xref: /dflybsd-src/sys/netinet6/ip6_input.c (revision 31c068aaf635ad9fa72dbc4c65b32d890ff7544d)
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 int ip6_forward_srcrt;			/* XXX */
137 int ip6_sourcecheck;			/* XXX */
138 int ip6_sourcecheck_interval;		/* XXX */
139 
140 int ip6_ours_check_algorithm;
141 
142 struct pfil_head inet6_pfil_hook;
143 
144 /* firewall hooks */
145 ip6_fw_chk_t *ip6_fw_chk_ptr;
146 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
147 int ip6_fw_enable = 1;
148 
149 struct ip6stat ip6stat;
150 
151 static void ip6_init2 (void *);
152 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
153 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
154 static void ip6_input(netmsg_t msg);
155 #ifdef PULLDOWN_TEST
156 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
157 #endif
158 static void transport6_processing_handler(netmsg_t netmsg);
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
165 ip6_init(void)
166 {
167 	struct protosw *pr;
168 	int i;
169 	struct timeval tv;
170 
171 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
172 	if (pr == NULL)
173 		panic("ip6_init");
174 	for (i = 0; i < IPPROTO_MAX; i++)
175 		ip6_protox[i] = pr - inet6sw;
176 	for (pr = inet6domain.dom_protosw;
177 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
178 		if (pr->pr_domain->dom_family == PF_INET6 &&
179 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
180 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
181 
182 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
183 	inet6_pfil_hook.ph_af = AF_INET6;
184 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
185 		kprintf("%s: WARNING: unable to register pfil hook, "
186 			"error %d\n", __func__, i);
187 	}
188 
189 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX hashfn */
190 	scope6_init();
191 	addrsel_policy_init();
192 	nd6_init();
193 	frag6_init();
194 	/*
195 	 * in many cases, random() here does NOT return random number
196 	 * as initialization during bootstrap time occur in fixed order.
197 	 */
198 	microtime(&tv);
199 	ip6_flow_seq = krandom() ^ tv.tv_usec;
200 	microtime(&tv);
201 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
202 }
203 
204 static void
205 ip6_init2(void *dummy)
206 {
207 	/* nd6_timer_init */
208 	callout_init(&nd6_timer_ch);
209 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
210 
211 	in6_tmpaddrtimer_init();
212 }
213 
214 /* cheat */
215 /* This must be after route_init(), which is now SI_ORDER_THIRD */
216 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
217 
218 extern struct	route_in6 ip6_forward_rt;
219 
220 static
221 void
222 ip6_input(netmsg_t msg)
223 {
224 	struct mbuf *m = msg->packet.nm_packet;
225 	struct ip6_hdr *ip6;
226 	int off = sizeof(struct ip6_hdr), nest;
227 	u_int32_t plen;
228 	u_int32_t rtalert = ~0;
229 	int nxt, ours = 0, rh_present = 0;
230 	struct ifnet *deliverifp = NULL;
231 	struct in6_addr odst;
232 	int srcrt = 0;
233 
234 #ifdef IPSEC
235 	/*
236 	 * should the inner packet be considered authentic?
237 	 * see comment in ah4_input().
238 	 */
239 	if (m) {
240 		m->m_flags &= ~M_AUTHIPHDR;
241 		m->m_flags &= ~M_AUTHIPDGM;
242 	}
243 #endif
244 
245 	/*
246 	 * make sure we don't have onion peering information into m_aux.
247 	 */
248 	ip6_delaux(m);
249 
250 	/*
251 	 * mbuf statistics
252 	 */
253 	if (m->m_flags & M_EXT) {
254 		if (m->m_next)
255 			ip6stat.ip6s_mext2m++;
256 		else
257 			ip6stat.ip6s_mext1++;
258 	} else {
259 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
260 		if (m->m_next) {
261 			if (m->m_flags & M_LOOP) {
262 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
263 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
264 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
265 			else
266 				ip6stat.ip6s_m2m[0]++;
267 		} else
268 			ip6stat.ip6s_m1++;
269 #undef M2MMAX
270 	}
271 
272 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
273 	ip6stat.ip6s_total++;
274 
275 #ifndef PULLDOWN_TEST
276 	/*
277 	 * L2 bridge code and some other code can return mbuf chain
278 	 * that does not conform to KAME requirement.  too bad.
279 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
280 	 */
281 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
282 		struct mbuf *n;
283 
284 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
285 		if (n == NULL)
286 			goto bad;
287 		M_MOVE_PKTHDR(n, m);
288 
289 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
290 		n->m_len = n->m_pkthdr.len;
291 		m_freem(m);
292 		m = n;
293 	}
294 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
295 #endif
296 
297 	if (m->m_len < sizeof(struct ip6_hdr)) {
298 		struct ifnet *inifp;
299 		inifp = m->m_pkthdr.rcvif;
300 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
301 			ip6stat.ip6s_toosmall++;
302 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
303 			goto bad2;
304 		}
305 	}
306 
307 	ip6 = mtod(m, struct ip6_hdr *);
308 
309 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
310 		ip6stat.ip6s_badvers++;
311 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
312 		goto bad;
313 	}
314 
315 	/*
316 	 * Run through list of hooks for input packets.
317 	 *
318 	 * NB: Beware of the destination address changing
319 	 *     (e.g. by NAT rewriting). When this happens,
320 	 *     tell ip6_forward to do the right thing.
321 	 */
322 	if (pfil_has_hooks(&inet6_pfil_hook)) {
323 		odst = ip6->ip6_dst;
324 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
325 		    m->m_pkthdr.rcvif, PFIL_IN)) {
326 			goto bad2;
327 		}
328 		if (m == NULL)			/* consumed by filter */
329 			goto bad2;
330 		ip6 = mtod(m, struct ip6_hdr *);
331 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
332 	}
333 
334 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
335 
336 #ifdef ALTQ
337 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
338 		/* packet is dropped by traffic conditioner */
339 		return;
340 	}
341 #endif
342 
343 	/*
344 	 * Check with the firewall...
