]> CyberLeo.Net >> Repos - FreeBSD/releng/9.2.git/blob - sys/netinet6/icmp6.c
- Copy stable/9 to releng/9.2 as part of the 9.2-RELEASE cycle.
[FreeBSD/releng/9.2.git] / sys / netinet6 / icmp6.c
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *      The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
61  */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/icmp6.h>
99 #include <netinet/tcp_var.h>
100
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/send.h>
109
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/key.h>
113 #endif
114
115 extern struct domain inet6domain;
116
117 VNET_DEFINE(struct icmp6stat, icmp6stat);
118
119 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
120 VNET_DECLARE(struct inpcbhead, ripcb);
121 VNET_DECLARE(int, icmp6errppslim);
122 static VNET_DEFINE(int, icmp6errpps_count) = 0;
123 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
124 VNET_DECLARE(int, icmp6_nodeinfo);
125
126 #define V_ripcbinfo                     VNET(ripcbinfo)
127 #define V_ripcb                         VNET(ripcb)
128 #define V_icmp6errppslim                VNET(icmp6errppslim)
129 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
130 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
131 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
132
133 static void icmp6_errcount(int, int);
134 static int icmp6_rip6_input(struct mbuf **, int);
135 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
136 static const char *icmp6_redirect_diag(struct in6_addr *,
137         struct in6_addr *, struct in6_addr *);
138 static struct mbuf *ni6_input(struct mbuf *, int);
139 static struct mbuf *ni6_nametodns(const char *, int, int);
140 static int ni6_dnsmatch(const char *, int, const char *, int);
141 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
142                           struct ifnet **, struct in6_addr *);
143 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
144                                 struct ifnet *, int);
145 static int icmp6_notify_error(struct mbuf **, int, int, int);
146
147 /*
148  * Kernel module interface for updating icmp6stat.  The argument is an index
149  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
150  * general layout of icmp6stat into the caller, it doesn't encode its
151  * location, so that future changes to add, for example, per-CPU stats
152  * support won't cause binary compatibility problems for kernel modules.
153  */
154 void
155 kmod_icmp6stat_inc(int statnum)
156 {
157
158         (*((u_quad_t *)&V_icmp6stat + statnum))++;
159 }
160
161 static void
162 icmp6_errcount(int type, int code)
163 {
164         switch (type) {
165         case ICMP6_DST_UNREACH:
166                 switch (code) {
167                 case ICMP6_DST_UNREACH_NOROUTE:
168                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
169                         return;
170                 case ICMP6_DST_UNREACH_ADMIN:
171                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
172                         return;
173                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
174                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
175                         return;
176                 case ICMP6_DST_UNREACH_ADDR:
177                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
178                         return;
179                 case ICMP6_DST_UNREACH_NOPORT:
180                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
181                         return;
182                 }
183                 break;
184         case ICMP6_PACKET_TOO_BIG:
185                 ICMP6STAT_INC(icp6s_opacket_too_big);
186                 return;
187         case ICMP6_TIME_EXCEEDED:
188                 switch (code) {
189                 case ICMP6_TIME_EXCEED_TRANSIT:
190                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
191                         return;
192                 case ICMP6_TIME_EXCEED_REASSEMBLY:
193                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
194                         return;
195                 }
196                 break;
197         case ICMP6_PARAM_PROB:
198                 switch (code) {
199                 case ICMP6_PARAMPROB_HEADER:
200                         ICMP6STAT_INC(icp6s_oparamprob_header);
201                         return;
202                 case ICMP6_PARAMPROB_NEXTHEADER:
203                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
204                         return;
205                 case ICMP6_PARAMPROB_OPTION:
206                         ICMP6STAT_INC(icp6s_oparamprob_option);
207                         return;
208                 }
209                 break;
210         case ND_REDIRECT:
211                 ICMP6STAT_INC(icp6s_oredirect);
212                 return;
213         }
214         ICMP6STAT_INC(icp6s_ounknown);
215 }
216
217 /*
218  * A wrapper function for icmp6_error() necessary when the erroneous packet
219  * may not contain enough scope zone information.
220  */
221 void
222 icmp6_error2(struct mbuf *m, int type, int code, int param,
223     struct ifnet *ifp)
224 {
225         struct ip6_hdr *ip6;
226
227         if (ifp == NULL)
228                 return;
229
230 #ifndef PULLDOWN_TEST
231         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
232 #else
233         if (m->m_len < sizeof(struct ip6_hdr)) {
234                 m = m_pullup(m, sizeof(struct ip6_hdr));
235                 if (m == NULL)
236                         return;
237         }
238 #endif
239
240         ip6 = mtod(m, struct ip6_hdr *);
241
242         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
243                 return;
244         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
245                 return;
246
247         icmp6_error(m, type, code, param);
248 }
249
250 /*
251  * Generate an error packet of type error in response to bad IP6 packet.
252  */
253 void
254 icmp6_error(struct mbuf *m, int type, int code, int param)
255 {
256         struct ip6_hdr *oip6, *nip6;
257         struct icmp6_hdr *icmp6;
258         u_int preplen;
259         int off;
260         int nxt;
261
262         ICMP6STAT_INC(icp6s_error);
263
264         /* count per-type-code statistics */
265         icmp6_errcount(type, code);
266
267 #ifdef M_DECRYPTED      /*not openbsd*/
268         if (m->m_flags & M_DECRYPTED) {
269                 ICMP6STAT_INC(icp6s_canterror);
270                 goto freeit;
271         }
272 #endif
273
274 #ifndef PULLDOWN_TEST
275         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
276 #else
277         if (m->m_len < sizeof(struct ip6_hdr)) {
278                 m = m_pullup(m, sizeof(struct ip6_hdr));
279                 if (m == NULL)
280                         return;
281         }
282 #endif
283         oip6 = mtod(m, struct ip6_hdr *);
284
285         /*
286          * If the destination address of the erroneous packet is a multicast
287          * address, or the packet was sent using link-layer multicast,
288          * we should basically suppress sending an error (RFC 2463, Section
289          * 2.4).
290          * We have two exceptions (the item e.2 in that section):
291          * - the Packet Too Big message can be sent for path MTU discovery.
292          * - the Parameter Problem Message that can be allowed an icmp6 error
293          *   in the option type field.  This check has been done in
294          *   ip6_unknown_opt(), so we can just check the type and code.
295          */
296         if ((m->m_flags & (M_BCAST|M_MCAST) ||
297              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
298             (type != ICMP6_PACKET_TOO_BIG &&
299              (type != ICMP6_PARAM_PROB ||
300               code != ICMP6_PARAMPROB_OPTION)))
301                 goto freeit;
302
303         /*
304          * RFC 2463, 2.4 (e.5): source address check.
305          * XXX: the case of anycast source?
306          */
307         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
308             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
309                 goto freeit;
310
311         /*
312          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
313          * don't do it.
314          */
315         nxt = -1;
316         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
317         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
318                 struct icmp6_hdr *icp;
319
320 #ifndef PULLDOWN_TEST
321                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
322                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
323 #else
324                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
325                         sizeof(*icp));
326                 if (icp == NULL) {
327                         ICMP6STAT_INC(icp6s_tooshort);
328                         return;
329                 }
330 #endif
331                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
332                     icp->icmp6_type == ND_REDIRECT) {
333                         /*
334                          * ICMPv6 error
335                          * Special case: for redirect (which is
336                          * informational) we must not send icmp6 error.
337                          */
338                         ICMP6STAT_INC(icp6s_canterror);
339                         goto freeit;
340                 } else {
341                         /* ICMPv6 informational - send the error */
342                 }
343         } else {
344                 /* non-ICMPv6 - send the error */
345         }
346
347         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
348
349         /* Finally, do rate limitation check. */
350         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
351                 ICMP6STAT_INC(icp6s_toofreq);
352                 goto freeit;
353         }
354
355         /*
356          * OK, ICMP6 can be generated.
357          */
358
359         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
360                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
361
362         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
363         M_PREPEND(m, preplen, M_DONTWAIT);      /* FIB is also copied over. */
364         if (m && m->m_len < preplen)
365                 m = m_pullup(m, preplen);
366         if (m == NULL) {
367                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
368                 return;
369         }
370
371         nip6 = mtod(m, struct ip6_hdr *);
372         nip6->ip6_src  = oip6->ip6_src;
373         nip6->ip6_dst  = oip6->ip6_dst;
374
375         in6_clearscope(&oip6->ip6_src);
376         in6_clearscope(&oip6->ip6_dst);
377
378         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
379         icmp6->icmp6_type = type;
380         icmp6->icmp6_code = code;
381         icmp6->icmp6_pptr = htonl((u_int32_t)param);
382
383         /*
384          * icmp6_reflect() is designed to be in the input path.
385          * icmp6_error() can be called from both input and output path,
386          * and if we are in output path rcvif could contain bogus value.
387          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
388          * information in ip header (nip6).
389          */
390         m->m_pkthdr.rcvif = NULL;
391
392         ICMP6STAT_INC(icp6s_outhist[type]);
393         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
394
395         return;
396
397   freeit:
398         /*
399          * If we can't tell whether or not we can generate ICMP6, free it.
400          */
401         m_freem(m);
402 }
403
404 /*
405  * Process a received ICMP6 message.
406  */
407 int
408 icmp6_input(struct mbuf **mp, int *offp, int proto)
409 {
410         struct mbuf *m = *mp, *n;
411         struct ifnet *ifp;
412         struct ip6_hdr *ip6, *nip6;
413         struct icmp6_hdr *icmp6, *nicmp6;
414         int off = *offp;
415         int icmp6len = m->m_pkthdr.len - *offp;
416         int code, sum, noff;
417         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
418         int ip6len, error;
419
420         ifp = m->m_pkthdr.rcvif;
421
422 #ifndef PULLDOWN_TEST
423         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
424         /* m might change if M_LOOP.  So, call mtod after this */
425 #endif
426
427         /*
428          * Locate icmp6 structure in mbuf, and check
429          * that not corrupted and of at least minimum length
430          */
431
432         ip6 = mtod(m, struct ip6_hdr *);
433         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
434         if (icmp6len < sizeof(struct icmp6_hdr)) {
435                 ICMP6STAT_INC(icp6s_tooshort);
436                 goto freeit;
437         }
438
439         /*
440          * Check multicast group membership.
441          * Note: SSM filters are not applied for ICMPv6 traffic.
442          */
443         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
444                 struct in6_multi        *inm;
445
446                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
447                 if (inm == NULL) {
448                         IP6STAT_INC(ip6s_notmember);
449                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
450                         goto freeit;
451                 }
452         }
453
454         /*
455          * calculate the checksum
456          */
457 #ifndef PULLDOWN_TEST
458         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
459 #else
460         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
461         if (icmp6 == NULL) {
462                 ICMP6STAT_INC(icp6s_tooshort);
463                 return IPPROTO_DONE;
464         }
465 #endif
466         code = icmp6->icmp6_code;
467
468         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
469                 nd6log((LOG_ERR,
470                     "ICMP6 checksum error(%d|%x) %s\n",
471                     icmp6->icmp6_type, sum,
472                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
473                 ICMP6STAT_INC(icp6s_checksum);
474                 goto freeit;
475         }
476
477         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
478                 /*
479                  * Deliver very specific ICMP6 type only.
