xref: /dragonfly/sys/netinet6/ip6_input.c (revision 23265324)
1 /*	$FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.15 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/ip6_input.c,v 1.32 2006/12/22 23:57:53 swildner Exp $	*/
3 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  */
68 
69 #include "opt_ip6fw.h"
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/syslog.h>
86 #include <sys/proc.h>
87 
88 #include <sys/thread2.h>
89 #include <sys/msgport2.h>
90 
91 #include <net/if.h>
92 #include <net/if_types.h>
93 #include <net/if_dl.h>
94 #include <net/route.h>
95 #include <net/netisr.h>
96 #include <net/pfil.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_systm.h>
100 #ifdef INET
101 #include <netinet/ip.h>
102 #include <netinet/ip_icmp.h>
103 #endif /* INET */
104 #include <netinet/ip6.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet/in_pcb.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/scope6_var.h>
110 #include <netinet6/in6_ifattach.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_prefix.h>
113 
114 #ifdef IPSEC
115 #include <netinet6/ipsec.h>
116 #ifdef INET6
117 #include <netinet6/ipsec6.h>
118 #endif
119 #endif
120 
121 #ifdef FAST_IPSEC
122 #include <netproto/ipsec/ipsec.h>
123 #include <netproto/ipsec/ipsec6.h>
124 #define	IPSEC
125 #endif /* FAST_IPSEC */
126 
127 #include <net/ip6fw/ip6_fw.h>
128 
129 #include <netinet6/ip6protosw.h>
130 
131 /* we need it for NLOOP. */
132 #include "use_loop.h"
133 
134 #include <net/net_osdep.h>
135 
136 extern struct domain inet6domain;
137 extern struct ip6protosw inet6sw[];
138 
139 u_char ip6_protox[IPPROTO_MAX];
140 struct in6_ifaddr *in6_ifaddr;
141 
142 extern struct callout in6_tmpaddrtimer_ch;
143 
144 int ip6_forward_srcrt;			/* XXX */
145 int ip6_sourcecheck;			/* XXX */
146 int ip6_sourcecheck_interval;		/* XXX */
147 
148 int ip6_ours_check_algorithm;
149 
150 struct pfil_head inet6_pfil_hook;
151 
152 /* firewall hooks */
153 ip6_fw_chk_t *ip6_fw_chk_ptr;
154 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
155 int ip6_fw_enable = 1;
156 
157 struct ip6stat ip6stat;
158 
159 static void ip6_init2 (void *);
160 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
161 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
162 static int ip6_input(struct netmsg *msg);
163 #ifdef PULLDOWN_TEST
164 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
165 #endif
166 
167 /*
168  * IP6 initialization: fill in IP6 protocol switch table.
169  * All protocols not implemented in kernel go to raw IP6 protocol handler.
170  */
171 void
172 ip6_init(void)
173 {
174 	struct ip6protosw *pr;
175 	int i;
176 	struct timeval tv;
177 
178 #ifdef DIAGNOSTIC
179 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
180 		panic("sizeof(protosw) != sizeof(ip6protosw)");
181 #endif
182 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
183 	if (pr == 0)
184 		panic("ip6_init");
185 	for (i = 0; i < IPPROTO_MAX; i++)
186 		ip6_protox[i] = pr - inet6sw;
187 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
188 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
189 		if (pr->pr_domain->dom_family == PF_INET6 &&
190 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
191 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
192 
193 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
194 	inet6_pfil_hook.ph_af = AF_INET6;
195 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
196 		kprintf("%s: WARNING: unable to register pfil hook, "
197 			"error %d\n", __func__, i);
198 	}
199 
200 	netisr_register(NETISR_IPV6, cpu0_portfn, ip6_input);
201 	scope6_init();
202 	nd6_init();
203 	frag6_init();
204 	/*
205 	 * in many cases, random() here does NOT return random number
206 	 * as initialization during bootstrap time occur in fixed order.
207 	 */
208 	microtime(&tv);
209 	ip6_flow_seq = krandom() ^ tv.tv_usec;
210 	microtime(&tv);
211 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
212 }
213 
214 static void
215 ip6_init2(void *dummy)
216 {
217 	/* nd6_timer_init */
218 	callout_init(&nd6_timer_ch);
219 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
220 
221 	/* router renumbering prefix list maintenance */
222 	callout_init(&in6_rr_timer_ch);
223 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
224 
225 	/* timer for regeneranation of temporary addresses randomize ID */
226 	callout_init(&in6_tmpaddrtimer_ch);
227 	callout_reset(&in6_tmpaddrtimer_ch,
228 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
229 		       ip6_temp_regen_advance) * hz,
230 		      in6_tmpaddrtimer, NULL);
231 }
232 
233 /* cheat */
234 /* This must be after route_init(), which is now SI_ORDER_THIRD */
235 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
236 
237 extern struct	route_in6 ip6_forward_rt;
238 
239 static
240 int
241 ip6_input(struct netmsg *msg)
242 {
243 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
244 	struct ip6_hdr *ip6;
245 	int off = sizeof(struct ip6_hdr), nest;
246 	u_int32_t plen;
247 	u_int32_t rtalert = ~0;
248 	int nxt, ours = 0;
249 	struct ifnet *deliverifp = NULL;
250 	struct in6_addr odst;
251 	int srcrt = 0;
252 
253 #ifdef IPSEC
254 	/*
255 	 * should the inner packet be considered authentic?
256 	 * see comment in ah4_input().
257 	 */
258 	if (m) {
259 		m->m_flags &= ~M_AUTHIPHDR;
260 		m->m_flags &= ~M_AUTHIPDGM;
261 	}
262 #endif
263 
264 	/*
265 	 * make sure we don't have onion peering information into m_aux.
