xref: /linux/net/ipv6/exthdrs.c (revision 6fdcba32711044c35c0e1b094cbd8f3f0b4472c9)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Extension Header handling for IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Andi Kleen		<ak@muc.de>
9  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
10  */
11 
12 /* Changes:
13  *	yoshfuji		: ensure not to overrun while parsing
14  *				  tlv options.
15  *	Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
16  *	YOSHIFUJI Hideaki @USAGI  Register inbound extension header
17  *				  handlers as inet6_protocol{}.
18  */
19 
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/netdevice.h>
26 #include <linux/in6.h>
27 #include <linux/icmpv6.h>
28 #include <linux/slab.h>
29 #include <linux/export.h>
30 
31 #include <net/dst.h>
32 #include <net/sock.h>
33 #include <net/snmp.h>
34 
35 #include <net/ipv6.h>
36 #include <net/protocol.h>
37 #include <net/transp_v6.h>
38 #include <net/rawv6.h>
39 #include <net/ndisc.h>
40 #include <net/ip6_route.h>
41 #include <net/addrconf.h>
42 #include <net/calipso.h>
43 #if IS_ENABLED(CONFIG_IPV6_MIP6)
44 #include <net/xfrm.h>
45 #endif
46 #include <linux/seg6.h>
47 #include <net/seg6.h>
48 #ifdef CONFIG_IPV6_SEG6_HMAC
49 #include <net/seg6_hmac.h>
50 #endif
51 
52 #include <linux/uaccess.h>
53 
54 /*
55  *	Parsing tlv encoded headers.
56  *
57  *	Parsing function "func" returns true, if parsing succeed
58  *	and false, if it failed.
59  *	It MUST NOT touch skb->h.
60  */
61 
62 struct tlvtype_proc {
63 	int	type;
64 	bool	(*func)(struct sk_buff *skb, int offset);
65 };
66 
67 /*********************
68   Generic functions
69  *********************/
70 
71 /* An unknown option is detected, decide what to do */
72 
73 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff,
74 			       bool disallow_unknowns)
75 {
76 	if (disallow_unknowns) {
77 		/* If unknown TLVs are disallowed by configuration
78 		 * then always silently drop packet. Note this also
79 		 * means no ICMP parameter problem is sent which
80 		 * could be a good property to mitigate a reflection DOS
81 		 * attack.
82 		 */
83 
84 		goto drop;
85 	}
86 
87 	switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
88 	case 0: /* ignore */
89 		return true;
90 
91 	case 1: /* drop packet */
92 		break;
93 
94 	case 3: /* Send ICMP if not a multicast address and drop packet */
95 		/* Actually, it is redundant check. icmp_send
96 		   will recheck in any case.
97 		 */
98 		if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
99 			break;
100 		/* fall through */
101 	case 2: /* send ICMP PARM PROB regardless and drop packet */
102 		icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
103 		return false;
104 	}
105 
106 drop:
107 	kfree_skb(skb);
108 	return false;
109 }
110 
111 /* Parse tlv encoded option header (hop-by-hop or destination) */
112 
113 static bool ip6_parse_tlv(const struct tlvtype_proc *procs,
114 			  struct sk_buff *skb,
115 			  int max_count)
116 {
117 	int len = (skb_transport_header(skb)[1] + 1) << 3;
118 	const unsigned char *nh = skb_network_header(skb);
119 	int off = skb_network_header_len(skb);
120 	const struct tlvtype_proc *curr;
121 	bool disallow_unknowns = false;
122 	int tlv_count = 0;
123 	int padlen = 0;
124 
125 	if (unlikely(max_count < 0)) {
126 		disallow_unknowns = true;
127 		max_count = -max_count;
128 	}
129 
130 	if (skb_transport_offset(skb) + len > skb_headlen(skb))
131 		goto bad;
132 
133 	off += 2;
134 	len -= 2;
135 
136 	while (len > 0) {
137 		int optlen = nh[off + 1] + 2;
138 		int i;
139 
140 		switch (nh[off]) {
141 		case IPV6_TLV_PAD1:
142 			optlen = 1;
143 			padlen++;
144 			if (padlen > 7)
145 				goto bad;
146 			break;
147 
148 		case IPV6_TLV_PADN:
149 			/* RFC 2460 states that the purpose of PadN is
150 			 * to align the containing header to multiples
151 			 * of 8. 7 is therefore the highest valid value.