480                  * This is important to deliver TOOBIG.  Otherwise PMTUD
481                  * will not work.
482                  */
483                 switch (icmp6->icmp6_type) {
484                 case ICMP6_DST_UNREACH:
485                 case ICMP6_PACKET_TOO_BIG:
486                 case ICMP6_TIME_EXCEEDED:
487                         break;
488                 default:
489                         goto freeit;
490                 }
491         }
492
493         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
494         icmp6_ifstat_inc(ifp, ifs6_in_msg);
495         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
496                 icmp6_ifstat_inc(ifp, ifs6_in_error);
497
498         switch (icmp6->icmp6_type) {
499         case ICMP6_DST_UNREACH:
500                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
501                 switch (code) {
502                 case ICMP6_DST_UNREACH_NOROUTE:
503                         code = PRC_UNREACH_NET;
504                         break;
505                 case ICMP6_DST_UNREACH_ADMIN:
506                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
507                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
508                         break;
509                 case ICMP6_DST_UNREACH_ADDR:
510                         code = PRC_HOSTDEAD;
511                         break;
512                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
513                         /* I mean "source address was incorrect." */
514                         code = PRC_PARAMPROB;
515                         break;
516                 case ICMP6_DST_UNREACH_NOPORT:
517                         code = PRC_UNREACH_PORT;
518                         break;
519                 default:
520                         goto badcode;
521                 }
522                 goto deliver;
523                 break;
524
525         case ICMP6_PACKET_TOO_BIG:
526                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
527
528                 /* validation is made in icmp6_mtudisc_update */
529
530                 code = PRC_MSGSIZE;
531
532                 /*
533                  * Updating the path MTU will be done after examining
534                  * intermediate extension headers.
535                  */
536                 goto deliver;
537                 break;
538
539         case ICMP6_TIME_EXCEEDED:
540                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
541                 switch (code) {
542                 case ICMP6_TIME_EXCEED_TRANSIT:
543                         code = PRC_TIMXCEED_INTRANS;
544                         break;
545                 case ICMP6_TIME_EXCEED_REASSEMBLY:
546                         code = PRC_TIMXCEED_REASS;
547                         break;
548                 default:
549                         goto badcode;
550                 }
551                 goto deliver;
552                 break;
553
554         case ICMP6_PARAM_PROB:
555                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
556                 switch (code) {
557                 case ICMP6_PARAMPROB_NEXTHEADER:
558                         code = PRC_UNREACH_PROTOCOL;
559                         break;
560                 case ICMP6_PARAMPROB_HEADER:
561                 case ICMP6_PARAMPROB_OPTION:
562                         code = PRC_PARAMPROB;
563                         break;
564                 default:
565                         goto badcode;
566                 }
567                 goto deliver;
568                 break;
569
570         case ICMP6_ECHO_REQUEST:
571                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
572                 if (code != 0)
573                         goto badcode;
574                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
575                         /* Give up remote */
576                         break;
577                 }
578                 if ((n->m_flags & M_EXT) != 0
579                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
580                         struct mbuf *n0 = n;
581                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
582                         int n0len;
583
584                         MGETHDR(n, M_DONTWAIT, n0->m_type);
585                         n0len = n0->m_pkthdr.len;       /* save for use below */
586                         if (n)
587                                 M_MOVE_PKTHDR(n, n0);   /* FIB copied. */
588                         if (n && maxlen >= MHLEN) {
589                                 MCLGET(n, M_DONTWAIT);
590                                 if ((n->m_flags & M_EXT) == 0) {
591                                         m_free(n);
592                                         n = NULL;
593                                 }
594                         }
595                         if (n == NULL) {
596                                 /* Give up remote */
597                                 m_freem(n0);
598                                 break;
599                         }
600                         /*
601                          * Copy IPv6 and ICMPv6 only.
602                          */
603                         nip6 = mtod(n, struct ip6_hdr *);
604                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
605                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
606                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
607                         noff = sizeof(struct ip6_hdr);
608                         /* new mbuf contains only ipv6+icmpv6 headers */
609                         n->m_len = noff + sizeof(struct icmp6_hdr);
610                         /*
611                          * Adjust mbuf.  ip6_plen will be adjusted in
612                          * ip6_output().
613                          */
614                         m_adj(n0, off + sizeof(struct icmp6_hdr));
615                         /* recalculate complete packet size */
616                         n->m_pkthdr.len = n0len + (noff - off);
617                         n->m_next = n0;
618                 } else {
619                         nip6 = mtod(n, struct ip6_hdr *);
620                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
621                             sizeof(*nicmp6));
622                         noff = off;
623                 }
624                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
625                 nicmp6->icmp6_code = 0;
626                 if (n) {
627                         ICMP6STAT_INC(icp6s_reflect);
628                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
629                         icmp6_reflect(n, noff);
630                 }
631                 break;
632
633         case ICMP6_ECHO_REPLY:
634                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
635                 if (code != 0)
636                         goto badcode;
637                 break;
638
639         case MLD_LISTENER_QUERY:
640         case MLD_LISTENER_REPORT:
641         case MLD_LISTENER_DONE:
642         case MLDV2_LISTENER_REPORT:
643                 /*
644                  * Drop MLD traffic which is not link-local, has a hop limit
645                  * of greater than 1 hop, or which does not have the
646                  * IPv6 HBH Router Alert option.
647                  * As IPv6 HBH options are stripped in ip6_input() we must
648                  * check an mbuf header flag.
649                  * XXX Should we also sanity check that these messages
650                  * were directed to a link-local multicast prefix?
651                  */
652                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
653                         goto freeit;
654                 if (mld_input(m, off, icmp6len) != 0)
655                         return (IPPROTO_DONE);
656                 /* m stays. */
657                 break;
658
659         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
660             {
661                 enum { WRU, FQDN } mode;
662
663                 if (!V_icmp6_nodeinfo)
664                         break;
665
666                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
667                         mode = WRU;
668                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
669                         mode = FQDN;
670                 else
671                         goto badlen;
672
673                 if (mode == FQDN) {
674 #ifndef PULLDOWN_TEST
675                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
676                             IPPROTO_DONE);
677 #endif
678                         n = m_copy(m, 0, M_COPYALL);
679                         if (n)
680                                 n = ni6_input(n, off);
681                         /* XXX meaningless if n == NULL */
682                         noff = sizeof(struct ip6_hdr);
683                 } else {
684                         struct prison *pr;
685                         u_char *p;
686                         int maxlen, maxhlen, hlen;
687
688                         /*
689                          * XXX: this combination of flags is pointless,
690                          * but should we keep this for compatibility?
691                          */
692                         if ((V_icmp6_nodeinfo & 5) != 5)
693                                 break;
694
695                         if (code != 0)
696                                 goto badcode;
697                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
698                         if (maxlen >= MCLBYTES) {
699                                 /* Give up remote */
700                                 break;
701                         }
702                         MGETHDR(n, M_DONTWAIT, m->m_type);
703                         if (n && maxlen > MHLEN) {
704                                 MCLGET(n, M_DONTWAIT);
705                                 if ((n->m_flags & M_EXT) == 0) {
706                                         m_free(n);
707                                         n = NULL;
708                                 }
709                         }
710                         if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
711                                 /*
712                                  * Previous code did a blind M_COPY_PKTHDR
713                                  * and said "just for rcvif".  If true, then
714                                  * we could tolerate the dup failing (due to
715                                  * the deep copy of the tag chain).  For now
716                                  * be conservative and just fail.
717                                  */
718                                 m_free(n);
719                                 n = NULL;
720                         }
721                         if (n == NULL) {
722                                 /* Give up remote */
723                                 break;
724                         }
725                         n->m_pkthdr.rcvif = NULL;
726                         n->m_len = 0;
727                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
728                         pr = curthread->td_ucred->cr_prison;
729                         mtx_lock(&pr->pr_mtx);
730                         hlen = strlen(pr->pr_hostname);
731                         if (maxhlen > hlen)
732                                 maxhlen = hlen;
733                         /*
734                          * Copy IPv6 and ICMPv6 only.
