xref: /openbsd/sys/netinet6/ip6_input.c (revision 9c4c1a6f)
1 /*	$OpenBSD: ip6_input.c,v 1.266 2024/07/19 16:58:32 bluhm Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.188 2001/03/29 05:34:31 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "pf.h"
65 #include "carp.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/mbuf.h>
70 #include <sys/domain.h>
71 #include <sys/sysctl.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/errno.h>
76 #include <sys/time.h>
77 #include <sys/timeout.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/task.h>
81 
82 #include <net/if.h>
83 #include <net/if_var.h>
84 #include <net/if_types.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 
88 #include <netinet/in.h>
89 
90 #include <netinet/ip.h>
91 
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet6/in6_var.h>
95 #include <netinet6/in6_ifattach.h>
96 #include <netinet/ip6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/nd6.h>
100 
101 #include "gif.h"
102 #include "bpfilter.h"
103 
104 #ifdef MROUTING
105 #include <netinet6/ip6_mroute.h>
106 #endif
107 
108 #if NPF > 0
109 #include <net/pfvar.h>
110 #endif
111 
112 #if NCARP > 0
113 #include <netinet/ip_carp.h>
114 #endif
115 
116 struct niqueue ip6intrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IPV6);
117 
118 struct cpumem *ip6counters;
119 
120 uint8_t ip6_soiikey[IP6_SOIIKEY_LEN];
121 
122 int ip6_ours(struct mbuf **, int *, int, int, int);
123 int ip6_check_rh0hdr(struct mbuf *, int *);
124 int ip6_hbhchcheck(struct mbuf **, int *, int *, int);
125 int ip6_hopopts_input(struct mbuf **, int *, u_int32_t *, u_int32_t *);
126 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
127 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
128 
129 static struct mbuf_queue	ip6send_mq;
130 
131 static void ip6_send_dispatch(void *);
132 static struct task ip6send_task =
133 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
134 
135 /*
136  * IP6 initialization: fill in IP6 protocol switch table.
137  * All protocols not implemented in kernel go to raw IP6 protocol handler.
138  */
139 void
ip6_init(void)140 ip6_init(void)
141 {
142 	const struct protosw *pr;
143 	int i;
144 
145 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
146 	if (pr == NULL)
147 		panic("%s", __func__);
148 	for (i = 0; i < IPPROTO_MAX; i++)
149 		ip6_protox[i] = pr - inet6sw;
150 	for (pr = inet6domain.dom_protosw;
151 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
152 		if (pr->pr_domain->dom_family == PF_INET6 &&
153 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
154 		    pr->pr_protocol < IPPROTO_MAX)
155 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
156 	ip6_randomid_init();
157 	nd6_init();
158 	frag6_init();
159 
160 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
161 
162 	ip6counters = counters_alloc(ip6s_ncounters);
163 #ifdef MROUTING
164 	rt_timer_queue_init(&ip6_mrouterq, MCAST_EXPIRE_TIMEOUT,
165 	    &mf6c_expire_route);
166 #endif
167 }
168 
169 /*
170  * Enqueue packet for local delivery.  Queuing is used as a boundary
171  * between the network layer (input/forward path) running with
172  * NET_LOCK_SHARED() and the transport layer needing it exclusively.
173  */
174 int
ip6_ours(struct mbuf ** mp,int * offp,int nxt,int af,int flags)175 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af, int flags)
176 {
177 	/* ip6_hbhchcheck() may be run before, then off and nxt are set */
178 	if (*offp == 0) {
179 		nxt = ip6_hbhchcheck(mp, offp, NULL, flags);
180 		if (nxt == IPPROTO_DONE)
181 			return IPPROTO_DONE;
182 	}
183 
184 	/* We are already in a IPv4/IPv6 local deliver loop. */
185 	if (af != AF_UNSPEC)
186 		return nxt;
187 
188 	nxt = ip_deliver(mp, offp, nxt, AF_INET6, 1);
189 	if (nxt == IPPROTO_DONE)
190 		return IPPROTO_DONE;
191 
192 	/* save values for later, use after dequeue */
193 	if (*offp != sizeof(struct ip6_hdr)) {
194 		struct m_tag *mtag;
195 		struct ipoffnxt *ion;
196 
197 		/* mbuf tags are expensive, but only used for header options */
198 		mtag = m_tag_get(PACKET_TAG_IP6_OFFNXT, sizeof(*ion),
199 		    M_NOWAIT);
200 		if (mtag == NULL) {
201 			ip6stat_inc(ip6s_idropped);
202 			m_freemp(mp);
203 			return IPPROTO_DONE;
204 		}
205 		ion = (struct ipoffnxt *)(mtag + 1);
206 		ion->ion_off = *offp;
207 		ion->ion_nxt = nxt;
208 
209 		m_tag_prepend(*mp, mtag);
210 	}
211 
212 	niq_enqueue(&ip6intrq, *mp);
213 	*mp = NULL;
214 	return IPPROTO_DONE;
215 }
216 
217 /*
218  * Dequeue and process locally delivered packets.
219  * This is called with exclusive NET_LOCK().
220  */
221 void
ip6intr(void)222 ip6intr(void)
223 {
224 	struct mbuf *m;
225 
226 	while ((m = niq_dequeue(&ip6intrq)) != NULL) {
227 		struct m_tag *mtag;
228 		int off, nxt;
229 
230 #ifdef DIAGNOSTIC
231 		if ((m->m_flags & M_PKTHDR) == 0)
232 			panic("ip6intr no HDR");
233 #endif
234 		mtag = m_tag_find(m, PACKET_TAG_IP6_OFFNXT, NULL);
235 		if (mtag != NULL) {
236 			struct ipoffnxt *ion;
237 
238 			ion = (struct ipoffnxt *)(mtag + 1);
239 			off = ion->ion_off;
240 			nxt = ion->ion_nxt;
241 
242 			m_tag_delete(m, mtag);
243 		} else {
244 			struct ip6_hdr *ip6;
245 
246 			ip6 = mtod(m, struct ip6_hdr *);
247 			off = sizeof(struct ip6_hdr);
248 			nxt = ip6->ip6_nxt;
249 		}
250 		nxt = ip_deliver(&m, &off, nxt, AF_INET6, 0);
251 		KASSERT(nxt == IPPROTO_DONE);
252 	}
253 }
254 
255 void
ipv6_input(struct ifnet * ifp,struct mbuf * m)256 ipv6_input(struct ifnet *ifp, struct mbuf *m)
257 {
258 	int off, nxt;
259 
260 	off = 0;
261 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
262 	KASSERT(nxt == IPPROTO_DONE);
263 }
264 
265 struct mbuf *
ipv6_check(struct ifnet * ifp,struct mbuf * m)266 ipv6_check(struct ifnet *ifp, struct mbuf *m)
267 {
268 	struct ip6_hdr *ip6;
269 
270 	if (m->m_len < sizeof(*ip6)) {
271 		m = m_pullup(m, sizeof(*ip6));
272 		if (m == NULL) {
273 			ip6stat_inc(ip6s_toosmall);
274 			return (NULL);
275 		}
276 	}
277 
278 	ip6 = mtod(m, struct ip6_hdr *);
279 
280 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
281 		ip6stat_inc(ip6s_badvers);
282 		goto bad;
283 	}
284 
285 	/*
286 	 * Check against address spoofing/corruption.
