xref: /dflybsd-src/sys/netinet6/ip6_input.c (revision d217426c317e15b34c1a61f020c85bd2b08f5c0d)
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 #include "opt_ipsec.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/kernel.h>
80 #include <sys/syslog.h>
81 #include <sys/proc.h>
82 #include <sys/priv.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_dl.h>
87 #include <net/route.h>
88 #include <net/netisr.h>
89 #include <net/pfil.h>
90 
91 #include <sys/thread2.h>
92 #include <sys/msgport2.h>
93 #include <net/netmsg2.h>
94 #include <net/netisr2.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_systm.h>
98 #ifdef INET
99 #include <netinet/ip.h>
100 #include <netinet/ip_icmp.h>
101 #endif /* INET */
102 #include <netinet/ip6.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 
111 #ifdef IPSEC
112 #include <netinet6/ipsec.h>
113 #ifdef INET6
114 #include <netinet6/ipsec6.h>
115 #endif
116 #endif
117 
118 #ifdef FAST_IPSEC
119 #include <netproto/ipsec/ipsec.h>
120 #include <netproto/ipsec/ipsec6.h>
121 #define	IPSEC
122 #endif /* FAST_IPSEC */
123 
124 #include <net/ip6fw/ip6_fw.h>
125 
126 #include <netinet6/ip6protosw.h>
127 
128 #include <net/net_osdep.h>
129 
130 extern struct domain inet6domain;
131 extern struct protosw inet6sw[];
132 
133 u_char ip6_protox[IPPROTO_MAX];
134 struct in6_ifaddr *in6_ifaddr;
135 
136 extern struct callout in6_tmpaddrtimer_ch;
137 
138 int ip6_forward_srcrt;			/* XXX */
139 int ip6_sourcecheck;			/* XXX */
140 int ip6_sourcecheck_interval;		/* XXX */
141 
142 int ip6_ours_check_algorithm;
143 
144 struct pfil_head inet6_pfil_hook;
145 
146 /* firewall hooks */
147 ip6_fw_chk_t *ip6_fw_chk_ptr;
148 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
149 int ip6_fw_enable = 1;
150 
151 struct ip6stat ip6stat;
152 
153 static void ip6_init2 (void *);
154 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
155 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
156 static void ip6_input(netmsg_t msg);
157 #ifdef PULLDOWN_TEST
158 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
159 #endif
160 static void transport6_processing_handler(netmsg_t netmsg);
161 
162 /*
163  * IP6 initialization: fill in IP6 protocol switch table.
164  * All protocols not implemented in kernel go to raw IP6 protocol handler.
165  */
166 void
167 ip6_init(void)
168 {
169 	struct protosw *pr;
170 	int i;
171 	struct timeval tv;
172 
173 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
174 	if (pr == NULL)
175 		panic("ip6_init");
176 	for (i = 0; i < IPPROTO_MAX; i++)
177 		ip6_protox[i] = pr - inet6sw;
178 	for (pr = inet6domain.dom_protosw;
179 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
180 		if (pr->pr_domain->dom_family == PF_INET6 &&
181 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
182 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
183 
184 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
185 	inet6_pfil_hook.ph_af = AF_INET6;
186 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
187 		kprintf("%s: WARNING: unable to register pfil hook, "
188 			"error %d\n", __func__, i);
189 	}
190 
191 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX hashfn */
192 	scope6_init();
193 	addrsel_policy_init();
194 	nd6_init();
195 	frag6_init();
196 	/*
197 	 * in many cases, random() here does NOT return random number
198 	 * as initialization during bootstrap time occur in fixed order.
199 	 */
200 	microtime(&tv);
201 	ip6_flow_seq = krandom() ^ tv.tv_usec;
202 	microtime(&tv);
203 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
204 }
205 
206 static void
207 ip6_init2(void *dummy)
208 {
209 	/* nd6_timer_init */
210 	callout_init(&nd6_timer_ch);
211 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
212 
213 	/* timer for regeneranation of temporary addresses randomize ID */
214 	callout_init(&in6_tmpaddrtimer_ch);
215 	callout_reset(&in6_tmpaddrtimer_ch,
216 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
217 		       ip6_temp_regen_advance) * hz,
218 		      in6_tmpaddrtimer, NULL);
219 }
220 
221 /* cheat */
222 /* This must be after route_init(), which is now SI_ORDER_THIRD */
223 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
224 
225 extern struct	route_in6 ip6_forward_rt;
226 
227 static
228 void
229 ip6_input(netmsg_t msg)
230 {
231 	struct mbuf *m = msg->packet.nm_packet;
232 	struct ip6_hdr *ip6;
233 	int off = sizeof(struct ip6_hdr), nest;
234 	u_int32_t plen;
235 	u_int32_t rtalert = ~0;
236 	int nxt, ours = 0, rh_present = 0;
237 	struct ifnet *deliverifp = NULL;
238 	struct in6_addr odst;
239 	int srcrt = 0;
240 
241 #ifdef IPSEC
242 	/*
243 	 * should the inner packet be considered authentic?
244 	 * see comment in ah4_input().
245 	 */
246 	if (m) {
247 		m->m_flags &= ~M_AUTHIPHDR;
248 		m->m_flags &= ~M_AUTHIPDGM;
249 	}
250 #endif
251 
252 	/*
253 	 * make sure we don't have onion peering information into m_aux.
254 	 */
255 	ip6_delaux(m);
256 
257 	/*
258 	 * mbuf statistics
259 	 */
260 	if (m->m_flags & M_EXT) {
261 		if (m->m_next)
262 			ip6stat.ip6s_mext2m++;
263 		else
264 			ip6stat.ip6s_mext1++;
265 	} else {
266 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
267 		if (m->m_next) {
268 			if (m->m_flags & M_LOOP) {
269 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
270 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
271 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
272 			else
273 				ip6stat.ip6s_m2m[0]++;
274 		} else
275 			ip6stat.ip6s_m1++;
276 #undef M2MMAX
277 	}
278 
279 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
280 	ip6stat.ip6s_total++;
281 
282 #ifndef PULLDOWN_TEST
283 	/*
284 	 * L2 bridge code and some other code can return mbuf chain
285 	 * that does not conform to KAME requirement.  too bad.