345 	 */
346 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
347 		u_short port = 0;
348 		/* If ipfw says divert, we have to just drop packet */
349 		/* use port as a dummy argument */
350 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
351 			m_freem(m);
352 			m = NULL;
353 		}
354 		if (!m)
355 			goto bad2;
356 	}
357 
358 	/*
359 	 * Check against address spoofing/corruption.
360 	 */
361 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
362 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
363 		/*
364 		 * XXX: "badscope" is not very suitable for a multicast source.
365 		 */
366 		ip6stat.ip6s_badscope++;
367 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
368 		goto bad;
369 	}
370 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
371 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
372 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
373 		ip6stat.ip6s_badscope++;
374 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
375 		goto bad;
376 	}
377 
378 	/*
379 	 * The following check is not documented in specs.  A malicious
380 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
381 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
382 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
383 	 *
384 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
385 	 * support IPv4-less kernel compilation, we cannot support SIIT
386 	 * environment at all.  So, it makes more sense for us to reject any
387 	 * malicious packets for non-SIIT environment, than try to do a
388 	 * partical support for SIIT environment.
389 	 */
390 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
391 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
392 		ip6stat.ip6s_badscope++;
393 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
394 		goto bad;
395 	}
396 #if 0
397 	/*
398 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
399 	 *
400 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
401 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
402 	 * is revised to forbid relaying case.
403 	 */
404 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
405 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
406 		ip6stat.ip6s_badscope++;
407 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
408 		goto bad;
409 	}
410 #endif
411 
412 	/* drop packets if interface ID portion is already filled */
413 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
414 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
415 		    ip6->ip6_src.s6_addr16[1]) {
416 			ip6stat.ip6s_badscope++;
417 			goto bad;
418 		}
419 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
420 		    ip6->ip6_dst.s6_addr16[1]) {
421 			ip6stat.ip6s_badscope++;
422 			goto bad;
423 		}
424 	}
425 
426 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
427 		ip6->ip6_src.s6_addr16[1]
428 			= htons(m->m_pkthdr.rcvif->if_index);
429 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
430 		ip6->ip6_dst.s6_addr16[1]
431 			= htons(m->m_pkthdr.rcvif->if_index);
432 
433 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
434 	/*
435 	 * We use rt->rt_ifp to determine if the address is ours or not.
436 	 * If rt_ifp is lo0, the address is ours.
437 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
438 	 * so any address under fe80::%lo0/64 will be mistakenly considered
439 	 * local.  The special case is supplied to handle the case properly
440 	 * by actually looking at interface addresses
441 	 * (using in6ifa_ifpwithaddr).
442 	 */
443 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
444 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
445 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
446 			icmp6_error(m, ICMP6_DST_UNREACH,
447 			    ICMP6_DST_UNREACH_ADDR, 0);
448 			/* m is already freed */
449 			goto bad2;
450 		}
451 
452 		ours = 1;
453 		deliverifp = m->m_pkthdr.rcvif;
454 		goto hbhcheck;
455 	}
456 #endif
457 
458 	/*
459 	 * Multicast check
460 	 *
461 	 * WARNING!  For general subnet proxying the interface hw will
462 	 *	     likely filter out the multicast solicitations, so
463 	 *	     the interface must be in promiscuous mode.
464 	 */
465 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
466 		ours = 1;
467 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
468 		deliverifp = m->m_pkthdr.rcvif;
469 		goto hbhcheck;
470 	}
471 
472 	/*
473 	 *  Unicast check
474 	 */
475 	switch (ip6_ours_check_algorithm) {
476 	default:
477 		/*
478 		 * XXX: I intentionally broke our indentation rule here,
479 		 *      since this switch-case is just for measurement and
480 		 *      therefore should soon be removed.
481 		 */
482 	if (ip6_forward_rt.ro_rt != NULL &&
483 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
484 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
485 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
486 		ip6stat.ip6s_forward_cachehit++;
487 	else {
488 		struct sockaddr_in6 *dst6;
489 
490 		if (ip6_forward_rt.ro_rt) {
491 			/* route is down or destination is different */
492 			ip6stat.ip6s_forward_cachemiss++;
493 			RTFREE(ip6_forward_rt.ro_rt);
494 			ip6_forward_rt.ro_rt = 0;
495 		}
496 
497 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
498 		dst6 = &ip6_forward_rt.ro_dst;
499 		dst6->sin6_len = sizeof(struct sockaddr_in6);
500 		dst6->sin6_family = AF_INET6;
501 		dst6->sin6_addr = ip6->ip6_dst;
502 
503 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
504 	}
505 
506 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
507 
508 	/*
509 	 * Accept the packet if the forwarding interface to the destination
510 	 * according to the routing table is the loopback interface,
511 	 * unless the associated route has a gateway.
512 	 * Note that this approach causes to accept a packet if there is a
513 	 * route to the loopback interface for the destination of the packet.
514 	 * But we think it's even useful in some situations, e.g. when using
515 	 * a special daemon which wants to intercept the packet.
516 	 *
517 	 * XXX: some OSes automatically make a cloned route for the destination
518 	 * of an outgoing packet.  If the outgoing interface of the packet
519 	 * is a loopback one, the kernel would consider the packet to be
520 	 * accepted, even if we have no such address assinged on the interface.
521 	 * We check the cloned flag of the route entry to reject such cases,
522 	 * assuming that route entries for our own addresses are not made by
523 	 * cloning (it should be true because in6_addloop explicitly installs
524 	 * the host route).  However, we might have to do an explicit check
525 	 * while it would be less efficient.  Or, should we rather install a
526 	 * reject route for such a case?