287 	 */
288 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
289 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
290 		/*
291 		 * XXX: "badscope" is not very suitable for a multicast source.
292 		 */
293 		ip6stat_inc(ip6s_badscope);
294 		goto bad;
295 	}
296 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
297 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
298 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
299 		ip6stat_inc(ip6s_badscope);
300 		goto bad;
301 	}
302 	/* Drop packets if interface ID portion is already filled. */
303 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
304 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
305 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
306 		ip6stat_inc(ip6s_badscope);
307 		goto bad;
308 	}
309 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
310 	    !(m->m_flags & M_LOOP)) {
311 		/*
312 		 * In this case, the packet should come from the loopback
313 		 * interface.  However, we cannot just check the if_flags,
314 		 * because ip6_mloopback() passes the "actual" interface
315 		 * as the outgoing/incoming interface.
316 		 */
317 		ip6stat_inc(ip6s_badscope);
318 		goto bad;
319 	}
320 
321 	/*
322 	 * The following check is not documented in specs.  A malicious
323 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
324 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
325 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
326 	 *
327 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
328 	 * support IPv4-less kernel compilation, we cannot support SIIT
329 	 * environment at all.  So, it makes more sense for us to reject any
330 	 * malicious packets for non-SIIT environment, than try to do a
331 	 * partial support for SIIT environment.
332 	 */
333 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
334 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
335 		ip6stat_inc(ip6s_badscope);
336 		goto bad;
337 	}
338 
339 	/*
340 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
341 	 *
342 	 * The code forbids automatic tunneling as per RFC4213.
343 	 */
344 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
345 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
346 		ip6stat_inc(ip6s_badscope);
347 		goto bad;
348 	}
349 
350 	return (m);
351 bad:
352 	m_freem(m);
353 	return (NULL);
354 }
355 
356 int
ip6_input_if(struct mbuf ** mp,int * offp,int nxt,int af,struct ifnet * ifp)357 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
358 {
359 	struct route ro;
360 	struct mbuf *m;
361 	struct ip6_hdr *ip6;
362 	struct rtentry *rt;
363 	int ours = 0;
364 	u_int16_t src_scope, dst_scope;
365 #if NPF > 0
366 	struct in6_addr odst;
367 #endif
368 	int flags = 0;
369 
370 	KASSERT(*offp == 0);
371 
372 	ro.ro_rt = NULL;
373 	ip6stat_inc(ip6s_total);
374 	m = *mp = ipv6_check(ifp, *mp);
375 	if (m == NULL)
376 		goto bad;
377 
378 	ip6 = mtod(m, struct ip6_hdr *);
379 
380 #if NCARP > 0
381 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
382 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
383 		goto bad;
384 #endif
385 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
386 
387 	/*
388 	 * If the packet has been received on a loopback interface it
389 	 * can be destined to any local address, not necessarily to
390 	 * an address configured on `ifp'.
391 	 */
392 	if (ifp->if_flags & IFF_LOOPBACK) {
393 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
394 			src_scope = ip6->ip6_src.s6_addr16[1];
395 			ip6->ip6_src.s6_addr16[1] = 0;
396 		}
397 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
398 			dst_scope = ip6->ip6_dst.s6_addr16[1];
399 			ip6->ip6_dst.s6_addr16[1] = 0;
400 		}
401 	}
402 
403 #if NPF > 0
404 	/*
405 	 * Packet filter
406 	 */
407 	odst = ip6->ip6_dst;
408 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
409 		goto bad;
410 	m = *mp;
411 	if (m == NULL)
412 		goto bad;
413 
414 	ip6 = mtod(m, struct ip6_hdr *);
415 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst))
416 		SET(flags, IPV6_REDIRECT);
417 #endif
418 
419 	switch (atomic_load_int(&ip6_forwarding)) {
420 	case 2:
421 		SET(flags, IPV6_FORWARDING_IPSEC);
422 		/* FALLTHROUGH */
423 	case 1:
424 		SET(flags, IPV6_FORWARDING);
425 		break;
426 	}
427 
428 	/*
429 	 * Without embedded scope ID we cannot find link-local
430 	 * addresses in the routing table.
431 	 */
432 	if (ifp->if_flags & IFF_LOOPBACK) {
433 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
434 			ip6->ip6_src.s6_addr16[1] = src_scope;
435 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
436 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
437 	} else {
438 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
439 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
440 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
441 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
442 	}
443 
444 	/*
445 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
446 	 * If pf has already scanned the header chain, do not do it twice.
447 	 */
448 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
449 	    ip6_check_rh0hdr(m, offp)) {
450 		ip6stat_inc(ip6s_badoptions);
451 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
452 		m = *mp = NULL;
453 		goto bad;
454 	}
455 
456 #if NPF > 0
457 	if (pf_ouraddr(m) == 1) {
458 		nxt = ip6_ours(mp, offp, nxt, af, flags);
459 		goto out;
460 	}
461 #endif
462 
463 	/*
464 	 * Multicast check
465 	 */
466 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
467 		/*
468 		 * Make sure M_MCAST is set.  It should theoretically
469 		 * already be there, but let's play safe because upper
470 		 * layers check for this flag.
471 		 */
472 		m->m_flags |= M_MCAST;
473 
474 		/*
475 		 * See if we belong to the destination multicast group on the
476 		 * arrival interface.
477 		 */
478 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
479 			ours = 1;
480 
481 #ifdef MROUTING
482 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
483 			int error;
484 
485 			nxt = ip6_hbhchcheck(&m, offp, &ours, flags);
486 			if (nxt == IPPROTO_DONE)
487 				goto out;
488 
489 			ip6 = mtod(m, struct ip6_hdr *);
490 
491 			/*
492 			 * If we are acting as a multicast router, all
493 			 * incoming multicast packets are passed to the
494 			 * kernel-level multicast forwarding function.
