xref: /dragonfly/sys/netinet6/ip6_input.c (revision 06937ef9)
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 	in6_tmpaddrtimer_init();
209 }
210 
211 /* cheat */
212 /* This must be after route_init(), which is now SI_ORDER_THIRD */
213 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
214 
215 extern struct	route_in6 ip6_forward_rt;
216 
217 static void
218 ip6_input(netmsg_t msg)
219 {
220 	struct mbuf *m = msg->packet.nm_packet;
221 	struct ip6_hdr *ip6;
222 	int off = sizeof(struct ip6_hdr), nest;
223 	u_int32_t plen;
224 	u_int32_t rtalert = ~0;
225 	int nxt, ours = 0, rh_present = 0;
226 	struct ifnet *deliverifp = NULL;
227 	struct in6_addr odst;
228 	int srcrt = 0;
229 
230 #ifdef IPSEC
231 	/*
232 	 * should the inner packet be considered authentic?
233 	 * see comment in ah4_input().
234 	 */
235 	if (m) {
236 		m->m_flags &= ~M_AUTHIPHDR;
237 		m->m_flags &= ~M_AUTHIPDGM;
238 	}
239 #endif
240 
241 	/*
242 	 * make sure we don't have onion peering information into m_aux.
243 	 */
244 	ip6_delaux(m);
245 
246 	/*
247 	 * mbuf statistics
248 	 */
249 	if (m->m_flags & M_EXT) {
250 		if (m->m_next)
251 			ip6stat.ip6s_mext2m++;
252 		else
253 			ip6stat.ip6s_mext1++;
254 	} else {
255 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
256 		if (m->m_next) {
257 			if (m->m_flags & M_LOOP) {
258 				ip6stat.ip6s_m2m[loif->if_index]++; /* XXX */
259 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
260 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
261 			else
262 				ip6stat.ip6s_m2m[0]++;
263 		} else
264 			ip6stat.ip6s_m1++;
265 #undef M2MMAX
266 	}
267 
268 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
269 	ip6stat.ip6s_total++;
270 
271 #ifndef PULLDOWN_TEST
272 	/*
273 	 * L2 bridge code and some other code can return mbuf chain
274 	 * that does not conform to KAME requirement.  too bad.
275 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
276 	 */
277 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
278 		struct mbuf *n;
279 
280 		n = m_getb(m->m_pkthdr.len, M_NOWAIT, MT_HEADER, M_PKTHDR);
281 		if (n == NULL)
282 			goto bad;
283 		M_MOVE_PKTHDR(n, m);
284 
285 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
286 		n->m_len = n->m_pkthdr.len;
287 		m_freem(m);
288 		m = n;
289 	}
290 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
291 #endif
292 
293 	if (m->m_len < sizeof(struct ip6_hdr)) {
294 		struct ifnet *inifp;
295 		inifp = m->m_pkthdr.rcvif;
296 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
297 			ip6stat.ip6s_toosmall++;
298 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
299 			goto bad2;
300 		}
301 	}
302 
303 	ip6 = mtod(m, struct ip6_hdr *);
304 
305 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
306 		ip6stat.ip6s_badvers++;
307 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
308 		goto bad;
309 	}
310 
311 	/*
312 	 * Run through list of hooks for input packets.
313 	 *
314 	 * NB: Beware of the destination address changing
315 	 *     (e.g. by NAT rewriting). When this happens,
316 	 *     tell ip6_forward to do the right thing.
317 	 */
318 	if (pfil_has_hooks(&inet6_pfil_hook)) {
319 		odst = ip6->ip6_dst;
320 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
321 		    m->m_pkthdr.rcvif, PFIL_IN)) {
322 			goto bad2;
323 		}
324 		if (m == NULL)			/* consumed by filter */
325 			goto bad2;
326 		ip6 = mtod(m, struct ip6_hdr *);
327 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
328 	}
329 
330 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
331 
332 #ifdef ALTQ
333 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
334 		/* packet is dropped by traffic conditioner */
335 		return;
336 	}
337 #endif
338 
339 	/*
340 	 * Check with the firewall...
341 	 */
342 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
343 		u_short port = 0;
344 		/* If ipfw says divert, we have to just drop packet */
345 		/* use port as a dummy argument */
346 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
347 			m_freem(m);
348 			m = NULL;
349 		}
350 		if (!m)
351 			goto bad2;
352 	}
353 
354 	/*
355 	 * Check against address spoofing/corruption.
356 	 */
357 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
358 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
359 		/*
360 		 * XXX: "badscope" is not very suitable for a multicast source.
361 		 */
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
367 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
368 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
369 		ip6stat.ip6s_badscope++;
370 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
371 		goto bad;
372 	}
373 
374 	/*
375 	 * The following check is not documented in specs.  A malicious
376 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
377 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
378 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
379 	 *
380 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
381 	 * support IPv4-less kernel compilation, we cannot support SIIT
382 	 * environment at all.  So, it makes more sense for us to reject any
383 	 * malicious packets for non-SIIT environment, than try to do a
384 	 * partial support for SIIT environment.
385 	 */
386 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
387 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
388 		ip6stat.ip6s_badscope++;
389 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
390 		goto bad;
391 	}
392 #if 0
393 	/*
394 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
395 	 *
396 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
397 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
398 	 * is revised to forbid relaying case.