286 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
287 	 */
288 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
289 		struct mbuf *n;
290 
291 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
292 		if (n == NULL)
293 			goto bad;
294 		M_MOVE_PKTHDR(n, m);
295 
296 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
297 		n->m_len = n->m_pkthdr.len;
298 		m_freem(m);
299 		m = n;
300 	}
301 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
302 #endif
303 
304 	if (m->m_len < sizeof(struct ip6_hdr)) {
305 		struct ifnet *inifp;
306 		inifp = m->m_pkthdr.rcvif;
307 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
308 			ip6stat.ip6s_toosmall++;
309 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
310 			goto bad2;
311 		}
312 	}
313 
314 	ip6 = mtod(m, struct ip6_hdr *);
315 
316 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
317 		ip6stat.ip6s_badvers++;
318 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
319 		goto bad;
320 	}
321 
322 	/*
323 	 * Run through list of hooks for input packets.
324 	 *
325 	 * NB: Beware of the destination address changing
326 	 *     (e.g. by NAT rewriting). When this happens,
327 	 *     tell ip6_forward to do the right thing.
328 	 */
329 	if (pfil_has_hooks(&inet6_pfil_hook)) {
330 		odst = ip6->ip6_dst;
331 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
332 		    m->m_pkthdr.rcvif, PFIL_IN)) {
333 			goto bad2;
334 		}
335 		if (m == NULL)			/* consumed by filter */
336 			goto bad2;
337 		ip6 = mtod(m, struct ip6_hdr *);
338 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
339 	}
340 
341 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
342 
343 #ifdef ALTQ
344 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
345 		/* packet is dropped by traffic conditioner */
346 		return;
347 	}
348 #endif
349 
350 	/*
351 	 * Check with the firewall...
352 	 */
353 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
354 		u_short port = 0;
355 		/* If ipfw says divert, we have to just drop packet */
356 		/* use port as a dummy argument */
357 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
358 			m_freem(m);
359 			m = NULL;
360 		}
361 		if (!m)
362 			goto bad2;
363 	}
364 
365 	/*
366 	 * Check against address spoofing/corruption.
367 	 */
368 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
369 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
370 		/*
371 		 * XXX: "badscope" is not very suitable for a multicast source.
372 		 */
373 		ip6stat.ip6s_badscope++;
374 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
375 		goto bad;
376 	}
377 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
378 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
379 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
380 		ip6stat.ip6s_badscope++;
381 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
382 		goto bad;
383 	}
384 
385 	/*
386 	 * The following check is not documented in specs.  A malicious
387 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
388 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
389 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
390 	 *
391 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
392 	 * support IPv4-less kernel compilation, we cannot support SIIT
393 	 * environment at all.  So, it makes more sense for us to reject any
394 	 * malicious packets for non-SIIT environment, than try to do a
395 	 * partical support for SIIT environment.
396 	 */
397 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
398 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
399 		ip6stat.ip6s_badscope++;
400 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
401 		goto bad;
402 	}
403 #if 0
404 	/*
405 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
406 	 *
407 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
408 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
409 	 * is revised to forbid relaying case.
410 	 */
411 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
412 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
413 		ip6stat.ip6s_badscope++;
414 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
415 		goto bad;
416 	}
417 #endif
418 
419 	/* drop packets if interface ID portion is already filled */
420 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
421 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
422 		    ip6->ip6_src.s6_addr16[1]) {
423 			ip6stat.ip6s_badscope++;
424 			goto bad;
425 		}
426 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
427 		    ip6->ip6_dst.s6_addr16[1]) {
428 			ip6stat.ip6s_badscope++;
429 			goto bad;
430 		}
431 	}
432 
433 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
434 		ip6->ip6_src.s6_addr16[1]
435 			= htons(m->m_pkthdr.rcvif->if_index);
436 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
437 		ip6->ip6_dst.s6_addr16[1]
438 			= htons(m->m_pkthdr.rcvif->if_index);
439 
440 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
441 	/*
442 	 * We use rt->rt_ifp to determine if the address is ours or not.
443 	 * If rt_ifp is lo0, the address is ours.
444 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
445 	 * so any address under fe80::%lo0/64 will be mistakenly considered
446 	 * local.  The special case is supplied to handle the case properly
447 	 * by actually looking at interface addresses
448 	 * (using in6ifa_ifpwithaddr).
449 	 */
450 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
451 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
452 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
453 			icmp6_error(m, ICMP6_DST_UNREACH,
454 			    ICMP6_DST_UNREACH_ADDR, 0);
455 			/* m is already freed */
456 			goto bad2;
457 		}
458 
459 		ours = 1;
460 		deliverifp = m->m_pkthdr.rcvif;
461 		goto hbhcheck;
462 	}
463 #endif
464 
465 	/*
466 	 * Multicast check
467 	 *
468 	 * WARNING!  For general subnet proxying the interface hw will
469 	 *	     likely filter out the multicast solicitations, so
470 	 *	     the interface must be in promiscuous mode.
471 	 */
472 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
473 		ours = 1;
474 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
475 		deliverifp = m->m_pkthdr.rcvif;
476 		goto hbhcheck;
477 	}
478 
479 	/*
480 	 *  Unicast check
481 	 */
482 	switch (ip6_ours_check_algorithm) {
483 	default:
484 		/*
485 		 * XXX: I intentionally broke our indentation rule here,
486 		 *      since this switch-case is just for measurement and
487 		 *      therefore should soon be removed.
488 		 */
489 	if (ip6_forward_rt.ro_rt != NULL &&
490 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
491 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
492 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
493 		ip6stat.ip6s_forward_cachehit++;
494 	else {
495 		struct sockaddr_in6 *dst6;
496 
497 		if (ip6_forward_rt.ro_rt) {
498 			/* route is down or destination is different */
499 			ip6stat.ip6s_forward_cachemiss++;
500 			RTFREE(ip6_forward_rt.ro_rt);
501 			ip6_forward_rt.ro_rt = 0;
502 		}
503 
504 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
505 		dst6 = &ip6_forward_rt.ro_dst;
506 		dst6->sin6_len = sizeof(struct sockaddr_in6);
507 		dst6->sin6_family = AF_INET6;
508 		dst6->sin6_addr = ip6->ip6_dst;
509 
510 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
511 	}
512 
513 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
514 
515 	/*
516 	 * Accept the packet if the forwarding interface to the destination
517 	 * according to the routing table is the loopback interface,
518 	 * unless the associated route has a gateway.