495 			 * The packet is returned (relatively) intact; if
496 			 * ip6_mforward() returns a non-zero value, the packet
497 			 * must be discarded, else it may be accepted below.
498 			 */
499 			KERNEL_LOCK();
500 			error = ip6_mforward(ip6, ifp, m, flags);
501 			KERNEL_UNLOCK();
502 			if (error) {
503 				ip6stat_inc(ip6s_cantforward);
504 				goto bad;
505 			}
506 
507 			if (ours) {
508 				if (af == AF_UNSPEC)
509 					nxt = ip6_ours(mp, offp, nxt, af,
510 					    flags);
511 				goto out;
512 			}
513 			goto bad;
514 		}
515 #endif
516 		if (!ours) {
517 			ip6stat_inc(ip6s_notmember);
518 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
519 				ip6stat_inc(ip6s_cantforward);
520 			goto bad;
521 		}
522 		nxt = ip6_ours(mp, offp, nxt, af, flags);
523 		goto out;
524 	}
525 
526 
527 	/*
528 	 *  Unicast check
529 	 */
530 	rt = route6_mpath(&ro, &ip6->ip6_dst, &ip6->ip6_src,
531 	    m->m_pkthdr.ph_rtableid);
532 
533 	/*
534 	 * Accept the packet if the route to the destination is marked
535 	 * as local.
536 	 */
537 	if (rt != NULL && ISSET(rt->rt_flags, RTF_LOCAL)) {
538 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
539 
540 		if (!ISSET(flags, IPV6_FORWARDING) &&
541 		    rt->rt_ifidx != ifp->if_index &&
542 		    !((ifp->if_flags & IFF_LOOPBACK) ||
543 		    (ifp->if_type == IFT_ENC) ||
544 		    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
545 			/* received on wrong interface */
546 #if NCARP > 0
547 			struct ifnet *out_if;
548 
549 			/*
550 			 * Virtual IPs on carp interfaces need to be checked
551 			 * also against the parent interface and other carp
552 			 * interfaces sharing the same parent.
553 			 */
554 			out_if = if_get(rt->rt_ifidx);
555 			if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
556 				ip6stat_inc(ip6s_wrongif);
557 				if_put(out_if);
558 				goto bad;
559 			}
560 			if_put(out_if);
561 #else
562 			ip6stat_inc(ip6s_wrongif);
563 			goto bad;
564 #endif
565 		}
566 		/*
567 		 * packets to a tentative, duplicated, or somehow invalid
568 		 * address must not be accepted.
569 		 */
570 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
571 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
572 
573 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
574 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
575 			/* address is not ready, so discard the packet. */
576 			nd6log((LOG_INFO,
577 			    "%s: packet to an unready address %s->%s\n",
578 			    __func__, src, dst));
579 
580 			goto bad;
581 		} else {
582 			nxt = ip6_ours(mp, offp, nxt, af, flags);
583 			goto out;
584 		}
585 	}
586 
587 #if NCARP > 0
588 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
589 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
590 	    ip6->ip6_dst.s6_addr32, 1))
591 		goto bad;
592 #endif
593 	/*
594 	 * Now there is no reason to process the packet if it's not our own
595 	 * and we're not a router.
596 	 */
597 	if (!ISSET(flags, IPV6_FORWARDING)) {
598 		ip6stat_inc(ip6s_cantforward);
599 		goto bad;
600 	}
601 
602 	nxt = ip6_hbhchcheck(&m, offp, &ours, flags);
603 	if (nxt == IPPROTO_DONE)
604 		goto out;
605 
606 	if (ours) {
607 		if (af == AF_UNSPEC)
608 			nxt = ip6_ours(mp, offp, nxt, af, flags);
609 		goto out;
610 	}
611 
612 #ifdef IPSEC
613 	if (ipsec_in_use) {
614 		int rv;
615 
616 		rv = ipsec_forward_check(m, *offp, AF_INET6);
617 		if (rv != 0) {
618 			ip6stat_inc(ip6s_cantforward);
619 			goto bad;
620 		}
621 		/*
622 		 * Fall through, forward packet. Outbound IPsec policy
623 		 * checking will occur in ip6_forward().
624 		 */
625 	}
626 #endif /* IPSEC */
627 
628 	ip6_forward(m, &ro, flags);
629 	*mp = NULL;
630 	rtfree(ro.ro_rt);
631 	return IPPROTO_DONE;
632  bad:
633 	nxt = IPPROTO_DONE;
634 	m_freemp(mp);
635  out:
636 	rtfree(ro.ro_rt);
637 	return nxt;
638 }
639 
640 /* On error free mbuf and return IPPROTO_DONE. */
641 int
ip6_hbhchcheck(struct mbuf ** mp,int * offp,int * oursp,int flags)642 ip6_hbhchcheck(struct mbuf **mp, int *offp, int *oursp, int flags)
643 {
644 	struct ip6_hdr *ip6;
645 	u_int32_t plen, rtalert = ~0;
646 	int nxt;
647 
648 	ip6 = mtod(*mp, struct ip6_hdr *);
649 
650 	/*
651 	 * Process Hop-by-Hop options header if it's contained.
652 	 * m may be modified in ip6_hopopts_input().
653 	 * If a JumboPayload option is included, plen will also be modified.
654 	 */
655 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
656 	*offp = sizeof(struct ip6_hdr);
657 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
658 		struct ip6_hbh *hbh;
659 
660 		if (ip6_hopopts_input(mp, offp, &plen, &rtalert))
661 			goto bad;	/* m have already been freed */
662 
663 		/* adjust pointer */
664 		ip6 = mtod(*mp, struct ip6_hdr *);
665 
666 		/*
667 		 * if the payload length field is 0 and the next header field
668 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
669 		 * option MUST be included.
670 		 */
671 		if (ip6->ip6_plen == 0 && plen == 0) {
672 			/*
673 			 * Note that if a valid jumbo payload option is
674 			 * contained, ip6_hopopts_input() must set a valid
675 			 * (non-zero) payload length to the variable plen.
676 			 */
677 			ip6stat_inc(ip6s_badoptions);
678 			icmp6_error(*mp, ICMP6_PARAM_PROB,
679 				    ICMP6_PARAMPROB_HEADER,
680 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
681 			goto bad;
682 		}
683 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
684 		    sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
685 		if (hbh == NULL) {
686 			ip6stat_inc(ip6s_tooshort);
687 			goto bad;
688 		}
689 		nxt = hbh->ip6h_nxt;
690 
691 		/*
692 		 * accept the packet if a router alert option is included
693 		 * and we act as an IPv6 router.