527 	 */
528 	if (ip6_forward_rt.ro_rt &&
529 	    (ip6_forward_rt.ro_rt->rt_flags &
530 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
531 #ifdef RTF_WASCLONED
532 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
533 #endif
534 #ifdef RTF_CLONED
535 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
536 #endif
537 #if 0
538 	    /*
539 	     * The check below is redundant since the comparison of
540 	     * the destination and the key of the rtentry has
541 	     * already done through looking up the routing table.
542 	     */
543 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
544 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
545 #endif
546 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
547 		struct in6_ifaddr *ia6 =
548 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
549 
550 		/*
551 		 * record address information into m_aux.
552 		 */
553 		ip6_setdstifaddr(m, ia6);
554 
555 		/*
556 		 * packets to a tentative, duplicated, or somehow invalid
557 		 * address must not be accepted.
558 		 */
559 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
560 			/* this address is ready */
561 			ours = 1;
562 			deliverifp = ia6->ia_ifp;	/* correct? */
563 			/* Count the packet in the ip address stats */
564 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
565 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
566 			goto hbhcheck;
567 		} else {
568 			/* address is not ready, so discard the packet. */
569 			nd6log((LOG_INFO,
570 			    "ip6_input: packet to an unready address %s->%s\n",
571 			    ip6_sprintf(&ip6->ip6_src),
572 			    ip6_sprintf(&ip6->ip6_dst)));
573 
574 			goto bad;
575 		}
576 	}
577 	} /* XXX indentation (see above) */
578 
579 	/*
580 	 * FAITH(Firewall Aided Internet Translator)
581 	 */
582 	if (ip6_keepfaith) {
583 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
584 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
585 			/* XXX do we need more sanity checks? */
586 			ours = 1;
587 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
588 			goto hbhcheck;
589 		}
590 	}
591 
592 	/*
593 	 * Now there is no reason to process the packet if it's not our own
594 	 * and we're not a router.
595 	 */
596 	if (!ip6_forwarding) {
597 		ip6stat.ip6s_cantforward++;
598 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
599 		goto bad;
600 	}
601 
602 hbhcheck:
603 	/*
604 	 * record address information into m_aux, if we don't have one yet.
605 	 * note that we are unable to record it, if the address is not listed
606 	 * as our interface address (e.g. multicast addresses, addresses
607 	 * within FAITH prefixes and such).
608 	 */
609 	if (deliverifp && !ip6_getdstifaddr(m)) {
610 		struct in6_ifaddr *ia6;
611 
612 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
613 		if (ia6) {
614 			if (!ip6_setdstifaddr(m, ia6)) {
615 				/*
616 				 * XXX maybe we should drop the packet here,
617 				 * as we could not provide enough information
618 				 * to the upper layers.
619 				 */
620 			}
621 		}
622 	}
623 
624 	/*
625 	 * Process Hop-by-Hop options header if it's contained.
626 	 * m may be modified in ip6_hopopts_input().
627 	 * If a JumboPayload option is included, plen will also be modified.
628 	 */
629 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
630 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
631 		struct ip6_hbh *hbh;
632 
633 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
634 #if 0	/*touches NULL pointer*/
635 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
636 #endif
637 			goto bad2;	/* m have already been freed */
638 		}
639 
640 		/* adjust pointer */
641 		ip6 = mtod(m, struct ip6_hdr *);
642 
643 		/*
644 		 * if the payload length field is 0 and the next header field
645 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
646 		 * option MUST be included.
647 		 */
648 		if (ip6->ip6_plen == 0 && plen == 0) {
649 			/*
650 			 * Note that if a valid jumbo payload option is
651 			 * contained, ip6_hoptops_input() must set a valid
652 			 * (non-zero) payload length to the variable plen.
653 			 */
654 			ip6stat.ip6s_badoptions++;
655 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
656 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
657 			icmp6_error(m, ICMP6_PARAM_PROB,
658 				    ICMP6_PARAMPROB_HEADER,
659 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
660 			goto bad2;
661 		}
662 #ifndef PULLDOWN_TEST
663 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
664 		hbh = (struct ip6_hbh *)(ip6 + 1);
665 #else
666 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
667 			sizeof(struct ip6_hbh));
668 		if (hbh == NULL) {
669 			ip6stat.ip6s_tooshort++;
670 			goto bad2;
671 		}
672 #endif
673 		nxt = hbh->ip6h_nxt;
674 
675 		/*
676 		 * If we are acting as a router and the packet contains a
677 		 * router alert option, see if we know the option value.
678 		 * Currently, we only support the option value for MLD, in which
679 		 * case we should pass the packet to the multicast routing
680 		 * daemon.
681 		 */
682 		if (rtalert != ~0 && ip6_forwarding) {
683 			switch (rtalert) {
684 			case IP6OPT_RTALERT_MLD:
685 				ours = 1;
686 				break;
687 			default:
688 				/*
689 				 * RFC2711 requires unrecognized values must be
690 				 * silently ignored.
691 				 */
692 				break;
693 			}
694 		}
695 	} else
696 		nxt = ip6->ip6_nxt;
697 
698 	/*
699 	 * Check that the amount of data in the buffers
700 	 * is as at least much as the IPv6 header would have us expect.
701 	 * Trim mbufs if longer than we expect.
702 	 * Drop packet if shorter than we expect.
703 	 */
704 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
705 		ip6stat.ip6s_tooshort++;
706 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
707 		goto bad;
708 	}
709 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
710 		if (m->m_len == m->m_pkthdr.len) {
711 			m->m_len = sizeof(struct ip6_hdr) + plen;
712 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
713 		} else
714 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
715 	}
716 
717 	/*
718 	 * Forward if desirable.
719 	 */
720 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
721 		/*
722 		 * If we are acting as a multicast router, all
723 		 * incoming multicast packets are passed to the
724 		 * kernel-level multicast forwarding function.
725 		 * The packet is returned (relatively) intact; if
726 		 * ip6_mforward() returns a non-zero value, the packet
727 		 * must be discarded, else it may be accepted below.
