xref: /dragonfly/sys/netinet6/ip6_input.c (revision 5337421c)
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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_ip6fw.h"
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 #include <sys/proc.h>
86 #include <sys/priv.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #include <net/pfil.h>
94 
95 #include <sys/thread2.h>
96 #include <sys/msgport2.h>
97 #include <net/netmsg2.h>
98 #include <net/netisr2.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_systm.h>
102 #ifdef INET
103 #include <netinet/ip.h>
104 #include <netinet/ip_icmp.h>
105 #endif /* INET */
106 #include <netinet/ip6.h>
107 #include <netinet6/in6_var.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/icmp6.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/in6_ifattach.h>
113 #include <netinet6/nd6.h>
114 #include <netinet6/in6_prefix.h>
115 
116 #ifdef IPSEC
117 #include <netinet6/ipsec.h>
118 #ifdef INET6
119 #include <netinet6/ipsec6.h>
120 #endif
121 #endif
122 
123 #ifdef FAST_IPSEC
124 #include <netproto/ipsec/ipsec.h>
125 #include <netproto/ipsec/ipsec6.h>
126 #define	IPSEC
127 #endif /* FAST_IPSEC */
128 
129 #include <net/ip6fw/ip6_fw.h>
130 
131 #include <netinet6/ip6protosw.h>
132 
133 #include <net/net_osdep.h>
134 
135 extern struct domain inet6domain;
136 extern struct protosw inet6sw[];
137 
138 u_char ip6_protox[IPPROTO_MAX];
139 struct in6_ifaddr *in6_ifaddr;
140 
141 extern struct callout in6_tmpaddrtimer_ch;
142 
143 int ip6_forward_srcrt;			/* XXX */
144 int ip6_sourcecheck;			/* XXX */
145 int ip6_sourcecheck_interval;		/* XXX */
146 
147 int ip6_ours_check_algorithm;
148 
149 struct pfil_head inet6_pfil_hook;
150 
151 /* firewall hooks */
152 ip6_fw_chk_t *ip6_fw_chk_ptr;
153 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
154 int ip6_fw_enable = 1;
155 
156 struct ip6stat ip6stat;
157 
158 static void ip6_init2 (void *);
159 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
160 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
161 static void ip6_input(netmsg_t msg);
162 #ifdef PULLDOWN_TEST
163 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
164 #endif
165 static void transport6_processing_handler(netmsg_t netmsg);
166 
167 /*
168  * IP6 initialization: fill in IP6 protocol switch table.
169  * All protocols not implemented in kernel go to raw IP6 protocol handler.
170  */
171 void
172 ip6_init(void)
173 {
174 	struct protosw *pr;
175 	int i;
176 	struct timeval tv;
177 
178 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
179 	if (pr == NULL)
180 		panic("ip6_init");
181 	for (i = 0; i < IPPROTO_MAX; i++)
182 		ip6_protox[i] = pr - inet6sw;
183 	for (pr = inet6domain.dom_protosw;
184 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
185 		if (pr->pr_domain->dom_family == PF_INET6 &&
186 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
187 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
188 
189 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
190 	inet6_pfil_hook.ph_af = AF_INET6;
191 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
192 		kprintf("%s: WARNING: unable to register pfil hook, "
193 			"error %d\n", __func__, i);
194 	}
195 
196 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX cpufn */
197 	scope6_init();
198 	addrsel_policy_init();
199 	nd6_init();
200 	frag6_init();
201 	/*
202 	 * in many cases, random() here does NOT return random number
203 	 * as initialization during bootstrap time occur in fixed order.
204 	 */
205 	microtime(&tv);
206 	ip6_flow_seq = krandom() ^ tv.tv_usec;
207 	microtime(&tv);
208 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
209 }
210 
211 static void
212 ip6_init2(void *dummy)
213 {
214 	/* nd6_timer_init */
215 	callout_init(&nd6_timer_ch);
216 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
217 
218 	/* router renumbering prefix list maintenance */
219 	callout_init(&in6_rr_timer_ch);
220 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
221 
222 	/* timer for regeneranation of temporary addresses randomize ID */
223 	callout_init(&in6_tmpaddrtimer_ch);
224 	callout_reset(&in6_tmpaddrtimer_ch,
225 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
226 		       ip6_temp_regen_advance) * hz,
227 		      in6_tmpaddrtimer, NULL);
228 }
229 
230 /* cheat */
231 /* This must be after route_init(), which is now SI_ORDER_THIRD */
232 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
233 
234 extern struct	route_in6 ip6_forward_rt;
235 
236 static
237 void
238 ip6_input(netmsg_t msg)
239 {
240 	struct mbuf *m = msg->packet.nm_packet;
241 	struct ip6_hdr *ip6;
242 	int off = sizeof(struct ip6_hdr), nest;
243 	u_int32_t plen;
244 	u_int32_t rtalert = ~0;
245 	int nxt, ours = 0, rh_present = 0;
246 	struct ifnet *deliverifp = NULL;
247 	struct in6_addr odst;
248 	int srcrt = 0;
249 
250 #ifdef IPSEC
251 	/*
252 	 * should the inner packet be considered authentic?
253 	 * see comment in ah4_input().
254 	 */
255 	if (m) {
256 		m->m_flags &= ~M_AUTHIPHDR;
257 		m->m_flags &= ~M_AUTHIPDGM;
258 	}
259 #endif
260 
261 	/*
262 	 * make sure we don't have onion peering information into m_aux.
263 	 */
264 	ip6_delaux(m);
265 
266 	/*
267 	 * mbuf statistics
268 	 */
269 	if (m->m_flags & M_EXT) {
270 		if (m->m_next)
271 			ip6stat.ip6s_mext2m++;
272 		else
273 			ip6stat.ip6s_mext1++;
274 	} else {
275 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
276 		if (m->m_next) {
277 			if (m->m_flags & M_LOOP) {
278 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
279 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
280 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
281 			else
282 				ip6stat.ip6s_m2m[0]++;
283 		} else
284 			ip6stat.ip6s_m1++;
285 #undef M2MMAX
286 	}
287 
288 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
289 	ip6stat.ip6s_total++;
290 
291 #ifndef PULLDOWN_TEST
292 	/*
293 	 * L2 bridge code and some other code can return mbuf chain
294 	 * that does not conform to KAME requirement.  too bad.
