xref: /openbmc/linux/net/ipv6/exthdrs.c (revision 151f4e2b)
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 #include <linux/seg6.h>
51 #include <net/seg6.h>
52 #ifdef CONFIG_IPV6_SEG6_HMAC
53 #include <net/seg6_hmac.h>
54 #endif
55 
56 #include <linux/uaccess.h>
57 
58 /*
59  *	Parsing tlv encoded headers.
60  *
61  *	Parsing function "func" returns true, if parsing succeed
62  *	and false, if it failed.
63  *	It MUST NOT touch skb->h.
64  */
65 
66 struct tlvtype_proc {
67 	int	type;
68 	bool	(*func)(struct sk_buff *skb, int offset);
69 };
70 
71 /*********************
72   Generic functions
73  *********************/
74 
75 /* An unknown option is detected, decide what to do */
76 
77 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff,
78 			       bool disallow_unknowns)
79 {
80 	if (disallow_unknowns) {
81 		/* If unknown TLVs are disallowed by configuration
82 		 * then always silently drop packet. Note this also
83 		 * means no ICMP parameter problem is sent which
84 		 * could be a good property to mitigate a reflection DOS
85 		 * attack.
86 		 */
87 
88 		goto drop;
89 	}
90 
91 	switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
92 	case 0: /* ignore */
93 		return true;
94 
95 	case 1: /* drop packet */
96 		break;
97 
98 	case 3: /* Send ICMP if not a multicast address and drop packet */
99 		/* Actually, it is redundant check. icmp_send
100 		   will recheck in any case.
101 		 */
102 		if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
103 			break;
104 		/* fall through */
105 	case 2: /* send ICMP PARM PROB regardless and drop packet */
106 		icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
107 		return false;
108 	}
109 
110 drop:
111 	kfree_skb(skb);
112 	return false;
113 }
114 
115 /* Parse tlv encoded option header (hop-by-hop or destination) */
116 
117 static bool ip6_parse_tlv(const struct tlvtype_proc *procs,
118 			  struct sk_buff *skb,
119 			  int max_count)
120 {
121 	int len = (skb_transport_header(skb)[1] + 1) << 3;
122 	const unsigned char *nh = skb_network_header(skb);
123 	int off = skb_network_header_len(skb);
124 	const struct tlvtype_proc *curr;
125 	bool disallow_unknowns = false;
126 	int tlv_count = 0;
127 	int padlen = 0;
128 
129 	if (unlikely(max_count < 0)) {
130 		disallow_unknowns = true;
131 		max_count = -max_count;
132 	}
133 
134 	if (skb_transport_offset(skb) + len > skb_headlen(skb))
135 		goto bad;
136 
137 	off += 2;
138 	len -= 2;
139 
140 	while (len > 0) {
141 		int optlen = nh[off + 1] + 2;
142 		int i;
143 
144 		switch (nh[off]) {
145 		case IPV6_TLV_PAD1:
146 			optlen = 1;
147 			padlen++;
148 			if (padlen > 7)
149 				goto bad;
150 			break;
151 
152 		case IPV6_TLV_PADN:
153 			/* RFC 2460 states that the purpose of PadN is
154 			 * to align the containing header to multiples
155 			 * of 8. 7 is therefore the highest valid value.
156 			 * See also RFC 4942, Section 2.1.9.5.
157 			 */
158 			padlen += optlen;
159 			if (padlen > 7)
160 				goto bad;
161 			/* RFC 4942 recommends receiving hosts to
162 			 * actively check PadN payload to contain
163 			 * only zeroes.