728 		 */
729 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
730 			ip6stat.ip6s_cantforward++;
731 			goto bad;
732 		}
733 		if (!ours)
734 			goto bad;
735 	} else if (!ours) {
736 		ip6_forward(m, srcrt);
737 		goto bad2;
738 	}
739 
740 	ip6 = mtod(m, struct ip6_hdr *);
741 
742 	/*
743 	 * Malicious party may be able to use IPv4 mapped addr to confuse
744 	 * tcp/udp stack and bypass security checks (act as if it was from
745 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
746 	 *
747 	 * For SIIT end node behavior, you may want to disable the check.
748 	 * However, you will  become vulnerable to attacks using IPv4 mapped
749 	 * source.
750 	 */
751 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
752 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
753 		ip6stat.ip6s_badscope++;
754 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
755 		goto bad;
756 	}
757 
758 	/*
759 	 * Tell launch routine the next header
760 	 */
761 	ip6stat.ip6s_delivered++;
762 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
763 	nest = 0;
764 
765 	rh_present = 0;
766 	while (nxt != IPPROTO_DONE) {
767 		struct protosw *sw6;
768 
769 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
770 			ip6stat.ip6s_toomanyhdr++;
771 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
772 			goto bad;
773 		}
774 
775 		/*
776 		 * protection against faulty packet - there should be
777 		 * more sanity checks in header chain processing.
778 		 */
779 		if (m->m_pkthdr.len < off) {
780 			ip6stat.ip6s_tooshort++;
781 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
782 			goto bad;
783 		}
784 
785 #if 0
786 		/*
787 		 * do we need to do it for every header?  yeah, other
788 		 * functions can play with it (like re-allocate and copy).
789 		 */
790 		mhist = ip6_addaux(m);
791 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
792 			hist = mtod(mhist, caddr_t) + mhist->m_len;
793 			bcopy(&nxt, hist, sizeof(nxt));
794 			mhist->m_len += sizeof(nxt);
795 		} else {
796 			ip6stat.ip6s_toomanyhdr++;
797 			goto bad;
798 		}
799 #endif
800 
801 		if (nxt == IPPROTO_ROUTING) {
802 			if (rh_present++) {
803 				in6_ifstat_inc(m->m_pkthdr.rcvif,
804 				    ifs6_in_hdrerr);
805 				ip6stat.ip6s_badoptions++;
806 				goto bad;
807 			}
808 		}
809 
810 		sw6 = &inet6sw[ip6_protox[nxt]];
811 #ifdef IPSEC
812 		/*
813 		 * enforce IPsec policy checking if we are seeing last header.
814 		 * note that we do not visit this with protocols with pcb layer
815 		 * code - like udp/tcp/raw ip.
816 		 */
817 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
818 			ipsec6stat.in_polvio++;
819 			goto bad;
820 		}
821 #endif
822 		/*
823 		 * If this is a terminal header forward to the port, otherwise
824 		 * process synchronously for more headers.
825 		 */
826 		if (sw6->pr_flags & PR_LASTHDR) {
827 			struct netmsg_packet *pmsg;
828 			lwkt_port_t port;
829 
830 			port = netisr_cpuport(0); /* XXX */
831 			KKASSERT(port != NULL);
832 			pmsg = &m->m_hdr.mh_netmsg;
833 			netmsg_init(&pmsg->base, NULL,
834 				    &netisr_apanic_rport,
835 				    0, transport6_processing_handler);
836 			pmsg->nm_packet = m;
837 			pmsg->nm_nxt = nxt;
838 			pmsg->base.lmsg.u.ms_result = off;
839 			lwkt_sendmsg(port, &pmsg->base.lmsg);
840 			/* done with m */
841 			nxt = IPPROTO_DONE;
842 		} else {
843 			nxt = sw6->pr_input(&m, &off, nxt);
844 		}
845 	}
846 	goto bad2;
847 bad:
848 	m_freem(m);
849 bad2:
850 	;
851 	/* msg was embedded in the mbuf, do not reply! */
852 }
853 
854 /*
855  * We have to call the pr_input() function from the correct protocol
856  * thread.  The sw6->pr_soport() request at the end of ip6_input()
857  * returns the port and we forward a netmsg to the port to execute
858  * this function.
859  */
860 static void
861 transport6_processing_handler(netmsg_t netmsg)
862 {
863 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
864 	struct protosw *sw6;
865 	int hlen;
866 	int nxt;
867 
868 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
869 	hlen = pmsg->base.lmsg.u.ms_result;
870 
871 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
872 	KKASSERT(nxt == IPPROTO_DONE);
873 	/* netmsg was embedded in the mbuf, do not reply! */
874 }
875 
876 /*
877  * set/grab in6_ifaddr correspond to IPv6 destination address.
878  * XXX backward compatibility wrapper
879  */
880 static struct ip6aux *
881 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
882 {
883 	struct ip6aux *n;
884 
885 	n = ip6_addaux(m);
886 	if (n)
887 		n->ip6a_dstia6 = ia6;
888 	return n;	/* NULL if failed to set */
889 }
890 
891 struct in6_ifaddr *
892 ip6_getdstifaddr(struct mbuf *m)
893 {
894 	struct ip6aux *n;
895 
896 	n = ip6_findaux(m);
897 	if (n)
898 		return n->ip6a_dstia6;
899 	else
900 		return NULL;
901 }
902 
903 /*
904  * Hop-by-Hop options header processing. If a valid jumbo payload option is
905  * included, the real payload length will be stored in plenp.
