xref: /dragonfly/sys/netinet6/ip6_input.c (revision de78d61c)
1 /*	$FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.15 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei 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 "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/domain.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/errno.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/proc.h>
81 #include <sys/priv.h>
82 
83 #include <net/if.h>
84 #include <net/if_types.h>
85 #include <net/if_dl.h>
86 #include <net/route.h>
87 #include <net/netisr.h>
88 #include <net/pfil.h>
89 
90 #include <sys/msgport2.h>
91 #include <net/netmsg2.h>
92 #include <net/netisr2.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_systm.h>
96 #ifdef INET
97 #include <netinet/ip.h>
98 #include <netinet/ip_icmp.h>
99 #endif /* INET */
100 #include <netinet/ip6.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/icmp6.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/nd6.h>
108 
109 #include <net/ip6fw/ip6_fw.h>
110 
111 #include <netinet6/ip6protosw.h>
112 
113 #include <net/net_osdep.h>
114 
115 extern struct domain inet6domain;
116 extern struct protosw inet6sw[];
117 
118 u_char ip6_protox[IPPROTO_MAX];
119 struct in6_ifaddr *in6_ifaddr;
120 
121 int ip6_forward_srcrt;			/* XXX */
122 int ip6_sourcecheck;			/* XXX */
123 int ip6_sourcecheck_interval;		/* XXX */
124 
125 int ip6_ours_check_algorithm;
126 
127 struct pfil_head inet6_pfil_hook;
128 
129 /* firewall hooks */
130 ip6_fw_chk_t *ip6_fw_chk_ptr;
131 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
132 int ip6_fw_enable = 1;
133 
134 struct ip6stat ip6stat;
135 
136 static void ip6_init2 (void *);
137 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
138 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
139 static void ip6_input(netmsg_t msg);
140 #ifdef PULLDOWN_TEST
141 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
142 #endif
143 static void transport6_processing_handler(netmsg_t netmsg);
144 
145 /*
146  * IP6 initialization: fill in IP6 protocol switch table.
147  * All protocols not implemented in kernel go to raw IP6 protocol handler.
148  */
149 void
150 ip6_init(void)
151 {
152 	struct protosw *pr;
153 	int i;
154 	struct timeval tv;
155 
156 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
157 	if (pr == NULL)
158 		panic("ip6_init");
159 	for (i = 0; i < IPPROTO_MAX; i++)
160 		ip6_protox[i] = pr - inet6sw;
161 	for (pr = inet6domain.dom_protosw;
162 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
163 		if (pr->pr_domain->dom_family == PF_INET6 &&
164 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
165 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
166 
167 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
168 	inet6_pfil_hook.ph_af = AF_INET6;
169 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
170 		kprintf("%s: WARNING: unable to register pfil hook, "
171 			"error %d\n", __func__, i);
172 	}
173 
174 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX hashfn */
175 	scope6_init();
176 	addrsel_policy_init();
177 	nd6_init();
178 	frag6_init();
179 	/*
180 	 * in many cases, random() here does NOT return random number
181 	 * as initialization during bootstrap time occur in fixed order.
182 	 */
183 	microtime(&tv);
184 	ip6_flow_seq = krandom() ^ tv.tv_usec;
185 	microtime(&tv);
186 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
187 }
188 
189 static void
190 ip6_init2(void *dummy)
191 {
192 	nd6_timer_init();
193 	in6_tmpaddrtimer_init();
194 }
195 
196 /* cheat */
197 /* This must be after route_init(), which is now SI_ORDER_THIRD */
198 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
199 
200 extern struct	route_in6 ip6_forward_rt;
201 
202 static void
203 ip6_input(netmsg_t msg)
204 {
205 	struct mbuf *m = msg->packet.nm_packet;
206 	struct ip6_hdr *ip6;
207 	int off = sizeof(struct ip6_hdr), nest;
208 	u_int32_t plen;
209 	u_int32_t rtalert = ~0;
210 	int nxt, ours = 0, rh_present = 0;
211 	struct ifnet *deliverifp = NULL;
212 	struct in6_addr odst;
213 	int srcrt = 0;
214 
215 	/*
216 	 * make sure we don't have onion peering information into m_aux.
217 	 */
218 	ip6_delaux(m);
219 
220 	/*
221 	 * mbuf statistics
222 	 */
223 	if (m->m_flags & M_EXT) {
224 		if (m->m_next)
225 			ip6stat.ip6s_mext2m++;
226 		else
227 			ip6stat.ip6s_mext1++;
228 	} else {
229 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
230 		if (m->m_next) {
231 			if (m->m_flags & M_LOOP) {
232 				ip6stat.ip6s_m2m[loif->if_index]++; /* XXX */
233 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
234 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
235 			else
236 				ip6stat.ip6s_m2m[0]++;
237 		} else
238 			ip6stat.ip6s_m1++;
239 #undef M2MMAX
240 	}
241 
242 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
243 	ip6stat.ip6s_total++;
244 
245 #ifndef PULLDOWN_TEST
246 	/*
247 	 * L2 bridge code and some other code can return mbuf chain
248 	 * that does not conform to KAME requirement.  too bad.
249 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
250 	 */
251 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
252 		struct mbuf *n;
253 
254 		n = m_getb(m->m_pkthdr.len, M_NOWAIT, MT_HEADER, M_PKTHDR);
255 		if (n == NULL)
256 			goto bad;
257 		M_MOVE_PKTHDR(n, m);
258 
259 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
260 		n->m_len = n->m_pkthdr.len;
261 		m_freem(m);
262 		m = n;
263 	}
264 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
265 #endif
266 
267 	if (m->m_len < sizeof(struct ip6_hdr)) {
268 		struct ifnet *inifp;
269 		inifp = m->m_pkthdr.rcvif;
270 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
271 			ip6stat.ip6s_toosmall++;
272 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
273 			goto bad2;
274 		}
275 	}
276 
277 	ip6 = mtod(m, struct ip6_hdr *);
278 
279 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
280 		ip6stat.ip6s_badvers++;
281 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
282 		goto bad;
283 	}
284 
285 	/*
286 	 * Run through list of hooks for input packets.