735                          */
736                         nip6 = mtod(n, struct ip6_hdr *);
737                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
738                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
739                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
740                         p = (u_char *)(nicmp6 + 1);
741                         bzero(p, 4);
742                         /* meaningless TTL */
743                         bcopy(pr->pr_hostname, p + 4, maxhlen);
744                         mtx_unlock(&pr->pr_mtx);
745                         noff = sizeof(struct ip6_hdr);
746                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
747                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
748                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
749                         nicmp6->icmp6_code = 0;
750                 }
751                 if (n) {
752                         ICMP6STAT_INC(icp6s_reflect);
753                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
754                         icmp6_reflect(n, noff);
755                 }
756                 break;
757             }
758
759         case ICMP6_WRUREPLY:
760                 if (code != 0)
761                         goto badcode;
762                 break;
763
764         case ND_ROUTER_SOLICIT:
765                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
766                 if (code != 0)
767                         goto badcode;
768                 if (icmp6len < sizeof(struct nd_router_solicit))
769                         goto badlen;
770                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
771                         /* give up local */
772
773                         /* Send incoming SeND packet to user space. */
774                         if (send_sendso_input_hook != NULL) {
775                                 IP6_EXTHDR_CHECK(m, off,
776                                     icmp6len, IPPROTO_DONE);
777                                 error = send_sendso_input_hook(m, ifp,
778                                     SND_IN, ip6len);
779                                 /* -1 == no app on SEND socket */
780                                 if (error == 0)
781                                         return (IPPROTO_DONE);
782                                 nd6_rs_input(m, off, icmp6len);
783                         } else
784                                 nd6_rs_input(m, off, icmp6len);
785                         m = NULL;
786                         goto freeit;
787                 }
788                 if (send_sendso_input_hook != NULL) {
789                         IP6_EXTHDR_CHECK(n, off,
790                             icmp6len, IPPROTO_DONE);
791                         error = send_sendso_input_hook(n, ifp,
792                             SND_IN, ip6len);
793                         if (error == 0)
794                                 goto freeit;
795                         /* -1 == no app on SEND socket */
796                         nd6_rs_input(n, off, icmp6len);
797                 } else
798                         nd6_rs_input(n, off, icmp6len);
799                 /* m stays. */
800                 break;
801
802         case ND_ROUTER_ADVERT:
803                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
804                 if (code != 0)
805                         goto badcode;
806                 if (icmp6len < sizeof(struct nd_router_advert))
807                         goto badlen;
808                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
809
810                         /* Send incoming SeND-protected/ND packet to user space. */
811                         if (send_sendso_input_hook != NULL) {
812                                 error = send_sendso_input_hook(m, ifp,
813                                     SND_IN, ip6len);
814                                 if (error == 0)
815                                         return (IPPROTO_DONE);
816                                 nd6_ra_input(m, off, icmp6len);
817                         } else
818                                 nd6_ra_input(m, off, icmp6len);
819                         m = NULL;
820                         goto freeit;
821                 }
822                 if (send_sendso_input_hook != NULL) {
823                         error = send_sendso_input_hook(n, ifp,
824                             SND_IN, ip6len);
825                         if (error == 0)
826                                 goto freeit;
827                         nd6_ra_input(n, off, icmp6len);
828                 } else
829                         nd6_ra_input(n, off, icmp6len);
830                 /* m stays. */
831                 break;
832
833         case ND_NEIGHBOR_SOLICIT:
834                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
835                 if (code != 0)
836                         goto badcode;
837                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
838                         goto badlen;
839                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
840                         if (send_sendso_input_hook != NULL) {
841                                 error = send_sendso_input_hook(m, ifp,
842                                     SND_IN, ip6len);
843                                 if (error == 0)
844                                         return (IPPROTO_DONE);
845                                 nd6_ns_input(m, off, icmp6len);
846                         } else
847                                 nd6_ns_input(m, off, icmp6len);
848                         m = NULL;
849                         goto freeit;
850                 }
851                 if (send_sendso_input_hook != NULL) {
852                         error = send_sendso_input_hook(n, ifp,
853                             SND_IN, ip6len);
854                         if (error == 0)
855                                 goto freeit;
856                         nd6_ns_input(n, off, icmp6len);
857                 } else
858                         nd6_ns_input(n, off, icmp6len);
859                 /* m stays. */
860                 break;
861
862         case ND_NEIGHBOR_ADVERT:
863                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
864                 if (code != 0)
865                         goto badcode;
866                 if (icmp6len < sizeof(struct nd_neighbor_advert))
867                         goto badlen;
868                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
869
870                         /* Send incoming SeND-protected/ND packet to user space. */
871                         if (send_sendso_input_hook != NULL) {
872                                 error = send_sendso_input_hook(m, ifp,
873                                     SND_IN, ip6len);
874                                 if (error == 0)
875                                         return (IPPROTO_DONE);
876                                 nd6_na_input(m, off, icmp6len);
877                         } else
878                                 nd6_na_input(m, off, icmp6len);
879                         m = NULL;
880                         goto freeit;
881                 }
882                 if (send_sendso_input_hook != NULL) {
883                         error = send_sendso_input_hook(n, ifp,
884                             SND_IN, ip6len);
885                         if (error == 0)
886                                 goto freeit;
887                         nd6_na_input(n, off, icmp6len);
888                 } else
889                         nd6_na_input(n, off, icmp6len);
890                 /* m stays. */
891                 break;
892
893         case ND_REDIRECT:
894                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
895                 if (code != 0)
896                         goto badcode;
897                 if (icmp6len < sizeof(struct nd_redirect))
898                         goto badlen;
899                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
900                         if (send_sendso_input_hook != NULL) {
901                                 error = send_sendso_input_hook(m, ifp,
902                                     SND_IN, ip6len);
903                                 if (error == 0)
904                                         return (IPPROTO_DONE);
905                             icmp6_redirect_input(m, off);
906                         } else
907                                 icmp6_redirect_input(m, off);
908                         m = NULL;
909                         goto freeit;
910                 }
911                 if (send_sendso_input_hook != NULL) {
912                         error = send_sendso_input_hook(n, ifp,
913                             SND_IN, ip6len);
914                         if (error == 0)
915                                 goto freeit;
916                         icmp6_redirect_input(n, off);
917                 } else
918                         icmp6_redirect_input(n, off);
919                 /* m stays. */
920                 break;
921
922         case ICMP6_ROUTER_RENUMBERING:
923                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
924                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
925                         goto badcode;
926                 if (icmp6len < sizeof(struct icmp6_router_renum))
927                         goto badlen;
928                 break;
929
930         default:
931                 nd6log((LOG_DEBUG,
932                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
933                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
934                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
935                     ifp ? ifp->if_index : 0));
936                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
937                         /* ICMPv6 error: MUST deliver it by spec... */
938                         code = PRC_NCMDS;
939                         /* deliver */
940                 } else {
941                         /* ICMPv6 informational: MUST not deliver */
942                         break;
943                 }
944         deliver:
945                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
946                         /* In this case, m should've been freed. */
947                         return (IPPROTO_DONE);
948                 }
949                 break;
950
951         badcode:
952                 ICMP6STAT_INC(icp6s_badcode);
953                 break;
954
955         badlen:
956                 ICMP6STAT_INC(icp6s_badlen);
957                 break;
958         }
959
960         /* deliver the packet to appropriate sockets */
961         icmp6_rip6_input(&m, *offp);
962
963         return IPPROTO_DONE;
964
965  freeit:
966         m_freem(m);
967         return IPPROTO_DONE;
968 }
969
970 static int
971 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
972 {
973         struct mbuf *m = *mp;
974         struct icmp6_hdr *icmp6;
975         struct ip6_hdr *eip6;
976         u_int32_t notifymtu;
977         struct sockaddr_in6 icmp6src, icmp6dst;
978
979         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
980                 ICMP6STAT_INC(icp6s_tooshort);
981                 goto freeit;
982         }
983 #ifndef PULLDOWN_TEST
984         IP6_EXTHDR_CHECK(m, off,
985             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
986         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
987 #else
988         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
989             sizeof(*icmp6) + sizeof(struct ip6_hdr));
990         if (icmp6 == NULL) {
991                 ICMP6STAT_INC(icp6s_tooshort);
992                 return (-1);
993         }
994 #endif
995         eip6 = (struct ip6_hdr *)(icmp6 + 1);
996
997         /* Detect the upper level protocol */
998         {
999                 void (*ctlfunc)(int, struct sockaddr *, void *);
1000                 u_int8_t nxt = eip6->ip6_nxt;
1001                 int eoff = off + sizeof(struct icmp6_hdr) +
1002                     sizeof(struct ip6_hdr);
1003                 struct ip6ctlparam ip6cp;
1004                 struct in6_addr *finaldst = NULL;
1005                 int icmp6type = icmp6->icmp6_type;
1006                 struct ip6_frag *fh;
1007                 struct ip6_rthdr *rth;
1008                 struct ip6_rthdr0 *rth0;
1009                 int rthlen;
1010
1011                 while (1) { /* XXX: should avoid infinite loop explicitly? */
1012                         struct ip6_ext *eh;
1013
1014                         switch (nxt) {
1015                         case IPPROTO_HOPOPTS:
1016                         case IPPROTO_DSTOPTS:
1017                         case IPPROTO_AH:
1018 #ifndef PULLDOWN_TEST
1019                                 IP6_EXTHDR_CHECK(m, 0,
1020                                     eoff + sizeof(struct ip6_ext), -1);
1021                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1022 #else
1023                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1024                                     eoff, sizeof(*eh));
1025                                 if (eh == NULL) {
1026                                         ICMP6STAT_INC(icp6s_tooshort);
1027                                         return (-1);
1028                                 }
1029 #endif
1030
1031                                 if (nxt == IPPROTO_AH)
1032                                         eoff += (eh->ip6e_len + 2) << 2;
1033                                 else
1034                                         eoff += (eh->ip6e_len + 1) << 3;
1035                                 nxt = eh->ip6e_nxt;
1036                                 break;
1037                         case IPPROTO_ROUTING:
1038                                 /*
1039                                  * When the erroneous packet contains a
1040                                  * routing header, we should examine the
1041                                  * header to determine the final destination.
1042                                  * Otherwise, we can't properly update
1043                                  * information that depends on the final
1044                                  * destination (e.g. path MTU).
1045                                  */
1046 #ifndef PULLDOWN_TEST
1047                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1048                                 rth = (struct ip6_rthdr *)
1049                                     (mtod(m, caddr_t) + eoff);
1050 #else
1051                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1052                                     eoff, sizeof(*rth));
1053                                 if (rth == NULL) {
1054                                         ICMP6STAT_INC(icp6s_tooshort);
1055                                         return (-1);
1056                                 }
1057 #endif
1058                                 rthlen = (rth->ip6r_len + 1) << 3;
1059                                 /*
1060                                  * XXX: currently there is no
1061                                  * officially defined type other
1062                                  * than type-0.
1063                                  * Note that if the segment left field
1064                                  * is 0, all intermediate hops must
1065                                  * have been passed.
1066                                  */
1067                                 if (rth->ip6r_segleft &&
1068                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1069                                         int hops;
1070
1071 #ifndef PULLDOWN_TEST
1072                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1073                                         rth0 = (struct ip6_rthdr0 *)
1074                                             (mtod(m, caddr_t) + eoff);
1075 #else
1076                                         IP6_EXTHDR_GET(rth0,
1077                                             struct ip6_rthdr0 *, m,
1078                                             eoff, rthlen);
1079                                         if (rth0 == NULL) {
1080                                                 ICMP6STAT_INC(icp6s_tooshort);
1081                                                 return (-1);
1082                                         }
1083 #endif
1084                                         /* just ignore a bogus header */
1085                                         if ((rth0->ip6r0_len % 2) == 0 &&
1086                                             (hops = rth0->ip6r0_len/2))
1087                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1088                                 }
1089                                 eoff += rthlen;
1090                                 nxt = rth->ip6r_nxt;
1091                                 break;
1092                         case IPPROTO_FRAGMENT:
1093 #ifndef PULLDOWN_TEST
1094                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1095                                     sizeof(struct ip6_frag), -1);
1096                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1097                                     eoff);
1098 #else
1099                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1100                                     eoff, sizeof(*fh));
1101                                 if (fh == NULL) {
1102                                         ICMP6STAT_INC(icp6s_tooshort);
1103                                         return (-1);
1104                                 }
1105 #endif
1106                                 /*
1107                                  * Data after a fragment header is meaningless
1108                                  * unless it is the first fragment, but
1109                                  * we'll go to the notify label for path MTU
1110                                  * discovery.
1111                                  */
1112                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1113                                         goto notify;
1114
1115                                 eoff += sizeof(struct ip6_frag);
1116                                 nxt = fh->ip6f_nxt;
1117                                 break;
1118                         default:
1119                                 /*
1120                                  * This case includes ESP and the No Next
1121                                  * Header.  In such cases going to the notify
1122                                  * label does not have any meaning
1123                                  * (i.e. ctlfunc will be NULL), but we go
1124                                  * anyway since we might have to update
1125                                  * path MTU information.
1126                                  */
1127                                 goto notify;
1128                         }
1129                 }
1130           notify:
1131 #ifndef PULLDOWN_TEST
1132                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1133 #else
1134                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1135                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1136                 if (icmp6 == NULL) {
1137                         ICMP6STAT_INC(icp6s_tooshort);
1138                         return (-1);
1139                 }
1140 #endif
1141
1142                 /*
1143                  * retrieve parameters from the inner IPv6 header, and convert
1144                  * them into sockaddr structures.
1145                  * XXX: there is no guarantee that the source or destination
1146                  * addresses of the inner packet are in the same scope as
1147                  * the addresses of the icmp packet.  But there is no other
1148                  * way to determine the zone.