906  */
907 static int
908 ip6_hopopts_input(u_int32_t *plenp,
909 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
910 		  struct mbuf **mp,
911 		  int *offp)
912 {
913 	struct mbuf *m = *mp;
914 	int off = *offp, hbhlen;
915 	struct ip6_hbh *hbh;
916 
917 	/* validation of the length of the header */
918 #ifndef PULLDOWN_TEST
919 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
920 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
921 	hbhlen = (hbh->ip6h_len + 1) << 3;
922 
923 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
924 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
925 #else
926 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
927 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
928 	if (hbh == NULL) {
929 		ip6stat.ip6s_tooshort++;
930 		return -1;
931 	}
932 	hbhlen = (hbh->ip6h_len + 1) << 3;
933 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
934 		hbhlen);
935 	if (hbh == NULL) {
936 		ip6stat.ip6s_tooshort++;
937 		return -1;
938 	}
939 #endif
940 	off += hbhlen;
941 	hbhlen -= sizeof(struct ip6_hbh);
942 
943 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
944 				hbhlen, rtalertp, plenp) < 0)
945 		return (-1);
946 
947 	*offp = off;
948 	*mp = m;
949 	return (0);
950 }
951 
952 /*
953  * Search header for all Hop-by-hop options and process each option.
954  * This function is separate from ip6_hopopts_input() in order to
955  * handle a case where the sending node itself process its hop-by-hop
956  * options header. In such a case, the function is called from ip6_output().
957  *
958  * The function assumes that hbh header is located right after the IPv6 header
959  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
960  * opthead + hbhlen is located in continuous memory region.
961  */
962 int
963 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
964 		    u_int32_t *rtalertp, u_int32_t *plenp)
965 {
966 	struct ip6_hdr *ip6;
967 	int optlen = 0;
968 	u_int8_t *opt = opthead;
969 	u_int16_t rtalert_val;
970 	u_int32_t jumboplen;
971 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
972 
973 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
974 		switch (*opt) {
975 		case IP6OPT_PAD1:
976 			optlen = 1;
977 			break;
978 		case IP6OPT_PADN:
979 			if (hbhlen < IP6OPT_MINLEN) {
980 				ip6stat.ip6s_toosmall++;
981 				goto bad;
982 			}
983 			optlen = *(opt + 1) + 2;
984 			break;
985 		case IP6OPT_RTALERT:
986 			/* XXX may need check for alignment */
987 			if (hbhlen < IP6OPT_RTALERT_LEN) {
988 				ip6stat.ip6s_toosmall++;
989 				goto bad;
990 			}
991 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
992 				/* XXX stat */
993 				icmp6_error(m, ICMP6_PARAM_PROB,
994 					    ICMP6_PARAMPROB_HEADER,
995 					    erroff + opt + 1 - opthead);
996 				return (-1);
997 			}
998 			optlen = IP6OPT_RTALERT_LEN;
999 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1000 			*rtalertp = ntohs(rtalert_val);
1001 			break;
1002 		case IP6OPT_JUMBO:
1003 			/* XXX may need check for alignment */
1004 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1005 				ip6stat.ip6s_toosmall++;
1006 				goto bad;
1007 			}
1008 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1009 				/* XXX stat */
1010 				icmp6_error(m, ICMP6_PARAM_PROB,
1011 					    ICMP6_PARAMPROB_HEADER,
1012 					    erroff + opt + 1 - opthead);
1013 				return (-1);
1014 			}
1015 			optlen = IP6OPT_JUMBO_LEN;
1016 
1017 			/*
1018 			 * IPv6 packets that have non 0 payload length
1019 			 * must not contain a jumbo payload option.
1020 			 */
1021 			ip6 = mtod(m, struct ip6_hdr *);
1022 			if (ip6->ip6_plen) {
1023 				ip6stat.ip6s_badoptions++;
1024 				icmp6_error(m, ICMP6_PARAM_PROB,
1025 					    ICMP6_PARAMPROB_HEADER,
1026 					    erroff + opt - opthead);
1027 				return (-1);
1028 			}
1029 
1030 			/*
1031 			 * We may see jumbolen in unaligned location, so
1032 			 * we'd need to perform bcopy().
1033 			 */
1034 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1035 			jumboplen = (u_int32_t)htonl(jumboplen);
1036 
1037 #if 1
1038 			/*
1039 			 * if there are multiple jumbo payload options,
1040 			 * *plenp will be non-zero and the packet will be
1041 			 * rejected.
1042 			 * the behavior may need some debate in ipngwg -
1043 			 * multiple options does not make sense, however,
1044 			 * there's no explicit mention in specification.
1045 			 */
1046 			if (*plenp != 0) {
1047 				ip6stat.ip6s_badoptions++;
1048 				icmp6_error(m, ICMP6_PARAM_PROB,
1049 					    ICMP6_PARAMPROB_HEADER,
1050 					    erroff + opt + 2 - opthead);
1051 				return (-1);
1052 			}
1053 #endif
1054 
1055 			/*
1056 			 * jumbo payload length must be larger than 65535.
1057 			 */
1058 			if (jumboplen <= IPV6_MAXPACKET) {
1059 				ip6stat.ip6s_badoptions++;
1060 				icmp6_error(m, ICMP6_PARAM_PROB,
1061 					    ICMP6_PARAMPROB_HEADER,
1062 					    erroff + opt + 2 - opthead);
1063 				return (-1);
1064 			}
1065 			*plenp = jumboplen;
1066 
1067 			break;
1068 		default:		/* unknown option */
1069 			if (hbhlen < IP6OPT_MINLEN) {
1070 				ip6stat.ip6s_toosmall++;
1071 				goto bad;
1072 			}
1073 			optlen = ip6_unknown_opt(opt, m,
1074 			    erroff + opt - opthead);
1075 			if (optlen == -1)
1076 				return (-1);
1077 			optlen += 2;
1078 			break;
1079 		}
1080 	}
1081 
1082 	return (0);
1083 
1084 bad:
1085 	m_freem(m);
1086 	return (-1);
1087 }
1088 
1089 /*
1090  * Unknown option processing.
1091  * The third argument `off' is the offset from the IPv6 header to the option,
1092  * which is necessary if the IPv6 header the and option header and IPv6 header
1093  * is not continuous in order to return an ICMPv6 error.