266 	 */
267 	ip6_delaux(m);
268 
269 	/*
270 	 * mbuf statistics
271 	 */
272 	if (m->m_flags & M_EXT) {
273 		if (m->m_next)
274 			ip6stat.ip6s_mext2m++;
275 		else
276 			ip6stat.ip6s_mext1++;
277 	} else {
278 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
279 		if (m->m_next) {
280 			if (m->m_flags & M_LOOP) {
281 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
282 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
283 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
284 			else
285 				ip6stat.ip6s_m2m[0]++;
286 		} else
287 			ip6stat.ip6s_m1++;
288 #undef M2MMAX
289 	}
290 
291 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
292 	ip6stat.ip6s_total++;
293 
294 #ifndef PULLDOWN_TEST
295 	/*
296 	 * L2 bridge code and some other code can return mbuf chain
297 	 * that does not conform to KAME requirement.  too bad.
298 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
299 	 */
300 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
301 		struct mbuf *n;
302 
303 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
304 		if (n == NULL)
305 			goto bad;
306 		M_MOVE_PKTHDR(n, m);
307 
308 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
309 		n->m_len = n->m_pkthdr.len;
310 		m_freem(m);
311 		m = n;
312 	}
313 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), EASYNC);
314 #endif
315 
316 	if (m->m_len < sizeof(struct ip6_hdr)) {
317 		struct ifnet *inifp;
318 		inifp = m->m_pkthdr.rcvif;
319 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
320 			ip6stat.ip6s_toosmall++;
321 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
322 			goto bad2;
323 		}
324 	}
325 
326 	ip6 = mtod(m, struct ip6_hdr *);
327 
328 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
329 		ip6stat.ip6s_badvers++;
330 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
331 		goto bad;
332 	}
333 
334 	/*
335 	 * Run through list of hooks for input packets.
336 	 *
337 	 * NB: Beware of the destination address changing
338 	 *     (e.g. by NAT rewriting). When this happens,
339 	 *     tell ip6_forward to do the right thing.
340 	 */
341 	if (pfil_has_hooks(&inet6_pfil_hook)) {
342 		odst = ip6->ip6_dst;
343 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
344 		    m->m_pkthdr.rcvif, PFIL_IN)) {
345 			goto bad2;
346 		}
347 		if (m == NULL)			/* consumed by filter */
348 			goto bad2;
349 		ip6 = mtod(m, struct ip6_hdr *);
350 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
351 	}
352 
353 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
354 
355 #ifdef ALTQ
356 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
357 		/* packet is dropped by traffic conditioner */
358 		return (ENOBUFS);
359 	}
360 #endif
361 
362 	/*
363 	 * Check with the firewall...
364 	 */
365 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
366 		u_short port = 0;
367 		/* If ipfw says divert, we have to just drop packet */
368 		/* use port as a dummy argument */
369 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
370 			m_freem(m);
371 			m = NULL;
372 		}
373 		if (!m)
374 			goto bad2;
375 	}
376 
377 	/*
378 	 * Check against address spoofing/corruption.
379 	 */
380 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
381 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
382 		/*
383 		 * XXX: "badscope" is not very suitable for a multicast source.
384 		 */
385 		ip6stat.ip6s_badscope++;
386 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
387 		goto bad;
388 	}
389 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
390 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
391 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
392 		ip6stat.ip6s_badscope++;
393 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
394 		goto bad;
395 	}
396 
397 	/*
398 	 * The following check is not documented in specs.  A malicious
399 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
400 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
401 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
402 	 *
403 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
404 	 * support IPv4-less kernel compilation, we cannot support SIIT
405 	 * environment at all.  So, it makes more sense for us to reject any
406 	 * malicious packets for non-SIIT environment, than try to do a
407 	 * partical support for SIIT environment.
408 	 */
409 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
410 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
411 		ip6stat.ip6s_badscope++;
412 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
413 		goto bad;
414 	}
415 #if 0
416 	/*
417 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
418 	 *
419 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
420 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
421 	 * is revised to forbid relaying case.
422 	 */
423 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
424 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
425 		ip6stat.ip6s_badscope++;
426 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
427 		goto bad;
428 	}
429 #endif
430 
431 	/* drop packets if interface ID portion is already filled */
432 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
433 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
434 		    ip6->ip6_src.s6_addr16[1]) {
435 			ip6stat.ip6s_badscope++;
436 			goto bad;
437 		}
438 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
439 		    ip6->ip6_dst.s6_addr16[1]) {
440 			ip6stat.ip6s_badscope++;
441 			goto bad;
442 		}
443 	}
444 
445 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
446 		ip6->ip6_src.s6_addr16[1]
447 			= htons(m->m_pkthdr.rcvif->if_index);
448 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
449 		ip6->ip6_dst.s6_addr16[1]
450 			= htons(m->m_pkthdr.rcvif->if_index);
451 
452 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
453 	/*
454 	 * We use rt->rt_ifp to determine if the address is ours or not.
455 	 * If rt_ifp is lo0, the address is ours.
456 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
457 	 * so any address under fe80::%lo0/64 will be mistakenly considered
458 	 * local.  The special case is supplied to handle the case properly
459 	 * by actually looking at interface addresses
460 	 * (using in6ifa_ifpwithaddr).
461 	 */
462 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
463 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
464 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
465 			icmp6_error(m, ICMP6_DST_UNREACH,
466 			    ICMP6_DST_UNREACH_ADDR, 0);
467 			/* m is already freed */
468 			goto bad2;
469 		}
470 
471 		ours = 1;
472 		deliverifp = m->m_pkthdr.rcvif;
473 		goto hbhcheck;
474 	}
475 #endif
476 
477 	/*
478 	 * Multicast check
479 	 */
480 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
481 	  	struct	in6_multi *in6m = 0;
482 
483 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
484 		/*
485 		 * See if we belong to the destination multicast group on the
486 		 * arrival interface.
