.h>
47 #if IS_ENABLED(CONFIG_IPV6_MIP6)
48 #include <net/xfrm.h>
49 #endif
50 
51 #include <linux/uaccess.h>
52 
53 /*
54  *	Parsing tlv encoded headers.
55  *
56  *	Parsing function "func" returns true, if parsing succeed
57  *	and false, if it failed.
58  *	It MUST NOT touch skb->h.
59  */
60 
61 struct tlvtype_proc {
62 	int	type;
63 	bool	(*func)(struct sk_buff *skb, int offset);
64 };
65 
66 /*********************
67   Generic functions
68  *********************/
69 
70 /* An unknown option is detected, decide what to do */
71 
72 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
73 {
74 	switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
75 	case 0: /* ignore */
76 		return true;
77 
78 	case 1: /* drop packet */
79 		break;
80 
81 	case 3: /* Send ICMP if not a multicast address and drop packet */
82 		/* Actually, it is redundant check. icmp_send
83 		   will recheck in any case.
84 		 */
85 		if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
86 			break;
87 	case 2: /* send ICMP PARM PROB regardless and drop packet */
88 		icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
89 		return false;
90 	}
91 
92 	kfree_skb(skb);
93 	return false;
94 }
95 
96 /* Parse tlv encoded option header (hop-by-hop or destination) */
97 
98 static bool ip6_parse_tlv(const struct tlvtype_proc *procs, struct sk_buff *skb)
99 {
100 	const struct tlvtype_proc *curr;
101 	const unsigned char *nh = skb_network_header(skb);
102 	int off = skb_network_header_len(skb);
103 	int len = (skb_transport_header(skb)[1] + 1) << 3;
104 	int padlen = 0;
105 
106 	if (skb_transport_offset(skb) + len > skb_headlen(skb))
107 		goto bad;
108 
109 	off += 2;
110 	len -= 2;
111 
112 	while (len > 0) {
113 		int optlen = nh[off + 1] + 2;
114 		int i;
115 
116 		switch (nh[off]) {
117 		case IPV6_TLV_PAD1:
118 			optlen = 1;
119 			padlen++;
120 			if (padlen > 7)
121 				goto bad;
122 			break;
123 
124 		case IPV6_TLV_PADN:
125 			/* RFC 2460 states that the purpose of PadN is
126 			 * to align the containing header to multiples
127 			 * of 8. 7 is therefore the highest valid value.
128 			 * See also RFC 4942, Section 2.1.9.5.
129 			 */
130 			padlen += optlen;
131 			if (padlen > 7)
132 				goto bad;
133 			/* RFC 4942 recommends receiving hosts to
134 			 * actively check PadN payload to contain
135 			 * only zeroes.
136 			 */
137 			for (i = 2; i < optlen; i++) {
138 				if (nh[off + i] != 0)
139 					goto bad;
140 			}
141 			break;
142 
143 		default: /* Other TLV code so scan list */
144 			if (optlen > len)
145 				goto bad;
146 			for (curr = procs; curr->type >= 0; curr++) {
147 				if (curr->type == nh[off]) {
148 					/* type specific length/alignment
149 					   checks will be performed in the
150 					   func(). */
151 					if (curr->func(skb, off) == false)
152 						return false;
153 					break;
154 				}
155 			}
156 			if (curr->type < 0) {
157 				if (ip6_tlvopt_unknown(skb, off) == 0)
158 					return false;
159 			}
160 			padlen = 0;
161 			break;
162 		}
163 		off += optlen;
164 		len -= optlen;
165 	}
166 
167 	if (len == 0)
168 		return true;
169 bad:
170 	kfree_skb(skb);
171 	return false;