1149                  */
1150                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1151
1152                 bzero(&icmp6dst, sizeof(icmp6dst));
1153                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1154                 icmp6dst.sin6_family = AF_INET6;
1155                 if (finaldst == NULL)
1156                         icmp6dst.sin6_addr = eip6->ip6_dst;
1157                 else
1158                         icmp6dst.sin6_addr = *finaldst;
1159                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1160                         goto freeit;
1161                 bzero(&icmp6src, sizeof(icmp6src));
1162                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1163                 icmp6src.sin6_family = AF_INET6;
1164                 icmp6src.sin6_addr = eip6->ip6_src;
1165                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1166                         goto freeit;
1167                 icmp6src.sin6_flowinfo =
1168                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1169
1170                 if (finaldst == NULL)
1171                         finaldst = &eip6->ip6_dst;
1172                 ip6cp.ip6c_m = m;
1173                 ip6cp.ip6c_icmp6 = icmp6;
1174                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1175                 ip6cp.ip6c_off = eoff;
1176                 ip6cp.ip6c_finaldst = finaldst;
1177                 ip6cp.ip6c_src = &icmp6src;
1178                 ip6cp.ip6c_nxt = nxt;
1179
1180                 m_addr_changed(m);
1181
1182                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1183                         notifymtu = ntohl(icmp6->icmp6_mtu);
1184                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1185                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1186                 }
1187
1188                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1189                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1190                 if (ctlfunc) {
1191                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1192                             &ip6cp);
1193                 }
1194         }
1195         *mp = m;
1196         return (0);
1197
1198   freeit:
1199         m_freem(m);
1200         return (-1);
1201 }
1202
1203 void
1204 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1205 {
1206         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1207         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1208         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1209         u_int mtu = ntohl(icmp6->icmp6_mtu);
1210         struct in_conninfo inc;
1211
1212 #if 0
1213         /*
1214          * RFC2460 section 5, last paragraph.
1215          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1216          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1217          * due to packet translator in the middle.
1218          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1219          * special handling.
1220          */
1221         if (mtu < IPV6_MMTU)
1222                 return;
1223 #endif
1224
1225         /*
1226          * we reject ICMPv6 too big with abnormally small value.
1227          * XXX what is the good definition of "abnormally small"?
1228          */
1229         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1230                 return;
1231
1232         if (!validated)
1233                 return;
1234
1235         /*
1236          * In case the suggested mtu is less than IPV6_MMTU, we
1237          * only need to remember that it was for above mentioned
1238          * "alwaysfrag" case.
1239          * Try to be as close to the spec as possible.
1240          */
1241         if (mtu < IPV6_MMTU)
1242                 mtu = IPV6_MMTU - 8;
1243
1244         bzero(&inc, sizeof(inc));
1245         inc.inc_flags |= INC_ISIPV6;
1246         inc.inc6_faddr = *dst;
1247         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1248                 return;
1249
1250         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1251                 tcp_hc_updatemtu(&inc, mtu);
1252                 ICMP6STAT_INC(icp6s_pmtuchg);
1253         }
1254 }
1255
1256 /*
1257  * Process a Node Information Query packet, based on
1258  * draft-ietf-ipngwg-icmp-name-lookups-07.
1259  *
1260  * Spec incompatibilities:
1261  * - IPv6 Subject address handling
1262  * - IPv4 Subject address handling support missing
1263  * - Proxy reply (answer even if it's not for me)
1264  * - joins NI group address at in6_ifattach() time only, does not cope
1265  *   with hostname changes by sethostname(3)
1266  */
1267 static struct mbuf *
1268 ni6_input(struct mbuf *m, int off)
1269 {
1270         struct icmp6_nodeinfo *ni6, *nni6;
1271         struct mbuf *n = NULL;
1272         struct prison *pr;
1273         u_int16_t qtype;
1274         int subjlen;
1275         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1276         struct ni_reply_fqdn *fqdn;
1277         int addrs;              /* for NI_QTYPE_NODEADDR */
1278         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1279         struct in6_addr in6_subj; /* subject address */
1280         struct ip6_hdr *ip6;
1281         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1282         char *subj = NULL;
1283         struct in6_ifaddr *ia6 = NULL;
1284
1285         ip6 = mtod(m, struct ip6_hdr *);
1286 #ifndef PULLDOWN_TEST
1287         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1288 #else
1289         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1290         if (ni6 == NULL) {
1291                 /* m is already reclaimed */
1292                 return (NULL);
1293         }
1294 #endif
1295
1296         /*
1297          * Validate IPv6 source address.
1298          * The default configuration MUST be to refuse answering queries from
1299          * global-scope addresses according to RFC4602.
1300          * Notes:
1301          *  - it's not very clear what "refuse" means; this implementation
1302          *    simply drops it.
1303          *  - it's not very easy to identify global-scope (unicast) addresses
1304          *    since there are many prefixes for them.  It should be safer
1305          *    and in practice sufficient to check "all" but loopback and
1306          *    link-local (note that site-local unicast was deprecated and
1307          *    ULA is defined as global scope-wise)
1308          */
1309         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1310             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1311             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1312                 goto bad;
1313
1314         /*
1315          * Validate IPv6 destination address.
1316          *
1317          * The Responder must discard the Query without further processing
1318          * unless it is one of the Responder's unicast or anycast addresses, or
1319          * a link-local scope multicast address which the Responder has joined.
1320          * [RFC4602, Section 5.]
1321          */
1322         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1323                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1324                         goto bad;
1325                 /* else it's a link-local multicast, fine */
1326         } else {                /* unicast or anycast */
1327                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1328                         goto bad; /* XXX impossible */
1329
1330                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1331                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1332                         ifa_free(&ia6->ia_ifa);
1333                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1334                                 "a temporary address in %s:%d",
1335                                __FILE__, __LINE__));
1336                         goto bad;
1337                 }
1338                 ifa_free(&ia6->ia_ifa);
1339         }
1340
1341         /* validate query Subject field. */
1342         qtype = ntohs(ni6->ni_qtype);
1343         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1344         switch (qtype) {
1345         case NI_QTYPE_NOOP:
1346         case NI_QTYPE_SUPTYPES:
1347                 /* 07 draft */
1348                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1349                         break;
1350                 /* FALLTHROUGH */
1351         case NI_QTYPE_FQDN:
1352         case NI_QTYPE_NODEADDR:
1353         case NI_QTYPE_IPV4ADDR:
1354                 switch (ni6->ni_code) {
1355                 case ICMP6_NI_SUBJ_IPV6:
1356 #if ICMP6_NI_SUBJ_IPV6 != 0
1357                 case 0:
1358 #endif
1359                         /*
1360                          * backward compatibility - try to accept 03 draft
1361                          * format, where no Subject is present.
1362                          */
1363                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1364                             subjlen == 0) {
1365                                 oldfqdn++;
1366                                 break;
1367                         }
1368 #if ICMP6_NI_SUBJ_IPV6 != 0
1369                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1370                                 goto bad;
1371 #endif
1372
1373                         if (subjlen != sizeof(struct in6_addr))
1374                                 goto bad;
1375
1376                         /*
1377                          * Validate Subject address.
1378                          *
1379                          * Not sure what exactly "address belongs to the node"
1380                          * means in the spec, is it just unicast, or what?
1381                          *
1382                          * At this moment we consider Subject address as
1383                          * "belong to the node" if the Subject address equals
1384                          * to the IPv6 destination address; validation for
1385                          * IPv6 destination address should have done enough
1386                          * check for us.
1387                          *
1388                          * We do not do proxy at this moment.
1389                          */
1390                         /* m_pulldown instead of copy? */
1391                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1392                             subjlen, (caddr_t)&in6_subj);
1393                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1394                                 goto bad;
1395
1396                         subj = (char *)&in6_subj;
1397                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1398                                 break;
1399
1400                         /*
1401                          * XXX if we are to allow other cases, we should really
1402                          * be careful about scope here.
1403                          * basically, we should disallow queries toward IPv6
1404                          * destination X with subject Y,
1405                          * if scope(X) > scope(Y).
1406                          * if we allow scope(X) > scope(Y), it will result in
1407                          * information leakage across scope boundary.
1408                          */
1409                         goto bad;
1410
1411                 case ICMP6_NI_SUBJ_FQDN:
1412                         /*
1413                          * Validate Subject name with gethostname(3).
1414                          *
1415                          * The behavior may need some debate, since:
1416                          * - we are not sure if the node has FQDN as
1417                          *   hostname (returned by gethostname(3)).
1418                          * - the code does wildcard match for truncated names.
1419                          *   however, we are not sure if we want to perform
1420                          *   wildcard match, if gethostname(3) side has
1421                          *   truncated hostname.
1422                          */
1423                         pr = curthread->td_ucred->cr_prison;
1424                         mtx_lock(&pr->pr_mtx);
1425                         n = ni6_nametodns(pr->pr_hostname,
1426                             strlen(pr->pr_hostname), 0);
1427                         mtx_unlock(&pr->pr_mtx);
1428                         if (!n || n->m_next || n->m_len == 0)
1429                                 goto bad;
1430                         IP6_EXTHDR_GET(subj, char *, m,
1431                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1432                         if (subj == NULL)
1433                                 goto bad;
1434                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1435                             n->m_len)) {
1436                                 goto bad;
1437                         }
1438                         m_freem(n);
1439                         n = NULL;
1440                         break;
1441
1442                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1443                 default:
1444                         goto bad;
1445                 }
1446                 break;
1447         }
1448
1449         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1450         switch (qtype) {
1451         case NI_QTYPE_FQDN:
1452                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1453                         goto bad;
1454                 break;
1455         case NI_QTYPE_NODEADDR:
1456         case NI_QTYPE_IPV4ADDR:
1457                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1458                         goto bad;
1459                 break;
1460         }
1461
1462         /* guess reply length */
1463         switch (qtype) {
1464         case NI_QTYPE_NOOP:
1465                 break;          /* no reply data */
1466         case NI_QTYPE_SUPTYPES:
1467                 replylen += sizeof(u_int32_t);
1468                 break;
1469         case NI_QTYPE_FQDN:
1470                 /* XXX will append an mbuf */
1471                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1472                 break;
1473         case NI_QTYPE_NODEADDR:
1474                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1475                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1476                     sizeof(u_int32_t))) > MCLBYTES)
1477                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1478                 break;
1479         case NI_QTYPE_IPV4ADDR:
1480                 /* unsupported - should respond with unknown Qtype? */
1481                 break;
1482         default:
1483                 /*
1484                  * XXX: We must return a reply with the ICMP6 code
1485                  * `unknown Qtype' in this case.  However we regard the case
1486                  * as an FQDN query for backward compatibility.
1487                  * Older versions set a random value to this field,
1488                  * so it rarely varies in the defined qtypes.
1489                  * But the mechanism is not reliable...
1490                  * maybe we should obsolete older versions.