164 			 */
165 			for (i = 2; i < optlen; i++) {
166 				if (nh[off + i] != 0)
167 					goto bad;
168 			}
169 			break;
170 
171 		default: /* Other TLV code so scan list */
172 			if (optlen > len)
173 				goto bad;
174 
175 			tlv_count++;
176 			if (tlv_count > max_count)
177 				goto bad;
178 
179 			for (curr = procs; curr->type >= 0; curr++) {
180 				if (curr->type == nh[off]) {
181 					/* type specific length/alignment
182 					   checks will be performed in the
183 					   func(). */
184 					if (curr->func(skb, off) == false)
185 						return false;
186 					break;
187 				}
188 			}
189 			if (curr->type < 0 &&
190 			    !ip6_tlvopt_unknown(skb, off, disallow_unknowns))
191 				return false;
192 
193 			padlen = 0;
194 			break;
195 		}
196 		off += optlen;
197 		len -= optlen;
198 	}
199 
200 	if (len == 0)
201 		return true;
202 bad:
203 	kfree_skb(skb);
204 	return false;
205 }
206 
207 /*****************************
208   Destination options header.
209  *****************************/
210 
211 #if IS_ENABLED(CONFIG_IPV6_MIP6)
212 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
213 {
214 	struct ipv6_destopt_hao *hao;
215 	struct inet6_skb_parm *opt = IP6CB(skb);
216 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
217 	int ret;
218 
219 	if (opt->dsthao) {
220 		net_dbg_ratelimited("hao duplicated\n");
221 		goto discard;
222 	}
223 	opt->dsthao = opt->dst1;
224 	opt->dst1 = 0;
225 
226 	hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
227 
228 	if (hao->length != 16) {
229 		net_dbg_ratelimited("hao invalid option length = %d\n",
230 				    hao->length);
231 		goto discard;
232 	}
233 
234 	if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
235 		net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
236 				    &hao->addr);
237 		goto discard;
238 	}
239 
240 	ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
241 			       (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
242 	if (unlikely(ret < 0))
243 		goto discard;
244 
245 	if (skb_cloned(skb)) {
246 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
247 			goto discard;
248 
249 		/* update all variable using below by copied skbuff */
250 		hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
251 						  optoff);
252 		ipv6h = ipv6_hdr(skb);
253 	}
254 
255 	if (skb->ip_summed == CHECKSUM_COMPLETE)
256 		skb->ip_summed = CHECKSUM_NONE;
257 
258 	swap(ipv6h->saddr, hao->addr);
259 
260 	if (skb->tstamp == 0)
261 		__net_timestamp(skb);
262 
263 	return true;
264 
265  discard:
266 	kfree_skb(skb);
267 	return false;
268 }
269 #endif
270 
271 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
272 #if IS_ENABLED(CONFIG_IPV6_MIP6)
273 	{
274 		.type	= IPV6_TLV_HAO,
275 		.func	= ipv6_dest_hao,
276 	},
277 #endif
278 	{-1,			NULL}
279 };
280 
281 static int ipv6_destopt_rcv(struct sk_buff *skb)
282 {
283 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
284 	struct inet6_skb_parm *opt = IP6CB(skb);
285 #if IS_ENABLED(CONFIG_IPV6_MIP6)
286 	__u16 dstbuf;
287 #endif
288 	struct dst_entry *dst = skb_dst(skb);
289 	struct net *net = dev_net(skb->dev);
290 	int extlen;
291 
292 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
293 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
294 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
295 		__IP6_INC_STATS(dev_net(dst->dev), idev,
296 				IPSTATS_MIB_INHDRERRORS);
297 fail_and_free:
298 		kfree_skb(skb);
299 		return -1;
300 	}
301 
302 	extlen = (skb_transport_header(skb)[1] + 1) << 3;
303 	if (extlen > net->ipv6.sysctl.max_dst_opts_len)
304 		goto fail_and_free;
305 
306 	opt->lastopt = opt->dst1 = skb_network_header_len(skb);
307 #if IS_ENABLED(CONFIG_IPV6_MIP6)
308 	dstbuf = opt->dst1;
309 #endif
310 
311 	if (ip6_parse_tlv(tlvprocdestopt_lst, skb,
312 			  init_net.ipv6.sysctl.max_dst_opts_cnt)) {
313 		skb->transport_header += extlen;
314 		opt = IP6CB(skb);
315 #if IS_ENABLED(CONFIG_IPV6_MIP6)