399 	 */
400 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
401 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
402 		ip6stat.ip6s_badscope++;
403 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
404 		goto bad;
405 	}
406 #endif
407 
408 	/* drop packets if interface ID portion is already filled */
409 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
410 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
411 		    ip6->ip6_src.s6_addr16[1]) {
412 			ip6stat.ip6s_badscope++;
413 			goto bad;
414 		}
415 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
416 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
417 		    ip6->ip6_dst.s6_addr16[1]) {
418 			ip6stat.ip6s_badscope++;
419 			goto bad;
420 		}
421 	}
422 
423 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
424 		ip6->ip6_src.s6_addr16[1]
425 			= htons(m->m_pkthdr.rcvif->if_index);
426 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
427 		ip6->ip6_dst.s6_addr16[1]
428 			= htons(m->m_pkthdr.rcvif->if_index);
429 
430 	/*
431 	 * Multicast check
432 	 *
433 	 * WARNING!  For general subnet proxying the interface hw will
434 	 *	     likely filter out the multicast solicitations, so
435 	 *	     the interface must be in promiscuous mode.
436 	 */
437 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
438 		ours = 1;
439 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
440 		deliverifp = m->m_pkthdr.rcvif;
441 		goto hbhcheck;
442 	}
443 
444 	/*
445 	 *  Unicast check
446 	 */
447 	if (ip6_forward_rt.ro_rt != NULL &&
448 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
449 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
450 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr)) {
451 		ip6stat.ip6s_forward_cachehit++;
452 	} else {
453 		struct sockaddr_in6 *dst6;
454 
455 		if (ip6_forward_rt.ro_rt) {
456 			/* route is down or destination is different */
457 			ip6stat.ip6s_forward_cachemiss++;
458 			RTFREE(ip6_forward_rt.ro_rt);
459 			ip6_forward_rt.ro_rt = 0;
460 		}
461 
462 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
463 		dst6 = &ip6_forward_rt.ro_dst;
464 		dst6->sin6_len = sizeof(struct sockaddr_in6);
465 		dst6->sin6_family = AF_INET6;
466 		dst6->sin6_addr = ip6->ip6_dst;
467 
468 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
469 	}
470 
471 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
472 
473 	/*
474 	 * Accept the packet if the forwarding interface to the destination
475 	 * according to the routing table is the loopback interface,
476 	 * unless the associated route has a gateway.
477 	 * Note that this approach causes to accept a packet if there is a
478 	 * route to the loopback interface for the destination of the packet.
479 	 * But we think it's even useful in some situations, e.g. when using
480 	 * a special daemon which wants to intercept the packet.
481 	 *
482 	 * XXX: some OSes automatically make a cloned route for the destination
483 	 * of an outgoing packet.  If the outgoing interface of the packet
484 	 * is a loopback one, the kernel would consider the packet to be
485 	 * accepted, even if we have no such address assinged on the interface.
486 	 * We check the cloned flag of the route entry to reject such cases,
487 	 * assuming that route entries for our own addresses are not made by
488 	 * cloning (it should be true because in6_addloop explicitly installs
489 	 * the host route).  However, we might have to do an explicit check
490 	 * while it would be less efficient.  Or, should we rather install a
491 	 * reject route for such a case?
492 	 */
493 	if (ip6_forward_rt.ro_rt &&
494 	    (ip6_forward_rt.ro_rt->rt_flags &
495 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
496 #ifdef RTF_WASCLONED
497 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
498 #endif
499 #ifdef RTF_CLONED
500 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
501 #endif
502 #if 0
503 	    /*
504 	     * The check below is redundant since the comparison of
505 	     * the destination and the key of the rtentry has
506 	     * already done through looking up the routing table.
507 	     */
508 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
509 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
510 #endif
511 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
512 		struct in6_ifaddr *ia6 =
513 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
514 
515 		/*
516 		 * record address information into m_aux.
517 		 */
518 		ip6_setdstifaddr(m, ia6);
519 
520 		/*
521 		 * packets to a tentative, duplicated, or somehow invalid
522 		 * address must not be accepted.
523 		 */
524 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
525 			/* this address is ready */
526 			ours = 1;
527 			deliverifp = ia6->ia_ifp;	/* correct? */
528 			/* Count the packet in the ip address stats */
529 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
530 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
531 			goto hbhcheck;
532 		} else {
533 			/* address is not ready, so discard the packet. */
534 			nd6log((LOG_INFO,
535 			    "ip6_input: packet to an unready address %s->%s\n",
536 			    ip6_sprintf(&ip6->ip6_src),
537 			    ip6_sprintf(&ip6->ip6_dst)));
538 
539 			goto bad;
540 		}
541 	}
542 
543 	/*
544 	 * Now there is no reason to process the packet if it's not our own
545 	 * and we're not a router.
546 	 */
547 	if (!ip6_forwarding) {
548 		ip6stat.ip6s_cantforward++;
549 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
550 		goto bad;
551 	}
552 
553 hbhcheck:
554 	/*
555 	 * record address information into m_aux, if we don't have one yet.
556 	 * note that we are unable to record it, if the address is not listed
557 	 * as our interface address (e.g. multicast addresses and such).
558 	 */
559 	if (deliverifp && !ip6_getdstifaddr(m)) {
560 		struct in6_ifaddr *ia6;
561 
562 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
563 		if (ia6) {
564 			if (!ip6_setdstifaddr(m, ia6)) {
565 				/*
566 				 * XXX maybe we should drop the packet here,
567 				 * as we could not provide enough information
568 				 * to the upper layers.
569 				 */
570 			}
571 		}
572 	}
573 
574 	/*
575 	 * Process Hop-by-Hop options header if it's contained.
576 	 * m may be modified in ip6_hopopts_input().