287 	 *
288 	 * NB: Beware of the destination address changing
289 	 *     (e.g. by NAT rewriting). When this happens,
290 	 *     tell ip6_forward to do the right thing.
291 	 */
292 	if (pfil_has_hooks(&inet6_pfil_hook)) {
293 		odst = ip6->ip6_dst;
294 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
295 		    m->m_pkthdr.rcvif, PFIL_IN)) {
296 			goto bad2;
297 		}
298 		if (m == NULL)			/* consumed by filter */
299 			goto bad2;
300 		ip6 = mtod(m, struct ip6_hdr *);
301 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
302 	}
303 
304 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
305 
306 #ifdef ALTQ
307 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
308 		/* packet is dropped by traffic conditioner */
309 		return;
310 	}
311 #endif
312 
313 	/*
314 	 * Check with the firewall...
315 	 */
316 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
317 		u_short port = 0;
318 		/* If ipfw says divert, we have to just drop packet */
319 		/* use port as a dummy argument */
320 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
321 			m_freem(m);
322 			m = NULL;
323 		}
324 		if (!m)
325 			goto bad2;
326 	}
327 
328 	/*
329 	 * Check against address spoofing/corruption.
330 	 */
331 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
332 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
333 		/*
334 		 * XXX: "badscope" is not very suitable for a multicast source.
335 		 */
336 		ip6stat.ip6s_badscope++;
337 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
338 		goto bad;
339 	}
340 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
341 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
342 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
343 		ip6stat.ip6s_badscope++;
344 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
345 		goto bad;
346 	}
347 
348 	/*
349 	 * The following check is not documented in specs.  A malicious
350 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
351 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
352 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
353 	 *
354 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
355 	 * support IPv4-less kernel compilation, we cannot support SIIT
356 	 * environment at all.  So, it makes more sense for us to reject any
357 	 * malicious packets for non-SIIT environment, than try to do a
358 	 * partial support for SIIT environment.
359 	 */
360 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
361 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 #if 0
367 	/*
368 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
369 	 *
370 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
371 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
372 	 * is revised to forbid relaying case.
373 	 */
374 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
375 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
376 		ip6stat.ip6s_badscope++;
377 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
378 		goto bad;
379 	}
380 #endif
381 
382 	/* drop packets if interface ID portion is already filled */
383 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
384 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
385 		    ip6->ip6_src.s6_addr16[1]) {
386 			ip6stat.ip6s_badscope++;
387 			goto bad;
388 		}
389 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
390 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
391 		    ip6->ip6_dst.s6_addr16[1]) {
392 			ip6stat.ip6s_badscope++;
393 			goto bad;
394 		}
395 	}
396 
397 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
398 		ip6->ip6_src.s6_addr16[1]
399 			= htons(m->m_pkthdr.rcvif->if_index);
400 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
401 		ip6->ip6_dst.s6_addr16[1]
402 			= htons(m->m_pkthdr.rcvif->if_index);
403 
404 	/*
405 	 * Multicast check
406 	 *
407 	 * WARNING!  For general subnet proxying the interface hw will
408 	 *	     likely filter out the multicast solicitations, so
409 	 *	     the interface must be in promiscuous mode.
410 	 */
411 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
412 		ours = 1;
413 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
414 		deliverifp = m->m_pkthdr.rcvif;
415 		goto hbhcheck;
416 	}
417 
418 	/*
419 	 *  Unicast check
420 	 */
421 	if (ip6_forward_rt.ro_rt != NULL &&
422 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
423 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
424 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr)) {
425 		ip6stat.ip6s_forward_cachehit++;
426 	} else {
427 		struct sockaddr_in6 *dst6;
428 
429 		if (ip6_forward_rt.ro_rt) {
430 			/* route is down or destination is different */
431 			ip6stat.ip6s_forward_cachemiss++;
432 			RTFREE(ip6_forward_rt.ro_rt);
433 			ip6_forward_rt.ro_rt = 0;
434 		}
435 
436 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
437 		dst6 = &ip6_forward_rt.ro_dst;
438 		dst6->sin6_len = sizeof(struct sockaddr_in6);
439 		dst6->sin6_family = AF_INET6;
440 		dst6->sin6_addr = ip6->ip6_dst;
441 
442 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
443 	}
444 
445 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
446 
447 	/*
448 	 * Accept the packet if the forwarding interface to the destination
449 	 * according to the routing table is the loopback interface,
450 	 * unless the associated route has a gateway.
451 	 * Note that this approach causes to accept a packet if there is a
452 	 * route to the loopback interface for the destination of the packet.
453 	 * But we think it's even useful in some situations, e.g. when using
454 	 * a special daemon which wants to intercept the packet.
455 	 *
456 	 * XXX: some OSes automatically make a cloned route for the destination
457 	 * of an outgoing packet.  If the outgoing interface of the packet
458 	 * is a loopback one, the kernel would consider the packet to be
459 	 * accepted, even if we have no such address assinged on the interface.
460 	 * We check the cloned flag of the route entry to reject such cases,
461 	 * assuming that route entries for our own addresses are not made by
462 	 * cloning (it should be true because in6_addloop explicitly installs
463 	 * the host route).  However, we might have to do an explicit check
464 	 * while it would be less efficient.  Or, should we rather install a
465 	 * reject route for such a case?