172 }
173 
174 /*****************************
175   Destination options header.
176  *****************************/
177 
178 #if IS_ENABLED(CONFIG_IPV6_MIP6)
179 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
180 {
181 	struct ipv6_destopt_hao *hao;
182 	struct inet6_skb_parm *opt = IP6CB(skb);
183 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
184 	struct in6_addr tmp_addr;
185 	int ret;
186 
187 	if (opt->dsthao) {
188 		net_dbg_ratelimited("hao duplicated\n");
189 		goto discard;
190 	}
191 	opt->dsthao = opt->dst1;
192 	opt->dst1 = 0;
193 
194 	hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
195 
196 	if (hao->length != 16) {
197 		net_dbg_ratelimited("hao invalid option length = %d\n",
198 				    hao->length);
199 		goto discard;
200 	}
201 
202 	if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
203 		net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
204 				    &hao->addr);
205 		goto discard;
206 	}
207 
208 	ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
209 			       (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
210 	if (unlikely(ret < 0))
211 		goto discard;
212 
213 	if (skb_cloned(skb)) {
214 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
215 			goto discard;
216 
217 		/* update all variable using below by copied skbuff */
218 		hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
219 						  optoff);
220 		ipv6h = ipv6_hdr(skb);
221 	}
222 
223 	if (skb->ip_summed == CHECKSUM_COMPLETE)
224 		skb->ip_summed = CHECKSUM_NONE;
225 
226 	tmp_addr = ipv6h->saddr;
227 	ipv6h->saddr = hao->addr;
228 	hao->addr = tmp_addr;
229 
230 	if (skb->tstamp.tv64 == 0)
231 		__net_timestamp(skb);
232 
233 	return true;
234 
235  discard:
236 	kfree_skb(skb);
237 	return false;
238 }
239 #endif
240 
241 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
242 #if IS_ENABLED(CONFIG_IPV6_MIP6)
243 	{
244 		.type	= IPV6_TLV_HAO,
245 		.func	= ipv6_dest_hao,
246 	},
247 #endif
248 	{-1,			NULL}
249 };
250 
251 static int ipv6_destopt_rcv(struct sk_buff *skb)
252 {
253 	struct inet6_skb_parm *opt = IP6CB(skb);
254 #if IS_ENABLED(CONFIG_IPV6_MIP6)
255 	__u16 dstbuf;
256 #endif
257 	struct dst_entry *dst = skb_dst(skb);
258 
259 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
260 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
261 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
262 		__IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
263 				IPSTATS_MIB_INHDRERRORS);
264 		kfree_skb(skb);
265 		return -1;
266 	}
267 
268 	opt->lastopt = opt->dst1 = skb_network_header_len(skb);
269 #if IS_ENABLED(CONFIG_IPV6_MIP6)
270 	dstbuf = opt->dst1;
271 #endif
272 
273 	if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
274 		skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
275 		opt = IP6CB(skb);
276 #if IS_ENABLED(CONFIG_IPV6_MIP6)
277 		opt->nhoff = dstbuf;
278 #else
279 		opt->nhoff = opt->dst1;
280 #endif
281 		return 1;
282 	}
283 
284 	__IP6_INC_STATS(dev_net(dst->dev),
285 			ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
286 	return -1;
287 }
288 
289 /********************************
290   Routing header.
291  ********************************/
292 
293 /* called with rcu_read_lock() */
294 static int ipv6_rthdr_rcv(struct sk_buff *skb)
295 {
296 	struct inet6_skb_parm *opt = IP6CB(skb);
297 	struct in6_addr *addr = NULL;
298 	struct in6_addr daddr;
299 	struct inet6_dev *idev;
300 	int n, i;
301 	struct ipv6_rt_hdr *hdr;
302 	struct rt0_hdr *rthdr;
303 	struct net *net = dev_net(skb->dev);
304 	int accept_source_route = net->ipv6.devconf_all->accept_source_route;
305 
306 	idev = __in6_dev_get(skb->dev);
307 	if (idev && accept_source_route > idev->cnf.accept_source_route)
308 		accept_source_route = idev->cnf.accept_source_route;
309 
310 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
311 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
312 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
313 		__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
314 				IPSTATS_MIB_INHDRERRORS);
315 		kfree_skb(skb);
316 		return -1;
317 	}
318 
319 	hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
320 
321 	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
322 	    skb->pkt_type != PACKET_HOST) {
323 		__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
324 				IPSTATS_MIB_INADDRERRORS);
325 		kfree_skb(skb);
326 		return -1;
327 	}
328 
329 looped_back:
330 	if (hdr->segments_left == 0) {
331 		switch (hdr->type) {
332 #if IS_ENABLED(CONFIG_IPV6_MIP6)
333 		case IPV6_SRCRT_TYPE_2:
334 			/* Silently discard type 2 header unless it was
335 			 * processed by own
336 			 */
337 			if (!addr) {
338 				__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
339 						IPSTATS_MIB_INADDRERRORS);
340 				kfree_skb(skb);
341 				return -1;
342 			}
343 			break;
344 #endif
345 		default:
346 			break;
347 		}
348 
349 		opt->lastopt = opt->srcrt = skb_network_header_len(skb);
350 		skb->transport_header += (hdr->hdrlen + 1) << 3;
351 		opt->dst0 = opt->dst1;
352 		opt->dst1 = 0;
353 		opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
354 		return 1;
355 	}
356 
357 	switch (hdr->type) {
358 #if IS_ENABLED(CONFIG_IPV6_MIP6)
359 	case IPV6_SRCRT_TYPE_2:
360 		if (accept_source_route < 0)
361 			goto unknown_rh;
362 		/* Silently discard invalid RTH type 2 */
363 		if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
364 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
365 					IPSTATS_MIB_INHDRERRORS);
366 			kfree_skb(skb);
367 			return -1;
368 		}
369 		break;
370 #endif
371 	default:
372 		goto unknown_rh;
373 	}
374 
375 	/*
376 	 *	This is the routing header forwarding algorithm from
377 	 *	RFC 2460, page 16.