577 	 * If a JumboPayload option is included, plen will also be modified.
578 	 */
579 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
580 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
581 		struct ip6_hbh *hbh;
582 
583 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
584 #if 0	/*touches NULL pointer*/
585 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
586 #endif
587 			goto bad2;	/* m have already been freed */
588 		}
589 
590 		/* adjust pointer */
591 		ip6 = mtod(m, struct ip6_hdr *);
592 
593 		/*
594 		 * if the payload length field is 0 and the next header field
595 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
596 		 * option MUST be included.
597 		 */
598 		if (ip6->ip6_plen == 0 && plen == 0) {
599 			/*
600 			 * Note that if a valid jumbo payload option is
601 			 * contained, ip6_hoptops_input() must set a valid
602 			 * (non-zero) payload length to the variable plen.
603 			 */
604 			ip6stat.ip6s_badoptions++;
605 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
606 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
607 			icmp6_error(m, ICMP6_PARAM_PROB,
608 				    ICMP6_PARAMPROB_HEADER,
609 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
610 			goto bad2;
611 		}
612 #ifndef PULLDOWN_TEST
613 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
614 		hbh = (struct ip6_hbh *)(ip6 + 1);
615 #else
616 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
617 			sizeof(struct ip6_hbh));
618 		if (hbh == NULL) {
619 			ip6stat.ip6s_tooshort++;
620 			goto bad2;
621 		}
622 #endif
623 		nxt = hbh->ip6h_nxt;
624 
625 		/*
626 		 * If we are acting as a router and the packet contains a
627 		 * router alert option, see if we know the option value.
628 		 * Currently, we only support the option value for MLD, in which
629 		 * case we should pass the packet to the multicast routing
630 		 * daemon.
631 		 */
632 		if (rtalert != ~0 && ip6_forwarding) {
633 			switch (rtalert) {
634 			case IP6OPT_RTALERT_MLD:
635 				ours = 1;
636 				break;
637 			default:
638 				/*
639 				 * RFC2711 requires unrecognized values must be
640 				 * silently ignored.
641 				 */
642 				break;
643 			}
644 		}
645 	} else
646 		nxt = ip6->ip6_nxt;
647 
648 	/*
649 	 * Check that the amount of data in the buffers
650 	 * is as at least much as the IPv6 header would have us expect.
651 	 * Trim mbufs if longer than we expect.
652 	 * Drop packet if shorter than we expect.
653 	 */
654 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
655 		ip6stat.ip6s_tooshort++;
656 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
657 		goto bad;
658 	}
659 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
660 		if (m->m_len == m->m_pkthdr.len) {
661 			m->m_len = sizeof(struct ip6_hdr) + plen;
662 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
663 		} else
664 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
665 	}
666 
667 	/*
668 	 * Forward if desirable.
669 	 */
670 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
671 		/*
672 		 * If we are acting as a multicast router, all
673 		 * incoming multicast packets are passed to the
674 		 * kernel-level multicast forwarding function.
675 		 * The packet is returned (relatively) intact; if
676 		 * ip6_mforward() returns a non-zero value, the packet
677 		 * must be discarded, else it may be accepted below.
678 		 */
679 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
680 			ip6stat.ip6s_cantforward++;
681 			goto bad;
682 		}
683 		if (!ours)
684 			goto bad;
685 	} else if (!ours) {
686 		ip6_forward(m, srcrt);
687 		goto bad2;
688 	}
689 
690 	ip6 = mtod(m, struct ip6_hdr *);
691 
692 	/*
693 	 * Malicious party may be able to use IPv4 mapped addr to confuse
694 	 * tcp/udp stack and bypass security checks (act as if it was from
695 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
696 	 *
697 	 * For SIIT end node behavior, you may want to disable the check.
698 	 * However, you will  become vulnerable to attacks using IPv4 mapped
699 	 * source.
700 	 */
701 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
702 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
703 		ip6stat.ip6s_badscope++;
704 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
705 		goto bad;
706 	}
707 
708 	/*
709 	 * Tell launch routine the next header
710 	 */
711 	ip6stat.ip6s_delivered++;
712 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
713 	nest = 0;
714 
715 	rh_present = 0;
716 	while (nxt != IPPROTO_DONE) {
717 		struct protosw *sw6;
718 
719 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
720 			ip6stat.ip6s_toomanyhdr++;
721 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
722 			goto bad;
723 		}
724 
725 		/*
726 		 * protection against faulty packet - there should be
727 		 * more sanity checks in header chain processing.
728 		 */
729 		if (m->m_pkthdr.len < off) {
730 			ip6stat.ip6s_tooshort++;
731 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
732 			goto bad;
733 		}
734 
735 #if 0
736 		/*
737 		 * do we need to do it for every header?  yeah, other
738 		 * functions can play with it (like re-allocate and copy).
739 		 */
740 		mhist = ip6_addaux(m);
741 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
742 			hist = mtod(mhist, caddr_t) + mhist->m_len;
743 			bcopy(&nxt, hist, sizeof(nxt));
744 			mhist->m_len += sizeof(nxt);
745 		} else {
746 			ip6stat.ip6s_toomanyhdr++;
747 			goto bad;
748 		}
749 #endif
750 
751 		if (nxt == IPPROTO_ROUTING) {
752 			if (rh_present++) {
753 				in6_ifstat_inc(m->m_pkthdr.rcvif,
754 				    ifs6_in_hdrerr);
755 				ip6stat.ip6s_badoptions++;
756 				goto bad;
757 			}
758 		}
759 
760 		sw6 = &inet6sw[ip6_protox[nxt]];
761 #ifdef IPSEC
762 		/*
763 		 * enforce IPsec policy checking if we are seeing last header.