1491                  */
1492                 qtype = NI_QTYPE_FQDN;
1493                 /* XXX will append an mbuf */
1494                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1495                 oldfqdn++;
1496                 break;
1497         }
1498
1499         /* allocate an mbuf to reply. */
1500         MGETHDR(n, M_DONTWAIT, m->m_type);
1501         if (n == NULL) {
1502                 m_freem(m);
1503                 return (NULL);
1504         }
1505         M_MOVE_PKTHDR(n, m); /* just for recvif and FIB */
1506         if (replylen > MHLEN) {
1507                 if (replylen > MCLBYTES) {
1508                         /*
1509                          * XXX: should we try to allocate more? But MCLBYTES
1510                          * is probably much larger than IPV6_MMTU...
1511                          */
1512                         goto bad;
1513                 }
1514                 MCLGET(n, M_DONTWAIT);
1515                 if ((n->m_flags & M_EXT) == 0) {
1516                         goto bad;
1517                 }
1518         }
1519         n->m_pkthdr.len = n->m_len = replylen;
1520
1521         /* copy mbuf header and IPv6 + Node Information base headers */
1522         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1523         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1524         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1525
1526         /* qtype dependent procedure */
1527         switch (qtype) {
1528         case NI_QTYPE_NOOP:
1529                 nni6->ni_code = ICMP6_NI_SUCCESS;
1530                 nni6->ni_flags = 0;
1531                 break;
1532         case NI_QTYPE_SUPTYPES:
1533         {
1534                 u_int32_t v;
1535                 nni6->ni_code = ICMP6_NI_SUCCESS;
1536                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1537                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1538                 v = (u_int32_t)htonl(0x0000000f);
1539                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1540                 break;
1541         }
1542         case NI_QTYPE_FQDN:
1543                 nni6->ni_code = ICMP6_NI_SUCCESS;
1544                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1545                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1546                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1547                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1548                 /*
1549                  * XXX do we really have FQDN in hostname?
1550                  */
1551                 pr = curthread->td_ucred->cr_prison;
1552                 mtx_lock(&pr->pr_mtx);
1553                 n->m_next = ni6_nametodns(pr->pr_hostname,
1554                     strlen(pr->pr_hostname), oldfqdn);
1555                 mtx_unlock(&pr->pr_mtx);
1556                 if (n->m_next == NULL)
1557                         goto bad;
1558                 /* XXX we assume that n->m_next is not a chain */
1559                 if (n->m_next->m_next != NULL)
1560                         goto bad;
1561                 n->m_pkthdr.len += n->m_next->m_len;
1562                 break;
1563         case NI_QTYPE_NODEADDR:
1564         {
1565                 int lenlim, copied;
1566
1567                 nni6->ni_code = ICMP6_NI_SUCCESS;
1568                 n->m_pkthdr.len = n->m_len =
1569                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1570                 lenlim = M_TRAILINGSPACE(n);
1571                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1572                 /* XXX: reset mbuf length */
1573                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1574                     sizeof(struct icmp6_nodeinfo) + copied;
1575                 break;
1576         }
1577         default:
1578                 break;          /* XXX impossible! */
1579         }
1580
1581         nni6->ni_type = ICMP6_NI_REPLY;
1582         m_freem(m);
1583         return (n);
1584
1585   bad:
1586         m_freem(m);
1587         if (n)
1588                 m_freem(n);
1589         return (NULL);
1590 }
1591
1592 /*
1593  * make a mbuf with DNS-encoded string.  no compression support.
1594  *
1595  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1596  * treated as truncated name (two \0 at the end).  this is a wild guess.
1597  *
1598  * old - return pascal string if non-zero
1599  */
1600 static struct mbuf *
1601 ni6_nametodns(const char *name, int namelen, int old)
1602 {
1603         struct mbuf *m;
1604         char *cp, *ep;
1605         const char *p, *q;
1606         int i, len, nterm;
1607
1608         if (old)
1609                 len = namelen + 1;
1610         else
1611                 len = MCLBYTES;
1612
1613         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1614         MGET(m, M_DONTWAIT, MT_DATA);
1615         if (m && len > MLEN) {
1616                 MCLGET(m, M_DONTWAIT);
1617                 if ((m->m_flags & M_EXT) == 0)
1618                         goto fail;
1619         }
1620         if (!m)
1621                 goto fail;
1622         m->m_next = NULL;
1623
1624         if (old) {
1625                 m->m_len = len;
1626                 *mtod(m, char *) = namelen;
1627                 bcopy(name, mtod(m, char *) + 1, namelen);
1628                 return m;
1629         } else {
1630                 m->m_len = 0;
1631                 cp = mtod(m, char *);
1632                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1633
1634                 /* if not certain about my name, return empty buffer */
1635                 if (namelen == 0)
1636                         return m;
1637
1638                 /*
1639                  * guess if it looks like shortened hostname, or FQDN.
1640                  * shortened hostname needs two trailing "\0".
1641                  */
1642                 i = 0;
1643                 for (p = name; p < name + namelen; p++) {
1644                         if (*p && *p == '.')
1645                                 i++;
1646                 }
1647                 if (i < 2)
1648                         nterm = 2;
1649                 else
1650                         nterm = 1;
1651
1652                 p = name;
1653                 while (cp < ep && p < name + namelen) {
1654                         i = 0;
1655                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1656                                 i++;
1657                         /* result does not fit into mbuf */
1658                         if (cp + i + 1 >= ep)
1659                                 goto fail;
1660                         /*
1661                          * DNS label length restriction, RFC1035 page 8.
1662                          * "i == 0" case is included here to avoid returning
1663                          * 0-length label on "foo..bar".
1664                          */
1665                         if (i <= 0 || i >= 64)
1666                                 goto fail;
1667                         *cp++ = i;
1668                         bcopy(p, cp, i);
1669                         cp += i;
1670                         p = q;
1671                         if (p < name + namelen && *p == '.')
1672                                 p++;
1673                 }
1674                 /* termination */
1675                 if (cp + nterm >= ep)
1676                         goto fail;
1677                 while (nterm-- > 0)
1678                         *cp++ = '\0';
1679                 m->m_len = cp - mtod(m, char *);
1680                 return m;
1681         }
1682
1683         panic("should not reach here");
1684         /* NOTREACHED */
1685
1686  fail:
1687         if (m)
1688                 m_freem(m);
1689         return NULL;
1690 }
1691
1692 /*
1693  * check if two DNS-encoded string matches.  takes care of truncated
1694  * form (with \0\0 at the end).  no compression support.
1695  * XXX upper/lowercase match (see RFC2065)
1696  */
1697 static int
1698 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1699 {
1700         const char *a0, *b0;
1701         int l;
1702
1703         /* simplest case - need validation? */
1704         if (alen == blen && bcmp(a, b, alen) == 0)
1705                 return 1;
1706
1707         a0 = a;
1708         b0 = b;
1709
1710         /* termination is mandatory */
1711         if (alen < 2 || blen < 2)
1712                 return 0;
1713         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1714                 return 0;
1715         alen--;
1716         blen--;
1717
1718         while (a - a0 < alen && b - b0 < blen) {
1719                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1720                         return 0;
1721
1722                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1723                         return 0;
1724                 /* we don't support compression yet */
1725                 if (a[0] >= 64 || b[0] >= 64)
1726                         return 0;
1727
1728                 /* truncated case */
1729                 if (a[0] == 0 && a - a0 == alen - 1)
1730                         return 1;
1731                 if (b[0] == 0 && b - b0 == blen - 1)
1732                         return 1;
1733                 if (a[0] == 0 || b[0] == 0)
1734                         return 0;
1735
1736                 if (a[0] != b[0])
1737                         return 0;
1738                 l = a[0];
1739                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1740                         return 0;
1741                 if (bcmp(a + 1, b + 1, l) != 0)
1742                         return 0;
1743
1744                 a += 1 + l;
1745                 b += 1 + l;
1746         }
1747
1748         if (a - a0 == alen && b - b0 == blen)
1749                 return 1;
1750         else
1751                 return 0;
1752 }
1753
1754 /*
1755  * calculate the number of addresses to be returned in the node info reply.
1756  */
1757 static int
1758 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1759     struct in6_addr *subj)
1760 {
1761         struct ifnet *ifp;
1762         struct in6_ifaddr *ifa6;
1763         struct ifaddr *ifa;
1764         int addrs = 0, addrsofif, iffound = 0;
1765         int niflags = ni6->ni_flags;
1766
1767         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1768                 switch (ni6->ni_code) {
1769                 case ICMP6_NI_SUBJ_IPV6:
1770                         if (subj == NULL) /* must be impossible... */
1771                                 return (0);
1772                         break;
1773                 default:
1774                         /*
1775                          * XXX: we only support IPv6 subject address for
1776                          * this Qtype.
1777                          */
1778                         return (0);
1779                 }
1780         }
1781
1782         IFNET_RLOCK_NOSLEEP();
1783         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1784                 addrsofif = 0;
1785                 IF_ADDR_RLOCK(ifp);
1786                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1787                         if (ifa->ifa_addr->sa_family != AF_INET6)
1788                                 continue;
1789                         ifa6 = (struct in6_ifaddr *)ifa;
1790
1791                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1792                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1793                                 iffound = 1;
1794
1795                         /*
1796                          * IPv4-mapped addresses can only be returned by a
1797                          * Node Information proxy, since they represent
1798                          * addresses of IPv4-only nodes, which perforce do
1799                          * not implement this protocol.
1800                          * [icmp-name-lookups-07, Section 5.4]
1801                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1802                          * this function at this moment.
1803                          */
1804
1805                         /* What do we have to do about ::1? */
1806                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1807                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1808                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1809                                         continue;
1810                                 break;
1811                         case IPV6_ADDR_SCOPE_SITELOCAL:
1812                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1813                                         continue;
1814                                 break;
1815                         case IPV6_ADDR_SCOPE_GLOBAL:
1816                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1817                                         continue;
1818                                 break;
1819                         default:
1820                                 continue;
1821                         }
1822
1823                         /*
1824                          * check if anycast is okay.
1825                          * XXX: just experimental.  not in the spec.
1826                          */
1827                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1828                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1829                                 continue; /* we need only unicast addresses */
1830                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1831                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1832                                 continue;
1833                         }
1834                         addrsofif++; /* count the address */
1835                 }
1836                 IF_ADDR_RUNLOCK(ifp);
1837                 if (iffound) {
1838                         *ifpp = ifp;
1839                         IFNET_RUNLOCK_NOSLEEP();
1840                         return (addrsofif);
1841                 }
1842
1843                 addrs += addrsofif;
1844         }
1845         IFNET_RUNLOCK_NOSLEEP();
1846
1847         return (addrs);
1848 }
1849
1850 static int
1851 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1852     struct ifnet *ifp0, int resid)
1853 {
1854         struct ifnet *ifp;
1855         struct in6_ifaddr *ifa6;
1856         struct ifaddr *ifa;
1857         struct ifnet *ifp_dep = NULL;
1858         int copied = 0, allow_deprecated = 0;
1859         u_char *cp = (u_char *)(nni6 + 1);
1860         int niflags = ni6->ni_flags;
1861         u_int32_t ltime;
1862
1863         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1864                 return (0);     /* needless to copy */
1865
1866         IFNET_RLOCK_NOSLEEP();
1867         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1868   again:
1869
1870         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1871                 IF_ADDR_RLOCK(ifp);
1872                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1873                         if (ifa->ifa_addr->sa_family != AF_INET6)
1874                                 continue;
1875                         ifa6 = (struct in6_ifaddr *)ifa;
1876
1877                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1878                             allow_deprecated == 0) {
1879                                 /*
1880                                  * prefererred address should be put before
1881                                  * deprecated addresses.