152 			 * See also RFC 4942, Section 2.1.9.5.
153 			 */
154 			padlen += optlen;
155 			if (padlen > 7)
156 				goto bad;
157 			/* RFC 4942 recommends receiving hosts to
158 			 * actively check PadN payload to contain
159 			 * only zeroes.
160 			 */
161 			for (i = 2; i < optlen; i++) {
162 				if (nh[off + i] != 0)
163 					goto bad;
164 			}
165 			break;
166 
167 		default: /* Other TLV code so scan list */
168 			if (optlen > len)
169 				goto bad;
170 
171 			tlv_count++;
172 			if (tlv_count > max_count)
173 				goto bad;
174 
175 			for (curr = procs; curr->type >= 0; curr++) {
176 				if (curr->type == nh[off]) {
177 					/* type specific length/alignment
178 					   checks will be performed in the
179 					   func(). */
180 					if (curr->func(skb, off) == false)
181 						return false;
182 					break;
183 				}
184 			}
185 			if (curr->type < 0 &&
186 			    !ip6_tlvopt_unknown(skb, off, disallow_unknowns))
187 				return false;
188 
189 			padlen = 0;
190 			break;
191 		}
192 		off += optlen;
193 		len -= optlen;
194 	}
195 
196 	if (len == 0)
197 		return true;
198 bad:
199 	kfree_skb(skb);
200 	return false;
201 }
202 
203 /*****************************
204   Destination options header.
205  *****************************/
206 
207 #if IS_ENABLED(CONFIG_IPV6_MIP6)
208 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
209 {
210 	struct ipv6_destopt_hao *hao;
211 	struct inet6_skb_parm *opt = IP6CB(skb);
212 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
213 	int ret;
214 
215 	if (opt->dsthao) {
216 		net_dbg_ratelimited("hao duplicated\n");
217 		goto discard;
218 	}
219 	opt->dsthao = opt->dst1;
220 	opt->dst1 = 0;
221 
222 	hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
223 
224 	if (hao->length != 16) {
225 		net_dbg_ratelimited("hao invalid option length = %d\n",
226 				    hao->length);
227 		goto discard;
228 	}
229 
230 	if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
231 		net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
232 				    &hao->addr);
233 		goto discard;
234 	}
235 
236 	ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
237 			       (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
238 	if (unlikely(ret < 0))
239 		goto discard;
240 
241 	if (skb_cloned(skb)) {
242 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
243 			goto discard;
244 
245 		/* update all variable using below by copied skbuff */
246 		hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
247 						  optoff);
248 		ipv6h = ipv6_hdr(skb);
249 	}
250 
251 	if (skb->ip_summed == CHECKSUM_COMPLETE)
252 		skb->ip_summed = CHECKSUM_NONE;
253 
254 	swap(ipv6h->saddr, hao->addr);
255 
256 	if (skb->tstamp == 0)
257 		__net_timestamp(skb);
258 
259 	return true;
260 
261  discard:
262 	kfree_skb(skb);
263 	return false;
264 }
265 #endif
266 
267 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
268 #if IS_ENABLED(CONFIG_IPV6_MIP6)
269 	{
270 		.type	= IPV6_TLV_HAO,
271 		.func	= ipv6_dest_hao,
272 	},
273 #endif
274 	{-1,			NULL}
275 };
276 
277 static int ipv6_destopt_rcv(struct sk_buff *skb)
278 {
279 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
280 	struct inet6_skb_parm *opt = IP6CB(skb);
281 #if IS_ENABLED(CONFIG_IPV6_MIP6)
282 	__u16 dstbuf;
283 #endif
284 	struct dst_entry *dst = skb_dst(skb);
285 	struct net *net = dev_net(skb->dev);
286 	int extlen;
287 
288 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
289 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
290 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
291 		__IP6_INC_STATS(dev_net(dst->dev), idev,
292 				IPSTATS_MIB_INHDRERRORS);
293 fail_and_free:
294 		kfree_skb(skb);
295 		return -1;
296 	}
297 
298 	extlen = (skb_transport_header(skb)[1] + 1) << 3;
299 	if (extlen > net->ipv6.sysctl.max_dst_opts_len)
300 		goto fail_and_free;
301 
302 	opt->lastopt = opt->dst1 = skb_network_header_len(skb);
303 #if IS_ENABLED(CONFIG_IPV6_MIP6)
304 	dstbuf = opt->dst1;
305 #endif
306 
307 	if (ip6_parse_tlv(tlvprocdestopt_lst, skb,
308 			  init_net.ipv6.sysctl.max_dst_opts_cnt)) {
309 		skb->transport_header += extlen;