295 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
296 	 */
297 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
298 		struct mbuf *n;
299 
300 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
301 		if (n == NULL)
302 			goto bad;
303 		M_MOVE_PKTHDR(n, m);
304 
305 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
306 		n->m_len = n->m_pkthdr.len;
307 		m_freem(m);
308 		m = n;
309 	}
310 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
311 #endif
312 
313 	if (m->m_len < sizeof(struct ip6_hdr)) {
314 		struct ifnet *inifp;
315 		inifp = m->m_pkthdr.rcvif;
316 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
317 			ip6stat.ip6s_toosmall++;
318 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
319 			goto bad2;
320 		}
321 	}
322 
323 	ip6 = mtod(m, struct ip6_hdr *);
324 
325 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
326 		ip6stat.ip6s_badvers++;
327 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
328 		goto bad;
329 	}
330 
331 	/*
332 	 * Run through list of hooks for input packets.
333 	 *
334 	 * NB: Beware of the destination address changing
335 	 *     (e.g. by NAT rewriting). When this happens,
336 	 *     tell ip6_forward to do the right thing.
337 	 */
338 	if (pfil_has_hooks(&inet6_pfil_hook)) {
339 		odst = ip6->ip6_dst;
340 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
341 		    m->m_pkthdr.rcvif, PFIL_IN)) {
342 			goto bad2;
343 		}
344 		if (m == NULL)			/* consumed by filter */
345 			goto bad2;
346 		ip6 = mtod(m, struct ip6_hdr *);
347 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
348 	}
349 
350 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
351 
352 #ifdef ALTQ
353 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
354 		/* packet is dropped by traffic conditioner */
355 		return;
356 	}
357 #endif
358 
359 	/*
360 	 * Check with the firewall...
361 	 */
362 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
363 		u_short port = 0;
364 		/* If ipfw says divert, we have to just drop packet */
365 		/* use port as a dummy argument */
366 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
367 			m_freem(m);
368 			m = NULL;
369 		}
370 		if (!m)
371 			goto bad2;
372 	}
373 
374 	/*
375 	 * Check against address spoofing/corruption.
376 	 */
377 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
378 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
379 		/*
380 		 * XXX: "badscope" is not very suitable for a multicast source.
381 		 */
382 		ip6stat.ip6s_badscope++;
383 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
384 		goto bad;
385 	}
386 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
387 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
388 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
389 		ip6stat.ip6s_badscope++;
390 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
391 		goto bad;
392 	}
393 
394 	/*
395 	 * The following check is not documented in specs.  A malicious
396 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
397 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
398 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
399 	 *
400 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
401 	 * support IPv4-less kernel compilation, we cannot support SIIT
402 	 * environment at all.  So, it makes more sense for us to reject any
403 	 * malicious packets for non-SIIT environment, than try to do a
404 	 * partical support for SIIT environment.
405 	 */
406 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
407 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
408 		ip6stat.ip6s_badscope++;
409 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
410 		goto bad;
411 	}
412 #if 0
413 	/*
414 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
415 	 *
416 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
417 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
418 	 * is revised to forbid relaying case.
419 	 */
420 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
421 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
422 		ip6stat.ip6s_badscope++;
423 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
424 		goto bad;
425 	}
426 #endif
427 
428 	/* drop packets if interface ID portion is already filled */
429 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
430 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
431 		    ip6->ip6_src.s6_addr16[1]) {
432 			ip6stat.ip6s_badscope++;
433 			goto bad;
434 		}
435 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
436 		    ip6->ip6_dst.s6_addr16[1]) {
437 			ip6stat.ip6s_badscope++;
438 			goto bad;
439 		}
440 	}
441 
442 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
443 		ip6->ip6_src.s6_addr16[1]
444 			= htons(m->m_pkthdr.rcvif->if_index);
445 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
446 		ip6->ip6_dst.s6_addr16[1]
447 			= htons(m->m_pkthdr.rcvif->if_index);
448 
449 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
450 	/*
451 	 * We use rt->rt_ifp to determine if the address is ours or not.
452 	 * If rt_ifp is lo0, the address is ours.
453 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
454 	 * so any address under fe80::%lo0/64 will be mistakenly considered
455 	 * local.  The special case is supplied to handle the case properly
456 	 * by actually looking at interface addresses
457 	 * (using in6ifa_ifpwithaddr).
458 	 */
459 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
460 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
461 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
462 			icmp6_error(m, ICMP6_DST_UNREACH,
463 			    ICMP6_DST_UNREACH_ADDR, 0);
464 			/* m is already freed */
465 			goto bad2;
466 		}
467 
468 		ours = 1;
469 		deliverifp = m->m_pkthdr.rcvif;
470 		goto hbhcheck;
471 	}
472 #endif
473 
474 	/*
475 	 * Multicast check
476 	 */
477 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
478 		struct	in6_multi *in6m = NULL;
479 
480 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
481 		/*
482 		 * See if we belong to the destination multicast group on the
483 		 * arrival interface.
484 		 */
485 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
486 		if (in6m)
487 			ours = 1;
488 		else if (!ip6_mrouter) {
489 			ip6stat.ip6s_notmember++;
490 			ip6stat.ip6s_cantforward++;
491 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
492 			goto bad;
493 		}
494 		deliverifp = m->m_pkthdr.rcvif;
495 		goto hbhcheck;
496 	}
497 
498 	/*
499 	 *  Unicast check
500 	 */
501 	switch (ip6_ours_check_algorithm) {
502 	default:
503 		/*
504 		 * XXX: I intentionally broke our indentation rule here,
505 		 *      since this switch-case is just for measurement and
506 		 *      therefore should soon be removed.