1094  */
1095 int
1096 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1097 {
1098 	struct ip6_hdr *ip6;
1099 
1100 	switch (IP6OPT_TYPE(*optp)) {
1101 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1102 		return ((int)*(optp + 1));
1103 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1104 		m_freem(m);
1105 		return (-1);
1106 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1107 		ip6stat.ip6s_badoptions++;
1108 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1109 		return (-1);
1110 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1111 		ip6stat.ip6s_badoptions++;
1112 		ip6 = mtod(m, struct ip6_hdr *);
1113 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1114 		    (m->m_flags & (M_BCAST|M_MCAST)))
1115 			m_freem(m);
1116 		else
1117 			icmp6_error(m, ICMP6_PARAM_PROB,
1118 				    ICMP6_PARAMPROB_OPTION, off);
1119 		return (-1);
1120 	}
1121 
1122 	m_freem(m);		/* XXX: NOTREACHED */
1123 	return (-1);
1124 }
1125 
1126 /*
1127  * Create the "control" list for this pcb.
1128  * The function will not modify mbuf chain at all.
1129  *
1130  * with KAME mbuf chain restriction:
1131  * The routine will be called from upper layer handlers like tcp6_input().
1132  * Thus the routine assumes that the caller (tcp6_input) have already
1133  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1134  * very first mbuf on the mbuf chain.
1135  */
1136 void
1137 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1138 		struct mbuf *m)
1139 {
1140 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1141 	struct thread *td = curthread;	/* XXX */
1142 	int privileged = 0;
1143 
1144 
1145 	if (priv_check(td, PRIV_ROOT) == 0)
1146  		privileged++;
1147 
1148 #ifdef SO_TIMESTAMP
1149 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1150 		struct timeval tv;
1151 
1152 		microtime(&tv);
1153 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1154 				      SCM_TIMESTAMP, SOL_SOCKET);
1155 		if (*mp) {
1156 			mp = &(*mp)->m_next;
1157 		}
1158 	}
1159 #endif
1160 
1161 	/* RFC 2292 sec. 5 */
1162 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1163 		struct in6_pktinfo pi6;
1164 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1165 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1166 			pi6.ipi6_addr.s6_addr16[1] = 0;
1167 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1168 					? m->m_pkthdr.rcvif->if_index
1169 					: 0;
1170 		*mp = sbcreatecontrol((caddr_t) &pi6,
1171 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1172 			IPPROTO_IPV6);
1173 		if (*mp)
1174 			mp = &(*mp)->m_next;
1175 	}
1176 
1177 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1178 		int hlim = ip6->ip6_hlim & 0xff;
1179 		*mp = sbcreatecontrol((caddr_t) &hlim,
1180 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1181 		if (*mp)
1182 			mp = &(*mp)->m_next;
1183 	}
1184 
1185 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1186 		u_int32_t flowinfo;
1187 		int tclass;
1188 
1189 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1190 		flowinfo >>= 20;
1191 
1192 		tclass = flowinfo & 0xff;
1193 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1194 		    IPV6_TCLASS, IPPROTO_IPV6);
1195 		if (*mp)
1196 			mp = &(*mp)->m_next;
1197 	}
1198 
1199 	/*
1200 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1201 	 * for the option, but it might be too strict, since there might
1202 	 * be some hop-by-hop options which can be returned to normal user.
1203 	 * See RFC 2292 section 6.
1204 	 */
1205 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1206 		/*
1207 		 * Check if a hop-by-hop options header is contatined in the
1208 		 * received packet, and if so, store the options as ancillary
1209 		 * data. Note that a hop-by-hop options header must be
1210 		 * just after the IPv6 header, which fact is assured through
1211 		 * the IPv6 input processing.
1212 		 */
1213 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1214 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1215 			struct ip6_hbh *hbh;
1216 			int hbhlen = 0;
1217 #ifdef PULLDOWN_TEST
1218 			struct mbuf *ext;
1219 #endif
1220 
1221 #ifndef PULLDOWN_TEST
1222 			hbh = (struct ip6_hbh *)(ip6 + 1);
1223 			hbhlen = (hbh->ip6h_len + 1) << 3;
1224 #else
1225 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1226 			    ip6->ip6_nxt);
1227 			if (ext == NULL) {
1228 				ip6stat.ip6s_tooshort++;
1229 				return;
1230 			}
1231 			hbh = mtod(ext, struct ip6_hbh *);
1232 			hbhlen = (hbh->ip6h_len + 1) << 3;
1233 			if (hbhlen != ext->m_len) {
1234 				m_freem(ext);
1235 				ip6stat.ip6s_tooshort++;
1236 				return;
1237 			}
1238 #endif
1239 
1240 			/*
1241 			 * XXX: We copy whole the header even if a jumbo
1242 			 * payload option is included, which option is to
1243 			 * be removed before returning in the RFC 2292.
1244 			 * Note: this constraint is removed in 2292bis.
1245 			 */
1246 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1247 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1248 			if (*mp)
1249 				mp = &(*mp)->m_next;
1250 #ifdef PULLDOWN_TEST
1251 			m_freem(ext);
1252 #endif
1253 		}
1254 	}
1255 
1256 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1257 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1258 		int proto, off, nxt;
1259 
1260 		/*
1261 		 * go through the header chain to see if a routing header is
1262 		 * contained in the packet. We need this information to store
1263 		 * destination options headers (if any) properly.
1264 		 * XXX: performance issue. We should record this info when
1265 		 * processing extension headers in incoming routine.
1266 		 * (todo) use m_aux?
1267 		 */
1268 		proto = IPPROTO_IPV6;
1269 		off = 0;
1270 		nxt = -1;
1271 		while (1) {
1272 			int newoff;
1273 
1274 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1275 			if (newoff < 0)
1276 				break;
1277 			if (newoff < off) /* invalid, check for safety */
1278 				break;
1279 			if ((proto = nxt) == IPPROTO_ROUTING)
1280 				break;
1281 			off = newoff;
1282 		}
1283 	}
1284 
1285 	if ((in6p->in6p_flags &
1286 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1287 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1288 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1289 
1290 		/*
1291 		 * Search for destination options headers or routing
1292 		 * header(s) through the header chain, and stores each
1293 		 * header as ancillary data.