764 		 * note that we do not visit this with protocols with pcb layer
765 		 * code - like udp/tcp/raw ip.
766 		 */
767 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
768 			ipsec6stat.in_polvio++;
769 			goto bad;
770 		}
771 #endif
772 		/*
773 		 * If this is a terminal header forward to the port, otherwise
774 		 * process synchronously for more headers.
775 		 */
776 		if (sw6->pr_flags & PR_LASTHDR) {
777 			struct netmsg_packet *pmsg;
778 			lwkt_port_t port;
779 
780 			port = netisr_cpuport(0); /* XXX */
781 			if ((m->m_flags & M_HASH) == 0)
782 				m_sethash(m, 0); /* XXX */
783 
784 			KKASSERT(port != NULL);
785 			pmsg = &m->m_hdr.mh_netmsg;
786 			netmsg_init(&pmsg->base, NULL,
787 				    &netisr_apanic_rport,
788 				    0, transport6_processing_handler);
789 			pmsg->nm_packet = m;
790 			pmsg->nm_nxt = nxt;
791 			pmsg->base.lmsg.u.ms_result = off;
792 			lwkt_sendmsg(port, &pmsg->base.lmsg);
793 			/* done with m */
794 			nxt = IPPROTO_DONE;
795 		} else {
796 			nxt = sw6->pr_input(&m, &off, nxt);
797 		}
798 	}
799 	goto bad2;
800 bad:
801 	m_freem(m);
802 bad2:
803 	;
804 	/* msg was embedded in the mbuf, do not reply! */
805 }
806 
807 /*
808  * We have to call the pr_input() function from the correct protocol
809  * thread.  The sw6->pr_soport() request at the end of ip6_input()
810  * returns the port and we forward a netmsg to the port to execute
811  * this function.
812  */
813 static void
814 transport6_processing_handler(netmsg_t netmsg)
815 {
816 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
817 	struct protosw *sw6;
818 	int hlen;
819 	int nxt;
820 
821 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
822 	hlen = pmsg->base.lmsg.u.ms_result;
823 
824 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
825 	KKASSERT(nxt == IPPROTO_DONE);
826 	/* netmsg was embedded in the mbuf, do not reply! */
827 }
828 
829 /*
830  * set/grab in6_ifaddr correspond to IPv6 destination address.
831  * XXX backward compatibility wrapper
832  */
833 static struct ip6aux *
834 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
835 {
836 	struct ip6aux *n;
837 
838 	n = ip6_addaux(m);
839 	if (n)
840 		n->ip6a_dstia6 = ia6;
841 	return n;	/* NULL if failed to set */
842 }
843 
844 struct in6_ifaddr *
845 ip6_getdstifaddr(struct mbuf *m)
846 {
847 	struct ip6aux *n;
848 
849 	n = ip6_findaux(m);
850 	if (n)
851 		return n->ip6a_dstia6;
852 	else
853 		return NULL;
854 }
855 
856 /*
857  * Hop-by-Hop options header processing. If a valid jumbo payload option is
858  * included, the real payload length will be stored in plenp.
859  */
860 static int
861 ip6_hopopts_input(u_int32_t *plenp,
862 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
863 		  struct mbuf **mp,
864 		  int *offp)
865 {
866 	struct mbuf *m = *mp;
867 	int off = *offp, hbhlen;
868 	struct ip6_hbh *hbh;
869 
870 	/* validation of the length of the header */
871 #ifndef PULLDOWN_TEST
872 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
873 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
874 	hbhlen = (hbh->ip6h_len + 1) << 3;
875 
876 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
877 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
878 #else
879 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
880 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
881 	if (hbh == NULL) {
882 		ip6stat.ip6s_tooshort++;
883 		return -1;
884 	}
885 	hbhlen = (hbh->ip6h_len + 1) << 3;
886 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
887 		hbhlen);
888 	if (hbh == NULL) {
889 		ip6stat.ip6s_tooshort++;
890 		return -1;
891 	}
892 #endif
893 	off += hbhlen;
894 	hbhlen -= sizeof(struct ip6_hbh);
895 
896 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
897 				hbhlen, rtalertp, plenp) < 0)
898 		return (-1);
899 
900 	*offp = off;
901 	*mp = m;
902 	return (0);
903 }
904 
905 /*
906  * Search header for all Hop-by-hop options and process each option.
907  * This function is separate from ip6_hopopts_input() in order to
908  * handle a case where the sending node itself process its hop-by-hop
909  * options header. In such a case, the function is called from ip6_output().
910  *
911  * The function assumes that hbh header is located right after the IPv6 header
912  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
913  * opthead + hbhlen is located in continuous memory region.
