]> CyberLeo.Net >> Repos - FreeBSD/stable/9.git/blob - sys/netinet6/icmp6.c
MFC r339587:
[FreeBSD/stable/9.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 == NULL) {
365                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
366                 return;
367         }
368
369         nip6 = mtod(m, struct ip6_hdr *);
370         nip6->ip6_src  = oip6->ip6_src;
371         nip6->ip6_dst  = oip6->ip6_dst;
372
373         in6_clearscope(&oip6->ip6_src);
374         in6_clearscope(&oip6->ip6_dst);
375
376         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
377         icmp6->icmp6_type = type;
378         icmp6->icmp6_code = code;
379         icmp6->icmp6_pptr = htonl((u_int32_t)param);
380
381         /*
382          * icmp6_reflect() is designed to be in the input path.
383          * icmp6_error() can be called from both input and output path,
384          * and if we are in output path rcvif could contain bogus value.
385          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
386          * information in ip header (nip6).
387          */
388         m->m_pkthdr.rcvif = NULL;
389
390         ICMP6STAT_INC(icp6s_outhist[type]);
391         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
392
393         return;
394
395   freeit:
396         /*
397          * If we can't tell whether or not we can generate ICMP6, free it.
398          */
399         m_freem(m);
400 }
401
402 /*
403  * Process a received ICMP6 message.
404  */
405 int
406 icmp6_input(struct mbuf **mp, int *offp, int proto)
407 {
408         struct mbuf *m = *mp, *n;
409         struct ifnet *ifp;
410         struct ip6_hdr *ip6, *nip6;
411         struct icmp6_hdr *icmp6, *nicmp6;
412         int off = *offp;
413         int icmp6len = m->m_pkthdr.len - *offp;
414         int code, sum, noff;
415         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
416         int ip6len, error;
417
418         ifp = m->m_pkthdr.rcvif;
419
420 #ifndef PULLDOWN_TEST
421         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
422         /* m might change if M_LOOP.  So, call mtod after this */
423 #endif
424
425         /*
426          * Locate icmp6 structure in mbuf, and check
427          * that not corrupted and of at least minimum length
428          */
429
430         ip6 = mtod(m, struct ip6_hdr *);
431         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
432         if (icmp6len < sizeof(struct icmp6_hdr)) {
433                 ICMP6STAT_INC(icp6s_tooshort);
434                 goto freeit;
435         }
436
437         /*
438          * Check multicast group membership.
439          * Note: SSM filters are not applied for ICMPv6 traffic.
440          */
441         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
442                 struct in6_multi        *inm;
443
444                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
445                 if (inm == NULL) {
446                         IP6STAT_INC(ip6s_notmember);
447                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
448                         goto freeit;
449                 }
450         }
451
452         /*
453          * calculate the checksum
454          */
455 #ifndef PULLDOWN_TEST
456         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
457 #else
458         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
459         if (icmp6 == NULL) {
460                 ICMP6STAT_INC(icp6s_tooshort);
461                 return IPPROTO_DONE;
462         }
463 #endif
464         code = icmp6->icmp6_code;
465
466         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
467                 nd6log((LOG_ERR,
468                     "ICMP6 checksum error(%d|%x) %s\n",
469                     icmp6->icmp6_type, sum,
470                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
471                 ICMP6STAT_INC(icp6s_checksum);
472                 goto freeit;
473         }
474
475         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
476                 /*
477                  * Deliver very specific ICMP6 type only.
478                  * This is important to deliver TOOBIG.  Otherwise PMTUD
479                  * will not work.