466 	 */
467 	if (ip6_forward_rt.ro_rt &&
468 	    (ip6_forward_rt.ro_rt->rt_flags &
469 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
470 #ifdef RTF_WASCLONED
471 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
472 #endif
473 #ifdef RTF_CLONED
474 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
475 #endif
476 #if 0
477 	    /*
478 	     * The check below is redundant since the comparison of
479 	     * the destination and the key of the rtentry has
480 	     * already done through looking up the routing table.
481 	     */
482 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
483 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
484 #endif
485 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
486 		struct in6_ifaddr *ia6 =
487 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
488 
489 		/*
490 		 * record address information into m_aux.
491 		 */
492 		ip6_setdstifaddr(m, ia6);
493 
494 		/*
495 		 * packets to a tentative, duplicated, or somehow invalid
496 		 * address must not be accepted.
497 		 */
498 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
499 			/* this address is ready */
500 			ours = 1;
501 			deliverifp = ia6->ia_ifp;	/* correct? */
502 			/* Count the packet in the ip address stats */
503 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
504 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
505 			goto hbhcheck;
506 		} else {
507 			/* address is not ready, so discard the packet. */
508 			nd6log((LOG_INFO,
509 			    "ip6_input: packet to an unready address %s->%s\n",
510 			    ip6_sprintf(&ip6->ip6_src),
511 			    ip6_sprintf(&ip6->ip6_dst)));
512 
513 			goto bad;
514 		}
515 	}
516 
517 	/*
518 	 * Now there is no reason to process the packet if it's not our own
519 	 * and we're not a router.
520 	 */
521 	if (!ip6_forwarding) {
522 		ip6stat.ip6s_cantforward++;
523 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
524 		goto bad;
525 	}
526 
527 hbhcheck:
528 	/*
529 	 * record address information into m_aux, if we don't have one yet.
530 	 * note that we are unable to record it, if the address is not listed
531 	 * as our interface address (e.g. multicast addresses and such).
532 	 */
533 	if (deliverifp && !ip6_getdstifaddr(m)) {
534 		struct in6_ifaddr *ia6;
535 
536 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
537 		if (ia6) {
538 			if (!ip6_setdstifaddr(m, ia6)) {
539 				/*
540 				 * XXX maybe we should drop the packet here,
541 				 * as we could not provide enough information
542 				 * to the upper layers.
543 				 */
544 			}
545 		}
546 	}
547 
548 	/*
549 	 * Process Hop-by-Hop options header if it's contained.
550 	 * m may be modified in ip6_hopopts_input().
551 	 * If a JumboPayload option is included, plen will also be modified.
552 	 */
553 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
554 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
555 		struct ip6_hbh *hbh;
556 
557 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
558 #if 0	/*touches NULL pointer*/
559 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
560 #endif
561 			goto bad2;	/* m have already been freed */
562 		}
563 
564 		/* adjust pointer */
565 		ip6 = mtod(m, struct ip6_hdr *);
566 
567 		/*
568 		 * if the payload length field is 0 and the next header field
569 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
570 		 * option MUST be included.
571 		 */
572 		if (ip6->ip6_plen == 0 && plen == 0) {
573 			/*
574 			 * Note that if a valid jumbo payload option is
575 			 * contained, ip6_hoptops_input() must set a valid
576 			 * (non-zero) payload length to the variable plen.
577 			 */
578 			ip6stat.ip6s_badoptions++;
579 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
580 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
581 			icmp6_error(m, ICMP6_PARAM_PROB,
582 				    ICMP6_PARAMPROB_HEADER,
583 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
584 			goto bad2;
585 		}
586 #ifndef PULLDOWN_TEST
587 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
588 		hbh = (struct ip6_hbh *)(ip6 + 1);
589 #else
590 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
591 			sizeof(struct ip6_hbh));
592 		if (hbh == NULL) {
593 			ip6stat.ip6s_tooshort++;
594 			goto bad2;
595 		}
596 #endif
597 		nxt = hbh->ip6h_nxt;
598 
599 		/*
600 		 * If we are acting as a router and the packet contains a
601 		 * router alert option, see if we know the option value.
602 		 * Currently, we only support the option value for MLD, in which
603 		 * case we should pass the packet to the multicast routing
604 		 * daemon.
605 		 */
606 		if (rtalert != ~0 && ip6_forwarding) {
607 			switch (rtalert) {
608 			case IP6OPT_RTALERT_MLD:
609 				ours = 1;
610 				break;
611 			default:
612 				/*
613 				 * RFC2711 requires unrecognized values must be
614 				 * silently ignored.
615 				 */
616 				break;
617 			}
618 		}
619 	} else
620 		nxt = ip6->ip6_nxt;
621 
622 	/*
623 	 * Check that the amount of data in the buffers
624 	 * is as at least much as the IPv6 header would have us expect.
625 	 * Trim mbufs if longer than we expect.
626 	 * Drop packet if shorter than we expect.
627 	 */
628 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
629 		ip6stat.ip6s_tooshort++;
630 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
631 		goto bad;
632 	}
633 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
634 		if (m->m_len == m->m_pkthdr.len) {
635 			m->m_len = sizeof(struct ip6_hdr) + plen;
636 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
637 		} else
638 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
639 	}
640 
641 	/*
642 	 * Forward if desirable.
643 	 */
644 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
645 		/*
646 		 * If we are acting as a multicast router, all
647 		 * incoming multicast packets are passed to the
648 		 * kernel-level multicast forwarding function.
649 		 * The packet is returned (relatively) intact; if
650 		 * ip6_mforward() returns a non-zero value, the packet
651 		 * must be discarded, else it may be accepted below.
652 		 */
653 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
654 			ip6stat.ip6s_cantforward++;
655 			goto bad;
656 		}
657 		if (!ours)
658 			goto bad;
659 	} else if (!ours) {
660 		ip6_forward(m, srcrt);
661 		goto bad2;
662 	}
663 
664 	ip6 = mtod(m, struct ip6_hdr *);
665 
666 	/*
667 	 * Malicious party may be able to use IPv4 mapped addr to confuse
668 	 * tcp/udp stack and bypass security checks (act as if it was from
669 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
670 	 *
671 	 * For SIIT end node behavior, you may want to disable the check.