310 		opt = IP6CB(skb);
311 #if IS_ENABLED(CONFIG_IPV6_MIP6)
312 		opt->nhoff = dstbuf;
313 #else
314 		opt->nhoff = opt->dst1;
315 #endif
316 		return 1;
317 	}
318 
319 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
320 	return -1;
321 }
322 
323 static void seg6_update_csum(struct sk_buff *skb)
324 {
325 	struct ipv6_sr_hdr *hdr;
326 	struct in6_addr *addr;
327 	__be32 from, to;
328 
329 	/* srh is at transport offset and seg_left is already decremented
330 	 * but daddr is not yet updated with next segment
331 	 */
332 
333 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
334 	addr = hdr->segments + hdr->segments_left;
335 
336 	hdr->segments_left++;
337 	from = *(__be32 *)hdr;
338 
339 	hdr->segments_left--;
340 	to = *(__be32 *)hdr;
341 
342 	/* update skb csum with diff resulting from seg_left decrement */
343 
344 	update_csum_diff4(skb, from, to);
345 
346 	/* compute csum diff between current and next segment and update */
347 
348 	update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
349 			   (__be32 *)addr);
350 }
351 
352 static int ipv6_srh_rcv(struct sk_buff *skb)
353 {
354 	struct inet6_skb_parm *opt = IP6CB(skb);
355 	struct net *net = dev_net(skb->dev);
356 	struct ipv6_sr_hdr *hdr;
357 	struct inet6_dev *idev;
358 	struct in6_addr *addr;
359 	int accept_seg6;
360 
361 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
362 
363 	idev = __in6_dev_get(skb->dev);
364 
365 	accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
366 	if (accept_seg6 > idev->cnf.seg6_enabled)
367 		accept_seg6 = idev->cnf.seg6_enabled;
368 
369 	if (!accept_seg6) {
370 		kfree_skb(skb);
371 		return -1;
372 	}
373 
374 #ifdef CONFIG_IPV6_SEG6_HMAC
375 	if (!seg6_hmac_validate_skb(skb)) {
376 		kfree_skb(skb);
377 		return -1;
378 	}
379 #endif
380 
381 looped_back:
382 	if (hdr->segments_left == 0) {
383 		if (hdr->nexthdr == NEXTHDR_IPV6) {
384 			int offset = (hdr->hdrlen + 1) << 3;
385 
386 			skb_postpull_rcsum(skb, skb_network_header(skb),
387 					   skb_network_header_len(skb));
388 
389 			if (!pskb_pull(skb, offset)) {
390 				kfree_skb(skb);
391 				return -1;
392 			}
393 			skb_postpull_rcsum(skb, skb_transport_header(skb),
394 					   offset);
395 
396 			skb_reset_network_header(skb);
397 			skb_reset_transport_header(skb);
398 			skb->encapsulation = 0;
399 
400 			__skb_tunnel_rx(skb, skb->dev, net);
401 
402 			netif_rx(skb);
403 			return -1;
404 		}
405 
406 		opt->srcrt = skb_network_header_len(skb);
407 		opt->lastopt = opt->srcrt;
408 		skb->transport_header += (hdr->hdrlen + 1) << 3;
409 		opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
410 
411 		return 1;
412 	}
413 
414 	if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
415 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
416 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
417 				  ((&hdr->segments_left) -
418 				   skb_network_header(skb)));
419 		return -1;
420 	}
421 
422 	if (skb_cloned(skb)) {
423 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
424 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
425 					IPSTATS_MIB_OUTDISCARDS);
426 			kfree_skb(skb);
427 			return -1;
428 		}
429 	}
430 
431 	hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
432 
433 	hdr->segments_left--;
434 	addr = hdr->segments + hdr->segments_left;
435 
436 	skb_push(skb, sizeof(struct ipv6hdr));
437 
438 	if (skb->ip_summed == CHECKSUM_COMPLETE)
439 		seg6_update_csum(skb);
440 
441 	ipv6_hdr(skb)->daddr = *addr;
442 
443 	skb_dst_drop(skb);
444 
445 	ip6_route_input(skb);
446 
447 	if (skb_dst(skb)->error) {
448 		dst_input(skb);
449 		return -1;
450 	}
451 
452 	if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
453 		if (ipv6_hdr(skb)->hop_limit <= 1) {
454 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
455 			icmpv6_send(skb, ICMPV6_TIME_EXCEED,
456 				    ICMPV6_EXC_HOPLIMIT, 0);
457 			kfree_skb(skb);
458 			return -1;
459 		}
460 		ipv6_hdr(skb)->hop_limit--;
461 
462 		skb_pull(skb, sizeof(struct ipv6hdr));
463 		goto looped_back;
464 	}
465 
466 	dst_input(skb);
467 
468 	return -1;