519 	 * Note that this approach causes to accept a packet if there is a
520 	 * route to the loopback interface for the destination of the packet.
521 	 * But we think it's even useful in some situations, e.g. when using
522 	 * a special daemon which wants to intercept the packet.
523 	 *
524 	 * XXX: some OSes automatically make a cloned route for the destination
525 	 * of an outgoing packet.  If the outgoing interface of the packet
526 	 * is a loopback one, the kernel would consider the packet to be
527 	 * accepted, even if we have no such address assinged on the interface.
528 	 * We check the cloned flag of the route entry to reject such cases,
529 	 * assuming that route entries for our own addresses are not made by
530 	 * cloning (it should be true because in6_addloop explicitly installs
531 	 * the host route).  However, we might have to do an explicit check
532 	 * while it would be less efficient.  Or, should we rather install a
533 	 * reject route for such a case?
534 	 */
535 	if (ip6_forward_rt.ro_rt &&
536 	    (ip6_forward_rt.ro_rt->rt_flags &
537 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
538 #ifdef RTF_WASCLONED
539 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
540 #endif
541 #ifdef RTF_CLONED
542 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
543 #endif
544 #if 0
545 	    /*
546 	     * The check below is redundant since the comparison of
547 	     * the destination and the key of the rtentry has
548 	     * already done through looking up the routing table.
549 	     */
550 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
551 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
552 #endif
553 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
554 		struct in6_ifaddr *ia6 =
555 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
556 
557 		/*
558 		 * record address information into m_aux.
559 		 */
560 		ip6_setdstifaddr(m, ia6);
561 
562 		/*
563 		 * packets to a tentative, duplicated, or somehow invalid
564 		 * address must not be accepted.
565 		 */
566 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
567 			/* this address is ready */
568 			ours = 1;
569 			deliverifp = ia6->ia_ifp;	/* correct? */
570 			/* Count the packet in the ip address stats */
571 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
572 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
573 			goto hbhcheck;
574 		} else {
575 			/* address is not ready, so discard the packet. */
576 			nd6log((LOG_INFO,
577 			    "ip6_input: packet to an unready address %s->%s\n",
578 			    ip6_sprintf(&ip6->ip6_src),
579 			    ip6_sprintf(&ip6->ip6_dst)));
580 
581 			goto bad;
582 		}
583 	}
584 	} /* XXX indentation (see above) */
585 
586 	/*
587 	 * FAITH(Firewall Aided Internet Translator)
588 	 */
589 	if (ip6_keepfaith) {
590 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
591 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
592 			/* XXX do we need more sanity checks? */
593 			ours = 1;
594 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
595 			goto hbhcheck;
596 		}
597 	}
598 
599 	/*
600 	 * Now there is no reason to process the packet if it's not our own
601 	 * and we're not a router.
602 	 */
603 	if (!ip6_forwarding) {
604 		ip6stat.ip6s_cantforward++;
605 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
606 		goto bad;
607 	}
608 
609 hbhcheck:
610 	/*
611 	 * record address information into m_aux, if we don't have one yet.
612 	 * note that we are unable to record it, if the address is not listed
613 	 * as our interface address (e.g. multicast addresses, addresses
614 	 * within FAITH prefixes and such).
615 	 */
616 	if (deliverifp && !ip6_getdstifaddr(m)) {
617 		struct in6_ifaddr *ia6;
618 
619 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
620 		if (ia6) {
621 			if (!ip6_setdstifaddr(m, ia6)) {
622 				/*
623 				 * XXX maybe we should drop the packet here,
624 				 * as we could not provide enough information
625 				 * to the upper layers.
626 				 */
627 			}
628 		}
629 	}
630 
631 	/*
632 	 * Process Hop-by-Hop options header if it's contained.
633 	 * m may be modified in ip6_hopopts_input().
634 	 * If a JumboPayload option is included, plen will also be modified.
635 	 */
636 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
637 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
638 		struct ip6_hbh *hbh;
639 
640 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
641 #if 0	/*touches NULL pointer*/
642 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
643 #endif
644 			goto bad2;	/* m have already been freed */
645 		}
646 
647 		/* adjust pointer */
648 		ip6 = mtod(m, struct ip6_hdr *);
649 
650 		/*
651 		 * if the payload length field is 0 and the next header field
652 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
653 		 * option MUST be included.
654 		 */
655 		if (ip6->ip6_plen == 0 && plen == 0) {
656 			/*
657 			 * Note that if a valid jumbo payload option is
658 			 * contained, ip6_hoptops_input() must set a valid
659 			 * (non-zero) payload length to the variable plen.
660 			 */
661 			ip6stat.ip6s_badoptions++;
662 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
663 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
664 			icmp6_error(m, ICMP6_PARAM_PROB,
665 				    ICMP6_PARAMPROB_HEADER,
666 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
667 			goto bad2;
668 		}
669 #ifndef PULLDOWN_TEST
670 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
671 		hbh = (struct ip6_hbh *)(ip6 + 1);
672 #else
673 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
674 			sizeof(struct ip6_hbh));
675 		if (hbh == NULL) {
676 			ip6stat.ip6s_tooshort++;
677 			goto bad2;
678 		}
679 #endif
680 		nxt = hbh->ip6h_nxt;
681 
682 		/*
683 		 * If we are acting as a router and the packet contains a
684 		 * router alert option, see if we know the option value.
685 		 * Currently, we only support the option value for MLD, in which
686 		 * case we should pass the packet to the multicast routing
687 		 * daemon.
688 		 */
689 		if (rtalert != ~0 && ip6_forwarding) {
690 			switch (rtalert) {
691 			case IP6OPT_RTALERT_MLD:
692 				ours = 1;
693 				break;
694 			default:
695 				/*
696 				 * RFC2711 requires unrecognized values must be
697 				 * silently ignored.