507 		 */
508 	if (ip6_forward_rt.ro_rt != NULL &&
509 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
510 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
511 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
512 		ip6stat.ip6s_forward_cachehit++;
513 	else {
514 		struct sockaddr_in6 *dst6;
515 
516 		if (ip6_forward_rt.ro_rt) {
517 			/* route is down or destination is different */
518 			ip6stat.ip6s_forward_cachemiss++;
519 			RTFREE(ip6_forward_rt.ro_rt);
520 			ip6_forward_rt.ro_rt = 0;
521 		}
522 
523 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
524 		dst6 = &ip6_forward_rt.ro_dst;
525 		dst6->sin6_len = sizeof(struct sockaddr_in6);
526 		dst6->sin6_family = AF_INET6;
527 		dst6->sin6_addr = ip6->ip6_dst;
528 
529 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
530 	}
531 
532 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
533 
534 	/*
535 	 * Accept the packet if the forwarding interface to the destination
536 	 * according to the routing table is the loopback interface,
537 	 * unless the associated route has a gateway.
538 	 * Note that this approach causes to accept a packet if there is a
539 	 * route to the loopback interface for the destination of the packet.
540 	 * But we think it's even useful in some situations, e.g. when using
541 	 * a special daemon which wants to intercept the packet.
542 	 *
543 	 * XXX: some OSes automatically make a cloned route for the destination
544 	 * of an outgoing packet.  If the outgoing interface of the packet
545 	 * is a loopback one, the kernel would consider the packet to be
546 	 * accepted, even if we have no such address assinged on the interface.
547 	 * We check the cloned flag of the route entry to reject such cases,
548 	 * assuming that route entries for our own addresses are not made by
549 	 * cloning (it should be true because in6_addloop explicitly installs
550 	 * the host route).  However, we might have to do an explicit check
551 	 * while it would be less efficient.  Or, should we rather install a
552 	 * reject route for such a case?
553 	 */
554 	if (ip6_forward_rt.ro_rt &&
555 	    (ip6_forward_rt.ro_rt->rt_flags &
556 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
557 #ifdef RTF_WASCLONED
558 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
559 #endif
560 #ifdef RTF_CLONED
561 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
562 #endif
563 #if 0
564 	    /*
565 	     * The check below is redundant since the comparison of
566 	     * the destination and the key of the rtentry has
567 	     * already done through looking up the routing table.
568 	     */
569 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
570 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
571 #endif
572 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
573 		struct in6_ifaddr *ia6 =
574 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
575 
576 		/*
577 		 * record address information into m_aux.
578 		 */
579 		ip6_setdstifaddr(m, ia6);
580 
581 		/*
582 		 * packets to a tentative, duplicated, or somehow invalid
583 		 * address must not be accepted.
584 		 */
585 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
586 			/* this address is ready */
587 			ours = 1;
588 			deliverifp = ia6->ia_ifp;	/* correct? */
589 			/* Count the packet in the ip address stats */
590 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
591 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
592 			goto hbhcheck;
593 		} else {
594 			/* address is not ready, so discard the packet. */
595 			nd6log((LOG_INFO,
596 			    "ip6_input: packet to an unready address %s->%s\n",
597 			    ip6_sprintf(&ip6->ip6_src),
598 			    ip6_sprintf(&ip6->ip6_dst)));
599 
600 			goto bad;
601 		}
602 	}
603 	} /* XXX indentation (see above) */
604 
605 	/*
606 	 * FAITH(Firewall Aided Internet Translator)
607 	 */
608 	if (ip6_keepfaith) {
609 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
610 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
611 			/* XXX do we need more sanity checks? */
612 			ours = 1;
613 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
614 			goto hbhcheck;
615 		}
616 	}
617 
618 	/*
619 	 * Now there is no reason to process the packet if it's not our own
620 	 * and we're not a router.
621 	 */
622 	if (!ip6_forwarding) {
623 		ip6stat.ip6s_cantforward++;
624 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
625 		goto bad;
626 	}
627 
628 hbhcheck:
629 	/*
630 	 * record address information into m_aux, if we don't have one yet.
631 	 * note that we are unable to record it, if the address is not listed
632 	 * as our interface address (e.g. multicast addresses, addresses
633 	 * within FAITH prefixes and such).
634 	 */
635 	if (deliverifp && !ip6_getdstifaddr(m)) {
636 		struct in6_ifaddr *ia6;
637 
638 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
639 		if (ia6) {
640 			if (!ip6_setdstifaddr(m, ia6)) {
641 				/*
642 				 * XXX maybe we should drop the packet here,
643 				 * as we could not provide enough information
644 				 * to the upper layers.
645 				 */
646 			}
647 		}
648 	}
649 
650 	/*
651 	 * Process Hop-by-Hop options header if it's contained.
652 	 * m may be modified in ip6_hopopts_input().
653 	 * If a JumboPayload option is included, plen will also be modified.
654 	 */
655 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
656 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
657 		struct ip6_hbh *hbh;
658 
659 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
660 #if 0	/*touches NULL pointer*/
661 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
662 #endif
663 			goto bad2;	/* m have already been freed */
664 		}
665 
666 		/* adjust pointer */
667 		ip6 = mtod(m, struct ip6_hdr *);
668 
669 		/*
670 		 * if the payload length field is 0 and the next header field
671 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
672 		 * option MUST be included.
673 		 */
674 		if (ip6->ip6_plen == 0 && plen == 0) {
675 			/*
676 			 * Note that if a valid jumbo payload option is
677 			 * contained, ip6_hoptops_input() must set a valid
678 			 * (non-zero) payload length to the variable plen.
679 			 */
680 			ip6stat.ip6s_badoptions++;
681 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
682 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
683 			icmp6_error(m, ICMP6_PARAM_PROB,
684 				    ICMP6_PARAMPROB_HEADER,
685 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
686 			goto bad2;
687 		}
688 #ifndef PULLDOWN_TEST
689 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
690 		hbh = (struct ip6_hbh *)(ip6 + 1);
691 #else
692 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
693 			sizeof(struct ip6_hbh));
694 		if (hbh == NULL) {
695 			ip6stat.ip6s_tooshort++;
696 			goto bad2;
697 		}
698 #endif
699 		nxt = hbh->ip6h_nxt;
700 
701 		/*
702 		 * If we are acting as a router and the packet contains a
703 		 * router alert option, see if we know the option value.
704 		 * Currently, we only support the option value for MLD, in which
705 		 * case we should pass the packet to the multicast routing
706 		 * daemon.