694 		 */
695 		if (rtalert != ~0 && ISSET(flags, IPV6_FORWARDING) &&
696 		    oursp != NULL)
697 			*oursp = 1;
698 	} else
699 		nxt = ip6->ip6_nxt;
700 
701 	/*
702 	 * Check that the amount of data in the buffers
703 	 * is as at least much as the IPv6 header would have us expect.
704 	 * Trim mbufs if longer than we expect.
705 	 * Drop packet if shorter than we expect.
706 	 */
707 	if ((*mp)->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
708 		ip6stat_inc(ip6s_tooshort);
709 		m_freemp(mp);
710 		goto bad;
711 	}
712 	if ((*mp)->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
713 		if ((*mp)->m_len == (*mp)->m_pkthdr.len) {
714 			(*mp)->m_len = sizeof(struct ip6_hdr) + plen;
715 			(*mp)->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
716 		} else {
717 			m_adj((*mp), sizeof(struct ip6_hdr) + plen -
718 			    (*mp)->m_pkthdr.len);
719 		}
720 	}
721 
722 	return nxt;
723  bad:
724 	return IPPROTO_DONE;
725 }
726 
727 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
728 int
ip6_check_rh0hdr(struct mbuf * m,int * offp)729 ip6_check_rh0hdr(struct mbuf *m, int *offp)
730 {
731 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
732 	struct ip6_rthdr rthdr;
733 	struct ip6_ext opt6;
734 	u_int8_t proto = ip6->ip6_nxt;
735 	int done = 0, lim, off, rh_cnt = 0;
736 
737 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
738 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
739 	do {
740 		switch (proto) {
741 		case IPPROTO_ROUTING:
742 			if (rh_cnt++) {
743 				/* more than one rh header present */
744 				*offp = off;
745 				return (1);
746 			}
747 
748 			if (off + sizeof(rthdr) > lim) {
749 				/* packet to short to make sense */
750 				*offp = off;
751 				return (1);
752 			}
753 
754 			m_copydata(m, off, sizeof(rthdr), &rthdr);
755 
756 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
757 				*offp = off +
758 				    offsetof(struct ip6_rthdr, ip6r_type);
759 				return (1);
760 			}
761 
762 			off += (rthdr.ip6r_len + 1) * 8;
763 			proto = rthdr.ip6r_nxt;
764 			break;
765 		case IPPROTO_AH:
766 		case IPPROTO_HOPOPTS:
767 		case IPPROTO_DSTOPTS:
768 			/* get next header and header length */
769 			if (off + sizeof(opt6) > lim) {
770 				/*
771 				 * Packet to short to make sense, we could
772 				 * reject the packet but as a router we
773 				 * should not do that so forward it.
774 				 */
775 				return (0);
776 			}
777 
778 			m_copydata(m, off, sizeof(opt6), &opt6);
779 
780 			if (proto == IPPROTO_AH)
781 				off += (opt6.ip6e_len + 2) * 4;
782 			else
783 				off += (opt6.ip6e_len + 1) * 8;
784 			proto = opt6.ip6e_nxt;
785 			break;
786 		case IPPROTO_FRAGMENT:
787 		default:
788 			/* end of header stack */
789 			done = 1;
790 			break;
791 		}
792 	} while (!done);
793 
794 	return (0);
795 }
796 
797 /*
798  * Hop-by-Hop options header processing. If a valid jumbo payload option is
799  * included, the real payload length will be stored in plenp.
800  * On error free mbuf and return -1.
801  *
802  * rtalertp - XXX: should be stored in a more smart way
803  */
804 int
ip6_hopopts_input(struct mbuf ** mp,int * offp,u_int32_t * plenp,u_int32_t * rtalertp)805 ip6_hopopts_input(struct mbuf **mp, int *offp, u_int32_t *plenp,
806     u_int32_t *rtalertp)
807 {
808 	int off = *offp, hbhlen;
809 	struct ip6_hbh *hbh;
810 
811 	/* validation of the length of the header */
812 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
813 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
814 	if (hbh == NULL) {
815 		ip6stat_inc(ip6s_tooshort);
816 		return -1;
817 	}
818 	hbhlen = (hbh->ip6h_len + 1) << 3;
819 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp, sizeof(struct ip6_hdr),
820 		hbhlen);
821 	if (hbh == NULL) {
822 		ip6stat_inc(ip6s_tooshort);
823 		return -1;
824 	}
825 	off += hbhlen;
826 	hbhlen -= sizeof(struct ip6_hbh);
827 
828 	if (ip6_process_hopopts(mp, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
829 				hbhlen, rtalertp, plenp) < 0)
830 		return (-1);
831 
832 	*offp = off;
833 	return (0);
834 }
835 
836 /*
837  * Search header for all Hop-by-hop options and process each option.
838  * This function is separate from ip6_hopopts_input() in order to
839  * handle a case where the sending node itself process its hop-by-hop
840  * options header. In such a case, the function is called from ip6_output().
841  * On error free mbuf and return -1.
842  *
843  * The function assumes that hbh header is located right after the IPv6 header
844  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
845  * opthead + hbhlen is located in continuous memory region.