914  */
915 int
916 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
917 		    u_int32_t *rtalertp, u_int32_t *plenp)
918 {
919 	struct ip6_hdr *ip6;
920 	int optlen = 0;
921 	u_int8_t *opt = opthead;
922 	u_int16_t rtalert_val;
923 	u_int32_t jumboplen;
924 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
925 
926 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
927 		switch (*opt) {
928 		case IP6OPT_PAD1:
929 			optlen = 1;
930 			break;
931 		case IP6OPT_PADN:
932 			if (hbhlen < IP6OPT_MINLEN) {
933 				ip6stat.ip6s_toosmall++;
934 				goto bad;
935 			}
936 			optlen = *(opt + 1) + 2;
937 			break;
938 		case IP6OPT_RTALERT:
939 			/* XXX may need check for alignment */
940 			if (hbhlen < IP6OPT_RTALERT_LEN) {
941 				ip6stat.ip6s_toosmall++;
942 				goto bad;
943 			}
944 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
945 				/* XXX stat */
946 				icmp6_error(m, ICMP6_PARAM_PROB,
947 				    ICMP6_PARAMPROB_HEADER,
948 				    erroff + opt + 1 - opthead);
949 				return (-1);
950 			}
951 			optlen = IP6OPT_RTALERT_LEN;
952 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
953 			*rtalertp = ntohs(rtalert_val);
954 			break;
955 		case IP6OPT_JUMBO:
956 			/* XXX may need check for alignment */
957 			if (hbhlen < IP6OPT_JUMBO_LEN) {
958 				ip6stat.ip6s_toosmall++;
959 				goto bad;
960 			}
961 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
962 				/* XXX stat */
963 				icmp6_error(m, ICMP6_PARAM_PROB,
964 				    ICMP6_PARAMPROB_HEADER,
965 				    erroff + opt + 1 - opthead);
966 				return (-1);
967 			}
968 			optlen = IP6OPT_JUMBO_LEN;
969 
970 			/*
971 			 * IPv6 packets that have non 0 payload length
972 			 * must not contain a jumbo payload option.
973 			 */
974 			ip6 = mtod(m, struct ip6_hdr *);
975 			if (ip6->ip6_plen) {
976 				ip6stat.ip6s_badoptions++;
977 				icmp6_error(m, ICMP6_PARAM_PROB,
978 				    ICMP6_PARAMPROB_HEADER,
979 				    erroff + opt - opthead);
980 				return (-1);
981 			}
982 
983 			/*
984 			 * We may see jumbolen in unaligned location, so
985 			 * we'd need to perform bcopy().
986 			 */
987 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
988 			jumboplen = (u_int32_t)htonl(jumboplen);
989 
990 #if 1
991 			/*
992 			 * if there are multiple jumbo payload options,
993 			 * *plenp will be non-zero and the packet will be
994 			 * rejected.
995 			 * the behavior may need some debate in ipngwg -
996 			 * multiple options does not make sense, however,
997 			 * there's no explicit mention in specification.
998 			 */
999 			if (*plenp != 0) {
1000 				ip6stat.ip6s_badoptions++;
1001 				icmp6_error(m, ICMP6_PARAM_PROB,
1002 				    ICMP6_PARAMPROB_HEADER,
1003 				    erroff + opt + 2 - opthead);
1004 				return (-1);
1005 			}
1006 #endif
1007 
1008 			/*
1009 			 * jumbo payload length must be larger than 65535.
1010 			 */
1011 			if (jumboplen <= IPV6_MAXPACKET) {
1012 				ip6stat.ip6s_badoptions++;
1013 				icmp6_error(m, ICMP6_PARAM_PROB,
1014 				    ICMP6_PARAMPROB_HEADER,
1015 				    erroff + opt + 2 - opthead);
1016 				return (-1);
1017 			}
1018 			*plenp = jumboplen;
1019 
1020 			break;
1021 		default:		/* unknown option */
1022 			if (hbhlen < IP6OPT_MINLEN) {
1023 				ip6stat.ip6s_toosmall++;
1024 				goto bad;
1025 			}
1026 			optlen = ip6_unknown_opt(opt, m,
1027 			    erroff + opt - opthead);
1028 			if (optlen == -1)
1029 				return (-1);
1030 			optlen += 2;
1031 			break;
1032 		}
1033 	}
1034 
1035 	return (0);
1036 
1037 bad:
1038 	m_freem(m);
1039 	return (-1);
1040 }
1041 
1042 /*
1043  * Unknown option processing.
1044  * The third argument `off' is the offset from the IPv6 header to the option,
1045  * which is necessary if the IPv6 header the and option header and IPv6 header
1046  * is not continuous in order to return an ICMPv6 error.
1047  */
1048 int
1049 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1050 {
1051 	struct ip6_hdr *ip6;
1052 
1053 	switch (IP6OPT_TYPE(*optp)) {
1054 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1055 		return ((int)*(optp + 1));
1056 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1057 		m_freem(m);
1058 		return (-1);
1059 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1060 		ip6stat.ip6s_badoptions++;
1061 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1062 		return (-1);
1063 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1064 		ip6stat.ip6s_badoptions++;
1065 		ip6 = mtod(m, struct ip6_hdr *);
1066 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1067 		    (m->m_flags & (M_BCAST|M_MCAST)))
1068 			m_freem(m);
1069 		else
1070 			icmp6_error(m, ICMP6_PARAM_PROB,
1071 				    ICMP6_PARAMPROB_OPTION, off);
1072 		return (-1);
1073 	}
1074 
1075 	m_freem(m);		/* XXX: NOTREACHED */
1076 	return (-1);
1077 }
1078 
1079 /*
1080  * Create the "control" list for this pcb.
1081  * The function will not modify mbuf chain at all.
1082  *
1083  * with KAME mbuf chain restriction:
1084  * The routine will be called from upper layer handlers like tcp6_input().
1085  * Thus the routine assumes that the caller (tcp6_input) have already
1086  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1087  * very first mbuf on the mbuf chain.