707 		 */
708 		if (rtalert != ~0 && ip6_forwarding) {
709 			switch (rtalert) {
710 			case IP6OPT_RTALERT_MLD:
711 				ours = 1;
712 				break;
713 			default:
714 				/*
715 				 * RFC2711 requires unrecognized values must be
716 				 * silently ignored.
717 				 */
718 				break;
719 			}
720 		}
721 	} else
722 		nxt = ip6->ip6_nxt;
723 
724 	/*
725 	 * Check that the amount of data in the buffers
726 	 * is as at least much as the IPv6 header would have us expect.
727 	 * Trim mbufs if longer than we expect.
728 	 * Drop packet if shorter than we expect.
729 	 */
730 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
731 		ip6stat.ip6s_tooshort++;
732 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
733 		goto bad;
734 	}
735 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
736 		if (m->m_len == m->m_pkthdr.len) {
737 			m->m_len = sizeof(struct ip6_hdr) + plen;
738 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
739 		} else
740 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
741 	}
742 
743 	/*
744 	 * Forward if desirable.
745 	 */
746 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
747 		/*
748 		 * If we are acting as a multicast router, all
749 		 * incoming multicast packets are passed to the
750 		 * kernel-level multicast forwarding function.
751 		 * The packet is returned (relatively) intact; if
752 		 * ip6_mforward() returns a non-zero value, the packet
753 		 * must be discarded, else it may be accepted below.
754 		 */
755 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
756 			ip6stat.ip6s_cantforward++;
757 			goto bad;
758 		}
759 		if (!ours)
760 			goto bad;
761 	} else if (!ours) {
762 		ip6_forward(m, srcrt);
763 		goto bad2;
764 	}
765 
766 	ip6 = mtod(m, struct ip6_hdr *);
767 
768 	/*
769 	 * Malicious party may be able to use IPv4 mapped addr to confuse
770 	 * tcp/udp stack and bypass security checks (act as if it was from
771 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
772 	 *
773 	 * For SIIT end node behavior, you may want to disable the check.
774 	 * However, you will  become vulnerable to attacks using IPv4 mapped
775 	 * source.
776 	 */
777 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
778 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
779 		ip6stat.ip6s_badscope++;
780 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
781 		goto bad;
782 	}
783 
784 	/*
785 	 * Tell launch routine the next header
786 	 */
787 	ip6stat.ip6s_delivered++;
788 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
789 	nest = 0;
790 
791 	rh_present = 0;
792 	while (nxt != IPPROTO_DONE) {
793 		struct protosw *sw6;
794 
795 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
796 			ip6stat.ip6s_toomanyhdr++;
797 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
798 			goto bad;
799 		}
800 
801 		/*
802 		 * protection against faulty packet - there should be
803 		 * more sanity checks in header chain processing.
804 		 */
805 		if (m->m_pkthdr.len < off) {
806 			ip6stat.ip6s_tooshort++;
807 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
808 			goto bad;
809 		}
810 
811 #if 0
812 		/*
813 		 * do we need to do it for every header?  yeah, other
814 		 * functions can play with it (like re-allocate and copy).
815 		 */
816 		mhist = ip6_addaux(m);
817 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
818 			hist = mtod(mhist, caddr_t) + mhist->m_len;
819 			bcopy(&nxt, hist, sizeof(nxt));
820 			mhist->m_len += sizeof(nxt);
821 		} else {
822 			ip6stat.ip6s_toomanyhdr++;
823 			goto bad;
824 		}
825 #endif
826 
827 		if (nxt == IPPROTO_ROUTING) {
828 			if (rh_present++) {
829 				in6_ifstat_inc(m->m_pkthdr.rcvif,
830 				    ifs6_in_hdrerr);
831 				ip6stat.ip6s_badoptions++;
832 				goto bad;
833 			}
834 		}
835 
836 		sw6 = &inet6sw[ip6_protox[nxt]];
837 #ifdef IPSEC
838 		/*
839 		 * enforce IPsec policy checking if we are seeing last header.
840 		 * note that we do not visit this with protocols with pcb layer
841 		 * code - like udp/tcp/raw ip.
842 		 */
843 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
844 			ipsec6stat.in_polvio++;
845 			goto bad;
846 		}
847 #endif
848 		/*
849 		 * If this is a terminal header forward to the port, otherwise
850 		 * process synchronously for more headers.
851 		 */
852 		if (sw6->pr_flags & PR_LASTHDR) {
853 			struct netmsg_packet *pmsg;
854 			lwkt_port_t port;
855 
856 			port = netisr_cpuport(0); /* XXX */
857 			KKASSERT(port != NULL);
858 			pmsg = &m->m_hdr.mh_netmsg;
859 			netmsg_init(&pmsg->base, NULL,
860 				    &netisr_apanic_rport,
861 				    0, transport6_processing_handler);
862 			pmsg->nm_packet = m;
863 			pmsg->nm_nxt = nxt;
864 			pmsg->base.lmsg.u.ms_result = off;
865 			lwkt_sendmsg(port, &pmsg->base.lmsg);
866 			/* done with m */
867 			nxt = IPPROTO_DONE;
868 		} else {
869 			nxt = sw6->pr_input(&m, &off, nxt);
870 		}
871 	}
872 	goto bad2;
873 bad:
874 	m_freem(m);
875 bad2:
876 	;
877 	/* msg was embedded in the mbuf, do not reply! */
878 }
879 
880 /*
881  * We have to call the pr_input() function from the correct protocol
882  * thread.  The sw6->pr_soport() request at the end of ip6_input()
883  * returns the port and we forward a netmsg to the port to execute
884  * this function.
885  */
886 static void
887 transport6_processing_handler(netmsg_t netmsg)
888 {
889 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
890 	struct protosw *sw6;
891 	int hlen;
892 	int nxt;
893 
894 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
895 	hlen = pmsg->base.lmsg.u.ms_result;
896 
897 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
898 	KKASSERT(nxt == IPPROTO_DONE);
899 	/* netmsg was embedded in the mbuf, do not reply! */
900 }
901 
902 /*
903  * set/grab in6_ifaddr correspond to IPv6 destination address.