316 		opt->nhoff = dstbuf;
317 #else
318 		opt->nhoff = opt->dst1;
319 #endif
320 		return 1;
321 	}
322 
323 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
324 	return -1;
325 }
326 
327 static void seg6_update_csum(struct sk_buff *skb)
328 {
329 	struct ipv6_sr_hdr *hdr;
330 	struct in6_addr *addr;
331 	__be32 from, to;
332 
333 	/* srh is at transport offset and seg_left is already decremented
334 	 * but daddr is not yet updated with next segment
335 	 */
336 
337 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
338 	addr = hdr->segments + hdr->segments_left;
339 
340 	hdr->segments_left++;
341 	from = *(__be32 *)hdr;
342 
343 	hdr->segments_left--;
344 	to = *(__be32 *)hdr;
345 
346 	/* update skb csum with diff resulting from seg_left decrement */
347 
348 	update_csum_diff4(skb, from, to);
349 
350 	/* compute csum diff between current and next segment and update */
351 
352 	update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
353 			   (__be32 *)addr);
354 }
355 
356 static int ipv6_srh_rcv(struct sk_buff *skb)
357 {
358 	struct inet6_skb_parm *opt = IP6CB(skb);
359 	struct net *net = dev_net(skb->dev);
360 	struct ipv6_sr_hdr *hdr;
361 	struct inet6_dev *idev;
362 	struct in6_addr *addr;
363 	int accept_seg6;
364 
365 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
366 
367 	idev = __in6_dev_get(skb->dev);
368 
369 	accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
370 	if (accept_seg6 > idev->cnf.seg6_enabled)
371 		accept_seg6 = idev->cnf.seg6_enabled;
372 
373 	if (!accept_seg6) {
374 		kfree_skb(skb);
375 		return -1;
376 	}
377 
378 #ifdef CONFIG_IPV6_SEG6_HMAC
379 	if (!seg6_hmac_validate_skb(skb)) {
380 		kfree_skb(skb);
381 		return -1;
382 	}
383 #endif
384 
385 looped_back:
386 	if (hdr->segments_left == 0) {
387 		if (hdr->nexthdr == NEXTHDR_IPV6) {
388 			int offset = (hdr->hdrlen + 1) << 3;
389 
390 			skb_postpull_rcsum(skb, skb_network_header(skb),
391 					   skb_network_header_len(skb));
392 
393 			if (!pskb_pull(skb, offset)) {
394 				kfree_skb(skb);
395 				return -1;
396 			}
397 			skb_postpull_rcsum(skb, skb_transport_header(skb),
398 					   offset);
399 
400 			skb_reset_network_header(skb);
401 			skb_reset_transport_header(skb);
402 			skb->encapsulation = 0;
403 
404 			__skb_tunnel_rx(skb, skb->dev, net);
405 
406 			netif_rx(skb);
407 			return -1;
408 		}
409 
410 		opt->srcrt = skb_network_header_len(skb);
411 		opt->lastopt = opt->srcrt;
412 		skb->transport_header += (hdr->hdrlen + 1) << 3;
413 		opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
414 
415 		return 1;
416 	}
417 
418 	if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
419 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
420 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
421 				  ((&hdr->segments_left) -
422 				   skb_network_header(skb)));
423 		return -1;
424 	}
425 
426 	if (skb_cloned(skb)) {
427 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
428 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
429 					IPSTATS_MIB_OUTDISCARDS);
430 			kfree_skb(skb);
431 			return -1;
432 		}
433 	}
434 
435 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
436 
437 	hdr->segments_left--;
438 	addr = hdr->segments + hdr->segments_left;
439 
440 	skb_push(skb, sizeof(struct ipv6hdr));
441 
442 	if (skb->ip_summed == CHECKSUM_COMPLETE)
443 		seg6_update_csum(skb);
444 
445 	ipv6_hdr(skb)->daddr = *addr;
446 
447 	skb_dst_drop(skb);
448 
449 	ip6_route_input(skb);
450 
451 	if (skb_dst(skb)->error) {
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, idev, IPSTATS_MIB_INHDRERRORS);
459 			icmpv6_send(skb, ICMPV6_TIME_EXCEED,
460 				    ICMPV6_EXC_HOPLIMIT, 0);
461 			kfree_skb(skb);
462 			return -1;
463 		}
464 		ipv6_hdr(skb)->hop_limit--;
465 
466 		skb_pull(skb, sizeof(struct ipv6hdr));
467 		goto looped_back;
468 	}
469 
470 	dst_input(skb);
471 
472 	return -1;