672 	 * However, you will  become vulnerable to attacks using IPv4 mapped
673 	 * source.
674 	 */
675 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
676 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
677 		ip6stat.ip6s_badscope++;
678 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
679 		goto bad;
680 	}
681 
682 	/*
683 	 * Tell launch routine the next header
684 	 */
685 	ip6stat.ip6s_delivered++;
686 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
687 	nest = 0;
688 
689 	rh_present = 0;
690 	while (nxt != IPPROTO_DONE) {
691 		struct protosw *sw6;
692 
693 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
694 			ip6stat.ip6s_toomanyhdr++;
695 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
696 			goto bad;
697 		}
698 
699 		/*
700 		 * protection against faulty packet - there should be
701 		 * more sanity checks in header chain processing.
702 		 */
703 		if (m->m_pkthdr.len < off) {
704 			ip6stat.ip6s_tooshort++;
705 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
706 			goto bad;
707 		}
708 
709 #if 0
710 		/*
711 		 * do we need to do it for every header?  yeah, other
712 		 * functions can play with it (like re-allocate and copy).
713 		 */
714 		mhist = ip6_addaux(m);
715 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
716 			hist = mtod(mhist, caddr_t) + mhist->m_len;
717 			bcopy(&nxt, hist, sizeof(nxt));
718 			mhist->m_len += sizeof(nxt);
719 		} else {
720 			ip6stat.ip6s_toomanyhdr++;
721 			goto bad;
722 		}
723 #endif
724 
725 		if (nxt == IPPROTO_ROUTING) {
726 			if (rh_present++) {
727 				in6_ifstat_inc(m->m_pkthdr.rcvif,
728 				    ifs6_in_hdrerr);
729 				ip6stat.ip6s_badoptions++;
730 				goto bad;
731 			}
732 		}
733 
734 		sw6 = &inet6sw[ip6_protox[nxt]];
735 		/*
736 		 * If this is a terminal header forward to the port, otherwise
737 		 * process synchronously for more headers.
738 		 */
739 		if (sw6->pr_flags & PR_LASTHDR) {
740 			struct netmsg_packet *pmsg;
741 			lwkt_port_t port;
742 
743 			port = netisr_cpuport(0); /* XXX */
744 			if ((m->m_flags & M_HASH) == 0)
745 				m_sethash(m, 0); /* XXX */
746 
747 			KKASSERT(port != NULL);
748 			pmsg = &m->m_hdr.mh_netmsg;
749 			netmsg_init(&pmsg->base, NULL,
750 				    &netisr_apanic_rport,
751 				    0, transport6_processing_handler);
752 			pmsg->nm_packet = m;
753 			pmsg->nm_nxt = nxt;
754 			pmsg->base.lmsg.u.ms_result = off;
755 			lwkt_sendmsg(port, &pmsg->base.lmsg);
756 			/* done with m */
757 			nxt = IPPROTO_DONE;
758 		} else {
759 			nxt = sw6->pr_input(&m, &off, nxt);
760 		}
761 	}
762 	goto bad2;
763 bad:
764 	m_freem(m);
765 bad2:
766 	;
767 	/* msg was embedded in the mbuf, do not reply! */
768 }
769 
770 /*
771  * We have to call the pr_input() function from the correct protocol
772  * thread.  The sw6->pr_soport() request at the end of ip6_input()
773  * returns the port and we forward a netmsg to the port to execute
774  * this function.
775  */
776 static void
777 transport6_processing_handler(netmsg_t netmsg)
778 {
779 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
780 	struct protosw *sw6;
781 	int hlen;
782 	int nxt;
783 
784 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
785 	hlen = pmsg->base.lmsg.u.ms_result;
786 
787 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
788 	KKASSERT(nxt == IPPROTO_DONE);
789 	/* netmsg was embedded in the mbuf, do not reply! */
790 }
791 
792 /*
793  * set/grab in6_ifaddr correspond to IPv6 destination address.
794  * XXX backward compatibility wrapper
795  */
796 static struct ip6aux *
797 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
798 {
799 	struct ip6aux *n;
800 
801 	n = ip6_addaux(m);
802 	if (n)
803 		n->ip6a_dstia6 = ia6;
804 	return n;	/* NULL if failed to set */
805 }
806 
807 struct in6_ifaddr *
808 ip6_getdstifaddr(struct mbuf *m)
809 {
810 	struct ip6aux *n;
811 
812 	n = ip6_findaux(m);
813 	if (n)
814 		return n->ip6a_dstia6;
815 	else
816 		return NULL;
817 }
818 
819 /*
820  * Hop-by-Hop options header processing. If a valid jumbo payload option is
821  * included, the real payload length will be stored in plenp.
822  */
823 static int
824 ip6_hopopts_input(u_int32_t *plenp,
825 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
826 		  struct mbuf **mp,
827 		  int *offp)
828 {
829 	struct mbuf *m = *mp;
830 	int off = *offp, hbhlen;
831 	struct ip6_hbh *hbh;
832 
833 	/* validation of the length of the header */
834 #ifndef PULLDOWN_TEST
835 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
836 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
837 	hbhlen = (hbh->ip6h_len + 1) << 3;
838 
839 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
840 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
841 #else
842 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
843 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
844 	if (hbh == NULL) {
845 		ip6stat.ip6s_tooshort++;
846 		return -1;
847 	}
848 	hbhlen = (hbh->ip6h_len + 1) << 3;
849 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
850 		hbhlen);
851 	if (hbh == NULL) {
852 		ip6stat.ip6s_tooshort++;
853 		return -1;
854 	}
855 #endif
856 	off += hbhlen;
857 	hbhlen -= sizeof(struct ip6_hbh);
858 
859 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
860 				hbhlen, rtalertp, plenp) < 0)
861 		return (-1);
862 
863 	*offp = off;
864 	*mp = m;
865 	return (0);
866 }
867 
868 /*
869  * Search header for all Hop-by-hop options and process each option.