469 }
470 
471 /********************************
472   Routing header.
473  ********************************/
474 
475 /* called with rcu_read_lock() */
476 static int ipv6_rthdr_rcv(struct sk_buff *skb)
477 {
478 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
479 	struct inet6_skb_parm *opt = IP6CB(skb);
480 	struct in6_addr *addr = NULL;
481 	struct in6_addr daddr;
482 	int n, i;
483 	struct ipv6_rt_hdr *hdr;
484 	struct rt0_hdr *rthdr;
485 	struct net *net = dev_net(skb->dev);
486 	int accept_source_route = net->ipv6.devconf_all->accept_source_route;
487 
488 	idev = __in6_dev_get(skb->dev);
489 	if (idev && accept_source_route > idev->cnf.accept_source_route)
490 		accept_source_route = idev->cnf.accept_source_route;
491 
492 	if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
493 	    !pskb_may_pull(skb, (skb_transport_offset(skb) +
494 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
495 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
496 		kfree_skb(skb);
497 		return -1;
498 	}
499 
500 	hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
501 
502 	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
503 	    skb->pkt_type != PACKET_HOST) {
504 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
505 		kfree_skb(skb);
506 		return -1;
507 	}
508 
509 	/* segment routing */
510 	if (hdr->type == IPV6_SRCRT_TYPE_4)
511 		return ipv6_srh_rcv(skb);
512 
513 looped_back:
514 	if (hdr->segments_left == 0) {
515 		switch (hdr->type) {
516 #if IS_ENABLED(CONFIG_IPV6_MIP6)
517 		case IPV6_SRCRT_TYPE_2:
518 			/* Silently discard type 2 header unless it was
519 			 * processed by own
520 			 */
521 			if (!addr) {
522 				__IP6_INC_STATS(net, idev,
523 						IPSTATS_MIB_INADDRERRORS);
524 				kfree_skb(skb);
525 				return -1;
526 			}
527 			break;
528 #endif
529 		default:
530 			break;
531 		}
532 
533 		opt->lastopt = opt->srcrt = skb_network_header_len(skb);
534 		skb->transport_header += (hdr->hdrlen + 1) << 3;
535 		opt->dst0 = opt->dst1;
536 		opt->dst1 = 0;
537 		opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
538 		return 1;
539 	}
540 
541 	switch (hdr->type) {
542 #if IS_ENABLED(CONFIG_IPV6_MIP6)
543 	case IPV6_SRCRT_TYPE_2:
544 		if (accept_source_route < 0)
545 			goto unknown_rh;
546 		/* Silently discard invalid RTH type 2 */
547 		if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
548 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
549 			kfree_skb(skb);
550 			return -1;
551 		}
552 		break;
553 #endif
554 	default:
555 		goto unknown_rh;
556 	}
557 
558 	/*
559 	 *	This is the routing header forwarding algorithm from
560 	 *	RFC 2460, page 16.
561 	 */
562 
563 	n = hdr->hdrlen >> 1;
564 
565 	if (hdr->segments_left > n) {
566 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
567 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
568 				  ((&hdr->segments_left) -
569 				   skb_network_header(skb)));
570 		return -1;
571 	}
572 
573 	/* We are about to mangle packet header. Be careful!