473 }
474 
475 /********************************
476   Routing header.
477  ********************************/
478 
479 /* called with rcu_read_lock() */
480 static int ipv6_rthdr_rcv(struct sk_buff *skb)
481 {
482 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
483 	struct inet6_skb_parm *opt = IP6CB(skb);
484 	struct in6_addr *addr = NULL;
485 	struct in6_addr daddr;
486 	int n, i;
487 	struct ipv6_rt_hdr *hdr;
488 	struct rt0_hdr *rthdr;
489 	struct net *net = dev_net(skb->dev);
490 	int accept_source_route = net->ipv6.devconf_all->accept_source_route;
491 
492 	idev = __in6_dev_get(skb->dev);
493 	if (idev && accept_source_route > idev->cnf.accept_source_route)
494 		accept_source_route = idev->cnf.accept_source_route;
495 
496 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
497 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
498 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
499 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
500 		kfree_skb(skb);
501 		return -1;
502 	}
503 
504 	hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
505 
506 	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
507 	    skb->pkt_type != PACKET_HOST) {
508 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
509 		kfree_skb(skb);
510 		return -1;
511 	}
512 
513 	/* segment routing */
514 	if (hdr->type == IPV6_SRCRT_TYPE_4)
515 		return ipv6_srh_rcv(skb);
516 
517 looped_back:
518 	if (hdr->segments_left == 0) {
519 		switch (hdr->type) {
520 #if IS_ENABLED(CONFIG_IPV6_MIP6)
521 		case IPV6_SRCRT_TYPE_2:
522 			/* Silently discard type 2 header unless it was
523 			 * processed by own
524 			 */
525 			if (!addr) {
526 				__IP6_INC_STATS(net, idev,
527 						IPSTATS_MIB_INADDRERRORS);
528 				kfree_skb(skb);
529 				return -1;
530 			}
531 			break;
532 #endif
533 		default:
534 			break;
535 		}
536 
537 		opt->lastopt = opt->srcrt = skb_network_header_len(skb);
538 		skb->transport_header += (hdr->hdrlen + 1) << 3;
539 		opt->dst0 = opt->dst1;
540 		opt->dst1 = 0;
541 		opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
542 		return 1;
543 	}
544 
545 	switch (hdr->type) {
546 #if IS_ENABLED(CONFIG_IPV6_MIP6)
547 	case IPV6_SRCRT_TYPE_2:
548 		if (accept_source_route < 0)
549 			goto unknown_rh;
550 		/* Silently discard invalid RTH type 2 */
551 		if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
552 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
553 			kfree_skb(skb);
554 			return -1;
555 		}
556 		break;
557 #endif
558 	default:
559 		goto unknown_rh;
560 	}
561 
562 	/*
563 	 *	This is the routing header forwarding algorithm from
564 	 *	RFC 2460, page 16.
565 	 */
566 
567 	n = hdr->hdrlen >> 1;
568 
569 	if (hdr->segments_left > n) {
570 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
571 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
572 				  ((&hdr->segments_left) -
573 				   skb_network_header(skb)));
574 		return -1;
575 	}
576 
577 	/* We are about to mangle packet header. Be careful!