1882                                  */
1883
1884                                 /* record the interface for later search */
1885                                 if (ifp_dep == NULL)
1886                                         ifp_dep = ifp;
1887
1888                                 continue;
1889                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1890                             allow_deprecated != 0)
1891                                 continue; /* we now collect deprecated addrs */
1892
1893                         /* What do we have to do about ::1? */
1894                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1895                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1896                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1897                                         continue;
1898                                 break;
1899                         case IPV6_ADDR_SCOPE_SITELOCAL:
1900                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1901                                         continue;
1902                                 break;
1903                         case IPV6_ADDR_SCOPE_GLOBAL:
1904                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1905                                         continue;
1906                                 break;
1907                         default:
1908                                 continue;
1909                         }
1910
1911                         /*
1912                          * check if anycast is okay.
1913                          * XXX: just experimental.  not in the spec.
1914                          */
1915                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1916                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1917                                 continue;
1918                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1919                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1920                                 continue;
1921                         }
1922
1923                         /* now we can copy the address */
1924                         if (resid < sizeof(struct in6_addr) +
1925                             sizeof(u_int32_t)) {
1926                                 IF_ADDR_RUNLOCK(ifp);
1927                                 /*
1928                                  * We give up much more copy.
1929                                  * Set the truncate flag and return.
1930                                  */
1931                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1932                                 IFNET_RUNLOCK_NOSLEEP();
1933                                 return (copied);
1934                         }
1935
1936                         /*
1937                          * Set the TTL of the address.
1938                          * The TTL value should be one of the following
1939                          * according to the specification:
1940                          *
1941                          * 1. The remaining lifetime of a DHCP lease on the
1942                          *    address, or
1943                          * 2. The remaining Valid Lifetime of a prefix from
1944                          *    which the address was derived through Stateless
1945                          *    Autoconfiguration.
1946                          *
1947                          * Note that we currently do not support stateful
1948                          * address configuration by DHCPv6, so the former
1949                          * case can't happen.
1950                          */
1951                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1952                                 ltime = ND6_INFINITE_LIFETIME;
1953                         else {
1954                                 if (ifa6->ia6_lifetime.ia6t_expire >
1955                                     time_second)
1956                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1957                                 else
1958                                         ltime = 0;
1959                         }
1960
1961                         bcopy(&ltime, cp, sizeof(u_int32_t));
1962                         cp += sizeof(u_int32_t);
1963
1964                         /* copy the address itself */
1965                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1966                             sizeof(struct in6_addr));
1967                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1968                         cp += sizeof(struct in6_addr);
1969
1970                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1971                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1972                 }
1973                 IF_ADDR_RUNLOCK(ifp);
1974                 if (ifp0)       /* we need search only on the specified IF */
1975                         break;
1976         }
1977
1978         if (allow_deprecated == 0 && ifp_dep != NULL) {
1979                 ifp = ifp_dep;
1980                 allow_deprecated = 1;
1981
1982                 goto again;
1983         }
1984
1985         IFNET_RUNLOCK_NOSLEEP();
1986
1987         return (copied);
1988 }
1989
1990 /*
1991  * XXX almost dup'ed code with rip6_input.
1992  */
1993 static int
1994 icmp6_rip6_input(struct mbuf **mp, int off)
1995 {
1996         struct mbuf *m = *mp;
1997         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1998         struct inpcb *in6p;
1999         struct inpcb *last = NULL;
2000         struct sockaddr_in6 fromsa;
2001         struct icmp6_hdr *icmp6;
2002         struct mbuf *opts = NULL;
2003
2004 #ifndef PULLDOWN_TEST
2005         /* this is assumed to be safe. */
2006         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
2007 #else
2008         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
2009         if (icmp6 == NULL) {
2010                 /* m is already reclaimed */
2011                 return (IPPROTO_DONE);
2012         }
2013 #endif
2014
2015         /*
2016          * XXX: the address may have embedded scope zone ID, which should be
2017          * hidden from applications.
2018          */
2019         bzero(&fromsa, sizeof(fromsa));
2020         fromsa.sin6_family = AF_INET6;
2021         fromsa.sin6_len = sizeof(struct sockaddr_in6);
2022         fromsa.sin6_addr = ip6->ip6_src;
2023         if (sa6_recoverscope(&fromsa)) {
2024                 m_freem(m);
2025                 return (IPPROTO_DONE);
2026         }
2027
2028         INP_INFO_RLOCK(&V_ripcbinfo);
2029         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2030                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2031                         continue;
2032                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2033                         continue;
2034                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2035                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2036                         continue;
2037                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2038                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2039                         continue;
2040                 INP_RLOCK(in6p);
2041                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2042                     in6p->in6p_icmp6filt)) {
2043                         INP_RUNLOCK(in6p);
2044                         continue;
2045                 }
2046                 if (last != NULL) {
2047                         struct  mbuf *n = NULL;
2048
2049                         /*
2050                          * Recent network drivers tend to allocate a single
2051                          * mbuf cluster, rather than to make a couple of
2052                          * mbufs without clusters.  Also, since the IPv6 code
2053                          * path tries to avoid m_pullup(), it is highly
2054                          * probable that we still have an mbuf cluster here
2055                          * even though the necessary length can be stored in an
2056                          * mbuf's internal buffer.
2057                          * Meanwhile, the default size of the receive socket
2058                          * buffer for raw sockets is not so large.  This means
2059                          * the possibility of packet loss is relatively higher
2060                          * than before.  To avoid this scenario, we copy the
2061                          * received data to a separate mbuf that does not use
2062                          * a cluster, if possible.
2063                          * XXX: it is better to copy the data after stripping
2064                          * intermediate headers.
2065                          */
2066                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2067                             m->m_len <= MHLEN) {
2068                                 MGET(n, M_DONTWAIT, m->m_type);
2069                                 if (n != NULL) {
2070                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2071                                                 bcopy(m->m_data, n->m_data,
2072                                                       m->m_len);
2073                                                 n->m_len = m->m_len;
2074                                         } else {
2075                                                 m_free(n);
2076                                                 n = NULL;
2077                                         }
2078                                 }
2079                         }
2080                         if (n != NULL ||
2081                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2082                                 if (last->inp_flags & INP_CONTROLOPTS)
2083                                         ip6_savecontrol(last, n, &opts);
2084                                 /* strip intermediate headers */
2085                                 m_adj(n, off);
2086                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2087                                 if (sbappendaddr_locked(
2088                                     &last->inp_socket->so_rcv,
2089                                     (struct sockaddr *)&fromsa, n, opts)
2090                                     == 0) {
2091                                         /* should notify about lost packet */
2092                                         m_freem(n);
2093                                         if (opts) {
2094                                                 m_freem(opts);
2095                                         }
2096                                         SOCKBUF_UNLOCK(
2097                                             &last->inp_socket->so_rcv);
2098                                 } else
2099                                         sorwakeup_locked(last->inp_socket);
2100                                 opts = NULL;
2101                         }
2102                         INP_RUNLOCK(last);
2103                 }
2104                 last = in6p;
2105         }
2106         INP_INFO_RUNLOCK(&V_ripcbinfo);
2107         if (last != NULL) {
2108                 if (last->inp_flags & INP_CONTROLOPTS)
2109                         ip6_savecontrol(last, m, &opts);
2110                 /* strip intermediate headers */
2111                 m_adj(m, off);
2112
2113                 /* avoid using mbuf clusters if possible (see above) */
2114                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2115                     m->m_len <= MHLEN) {
2116                         struct mbuf *n;
2117
2118                         MGET(n, M_DONTWAIT, m->m_type);
2119                         if (n != NULL) {
2120                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2121                                         bcopy(m->m_data, n->m_data, m->m_len);
2122                                         n->m_len = m->m_len;
2123
2124                                         m_freem(m);
2125                                         m = n;
2126                                 } else {
2127                                         m_freem(n);
2128                                         n = NULL;
2129                                 }
2130                         }
2131                 }
2132                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2133                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2134                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2135                         m_freem(m);
2136                         if (opts)
2137                                 m_freem(opts);
2138                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2139                 } else
2140                         sorwakeup_locked(last->inp_socket);
2141                 INP_RUNLOCK(last);
2142         } else {
2143                 m_freem(m);
2144                 IP6STAT_DEC(ip6s_delivered);
2145         }
2146         return IPPROTO_DONE;
2147 }
2148
2149 /*
2150  * Reflect the ip6 packet back to the source.
2151  * OFF points to the icmp6 header, counted from the top of the mbuf.
2152  */
2153 void
2154 icmp6_reflect(struct mbuf *m, size_t off)
2155 {
2156         struct ip6_hdr *ip6;
2157         struct icmp6_hdr *icmp6;
2158         struct in6_ifaddr *ia = NULL;
2159         int plen;
2160         int type, code;
2161         struct ifnet *outif = NULL;
2162         struct in6_addr origdst, src, *srcp = NULL;
2163
2164         /* too short to reflect */
2165         if (off < sizeof(struct ip6_hdr)) {
2166                 nd6log((LOG_DEBUG,
2167                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2168                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2169                     __FILE__, __LINE__));
2170                 goto bad;
2171         }
2172
2173         /*
2174          * If there are extra headers between IPv6 and ICMPv6, strip
2175          * off that header first.
2176          */
2177 #ifdef DIAGNOSTIC
2178         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2179                 panic("assumption failed in icmp6_reflect");
2180 #endif
2181         if (off > sizeof(struct ip6_hdr)) {
2182                 size_t l;
2183                 struct ip6_hdr nip6;
2184
2185                 l = off - sizeof(struct ip6_hdr);
2186                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2187                 m_adj(m, l);
2188                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2189                 if (m->m_len < l) {
2190                         if ((m = m_pullup(m, l)) == NULL)
2191                                 return;
2192                 }
2193                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2194         } else /* off == sizeof(struct ip6_hdr) */ {
2195                 size_t l;
2196                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2197                 if (m->m_len < l) {
2198                         if ((m = m_pullup(m, l)) == NULL)
2199                                 return;
2200                 }
2201         }
2202         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2203         ip6 = mtod(m, struct ip6_hdr *);
2204         ip6->ip6_nxt = IPPROTO_ICMPV6;
2205         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2206         type = icmp6->icmp6_type; /* keep type for statistics */
2207         code = icmp6->icmp6_code; /* ditto. */
2208
2209         origdst = ip6->ip6_dst;
2210         /*
2211          * ip6_input() drops a packet if its src is multicast.
2212          * So, the src is never multicast.
2213          */
2214         ip6->ip6_dst = ip6->ip6_src;
2215
2216         /*
2217          * If the incoming packet was addressed directly to us (i.e. unicast),
2218          * use dst as the src for the reply.