870  * This function is separate from ip6_hopopts_input() in order to
871  * handle a case where the sending node itself process its hop-by-hop
872  * options header. In such a case, the function is called from ip6_output().
873  *
874  * The function assumes that hbh header is located right after the IPv6 header
875  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
876  * opthead + hbhlen is located in continuous memory region.
877  */
878 int
879 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
880 		    u_int32_t *rtalertp, u_int32_t *plenp)
881 {
882 	struct ip6_hdr *ip6;
883 	int optlen = 0;
884 	u_int8_t *opt = opthead;
885 	u_int16_t rtalert_val;
886 	u_int32_t jumboplen;
887 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
888 
889 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
890 		switch (*opt) {
891 		case IP6OPT_PAD1:
892 			optlen = 1;
893 			break;
894 		case IP6OPT_PADN:
895 			if (hbhlen < IP6OPT_MINLEN) {
896 				ip6stat.ip6s_toosmall++;
897 				goto bad;
898 			}
899 			optlen = *(opt + 1) + 2;
900 			break;
901 		case IP6OPT_RTALERT:
902 			/* XXX may need check for alignment */
903 			if (hbhlen < IP6OPT_RTALERT_LEN) {
904 				ip6stat.ip6s_toosmall++;
905 				goto bad;
906 			}
907 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
908 				/* XXX stat */
909 				icmp6_error(m, ICMP6_PARAM_PROB,
910 				    ICMP6_PARAMPROB_HEADER,
911 				    erroff + opt + 1 - opthead);
912 				return (-1);
913 			}
914 			optlen = IP6OPT_RTALERT_LEN;
915 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
916 			*rtalertp = ntohs(rtalert_val);
917 			break;
918 		case IP6OPT_JUMBO:
919 			/* XXX may need check for alignment */
920 			if (hbhlen < IP6OPT_JUMBO_LEN) {
921 				ip6stat.ip6s_toosmall++;
922 				goto bad;
923 			}
924 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
925 				/* XXX stat */
926 				icmp6_error(m, ICMP6_PARAM_PROB,
927 				    ICMP6_PARAMPROB_HEADER,
928 				    erroff + opt + 1 - opthead);
929 				return (-1);
930 			}
931 			optlen = IP6OPT_JUMBO_LEN;
932 
933 			/*
934 			 * IPv6 packets that have non 0 payload length
935 			 * must not contain a jumbo payload option.
936 			 */
937 			ip6 = mtod(m, struct ip6_hdr *);
938 			if (ip6->ip6_plen) {
939 				ip6stat.ip6s_badoptions++;
940 				icmp6_error(m, ICMP6_PARAM_PROB,
941 				    ICMP6_PARAMPROB_HEADER,
942 				    erroff + opt - opthead);
943 				return (-1);
944 			}
945 
946 			/*
947 			 * We may see jumbolen in unaligned location, so
948 			 * we'd need to perform bcopy().
949 			 */
950 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
951 			jumboplen = (u_int32_t)htonl(jumboplen);
952 
953 #if 1
954 			/*
955 			 * if there are multiple jumbo payload options,
956 			 * *plenp will be non-zero and the packet will be
957 			 * rejected.
958 			 * the behavior may need some debate in ipngwg -
959 			 * multiple options does not make sense, however,
960 			 * there's no explicit mention in specification.
961 			 */
962 			if (*plenp != 0) {
963 				ip6stat.ip6s_badoptions++;
964 				icmp6_error(m, ICMP6_PARAM_PROB,
965 				    ICMP6_PARAMPROB_HEADER,
966 				    erroff + opt + 2 - opthead);
967 				return (-1);
968 			}
969 #endif
970 
971 			/*
972 			 * jumbo payload length must be larger than 65535.
973 			 */
974 			if (jumboplen <= IPV6_MAXPACKET) {
975 				ip6stat.ip6s_badoptions++;
976 				icmp6_error(m, ICMP6_PARAM_PROB,
977 				    ICMP6_PARAMPROB_HEADER,
978 				    erroff + opt + 2 - opthead);
979 				return (-1);
980 			}
981 			*plenp = jumboplen;
982 
983 			break;
984 		default:		/* unknown option */
985 			if (hbhlen < IP6OPT_MINLEN) {
986 				ip6stat.ip6s_toosmall++;
987 				goto bad;
988 			}
989 			optlen = ip6_unknown_opt(opt, m,
990 			    erroff + opt - opthead);
991 			if (optlen == -1)
992 				return (-1);
993 			optlen += 2;
994 			break;
995 		}
996 	}
997 
998 	return (0);
999 
1000 bad:
1001 	m_freem(m);
1002 	return (-1);
1003 }
1004 
1005 /*
1006  * Unknown option processing.
1007  * The third argument `off' is the offset from the IPv6 header to the option,
1008  * which is necessary if the IPv6 header the and option header and IPv6 header
1009  * is not continuous in order to return an ICMPv6 error.
1010  */
1011 int
1012 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1013 {
1014 	struct ip6_hdr *ip6;
1015 
1016 	switch (IP6OPT_TYPE(*optp)) {
1017 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1018 		return ((int)*(optp + 1));
1019 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1020 		m_freem(m);
1021 		return (-1);
1022 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1023 		ip6stat.ip6s_badoptions++;
1024 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1025 		return (-1);
1026 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1027 		ip6stat.ip6s_badoptions++;
1028 		ip6 = mtod(m, struct ip6_hdr *);
1029 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1030 		    (m->m_flags & (M_BCAST|M_MCAST)))
1031 			m_freem(m);
1032 		else
1033 			icmp6_error(m, ICMP6_PARAM_PROB,
1034 				    ICMP6_PARAMPROB_OPTION, off);
1035 		return (-1);
1036 	}
1037 
1038 	m_freem(m);		/* XXX: NOTREACHED */
1039 	return (-1);
1040 }
1041 
1042 /*
1043  * Create the "control" list for this pcb.