578 	   Do not damage packets queued somewhere.
579 	 */
580 	if (skb_cloned(skb)) {
581 		/* the copy is a forwarded packet */
582 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
583 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
584 					IPSTATS_MIB_OUTDISCARDS);
585 			kfree_skb(skb);
586 			return -1;
587 		}
588 		hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
589 	}
590 
591 	if (skb->ip_summed == CHECKSUM_COMPLETE)
592 		skb->ip_summed = CHECKSUM_NONE;
593 
594 	i = n - --hdr->segments_left;
595 
596 	rthdr = (struct rt0_hdr *) hdr;
597 	addr = rthdr->addr;
598 	addr += i - 1;
599 
600 	switch (hdr->type) {
601 #if IS_ENABLED(CONFIG_IPV6_MIP6)
602 	case IPV6_SRCRT_TYPE_2:
603 		if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
604 				     (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
605 				     IPPROTO_ROUTING) < 0) {
606 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
607 			kfree_skb(skb);
608 			return -1;
609 		}
610 		if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
611 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612 			kfree_skb(skb);
613 			return -1;
614 		}
615 		break;
616 #endif
617 	default:
618 		break;
619 	}
620 
621 	if (ipv6_addr_is_multicast(addr)) {
622 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
623 		kfree_skb(skb);
624 		return -1;
625 	}
626 
627 	daddr = *addr;
628 	*addr = ipv6_hdr(skb)->daddr;
629 	ipv6_hdr(skb)->daddr = daddr;
630 
631 	skb_dst_drop(skb);
632 	ip6_route_input(skb);
633 	if (skb_dst(skb)->error) {
634 		skb_push(skb, skb->data - skb_network_header(skb));
635 		dst_input(skb);
636 		return -1;
637 	}
638 
639 	if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
640 		if (ipv6_hdr(skb)->hop_limit <= 1) {
641 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
642 			icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
643 				    0);
644 			kfree_skb(skb);
645 			return -1;
646 		}
647 		ipv6_hdr(skb)->hop_limit--;
648 		goto looped_back;
649 	}
650 
651 	skb_push(skb, skb->data - skb_network_header(skb));
652 	dst_input(skb);
653 	return -1;
654 
655 unknown_rh:
656 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
657 	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
658 			  (&hdr->type) - skb_network_header(skb));
659 	return -1;
660 }
661 
662 static const struct inet6_protocol rthdr_protocol = {
663 	.handler	=	ipv6_rthdr_rcv,
664 	.flags		=	INET6_PROTO_NOPOLICY,
665 };
666 
667 static const struct inet6_protocol destopt_protocol = {
668 	.handler	=	ipv6_destopt_rcv,
669 	.flags		=	INET6_PROTO_NOPOLICY,
670 };
671 
672 static const struct inet6_protocol nodata_protocol = {
673 	.handler	=	dst_discard,
674 	.flags		=	INET6_PROTO_NOPOLICY,
675 };
676 
677 int __init ipv6_exthdrs_init(void)
678 {
679 	int ret;
680 
681 	ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
682 	if (ret)
683 		goto out;
684 
685 	ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
686 	if (ret)
687 		goto out_rthdr;
688 
689 	ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
690 	if (ret)
691 		goto out_destopt;
692 
693 out:
694 	return ret;
695 out_destopt:
696 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
697 out_rthdr:
698 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
699 	goto out;
700 };
701 
702 void ipv6_exthdrs_exit(void)
703 {
704 	inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
705 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
706 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
707 }
708 
709 /**********************************
710   Hop-by-hop options.
711  **********************************/
712 
713 /*
714  * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
715  */
716 static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
717 {
718 	return skb_dst(skb) ? ip6_dst_idev(skb_dst(skb)) : __in6_dev_get(skb->dev);
719 }
720 
721 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
722 {
723 	return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
724 }
725 
726 /* Router Alert as of RFC 2711 */
727 
728 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
729 {
730 	const unsigned char *nh = skb_network_header(skb);
731 
732 	if (nh[optoff + 1] == 2) {
733 		IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