378 	 */
379 
380 	n = hdr->hdrlen >> 1;
381 
382 	if (hdr->segments_left > n) {
383 		__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
384 				IPSTATS_MIB_INHDRERRORS);
385 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
386 				  ((&hdr->segments_left) -
387 				   skb_network_header(skb)));
388 		return -1;
389 	}
390 
391 	/* We are about to mangle packet header. Be careful!
392 	   Do not damage packets queued somewhere.
393 	 */
394 	if (skb_cloned(skb)) {
395 		/* the copy is a forwarded packet */
396 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
397 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
398 					IPSTATS_MIB_OUTDISCARDS);
399 			kfree_skb(skb);
400 			return -1;
401 		}
402 		hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
403 	}
404 
405 	if (skb->ip_summed == CHECKSUM_COMPLETE)
406 		skb->ip_summed = CHECKSUM_NONE;
407 
408 	i = n - --hdr->segments_left;
409 
410 	rthdr = (struct rt0_hdr *) hdr;
411 	addr = rthdr->addr;
412 	addr += i - 1;
413 
414 	switch (hdr->type) {
415 #if IS_ENABLED(CONFIG_IPV6_MIP6)
416 	case IPV6_SRCRT_TYPE_2:
417 		if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
418 				     (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
419 				     IPPROTO_ROUTING) < 0) {
420 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
421 					IPSTATS_MIB_INADDRERRORS);
422 			kfree_skb(skb);
423 			return -1;
424 		}
425 		if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
426 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
427 					IPSTATS_MIB_INADDRERRORS);
428 			kfree_skb(skb);
429 			return -1;
430 		}
431 		break;
432 #endif
433 	default:
434 		break;
435 	}
436 
437 	if (ipv6_addr_is_multicast(addr)) {
438 		__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
439 				IPSTATS_MIB_INADDRERRORS);
440 		kfree_skb(skb);
441 		return -1;
442 	}
443 
444 	daddr = *addr;
445 	*addr = ipv6_hdr(skb)->daddr;
446 	ipv6_hdr(skb)->daddr = daddr;
447 
448 	skb_dst_drop(skb);
449 	ip6_route_input(skb);
450 	if (skb_dst(skb)->error) {
451 		skb_push(skb, skb->data - skb_network_header(skb));
452 		dst_input(skb);
453 		return -1;
454 	}
455 
456 	if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
457 		if (ipv6_hdr(skb)->hop_limit <= 1) {
458 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
459 					IPSTATS_MIB_INHDRERRORS);
460 			icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
461 				    0);
462 			kfree_skb(skb);
463 			return -1;
464 		}
465 		ipv6_hdr(skb)->hop_limit--;
466 		goto looped_back;
467 	}
468 
469 	skb_push(skb, skb->data - skb_network_header(skb));
470 	dst_input(skb);
471 	return -1;
472 
473 unknown_rh:
474 	__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
475 	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
476 			  (&hdr->type) - skb_network_header(skb));
477 	return -1;
478 }
479 
480 static const struct inet6_protocol rthdr_protocol = {
481 	.handler	=	ipv6_rthdr_rcv,
482 	.flags		=	INET6_PROTO_NOPOLICY,
483 };
484 
485 static const struct inet6_protocol destopt_protocol = {
486 	.handler	=	ipv6_destopt_rcv,
487 	.flags		=	INET6_PROTO_NOPOLICY,
488 };
489 
490 static const struct inet6_protocol nodata_protocol = {
491 	.handler	=	dst_discard,
492 	.flags		=	INET6_PROTO_NOPOLICY,
493 };
494 
495 int __init ipv6_exthdrs_init(void)
496 {
497 	int ret;
498 
499 	ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
500 	if (ret)
501 		goto out;
502 
503 	ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
504 	if (ret)
505 		goto out_rthdr;
506 
507 	ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
508 	if (ret)
509 		goto out_destopt;
510 
511 out:
512 	return ret;
513 out_destopt:
514 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
515 out_rthdr:
516 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
517 	goto out;
518 };
519 
520 void ipv6_exthdrs_exit(void)