1088  */
1089 void
1090 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1091 		struct mbuf *m)
1092 {
1093 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1094 	struct thread *td = curthread;	/* XXX */
1095 	int privileged = 0;
1096 
1097 
1098 	if (priv_check(td, PRIV_ROOT) == 0)
1099 		privileged++;
1100 
1101 #ifdef SO_TIMESTAMP
1102 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1103 		struct timeval tv;
1104 
1105 		microtime(&tv);
1106 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1107 		    SCM_TIMESTAMP, SOL_SOCKET);
1108 		if (*mp) {
1109 			mp = &(*mp)->m_next;
1110 		}
1111 	}
1112 #endif
1113 
1114 	/* RFC 2292 sec. 5 */
1115 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1116 		struct in6_pktinfo pi6;
1117 
1118 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1119 		in6_clearscope(&pi6.ipi6_addr); /* XXX */
1120 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif) ?
1121 		    m->m_pkthdr.rcvif->if_index : 0;
1122 		*mp = sbcreatecontrol((caddr_t)&pi6, sizeof(struct in6_pktinfo),
1123 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1124 		if (*mp)
1125 			mp = &(*mp)->m_next;
1126 	}
1127 
1128 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1129 		int hlim = ip6->ip6_hlim & 0xff;
1130 
1131 		*mp = sbcreatecontrol((caddr_t)&hlim, sizeof(int),
1132 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1133 		if (*mp)
1134 			mp = &(*mp)->m_next;
1135 	}
1136 
1137 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1138 		u_int32_t flowinfo;
1139 		int tclass;
1140 
1141 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1142 		flowinfo >>= 20;
1143 
1144 		tclass = flowinfo & 0xff;
1145 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1146 		    IPV6_TCLASS, IPPROTO_IPV6);
1147 		if (*mp)
1148 			mp = &(*mp)->m_next;
1149 	}
1150 
1151 	/*
1152 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1153 	 * for the option, but it might be too strict, since there might
1154 	 * be some hop-by-hop options which can be returned to normal user.
1155 	 * See RFC 2292 section 6.
1156 	 */
1157 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1158 		/*
1159 		 * Check if a hop-by-hop options header is contatined in the
1160 		 * received packet, and if so, store the options as ancillary
1161 		 * data. Note that a hop-by-hop options header must be
1162 		 * just after the IPv6 header, which is assured through the
1163 		 * IPv6 input processing.
1164 		 */
1165 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1166 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1167 			struct ip6_hbh *hbh;
1168 			int hbhlen = 0;
1169 #ifdef PULLDOWN_TEST
1170 			struct mbuf *ext;
1171 #endif
1172 
1173 #ifndef PULLDOWN_TEST
1174 			hbh = (struct ip6_hbh *)(ip6 + 1);
1175 			hbhlen = (hbh->ip6h_len + 1) << 3;
1176 #else
1177 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1178 			    ip6->ip6_nxt);
1179 			if (ext == NULL) {
1180 				ip6stat.ip6s_tooshort++;
1181 				return;
1182 			}
1183 			hbh = mtod(ext, struct ip6_hbh *);
1184 			hbhlen = (hbh->ip6h_len + 1) << 3;
1185 			if (hbhlen != ext->m_len) {
1186 				m_freem(ext);
1187 				ip6stat.ip6s_tooshort++;
1188 				return;
1189 			}
1190 #endif
1191 
1192 			/*
1193 			 * XXX: We copy the whole header even if a
1194 			 * jumbo payload option is included, the option which
1195 			 * is to be removed before returning according to
1196 			 * RFC 2292.
1197 			 * Note: this constraint is removed in 2292bis.
1198 			 */
1199 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1200 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1201 			    IPPROTO_IPV6);
1202 			if (*mp)
1203 				mp = &(*mp)->m_next;
1204 #ifdef PULLDOWN_TEST
1205 			m_freem(ext);
1206 #endif
1207 		}
1208 	}
1209 
1210 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1211 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1212 		int proto, off, nxt;
1213 
1214 		/*
1215 		 * go through the header chain to see if a routing header is
1216 		 * contained in the packet. We need this information to store
1217 		 * destination options headers (if any) properly.
1218 		 * XXX: performance issue. We should record this info when
1219 		 * processing extension headers in incoming routine.
1220 		 * (todo) use m_aux?
1221 		 */
1222 		proto = IPPROTO_IPV6;
1223 		off = 0;
1224 		nxt = -1;
1225 		while (1) {
1226 			int newoff;
1227 
1228 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1229 			if (newoff < 0)
1230 				break;
1231 			if (newoff < off) /* invalid, check for safety */
1232 				break;
1233 			if ((proto = nxt) == IPPROTO_ROUTING)
1234 				break;
1235 			off = newoff;
1236 		}
1237 	}
1238 
1239 	if ((in6p->in6p_flags &
1240 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1241 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1242 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1243 
1244 		/*
1245 		 * Search for destination options headers or routing
1246 		 * header(s) through the header chain, and stores each
1247 		 * header as ancillary data.
1248 		 * Note that the order of the headers remains in
1249 		 * the chain of ancillary data.
1250 		 */
1251 		while (1) {	/* is explicit loop prevention necessary? */
1252 			struct ip6_ext *ip6e = NULL;
1253 			int elen;
1254 #ifdef PULLDOWN_TEST
1255 			struct mbuf *ext = NULL;
1256 #endif
1257 
1258 			/*
1259 			 * if it is not an extension header, don't try to
1260 			 * pull it from the chain.