480                  */
481                 switch (icmp6->icmp6_type) {
482                 case ICMP6_DST_UNREACH:
483                 case ICMP6_PACKET_TOO_BIG:
484                 case ICMP6_TIME_EXCEEDED:
485                         break;
486                 default:
487                         goto freeit;
488                 }
489         }
490
491         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
492         icmp6_ifstat_inc(ifp, ifs6_in_msg);
493         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
494                 icmp6_ifstat_inc(ifp, ifs6_in_error);
495
496         switch (icmp6->icmp6_type) {
497         case ICMP6_DST_UNREACH:
498                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
499                 switch (code) {
500                 case ICMP6_DST_UNREACH_NOROUTE:
501                         code = PRC_UNREACH_NET;
502                         break;
503                 case ICMP6_DST_UNREACH_ADMIN:
504                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
505                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
506                         break;
507                 case ICMP6_DST_UNREACH_ADDR:
508                         code = PRC_HOSTDEAD;
509                         break;
510                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
511                         /* I mean "source address was incorrect." */
512                         code = PRC_PARAMPROB;
513                         break;
514                 case ICMP6_DST_UNREACH_NOPORT:
515                         code = PRC_UNREACH_PORT;
516                         break;
517                 default:
518                         goto badcode;
519                 }
520                 goto deliver;
521                 break;
522
523         case ICMP6_PACKET_TOO_BIG:
524                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
525
526                 /* validation is made in icmp6_mtudisc_update */
527
528                 code = PRC_MSGSIZE;
529
530                 /*
531                  * Updating the path MTU will be done after examining
532                  * intermediate extension headers.
533                  */
534                 goto deliver;
535                 break;
536
537         case ICMP6_TIME_EXCEEDED:
538                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
539                 switch (code) {
540                 case ICMP6_TIME_EXCEED_TRANSIT:
541                         code = PRC_TIMXCEED_INTRANS;
542                         break;
543                 case ICMP6_TIME_EXCEED_REASSEMBLY:
544                         code = PRC_TIMXCEED_REASS;
545                         break;
546                 default:
547                         goto badcode;
548                 }
549                 goto deliver;
550                 break;
551
552         case ICMP6_PARAM_PROB:
553                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
554                 switch (code) {
555                 case ICMP6_PARAMPROB_NEXTHEADER:
556                         code = PRC_UNREACH_PROTOCOL;
557                         break;
558                 case ICMP6_PARAMPROB_HEADER:
559                 case ICMP6_PARAMPROB_OPTION:
560                         code = PRC_PARAMPROB;
561                         break;
562                 default:
563                         goto badcode;
564                 }
565                 goto deliver;
566                 break;
567
568         case ICMP6_ECHO_REQUEST:
569                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
570                 if (code != 0)
571                         goto badcode;
572                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
573                         /* Give up remote */
574                         break;
575                 }
576                 if ((n->m_flags & M_EXT) != 0
577                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
578                         struct mbuf *n0 = n;
579                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
580                         int n0len;
581
582                         MGETHDR(n, M_DONTWAIT, n0->m_type);
583                         n0len = n0->m_pkthdr.len;       /* save for use below */
584                         if (n)
585                                 M_MOVE_PKTHDR(n, n0);   /* FIB copied. */
586                         if (n && maxlen >= MHLEN) {
587                                 MCLGET(n, M_DONTWAIT);
588                                 if ((n->m_flags & M_EXT) == 0) {
589                                         m_free(n);
590                                         n = NULL;
591                                 }
592                         }
593                         if (n == NULL) {
594                                 /* Give up remote */
595                                 m_freem(n0);
596                                 break;
597                         }
598                         /*
599                          * Copy IPv6 and ICMPv6 only.
600                          */
601                         nip6 = mtod(n, struct ip6_hdr *);
602                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
603                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
604                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
605                         noff = sizeof(struct ip6_hdr);
606                         /* new mbuf contains only ipv6+icmpv6 headers */
607                         n->m_len = noff + sizeof(struct icmp6_hdr);
608                         /*
609                          * Adjust mbuf.  ip6_plen will be adjusted in
610                          * ip6_output().