734 		memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
735 		return true;
736 	}
737 	net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
738 			    nh[optoff + 1]);
739 	kfree_skb(skb);
740 	return false;
741 }
742 
743 /* Jumbo payload */
744 
745 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
746 {
747 	const unsigned char *nh = skb_network_header(skb);
748 	struct inet6_dev *idev = __in6_dev_get_safely(skb->dev);
749 	struct net *net = ipv6_skb_net(skb);
750 	u32 pkt_len;
751 
752 	if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
753 		net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
754 				    nh[optoff+1]);
755 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
756 		goto drop;
757 	}
758 
759 	pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
760 	if (pkt_len <= IPV6_MAXPLEN) {
761 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
762 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
763 		return false;
764 	}
765 	if (ipv6_hdr(skb)->payload_len) {
766 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
767 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
768 		return false;
769 	}
770 
771 	if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
772 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INTRUNCATEDPKTS);
773 		goto drop;
774 	}
775 
776 	if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
777 		goto drop;
778 
779 	IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
780 	return true;
781 
782 drop:
783 	kfree_skb(skb);
784 	return false;
785 }
786 
787 /* CALIPSO RFC 5570 */
788 
789 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
790 {
791 	const unsigned char *nh = skb_network_header(skb);
792 
793 	if (nh[optoff + 1] < 8)
794 		goto drop;
795 
796 	if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
797 		goto drop;
798 
799 	if (!calipso_validate(skb, nh + optoff))
800 		goto drop;
801 
802 	return true;
803 
804 drop:
805 	kfree_skb(skb);
806 	return false;
807 }
808 
809 static const struct tlvtype_proc tlvprochopopt_lst[] = {
810 	{
811 		.type	= IPV6_TLV_ROUTERALERT,
812 		.func	= ipv6_hop_ra,
813 	},
814 	{
815 		.type	= IPV6_TLV_JUMBO,
816 		.func	= ipv6_hop_jumbo,
817 	},
818 	{
819 		.type	= IPV6_TLV_CALIPSO,
820 		.func	= ipv6_hop_calipso,
821 	},
822 	{ -1, }
823 };
824 
825 int ipv6_parse_hopopts(struct sk_buff *skb)
826 {
827 	struct inet6_skb_parm *opt = IP6CB(skb);
828 	struct net *net = dev_net(skb->dev);
829 	int extlen;
830 
831 	/*
832 	 * skb_network_header(skb) is equal to skb->data, and
833 	 * skb_network_header_len(skb) is always equal to
834 	 * sizeof(struct ipv6hdr) by definition of
835 	 * hop-by-hop options.
836 	 */
837 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
838 	    !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
839 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
840 fail_and_free:
841 		kfree_skb(skb);
842 		return -1;
843 	}
844 
845 	extlen = (skb_transport_header(skb)[1] + 1) << 3;
846 	if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
847 		goto fail_and_free;
848 
849 	opt->flags |= IP6SKB_HOPBYHOP;
850 	if (ip6_parse_tlv(tlvprochopopt_lst, skb,
851 			  init_net.ipv6.sysctl.max_hbh_opts_cnt)) {
852 		skb->transport_header += extlen;
853 		opt = IP6CB(skb);
854 		opt->nhoff = sizeof(struct ipv6hdr);
855 		return 1;
856 	}
857 	return -1;
858 }
859 
860 /*
861  *	Creating outbound headers.
862  *
863  *	"build" functions work when skb is filled from head to tail (datagram)
864  *	"push"	functions work when headers are added from tail to head (tcp)
865  *
866  *	In both cases we assume, that caller reserved enough room
867  *	for headers.
868  */
869 
870 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
871 			     struct ipv6_rt_hdr *opt,
872 			     struct in6_addr **addr_p, struct in6_addr *saddr)
873 {
874 	struct rt0_hdr *phdr, *ihdr;
875 	int hops;
876 
877 	ihdr = (struct rt0_hdr *) opt;
878 
879 	phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
880 	memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
881 
882 	hops = ihdr->rt_hdr.hdrlen >> 1;
883 
884 	if (hops > 1)
885 		memcpy(phdr->addr, ihdr->addr + 1,
886 		       (hops - 1) * sizeof(struct in6_addr));