1261 			 */
1262 			switch (nxt) {
1263 			case IPPROTO_DSTOPTS:
1264 			case IPPROTO_ROUTING:
1265 			case IPPROTO_HOPOPTS:
1266 			case IPPROTO_AH: /* is it possible? */
1267 				break;
1268 			default:
1269 				goto loopend;
1270 			}
1271 
1272 #ifndef PULLDOWN_TEST
1273 			if (off + sizeof(*ip6e) > m->m_len)
1274 				goto loopend;
1275 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1276 			if (nxt == IPPROTO_AH)
1277 				elen = (ip6e->ip6e_len + 2) << 2;
1278 			else
1279 				elen = (ip6e->ip6e_len + 1) << 3;
1280 			if (off + elen > m->m_len)
1281 				goto loopend;
1282 #else
1283 			ext = ip6_pullexthdr(m, off, nxt);
1284 			if (ext == NULL) {
1285 				ip6stat.ip6s_tooshort++;
1286 				return;
1287 			}
1288 			ip6e = mtod(ext, struct ip6_ext *);
1289 			if (nxt == IPPROTO_AH)
1290 				elen = (ip6e->ip6e_len + 2) << 2;
1291 			else
1292 				elen = (ip6e->ip6e_len + 1) << 3;
1293 			if (elen != ext->m_len) {
1294 				m_freem(ext);
1295 				ip6stat.ip6s_tooshort++;
1296 				return;
1297 			}
1298 #endif
1299 
1300 			switch (nxt) {
1301 			case IPPROTO_DSTOPTS:
1302 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1303 					break;
1304 
1305 				/*
1306 				 * We also require super-user privilege for
1307 				 * the option.  See comments on IN6_HOPOPTS.
1308 				 */
1309 				if (!privileged)
1310 					break;
1311 
1312 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1313 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1314 				    IPPROTO_IPV6);
1315 				if (*mp)
1316 					mp = &(*mp)->m_next;
1317 				break;
1318 			case IPPROTO_ROUTING:
1319 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1320 					break;
1321 
1322 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1323 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1324 				    IPPROTO_IPV6);
1325 				if (*mp)
1326 					mp = &(*mp)->m_next;
1327 				break;
1328 			case IPPROTO_HOPOPTS:
1329 			case IPPROTO_AH: /* is it possible? */
1330 				break;
1331 
1332 			default:
1333 				/*
1334 			 	 * other cases have been filtered in the above.
1335 				 * none will visit this case.  here we supply
1336 				 * the code just in case (nxt overwritten or
1337 				 * other cases).
1338 				 */
1339 #ifdef PULLDOWN_TEST
1340 				m_freem(ext);
1341 #endif
1342 				goto loopend;
1343 
1344 			}
1345 
1346 			/* proceed with the next header. */
1347 			off += elen;
1348 			nxt = ip6e->ip6e_nxt;
1349 			ip6e = NULL;
1350 #ifdef PULLDOWN_TEST
1351 			m_freem(ext);
1352 			ext = NULL;
1353 #endif
1354 		}
1355 	  loopend:
1356 		;
1357 	}
1358 #undef IS2292
1359 }
1360 
1361 void
1362 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1363 {
1364 	struct socket *so;
1365 	struct mbuf *m_mtu;
1366 	struct ip6_mtuinfo mtuctl;
1367 
1368 	so =  in6p->inp_socket;
1369 
1370 	if (mtu == NULL)
1371 		return;
1372 
1373 #ifdef DIAGNOSTIC
1374 	if (so == NULL)		/* I believe this is impossible */
1375 		panic("ip6_notify_pmtu: socket is NULL");
1376 #endif
1377 
1378 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1379 	mtuctl.ip6m_mtu = *mtu;
1380 	mtuctl.ip6m_addr = *dst;
1381 
1382 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1383 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1384 		return;
1385 
1386 	lwkt_gettoken(&so->so_rcv.ssb_token);
1387 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1388 	    == 0) {
1389 		m_freem(m_mtu);
1390 		/* XXX: should count statistics */
1391 	} else {
1392 		sorwakeup(so);
1393 	}
1394 	lwkt_reltoken(&so->so_rcv.ssb_token);
1395 }
1396 
1397 #ifdef PULLDOWN_TEST
1398 /*
1399  * pull single extension header from mbuf chain.  returns single mbuf that
1400  * contains the result, or NULL on error.
1401  */
1402 static struct mbuf *
1403 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1404 {
1405 	struct ip6_ext ip6e;
1406 	size_t elen;
1407 	struct mbuf *n;
1408 
1409 #ifdef DIAGNOSTIC
1410 	switch (nxt) {
1411 	case IPPROTO_DSTOPTS:
1412 	case IPPROTO_ROUTING:
1413 	case IPPROTO_HOPOPTS:
1414 	case IPPROTO_AH: /* is it possible? */
1415 		break;
1416 	default:
1417 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1418 	}
1419 #endif
1420 
1421 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1422 	if (nxt == IPPROTO_AH)
1423 		elen = (ip6e.ip6e_len + 2) << 2;
1424 	else
1425 		elen = (ip6e.ip6e_len + 1) << 3;
1426 
1427 	n = m_getb(elen, M_NOWAIT, MT_DATA, 0);
1428 	if (n == NULL)
1429 		return NULL;
1430 	n->m_len = 0;
1431 
1432 	if (elen >= M_TRAILINGSPACE(n)) {
1433 		m_free(n);
1434 		return NULL;
1435 	}
1436 
1437 	m_copydata(m, off, elen, mtod(n, caddr_t));
1438 	n->m_len = elen;
1439 	return n;
1440 }
1441 #endif
1442 
1443 /*
1444  * Get pointer to the previous header followed by the header
1445  * currently processed.