2219          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2220          * (for example) when we encounter an error while forwarding procedure
2221          * destined to a duplicated address of ours.
2222          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2223          * procedure of an outgoing packet of our own, in which case we need
2224          * to search in the ifaddr list.
2225          */
2226         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2227                 if ((ia = ip6_getdstifaddr(m))) {
2228                         if (!(ia->ia6_flags &
2229                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2230                                 srcp = &ia->ia_addr.sin6_addr;
2231                 } else {
2232                         struct sockaddr_in6 d;
2233
2234                         bzero(&d, sizeof(d));
2235                         d.sin6_family = AF_INET6;
2236                         d.sin6_len = sizeof(d);
2237                         d.sin6_addr = origdst;
2238                         ia = (struct in6_ifaddr *)
2239                             ifa_ifwithaddr((struct sockaddr *)&d);
2240                         if (ia &&
2241                             !(ia->ia6_flags &
2242                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2243                                 srcp = &ia->ia_addr.sin6_addr;
2244                         }
2245                 }
2246         }
2247
2248         if (srcp == NULL) {
2249                 int e;
2250                 struct sockaddr_in6 sin6;
2251                 struct route_in6 ro;
2252
2253                 /*
2254                  * This case matches to multicasts, our anycast, or unicasts
2255                  * that we do not own.  Select a source address based on the
2256                  * source address of the erroneous packet.
2257                  */
2258                 bzero(&sin6, sizeof(sin6));
2259                 sin6.sin6_family = AF_INET6;
2260                 sin6.sin6_len = sizeof(sin6);
2261                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2262
2263                 bzero(&ro, sizeof(ro));
2264                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2265                 if (ro.ro_rt)
2266                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2267                 if (e) {
2268                         char ip6buf[INET6_ADDRSTRLEN];
2269                         nd6log((LOG_DEBUG,
2270                             "icmp6_reflect: source can't be determined: "
2271                             "dst=%s, error=%d\n",
2272                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2273                         goto bad;
2274                 }
2275                 srcp = &src;
2276         }
2277
2278         ip6->ip6_src = *srcp;
2279         ip6->ip6_flow = 0;
2280         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2281         ip6->ip6_vfc |= IPV6_VERSION;
2282         ip6->ip6_nxt = IPPROTO_ICMPV6;
2283         if (outif)
2284                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2285         else if (m->m_pkthdr.rcvif) {
2286                 /* XXX: This may not be the outgoing interface */
2287                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2288         } else
2289                 ip6->ip6_hlim = V_ip6_defhlim;
2290
2291         icmp6->icmp6_cksum = 0;
2292         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2293             sizeof(struct ip6_hdr), plen);
2294
2295         /*
2296          * XXX option handling
2297          */
2298
2299         m->m_flags &= ~(M_BCAST|M_MCAST);
2300
2301         m_addr_changed(m);
2302
2303         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2304         if (outif)
2305                 icmp6_ifoutstat_inc(outif, type, code);
2306
2307         if (ia != NULL)
2308                 ifa_free(&ia->ia_ifa);
2309         return;
2310
2311  bad:
2312         if (ia != NULL)
2313                 ifa_free(&ia->ia_ifa);
2314         m_freem(m);
2315         return;
2316 }
2317
2318 void
2319 icmp6_fasttimo(void)
2320 {
2321
2322         mld_fasttimo();
2323 }
2324
2325 void
2326 icmp6_slowtimo(void)
2327 {
2328
2329         mld_slowtimo();
2330 }
2331
2332 static const char *
2333 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2334     struct in6_addr *tgt6)
2335 {
2336         static char buf[1024];
2337         char ip6bufs[INET6_ADDRSTRLEN];
2338         char ip6bufd[INET6_ADDRSTRLEN];
2339         char ip6buft[INET6_ADDRSTRLEN];
2340         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2341             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2342             ip6_sprintf(ip6buft, tgt6));
2343         return buf;
2344 }
2345
2346 void
2347 icmp6_redirect_input(struct mbuf *m, int off)
2348 {
2349         struct ifnet *ifp;
2350         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2351         struct nd_redirect *nd_rd;
2352         int icmp6len = ntohs(ip6->ip6_plen);
2353         char *lladdr = NULL;
2354         int lladdrlen = 0;
2355         struct rtentry *rt = NULL;
2356         int is_router;
2357         int is_onlink;
2358         struct in6_addr src6 = ip6->ip6_src;
2359         struct in6_addr redtgt6;
2360         struct in6_addr reddst6;
2361         union nd_opts ndopts;
2362         char ip6buf[INET6_ADDRSTRLEN];
2363
2364         M_ASSERTPKTHDR(m);
2365         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2366
2367         ifp = m->m_pkthdr.rcvif;
2368
2369         /* XXX if we are router, we don't update route by icmp6 redirect */
2370         if (V_ip6_forwarding)
2371                 goto freeit;
2372         if (!V_icmp6_rediraccept)
2373                 goto freeit;
2374
2375 #ifndef PULLDOWN_TEST
2376         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2377         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2378 #else
2379         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2380         if (nd_rd == NULL) {
2381                 ICMP6STAT_INC(icp6s_tooshort);
2382                 return;
2383         }
2384 #endif
2385         redtgt6 = nd_rd->nd_rd_target;
2386         reddst6 = nd_rd->nd_rd_dst;
2387
2388         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2389             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2390                 goto freeit;
2391         }
2392
2393         /* validation */
2394         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2395                 nd6log((LOG_ERR,
2396                     "ICMP6 redirect sent from %s rejected; "
2397                     "must be from linklocal\n",
2398                     ip6_sprintf(ip6buf, &src6)));
2399                 goto bad;
2400         }
2401         if (ip6->ip6_hlim != 255) {
2402                 nd6log((LOG_ERR,
2403                     "ICMP6 redirect sent from %s rejected; "
2404                     "hlim=%d (must be 255)\n",
2405                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2406                 goto bad;
2407         }
2408     {
2409         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2410         struct sockaddr_in6 sin6;
2411         struct in6_addr *gw6;
2412
2413         bzero(&sin6, sizeof(sin6));
2414         sin6.sin6_family = AF_INET6;
2415         sin6.sin6_len = sizeof(struct sockaddr_in6);
2416         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2417         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2418         if (rt) {
2419                 if (rt->rt_gateway == NULL ||
2420                     rt->rt_gateway->sa_family != AF_INET6) {
2421                         RTFREE_LOCKED(rt);
2422                         nd6log((LOG_ERR,
2423                             "ICMP6 redirect rejected; no route "
2424                             "with inet6 gateway found for redirect dst: %s\n",
2425                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2426                         goto bad;
2427                 }
2428
2429                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2430                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2431                         RTFREE_LOCKED(rt);
2432                         nd6log((LOG_ERR,
2433                             "ICMP6 redirect rejected; "
2434                             "not equal to gw-for-src=%s (must be same): "
2435                             "%s\n",
2436                             ip6_sprintf(ip6buf, gw6),
2437                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2438                         goto bad;
2439                 }
2440         } else {
2441                 nd6log((LOG_ERR,
2442                     "ICMP6 redirect rejected; "
2443                     "no route found for redirect dst: %s\n",
2444                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2445                 goto bad;
2446         }
2447         RTFREE_LOCKED(rt);
2448         rt = NULL;
2449     }
2450         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2451                 nd6log((LOG_ERR,
2452                     "ICMP6 redirect rejected; "
2453                     "redirect dst must be unicast: %s\n",
2454                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2455                 goto bad;
2456         }
2457
2458         is_router = is_onlink = 0;
2459         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2460                 is_router = 1;  /* router case */
2461         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2462                 is_onlink = 1;  /* on-link destination case */
2463         if (!is_router && !is_onlink) {
2464                 nd6log((LOG_ERR,
2465                     "ICMP6 redirect rejected; "
2466                     "neither router case nor onlink case: %s\n",
2467                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2468                 goto bad;
2469         }
2470         /* validation passed */
2471
2472         icmp6len -= sizeof(*nd_rd);
2473         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2474         if (nd6_options(&ndopts) < 0) {
2475                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2476                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2477                 /* nd6_options have incremented stats */
2478                 goto freeit;
2479         }
2480
2481         if (ndopts.nd_opts_tgt_lladdr) {
2482                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2483                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2484         }
2485
2486         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2487                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2488                     "(if %d, icmp6 packet %d): %s\n",
2489                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2490                     ifp->if_addrlen, lladdrlen - 2,
2491                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2492                 goto bad;
2493         }
2494
2495         /* RFC 2461 8.3 */
2496         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2497             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2498
2499         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2500                 /* perform rtredirect */
2501                 struct sockaddr_in6 sdst;
2502                 struct sockaddr_in6 sgw;
2503                 struct sockaddr_in6 ssrc;
2504                 u_int fibnum;
2505
2506                 bzero(&sdst, sizeof(sdst));
2507                 bzero(&sgw, sizeof(sgw));
2508                 bzero(&ssrc, sizeof(ssrc));
2509                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2510                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2511                         sizeof(struct sockaddr_in6);
2512                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2513                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2514                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2515                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2516                         in6_rtredirect((struct sockaddr *)&sdst,
2517                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2518                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2519                             fibnum);
2520         }
2521         /* finally update cached route in each socket via pfctlinput */
2522     {
2523         struct sockaddr_in6 sdst;
2524
2525         bzero(&sdst, sizeof(sdst));
2526         sdst.sin6_family = AF_INET6;
2527         sdst.sin6_len = sizeof(struct sockaddr_in6);
2528         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2529         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2530 #ifdef IPSEC
2531         key_sa_routechange((struct sockaddr *)&sdst);
2532 #endif /* IPSEC */
2533     }
2534
2535  freeit:
2536         m_freem(m);
2537         return;
2538
2539  bad:
2540         ICMP6STAT_INC(icp6s_badredirect);
2541         m_freem(m);
2542 }
2543
2544 void
2545 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2546 {
2547         struct ifnet *ifp;      /* my outgoing interface */
2548         struct in6_addr *ifp_ll6;
2549         struct in6_addr *router_ll6;
2550         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2551         struct mbuf *m = NULL;  /* newly allocated one */
2552         struct m_tag *mtag;
2553         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2554         struct nd_redirect *nd_rd;
2555         struct llentry *ln = NULL;
2556         size_t maxlen;
2557         u_char *p;
2558         struct ifnet *outif = NULL;
2559         struct sockaddr_in6 src_sa;
2560
2561         icmp6_errcount(ND_REDIRECT, 0);
2562
2563         /* if we are not router, we don't send icmp6 redirect */
2564         if (!V_ip6_forwarding)
2565                 goto fail;
2566
2567         /* sanity check */
2568         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2569                 goto fail;
2570
2571         /*
2572          * Address check:
2573          *  the source address must identify a neighbor, and
2574          *  the destination address must not be a multicast address
2575          *  [RFC 2461, sec 8.2]
2576          */
2577         sip6 = mtod(m0, struct ip6_hdr *);
2578         bzero(&src_sa, sizeof(src_sa));
2579         src_sa.sin6_family = AF_INET6;
2580         src_sa.sin6_len = sizeof(src_sa);
2581         src_sa.sin6_addr = sip6->ip6_src;
2582         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2583                 goto fail;
2584         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2585                 goto fail;      /* what should we do here? */
2586
2587         /* rate limit */
2588         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2589                 goto fail;
2590
2591         /*
2592          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2593          * we almost always ask for an mbuf cluster for simplicity.