698 				 */
699 				break;
700 			}
701 		}
702 	} else
703 		nxt = ip6->ip6_nxt;
704 
705 	/*
706 	 * Check that the amount of data in the buffers
707 	 * is as at least much as the IPv6 header would have us expect.
708 	 * Trim mbufs if longer than we expect.
709 	 * Drop packet if shorter than we expect.
710 	 */
711 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
712 		ip6stat.ip6s_tooshort++;
713 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
714 		goto bad;
715 	}
716 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
717 		if (m->m_len == m->m_pkthdr.len) {
718 			m->m_len = sizeof(struct ip6_hdr) + plen;
719 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
720 		} else
721 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
722 	}
723 
724 	/*
725 	 * Forward if desirable.
726 	 */
727 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
728 		/*
729 		 * If we are acting as a multicast router, all
730 		 * incoming multicast packets are passed to the
731 		 * kernel-level multicast forwarding function.
732 		 * The packet is returned (relatively) intact; if
733 		 * ip6_mforward() returns a non-zero value, the packet
734 		 * must be discarded, else it may be accepted below.
735 		 */
736 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
737 			ip6stat.ip6s_cantforward++;
738 			goto bad;
739 		}
740 		if (!ours)
741 			goto bad;
742 	} else if (!ours) {
743 		ip6_forward(m, srcrt);
744 		goto bad2;
745 	}
746 
747 	ip6 = mtod(m, struct ip6_hdr *);
748 
749 	/*
750 	 * Malicious party may be able to use IPv4 mapped addr to confuse
751 	 * tcp/udp stack and bypass security checks (act as if it was from
752 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
753 	 *
754 	 * For SIIT end node behavior, you may want to disable the check.
755 	 * However, you will  become vulnerable to attacks using IPv4 mapped
756 	 * source.
757 	 */
758 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
759 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
760 		ip6stat.ip6s_badscope++;
761 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
762 		goto bad;
763 	}
764 
765 	/*
766 	 * Tell launch routine the next header
767 	 */
768 	ip6stat.ip6s_delivered++;
769 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
770 	nest = 0;
771 
772 	rh_present = 0;
773 	while (nxt != IPPROTO_DONE) {
774 		struct protosw *sw6;
775 
776 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
777 			ip6stat.ip6s_toomanyhdr++;
778 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
779 			goto bad;
780 		}
781 
782 		/*
783 		 * protection against faulty packet - there should be
784 		 * more sanity checks in header chain processing.
785 		 */
786 		if (m->m_pkthdr.len < off) {
787 			ip6stat.ip6s_tooshort++;
788 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
789 			goto bad;
790 		}
791 
792 #if 0
793 		/*
794 		 * do we need to do it for every header?  yeah, other
795 		 * functions can play with it (like re-allocate and copy).
796 		 */
797 		mhist = ip6_addaux(m);
798 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
799 			hist = mtod(mhist, caddr_t) + mhist->m_len;
800 			bcopy(&nxt, hist, sizeof(nxt));
801 			mhist->m_len += sizeof(nxt);
802 		} else {
803 			ip6stat.ip6s_toomanyhdr++;
804 			goto bad;
805 		}
806 #endif
807 
808 		if (nxt == IPPROTO_ROUTING) {
809 			if (rh_present++) {
810 				in6_ifstat_inc(m->m_pkthdr.rcvif,
811 				    ifs6_in_hdrerr);
812 				ip6stat.ip6s_badoptions++;
813 				goto bad;
814 			}
815 		}
816 
817 		sw6 = &inet6sw[ip6_protox[nxt]];
818 #ifdef IPSEC
819 		/*
820 		 * enforce IPsec policy checking if we are seeing last header.
821 		 * note that we do not visit this with protocols with pcb layer
822 		 * code - like udp/tcp/raw ip.
823 		 */
824 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
825 			ipsec6stat.in_polvio++;
826 			goto bad;
827 		}
828 #endif
829 		/*
830 		 * If this is a terminal header forward to the port, otherwise
831 		 * process synchronously for more headers.
832 		 */
833 		if (sw6->pr_flags & PR_LASTHDR) {
834 			struct netmsg_packet *pmsg;
835 			lwkt_port_t port;
836 
837 			port = netisr_cpuport(0); /* XXX */
838 			KKASSERT(port != NULL);
839 			pmsg = &m->m_hdr.mh_netmsg;
840 			netmsg_init(&pmsg->base, NULL,
841 				    &netisr_apanic_rport,
842 				    0, transport6_processing_handler);
843 			pmsg->nm_packet = m;
844 			pmsg->nm_nxt = nxt;
845 			pmsg->base.lmsg.u.ms_result = off;
846 			lwkt_sendmsg(port, &pmsg->base.lmsg);
847 			/* done with m */
848 			nxt = IPPROTO_DONE;
849 		} else {
850 			nxt = sw6->pr_input(&m, &off, nxt);
851 		}
852 	}
853 	goto bad2;
854 bad:
855 	m_freem(m);
856 bad2:
857 	;
858 	/* msg was embedded in the mbuf, do not reply! */
859 }
860 
861 /*
862  * We have to call the pr_input() function from the correct protocol
863  * thread.  The sw6->pr_soport() request at the end of ip6_input()
864  * returns the port and we forward a netmsg to the port to execute
865  * this function.
866  */
867 static void
868 transport6_processing_handler(netmsg_t netmsg)
869 {
870 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
871 	struct protosw *sw6;
872 	int hlen;
873 	int nxt;
874 
875 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
876 	hlen = pmsg->base.lmsg.u.ms_result;
877 
878 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
879 	KKASSERT(nxt == IPPROTO_DONE);
880 	/* netmsg was embedded in the mbuf, do not reply! */
881 }
882 
883 /*
884  * set/grab in6_ifaddr correspond to IPv6 destination address.