574 	   Do not damage packets queued somewhere.
575 	 */
576 	if (skb_cloned(skb)) {
577 		/* the copy is a forwarded packet */
578 		if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
579 			__IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
580 					IPSTATS_MIB_OUTDISCARDS);
581 			kfree_skb(skb);
582 			return -1;
583 		}
584 		hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
585 	}
586 
587 	if (skb->ip_summed == CHECKSUM_COMPLETE)
588 		skb->ip_summed = CHECKSUM_NONE;
589 
590 	i = n - --hdr->segments_left;
591 
592 	rthdr = (struct rt0_hdr *) hdr;
593 	addr = rthdr->addr;
594 	addr += i - 1;
595 
596 	switch (hdr->type) {
597 #if IS_ENABLED(CONFIG_IPV6_MIP6)
598 	case IPV6_SRCRT_TYPE_2:
599 		if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
600 				     (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
601 				     IPPROTO_ROUTING) < 0) {
602 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
603 			kfree_skb(skb);
604 			return -1;
605 		}
606 		if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
607 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
608 			kfree_skb(skb);
609 			return -1;
610 		}
611 		break;
612 #endif
613 	default:
614 		break;
615 	}
616 
617 	if (ipv6_addr_is_multicast(addr)) {
618 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
619 		kfree_skb(skb);
620 		return -1;
621 	}
622 
623 	daddr = *addr;
624 	*addr = ipv6_hdr(skb)->daddr;
625 	ipv6_hdr(skb)->daddr = daddr;
626 
627 	skb_dst_drop(skb);
628 	ip6_route_input(skb);
629 	if (skb_dst(skb)->error) {
630 		skb_push(skb, skb->data - skb_network_header(skb));
631 		dst_input(skb);
632 		return -1;
633 	}
634 
635 	if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
636 		if (ipv6_hdr(skb)->hop_limit <= 1) {
637 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
638 			icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
639 				    0);
640 			kfree_skb(skb);
641 			return -1;
642 		}
643 		ipv6_hdr(skb)->hop_limit--;
644 		goto looped_back;
645 	}
646 
647 	skb_push(skb, skb->data - skb_network_header(skb));
648 	dst_input(skb);
649 	return -1;
650 
651 unknown_rh:
652 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
653 	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
654 			  (&hdr->type) - skb_network_header(skb));
655 	return -1;
656 }
657 
658 static const struct inet6_protocol rthdr_protocol = {
659 	.handler	=	ipv6_rthdr_rcv,
660 	.flags		=	INET6_PROTO_NOPOLICY,
661 };
662 
663 static const struct inet6_protocol destopt_protocol = {
664 	.handler	=	ipv6_destopt_rcv,
665 	.flags		=	INET6_PROTO_NOPOLICY,
666 };
667 
668 static const struct inet6_protocol nodata_protocol = {
669 	.handler	=	dst_discard,
670 	.flags		=	INET6_PROTO_NOPOLICY,
671 };
672 
673 int __init ipv6_exthdrs_init(void)
674 {
675 	int ret;
676 
677 	ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
678 	if (ret)
679 		goto out;
680 
681 	ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
682 	if (ret)
683 		goto out_rthdr;
684 
685 	ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
686 	if (ret)
687 		goto out_destopt;
688 
689 out:
690 	return ret;
691 out_destopt:
692 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
693 out_rthdr:
694 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
695 	goto out;
696 };
697 
698 void ipv6_exthdrs_exit(void)