2594          * (MHLEN < IPV6_MMTU is almost always true)
2595          */
2596 #if IPV6_MMTU >= MCLBYTES
2597 # error assumption failed about IPV6_MMTU and MCLBYTES
2598 #endif
2599         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2600         if (m && IPV6_MMTU >= MHLEN)
2601                 MCLGET(m, M_DONTWAIT);
2602         if (!m)
2603                 goto fail;
2604         M_SETFIB(m, rt->rt_fibnum);
2605         m->m_pkthdr.rcvif = NULL;
2606         m->m_len = 0;
2607         maxlen = M_TRAILINGSPACE(m);
2608         maxlen = min(IPV6_MMTU, maxlen);
2609         /* just for safety */
2610         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2611             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2612                 goto fail;
2613         }
2614
2615         {
2616                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2617                 struct in6_ifaddr *ia;
2618                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2619                                                  IN6_IFF_NOTREADY|
2620                                                  IN6_IFF_ANYCAST)) == NULL)
2621                         goto fail;
2622                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2623                 /* XXXRW: reference released prematurely. */
2624                 ifa_free(&ia->ia_ifa);
2625         }
2626
2627         /* get ip6 linklocal address for the router. */
2628         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2629                 struct sockaddr_in6 *sin6;
2630                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2631                 router_ll6 = &sin6->sin6_addr;
2632                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2633                         router_ll6 = (struct in6_addr *)NULL;
2634         } else
2635                 router_ll6 = (struct in6_addr *)NULL;
2636
2637         /* ip6 */
2638         ip6 = mtod(m, struct ip6_hdr *);
2639         ip6->ip6_flow = 0;
2640         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2641         ip6->ip6_vfc |= IPV6_VERSION;
2642         /* ip6->ip6_plen will be set later */
2643         ip6->ip6_nxt = IPPROTO_ICMPV6;
2644         ip6->ip6_hlim = 255;
2645         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2646         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2647         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2648
2649         /* ND Redirect */
2650         nd_rd = (struct nd_redirect *)(ip6 + 1);
2651         nd_rd->nd_rd_type = ND_REDIRECT;
2652         nd_rd->nd_rd_code = 0;
2653         nd_rd->nd_rd_reserved = 0;
2654         if (rt->rt_flags & RTF_GATEWAY) {
2655                 /*
2656                  * nd_rd->nd_rd_target must be a link-local address in
2657                  * better router cases.
2658                  */
2659                 if (!router_ll6)
2660                         goto fail;
2661                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2662                     sizeof(nd_rd->nd_rd_target));
2663                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2664                     sizeof(nd_rd->nd_rd_dst));
2665         } else {
2666                 /* make sure redtgt == reddst */
2667                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2668                     sizeof(nd_rd->nd_rd_target));
2669                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2670                     sizeof(nd_rd->nd_rd_dst));
2671         }
2672
2673         p = (u_char *)(nd_rd + 1);
2674
2675         if (!router_ll6)
2676                 goto nolladdropt;
2677
2678         {
2679                 /* target lladdr option */
2680                 int len;
2681                 struct nd_opt_hdr *nd_opt;
2682                 char *lladdr;
2683
2684                 IF_AFDATA_RLOCK(ifp);
2685                 ln = nd6_lookup(router_ll6, 0, ifp);
2686                 IF_AFDATA_RUNLOCK(ifp);
2687                 if (ln == NULL)
2688                         goto nolladdropt;
2689
2690                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2691                 len = (len + 7) & ~7;   /* round by 8 */
2692                 /* safety check */
2693                 if (len + (p - (u_char *)ip6) > maxlen)                         
2694                         goto nolladdropt;
2695
2696                 if (ln->la_flags & LLE_VALID) {
2697                         nd_opt = (struct nd_opt_hdr *)p;
2698                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2699                         nd_opt->nd_opt_len = len >> 3;
2700                         lladdr = (char *)(nd_opt + 1);
2701                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2702                         p += len;
2703                 }
2704         }
2705 nolladdropt:
2706         if (ln != NULL)
2707                 LLE_RUNLOCK(ln);
2708                 
2709         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2710
2711         /* just to be safe */
2712 #ifdef M_DECRYPTED      /*not openbsd*/
2713         if (m0->m_flags & M_DECRYPTED)
2714                 goto noredhdropt;
2715 #endif
2716         if (p - (u_char *)ip6 > maxlen)
2717                 goto noredhdropt;
2718
2719         {
2720                 /* redirected header option */
2721                 int len;
2722                 struct nd_opt_rd_hdr *nd_opt_rh;
2723
2724                 /*
2725                  * compute the maximum size for icmp6 redirect header option.
2726                  * XXX room for auth header?
2727                  */
2728                 len = maxlen - (p - (u_char *)ip6);
2729                 len &= ~7;
2730
2731                 /* This is just for simplicity. */
2732                 if (m0->m_pkthdr.len != m0->m_len) {
2733                         if (m0->m_next) {
2734                                 m_freem(m0->m_next);
2735                                 m0->m_next = NULL;
2736                         }
2737                         m0->m_pkthdr.len = m0->m_len;
2738                 }
2739
2740                 /*
2741                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2742                  * about padding/truncate rule for the original IP packet.
2743                  * From the discussion on IPv6imp in Feb 1999,
2744                  * the consensus was:
2745                  * - "attach as much as possible" is the goal
2746                  * - pad if not aligned (original size can be guessed by
2747                  *   original ip6 header)
2748                  * Following code adds the padding if it is simple enough,
2749                  * and truncates if not.
2750                  */
2751                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2752                         panic("assumption failed in %s:%d", __FILE__,
2753                             __LINE__);
2754
2755                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2756                         /* not enough room, truncate */
2757                         m0->m_pkthdr.len = m0->m_len = len -
2758                             sizeof(*nd_opt_rh);
2759                 } else {
2760                         /* enough room, pad or truncate */
2761                         size_t extra;
2762
2763                         extra = m0->m_pkthdr.len % 8;
2764                         if (extra) {
2765                                 /* pad if easy enough, truncate if not */
2766                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2767                                         /* pad */
2768                                         m0->m_len += (8 - extra);
2769                                         m0->m_pkthdr.len += (8 - extra);
2770                                 } else {
2771                                         /* truncate */
2772                                         m0->m_pkthdr.len -= extra;
2773                                         m0->m_len -= extra;
2774                                 }
2775                         }
2776                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2777                         m0->m_pkthdr.len = m0->m_len = len -
2778                             sizeof(*nd_opt_rh);
2779                 }
2780
2781                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2782                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2783                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2784                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2785                 p += sizeof(*nd_opt_rh);
2786                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2787
2788                 /* connect m0 to m */
2789                 m_tag_delete_chain(m0, NULL);
2790                 m0->m_flags &= ~M_PKTHDR;
2791                 m->m_next = m0;
2792                 m->m_pkthdr.len = m->m_len + m0->m_len;
2793                 m0 = NULL;
2794         }
2795 noredhdropt:;
2796         if (m0) {
2797                 m_freem(m0);
2798                 m0 = NULL;
2799         }
2800
2801         /* XXX: clear embedded link IDs in the inner header */
2802         in6_clearscope(&sip6->ip6_src);
2803         in6_clearscope(&sip6->ip6_dst);
2804         in6_clearscope(&nd_rd->nd_rd_target);
2805         in6_clearscope(&nd_rd->nd_rd_dst);
2806
2807         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2808
2809         nd_rd->nd_rd_cksum = 0;
2810         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2811             sizeof(*ip6), ntohs(ip6->ip6_plen));
2812
2813         if (send_sendso_input_hook != NULL) {
2814                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2815                         M_NOWAIT);
2816                 if (mtag == NULL)
2817                         goto fail;
2818                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2819                 m_tag_prepend(m, mtag);
2820         }
2821
2822         /* send the packet to outside... */
2823         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2824         if (outif) {
2825                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2826                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2827         }
2828         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2829
2830         return;
2831
2832 fail:
2833         if (m)
2834                 m_freem(m);
2835         if (m0)
2836                 m_freem(m0);
2837 }
2838
2839 /*
2840  * ICMPv6 socket option processing.
2841  */
2842 int
2843 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2844 {
2845         int error = 0;
2846         int optlen;
2847         struct inpcb *inp = sotoinpcb(so);
2848         int level, op, optname;
2849
2850         if (sopt) {
2851                 level = sopt->sopt_level;
2852                 op = sopt->sopt_dir;
2853                 optname = sopt->sopt_name;
2854                 optlen = sopt->sopt_valsize;
2855         } else
2856                 level = op = optname = optlen = 0;
2857
2858         if (level != IPPROTO_ICMPV6) {
2859                 return EINVAL;
2860         }
2861
2862         switch (op) {
2863         case PRCO_SETOPT:
2864                 switch (optname) {
2865                 case ICMP6_FILTER:
2866                     {
2867                         struct icmp6_filter ic6f;
2868
2869                         if (optlen != sizeof(ic6f)) {
2870                                 error = EMSGSIZE;
2871                                 break;
2872                         }
2873                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2874                         if (error == 0) {
2875                                 INP_WLOCK(inp);
2876                                 *inp->in6p_icmp6filt = ic6f;
2877                                 INP_WUNLOCK(inp);
2878                         }
2879                         break;
2880                     }
2881
2882                 default:
2883                         error = ENOPROTOOPT;
2884                         break;
2885                 }
2886                 break;
2887
2888         case PRCO_GETOPT:
2889                 switch (optname) {
2890                 case ICMP6_FILTER:
2891                     {
2892                         struct icmp6_filter ic6f;
2893
2894                         INP_RLOCK(inp);
2895                         ic6f = *inp->in6p_icmp6filt;
2896                         INP_RUNLOCK(inp);
2897                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2898                         break;
2899                     }
2900
2901                 default:
2902                         error = ENOPROTOOPT;
2903                         break;
2904                 }
2905                 break;
2906         }
2907
2908         return (error);
2909 }
2910
2911 /*
2912  * Perform rate limit check.
2913  * Returns 0 if it is okay to send the icmp6 packet.
2914  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2915  * limitation.
2916  *
2917  * XXX per-destination/type check necessary?
2918  *
2919  * dst - not used at this moment
2920  * type - not used at this moment
2921  * code - not used at this moment
2922  */
2923 static int
2924 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2925     const int code)
2926 {
2927         int ret;
2928
2929         ret = 0;        /* okay to send */
2930
2931         /* PPS limit */
2932         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2933             V_icmp6errppslim)) {
2934                 /* The packet is subject to rate limit */
2935                 ret++;
2936         }
2937
2938         return ret;
2939 }