885  * XXX backward compatibility wrapper
886  */
887 static struct ip6aux *
888 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
889 {
890 	struct ip6aux *n;
891 
892 	n = ip6_addaux(m);
893 	if (n)
894 		n->ip6a_dstia6 = ia6;
895 	return n;	/* NULL if failed to set */
896 }
897 
898 struct in6_ifaddr *
899 ip6_getdstifaddr(struct mbuf *m)
900 {
901 	struct ip6aux *n;
902 
903 	n = ip6_findaux(m);
904 	if (n)
905 		return n->ip6a_dstia6;
906 	else
907 		return NULL;
908 }
909 
910 /*
911  * Hop-by-Hop options header processing. If a valid jumbo payload option is
912  * included, the real payload length will be stored in plenp.
913  */
914 static int
915 ip6_hopopts_input(u_int32_t *plenp,
916 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
917 		  struct mbuf **mp,
918 		  int *offp)
919 {
920 	struct mbuf *m = *mp;
921 	int off = *offp, hbhlen;
922 	struct ip6_hbh *hbh;
923 
924 	/* validation of the length of the header */
925 #ifndef PULLDOWN_TEST
926 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
927 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
928 	hbhlen = (hbh->ip6h_len + 1) << 3;
929 
930 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
931 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
932 #else
933 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
934 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
935 	if (hbh == NULL) {
936 		ip6stat.ip6s_tooshort++;
937 		return -1;
938 	}
939 	hbhlen = (hbh->ip6h_len + 1) << 3;
940 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
941 		hbhlen);
942 	if (hbh == NULL) {
943 		ip6stat.ip6s_tooshort++;
944 		return -1;
945 	}
946 #endif
947 	off += hbhlen;
948 	hbhlen -= sizeof(struct ip6_hbh);
949 
950 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
951 				hbhlen, rtalertp, plenp) < 0)
952 		return (-1);
953 
954 	*offp = off;
955 	*mp = m;
956 	return (0);
957 }
958 
959 /*
960  * Search header for all Hop-by-hop options and process each option.
961  * This function is separate from ip6_hopopts_input() in order to
962  * handle a case where the sending node itself process its hop-by-hop
963  * options header. In such a case, the function is called from ip6_output().
964  *
965  * The function assumes that hbh header is located right after the IPv6 header
966  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
967  * opthead + hbhlen is located in continuous memory region.
968  */
969 int
970 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
971 		    u_int32_t *rtalertp, u_int32_t *plenp)
972 {
973 	struct ip6_hdr *ip6;
974 	int optlen = 0;
975 	u_int8_t *opt = opthead;
976 	u_int16_t rtalert_val;
977 	u_int32_t jumboplen;
978 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
979 
980 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
981 		switch (*opt) {
982 		case IP6OPT_PAD1:
983 			optlen = 1;
984 			break;
985 		case IP6OPT_PADN:
986 			if (hbhlen < IP6OPT_MINLEN) {
987 				ip6stat.ip6s_toosmall++;
988 				goto bad;
989 			}
990 			optlen = *(opt + 1) + 2;
991 			break;
992 		case IP6OPT_RTALERT:
993 			/* XXX may need check for alignment */
994 			if (hbhlen < IP6OPT_RTALERT_LEN) {
995 				ip6stat.ip6s_toosmall++;
996 				goto bad;
997 			}
998 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
999 				/* XXX stat */
1000 				icmp6_error(m, ICMP6_PARAM_PROB,
1001 					    ICMP6_PARAMPROB_HEADER,
1002 					    erroff + opt + 1 - opthead);
1003 				return (-1);
1004 			}
1005 			optlen = IP6OPT_RTALERT_LEN;
1006 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1007 			*rtalertp = ntohs(rtalert_val);
1008 			break;
1009 		case IP6OPT_JUMBO:
1010 			/* XXX may need check for alignment */
1011 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1012 				ip6stat.ip6s_toosmall++;
1013 				goto bad;
1014 			}
1015 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1016 				/* XXX stat */
1017 				icmp6_error(m, ICMP6_PARAM_PROB,
1018 					    ICMP6_PARAMPROB_HEADER,
1019 					    erroff + opt + 1 - opthead);
1020 				return (-1);
1021 			}
1022 			optlen = IP6OPT_JUMBO_LEN;
1023 
1024 			/*
1025 			 * IPv6 packets that have non 0 payload length
1026 			 * must not contain a jumbo payload option.
1027 			 */
1028 			ip6 = mtod(m, struct ip6_hdr *);
1029 			if (ip6->ip6_plen) {
1030 				ip6stat.ip6s_badoptions++;
1031 				icmp6_error(m, ICMP6_PARAM_PROB,
1032 					    ICMP6_PARAMPROB_HEADER,
1033 					    erroff + opt - opthead);
1034 				return (-1);
1035 			}
1036 
1037 			/*
1038 			 * We may see jumbolen in unaligned location, so
1039 			 * we'd need to perform bcopy().
1040 			 */
1041 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1042 			jumboplen = (u_int32_t)htonl(jumboplen);
1043 
1044 #if 1
1045 			/*
1046 			 * if there are multiple jumbo payload options,
1047 			 * *plenp will be non-zero and the packet will be
1048 			 * rejected.
1049 			 * the behavior may need some debate in ipngwg -
1050 			 * multiple options does not make sense, however,
1051 			 * there's no explicit mention in specification.
1052 			 */
1053 			if (*plenp != 0) {
1054 				ip6stat.ip6s_badoptions++;
1055 				icmp6_error(m, ICMP6_PARAM_PROB,
1056 					    ICMP6_PARAMPROB_HEADER,
1057 					    erroff + opt + 2 - opthead);
1058 				return (-1);
1059 			}
1060 #endif
1061 
1062 			/*
1063 			 * jumbo payload length must be larger than 65535.
1064 			 */
1065 			if (jumboplen <= IPV6_MAXPACKET) {
1066 				ip6stat.ip6s_badoptions++;
1067 				icmp6_error(m, ICMP6_PARAM_PROB,
1068 					    ICMP6_PARAMPROB_HEADER,
1069 					    erroff + opt + 2 - opthead);
1070 				return (-1);
1071 			}
1072 			*plenp = jumboplen;
1073 
1074 			break;
1075 		default:		/* unknown option */
1076 			if (hbhlen < IP6OPT_MINLEN) {
1077 				ip6stat.ip6s_toosmall++;
1078 				goto bad;
1079 			}
1080 			optlen = ip6_unknown_opt(opt, m,
1081 			    erroff + opt - opthead);
1082 			if (optlen == -1)
1083 				return (-1);
1084 			optlen += 2;
1085 			break;
1086 		}
1087 	}
1088 
1089 	return (0);
1090 
1091 bad:
1092 	m_freem(m);
1093 	return (-1);
1094 }
1095 
1096 /*
1097  * Unknown option processing.