487 		 */
488 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
489 		if (in6m)
490 			ours = 1;
491 		else if (!ip6_mrouter) {
492 			ip6stat.ip6s_notmember++;
493 			ip6stat.ip6s_cantforward++;
494 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
495 			goto bad;
496 		}
497 		deliverifp = m->m_pkthdr.rcvif;
498 		goto hbhcheck;
499 	}
500 
501 	/*
502 	 *  Unicast check
503 	 */
504 	switch (ip6_ours_check_algorithm) {
505 	default:
506 		/*
507 		 * XXX: I intentionally broke our indentation rule here,
508 		 *      since this switch-case is just for measurement and
509 		 *      therefore should soon be removed.
510 		 */
511 	if (ip6_forward_rt.ro_rt != NULL &&
512 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
513 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
514 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
515 		ip6stat.ip6s_forward_cachehit++;
516 	else {
517 		struct sockaddr_in6 *dst6;
518 
519 		if (ip6_forward_rt.ro_rt) {
520 			/* route is down or destination is different */
521 			ip6stat.ip6s_forward_cachemiss++;
522 			RTFREE(ip6_forward_rt.ro_rt);
523 			ip6_forward_rt.ro_rt = 0;
524 		}
525 
526 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
527 		dst6 = &ip6_forward_rt.ro_dst;
528 		dst6->sin6_len = sizeof(struct sockaddr_in6);
529 		dst6->sin6_family = AF_INET6;
530 		dst6->sin6_addr = ip6->ip6_dst;
531 
532 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
533 	}
534 
535 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
536 
537 	/*
538 	 * Accept the packet if the forwarding interface to the destination
539 	 * according to the routing table is the loopback interface,
540 	 * unless the associated route has a gateway.
541 	 * Note that this approach causes to accept a packet if there is a
542 	 * route to the loopback interface for the destination of the packet.
543 	 * But we think it's even useful in some situations, e.g. when using
544 	 * a special daemon which wants to intercept the packet.
545 	 *
546 	 * XXX: some OSes automatically make a cloned route for the destination
547 	 * of an outgoing packet.  If the outgoing interface of the packet
548 	 * is a loopback one, the kernel would consider the packet to be
549 	 * accepted, even if we have no such address assinged on the interface.
550 	 * We check the cloned flag of the route entry to reject such cases,
551 	 * assuming that route entries for our own addresses are not made by
552 	 * cloning (it should be true because in6_addloop explicitly installs
553 	 * the host route).  However, we might have to do an explicit check
554 	 * while it would be less efficient.  Or, should we rather install a
555 	 * reject route for such a case?
556 	 */
557 	if (ip6_forward_rt.ro_rt &&
558 	    (ip6_forward_rt.ro_rt->rt_flags &
559 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
560 #ifdef RTF_WASCLONED
561 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
562 #endif
563 #ifdef RTF_CLONED
564 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
565 #endif
566 #if 0
567 	    /*
568 	     * The check below is redundant since the comparison of
569 	     * the destination and the key of the rtentry has
570 	     * already done through looking up the routing table.
571 	     */
572 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
573 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
574 #endif
575 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
576 		struct in6_ifaddr *ia6 =
577 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
578 
579 		/*
580 		 * record address information into m_aux.
581 		 */
582 		ip6_setdstifaddr(m, ia6);
583 
584 		/*
585 		 * packets to a tentative, duplicated, or somehow invalid
586 		 * address must not be accepted.
587 		 */
588 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
589 			/* this address is ready */
590 			ours = 1;
591 			deliverifp = ia6->ia_ifp;	/* correct? */
592 			/* Count the packet in the ip address stats */
593 			ia6->ia_ifa.if_ipackets++;
594 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
595 			goto hbhcheck;
596 		} else {
597 			/* address is not ready, so discard the packet. */
598 			nd6log((LOG_INFO,
599 			    "ip6_input: packet to an unready address %s->%s\n",
600 			    ip6_sprintf(&ip6->ip6_src),
601 			    ip6_sprintf(&ip6->ip6_dst)));
602 
603 			goto bad;
604 		}
605 	}
606 	} /* XXX indentation (see above) */
607 
608 	/*
609 	 * FAITH(Firewall Aided Internet Translator)
610 	 */
611 	if (ip6_keepfaith) {
612 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
613 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
614 			/* XXX do we need more sanity checks? */
615 			ours = 1;
616 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
617 			goto hbhcheck;
618 		}
619 	}
620 
621 	/*
622 	 * Now there is no reason to process the packet if it's not our own
623 	 * and we're not a router.
624 	 */
625 	if (!ip6_forwarding) {
626 		ip6stat.ip6s_cantforward++;
627 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
628 		goto bad;
629 	}
630 
631 hbhcheck:
632 	/*
633 	 * record address information into m_aux, if we don't have one yet.
634 	 * note that we are unable to record it, if the address is not listed
635 	 * as our interface address (e.g. multicast addresses, addresses
636 	 * within FAITH prefixes and such).
637 	 */
638 	if (deliverifp && !ip6_getdstifaddr(m)) {
639 		struct in6_ifaddr *ia6;
640 
641 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
642 		if (ia6) {
643 			if (!ip6_setdstifaddr(m, ia6)) {
644 				/*
645 				 * XXX maybe we should drop the packet here,
646 				 * as we could not provide enough information
647 				 * to the upper layers.
648 				 */
649 			}
650 		}
651 	}
652 
653 	/*
654 	 * Process Hop-by-Hop options header if it's contained.
655 	 * m may be modified in ip6_hopopts_input().
656 	 * If a JumboPayload option is included, plen will also be modified.