699 {
700 	inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
701 	inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
702 	inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
703 }
704 
705 /**********************************
706   Hop-by-hop options.
707  **********************************/
708 
709 /*
710  * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
711  */
712 static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
713 {
714 	return skb_dst(skb) ? ip6_dst_idev(skb_dst(skb)) : __in6_dev_get(skb->dev);
715 }
716 
717 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
718 {
719 	return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
720 }
721 
722 /* Router Alert as of RFC 2711 */
723 
724 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
725 {
726 	const unsigned char *nh = skb_network_header(skb);
727 
728 	if (nh[optoff + 1] == 2) {
729 		IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
730 		memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
731 		return true;
732 	}
733 	net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
734 			    nh[optoff + 1]);
735 	kfree_skb(skb);
736 	return false;
737 }
738 
739 /* Jumbo payload */
740 
741 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
742 {
743 	const unsigned char *nh = skb_network_header(skb);
744 	struct inet6_dev *idev = __in6_dev_get_safely(skb->dev);
745 	struct net *net = ipv6_skb_net(skb);
746 	u32 pkt_len;
747 
748 	if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
749 		net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
750 				    nh[optoff+1]);
751 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
752 		goto drop;
753 	}
754 
755 	pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
756 	if (pkt_len <= IPV6_MAXPLEN) {
757 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
758 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
759 		return false;
760 	}
761 	if (ipv6_hdr(skb)->payload_len) {
762 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
763 		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
764 		return false;
765 	}
766 
767 	if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
768 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INTRUNCATEDPKTS);
769 		goto drop;
770 	}
771 
772 	if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
773 		goto drop;
774 
775 	IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
776 	return true;
777 
778 drop:
779 	kfree_skb(skb);
780 	return false;
781 }
782 
783 /* CALIPSO RFC 5570 */
784 
785 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
786 {
787 	const unsigned char *nh = skb_network_header(skb);
788 
789 	if (nh[optoff + 1] < 8)
790 		goto drop;
791 
792 	if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
793 		goto drop;
794 
795 	if (!calipso_validate(skb, nh + optoff))
796 		goto drop;
797 
798 	return true;
799 
800 drop:
801 	kfree_skb(skb);
802 	return false;
803 }
804 
805 static const struct tlvtype_proc tlvprochopopt_lst[] = {
806 	{
807 		.type	= IPV6_TLV_ROUTERALERT,
808 		.func	= ipv6_hop_ra,
809 	},
810 	{
811 		.type	= IPV6_TLV_JUMBO,
812 		.func	= ipv6_hop_jumbo,
813 	},
814 	{
815 		.type	= IPV6_TLV_CALIPSO,
816 		.func	= ipv6_hop_calipso,
817 	},
818 	{ -1, }
819 };
820 
821 int ipv6_parse_hopopts(struct sk_buff *skb)
822 {
823 	struct inet6_skb_parm *opt = IP6CB(skb);
824 	struct net *net = dev_net(skb->dev);
825 	int extlen;
826 
827 	/*
828 	 * skb_network_header(skb) is equal to skb->data, and
829 	 * skb_network_header_len(skb) is always equal to
830 	 * sizeof(struct ipv6hdr) by definition of
831 	 * hop-by-hop options.
832 	 */
833 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
834 	    !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
835 				 ((skb_transport_header(skb)[1] + 1) << 3)))) {
836 fail_and_free:
837 		kfree_skb(skb);
838 		return -1;
839 	}
840 
841 	extlen = (skb_transport_header(skb)[1] + 1) << 3;
842 	if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
843 		goto fail_and_free;
844 
845 	opt->flags |= IP6SKB_HOPBYHOP;
846 	if (ip6_parse_tlv(tlvprochopopt_lst, skb,
847 			  init_net.ipv6.sysctl.max_hbh_opts_cnt)) {
848 		skb->transport_header += extlen;
849 		opt = IP6CB(skb);
850 		opt->nhoff = sizeof(struct ipv6hdr);
851 		return 1;
852 	}
853 	return -1;
854 }
855 
856 /*
857  *	Creating outbound headers.
858  *
859  *	"build" functions work when skb is filled from head to tail (datagram)
860  *	"push"	functions work when headers are added from tail to head (tcp)
861  *
862  *	In both cases we assume, that caller reserved enough room
863  *	for headers.