1098  * The third argument `off' is the offset from the IPv6 header to the option,
1099  * which is necessary if the IPv6 header the and option header and IPv6 header
1100  * is not continuous in order to return an ICMPv6 error.
1101  */
1102 int
1103 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1104 {
1105 	struct ip6_hdr *ip6;
1106 
1107 	switch (IP6OPT_TYPE(*optp)) {
1108 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1109 		return ((int)*(optp + 1));
1110 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1111 		m_freem(m);
1112 		return (-1);
1113 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1114 		ip6stat.ip6s_badoptions++;
1115 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1116 		return (-1);
1117 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1118 		ip6stat.ip6s_badoptions++;
1119 		ip6 = mtod(m, struct ip6_hdr *);
1120 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1121 		    (m->m_flags & (M_BCAST|M_MCAST)))
1122 			m_freem(m);
1123 		else
1124 			icmp6_error(m, ICMP6_PARAM_PROB,
1125 				    ICMP6_PARAMPROB_OPTION, off);
1126 		return (-1);
1127 	}
1128 
1129 	m_freem(m);		/* XXX: NOTREACHED */
1130 	return (-1);
1131 }
1132 
1133 /*
1134  * Create the "control" list for this pcb.
1135  * The function will not modify mbuf chain at all.
1136  *
1137  * with KAME mbuf chain restriction:
1138  * The routine will be called from upper layer handlers like tcp6_input().
1139  * Thus the routine assumes that the caller (tcp6_input) have already
1140  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1141  * very first mbuf on the mbuf chain.
1142  */
1143 void
1144 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1145 		struct mbuf *m)
1146 {
1147 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1148 	struct thread *td = curthread;	/* XXX */
1149 	int privileged = 0;
1150 
1151 
1152 	if (priv_check(td, PRIV_ROOT) == 0)
1153  		privileged++;
1154 
1155 #ifdef SO_TIMESTAMP
1156 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1157 		struct timeval tv;
1158 
1159 		microtime(&tv);
1160 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1161 				      SCM_TIMESTAMP, SOL_SOCKET);
1162 		if (*mp) {
1163 			mp = &(*mp)->m_next;
1164 		}
1165 	}
1166 #endif
1167 
1168 	/* RFC 2292 sec. 5 */
1169 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1170 		struct in6_pktinfo pi6;
1171 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1172 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1173 			pi6.ipi6_addr.s6_addr16[1] = 0;
1174 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1175 					? m->m_pkthdr.rcvif->if_index
1176 					: 0;
1177 		*mp = sbcreatecontrol((caddr_t) &pi6,
1178 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1179 			IPPROTO_IPV6);
1180 		if (*mp)
1181 			mp = &(*mp)->m_next;
1182 	}
1183 
1184 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1185 		int hlim = ip6->ip6_hlim & 0xff;
1186 		*mp = sbcreatecontrol((caddr_t) &hlim,
1187 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1188 		if (*mp)
1189 			mp = &(*mp)->m_next;
1190 	}
1191 
1192 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1193 		u_int32_t flowinfo;
1194 		int tclass;
1195 
1196 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1197 		flowinfo >>= 20;
1198 
1199 		tclass = flowinfo & 0xff;
1200 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1201 		    IPV6_TCLASS, IPPROTO_IPV6);
1202 		if (*mp)
1203 			mp = &(*mp)->m_next;
1204 	}
1205 
1206 	/*
1207 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1208 	 * for the option, but it might be too strict, since there might
1209 	 * be some hop-by-hop options which can be returned to normal user.
1210 	 * See RFC 2292 section 6.
1211 	 */
1212 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1213 		/*
1214 		 * Check if a hop-by-hop options header is contatined in the
1215 		 * received packet, and if so, store the options as ancillary
1216 		 * data. Note that a hop-by-hop options header must be
1217 		 * just after the IPv6 header, which fact is assured through
1218 		 * the IPv6 input processing.
1219 		 */
1220 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1221 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1222 			struct ip6_hbh *hbh;
1223 			int hbhlen = 0;
1224 #ifdef PULLDOWN_TEST
1225 			struct mbuf *ext;
1226 #endif
1227 
1228 #ifndef PULLDOWN_TEST
1229 			hbh = (struct ip6_hbh *)(ip6 + 1);
1230 			hbhlen = (hbh->ip6h_len + 1) << 3;
1231 #else
1232 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1233 			    ip6->ip6_nxt);
1234 			if (ext == NULL) {
1235 				ip6stat.ip6s_tooshort++;
1236 				return;
1237 			}
1238 			hbh = mtod(ext, struct ip6_hbh *);
1239 			hbhlen = (hbh->ip6h_len + 1) << 3;
1240 			if (hbhlen != ext->m_len) {
1241 				m_freem(ext);
1242 				ip6stat.ip6s_tooshort++;
1243 				return;
1244 			}
1245 #endif
1246 
1247 			/*
1248 			 * XXX: We copy whole the header even if a jumbo
1249 			 * payload option is included, which option is to
1250 			 * be removed before returning in the RFC 2292.
1251 			 * Note: this constraint is removed in 2292bis.
1252 			 */
1253 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1254 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1255 			if (*mp)
1256 				mp = &(*mp)->m_next;
1257 #ifdef PULLDOWN_TEST
1258 			m_freem(ext);
1259 #endif
1260 		}
1261 	}
1262 
1263 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1264 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1265 		int proto, off, nxt;
1266 
1267 		/*
1268 		 * go through the header chain to see if a routing header is
1269 		 * contained in the packet. We need this information to store
1270 		 * destination options headers (if any) properly.
1271 		 * XXX: performance issue. We should record this info when
1272 		 * processing extension headers in incoming routine.