846  */
847 int
ip6_process_hopopts(struct mbuf ** mp,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp,u_int32_t * plenp)848 ip6_process_hopopts(struct mbuf **mp, u_int8_t *opthead, int hbhlen,
849     u_int32_t *rtalertp, u_int32_t *plenp)
850 {
851 	struct ip6_hdr *ip6;
852 	int optlen = 0;
853 	u_int8_t *opt = opthead;
854 	u_int16_t rtalert_val;
855 	u_int32_t jumboplen;
856 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
857 
858 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
859 		switch (*opt) {
860 		case IP6OPT_PAD1:
861 			optlen = 1;
862 			break;
863 		case IP6OPT_PADN:
864 			if (hbhlen < IP6OPT_MINLEN) {
865 				ip6stat_inc(ip6s_toosmall);
866 				goto bad;
867 			}
868 			optlen = *(opt + 1) + 2;
869 			break;
870 		case IP6OPT_ROUTER_ALERT:
871 			/* XXX may need check for alignment */
872 			if (hbhlen < IP6OPT_RTALERT_LEN) {
873 				ip6stat_inc(ip6s_toosmall);
874 				goto bad;
875 			}
876 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
877 				/* XXX stat */
878 				icmp6_error(*mp, ICMP6_PARAM_PROB,
879 				    ICMP6_PARAMPROB_HEADER,
880 				    erroff + opt + 1 - opthead);
881 				return (-1);
882 			}
883 			optlen = IP6OPT_RTALERT_LEN;
884 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
885 			*rtalertp = ntohs(rtalert_val);
886 			break;
887 		case IP6OPT_JUMBO:
888 			/* XXX may need check for alignment */
889 			if (hbhlen < IP6OPT_JUMBO_LEN) {
890 				ip6stat_inc(ip6s_toosmall);
891 				goto bad;
892 			}
893 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
894 				/* XXX stat */
895 				icmp6_error(*mp, ICMP6_PARAM_PROB,
896 				    ICMP6_PARAMPROB_HEADER,
897 				    erroff + opt + 1 - opthead);
898 				return (-1);
899 			}
900 			optlen = IP6OPT_JUMBO_LEN;
901 
902 			/*
903 			 * IPv6 packets that have non 0 payload length
904 			 * must not contain a jumbo payload option.
905 			 */
906 			ip6 = mtod(*mp, struct ip6_hdr *);
907 			if (ip6->ip6_plen) {
908 				ip6stat_inc(ip6s_badoptions);
909 				icmp6_error(*mp, ICMP6_PARAM_PROB,
910 				    ICMP6_PARAMPROB_HEADER,
911 				    erroff + opt - opthead);
912 				return (-1);
913 			}
914 
915 			/*
916 			 * We may see jumbolen in unaligned location, so
917 			 * we'd need to perform memcpy().
918 			 */
919 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
920 			jumboplen = (u_int32_t)htonl(jumboplen);
921 
922 #if 1
923 			/*
924 			 * if there are multiple jumbo payload options,
925 			 * *plenp will be non-zero and the packet will be
926 			 * rejected.
927 			 * the behavior may need some debate in ipngwg -
928 			 * multiple options does not make sense, however,
929 			 * there's no explicit mention in specification.
930 			 */
931 			if (*plenp != 0) {
932 				ip6stat_inc(ip6s_badoptions);
933 				icmp6_error(*mp, ICMP6_PARAM_PROB,
934 				    ICMP6_PARAMPROB_HEADER,
935 				    erroff + opt + 2 - opthead);
936 				return (-1);
937 			}
938 #endif
939 
940 			/*
941 			 * jumbo payload length must be larger than 65535.
942 			 */
943 			if (jumboplen <= IPV6_MAXPACKET) {
944 				ip6stat_inc(ip6s_badoptions);
945 				icmp6_error(*mp, ICMP6_PARAM_PROB,
946 				    ICMP6_PARAMPROB_HEADER,
947 				    erroff + opt + 2 - opthead);
948 				return (-1);
949 			}
950 			*plenp = jumboplen;
951 
952 			break;
953 		default:		/* unknown option */
954 			if (hbhlen < IP6OPT_MINLEN) {
955 				ip6stat_inc(ip6s_toosmall);
956 				goto bad;
957 			}
958 			optlen = ip6_unknown_opt(mp, opt,
959 			    erroff + opt - opthead);
960 			if (optlen == -1)
961 				return (-1);
962 			optlen += 2;
963 			break;
964 		}
965 	}
966 
967 	return (0);
968 
969   bad:
970 	m_freemp(mp);
971 	return (-1);
972 }
973 
974 /*
975  * Unknown option processing.
976  * The third argument `off' is the offset from the IPv6 header to the option,
977  * which allows returning an ICMPv6 error even if the IPv6 header and the
978  * option header are not continuous.
979  * On error free mbuf and return -1.
980  */
981 int
ip6_unknown_opt(struct mbuf ** mp,u_int8_t * optp,int off)982 ip6_unknown_opt(struct mbuf **mp, u_int8_t *optp, int off)
983 {
984 	struct ip6_hdr *ip6;
985 
986 	switch (IP6OPT_TYPE(*optp)) {
987 	case IP6OPT_TYPE_SKIP: /* ignore the option */
988 		return ((int)*(optp + 1));
989 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
990 		m_freemp(mp);
991 		return (-1);
992 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
993 		ip6stat_inc(ip6s_badoptions);
994 		icmp6_error(*mp, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
995 		return (-1);
996 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
997 		ip6stat_inc(ip6s_badoptions);
998 		ip6 = mtod(*mp, struct ip6_hdr *);
999 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1000 		    ((*mp)->m_flags & (M_BCAST|M_MCAST)))
1001 			m_freemp(mp);
1002 		else
1003 			icmp6_error(*mp, ICMP6_PARAM_PROB,
1004 				    ICMP6_PARAMPROB_OPTION, off);
1005 		return (-1);
1006 	}
1007 
1008 	m_freemp(mp);		/* XXX: NOTREACHED */
1009 	return (-1);
1010 }
1011 
1012 /*
1013  * Create the "control" list for this pcb.
1014  *
1015  * The routine will be called from upper layer handlers like udp_input().
1016  * Thus the routine assumes that the caller (udp_input) have already
1017  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1018  * very first mbuf on the mbuf chain.
1019  * We may want to add some infinite loop prevention or sanity checks for safety.
1020  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1021  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1022  */
1023 void
ip6_savecontrol(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp)1024 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1025 {
1026 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1027 
1028 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1029 		struct timeval tv;
1030 
1031 		m_microtime(m, &tv);
1032 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1033 		    SCM_TIMESTAMP, SOL_SOCKET);
1034 		if (*mp)
1035 			mp = &(*mp)->m_next;
1036 	}
1037 
1038 	/* RFC 2292 sec. 5 */
1039 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1040 		struct in6_pktinfo pi6;
1041 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
1042 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
1043 			pi6.ipi6_addr.s6_addr16[1] = 0;
1044 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
1045 		*mp = sbcreatecontrol((caddr_t) &pi6,
1046 		    sizeof(struct in6_pktinfo),
1047 		    IPV6_PKTINFO, IPPROTO_IPV6);
1048 		if (*mp)
1049 			mp = &(*mp)->m_next;
1050 	}
1051 
1052 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1053 		int hlim = ip6->ip6_hlim & 0xff;
1054 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1055 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
1056 		if (*mp)
1057 			mp = &(*mp)->m_next;
1058 	}
1059 
1060 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1061 		u_int32_t flowinfo;
1062 		int tclass;
1063 
1064 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1065 		flowinfo >>= 20;
1066 
1067 		tclass = flowinfo & 0xff;
1068 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
1069 		    IPV6_TCLASS, IPPROTO_IPV6);
1070 		if (*mp)
1071 			mp = &(*mp)->m_next;
1072 	}
1073 
1074 	/*
1075 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1076 	 * privilege for the option (see ip6_ctloutput), but it might be too
1077 	 * strict, since there might be some hop-by-hop options which can be
1078 	 * returned to normal user.