864  */
865 
866 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
867 			     struct ipv6_rt_hdr *opt,
868 			     struct in6_addr **addr_p, struct in6_addr *saddr)
869 {
870 	struct rt0_hdr *phdr, *ihdr;
871 	int hops;
872 
873 	ihdr = (struct rt0_hdr *) opt;
874 
875 	phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
876 	memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
877 
878 	hops = ihdr->rt_hdr.hdrlen >> 1;
879 
880 	if (hops > 1)
881 		memcpy(phdr->addr, ihdr->addr + 1,
882 		       (hops - 1) * sizeof(struct in6_addr));
883 
884 	phdr->addr[hops - 1] = **addr_p;
885 	*addr_p = ihdr->addr;
886 
887 	phdr->rt_hdr.nexthdr = *proto;
888 	*proto = NEXTHDR_ROUTING;
889 }
890 
891 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
892 			     struct ipv6_rt_hdr *opt,
893 			     struct in6_addr **addr_p, struct in6_addr *saddr)
894 {
895 	struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
896 	int plen, hops;
897 
898 	sr_ihdr = (struct ipv6_sr_hdr *)opt;
899 	plen = (sr_ihdr->hdrlen + 1) << 3;
900 
901 	sr_phdr = skb_push(skb, plen);
902 	memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
903 
904 	hops = sr_ihdr->first_segment + 1;
905 	memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
906 	       (hops - 1) * sizeof(struct in6_addr));
907 
908 	sr_phdr->segments[0] = **addr_p;
909 	*addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
910 
911 	if (sr_ihdr->hdrlen > hops * 2) {
912 		int tlvs_offset, tlvs_length;
913 
914 		tlvs_offset = (1 + hops * 2) << 3;
915 		tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
916 		memcpy((char *)sr_phdr + tlvs_offset,
917 		       (char *)sr_ihdr + tlvs_offset, tlvs_length);
918 	}
919 
920 #ifdef CONFIG_IPV6_SEG6_HMAC
921 	if (sr_has_hmac(sr_phdr)) {
922 		struct net *net = NULL;
923 
924 		if (skb->dev)
925 			net = dev_net(skb->dev);
926 		else if (skb->sk)
927 			net = sock_net(skb->sk);
928 
929 		WARN_ON(!net);
930 
931 		if (net)
932 			seg6_push_hmac(net, saddr, sr_phdr);
933 	}
934 #endif
935 
936 	sr_phdr->nexthdr = *proto;
937 	*proto = NEXTHDR_ROUTING;
938 }
939 
940 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
941 			    struct ipv6_rt_hdr *opt,
942 			    struct in6_addr **addr_p, struct in6_addr *saddr)
943 {
944 	switch (opt->type) {
945 	case IPV6_SRCRT_TYPE_0:
946 	case IPV6_SRCRT_STRICT:
947 	case IPV6_SRCRT_TYPE_2:
948 		ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
949 		break;
950 	case IPV6_SRCRT_TYPE_4:
951 		ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
952 		break;
953 	default:
954 		break;
955 	}
956 }
957 
958 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
959 {
960 	struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
961 
962 	memcpy(h, opt, ipv6_optlen(opt));
963 	h->nexthdr = *proto;
964 	*proto = type;
965 }
966 
967 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
968 			  u8 *proto,
969 			  struct in6_addr **daddr, struct in6_addr *saddr)
970 {
971 	if (opt->srcrt) {
972 		ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
973 		/*
974 		 * IPV6_RTHDRDSTOPTS is ignored
975 		 * unless IPV6_RTHDR is set (RFC3542).
976 		 */
977 		if (opt->dst0opt)
978 			ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
979 	}
980 	if (opt->hopopt)
981 		ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
982 }
983 
984 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
985 {
986 	if (opt->dst1opt)
987 		ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
988 }
989 EXPORT_SYMBOL(ipv6_push_frag_opts);
990 
991 struct ipv6_txoptions *
992 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
993 {
994 	struct ipv6_txoptions *opt2;
995 
996 	opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
997 	if (opt2) {
998 		long dif = (char *)opt2 - (char *)opt;
999 		memcpy(opt2, opt, opt->tot_len);
1000 		if (opt2->hopopt)
1001 			*((char **)&opt2->hopopt) += dif;
1002 		if (opt2->dst0opt)
1003 			*((char **)&opt2->dst0opt) += dif;
1004 		if (opt2->dst1opt)
1005 			*((char **)&opt2->dst1opt) += dif;
1006 		if (opt2->srcrt)
1007 			*((char **)&opt2->srcrt) += dif;
1008 		refcount_set(&opt2->refcnt, 1);
1009 	}
1010 	return opt2;
1011 }
1012 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1013 
1014 static void ipv6_renew_option(int renewtype,
1015 			      struct ipv6_opt_hdr **dest,
1016 			      struct ipv6_opt_hdr *old,
1017 			      struct ipv6_opt_hdr *new,
1018 			      int newtype, char **p)
1019 {
1020 	struct ipv6_opt_hdr *src;
1021 
1022 	src = (renewtype == newtype ? new : old);
1023 	if (!src)
1024 		return;
1025 
1026 	memcpy(*p, src, ipv6_optlen(src));
1027 	*dest = (struct ipv6_opt_hdr *)*p;
1028 	*p += CMSG_ALIGN(ipv6_optlen(*dest));
1029 }
1030 
1031 /**
1032  * ipv6_renew_options - replace a specific ext hdr with a new one.