1294 		 * Note that the order of the headers remains in
1295 		 * the chain of ancillary data.
1296 		 */
1297 		while (1) {	/* is explicit loop prevention necessary? */
1298 			struct ip6_ext *ip6e = NULL;
1299 			int elen;
1300 #ifdef PULLDOWN_TEST
1301 			struct mbuf *ext = NULL;
1302 #endif
1303 
1304 			/*
1305 			 * if it is not an extension header, don't try to
1306 			 * pull it from the chain.
1307 			 */
1308 			switch (nxt) {
1309 			case IPPROTO_DSTOPTS:
1310 			case IPPROTO_ROUTING:
1311 			case IPPROTO_HOPOPTS:
1312 			case IPPROTO_AH: /* is it possible? */
1313 				break;
1314 			default:
1315 				goto loopend;
1316 			}
1317 
1318 #ifndef PULLDOWN_TEST
1319 			if (off + sizeof(*ip6e) > m->m_len)
1320 				goto loopend;
1321 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1322 			if (nxt == IPPROTO_AH)
1323 				elen = (ip6e->ip6e_len + 2) << 2;
1324 			else
1325 				elen = (ip6e->ip6e_len + 1) << 3;
1326 			if (off + elen > m->m_len)
1327 				goto loopend;
1328 #else
1329 			ext = ip6_pullexthdr(m, off, nxt);
1330 			if (ext == NULL) {
1331 				ip6stat.ip6s_tooshort++;
1332 				return;
1333 			}
1334 			ip6e = mtod(ext, struct ip6_ext *);
1335 			if (nxt == IPPROTO_AH)
1336 				elen = (ip6e->ip6e_len + 2) << 2;
1337 			else
1338 				elen = (ip6e->ip6e_len + 1) << 3;
1339 			if (elen != ext->m_len) {
1340 				m_freem(ext);
1341 				ip6stat.ip6s_tooshort++;
1342 				return;
1343 			}
1344 #endif
1345 
1346 			switch (nxt) {
1347 			case IPPROTO_DSTOPTS:
1348 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1349 					break;
1350 
1351 				/*
1352 				 * We also require super-user privilege for
1353 				 * the option.
1354 				 * See the comments on IN6_HOPOPTS.
1355 				 */
1356 				if (!privileged)
1357 					break;
1358 
1359 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1360 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1361 						      IPPROTO_IPV6);
1362 				if (*mp)
1363 					mp = &(*mp)->m_next;
1364 				break;
1365 			case IPPROTO_ROUTING:
1366 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1367 					break;
1368 
1369 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1370 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1371 						      IPPROTO_IPV6);
1372 				if (*mp)
1373 					mp = &(*mp)->m_next;
1374 				break;
1375 			case IPPROTO_HOPOPTS:
1376 			case IPPROTO_AH: /* is it possible? */
1377 				break;
1378 
1379 			default:
1380 				/*
1381 			 	 * other cases have been filtered in the above.
1382 				 * none will visit this case.  here we supply
1383 				 * the code just in case (nxt overwritten or
1384 				 * other cases).
1385 				 */
1386 #ifdef PULLDOWN_TEST
1387 				m_freem(ext);
1388 #endif
1389 				goto loopend;
1390 
1391 			}
1392 
1393 			/* proceed with the next header. */
1394 			off += elen;
1395 			nxt = ip6e->ip6e_nxt;
1396 			ip6e = NULL;
1397 #ifdef PULLDOWN_TEST
1398 			m_freem(ext);
1399 			ext = NULL;
1400 #endif
1401 		}
1402 	  loopend:
1403 		;
1404 	}
1405 #undef IS2292
1406 }
1407 
1408 void
1409 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1410 {
1411 	struct socket *so;
1412 	struct mbuf *m_mtu;
1413 	struct ip6_mtuinfo mtuctl;
1414 
1415 	so =  in6p->inp_socket;
1416 
1417 	if (mtu == NULL)
1418 		return;
1419 
1420 #ifdef DIAGNOSTIC
1421 	if (so == NULL)		/* I believe this is impossible */
1422 		panic("ip6_notify_pmtu: socket is NULL");
1423 #endif
1424 
1425 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1426 	mtuctl.ip6m_mtu = *mtu;
1427 	mtuctl.ip6m_addr = *dst;
1428 
1429 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1430 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1431 		return;
1432 
1433 	lwkt_gettoken(&so->so_rcv.ssb_token);
1434 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1435 	    == 0) {
1436 		m_freem(m_mtu);
1437 		/* XXX: should count statistics */
1438 	} else {
1439 		sorwakeup(so);
1440 	}
1441 	lwkt_reltoken(&so->so_rcv.ssb_token);
1442 }
1443 
1444 #ifdef PULLDOWN_TEST
1445 /*
1446  * pull single extension header from mbuf chain.  returns single mbuf that
1447  * contains the result, or NULL on error.
1448  */
1449 static struct mbuf *
1450 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1451 {
1452 	struct ip6_ext ip6e;
1453 	size_t elen;
1454 	struct mbuf *n;
1455 
1456 #ifdef DIAGNOSTIC
1457 	switch (nxt) {
1458 	case IPPROTO_DSTOPTS:
1459 	case IPPROTO_ROUTING:
1460 	case IPPROTO_HOPOPTS:
1461 	case IPPROTO_AH: /* is it possible? */
1462 		break;
1463 	default:
1464 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1465 	}
1466 #endif
1467 
1468 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1469 	if (nxt == IPPROTO_AH)
1470 		elen = (ip6e.ip6e_len + 2) << 2;
1471 	else
1472 		elen = (ip6e.ip6e_len + 1) << 3;
1473 
1474 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1475 	if (n == NULL)
1476 		return NULL;
1477 	n->m_len = 0;
1478 
1479 	if (elen >= M_TRAILINGSPACE(n)) {
1480 		m_free(n);
1481 		return NULL;
1482 	}
1483 
1484 	m_copydata(m, off, elen, mtod(n, caddr_t));
1485 	n->m_len = elen;
1486 	return n;
1487 }
1488 #endif
1489 
1490 /*
1491  * Get pointer to the previous header followed by the header
1492  * currently processed.