904  * XXX backward compatibility wrapper
905  */
906 static struct ip6aux *
907 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
908 {
909 	struct ip6aux *n;
910 
911 	n = ip6_addaux(m);
912 	if (n)
913 		n->ip6a_dstia6 = ia6;
914 	return n;	/* NULL if failed to set */
915 }
916 
917 struct in6_ifaddr *
918 ip6_getdstifaddr(struct mbuf *m)
919 {
920 	struct ip6aux *n;
921 
922 	n = ip6_findaux(m);
923 	if (n)
924 		return n->ip6a_dstia6;
925 	else
926 		return NULL;
927 }
928 
929 /*
930  * Hop-by-Hop options header processing. If a valid jumbo payload option is
931  * included, the real payload length will be stored in plenp.
932  */
933 static int
934 ip6_hopopts_input(u_int32_t *plenp,
935 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
936 		  struct mbuf **mp,
937 		  int *offp)
938 {
939 	struct mbuf *m = *mp;
940 	int off = *offp, hbhlen;
941 	struct ip6_hbh *hbh;
942 
943 	/* validation of the length of the header */
944 #ifndef PULLDOWN_TEST
945 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
946 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
947 	hbhlen = (hbh->ip6h_len + 1) << 3;
948 
949 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
950 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
951 #else
952 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
953 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
954 	if (hbh == NULL) {
955 		ip6stat.ip6s_tooshort++;
956 		return -1;
957 	}
958 	hbhlen = (hbh->ip6h_len + 1) << 3;
959 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
960 		hbhlen);
961 	if (hbh == NULL) {
962 		ip6stat.ip6s_tooshort++;
963 		return -1;
964 	}
965 #endif
966 	off += hbhlen;
967 	hbhlen -= sizeof(struct ip6_hbh);
968 
969 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
970 				hbhlen, rtalertp, plenp) < 0)
971 		return (-1);
972 
973 	*offp = off;
974 	*mp = m;
975 	return (0);
976 }
977 
978 /*
979  * Search header for all Hop-by-hop options and process each option.
980  * This function is separate from ip6_hopopts_input() in order to
981  * handle a case where the sending node itself process its hop-by-hop
982  * options header. In such a case, the function is called from ip6_output().
983  *
984  * The function assumes that hbh header is located right after the IPv6 header
985  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
986  * opthead + hbhlen is located in continuous memory region.
987  */
988 int
989 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
990 		    u_int32_t *rtalertp, u_int32_t *plenp)
991 {
992 	struct ip6_hdr *ip6;
993 	int optlen = 0;
994 	u_int8_t *opt = opthead;
995 	u_int16_t rtalert_val;
996 	u_int32_t jumboplen;
997 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
998 
999 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1000 		switch (*opt) {
1001 		case IP6OPT_PAD1:
1002 			optlen = 1;
1003 			break;
1004 		case IP6OPT_PADN:
1005 			if (hbhlen < IP6OPT_MINLEN) {
1006 				ip6stat.ip6s_toosmall++;
1007 				goto bad;
1008 			}
1009 			optlen = *(opt + 1) + 2;
1010 			break;
1011 		case IP6OPT_RTALERT:
1012 			/* XXX may need check for alignment */
1013 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1014 				ip6stat.ip6s_toosmall++;
1015 				goto bad;
1016 			}
1017 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1018 				/* XXX stat */
1019 				icmp6_error(m, ICMP6_PARAM_PROB,
1020 					    ICMP6_PARAMPROB_HEADER,
1021 					    erroff + opt + 1 - opthead);
1022 				return (-1);
1023 			}
1024 			optlen = IP6OPT_RTALERT_LEN;
1025 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1026 			*rtalertp = ntohs(rtalert_val);
1027 			break;
1028 		case IP6OPT_JUMBO:
1029 			/* XXX may need check for alignment */
1030 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1031 				ip6stat.ip6s_toosmall++;
1032 				goto bad;
1033 			}
1034 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1035 				/* XXX stat */
1036 				icmp6_error(m, ICMP6_PARAM_PROB,
1037 					    ICMP6_PARAMPROB_HEADER,
1038 					    erroff + opt + 1 - opthead);
1039 				return (-1);
1040 			}
1041 			optlen = IP6OPT_JUMBO_LEN;
1042 
1043 			/*
1044 			 * IPv6 packets that have non 0 payload length
1045 			 * must not contain a jumbo payload option.
1046 			 */
1047 			ip6 = mtod(m, struct ip6_hdr *);
1048 			if (ip6->ip6_plen) {
1049 				ip6stat.ip6s_badoptions++;
1050 				icmp6_error(m, ICMP6_PARAM_PROB,
1051 					    ICMP6_PARAMPROB_HEADER,
1052 					    erroff + opt - opthead);
1053 				return (-1);
1054 			}
1055 
1056 			/*
1057 			 * We may see jumbolen in unaligned location, so
1058 			 * we'd need to perform bcopy().
1059 			 */
1060 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1061 			jumboplen = (u_int32_t)htonl(jumboplen);
1062 
1063 #if 1
1064 			/*
1065 			 * if there are multiple jumbo payload options,
1066 			 * *plenp will be non-zero and the packet will be
1067 			 * rejected.
1068 			 * the behavior may need some debate in ipngwg -
1069 			 * multiple options does not make sense, however,
1070 			 * there's no explicit mention in specification.
1071 			 */
1072 			if (*plenp != 0) {
1073 				ip6stat.ip6s_badoptions++;
1074 				icmp6_error(m, ICMP6_PARAM_PROB,
1075 					    ICMP6_PARAMPROB_HEADER,
1076 					    erroff + opt + 2 - opthead);
1077 				return (-1);
1078 			}
1079 #endif
1080 
1081 			/*
1082 			 * jumbo payload length must be larger than 65535.