657 	 */
658 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
659 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
660 		struct ip6_hbh *hbh;
661 
662 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
663 #if 0	/*touches NULL pointer*/
664 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
665 #endif
666 			goto bad2;	/* m have already been freed */
667 		}
668 
669 		/* adjust pointer */
670 		ip6 = mtod(m, struct ip6_hdr *);
671 
672 		/*
673 		 * if the payload length field is 0 and the next header field
674 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
675 		 * option MUST be included.
676 		 */
677 		if (ip6->ip6_plen == 0 && plen == 0) {
678 			/*
679 			 * Note that if a valid jumbo payload option is
680 			 * contained, ip6_hoptops_input() must set a valid
681 			 * (non-zero) payload length to the variable plen.
682 			 */
683 			ip6stat.ip6s_badoptions++;
684 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
685 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
686 			icmp6_error(m, ICMP6_PARAM_PROB,
687 				    ICMP6_PARAMPROB_HEADER,
688 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
689 			goto bad2;
690 		}
691 #ifndef PULLDOWN_TEST
692 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
693 		hbh = (struct ip6_hbh *)(ip6 + 1);
694 #else
695 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
696 			sizeof(struct ip6_hbh));
697 		if (hbh == NULL) {
698 			ip6stat.ip6s_tooshort++;
699 			goto bad2;
700 		}
701 #endif
702 		nxt = hbh->ip6h_nxt;
703 
704 		/*
705 		 * accept the packet if a router alert option is included
706 		 * and we act as an IPv6 router.
707 		 */
708 		if (rtalert != ~0 && ip6_forwarding)
709 			ours = 1;
710 	} else
711 		nxt = ip6->ip6_nxt;
712 
713 	/*
714 	 * Check that the amount of data in the buffers
715 	 * is as at least much as the IPv6 header would have us expect.
716 	 * Trim mbufs if longer than we expect.
717 	 * Drop packet if shorter than we expect.
718 	 */
719 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
720 		ip6stat.ip6s_tooshort++;
721 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
722 		goto bad;
723 	}
724 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
725 		if (m->m_len == m->m_pkthdr.len) {
726 			m->m_len = sizeof(struct ip6_hdr) + plen;
727 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
728 		} else
729 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
730 	}
731 
732 	/*
733 	 * Forward if desirable.
734 	 */
735 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
736 		/*
737 		 * If we are acting as a multicast router, all
738 		 * incoming multicast packets are passed to the
739 		 * kernel-level multicast forwarding function.
740 		 * The packet is returned (relatively) intact; if
741 		 * ip6_mforward() returns a non-zero value, the packet
742 		 * must be discarded, else it may be accepted below.
743 		 */
744 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
745 			ip6stat.ip6s_cantforward++;
746 			goto bad;
747 		}
748 		if (!ours)
749 			goto bad;
750 	} else if (!ours) {
751 		ip6_forward(m, srcrt);
752 		goto bad2;
753 	}
754 
755 	ip6 = mtod(m, struct ip6_hdr *);
756 
757 	/*
758 	 * Malicious party may be able to use IPv4 mapped addr to confuse
759 	 * tcp/udp stack and bypass security checks (act as if it was from
760 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
761 	 *
762 	 * For SIIT end node behavior, you may want to disable the check.
763 	 * However, you will  become vulnerable to attacks using IPv4 mapped
764 	 * source.
765 	 */
766 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
767 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
768 		ip6stat.ip6s_badscope++;
769 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
770 		goto bad;
771 	}
772 
773 	/*
774 	 * Tell launch routine the next header
775 	 */
776 	ip6stat.ip6s_delivered++;
777 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
778 	nest = 0;
779 
780 	while (nxt != IPPROTO_DONE) {
781 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
782 			ip6stat.ip6s_toomanyhdr++;
783 			goto bad;
784 		}
785 
786 		/*
787 		 * protection against faulty packet - there should be
788 		 * more sanity checks in header chain processing.
789 		 */
790 		if (m->m_pkthdr.len < off) {
791 			ip6stat.ip6s_tooshort++;
792 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
793 			goto bad;
794 		}
795 
796 #if 0
797 		/*
798 		 * do we need to do it for every header?  yeah, other
799 		 * functions can play with it (like re-allocate and copy).
800 		 */
801 		mhist = ip6_addaux(m);
802 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
803 			hist = mtod(mhist, caddr_t) + mhist->m_len;
804 			bcopy(&nxt, hist, sizeof(nxt));
805 			mhist->m_len += sizeof(nxt);
806 		} else {
807 			ip6stat.ip6s_toomanyhdr++;
808 			goto bad;
809 		}
810 #endif
811 
812 #ifdef IPSEC
813 		/*
814 		 * enforce IPsec policy checking if we are seeing last header.
815 		 * note that we do not visit this with protocols with pcb layer
816 		 * code - like udp/tcp/raw ip.
817 		 */
818 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) &&
819 		    ipsec6_in_reject(m, NULL)) {
820 			ipsec6stat.in_polvio++;
821 			goto bad;
822 		}
823 #endif
824 
825 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
826 	}
827 	goto bad2;
828 bad:
829 	m_freem(m);
830 bad2:
831 	/* msg was embedded in the mbuf, do not reply! */
832 	return (EASYNC);
833 }
834 
835 /*
836  * set/grab in6_ifaddr correspond to IPv6 destination address.
837  * XXX backward compatibility wrapper
838  */
839 static struct ip6aux *
840 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
841 {
842 	struct ip6aux *n;
843 
844 	n = ip6_addaux(m);
845 	if (n)
846 		n->ip6a_dstia6 = ia6;
847 	return n;	/* NULL if failed to set */
848 }
849 
850 struct in6_ifaddr *
851 ip6_getdstifaddr(struct mbuf *m)
852 {
853 	struct ip6aux *n;
854 
855 	n = ip6_findaux(m);
856 	if (n)
857 		return n->ip6a_dstia6;
858 	else
859 		return NULL;
860 }
861 
862 /*
863  * Hop-by-Hop options header processing. If a valid jumbo payload option is
864  * included, the real payload length will be stored in plenp.