1273 		 * (todo) use m_aux?
1274 		 */
1275 		proto = IPPROTO_IPV6;
1276 		off = 0;
1277 		nxt = -1;
1278 		while (1) {
1279 			int newoff;
1280 
1281 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1282 			if (newoff < 0)
1283 				break;
1284 			if (newoff < off) /* invalid, check for safety */
1285 				break;
1286 			if ((proto = nxt) == IPPROTO_ROUTING)
1287 				break;
1288 			off = newoff;
1289 		}
1290 	}
1291 
1292 	if ((in6p->in6p_flags &
1293 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1294 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1295 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1296 
1297 		/*
1298 		 * Search for destination options headers or routing
1299 		 * header(s) through the header chain, and stores each
1300 		 * header as ancillary data.
1301 		 * Note that the order of the headers remains in
1302 		 * the chain of ancillary data.
1303 		 */
1304 		while (1) {	/* is explicit loop prevention necessary? */
1305 			struct ip6_ext *ip6e = NULL;
1306 			int elen;
1307 #ifdef PULLDOWN_TEST
1308 			struct mbuf *ext = NULL;
1309 #endif
1310 
1311 			/*
1312 			 * if it is not an extension header, don't try to
1313 			 * pull it from the chain.
1314 			 */
1315 			switch (nxt) {
1316 			case IPPROTO_DSTOPTS:
1317 			case IPPROTO_ROUTING:
1318 			case IPPROTO_HOPOPTS:
1319 			case IPPROTO_AH: /* is it possible? */
1320 				break;
1321 			default:
1322 				goto loopend;
1323 			}
1324 
1325 #ifndef PULLDOWN_TEST
1326 			if (off + sizeof(*ip6e) > m->m_len)
1327 				goto loopend;
1328 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1329 			if (nxt == IPPROTO_AH)
1330 				elen = (ip6e->ip6e_len + 2) << 2;
1331 			else
1332 				elen = (ip6e->ip6e_len + 1) << 3;
1333 			if (off + elen > m->m_len)
1334 				goto loopend;
1335 #else
1336 			ext = ip6_pullexthdr(m, off, nxt);
1337 			if (ext == NULL) {
1338 				ip6stat.ip6s_tooshort++;
1339 				return;
1340 			}
1341 			ip6e = mtod(ext, struct ip6_ext *);
1342 			if (nxt == IPPROTO_AH)
1343 				elen = (ip6e->ip6e_len + 2) << 2;
1344 			else
1345 				elen = (ip6e->ip6e_len + 1) << 3;
1346 			if (elen != ext->m_len) {
1347 				m_freem(ext);
1348 				ip6stat.ip6s_tooshort++;
1349 				return;
1350 			}
1351 #endif
1352 
1353 			switch (nxt) {
1354 			case IPPROTO_DSTOPTS:
1355 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1356 					break;
1357 
1358 				/*
1359 				 * We also require super-user privilege for
1360 				 * the option.
1361 				 * See the comments on IN6_HOPOPTS.
1362 				 */
1363 				if (!privileged)
1364 					break;
1365 
1366 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1367 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1368 						      IPPROTO_IPV6);
1369 				if (*mp)
1370 					mp = &(*mp)->m_next;
1371 				break;
1372 			case IPPROTO_ROUTING:
1373 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1374 					break;
1375 
1376 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1377 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1378 						      IPPROTO_IPV6);
1379 				if (*mp)
1380 					mp = &(*mp)->m_next;
1381 				break;
1382 			case IPPROTO_HOPOPTS:
1383 			case IPPROTO_AH: /* is it possible? */
1384 				break;
1385 
1386 			default:
1387 				/*
1388 			 	 * other cases have been filtered in the above.
1389 				 * none will visit this case.  here we supply
1390 				 * the code just in case (nxt overwritten or
1391 				 * other cases).
1392 				 */
1393 #ifdef PULLDOWN_TEST
1394 				m_freem(ext);
1395 #endif
1396 				goto loopend;
1397 
1398 			}
1399 
1400 			/* proceed with the next header. */
1401 			off += elen;
1402 			nxt = ip6e->ip6e_nxt;
1403 			ip6e = NULL;
1404 #ifdef PULLDOWN_TEST
1405 			m_freem(ext);
1406 			ext = NULL;
1407 #endif
1408 		}
1409 	  loopend:
1410 		;
1411 	}
1412 #undef IS2292
1413 }
1414 
1415 void
1416 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1417 {
1418 	struct socket *so;
1419 	struct mbuf *m_mtu;
1420 	struct ip6_mtuinfo mtuctl;
1421 
1422 	so =  in6p->inp_socket;
1423 
1424 	if (mtu == NULL)
1425 		return;
1426 
1427 #ifdef DIAGNOSTIC
1428 	if (so == NULL)		/* I believe this is impossible */
1429 		panic("ip6_notify_pmtu: socket is NULL");
1430 #endif
1431 
1432 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1433 	mtuctl.ip6m_mtu = *mtu;
1434 	mtuctl.ip6m_addr = *dst;
1435 
1436 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1437 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1438 		return;
1439 
1440 	lwkt_gettoken(&so->so_rcv.ssb_token);
1441 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1442 	    == 0) {
1443 		m_freem(m_mtu);
1444 		/* XXX: should count statistics */
1445 	} else {
1446 		sorwakeup(so);
1447 	}
1448 	lwkt_reltoken(&so->so_rcv.ssb_token);
1449 }
1450 
1451 #ifdef PULLDOWN_TEST
1452 /*
1453  * pull single extension header from mbuf chain.  returns single mbuf that
1454  * contains the result, or NULL on error.