521 {
522 	inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
523 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
524 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
525 }
526 
527 /**********************************
528   Hop-by-hop options.
529  **********************************/
530 
531 /*
532  * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
533  */
534 static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
535 {
536 	return skb_dst(skb) ? ip6_dst_idev(skb_dst(skb)) : __in6_dev_get(skb->dev);
537 }
538 
539 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
540 {
541 	return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
542 }
543 
544 /* Router Alert as of RFC 2711 */
545 
546 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
547 {
548 	const unsigned char *nh = skb_network_header(skb);
549 
550 	if (nh[optoff + 1] == 2) {
551 		IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
552 		memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
553 		return true;
554 	}
555 	net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
556 			    nh[optoff + 1]);
557 	kfree_skb(skb);
558 	return false;
559 }
560 
561 /* Jumbo payload */
562 
563 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
564 {
565 	const unsigned char *nh = skb_network_header(skb);
566 	struct net *net = ipv6_skb_net(skb);
567 	u32 pkt_len;
568 
569 	if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
570 		net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
571 				    nh[optoff+1]);
572 		__IP6_INC_STATS(net, ipv6_skb_idev(skb),
573 				IPSTATS_MIB_INHDRERRORS);
574 		goto drop;
575 	}
576 
577 	pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
578 	if (pkt_len <= IPV6_MAXPLEN) {
579 		__IP6_INC_STATS(net, ipv6_skb_idev(skb),
580 				IPSTATS_MIB_INHDRERRORS);
581 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
582 		return false;
583 	}
584 	if (ipv6_hdr(skb)->payload_len) {
585 		__IP6_INC_STATS(net, ipv6_skb_idev(skb),
586 				IPSTATS_MIB_INHDRERRORS);
587 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
588 		return false;
589 	}
590 
591 	if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
592 		__IP6_INC_STATS(net, ipv6_skb_idev(skb),
593 				IPSTATS_MIB_INTRUNCATEDPKTS);
594 		goto drop;
595 	}
596 
597 	if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
598 		goto drop;
599 
600 	return true;
601 
602 drop:
603 	kfree_skb(skb);
604 	return false;
605 }
606 
607 /* CALIPSO RFC 5570 */
608 
609 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
610 {
611 	const unsigned char *nh = skb_network_header(skb);
612 
613 	if (nh[optoff + 1] < 8)
614 		goto drop;
615 
616 	if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
617 		goto drop;
618 
619 	if (!calipso_validate(skb, nh + optoff))
620 		goto drop;
621 
622 	return true;
623 
624 drop:
625 	kfree_skb(skb);
626 	return false;
627 }
628 
629 static const struct tlvtype_proc tlvprochopopt_lst[] = {
630 	{
631 		.type	= IPV6_TLV_ROUTERALERT,
632 		.func	= ipv6_hop_ra,
633 	},
634 	{
635 		.type	= IPV6_TLV_JUMBO,
636 		.func	= ipv6_hop_jumbo,
637 	},
638 	{
639 		.type	= IPV6_TLV_CALIPSO,
640 		.func	= ipv6_hop_calipso,
641 	},
642 	{ -1, }
643 };
644 
645 int ipv6_parse_hopopts(struct sk_buff *skb)
646 {
647 	struct inet6_skb_parm *opt = IP6CB(skb);
648 
649 	/*
650 	 * skb_network_header(skb) is equal to skb->data, and
651 	 * skb_network_header_len(skb) is always equal to
652 	 * sizeof(struct ipv6hdr) by definition of
653 	 * hop-by-hop options.
654 	 */
655 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
656 	    !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
657 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
658 		kfree_skb(skb);
659 		return -1;
660 	}
661 
662 	opt->flags |= IP6SKB_HOPBYHOP;
663 	if (ip6_parse_tlv(tlvprochopopt_lst, skb)) {
664 		skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
665 		opt = IP6CB(skb);
666 		opt->nhoff = sizeof(struct ipv6hdr);
667 		return 1;
668 	}
669 	return -1;
670 }
671 
672 /*
673  *	Creating outbound headers.
674  *
675  *	"build" functions work when skb is filled from head to tail (datagram)
676  *	"push"	functions work when headers are added from tail to head (tcp)
677  *
678  *	In both cases we assume, that caller reserved enough room
679  *	for headers.