1044  * The function will not modify mbuf chain at all.
1045  *
1046  * with KAME mbuf chain restriction:
1047  * The routine will be called from upper layer handlers like tcp6_input().
1048  * Thus the routine assumes that the caller (tcp6_input) have already
1049  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1050  * very first mbuf on the mbuf chain.
1051  */
1052 void
1053 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1054 		struct mbuf *m)
1055 {
1056 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1057 	struct thread *td = curthread;	/* XXX */
1058 	int privileged = 0;
1059 
1060 
1061 	if (priv_check(td, PRIV_ROOT) == 0)
1062 		privileged++;
1063 
1064 #ifdef SO_TIMESTAMP
1065 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1066 		struct timeval tv;
1067 
1068 		microtime(&tv);
1069 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1070 		    SCM_TIMESTAMP, SOL_SOCKET);
1071 		if (*mp) {
1072 			mp = &(*mp)->m_next;
1073 		}
1074 	}
1075 #endif
1076 
1077 	/* RFC 2292 sec. 5 */
1078 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1079 		struct in6_pktinfo pi6;
1080 
1081 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1082 		in6_clearscope(&pi6.ipi6_addr); /* XXX */
1083 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif) ?
1084 		    m->m_pkthdr.rcvif->if_index : 0;
1085 		*mp = sbcreatecontrol((caddr_t)&pi6, sizeof(struct in6_pktinfo),
1086 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1087 		if (*mp)
1088 			mp = &(*mp)->m_next;
1089 	}
1090 
1091 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1092 		int hlim = ip6->ip6_hlim & 0xff;
1093 
1094 		*mp = sbcreatecontrol((caddr_t)&hlim, sizeof(int),
1095 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1096 		if (*mp)
1097 			mp = &(*mp)->m_next;
1098 	}
1099 
1100 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1101 		u_int32_t flowinfo;
1102 		int tclass;
1103 
1104 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1105 		flowinfo >>= 20;
1106 
1107 		tclass = flowinfo & 0xff;
1108 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1109 		    IPV6_TCLASS, IPPROTO_IPV6);
1110 		if (*mp)
1111 			mp = &(*mp)->m_next;
1112 	}
1113 
1114 	/*
1115 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1116 	 * for the option, but it might be too strict, since there might
1117 	 * be some hop-by-hop options which can be returned to normal user.
1118 	 * See RFC 2292 section 6.
1119 	 */
1120 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1121 		/*
1122 		 * Check if a hop-by-hop options header is contatined in the
1123 		 * received packet, and if so, store the options as ancillary
1124 		 * data. Note that a hop-by-hop options header must be
1125 		 * just after the IPv6 header, which is assured through the
1126 		 * IPv6 input processing.
1127 		 */
1128 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1129 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1130 			struct ip6_hbh *hbh;
1131 			int hbhlen = 0;
1132 #ifdef PULLDOWN_TEST
1133 			struct mbuf *ext;
1134 #endif
1135 
1136 #ifndef PULLDOWN_TEST
1137 			hbh = (struct ip6_hbh *)(ip6 + 1);
1138 			hbhlen = (hbh->ip6h_len + 1) << 3;
1139 #else
1140 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1141 			    ip6->ip6_nxt);
1142 			if (ext == NULL) {
1143 				ip6stat.ip6s_tooshort++;
1144 				return;
1145 			}
1146 			hbh = mtod(ext, struct ip6_hbh *);
1147 			hbhlen = (hbh->ip6h_len + 1) << 3;
1148 			if (hbhlen != ext->m_len) {
1149 				m_freem(ext);
1150 				ip6stat.ip6s_tooshort++;
1151 				return;
1152 			}
1153 #endif
1154 
1155 			/*
1156 			 * XXX: We copy the whole header even if a
1157 			 * jumbo payload option is included, the option which
1158 			 * is to be removed before returning according to
1159 			 * RFC 2292.
1160 			 * Note: this constraint is removed in 2292bis.
1161 			 */
1162 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1163 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1164 			    IPPROTO_IPV6);
1165 			if (*mp)
1166 				mp = &(*mp)->m_next;
1167 #ifdef PULLDOWN_TEST
1168 			m_freem(ext);
1169 #endif
1170 		}
1171 	}
1172 
1173 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1174 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1175 		int proto, off, nxt;
1176 
1177 		/*
1178 		 * go through the header chain to see if a routing header is
1179 		 * contained in the packet. We need this information to store
1180 		 * destination options headers (if any) properly.
1181 		 * XXX: performance issue. We should record this info when
1182 		 * processing extension headers in incoming routine.
1183 		 * (todo) use m_aux?
1184 		 */
1185 		proto = IPPROTO_IPV6;
1186 		off = 0;
1187 		nxt = -1;
1188 		while (1) {
1189 			int newoff;
1190 
1191 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1192 			if (newoff < 0)
1193 				break;
1194 			if (newoff < off) /* invalid, check for safety */
1195 				break;
1196 			if ((proto = nxt) == IPPROTO_ROUTING)
1197 				break;
1198 			off = newoff;
1199 		}
1200 	}
1201 
1202 	if ((in6p->in6p_flags &
1203 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1204 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1205 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1206 
1207 		/*
1208 		 * Search for destination options headers or routing
1209 		 * header(s) through the header chain, and stores each
1210 		 * header as ancillary data.
1211 		 * Note that the order of the headers remains in
1212 		 * the chain of ancillary data.