865  */
866 static int
867 ip6_hopopts_input(u_int32_t *plenp,
868 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
869 		  struct mbuf **mp,
870 		  int *offp)
871 {
872 	struct mbuf *m = *mp;
873 	int off = *offp, hbhlen;
874 	struct ip6_hbh *hbh;
875 	u_int8_t *opt;
876 
877 	/* validation of the length of the header */
878 #ifndef PULLDOWN_TEST
879 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
880 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
881 	hbhlen = (hbh->ip6h_len + 1) << 3;
882 
883 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
884 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
885 #else
886 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
887 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
888 	if (hbh == NULL) {
889 		ip6stat.ip6s_tooshort++;
890 		return -1;
891 	}
892 	hbhlen = (hbh->ip6h_len + 1) << 3;
893 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
894 		hbhlen);
895 	if (hbh == NULL) {
896 		ip6stat.ip6s_tooshort++;
897 		return -1;
898 	}
899 #endif
900 	off += hbhlen;
901 	hbhlen -= sizeof(struct ip6_hbh);
902 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
903 
904 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
905 				hbhlen, rtalertp, plenp) < 0)
906 		return (-1);
907 
908 	*offp = off;
909 	*mp = m;
910 	return (0);
911 }
912 
913 /*
914  * Search header for all Hop-by-hop options and process each option.
915  * This function is separate from ip6_hopopts_input() in order to
916  * handle a case where the sending node itself process its hop-by-hop
917  * options header. In such a case, the function is called from ip6_output().
918  *
919  * The function assumes that hbh header is located right after the IPv6 header
920  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
921  * opthead + hbhlen is located in continuous memory region.
922  */
923 int
924 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
925 		    u_int32_t *rtalertp, u_int32_t *plenp)
926 {
927 	struct ip6_hdr *ip6;
928 	int optlen = 0;
929 	u_int8_t *opt = opthead;
930 	u_int16_t rtalert_val;
931 	u_int32_t jumboplen;
932 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
933 
934 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
935 		switch (*opt) {
936 		case IP6OPT_PAD1:
937 			optlen = 1;
938 			break;
939 		case IP6OPT_PADN:
940 			if (hbhlen < IP6OPT_MINLEN) {
941 				ip6stat.ip6s_toosmall++;
942 				goto bad;
943 			}
944 			optlen = *(opt + 1) + 2;
945 			break;
946 		case IP6OPT_RTALERT:
947 			/* XXX may need check for alignment */
948 			if (hbhlen < IP6OPT_RTALERT_LEN) {
949 				ip6stat.ip6s_toosmall++;
950 				goto bad;
951 			}
952 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
953 				/* XXX stat */
954 				icmp6_error(m, ICMP6_PARAM_PROB,
955 					    ICMP6_PARAMPROB_HEADER,
956 					    erroff + opt + 1 - opthead);
957 				return (-1);
958 			}
959 			optlen = IP6OPT_RTALERT_LEN;
960 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
961 			*rtalertp = ntohs(rtalert_val);
962 			break;
963 		case IP6OPT_JUMBO:
964 			/* XXX may need check for alignment */
965 			if (hbhlen < IP6OPT_JUMBO_LEN) {
966 				ip6stat.ip6s_toosmall++;
967 				goto bad;
968 			}
969 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
970 				/* XXX stat */
971 				icmp6_error(m, ICMP6_PARAM_PROB,
972 					    ICMP6_PARAMPROB_HEADER,
973 					    erroff + opt + 1 - opthead);
974 				return (-1);
975 			}
976 			optlen = IP6OPT_JUMBO_LEN;
977 
978 			/*
979 			 * IPv6 packets that have non 0 payload length
980 			 * must not contain a jumbo payload option.
981 			 */
982 			ip6 = mtod(m, struct ip6_hdr *);
983 			if (ip6->ip6_plen) {
984 				ip6stat.ip6s_badoptions++;
985 				icmp6_error(m, ICMP6_PARAM_PROB,
986 					    ICMP6_PARAMPROB_HEADER,
987 					    erroff + opt - opthead);
988 				return (-1);
989 			}
990 
991 			/*
992 			 * We may see jumbolen in unaligned location, so
993 			 * we'd need to perform bcopy().
994 			 */
995 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
996 			jumboplen = (u_int32_t)htonl(jumboplen);
997 
998 #if 1
999 			/*
1000 			 * if there are multiple jumbo payload options,
1001 			 * *plenp will be non-zero and the packet will be
1002 			 * rejected.
1003 			 * the behavior may need some debate in ipngwg -
1004 			 * multiple options does not make sense, however,
1005 			 * there's no explicit mention in specification.
1006 			 */
1007 			if (*plenp != 0) {
1008 				ip6stat.ip6s_badoptions++;
1009 				icmp6_error(m, ICMP6_PARAM_PROB,
1010 					    ICMP6_PARAMPROB_HEADER,
1011 					    erroff + opt + 2 - opthead);
1012 				return (-1);
1013 			}
1014 #endif
1015 
1016 			/*
1017 			 * jumbo payload length must be larger than 65535.