887 
888 	phdr->addr[hops - 1] = **addr_p;
889 	*addr_p = ihdr->addr;
890 
891 	phdr->rt_hdr.nexthdr = *proto;
892 	*proto = NEXTHDR_ROUTING;
893 }
894 
895 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
896 			     struct ipv6_rt_hdr *opt,
897 			     struct in6_addr **addr_p, struct in6_addr *saddr)
898 {
899 	struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
900 	int plen, hops;
901 
902 	sr_ihdr = (struct ipv6_sr_hdr *)opt;
903 	plen = (sr_ihdr->hdrlen + 1) << 3;
904 
905 	sr_phdr = skb_push(skb, plen);
906 	memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
907 
908 	hops = sr_ihdr->first_segment + 1;
909 	memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
910 	       (hops - 1) * sizeof(struct in6_addr));
911 
912 	sr_phdr->segments[0] = **addr_p;
913 	*addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
914 
915 	if (sr_ihdr->hdrlen > hops * 2) {
916 		int tlvs_offset, tlvs_length;
917 
918 		tlvs_offset = (1 + hops * 2) << 3;
919 		tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
920 		memcpy((char *)sr_phdr + tlvs_offset,
921 		       (char *)sr_ihdr + tlvs_offset, tlvs_length);
922 	}
923 
924 #ifdef CONFIG_IPV6_SEG6_HMAC
925 	if (sr_has_hmac(sr_phdr)) {
926 		struct net *net = NULL;
927 
928 		if (skb->dev)
929 			net = dev_net(skb->dev);
930 		else if (skb->sk)
931 			net = sock_net(skb->sk);
932 
933 		WARN_ON(!net);
934 
935 		if (net)
936 			seg6_push_hmac(net, saddr, sr_phdr);
937 	}
938 #endif
939 
940 	sr_phdr->nexthdr = *proto;
941 	*proto = NEXTHDR_ROUTING;
942 }
943 
944 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
945 			    struct ipv6_rt_hdr *opt,
946 			    struct in6_addr **addr_p, struct in6_addr *saddr)
947 {
948 	switch (opt->type) {
949 	case IPV6_SRCRT_TYPE_0:
950 	case IPV6_SRCRT_STRICT:
951 	case IPV6_SRCRT_TYPE_2:
952 		ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
953 		break;
954 	case IPV6_SRCRT_TYPE_4:
955 		ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
956 		break;
957 	default:
958 		break;
959 	}
960 }
961 
962 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
963 {
964 	struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
965 
966 	memcpy(h, opt, ipv6_optlen(opt));
967 	h->nexthdr = *proto;
968 	*proto = type;
969 }
970 
971 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
972 			  u8 *proto,
973 			  struct in6_addr **daddr, struct in6_addr *saddr)
974 {
975 	if (opt->srcrt) {
976 		ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
977 		/*
978 		 * IPV6_RTHDRDSTOPTS is ignored
979 		 * unless IPV6_RTHDR is set (RFC3542).
980 		 */
981 		if (opt->dst0opt)
982 			ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
983 	}
984 	if (opt->hopopt)
985 		ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
986 }
987 
988 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
989 {
990 	if (opt->dst1opt)
991 		ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
992 }
993 EXPORT_SYMBOL(ipv6_push_frag_opts);
994 
995 struct ipv6_txoptions *
996 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
997 {
998 	struct ipv6_txoptions *opt2;
999 
1000 	opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1001 	if (opt2) {
1002 		long dif = (char *)opt2 - (char *)opt;
1003 		memcpy(opt2, opt, opt->tot_len);
1004 		if (opt2->hopopt)
1005 			*((char **)&opt2->hopopt) += dif;
1006 		if (opt2->dst0opt)
1007 			*((char **)&opt2->dst0opt) += dif;
1008 		if (opt2->dst1opt)
1009 			*((char **)&opt2->dst1opt) += dif;
1010 		if (opt2->srcrt)
1011 			*((char **)&opt2->srcrt) += dif;
1012 		refcount_set(&opt2->refcnt, 1);
1013 	}
1014 	return opt2;
1015 }
1016 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1017 
1018 static void ipv6_renew_option(int renewtype,
1019 			      struct ipv6_opt_hdr **dest,
1020 			      struct ipv6_opt_hdr *old,
1021 			      struct ipv6_opt_hdr *new,
1022 			      int newtype, char **p)
1023 {
1024 	struct ipv6_opt_hdr *src;
1025 
1026 	src = (renewtype == newtype ? new : old);
1027 	if (!src)
1028 		return;
1029 
1030 	memcpy(*p, src, ipv6_optlen(src));
1031 	*dest = (struct ipv6_opt_hdr *)*p;
1032 	*p += CMSG_ALIGN(ipv6_optlen(*dest));
1033 }
1034 
1035 /**
1036  * ipv6_renew_options - replace a specific ext hdr with a new one.