1079 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1080 	 */
1081 	if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1082 		/*
1083 		 * Check if a hop-by-hop options header is contained in the
1084 		 * received packet, and if so, store the options as ancillary
1085 		 * data. Note that a hop-by-hop options header must be
1086 		 * just after the IPv6 header, which is assured through the
1087 		 * IPv6 input processing.
1088 		 */
1089 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1090 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1091 			struct ip6_hbh *hbh;
1092 			int hbhlen = 0;
1093 			struct mbuf *ext;
1094 
1095 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1096 			    ip6->ip6_nxt);
1097 			if (ext == NULL) {
1098 				ip6stat_inc(ip6s_tooshort);
1099 				return;
1100 			}
1101 			hbh = mtod(ext, struct ip6_hbh *);
1102 			hbhlen = (hbh->ip6h_len + 1) << 3;
1103 			if (hbhlen != ext->m_len) {
1104 				m_freem(ext);
1105 				ip6stat_inc(ip6s_tooshort);
1106 				return;
1107 			}
1108 
1109 			/*
1110 			 * XXX: We copy the whole header even if a
1111 			 * jumbo payload option is included, the option which
1112 			 * is to be removed before returning according to
1113 			 * RFC2292.
1114 			 * Note: this constraint is removed in RFC3542.
1115 			 */
1116 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1117 			    IPV6_HOPOPTS,
1118 			    IPPROTO_IPV6);
1119 			if (*mp)
1120 				mp = &(*mp)->m_next;
1121 			m_freem(ext);
1122 		}
1123 	}
1124 
1125 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1126 	if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1127 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1128 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1129 
1130 		/*
1131 		 * Search for destination options headers or routing
1132 		 * header(s) through the header chain, and stores each
1133 		 * header as ancillary data.
1134 		 * Note that the order of the headers remains in
1135 		 * the chain of ancillary data.
1136 		 */
1137 		while (1) {	/* is explicit loop prevention necessary? */
1138 			struct ip6_ext *ip6e = NULL;
1139 			int elen;
1140 			struct mbuf *ext = NULL;
1141 
1142 			/*
1143 			 * if it is not an extension header, don't try to
1144 			 * pull it from the chain.
1145 			 */
1146 			switch (nxt) {
1147 			case IPPROTO_DSTOPTS:
1148 			case IPPROTO_ROUTING:
1149 			case IPPROTO_HOPOPTS:
1150 			case IPPROTO_AH: /* is it possible? */
1151 				break;
1152 			default:
1153 				goto loopend;
1154 			}
1155 
1156 			ext = ip6_pullexthdr(m, off, nxt);
1157 			if (ext == NULL) {
1158 				ip6stat_inc(ip6s_tooshort);
1159 				return;
1160 			}
1161 			ip6e = mtod(ext, struct ip6_ext *);
1162 			if (nxt == IPPROTO_AH)
1163 				elen = (ip6e->ip6e_len + 2) << 2;
1164 			else
1165 				elen = (ip6e->ip6e_len + 1) << 3;
1166 			if (elen != ext->m_len) {
1167 				m_freem(ext);
1168 				ip6stat_inc(ip6s_tooshort);
1169 				return;
1170 			}
1171 
1172 			switch (nxt) {
1173 			case IPPROTO_DSTOPTS:
1174 				if (!(inp->inp_flags & IN6P_DSTOPTS))
1175 					break;
1176 
1177 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1178 				    IPV6_DSTOPTS,
1179 				    IPPROTO_IPV6);
1180 				if (*mp)
1181 					mp = &(*mp)->m_next;
1182 				break;
1183 
1184 			case IPPROTO_ROUTING:
1185 				if (!(inp->inp_flags & IN6P_RTHDR))
1186 					break;
1187 
1188 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1189 				    IPV6_RTHDR,
1190 				    IPPROTO_IPV6);
1191 				if (*mp)
1192 					mp = &(*mp)->m_next;
1193 				break;
1194 
1195 			case IPPROTO_HOPOPTS:
1196 			case IPPROTO_AH: /* is it possible? */
1197 				break;
1198 
1199 			default:
1200 				/*
1201 				 * other cases have been filtered in the above.
1202 				 * none will visit this case.  here we supply
1203 				 * the code just in case (nxt overwritten or
1204 				 * other cases).
1205 				 */
1206 				m_freem(ext);
1207 				goto loopend;
1208 
1209 			}
1210 
1211 			/* proceed with the next header. */
1212 			off += elen;
1213 			nxt = ip6e->ip6e_nxt;
1214 			ip6e = NULL;
1215 			m_freem(ext);
1216 			ext = NULL;
1217 		}
1218 loopend:
1219 		;
1220 	}
1221 }
1222 
1223 /*
1224  * pull single extension header from mbuf chain.  returns single mbuf that
1225  * contains the result, or NULL on error.
1226  */
1227 struct mbuf *
ip6_pullexthdr(struct mbuf * m,size_t off,int nxt)1228 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1229 {
1230 	struct ip6_ext ip6e;
1231 	size_t elen;
1232 	struct mbuf *n;
1233 
1234 #ifdef DIAGNOSTIC
1235 	switch (nxt) {
1236 	case IPPROTO_DSTOPTS:
1237 	case IPPROTO_ROUTING:
1238 	case IPPROTO_HOPOPTS:
1239 	case IPPROTO_AH: /* is it possible? */
1240 		break;
1241 	default:
1242 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1243 	}
1244 #endif
1245 
1246 	if (off + sizeof(ip6e) > m->m_pkthdr.len)
1247 		return NULL;
1248 
1249 	m_copydata(m, off, sizeof(ip6e), &ip6e);
1250 	if (nxt == IPPROTO_AH)
1251 		elen = (ip6e.ip6e_len + 2) << 2;
1252 	else
1253 		elen = (ip6e.ip6e_len + 1) << 3;
1254 
1255 	if (off + elen > m->m_pkthdr.len)
1256 		return NULL;
1257 
1258 	MGET(n, M_DONTWAIT, MT_DATA);
1259 	if (n && elen >= MLEN) {
1260 		MCLGET(n, M_DONTWAIT);
1261 		if ((n->m_flags & M_EXT) == 0) {
1262 			m_free(n);
1263 			n = NULL;
1264 		}
1265 	}
1266 	if (n == NULL) {
1267 		ip6stat_inc(ip6s_idropped);
1268 		return NULL;
1269 	}
1270 
1271 	n->m_len = 0;
1272 	if (elen >= m_trailingspace(n)) {
1273 		m_free(n);
1274 		return NULL;
1275 	}
1276 
1277 	m_copydata(m, off, elen, mtod(n, caddr_t));
1278 	n->m_len = elen;
1279 	return n;
1280 }
1281 
1282 /*
1283  * Get offset to the previous header followed by the header
1284  * currently processed.