1018 			 */
1019 			if (jumboplen <= IPV6_MAXPACKET) {
1020 				ip6stat.ip6s_badoptions++;
1021 				icmp6_error(m, ICMP6_PARAM_PROB,
1022 					    ICMP6_PARAMPROB_HEADER,
1023 					    erroff + opt + 2 - opthead);
1024 				return (-1);
1025 			}
1026 			*plenp = jumboplen;
1027 
1028 			break;
1029 		default:		/* unknown option */
1030 			if (hbhlen < IP6OPT_MINLEN) {
1031 				ip6stat.ip6s_toosmall++;
1032 				goto bad;
1033 			}
1034 			optlen = ip6_unknown_opt(opt, m,
1035 			    erroff + opt - opthead);
1036 			if (optlen == -1)
1037 				return (-1);
1038 			optlen += 2;
1039 			break;
1040 		}
1041 	}
1042 
1043 	return (0);
1044 
1045 bad:
1046 	m_freem(m);
1047 	return (-1);
1048 }
1049 
1050 /*
1051  * Unknown option processing.
1052  * The third argument `off' is the offset from the IPv6 header to the option,
1053  * which is necessary if the IPv6 header the and option header and IPv6 header
1054  * is not continuous in order to return an ICMPv6 error.
1055  */
1056 int
1057 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1058 {
1059 	struct ip6_hdr *ip6;
1060 
1061 	switch (IP6OPT_TYPE(*optp)) {
1062 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1063 		return ((int)*(optp + 1));
1064 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1065 		m_freem(m);
1066 		return (-1);
1067 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1068 		ip6stat.ip6s_badoptions++;
1069 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1070 		return (-1);
1071 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1072 		ip6stat.ip6s_badoptions++;
1073 		ip6 = mtod(m, struct ip6_hdr *);
1074 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1075 		    (m->m_flags & (M_BCAST|M_MCAST)))
1076 			m_freem(m);
1077 		else
1078 			icmp6_error(m, ICMP6_PARAM_PROB,
1079 				    ICMP6_PARAMPROB_OPTION, off);
1080 		return (-1);
1081 	}
1082 
1083 	m_freem(m);		/* XXX: NOTREACHED */
1084 	return (-1);
1085 }
1086 
1087 /*
1088  * Create the "control" list for this pcb.
1089  * The function will not modify mbuf chain at all.
1090  *
1091  * with KAME mbuf chain restriction:
1092  * The routine will be called from upper layer handlers like tcp6_input().
1093  * Thus the routine assumes that the caller (tcp6_input) have already
1094  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1095  * very first mbuf on the mbuf chain.
1096  */
1097 void
1098 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1099 		struct mbuf *m)
1100 {
1101 	struct thread *td = curthread;	/* XXX */
1102 	int privileged = 0;
1103 	int rthdr_exist = 0;
1104 
1105 
1106 	if (suser(td) == 0)
1107  		privileged++;
1108 
1109 #ifdef SO_TIMESTAMP
1110 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1111 		struct timeval tv;
1112 
1113 		microtime(&tv);
1114 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1115 				      SCM_TIMESTAMP, SOL_SOCKET);
1116 		if (*mp) {
1117 			mp = &(*mp)->m_next;
1118 		}
1119 	}
1120 #endif
1121 
1122 	/* RFC 2292 sec. 5 */
1123 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1124 		struct in6_pktinfo pi6;
1125 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1126 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1127 			pi6.ipi6_addr.s6_addr16[1] = 0;
1128 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1129 					? m->m_pkthdr.rcvif->if_index
1130 					: 0;
1131 		*mp = sbcreatecontrol((caddr_t) &pi6,
1132 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1133 			IPPROTO_IPV6);
1134 		if (*mp)
1135 			mp = &(*mp)->m_next;
1136 	}
1137 
1138 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1139 		int hlim = ip6->ip6_hlim & 0xff;
1140 		*mp = sbcreatecontrol((caddr_t) &hlim,
1141 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1142 		if (*mp)
1143 			mp = &(*mp)->m_next;
1144 	}
1145 
1146 	/*
1147 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1148 	 * for the option, but it might be too strict, since there might
1149 	 * be some hop-by-hop options which can be returned to normal user.
1150 	 * See RFC 2292 section 6.
1151 	 */
1152 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1153 		/*
1154 		 * Check if a hop-by-hop options header is contatined in the
1155 		 * received packet, and if so, store the options as ancillary
1156 		 * data. Note that a hop-by-hop options header must be
1157 		 * just after the IPv6 header, which fact is assured through
1158 		 * the IPv6 input processing.
1159 		 */
1160 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1161 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1162 			struct ip6_hbh *hbh;
1163 			int hbhlen = 0;
1164 #ifdef PULLDOWN_TEST
1165 			struct mbuf *ext;
1166 #endif
1167 
1168 #ifndef PULLDOWN_TEST
1169 			hbh = (struct ip6_hbh *)(ip6 + 1);
1170 			hbhlen = (hbh->ip6h_len + 1) << 3;
1171 #else
1172 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1173 			    ip6->ip6_nxt);
1174 			if (ext == NULL) {
1175 				ip6stat.ip6s_tooshort++;
1176 				return;
1177 			}
1178 			hbh = mtod(ext, struct ip6_hbh *);
1179 			hbhlen = (hbh->ip6h_len + 1) << 3;
1180 			if (hbhlen != ext->m_len) {
1181 				m_freem(ext);
1182 				ip6stat.ip6s_tooshort++;
1183 				return;
1184 			}
1185 #endif
1186 
1187 			/*
1188 			 * XXX: We copy whole the header even if a jumbo
1189 			 * payload option is included, which option is to
1190 			 * be removed before returning in the RFC 2292.
1191 			 * Note: this constraint is removed in 2292bis.
1192 			 */
1193 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1194 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1195 			if (*mp)
1196 				mp = &(*mp)->m_next;
1197 #ifdef PULLDOWN_TEST
1198 			m_freem(ext);
1199 #endif
1200 		}
1201 	}
1202 
1203 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1204 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1205 		int proto, off, nxt;
1206 
1207 		/*
1208 		 * go through the header chain to see if a routing header is
1209 		 * contained in the packet. We need this information to store
1210 		 * destination options headers (if any) properly.