1037  *
1038  * @sk: sock from which to allocate memory
1039  * @opt: original options
1040  * @newtype: option type to replace in @opt
1041  * @newopt: new option of type @newtype to replace (user-mem)
1042  * @newoptlen: length of @newopt
1043  *
1044  * Returns a new set of options which is a copy of @opt with the
1045  * option type @newtype replaced with @newopt.
1046  *
1047  * @opt may be NULL, in which case a new set of options is returned
1048  * containing just @newopt.
1049  *
1050  * @newopt may be NULL, in which case the specified option type is
1051  * not copied into the new set of options.
1052  *
1053  * The new set of options is allocated from the socket option memory
1054  * buffer of @sk.
1055  */
1056 struct ipv6_txoptions *
1057 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1058 		   int newtype, struct ipv6_opt_hdr *newopt)
1059 {
1060 	int tot_len = 0;
1061 	char *p;
1062 	struct ipv6_txoptions *opt2;
1063 
1064 	if (opt) {
1065 		if (newtype != IPV6_HOPOPTS && opt->hopopt)
1066 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1067 		if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1068 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1069 		if (newtype != IPV6_RTHDR && opt->srcrt)
1070 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1071 		if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1072 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1073 	}
1074 
1075 	if (newopt)
1076 		tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1077 
1078 	if (!tot_len)
1079 		return NULL;
1080 
1081 	tot_len += sizeof(*opt2);
1082 	opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1083 	if (!opt2)
1084 		return ERR_PTR(-ENOBUFS);
1085 
1086 	memset(opt2, 0, tot_len);
1087 	refcount_set(&opt2->refcnt, 1);
1088 	opt2->tot_len = tot_len;
1089 	p = (char *)(opt2 + 1);
1090 
1091 	ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1092 			  (opt ? opt->hopopt : NULL),
1093 			  newopt, newtype, &p);
1094 	ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1095 			  (opt ? opt->dst0opt : NULL),
1096 			  newopt, newtype, &p);
1097 	ipv6_renew_option(IPV6_RTHDR,
1098 			  (struct ipv6_opt_hdr **)&opt2->srcrt,
1099 			  (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1100 			  newopt, newtype, &p);
1101 	ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1102 			  (opt ? opt->dst1opt : NULL),
1103 			  newopt, newtype, &p);
1104 
1105 	opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1106 			  (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1107 			  (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1108 	opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1109 
1110 	return opt2;
1111 }
1112 
1113 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1114 					  struct ipv6_txoptions *opt)
1115 {
1116 	/*
1117 	 * ignore the dest before srcrt unless srcrt is being included.
1118 	 * --yoshfuji
1119 	 */
1120 	if (opt && opt->dst0opt && !opt->srcrt) {
1121 		if (opt_space != opt) {
1122 			memcpy(opt_space, opt, sizeof(*opt_space));
1123 			opt = opt_space;
1124 		}
1125 		opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1126 		opt->dst0opt = NULL;
1127 	}
1128 
1129 	return opt;
1130 }
1131 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
1132 
1133 /**
1134  * fl6_update_dst - update flowi destination address with info given
1135  *                  by srcrt option, if any.
1136  *
1137  * @fl6: flowi6 for which daddr is to be updated
1138  * @opt: struct ipv6_txoptions in which to look for srcrt opt
1139  * @orig: copy of original daddr address if modified
1140  *
1141  * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1142  * and initial value of fl6->daddr set in orig
1143  */
1144 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1145 				const struct ipv6_txoptions *opt,
1146 				struct in6_addr *orig)
1147 {
1148 	if (!opt || !opt->srcrt)
1149 		return NULL;
1150 
1151 	*orig = fl6->daddr;
1152 
1153 	switch (opt->srcrt->type) {
1154 	case IPV6_SRCRT_TYPE_0:
1155 	case IPV6_SRCRT_STRICT:
1156 	case IPV6_SRCRT_TYPE_2:
1157 		fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1158 		break;
1159 	case IPV6_SRCRT_TYPE_4:
1160 	{
1161 		struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1162 
1163 		fl6->daddr = srh->segments[srh->segments_left];
1164 		break;
1165 	}
1166 	default:
1167 		return NULL;
1168 	}
1169 
1170 	return orig;
1171 }
1172 EXPORT_SYMBOL_GPL(fl6_update_dst);
1173