1455  */
1456 static struct mbuf *
1457 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1458 {
1459 	struct ip6_ext ip6e;
1460 	size_t elen;
1461 	struct mbuf *n;
1462 
1463 #ifdef DIAGNOSTIC
1464 	switch (nxt) {
1465 	case IPPROTO_DSTOPTS:
1466 	case IPPROTO_ROUTING:
1467 	case IPPROTO_HOPOPTS:
1468 	case IPPROTO_AH: /* is it possible? */
1469 		break;
1470 	default:
1471 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1472 	}
1473 #endif
1474 
1475 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1476 	if (nxt == IPPROTO_AH)
1477 		elen = (ip6e.ip6e_len + 2) << 2;
1478 	else
1479 		elen = (ip6e.ip6e_len + 1) << 3;
1480 
1481 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1482 	if (n == NULL)
1483 		return NULL;
1484 	n->m_len = 0;
1485 
1486 	if (elen >= M_TRAILINGSPACE(n)) {
1487 		m_free(n);
1488 		return NULL;
1489 	}
1490 
1491 	m_copydata(m, off, elen, mtod(n, caddr_t));
1492 	n->m_len = elen;
1493 	return n;
1494 }
1495 #endif
1496 
1497 /*
1498  * Get pointer to the previous header followed by the header
1499  * currently processed.
1500  * XXX: This function supposes that
1501  *	M includes all headers,
1502  *	the next header field and the header length field of each header
1503  *	are valid, and
1504  *	the sum of each header length equals to OFF.
1505  * Because of these assumptions, this function must be called very
1506  * carefully. Moreover, it will not be used in the near future when
1507  * we develop `neater' mechanism to process extension headers.
1508  */
1509 char *
1510 ip6_get_prevhdr(struct mbuf *m, int off)
1511 {
1512 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1513 
1514 	if (off == sizeof(struct ip6_hdr))
1515 		return (&ip6->ip6_nxt);
1516 	else {
1517 		int len, nxt;
1518 		struct ip6_ext *ip6e = NULL;
1519 
1520 		nxt = ip6->ip6_nxt;
1521 		len = sizeof(struct ip6_hdr);
1522 		while (len < off) {
1523 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1524 
1525 			switch (nxt) {
1526 			case IPPROTO_FRAGMENT:
1527 				len += sizeof(struct ip6_frag);
1528 				break;
1529 			case IPPROTO_AH:
1530 				len += (ip6e->ip6e_len + 2) << 2;
1531 				break;
1532 			default:
1533 				len += (ip6e->ip6e_len + 1) << 3;
1534 				break;
1535 			}
1536 			nxt = ip6e->ip6e_nxt;
1537 		}
1538 		if (ip6e)
1539 			return (&ip6e->ip6e_nxt);
1540 		else
1541 			return NULL;
1542 	}
1543 }
1544 
1545 /*
1546  * get next header offset.  m will be retained.
1547  */
1548 int
1549 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1550 {
1551 	struct ip6_hdr ip6;
1552 	struct ip6_ext ip6e;
1553 	struct ip6_frag fh;
1554 
1555 	/* just in case */
1556 	if (m == NULL)
1557 		panic("ip6_nexthdr: m == NULL");
1558 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1559 		return -1;
1560 
1561 	switch (proto) {
1562 	case IPPROTO_IPV6:
1563 		if (m->m_pkthdr.len < off + sizeof(ip6))
1564 			return -1;
1565 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1566 		if (nxtp)
1567 			*nxtp = ip6.ip6_nxt;
1568 		off += sizeof(ip6);
1569 		return off;
1570 
1571 	case IPPROTO_FRAGMENT:
1572 		/*
1573 		 * terminate parsing if it is not the first fragment,
1574 		 * it does not make sense to parse through it.
1575 		 */
1576 		if (m->m_pkthdr.len < off + sizeof(fh))
1577 			return -1;
1578 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1579 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1580 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1581 			return -1;
1582 		if (nxtp)
1583 			*nxtp = fh.ip6f_nxt;
1584 		off += sizeof(struct ip6_frag);
1585 		return off;
1586 
1587 	case IPPROTO_AH:
1588 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1589 			return -1;
1590 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1591 		if (nxtp)
1592 			*nxtp = ip6e.ip6e_nxt;
1593 		off += (ip6e.ip6e_len + 2) << 2;
1594 		return off;
1595 
1596 	case IPPROTO_HOPOPTS:
1597 	case IPPROTO_ROUTING:
1598 	case IPPROTO_DSTOPTS:
1599 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1600 			return -1;
1601 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1602 		if (nxtp)
1603 			*nxtp = ip6e.ip6e_nxt;
1604 		off += (ip6e.ip6e_len + 1) << 3;
1605 		return off;
1606 
1607 	case IPPROTO_NONE:
1608 	case IPPROTO_ESP:
1609 	case IPPROTO_IPCOMP:
1610 		/* give up */
1611 		return -1;
1612 
1613 	default:
1614 		return -1;
1615 	}
1616 
1617 	return -1;
1618 }
1619 
1620 /*
1621  * get offset for the last header in the chain.  m will be kept untainted.
1622  */
1623 int
1624 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1625 {
1626 	int newoff;
1627 	int nxt;
1628 
1629 	if (!nxtp) {
1630 		nxt = -1;
1631 		nxtp = &nxt;
1632 	}
1633 	while (1) {
1634 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1635 		if (newoff < 0)
1636 			return off;
1637 		else if (newoff < off)
1638 			return -1;	/* invalid */
1639 		else if (newoff == off)
1640 			return newoff;
1641 
1642 		off = newoff;
1643 		proto = *nxtp;
1644 	}
1645 }
1646 
1647 struct ip6aux *
1648 ip6_addaux(struct mbuf *m)
1649 {
1650 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1651 	if (!tag) {
1652 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1653 				sizeof (struct ip6aux),
1654 				MB_DONTWAIT);
1655 		if (tag)
1656 			m_tag_prepend(m, tag);
1657 	}
1658 	if (tag)
1659 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1660 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1661 }
1662 
1663 struct ip6aux *
1664 ip6_findaux(struct mbuf *m)
1665 {
1666 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1667 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1668 }
1669 
1670 void
1671 ip6_delaux(struct mbuf *m)
1672 {
1673 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1674 	if (tag)
1675 		m_tag_delete(m, tag);
1676 }
1677 
1678 /*
1679  * System control for IP6
1680  */
1681 
1682 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1683 	0,		0,		0,		0,
1684 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1685 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1686 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1687 	0,		0,		0,		0,
1688 	ENOPROTOOPT
1689 };
1690