611                          */
612                         m_adj(n0, off + sizeof(struct icmp6_hdr));
613                         /* recalculate complete packet size */
614                         n->m_pkthdr.len = n0len + (noff - off);
615                         n->m_next = n0;
616                 } else {
617                         nip6 = mtod(n, struct ip6_hdr *);
618                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
619                             sizeof(*nicmp6));
620                         noff = off;
621                 }
622                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
623                 nicmp6->icmp6_code = 0;
624                 if (n) {
625                         ICMP6STAT_INC(icp6s_reflect);
626                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
627                         icmp6_reflect(n, noff);
628                 }
629                 break;
630
631         case ICMP6_ECHO_REPLY:
632                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
633                 if (code != 0)
634                         goto badcode;
635                 break;
636
637         case MLD_LISTENER_QUERY:
638         case MLD_LISTENER_REPORT:
639         case MLD_LISTENER_DONE:
640         case MLDV2_LISTENER_REPORT:
641                 /*
642                  * Drop MLD traffic which is not link-local, has a hop limit
643                  * of greater than 1 hop, or which does not have the
644                  * IPv6 HBH Router Alert option.
645                  * As IPv6 HBH options are stripped in ip6_input() we must
646                  * check an mbuf header flag.
647                  * XXX Should we also sanity check that these messages
648                  * were directed to a link-local multicast prefix?
649                  */
650                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
651                         goto freeit;
652                 if (mld_input(m, off, icmp6len) != 0)
653                         return (IPPROTO_DONE);
654                 /* m stays. */
655                 break;
656
657         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
658             {
659                 enum { WRU, FQDN } mode;
660
661                 if (!V_icmp6_nodeinfo)
662                         break;
663
664                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
665                         mode = WRU;
666                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
667                         mode = FQDN;
668                 else
669                         goto badlen;
670
671                 if (mode == FQDN) {
672 #ifndef PULLDOWN_TEST
673                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
674                             IPPROTO_DONE);
675 #endif
676                         n = m_copy(m, 0, M_COPYALL);
677                         if (n)
678                                 n = ni6_input(n, off);
679                         /* XXX meaningless if n == NULL */
680                         noff = sizeof(struct ip6_hdr);
681                 } else {
682                         struct prison *pr;
683                         u_char *p;
684                         int maxlen, maxhlen, hlen;
685
686                         /*
687                          * XXX: this combination of flags is pointless,
688                          * but should we keep this for compatibility?
689                          */
690                         if ((V_icmp6_nodeinfo & 5) != 5)
691                                 break;
692
693                         if (code != 0)
694                                 goto badcode;
695                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
696                         if (maxlen >= MCLBYTES) {
697                                 /* Give up remote */
698                                 break;
699                         }
700                         MGETHDR(n, M_DONTWAIT, m->m_type);
701                         if (n && maxlen > MHLEN) {
702                                 MCLGET(n, M_DONTWAIT);
703                                 if ((n->m_flags & M_EXT) == 0) {
704                                         m_free(n);
705                                         n = NULL;
706                                 }
707                         }
708                         if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
709                                 /*
710                                  * Previous code did a blind M_COPY_PKTHDR
711                                  * and said "just for rcvif".  If true, then
712                                  * we could tolerate the dup failing (due to
713                                  * the deep copy of the tag chain).  For now
714                                  * be conservative and just fail.
715                                  */
716                                 m_free(n);
717                                 n = NULL;
718                         }
719                         if (n == NULL) {
720                                 /* Give up remote */
721                                 break;
722                         }
723                         n->m_pkthdr.rcvif = NULL;
724                         n->m_len = 0;
725                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
726                         pr = curthread->td_ucred->cr_prison;
727                         mtx_lock(&pr->pr_mtx);
728                         hlen = strlen(pr->pr_hostname);
729                         if (maxhlen > hlen)
730                                 maxhlen = hlen;
731                         /*
732                          * Copy IPv6 and ICMPv6 only.