1493  * XXX: This function supposes that
1494  *	M includes all headers,
1495  *	the next header field and the header length field of each header
1496  *	are valid, and
1497  *	the sum of each header length equals to OFF.
1498  * Because of these assumptions, this function must be called very
1499  * carefully. Moreover, it will not be used in the near future when
1500  * we develop `neater' mechanism to process extension headers.
1501  */
1502 char *
1503 ip6_get_prevhdr(struct mbuf *m, int off)
1504 {
1505 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1506 
1507 	if (off == sizeof(struct ip6_hdr))
1508 		return (&ip6->ip6_nxt);
1509 	else {
1510 		int len, nxt;
1511 		struct ip6_ext *ip6e = NULL;
1512 
1513 		nxt = ip6->ip6_nxt;
1514 		len = sizeof(struct ip6_hdr);
1515 		while (len < off) {
1516 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1517 
1518 			switch (nxt) {
1519 			case IPPROTO_FRAGMENT:
1520 				len += sizeof(struct ip6_frag);
1521 				break;
1522 			case IPPROTO_AH:
1523 				len += (ip6e->ip6e_len + 2) << 2;
1524 				break;
1525 			default:
1526 				len += (ip6e->ip6e_len + 1) << 3;
1527 				break;
1528 			}
1529 			nxt = ip6e->ip6e_nxt;
1530 		}
1531 		if (ip6e)
1532 			return (&ip6e->ip6e_nxt);
1533 		else
1534 			return NULL;
1535 	}
1536 }
1537 
1538 /*
1539  * get next header offset.  m will be retained.
1540  */
1541 int
1542 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1543 {
1544 	struct ip6_hdr ip6;
1545 	struct ip6_ext ip6e;
1546 	struct ip6_frag fh;
1547 
1548 	/* just in case */
1549 	if (m == NULL)
1550 		panic("ip6_nexthdr: m == NULL");
1551 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1552 		return -1;
1553 
1554 	switch (proto) {
1555 	case IPPROTO_IPV6:
1556 		if (m->m_pkthdr.len < off + sizeof(ip6))
1557 			return -1;
1558 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1559 		if (nxtp)
1560 			*nxtp = ip6.ip6_nxt;
1561 		off += sizeof(ip6);
1562 		return off;
1563 
1564 	case IPPROTO_FRAGMENT:
1565 		/*
1566 		 * terminate parsing if it is not the first fragment,
1567 		 * it does not make sense to parse through it.
1568 		 */
1569 		if (m->m_pkthdr.len < off + sizeof(fh))
1570 			return -1;
1571 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1572 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1573 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1574 			return -1;
1575 		if (nxtp)
1576 			*nxtp = fh.ip6f_nxt;
1577 		off += sizeof(struct ip6_frag);
1578 		return off;
1579 
1580 	case IPPROTO_AH:
1581 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1582 			return -1;
1583 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1584 		if (nxtp)
1585 			*nxtp = ip6e.ip6e_nxt;
1586 		off += (ip6e.ip6e_len + 2) << 2;
1587 		return off;
1588 
1589 	case IPPROTO_HOPOPTS:
1590 	case IPPROTO_ROUTING:
1591 	case IPPROTO_DSTOPTS:
1592 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1593 			return -1;
1594 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1595 		if (nxtp)
1596 			*nxtp = ip6e.ip6e_nxt;
1597 		off += (ip6e.ip6e_len + 1) << 3;
1598 		return off;
1599 
1600 	case IPPROTO_NONE:
1601 	case IPPROTO_ESP:
1602 	case IPPROTO_IPCOMP:
1603 		/* give up */
1604 		return -1;
1605 
1606 	default:
1607 		return -1;
1608 	}
1609 
1610 	return -1;
1611 }
1612 
1613 /*
1614  * get offset for the last header in the chain.  m will be kept untainted.
1615  */
1616 int
1617 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1618 {
1619 	int newoff;
1620 	int nxt;
1621 
1622 	if (!nxtp) {
1623 		nxt = -1;
1624 		nxtp = &nxt;
1625 	}
1626 	while (1) {
1627 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1628 		if (newoff < 0)
1629 			return off;
1630 		else if (newoff < off)
1631 			return -1;	/* invalid */
1632 		else if (newoff == off)
1633 			return newoff;
1634 
1635 		off = newoff;
1636 		proto = *nxtp;
1637 	}
1638 }
1639 
1640 struct ip6aux *
1641 ip6_addaux(struct mbuf *m)
1642 {
1643 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1644 	if (!tag) {
1645 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1646 				sizeof (struct ip6aux),
1647 				MB_DONTWAIT);
1648 		if (tag)
1649 			m_tag_prepend(m, tag);
1650 	}
1651 	if (tag)
1652 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1653 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1654 }
1655 
1656 struct ip6aux *
1657 ip6_findaux(struct mbuf *m)
1658 {
1659 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1660 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1661 }
1662 
1663 void
1664 ip6_delaux(struct mbuf *m)
1665 {
1666 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1667 	if (tag)
1668 		m_tag_delete(m, tag);
1669 }
1670 
1671 /*
1672  * System control for IP6
1673  */
1674 
1675 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1676 	0,		0,		0,		0,
1677 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1678 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1679 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1680 	0,		0,		0,		0,
1681 	ENOPROTOOPT
1682 };
1683