1083 			 */
1084 			if (jumboplen <= IPV6_MAXPACKET) {
1085 				ip6stat.ip6s_badoptions++;
1086 				icmp6_error(m, ICMP6_PARAM_PROB,
1087 					    ICMP6_PARAMPROB_HEADER,
1088 					    erroff + opt + 2 - opthead);
1089 				return (-1);
1090 			}
1091 			*plenp = jumboplen;
1092 
1093 			break;
1094 		default:		/* unknown option */
1095 			if (hbhlen < IP6OPT_MINLEN) {
1096 				ip6stat.ip6s_toosmall++;
1097 				goto bad;
1098 			}
1099 			optlen = ip6_unknown_opt(opt, m,
1100 			    erroff + opt - opthead);
1101 			if (optlen == -1)
1102 				return (-1);
1103 			optlen += 2;
1104 			break;
1105 		}
1106 	}
1107 
1108 	return (0);
1109 
1110 bad:
1111 	m_freem(m);
1112 	return (-1);
1113 }
1114 
1115 /*
1116  * Unknown option processing.
1117  * The third argument `off' is the offset from the IPv6 header to the option,
1118  * which is necessary if the IPv6 header the and option header and IPv6 header
1119  * is not continuous in order to return an ICMPv6 error.
1120  */
1121 int
1122 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1123 {
1124 	struct ip6_hdr *ip6;
1125 
1126 	switch (IP6OPT_TYPE(*optp)) {
1127 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1128 		return ((int)*(optp + 1));
1129 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1130 		m_freem(m);
1131 		return (-1);
1132 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1133 		ip6stat.ip6s_badoptions++;
1134 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1135 		return (-1);
1136 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1137 		ip6stat.ip6s_badoptions++;
1138 		ip6 = mtod(m, struct ip6_hdr *);
1139 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1140 		    (m->m_flags & (M_BCAST|M_MCAST)))
1141 			m_freem(m);
1142 		else
1143 			icmp6_error(m, ICMP6_PARAM_PROB,
1144 				    ICMP6_PARAMPROB_OPTION, off);
1145 		return (-1);
1146 	}
1147 
1148 	m_freem(m);		/* XXX: NOTREACHED */
1149 	return (-1);
1150 }
1151 
1152 /*
1153  * Create the "control" list for this pcb.
1154  * The function will not modify mbuf chain at all.
1155  *
1156  * with KAME mbuf chain restriction:
1157  * The routine will be called from upper layer handlers like tcp6_input().
1158  * Thus the routine assumes that the caller (tcp6_input) have already
1159  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1160  * very first mbuf on the mbuf chain.
1161  */
1162 void
1163 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1164 		struct mbuf *m)
1165 {
1166 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1167 	struct thread *td = curthread;	/* XXX */
1168 	int privileged = 0;
1169 
1170 
1171 	if (priv_check(td, PRIV_ROOT) == 0)
1172  		privileged++;
1173 
1174 #ifdef SO_TIMESTAMP
1175 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1176 		struct timeval tv;
1177 
1178 		microtime(&tv);
1179 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1180 				      SCM_TIMESTAMP, SOL_SOCKET);
1181 		if (*mp) {
1182 			mp = &(*mp)->m_next;
1183 		}
1184 	}
1185 #endif
1186 
1187 	/* RFC 2292 sec. 5 */
1188 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1189 		struct in6_pktinfo pi6;
1190 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1191 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1192 			pi6.ipi6_addr.s6_addr16[1] = 0;
1193 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1194 					? m->m_pkthdr.rcvif->if_index
1195 					: 0;
1196 		*mp = sbcreatecontrol((caddr_t) &pi6,
1197 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1198 			IPPROTO_IPV6);
1199 		if (*mp)
1200 			mp = &(*mp)->m_next;
1201 	}
1202 
1203 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1204 		int hlim = ip6->ip6_hlim & 0xff;
1205 		*mp = sbcreatecontrol((caddr_t) &hlim,
1206 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1207 		if (*mp)
1208 			mp = &(*mp)->m_next;
1209 	}
1210 
1211 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1212 		u_int32_t flowinfo;
1213 		int tclass;
1214 
1215 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1216 		flowinfo >>= 20;
1217 
1218 		tclass = flowinfo & 0xff;
1219 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1220 		    IPV6_TCLASS, IPPROTO_IPV6);
1221 		if (*mp)
1222 			mp = &(*mp)->m_next;
1223 	}
1224 
1225 	/*
1226 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1227 	 * for the option, but it might be too strict, since there might
1228 	 * be some hop-by-hop options which can be returned to normal user.
1229 	 * See RFC 2292 section 6.
1230 	 */
1231 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1232 		/*
1233 		 * Check if a hop-by-hop options header is contatined in the
1234 		 * received packet, and if so, store the options as ancillary
1235 		 * data. Note that a hop-by-hop options header must be
1236 		 * just after the IPv6 header, which fact is assured through
1237 		 * the IPv6 input processing.
1238 		 */
1239 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1240 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1241 			struct ip6_hbh *hbh;
1242 			int hbhlen = 0;
1243 #ifdef PULLDOWN_TEST
1244 			struct mbuf *ext;
1245 #endif
1246 
1247 #ifndef PULLDOWN_TEST
1248 			hbh = (struct ip6_hbh *)(ip6 + 1);
1249 			hbhlen = (hbh->ip6h_len + 1) << 3;
1250 #else
1251 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1252 			    ip6->ip6_nxt);
1253 			if (ext == NULL) {
1254 				ip6stat.ip6s_tooshort++;
1255 				return;
1256 			}
1257 			hbh = mtod(ext, struct ip6_hbh *);
1258 			hbhlen = (hbh->ip6h_len + 1) << 3;
1259 			if (hbhlen != ext->m_len) {
1260 				m_freem(ext);
1261 				ip6stat.ip6s_tooshort++;
1262 				return;
1263 			}
1264 #endif
1265 
1266 			/*
1267 			 * XXX: We copy whole the header even if a jumbo
1268 			 * payload option is included, which option is to
1269 			 * be removed before returning in the RFC 2292.
1270 			 * Note: this constraint is removed in 2292bis.
1271 			 */
1272 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1273 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1274 			if (*mp)
1275 				mp = &(*mp)->m_next;
1276 #ifdef PULLDOWN_TEST
1277 			m_freem(ext);
1278 #endif
1279 		}
1280 	}
1281 
1282 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1283 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1284 		int proto, off, nxt;
1285 
1286 		/*
1287 		 * go through the header chain to see if a routing header is
1288 		 * contained in the packet. We need this information to store
1289 		 * destination options headers (if any) properly.