1446  * XXX: This function supposes that
1447  *	M includes all headers,
1448  *	the next header field and the header length field of each header
1449  *	are valid, and
1450  *	the sum of each header length equals to OFF.
1451  * Because of these assumptions, this function must be called very
1452  * carefully. Moreover, it will not be used in the near future when
1453  * we develop `neater' mechanism to process extension headers.
1454  */
1455 char *
1456 ip6_get_prevhdr(struct mbuf *m, int off)
1457 {
1458 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1459 
1460 	if (off == sizeof(struct ip6_hdr))
1461 		return (&ip6->ip6_nxt);
1462 	else {
1463 		int len, nxt;
1464 		struct ip6_ext *ip6e = NULL;
1465 
1466 		nxt = ip6->ip6_nxt;
1467 		len = sizeof(struct ip6_hdr);
1468 		while (len < off) {
1469 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1470 
1471 			switch (nxt) {
1472 			case IPPROTO_FRAGMENT:
1473 				len += sizeof(struct ip6_frag);
1474 				break;
1475 			case IPPROTO_AH:
1476 				len += (ip6e->ip6e_len + 2) << 2;
1477 				break;
1478 			default:
1479 				len += (ip6e->ip6e_len + 1) << 3;
1480 				break;
1481 			}
1482 			nxt = ip6e->ip6e_nxt;
1483 		}
1484 		if (ip6e)
1485 			return (&ip6e->ip6e_nxt);
1486 		else
1487 			return NULL;
1488 	}
1489 }
1490 
1491 /*
1492  * get next header offset.  m will be retained.
1493  */
1494 int
1495 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1496 {
1497 	struct ip6_hdr ip6;
1498 	struct ip6_ext ip6e;
1499 	struct ip6_frag fh;
1500 
1501 	/* just in case */
1502 	if (m == NULL)
1503 		panic("ip6_nexthdr: m == NULL");
1504 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1505 		return -1;
1506 
1507 	switch (proto) {
1508 	case IPPROTO_IPV6:
1509 		if (m->m_pkthdr.len < off + sizeof(ip6))
1510 			return -1;
1511 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1512 		if (nxtp)
1513 			*nxtp = ip6.ip6_nxt;
1514 		off += sizeof(ip6);
1515 		return off;
1516 
1517 	case IPPROTO_FRAGMENT:
1518 		/*
1519 		 * terminate parsing if it is not the first fragment,
1520 		 * it does not make sense to parse through it.
1521 		 */
1522 		if (m->m_pkthdr.len < off + sizeof(fh))
1523 			return -1;
1524 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1525 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1526 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1527 			return -1;
1528 		if (nxtp)
1529 			*nxtp = fh.ip6f_nxt;
1530 		off += sizeof(struct ip6_frag);
1531 		return off;
1532 
1533 	case IPPROTO_AH:
1534 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1535 			return -1;
1536 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1537 		if (nxtp)
1538 			*nxtp = ip6e.ip6e_nxt;
1539 		off += (ip6e.ip6e_len + 2) << 2;
1540 		return off;
1541 
1542 	case IPPROTO_HOPOPTS:
1543 	case IPPROTO_ROUTING:
1544 	case IPPROTO_DSTOPTS:
1545 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1546 			return -1;
1547 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1548 		if (nxtp)
1549 			*nxtp = ip6e.ip6e_nxt;
1550 		off += (ip6e.ip6e_len + 1) << 3;
1551 		return off;
1552 
1553 	case IPPROTO_NONE:
1554 	case IPPROTO_ESP:
1555 	case IPPROTO_IPCOMP:
1556 		/* give up */
1557 		return -1;
1558 
1559 	default:
1560 		return -1;
1561 	}
1562 
1563 	return -1;
1564 }
1565 
1566 /*
1567  * get offset for the last header in the chain.  m will be kept untainted.
1568  */
1569 int
1570 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1571 {
1572 	int newoff;
1573 	int nxt;
1574 
1575 	if (!nxtp) {
1576 		nxt = -1;
1577 		nxtp = &nxt;
1578 	}
1579 	while (1) {
1580 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1581 		if (newoff < 0)
1582 			return off;
1583 		else if (newoff < off)
1584 			return -1;	/* invalid */
1585 		else if (newoff == off)
1586 			return newoff;
1587 
1588 		off = newoff;
1589 		proto = *nxtp;
1590 	}
1591 }
1592 
1593 struct ip6aux *
1594 ip6_addaux(struct mbuf *m)
1595 {
1596 	struct m_tag *mtag;
1597 
1598 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1599 	if (!mtag) {
1600 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1601 		    M_NOWAIT);
1602 		if (mtag)
1603 			m_tag_prepend(m, mtag);
1604 	}
1605 	if (mtag)
1606 		bzero(m_tag_data(mtag), sizeof(struct ip6aux));
1607 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1608 }
1609 
1610 struct ip6aux *
1611 ip6_findaux(struct mbuf *m)
1612 {
1613 	struct m_tag *mtag;
1614 
1615 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1616 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1617 }
1618 
1619 void
1620 ip6_delaux(struct mbuf *m)
1621 {
1622 	struct m_tag *mtag;
1623 
1624 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1625 	if (mtag)
1626 		m_tag_delete(m, mtag);
1627 }
1628 
1629 /*
1630  * System control for IP6
1631  */
1632 
1633 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1634 	0,		0,		0,		0,
1635 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1636 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1637 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1638 	0,		0,		0,		0,
1639 	ENOPROTOOPT
1640 };
1641