680  */
681 
682 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
683 			    struct ipv6_rt_hdr *opt,
684 			    struct in6_addr **addr_p)
685 {
686 	struct rt0_hdr *phdr, *ihdr;
687 	int hops;
688 
689 	ihdr = (struct rt0_hdr *) opt;
690 
691 	phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
692 	memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
693 
694 	hops = ihdr->rt_hdr.hdrlen >> 1;
695 
696 	if (hops > 1)
697 		memcpy(phdr->addr, ihdr->addr + 1,
698 		       (hops - 1) * sizeof(struct in6_addr));
699 
700 	phdr->addr[hops - 1] = **addr_p;
701 	*addr_p = ihdr->addr;
702 
703 	phdr->rt_hdr.nexthdr = *proto;
704 	*proto = NEXTHDR_ROUTING;
705 }
706 
707 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
708 {
709 	struct ipv6_opt_hdr *h = (struct ipv6_opt_hdr *)skb_push(skb, ipv6_optlen(opt));
710 
711 	memcpy(h, opt, ipv6_optlen(opt));
712 	h->nexthdr = *proto;
713 	*proto = type;
714 }
715 
716 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
717 			  u8 *proto,
718 			  struct in6_addr **daddr)
719 {
720 	if (opt->srcrt) {
721 		ipv6_push_rthdr(skb, proto, opt->srcrt, daddr);
722 		/*
723 		 * IPV6_RTHDRDSTOPTS is ignored
724 		 * unless IPV6_RTHDR is set (RFC3542).
725 		 */
726 		if (opt->dst0opt)
727 			ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
728 	}
729 	if (opt->hopopt)
730 		ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
731 }
732 EXPORT_SYMBOL(ipv6_push_nfrag_opts);
733 
734 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
735 {
736 	if (opt->dst1opt)
737 		ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
738 }
739 
740 struct ipv6_txoptions *
741 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
742 {
743 	struct ipv6_txoptions *opt2;
744 
745 	opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
746 	if (opt2) {
747 		long dif = (char *)opt2 - (char *)opt;
748 		memcpy(opt2, opt, opt->tot_len);
749 		if (opt2->hopopt)
750 			*((char **)&opt2->hopopt) += dif;
751 		if (opt2->dst0opt)
752 			*((char **)&opt2->dst0opt) += dif;
753 		if (opt2->dst1opt)
754 			*((char **)&opt2->dst1opt) += dif;
755 		if (opt2->srcrt)
756 			*((char **)&opt2->srcrt) += dif;
757 		atomic_set(&opt2->refcnt, 1);
758 	}
759 	return opt2;
760 }
761 EXPORT_SYMBOL_GPL(ipv6_dup_options);
762 
763 static int ipv6_renew_option(void *ohdr,
764 			     struct ipv6_opt_hdr __user *newopt, int newoptlen,
765 			     int inherit,
766 			     struct ipv6_opt_hdr **hdr,
767 			     char **p)
768 {
769 	if (inherit) {
770 		if (ohdr) {
771 			memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
772 			*hdr = (struct ipv6_opt_hdr *)*p;
773 			*p += CMSG_ALIGN(ipv6_optlen(*hdr));
774 		}
775 	} else {
776 		if (newopt) {
777 			if (copy_from_user(*p, newopt, newoptlen))
778 				return -EFAULT;
779 			*hdr = (struct ipv6_opt_hdr *)*p;
780 			if (ipv6_optlen(*hdr) > newoptlen)
781 				return -EINVAL;
782 			*p += CMSG_ALIGN(newoptlen);
783 		}
784 	}
785 	return 0;
786 }
787 
788 /**
789  * ipv6_renew_options - replace a specific ext hdr with a new one.
790  *
791  * @sk: sock from which to allocate memory
792  * @opt: original options
793  * @newtype: option type to replace in @opt
794  * @newopt: new option of type @newtype to replace (user-mem)
795  * @newoptlen: length of @newopt
796  *
797  * Returns a new set of options which is a copy of @opt with the
798  * option type @newtype replaced with @newopt.
799  *
800  * @opt may be NULL, in which case a new set of options is returned
801  * containing just @newopt.
802  *
803  * @newopt may be NULL, in which case the specified option type is
804  * not copied into the new set of options.
805  *
806  * The new set of options is allocated from the socket option memory
807  * buffer of @sk.