1285  */
1286 int
ip6_get_prevhdr(struct mbuf * m,int off)1287 ip6_get_prevhdr(struct mbuf *m, int off)
1288 {
1289 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1290 
1291 	if (off == sizeof(struct ip6_hdr)) {
1292 		return offsetof(struct ip6_hdr, ip6_nxt);
1293 	} else if (off < sizeof(struct ip6_hdr)) {
1294 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1295 	} else {
1296 		int len, nlen, nxt;
1297 		struct ip6_ext ip6e;
1298 
1299 		nxt = ip6->ip6_nxt;
1300 		len = sizeof(struct ip6_hdr);
1301 		nlen = 0;
1302 		while (len < off) {
1303 			m_copydata(m, len, sizeof(ip6e), &ip6e);
1304 
1305 			switch (nxt) {
1306 			case IPPROTO_FRAGMENT:
1307 				nlen = sizeof(struct ip6_frag);
1308 				break;
1309 			case IPPROTO_AH:
1310 				nlen = (ip6e.ip6e_len + 2) << 2;
1311 				break;
1312 			default:
1313 				nlen = (ip6e.ip6e_len + 1) << 3;
1314 				break;
1315 			}
1316 			len += nlen;
1317 			nxt = ip6e.ip6e_nxt;
1318 		}
1319 
1320 		return (len - nlen);
1321 	}
1322 }
1323 
1324 /*
1325  * get next header offset.  m will be retained.
1326  */
1327 int
ip6_nexthdr(struct mbuf * m,int off,int proto,int * nxtp)1328 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1329 {
1330 	struct ip6_hdr ip6;
1331 	struct ip6_ext ip6e;
1332 	struct ip6_frag fh;
1333 
1334 	/* just in case */
1335 	if (m == NULL)
1336 		panic("%s: m == NULL", __func__);
1337 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1338 		return -1;
1339 
1340 	switch (proto) {
1341 	case IPPROTO_IPV6:
1342 		if (m->m_pkthdr.len < off + sizeof(ip6))
1343 			return -1;
1344 		m_copydata(m, off, sizeof(ip6), &ip6);
1345 		if (nxtp)
1346 			*nxtp = ip6.ip6_nxt;
1347 		off += sizeof(ip6);
1348 		return off;
1349 
1350 	case IPPROTO_FRAGMENT:
1351 		/*
1352 		 * terminate parsing if it is not the first fragment,
1353 		 * it does not make sense to parse through it.
1354 		 */
1355 		if (m->m_pkthdr.len < off + sizeof(fh))
1356 			return -1;
1357 		m_copydata(m, off, sizeof(fh), &fh);
1358 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1359 			return -1;
1360 		if (nxtp)
1361 			*nxtp = fh.ip6f_nxt;
1362 		off += sizeof(struct ip6_frag);
1363 		return off;
1364 
1365 	case IPPROTO_AH:
1366 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1367 			return -1;
1368 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1369 		if (nxtp)
1370 			*nxtp = ip6e.ip6e_nxt;
1371 		off += (ip6e.ip6e_len + 2) << 2;
1372 		if (m->m_pkthdr.len < off)
1373 			return -1;
1374 		return off;
1375 
1376 	case IPPROTO_HOPOPTS:
1377 	case IPPROTO_ROUTING:
1378 	case IPPROTO_DSTOPTS:
1379 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1380 			return -1;
1381 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1382 		if (nxtp)
1383 			*nxtp = ip6e.ip6e_nxt;
1384 		off += (ip6e.ip6e_len + 1) << 3;
1385 		if (m->m_pkthdr.len < off)
1386 			return -1;
1387 		return off;
1388 
1389 	case IPPROTO_NONE:
1390 	case IPPROTO_ESP:
1391 	case IPPROTO_IPCOMP:
1392 		/* give up */
1393 		return -1;
1394 
1395 	default:
1396 		return -1;
1397 	}
1398 
1399 	return -1;
1400 }
1401 
1402 /*
1403  * get offset for the last header in the chain.  m will be kept untainted.