1290 		 * XXX: performance issue. We should record this info when
1291 		 * processing extension headers in incoming routine.
1292 		 * (todo) use m_aux?
1293 		 */
1294 		proto = IPPROTO_IPV6;
1295 		off = 0;
1296 		nxt = -1;
1297 		while (1) {
1298 			int newoff;
1299 
1300 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1301 			if (newoff < 0)
1302 				break;
1303 			if (newoff < off) /* invalid, check for safety */
1304 				break;
1305 			if ((proto = nxt) == IPPROTO_ROUTING)
1306 				break;
1307 			off = newoff;
1308 		}
1309 	}
1310 
1311 	if ((in6p->in6p_flags &
1312 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1313 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1314 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1315 
1316 		/*
1317 		 * Search for destination options headers or routing
1318 		 * header(s) through the header chain, and stores each
1319 		 * header as ancillary data.
1320 		 * Note that the order of the headers remains in
1321 		 * the chain of ancillary data.
1322 		 */
1323 		while (1) {	/* is explicit loop prevention necessary? */
1324 			struct ip6_ext *ip6e = NULL;
1325 			int elen;
1326 #ifdef PULLDOWN_TEST
1327 			struct mbuf *ext = NULL;
1328 #endif
1329 
1330 			/*
1331 			 * if it is not an extension header, don't try to
1332 			 * pull it from the chain.
1333 			 */
1334 			switch (nxt) {
1335 			case IPPROTO_DSTOPTS:
1336 			case IPPROTO_ROUTING:
1337 			case IPPROTO_HOPOPTS:
1338 			case IPPROTO_AH: /* is it possible? */
1339 				break;
1340 			default:
1341 				goto loopend;
1342 			}
1343 
1344 #ifndef PULLDOWN_TEST
1345 			if (off + sizeof(*ip6e) > m->m_len)
1346 				goto loopend;
1347 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1348 			if (nxt == IPPROTO_AH)
1349 				elen = (ip6e->ip6e_len + 2) << 2;
1350 			else
1351 				elen = (ip6e->ip6e_len + 1) << 3;
1352 			if (off + elen > m->m_len)
1353 				goto loopend;
1354 #else
1355 			ext = ip6_pullexthdr(m, off, nxt);
1356 			if (ext == NULL) {
1357 				ip6stat.ip6s_tooshort++;
1358 				return;
1359 			}
1360 			ip6e = mtod(ext, struct ip6_ext *);
1361 			if (nxt == IPPROTO_AH)
1362 				elen = (ip6e->ip6e_len + 2) << 2;
1363 			else
1364 				elen = (ip6e->ip6e_len + 1) << 3;
1365 			if (elen != ext->m_len) {
1366 				m_freem(ext);
1367 				ip6stat.ip6s_tooshort++;
1368 				return;
1369 			}
1370 #endif
1371 
1372 			switch (nxt) {
1373 			case IPPROTO_DSTOPTS:
1374 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1375 					break;
1376 
1377 				/*
1378 				 * We also require super-user privilege for
1379 				 * the option.
1380 				 * See the comments on IN6_HOPOPTS.
1381 				 */
1382 				if (!privileged)
1383 					break;
1384 
1385 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1386 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1387 						      IPPROTO_IPV6);
1388 				if (*mp)
1389 					mp = &(*mp)->m_next;
1390 				break;
1391 			case IPPROTO_ROUTING:
1392 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1393 					break;
1394 
1395 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1396 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1397 						      IPPROTO_IPV6);
1398 				if (*mp)
1399 					mp = &(*mp)->m_next;
1400 				break;
1401 			case IPPROTO_HOPOPTS:
1402 			case IPPROTO_AH: /* is it possible? */
1403 				break;
1404 
1405 			default:
1406 				/*
1407 			 	 * other cases have been filtered in the above.
1408 				 * none will visit this case.  here we supply
1409 				 * the code just in case (nxt overwritten or
1410 				 * other cases).
1411 				 */
1412 #ifdef PULLDOWN_TEST
1413 				m_freem(ext);
1414 #endif
1415 				goto loopend;
1416 
1417 			}
1418 
1419 			/* proceed with the next header. */
1420 			off += elen;
1421 			nxt = ip6e->ip6e_nxt;
1422 			ip6e = NULL;
1423 #ifdef PULLDOWN_TEST
1424 			m_freem(ext);
1425 			ext = NULL;
1426 #endif
1427 		}
1428 	  loopend:
1429 		;
1430 	}
1431 #undef IS2292
1432 }
1433 
1434 void
1435 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1436 {
1437 	struct socket *so;
1438 	struct mbuf *m_mtu;
1439 	struct ip6_mtuinfo mtuctl;
1440 
1441 	so =  in6p->inp_socket;
1442 
1443 	if (mtu == NULL)
1444 		return;
1445 
1446 #ifdef DIAGNOSTIC
1447 	if (so == NULL)		/* I believe this is impossible */
1448 		panic("ip6_notify_pmtu: socket is NULL");
1449 #endif
1450 
1451 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1452 	mtuctl.ip6m_mtu = *mtu;
1453 	mtuctl.ip6m_addr = *dst;
1454 
1455 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1456 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1457 		return;
1458 
1459 	lwkt_gettoken(&so->so_rcv.ssb_token);
1460 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1461 	    == 0) {
1462 		m_freem(m_mtu);
1463 		/* XXX: should count statistics */
1464 	} else {
1465 		sorwakeup(so);
1466 	}
1467 	lwkt_reltoken(&so->so_rcv.ssb_token);
1468 }
1469 
1470 #ifdef PULLDOWN_TEST
1471 /*
1472  * pull single extension header from mbuf chain.  returns single mbuf that
1473  * contains the result, or NULL on error.