808  */
809 struct ipv6_txoptions *
810 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
811 		   int newtype,
812 		   struct ipv6_opt_hdr __user *newopt, int newoptlen)
813 {
814 	int tot_len = 0;
815 	char *p;
816 	struct ipv6_txoptions *opt2;
817 	int err;
818 
819 	if (opt) {
820 		if (newtype != IPV6_HOPOPTS && opt->hopopt)
821 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
822 		if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
823 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
824 		if (newtype != IPV6_RTHDR && opt->srcrt)
825 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
826 		if (newtype != IPV6_DSTOPTS && opt->dst1opt)
827 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
828 	}
829 
830 	if (newopt && newoptlen)
831 		tot_len += CMSG_ALIGN(newoptlen);
832 
833 	if (!tot_len)
834 		return NULL;
835 
836 	tot_len += sizeof(*opt2);
837 	opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
838 	if (!opt2)
839 		return ERR_PTR(-ENOBUFS);
840 
841 	memset(opt2, 0, tot_len);
842 	atomic_set(&opt2->refcnt, 1);
843 	opt2->tot_len = tot_len;
844 	p = (char *)(opt2 + 1);
845 
846 	err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
847 				newtype != IPV6_HOPOPTS,
848 				&opt2->hopopt, &p);
849 	if (err)
850 		goto out;
851 
852 	err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
853 				newtype != IPV6_RTHDRDSTOPTS,
854 				&opt2->dst0opt, &p);
855 	if (err)
856 		goto out;
857 
858 	err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
859 				newtype != IPV6_RTHDR,
860 				(struct ipv6_opt_hdr **)&opt2->srcrt, &p);
861 	if (err)
862 		goto out;
863 
864 	err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
865 				newtype != IPV6_DSTOPTS,
866 				&opt2->dst1opt, &p);
867 	if (err)
868 		goto out;
869 
870 	opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
871 			  (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
872 			  (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
873 	opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
874 
875 	return opt2;
876 out:
877 	sock_kfree_s(sk, opt2, opt2->tot_len);
878 	return ERR_PTR(err);
879 }
880 
881 /**
882  * ipv6_renew_options_kern - replace a specific ext hdr with a new one.
883  *
884  * @sk: sock from which to allocate memory
885  * @opt: original options
886  * @newtype: option type to replace in @opt
887  * @newopt: new option of type @newtype to replace (kernel-mem)
888  * @newoptlen: length of @newopt
889  *
890  * See ipv6_renew_options().  The difference is that @newopt is
891  * kernel memory, rather than user memory.
892  */
893 struct ipv6_txoptions *
894 ipv6_renew_options_kern(struct sock *sk, struct ipv6_txoptions *opt,
895 			int newtype, struct ipv6_opt_hdr *newopt,
896 			int newoptlen)
897 {
898 	struct ipv6_txoptions *ret_val;
899 	const mm_segment_t old_fs = get_fs();
900 
901 	set_fs(KERNEL_DS);
902 	ret_val = ipv6_renew_options(sk, opt, newtype,
903 				     (struct ipv6_opt_hdr __user *)newopt,
904 				     newoptlen);
905 	set_fs(old_fs);
906 	return ret_val;
907 }
908 
909 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
910 					  struct ipv6_txoptions *opt)
911 {
912 	/*
913 	 * ignore the dest before srcrt unless srcrt is being included.
914 	 * --yoshfuji
915 	 */
916 	if (opt && opt->dst0opt && !opt->srcrt) {
917 		if (opt_space != opt) {
918 			memcpy(opt_space, opt, sizeof(*opt_space));
919 			opt = opt_space;
920 		}
921 		opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
922 		opt->dst0opt = NULL;
923 	}
924 
925 	return opt;
926 }
927 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
928 
929 /**
930  * fl6_update_dst - update flowi destination address with info given
931  *                  by srcrt option, if any.
932  *
933  * @fl6: flowi6 for which daddr is to be updated
934  * @opt: struct ipv6_txoptions in which to look for srcrt opt
935  * @orig: copy of original daddr address if modified
936  *
937  * Returns NULL if no txoptions or no srcrt, otherwise returns orig
938  * and initial value of fl6->daddr set in orig
939  */
940 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
941 				const struct ipv6_txoptions *opt,
942 				struct in6_addr *orig)
943 {
944 	if (!opt || !opt->srcrt)
945 		return NULL;
946 
947 	*orig = fl6->daddr;
948 	fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
949 	return orig;
950 }
951 EXPORT_SYMBOL_GPL(fl6_update_dst);
952