1211 		 * XXX: performance issue. We should record this info when
1212 		 * processing extension headers in incoming routine.
1213 		 * (todo) use m_aux?
1214 		 */
1215 		proto = IPPROTO_IPV6;
1216 		off = 0;
1217 		nxt = -1;
1218 		while (1) {
1219 			int newoff;
1220 
1221 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1222 			if (newoff < 0)
1223 				break;
1224 			if (newoff < off) /* invalid, check for safety */
1225 				break;
1226 			if ((proto = nxt) == IPPROTO_ROUTING) {
1227 				rthdr_exist = 1;
1228 				break;
1229 			}
1230 			off = newoff;
1231 		}
1232 	}
1233 
1234 	if ((in6p->in6p_flags &
1235 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1236 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1237 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1238 
1239 		/*
1240 		 * Search for destination options headers or routing
1241 		 * header(s) through the header chain, and stores each
1242 		 * header as ancillary data.
1243 		 * Note that the order of the headers remains in
1244 		 * the chain of ancillary data.
1245 		 */
1246 		while (1) {	/* is explicit loop prevention necessary? */
1247 			struct ip6_ext *ip6e = NULL;
1248 			int elen;
1249 #ifdef PULLDOWN_TEST
1250 			struct mbuf *ext = NULL;
1251 #endif
1252 
1253 			/*
1254 			 * if it is not an extension header, don't try to
1255 			 * pull it from the chain.
1256 			 */
1257 			switch (nxt) {
1258 			case IPPROTO_DSTOPTS:
1259 			case IPPROTO_ROUTING:
1260 			case IPPROTO_HOPOPTS:
1261 			case IPPROTO_AH: /* is it possible? */
1262 				break;
1263 			default:
1264 				goto loopend;
1265 			}
1266 
1267 #ifndef PULLDOWN_TEST
1268 			if (off + sizeof(*ip6e) > m->m_len)
1269 				goto loopend;
1270 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1271 			if (nxt == IPPROTO_AH)
1272 				elen = (ip6e->ip6e_len + 2) << 2;
1273 			else
1274 				elen = (ip6e->ip6e_len + 1) << 3;
1275 			if (off + elen > m->m_len)
1276 				goto loopend;
1277 #else
1278 			ext = ip6_pullexthdr(m, off, nxt);
1279 			if (ext == NULL) {
1280 				ip6stat.ip6s_tooshort++;
1281 				return;
1282 			}
1283 			ip6e = mtod(ext, struct ip6_ext *);
1284 			if (nxt == IPPROTO_AH)
1285 				elen = (ip6e->ip6e_len + 2) << 2;
1286 			else
1287 				elen = (ip6e->ip6e_len + 1) << 3;
1288 			if (elen != ext->m_len) {
1289 				m_freem(ext);
1290 				ip6stat.ip6s_tooshort++;
1291 				return;
1292 			}
1293 #endif
1294 
1295 			switch (nxt) {
1296 			case IPPROTO_DSTOPTS:
1297 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1298 					break;
1299 
1300 				/*
1301 				 * We also require super-user privilege for
1302 				 * the option.
1303 				 * See the comments on IN6_HOPOPTS.
1304 				 */
1305 				if (!privileged)
1306 					break;
1307 
1308 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1309 						      IPV6_DSTOPTS,
1310 						      IPPROTO_IPV6);
1311 				if (*mp)
1312 					mp = &(*mp)->m_next;
1313 				break;
1314 			case IPPROTO_ROUTING:
1315 				if (!in6p->in6p_flags & IN6P_RTHDR)
1316 					break;
1317 
1318 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1319 						      IPV6_RTHDR,
1320 						      IPPROTO_IPV6);
1321 				if (*mp)
1322 					mp = &(*mp)->m_next;
1323 				break;
1324 			case IPPROTO_HOPOPTS:
1325 			case IPPROTO_AH: /* is it possible? */
1326 				break;
1327 
1328 			default:
1329 				/*
1330 			 	 * other cases have been filtered in the above.
1331 				 * none will visit this case.  here we supply
1332 				 * the code just in case (nxt overwritten or
1333 				 * other cases).
1334 				 */
1335 #ifdef PULLDOWN_TEST
1336 				m_freem(ext);
1337 #endif
1338 				goto loopend;
1339 
1340 			}
1341 
1342 			/* proceed with the next header. */
1343 			off += elen;
1344 			nxt = ip6e->ip6e_nxt;
1345 			ip6e = NULL;
1346 #ifdef PULLDOWN_TEST
1347 			m_freem(ext);
1348 			ext = NULL;
1349 #endif
1350 		}
1351 	  loopend:
1352 		;
1353 	}
1354 
1355 }
1356 
1357 #ifdef PULLDOWN_TEST
1358 /*
1359  * pull single extension header from mbuf chain.  returns single mbuf that
1360  * contains the result, or NULL on error.
1361  */
1362 static struct mbuf *
1363 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1364 {
1365 	struct ip6_ext ip6e;
1366 	size_t elen;
1367 	struct mbuf *n;
1368 
1369 #ifdef DIAGNOSTIC
1370 	switch (nxt) {
1371 	case IPPROTO_DSTOPTS:
1372 	case IPPROTO_ROUTING:
1373 	case IPPROTO_HOPOPTS:
1374 	case IPPROTO_AH: /* is it possible? */
1375 		break;
1376 	default:
1377 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1378 	}
1379 #endif
1380 
1381 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1382 	if (nxt == IPPROTO_AH)
1383 		elen = (ip6e.ip6e_len + 2) << 2;
1384 	else
1385 		elen = (ip6e.ip6e_len + 1) << 3;
1386 
1387 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1388 	if (n == NULL)
1389 		return NULL;
1390 	n->m_len = 0;
1391 
1392 	if (elen >= M_TRAILINGSPACE(n)) {
1393 		m_free(n);
1394 		return NULL;
1395 	}
1396 
1397 	m_copydata(m, off, elen, mtod(n, caddr_t));
1398 	n->m_len = elen;
1399 	return n;
1400 }
1401 #endif
1402 
1403 /*
1404  * Get pointer to the previous header followed by the header
1405  * currently processed.