1213 		 */
1214 		while (1) {	/* is explicit loop prevention necessary? */
1215 			struct ip6_ext *ip6e = NULL;
1216 			int elen;
1217 #ifdef PULLDOWN_TEST
1218 			struct mbuf *ext = NULL;
1219 #endif
1220 
1221 			/*
1222 			 * if it is not an extension header, don't try to
1223 			 * pull it from the chain.
1224 			 */
1225 			switch (nxt) {
1226 			case IPPROTO_DSTOPTS:
1227 			case IPPROTO_ROUTING:
1228 			case IPPROTO_HOPOPTS:
1229 			case IPPROTO_AH: /* is it possible? */
1230 				break;
1231 			default:
1232 				goto loopend;
1233 			}
1234 
1235 #ifndef PULLDOWN_TEST
1236 			if (off + sizeof(*ip6e) > m->m_len)
1237 				goto loopend;
1238 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1239 			if (nxt == IPPROTO_AH)
1240 				elen = (ip6e->ip6e_len + 2) << 2;
1241 			else
1242 				elen = (ip6e->ip6e_len + 1) << 3;
1243 			if (off + elen > m->m_len)
1244 				goto loopend;
1245 #else
1246 			ext = ip6_pullexthdr(m, off, nxt);
1247 			if (ext == NULL) {
1248 				ip6stat.ip6s_tooshort++;
1249 				return;
1250 			}
1251 			ip6e = mtod(ext, struct ip6_ext *);
1252 			if (nxt == IPPROTO_AH)
1253 				elen = (ip6e->ip6e_len + 2) << 2;
1254 			else
1255 				elen = (ip6e->ip6e_len + 1) << 3;
1256 			if (elen != ext->m_len) {
1257 				m_freem(ext);
1258 				ip6stat.ip6s_tooshort++;
1259 				return;
1260 			}
1261 #endif
1262 
1263 			switch (nxt) {
1264 			case IPPROTO_DSTOPTS:
1265 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1266 					break;
1267 
1268 				/*
1269 				 * We also require super-user privilege for
1270 				 * the option.  See comments on IN6_HOPOPTS.
1271 				 */
1272 				if (!privileged)
1273 					break;
1274 
1275 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1276 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1277 				    IPPROTO_IPV6);
1278 				if (*mp)
1279 					mp = &(*mp)->m_next;
1280 				break;
1281 			case IPPROTO_ROUTING:
1282 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1283 					break;
1284 
1285 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1286 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1287 				    IPPROTO_IPV6);
1288 				if (*mp)
1289 					mp = &(*mp)->m_next;
1290 				break;
1291 			case IPPROTO_HOPOPTS:
1292 			case IPPROTO_AH: /* is it possible? */
1293 				break;
1294 
1295 			default:
1296 				/*
1297 			 	 * other cases have been filtered in the above.
1298 				 * none will visit this case.  here we supply
1299 				 * the code just in case (nxt overwritten or
1300 				 * other cases).
1301 				 */
1302 #ifdef PULLDOWN_TEST
1303 				m_freem(ext);
1304 #endif
1305 				goto loopend;
1306 
1307 			}
1308 
1309 			/* proceed with the next header. */
1310 			off += elen;
1311 			nxt = ip6e->ip6e_nxt;
1312 			ip6e = NULL;
1313 #ifdef PULLDOWN_TEST
1314 			m_freem(ext);
1315 			ext = NULL;
1316 #endif
1317 		}
1318 	  loopend:
1319 		;
1320 	}
1321 #undef IS2292
1322 }
1323 
1324 void
1325 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1326 {
1327 	struct socket *so;
1328 	struct mbuf *m_mtu;
1329 	struct ip6_mtuinfo mtuctl;
1330 
1331 	so =  in6p->inp_socket;
1332 
1333 	if (mtu == NULL)
1334 		return;
1335 
1336 #ifdef DIAGNOSTIC
1337 	if (so == NULL)		/* I believe this is impossible */
1338 		panic("ip6_notify_pmtu: socket is NULL");
1339 #endif
1340 
1341 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1342 	mtuctl.ip6m_mtu = *mtu;
1343 	mtuctl.ip6m_addr = *dst;
1344 
1345 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1346 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1347 		return;
1348 
1349 	lwkt_gettoken(&so->so_rcv.ssb_token);
1350 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1351 	    == 0) {
1352 		m_freem(m_mtu);
1353 		/* XXX: should count statistics */
1354 	} else {
1355 		sorwakeup(so);
1356 	}
1357 	lwkt_reltoken(&so->so_rcv.ssb_token);
1358 }
1359 
1360 #ifdef PULLDOWN_TEST
1361 /*
1362  * pull single extension header from mbuf chain.  returns single mbuf that
1363  * contains the result, or NULL on error.
1364  */
1365 static struct mbuf *
1366 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1367 {
1368 	struct ip6_ext ip6e;
1369 	size_t elen;
1370 	struct mbuf *n;
1371 
1372 #ifdef DIAGNOSTIC
1373 	switch (nxt) {
1374 	case IPPROTO_DSTOPTS:
1375 	case IPPROTO_ROUTING:
1376 	case IPPROTO_HOPOPTS:
1377 	case IPPROTO_AH: /* is it possible? */
1378 		break;
1379 	default:
1380 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1381 	}
1382 #endif
1383 
1384 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1385 	if (nxt == IPPROTO_AH)
1386 		elen = (ip6e.ip6e_len + 2) << 2;
1387 	else
1388 		elen = (ip6e.ip6e_len + 1) << 3;
1389 
1390 	n = m_getb(elen, M_NOWAIT, MT_DATA, 0);
1391 	if (n == NULL)
1392 		return NULL;
1393 	n->m_len = 0;
1394 
1395 	if (elen >= M_TRAILINGSPACE(n)) {
1396 		m_free(n);
1397 		return NULL;
1398 	}
1399 
1400 	m_copydata(m, off, elen, mtod(n, caddr_t));
1401 	n->m_len = elen;
1402 	return n;
1403 }
1404 #endif
1405 
1406 /*
1407  * Get pointer to the previous header followed by the header
1408  * currently processed.