1474  */
1475 static struct mbuf *
1476 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1477 {
1478 	struct ip6_ext ip6e;
1479 	size_t elen;
1480 	struct mbuf *n;
1481 
1482 #ifdef DIAGNOSTIC
1483 	switch (nxt) {
1484 	case IPPROTO_DSTOPTS:
1485 	case IPPROTO_ROUTING:
1486 	case IPPROTO_HOPOPTS:
1487 	case IPPROTO_AH: /* is it possible? */
1488 		break;
1489 	default:
1490 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1491 	}
1492 #endif
1493 
1494 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1495 	if (nxt == IPPROTO_AH)
1496 		elen = (ip6e.ip6e_len + 2) << 2;
1497 	else
1498 		elen = (ip6e.ip6e_len + 1) << 3;
1499 
1500 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1501 	if (n == NULL)
1502 		return NULL;
1503 	n->m_len = 0;
1504 
1505 	if (elen >= M_TRAILINGSPACE(n)) {
1506 		m_free(n);
1507 		return NULL;
1508 	}
1509 
1510 	m_copydata(m, off, elen, mtod(n, caddr_t));
1511 	n->m_len = elen;
1512 	return n;
1513 }
1514 #endif
1515 
1516 /*
1517  * Get pointer to the previous header followed by the header
1518  * currently processed.
1519  * XXX: This function supposes that
1520  *	M includes all headers,
1521  *	the next header field and the header length field of each header
1522  *	are valid, and
1523  *	the sum of each header length equals to OFF.
1524  * Because of these assumptions, this function must be called very
1525  * carefully. Moreover, it will not be used in the near future when
1526  * we develop `neater' mechanism to process extension headers.
1527  */
1528 char *
1529 ip6_get_prevhdr(struct mbuf *m, int off)
1530 {
1531 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1532 
1533 	if (off == sizeof(struct ip6_hdr))
1534 		return (&ip6->ip6_nxt);
1535 	else {
1536 		int len, nxt;
1537 		struct ip6_ext *ip6e = NULL;
1538 
1539 		nxt = ip6->ip6_nxt;
1540 		len = sizeof(struct ip6_hdr);
1541 		while (len < off) {
1542 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1543 
1544 			switch (nxt) {
1545 			case IPPROTO_FRAGMENT:
1546 				len += sizeof(struct ip6_frag);
1547 				break;
1548 			case IPPROTO_AH:
1549 				len += (ip6e->ip6e_len + 2) << 2;
1550 				break;
1551 			default:
1552 				len += (ip6e->ip6e_len + 1) << 3;
1553 				break;
1554 			}
1555 			nxt = ip6e->ip6e_nxt;
1556 		}
1557 		if (ip6e)
1558 			return (&ip6e->ip6e_nxt);
1559 		else
1560 			return NULL;
1561 	}
1562 }
1563 
1564 /*
1565  * get next header offset.  m will be retained.
1566  */
1567 int
1568 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1569 {
1570 	struct ip6_hdr ip6;
1571 	struct ip6_ext ip6e;
1572 	struct ip6_frag fh;
1573 
1574 	/* just in case */
1575 	if (m == NULL)
1576 		panic("ip6_nexthdr: m == NULL");
1577 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1578 		return -1;
1579 
1580 	switch (proto) {
1581 	case IPPROTO_IPV6:
1582 		if (m->m_pkthdr.len < off + sizeof(ip6))
1583 			return -1;
1584 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1585 		if (nxtp)
1586 			*nxtp = ip6.ip6_nxt;
1587 		off += sizeof(ip6);
1588 		return off;
1589 
1590 	case IPPROTO_FRAGMENT:
1591 		/*
1592 		 * terminate parsing if it is not the first fragment,
1593 		 * it does not make sense to parse through it.
1594 		 */
1595 		if (m->m_pkthdr.len < off + sizeof(fh))
1596 			return -1;
1597 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1598 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1599 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1600 			return -1;
1601 		if (nxtp)
1602 			*nxtp = fh.ip6f_nxt;
1603 		off += sizeof(struct ip6_frag);
1604 		return off;
1605 
1606 	case IPPROTO_AH:
1607 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1608 			return -1;
1609 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1610 		if (nxtp)
1611 			*nxtp = ip6e.ip6e_nxt;
1612 		off += (ip6e.ip6e_len + 2) << 2;
1613 		return off;
1614 
1615 	case IPPROTO_HOPOPTS:
1616 	case IPPROTO_ROUTING:
1617 	case IPPROTO_DSTOPTS:
1618 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1619 			return -1;
1620 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1621 		if (nxtp)
1622 			*nxtp = ip6e.ip6e_nxt;
1623 		off += (ip6e.ip6e_len + 1) << 3;
1624 		return off;
1625 
1626 	case IPPROTO_NONE:
1627 	case IPPROTO_ESP:
1628 	case IPPROTO_IPCOMP:
1629 		/* give up */
1630 		return -1;
1631 
1632 	default:
1633 		return -1;
1634 	}
1635 
1636 	return -1;
1637 }
1638 
1639 /*
1640  * get offset for the last header in the chain.  m will be kept untainted.
1641  */
1642 int
1643 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1644 {
1645 	int newoff;
1646 	int nxt;
1647 
1648 	if (!nxtp) {
1649 		nxt = -1;
1650 		nxtp = &nxt;
1651 	}
1652 	while (1) {
1653 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1654 		if (newoff < 0)
1655 			return off;
1656 		else if (newoff < off)
1657 			return -1;	/* invalid */
1658 		else if (newoff == off)
1659 			return newoff;
1660 
1661 		off = newoff;
1662 		proto = *nxtp;
1663 	}
1664 }
1665 
1666 struct ip6aux *
1667 ip6_addaux(struct mbuf *m)
1668 {
1669 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1670 	if (!tag) {
1671 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1672 				sizeof (struct ip6aux),
1673 				MB_DONTWAIT);
1674 		if (tag)
1675 			m_tag_prepend(m, tag);
1676 	}
1677 	if (tag)
1678 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1679 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1680 }
1681 
1682 struct ip6aux *
1683 ip6_findaux(struct mbuf *m)
1684 {
1685 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1686 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1687 }
1688 
1689 void
1690 ip6_delaux(struct mbuf *m)
1691 {
1692 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1693 	if (tag)
1694 		m_tag_delete(m, tag);
1695 }
1696 
1697 /*
1698  * System control for IP6
1699  */
1700 
1701 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1702 	0,		0,		0,		0,
1703 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1704 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1705 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1706 	0,		0,		0,		0,
1707 	ENOPROTOOPT
1708 };
1709