1404  */
1405 int
ip6_lasthdr(struct mbuf * m,int off,int proto,int * nxtp)1406 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1407 {
1408 	int newoff;
1409 	int nxt;
1410 
1411 	if (!nxtp) {
1412 		nxt = -1;
1413 		nxtp = &nxt;
1414 	}
1415 	while (1) {
1416 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1417 		if (newoff < 0)
1418 			return off;
1419 		else if (newoff < off)
1420 			return -1;	/* invalid */
1421 		else if (newoff == off)
1422 			return newoff;
1423 
1424 		off = newoff;
1425 		proto = *nxtp;
1426 	}
1427 }
1428 
1429 /*
1430  * System control for IP6
1431  */
1432 
1433 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1434 	0,		0,		0,		0,
1435 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1436 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1437 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1438 	0,		0,		0,		0,
1439 	ENOPROTOOPT
1440 };
1441 
1442 #ifdef MROUTING
1443 extern int ip6_mrtproto;
1444 #endif
1445 
1446 const struct sysctl_bounded_args ipv6ctl_vars_unlocked[] = {
1447 	{ IPV6CTL_FORWARDING, &ip6_forwarding, 0, 2 },
1448 	{ IPV6CTL_SENDREDIRECTS, &ip6_sendredirects, 0, 1 },
1449 };
1450 
1451 const struct sysctl_bounded_args ipv6ctl_vars[] = {
1452 	{ IPV6CTL_DAD_PENDING, &ip6_dad_pending, SYSCTL_INT_READONLY },
1453 #ifdef MROUTING
1454 	{ IPV6CTL_MRTPROTO, &ip6_mrtproto, SYSCTL_INT_READONLY },
1455 #endif
1456 	{ IPV6CTL_DEFHLIM, &ip6_defhlim, 0, 255 },
1457 	{ IPV6CTL_MAXFRAGPACKETS, &ip6_maxfragpackets, 0, 1000 },
1458 	{ IPV6CTL_LOG_INTERVAL, &ip6_log_interval, 0, INT_MAX },
1459 	{ IPV6CTL_HDRNESTLIMIT, &ip6_hdrnestlimit, 0, 100 },
1460 	{ IPV6CTL_DAD_COUNT, &ip6_dad_count, 0, 10 },
1461 	{ IPV6CTL_AUTO_FLOWLABEL, &ip6_auto_flowlabel, 0, 1 },
1462 	{ IPV6CTL_DEFMCASTHLIM, &ip6_defmcasthlim, 0, 255 },
1463 	{ IPV6CTL_USE_DEPRECATED, &ip6_use_deprecated, 0, 1 },
1464 	{ IPV6CTL_MAXFRAGS, &ip6_maxfrags, 0, 1000 },
1465 	{ IPV6CTL_MFORWARDING, &ip6_mforwarding, 0, 1 },
1466 	{ IPV6CTL_MCAST_PMTU, &ip6_mcast_pmtu, 0, 1 },
1467 	{ IPV6CTL_NEIGHBORGCTHRESH, &ip6_neighborgcthresh, -1, 5 * 2048 },
1468 	{ IPV6CTL_MAXDYNROUTES, &ip6_maxdynroutes, -1, 5 * 4096 },
1469 };
1470 
1471 int
ip6_sysctl_ip6stat(void * oldp,size_t * oldlenp,void * newp)1472 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1473 {
1474 	struct ip6stat *ip6stat;
1475 	int ret;
1476 
1477 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1478 
1479 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1480 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters, NULL);
1481 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1482 	    ip6stat, sizeof(*ip6stat));
1483 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1484 
1485 	return (ret);
1486 }
1487 
1488 int
ip6_sysctl_soiikey(void * oldp,size_t * oldlenp,void * newp,size_t newlen)1489 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1490 {
1491 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1492 	int error;
1493 
1494 	error = suser(curproc);
1495 	if (error != 0)
1496 		return (error);
1497 
1498 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1499 
1500 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1501 	    sizeof(ip6_soiikey));
1502 
1503 	return (error);
1504 }
1505 
1506 int
ip6_sysctl(int * name,u_int namelen,void * oldp,size_t * oldlenp,void * newp,size_t newlen)1507 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1508     void *newp, size_t newlen)
1509 {
1510 #ifdef MROUTING
1511 	extern struct mrt6stat mrt6stat;
1512 #endif
1513 	int oldval, error;
1514 
1515 	/* Almost all sysctl names at this level are terminal. */
1516 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1517 		return (ENOTDIR);
1518 
1519 	switch (name[0]) {
1520 	case IPV6CTL_STATS:
1521 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1522 #ifdef MROUTING
1523 	case IPV6CTL_MRTSTATS:
1524 		if (newp != NULL)
1525 			return (EPERM);
1526 		NET_LOCK();
1527 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1528 		    &mrt6stat, sizeof(mrt6stat));
1529 		NET_UNLOCK();
1530 		return (error);
1531 	case IPV6CTL_MRTMIF:
1532 		if (newp)
1533 			return (EPERM);
1534 		NET_LOCK();
1535 		error = mrt6_sysctl_mif(oldp, oldlenp);
1536 		NET_UNLOCK();
1537 		return (error);
1538 	case IPV6CTL_MRTMFC:
1539 		if (newp)
1540 			return (EPERM);
1541 		NET_LOCK();
1542 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1543 		NET_UNLOCK();
1544 		return (error);
1545 #else
1546 	case IPV6CTL_MRTSTATS:
1547 	case IPV6CTL_MRTPROTO:
1548 	case IPV6CTL_MRTMIF:
1549 	case IPV6CTL_MRTMFC:
1550 		return (EOPNOTSUPP);
1551 #endif
1552 	case IPV6CTL_MTUDISCTIMEOUT:
1553 		NET_LOCK();
1554 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1555 		    &ip6_mtudisc_timeout, 0, INT_MAX);
1556 		rt_timer_queue_change(&icmp6_mtudisc_timeout_q,
1557 		    ip6_mtudisc_timeout);
1558 		NET_UNLOCK();
1559 		return (error);
1560 	case IPV6CTL_IFQUEUE:
1561 		return (sysctl_niq(name + 1, namelen - 1,
1562 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1563 	case IPV6CTL_SOIIKEY:
1564 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1565 	case IPV6CTL_MULTIPATH:
1566 		NET_LOCK();
1567 		oldval = ip6_multipath;
1568 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1569 		    &ip6_multipath, 0, 1);
1570 		if (oldval != ip6_multipath)
1571 			atomic_inc_long(&rtgeneration);
1572 		NET_UNLOCK();
1573 		return (error);
1574 	case IPV6CTL_FORWARDING:
1575 	case IPV6CTL_SENDREDIRECTS:
1576 		return (sysctl_bounded_arr(
1577 		    ipv6ctl_vars_unlocked, nitems(ipv6ctl_vars_unlocked),
1578 		    name, namelen, oldp, oldlenp, newp, newlen));
1579 	default:
1580 		NET_LOCK();
1581 		error = sysctl_bounded_arr(ipv6ctl_vars, nitems(ipv6ctl_vars),
1582 		    name, namelen, oldp, oldlenp, newp, newlen);
1583 		NET_UNLOCK();
1584 		return (error);
1585 	}
1586 	/* NOTREACHED */
1587 }
1588 
1589 void
ip6_send_dispatch(void * xmq)1590 ip6_send_dispatch(void *xmq)
1591 {
1592 	struct mbuf_queue *mq = xmq;
1593 	struct mbuf *m;
1594 	struct mbuf_list ml;
1595 
1596 	mq_delist(mq, &ml);
1597 	if (ml_empty(&ml))
1598 		return;
1599 
1600 	NET_LOCK_SHARED();
1601 	while ((m = ml_dequeue(&ml)) != NULL) {
1602 		ip6_output(m, NULL, NULL, 0, NULL, NULL);
1603 	}
1604 	NET_UNLOCK_SHARED();
1605 }
1606 
1607 void
ip6_send(struct mbuf * m)1608 ip6_send(struct mbuf *m)
1609 {
1610 	mq_enqueue(&ip6send_mq, m);
1611 	task_add(net_tq(0), &ip6send_task);
1612 }
1613