1406  * XXX: This function supposes that
1407  *	M includes all headers,
1408  *	the next header field and the header length field of each header
1409  *	are valid, and
1410  *	the sum of each header length equals to OFF.
1411  * Because of these assumptions, this function must be called very
1412  * carefully. Moreover, it will not be used in the near future when
1413  * we develop `neater' mechanism to process extension headers.
1414  */
1415 char *
1416 ip6_get_prevhdr(struct mbuf *m, int off)
1417 {
1418 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1419 
1420 	if (off == sizeof(struct ip6_hdr))
1421 		return (&ip6->ip6_nxt);
1422 	else {
1423 		int len, nxt;
1424 		struct ip6_ext *ip6e = NULL;
1425 
1426 		nxt = ip6->ip6_nxt;
1427 		len = sizeof(struct ip6_hdr);
1428 		while (len < off) {
1429 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1430 
1431 			switch (nxt) {
1432 			case IPPROTO_FRAGMENT:
1433 				len += sizeof(struct ip6_frag);
1434 				break;
1435 			case IPPROTO_AH:
1436 				len += (ip6e->ip6e_len + 2) << 2;
1437 				break;
1438 			default:
1439 				len += (ip6e->ip6e_len + 1) << 3;
1440 				break;
1441 			}
1442 			nxt = ip6e->ip6e_nxt;
1443 		}
1444 		if (ip6e)
1445 			return (&ip6e->ip6e_nxt);
1446 		else
1447 			return NULL;
1448 	}
1449 }
1450 
1451 /*
1452  * get next header offset.  m will be retained.
1453  */
1454 int
1455 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1456 {
1457 	struct ip6_hdr ip6;
1458 	struct ip6_ext ip6e;
1459 	struct ip6_frag fh;
1460 
1461 	/* just in case */
1462 	if (m == NULL)
1463 		panic("ip6_nexthdr: m == NULL");
1464 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1465 		return -1;
1466 
1467 	switch (proto) {
1468 	case IPPROTO_IPV6:
1469 		if (m->m_pkthdr.len < off + sizeof(ip6))
1470 			return -1;
1471 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1472 		if (nxtp)
1473 			*nxtp = ip6.ip6_nxt;
1474 		off += sizeof(ip6);
1475 		return off;
1476 
1477 	case IPPROTO_FRAGMENT:
1478 		/*
1479 		 * terminate parsing if it is not the first fragment,
1480 		 * it does not make sense to parse through it.
1481 		 */
1482 		if (m->m_pkthdr.len < off + sizeof(fh))
1483 			return -1;
1484 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1485 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1486 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1487 			return -1;
1488 		if (nxtp)
1489 			*nxtp = fh.ip6f_nxt;
1490 		off += sizeof(struct ip6_frag);
1491 		return off;
1492 
1493 	case IPPROTO_AH:
1494 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1495 			return -1;
1496 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1497 		if (nxtp)
1498 			*nxtp = ip6e.ip6e_nxt;
1499 		off += (ip6e.ip6e_len + 2) << 2;
1500 		return off;
1501 
1502 	case IPPROTO_HOPOPTS:
1503 	case IPPROTO_ROUTING:
1504 	case IPPROTO_DSTOPTS:
1505 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1506 			return -1;
1507 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1508 		if (nxtp)
1509 			*nxtp = ip6e.ip6e_nxt;
1510 		off += (ip6e.ip6e_len + 1) << 3;
1511 		return off;
1512 
1513 	case IPPROTO_NONE:
1514 	case IPPROTO_ESP:
1515 	case IPPROTO_IPCOMP:
1516 		/* give up */
1517 		return -1;
1518 
1519 	default:
1520 		return -1;
1521 	}
1522 
1523 	return -1;
1524 }
1525 
1526 /*
1527  * get offset for the last header in the chain.  m will be kept untainted.
1528  */
1529 int
1530 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1531 {
1532 	int newoff;
1533 	int nxt;
1534 
1535 	if (!nxtp) {
1536 		nxt = -1;
1537 		nxtp = &nxt;
1538 	}
1539 	while (1) {
1540 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1541 		if (newoff < 0)
1542 			return off;
1543 		else if (newoff < off)
1544 			return -1;	/* invalid */
1545 		else if (newoff == off)
1546 			return newoff;
1547 
1548 		off = newoff;
1549 		proto = *nxtp;
1550 	}
1551 }
1552 
1553 struct ip6aux *
1554 ip6_addaux(struct mbuf *m)
1555 {
1556 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1557 	if (!tag) {
1558 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1559 				sizeof (struct ip6aux),
1560 				MB_DONTWAIT);
1561 		if (tag)
1562 			m_tag_prepend(m, tag);
1563 	}
1564 	if (tag)
1565 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1566 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1567 }
1568 
1569 struct ip6aux *
1570 ip6_findaux(struct mbuf *m)
1571 {
1572 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1573 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1574 }
1575 
1576 void
1577 ip6_delaux(struct mbuf *m)
1578 {
1579 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1580 	if (tag)
1581 		m_tag_delete(m, tag);
1582 }
1583 
1584 /*
1585  * System control for IP6
1586  */
1587 
1588 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1589 	0,		0,		0,		0,
1590 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1591 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1592 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1593 	0,		0,		0,		0,
1594 	ENOPROTOOPT
1595 };
1596