1033  *
1034  * @sk: sock from which to allocate memory
1035  * @opt: original options
1036  * @newtype: option type to replace in @opt
1037  * @newopt: new option of type @newtype to replace (user-mem)
1038  * @newoptlen: length of @newopt
1039  *
1040  * Returns a new set of options which is a copy of @opt with the
1041  * option type @newtype replaced with @newopt.
1042  *
1043  * @opt may be NULL, in which case a new set of options is returned
1044  * containing just @newopt.
1045  *
1046  * @newopt may be NULL, in which case the specified option type is
1047  * not copied into the new set of options.
1048  *
1049  * The new set of options is allocated from the socket option memory
1050  * buffer of @sk.
1051  */
1052 struct ipv6_txoptions *
1053 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1054 		   int newtype, struct ipv6_opt_hdr *newopt)
1055 {
1056 	int tot_len = 0;
1057 	char *p;
1058 	struct ipv6_txoptions *opt2;
1059 
1060 	if (opt) {
1061 		if (newtype != IPV6_HOPOPTS && opt->hopopt)
1062 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1063 		if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1064 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1065 		if (newtype != IPV6_RTHDR && opt->srcrt)
1066 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1067 		if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1068 			tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1069 	}
1070 
1071 	if (newopt)
1072 		tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1073 
1074 	if (!tot_len)
1075 		return NULL;
1076 
1077 	tot_len += sizeof(*opt2);
1078 	opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1079 	if (!opt2)
1080 		return ERR_PTR(-ENOBUFS);
1081 
1082 	memset(opt2, 0, tot_len);
1083 	refcount_set(&opt2->refcnt, 1);
1084 	opt2->tot_len = tot_len;
1085 	p = (char *)(opt2 + 1);
1086 
1087 	ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1088 			  (opt ? opt->hopopt : NULL),
1089 			  newopt, newtype, &p);
1090 	ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1091 			  (opt ? opt->dst0opt : NULL),
1092 			  newopt, newtype, &p);
1093 	ipv6_renew_option(IPV6_RTHDR,
1094 			  (struct ipv6_opt_hdr **)&opt2->srcrt,
1095 			  (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1096 			  newopt, newtype, &p);
1097 	ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1098 			  (opt ? opt->dst1opt : NULL),
1099 			  newopt, newtype, &p);
1100 
1101 	opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1102 			  (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1103 			  (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1104 	opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1105 
1106 	return opt2;
1107 }
1108 
1109 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1110 					  struct ipv6_txoptions *opt)
1111 {
1112 	/*
1113 	 * ignore the dest before srcrt unless srcrt is being included.
1114 	 * --yoshfuji
1115 	 */
1116 	if (opt && opt->dst0opt && !opt->srcrt) {
1117 		if (opt_space != opt) {
1118 			memcpy(opt_space, opt, sizeof(*opt_space));
1119 			opt = opt_space;
1120 		}
1121 		opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1122 		opt->dst0opt = NULL;
1123 	}
1124 
1125 	return opt;
1126 }
1127 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
1128 
1129 /**
1130  * fl6_update_dst - update flowi destination address with info given
1131  *                  by srcrt option, if any.
1132  *
1133  * @fl6: flowi6 for which daddr is to be updated
1134  * @opt: struct ipv6_txoptions in which to look for srcrt opt
1135  * @orig: copy of original daddr address if modified
1136  *
1137  * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1138  * and initial value of fl6->daddr set in orig
1139  */
1140 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1141 				const struct ipv6_txoptions *opt,
1142 				struct in6_addr *orig)
1143 {
1144 	if (!opt || !opt->srcrt)
1145 		return NULL;
1146 
1147 	*orig = fl6->daddr;
1148 
1149 	switch (opt->srcrt->type) {
1150 	case IPV6_SRCRT_TYPE_0:
1151 	case IPV6_SRCRT_STRICT:
1152 	case IPV6_SRCRT_TYPE_2:
1153 		fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1154 		break;
1155 	case IPV6_SRCRT_TYPE_4:
1156 	{
1157 		struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1158 
1159 		fl6->daddr = srh->segments[srh->segments_left];
1160 		break;
1161 	}
1162 	default:
1163 		return NULL;
1164 	}
1165 
1166 	return orig;
1167 }
1168 EXPORT_SYMBOL_GPL(fl6_update_dst);
1169