1409  * XXX: This function supposes that
1410  *	M includes all headers,
1411  *	the next header field and the header length field of each header
1412  *	are valid, and
1413  *	the sum of each header length equals to OFF.
1414  * Because of these assumptions, this function must be called very
1415  * carefully. Moreover, it will not be used in the near future when
1416  * we develop `neater' mechanism to process extension headers.
1417  */
1418 char *
1419 ip6_get_prevhdr(struct mbuf *m, int off)
1420 {
1421 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1422 
1423 	if (off == sizeof(struct ip6_hdr))
1424 		return (&ip6->ip6_nxt);
1425 	else {
1426 		int len, nxt;
1427 		struct ip6_ext *ip6e = NULL;
1428 
1429 		nxt = ip6->ip6_nxt;
1430 		len = sizeof(struct ip6_hdr);
1431 		while (len < off) {
1432 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1433 
1434 			switch (nxt) {
1435 			case IPPROTO_FRAGMENT:
1436 				len += sizeof(struct ip6_frag);
1437 				break;
1438 			case IPPROTO_AH:
1439 				len += (ip6e->ip6e_len + 2) << 2;
1440 				break;
1441 			default:
1442 				len += (ip6e->ip6e_len + 1) << 3;
1443 				break;
1444 			}
1445 			nxt = ip6e->ip6e_nxt;
1446 		}
1447 		if (ip6e)
1448 			return (&ip6e->ip6e_nxt);
1449 		else
1450 			return NULL;
1451 	}
1452 }
1453 
1454 /*
1455  * get next header offset.  m will be retained.
1456  */
1457 int
1458 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1459 {
1460 	struct ip6_hdr ip6;
1461 	struct ip6_ext ip6e;
1462 	struct ip6_frag fh;
1463 
1464 	/* just in case */
1465 	if (m == NULL)
1466 		panic("ip6_nexthdr: m == NULL");
1467 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1468 		return -1;
1469 
1470 	switch (proto) {
1471 	case IPPROTO_IPV6:
1472 		if (m->m_pkthdr.len < off + sizeof(ip6))
1473 			return -1;
1474 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1475 		if (nxtp)
1476 			*nxtp = ip6.ip6_nxt;
1477 		off += sizeof(ip6);
1478 		return off;
1479 
1480 	case IPPROTO_FRAGMENT:
1481 		/*
1482 		 * terminate parsing if it is not the first fragment,
1483 		 * it does not make sense to parse through it.
1484 		 */
1485 		if (m->m_pkthdr.len < off + sizeof(fh))
1486 			return -1;
1487 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1488 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1489 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1490 			return -1;
1491 		if (nxtp)
1492 			*nxtp = fh.ip6f_nxt;
1493 		off += sizeof(struct ip6_frag);
1494 		return off;
1495 
1496 	case IPPROTO_AH:
1497 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1498 			return -1;
1499 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1500 		if (nxtp)
1501 			*nxtp = ip6e.ip6e_nxt;
1502 		off += (ip6e.ip6e_len + 2) << 2;
1503 		return off;
1504 
1505 	case IPPROTO_HOPOPTS:
1506 	case IPPROTO_ROUTING:
1507 	case IPPROTO_DSTOPTS:
1508 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1509 			return -1;
1510 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1511 		if (nxtp)
1512 			*nxtp = ip6e.ip6e_nxt;
1513 		off += (ip6e.ip6e_len + 1) << 3;
1514 		return off;
1515 
1516 	case IPPROTO_NONE:
1517 	case IPPROTO_ESP:
1518 	case IPPROTO_IPCOMP:
1519 		/* give up */
1520 		return -1;
1521 
1522 	default:
1523 		return -1;
1524 	}
1525 
1526 	return -1;
1527 }
1528 
1529 /*
1530  * get offset for the last header in the chain.  m will be kept untainted.
1531  */
1532 int
1533 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1534 {
1535 	int newoff;
1536 	int nxt;
1537 
1538 	if (!nxtp) {
1539 		nxt = -1;
1540 		nxtp = &nxt;
1541 	}
1542 	while (1) {
1543 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1544 		if (newoff < 0)
1545 			return off;
1546 		else if (newoff < off)
1547 			return -1;	/* invalid */
1548 		else if (newoff == off)
1549 			return newoff;
1550 
1551 		off = newoff;
1552 		proto = *nxtp;
1553 	}
1554 }
1555 
1556 struct ip6aux *
1557 ip6_addaux(struct mbuf *m)
1558 {
1559 	struct m_tag *mtag;
1560 
1561 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1562 	if (!mtag) {
1563 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1564 		    M_NOWAIT);
1565 		if (mtag)
1566 			m_tag_prepend(m, mtag);
1567 	}
1568 	if (mtag)
1569 		bzero(m_tag_data(mtag), sizeof(struct ip6aux));
1570 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1571 }
1572 
1573 struct ip6aux *
1574 ip6_findaux(struct mbuf *m)
1575 {
1576 	struct m_tag *mtag;
1577 
1578 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1579 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1580 }
1581 
1582 void
1583 ip6_delaux(struct mbuf *m)
1584 {
1585 	struct m_tag *mtag;
1586 
1587 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1588 	if (mtag)
1589 		m_tag_delete(m, mtag);
1590 }
1591 
1592 /*
1593  * System control for IP6
1594  */
1595 
1596 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1597 	0,		0,		0,		0,
1598 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1599 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1600 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1601 	0,		0,		0,		0,
1602 	ENOPROTOOPT
1603 };
1604