733                          */
734                         nip6 = mtod(n, struct ip6_hdr *);
735                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
736                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
737                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
738                         p = (u_char *)(nicmp6 + 1);
739                         bzero(p, 4);
740                         /* meaningless TTL */
741                         bcopy(pr->pr_hostname, p + 4, maxhlen);
742                         mtx_unlock(&pr->pr_mtx);
743                         noff = sizeof(struct ip6_hdr);
744                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
745                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
746                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
747                         nicmp6->icmp6_code = 0;
748                 }
749                 if (n) {
750                         ICMP6STAT_INC(icp6s_reflect);
751                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
752                         icmp6_reflect(n, noff);
753                 }
754                 break;
755             }
756
757         case ICMP6_WRUREPLY:
758                 if (code != 0)
759                         goto badcode;
760                 break;
761
762         case ND_ROUTER_SOLICIT:
763                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
764                 if (code != 0)
765                         goto badcode;
766                 if (icmp6len < sizeof(struct nd_router_solicit))
767                         goto badlen;
768                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
769                         /* give up local */
770
771                         /* Send incoming SeND packet to user space. */
772                         if (send_sendso_input_hook != NULL) {
773                                 IP6_EXTHDR_CHECK(m, off,
774                                     icmp6len, IPPROTO_DONE);
775                                 error = send_sendso_input_hook(m, ifp,
776                                     SND_IN, ip6len);
777                                 /* -1 == no app on SEND socket */
778                                 if (error == 0)
779                                         return (IPPROTO_DONE);
780                                 nd6_rs_input(m, off, icmp6len);
781                         } else
782                                 nd6_rs_input(m, off, icmp6len);
783                         m = NULL;
784                         goto freeit;
785                 }
786                 if (send_sendso_input_hook != NULL) {
787                         IP6_EXTHDR_CHECK(n, off,
788                             icmp6len, IPPROTO_DONE);
789                         error = send_sendso_input_hook(n, ifp,
790                             SND_IN, ip6len);
791                         if (error == 0)
792                                 goto freeit;
793                         /* -1 == no app on SEND socket */
794                         nd6_rs_input(n, off, icmp6len);
795                 } else
796                         nd6_rs_input(n, off, icmp6len);
797                 /* m stays. */
798                 break;
799
800         case ND_ROUTER_ADVERT:
801                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
802                 if (code != 0)
803                         goto badcode;
804                 if (icmp6len < sizeof(struct nd_router_advert))
805                         goto badlen;
806                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
807
808                         /* Send incoming SeND-protected/ND packet to user space. */
809                         if (send_sendso_input_hook != NULL) {
810                                 error = send_sendso_input_hook(m, ifp,
811                                     SND_IN, ip6len);
812                                 if (error == 0)
813                                         return (IPPROTO_DONE);
814                                 nd6_ra_input(m, off, icmp6len);
815                         } else
816                                 nd6_ra_input(m, off, icmp6len);
817                         m = NULL;
818                         goto freeit;
819                 }
820                 if (send_sendso_input_hook != NULL) {
821                         error = send_sendso_input_hook(n, ifp,
822                             SND_IN, ip6len);
823                         if (error == 0)
824                                 goto freeit;
825                         nd6_ra_input(n, off, icmp6len);
826                 } else
827                         nd6_ra_input(n, off, icmp6len);
828                 /* m stays. */
829                 break;
830
831         case ND_NEIGHBOR_SOLICIT:
832                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
833                 if (code != 0)
834                         goto badcode;
835                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
836                         goto badlen;
837                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
838                         if (send_sendso_input_hook != NULL) {
839                                 error = send_sendso_input_hook(m, ifp,
840                                     SND_IN, ip6len);
841                                 if (error == 0)
842                                         return (IPPROTO_DONE);
843                                 nd6_ns_input(m, off, icmp6len);
844                         } else
845                                 nd6_ns_input(m, off, icmp6len);
846                         m = NULL;
847                         goto freeit;
848                 }
849                 if (send_sendso_input_hook != NULL) {
850                         error = send_sendso_input_hook(n, ifp,
851                             SND_IN, ip6len);
852                         if (error == 0)
853                                 goto freeit;
854                         nd6_ns_input(n, off, icmp6len);
855                 } else
856                         nd6_ns_input(n, off, icmp6len);
857                 /* m stays. */
858                 break;
859
860         case ND_NEIGHBOR_ADVERT:
861                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
862                 if (code != 0)
863                         goto badcode;
864                 if (icmp6len < sizeof(struct nd_neighbor_advert))
865                         goto badlen;
866                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
867
868                         /* Send incoming SeND-protected/ND packet to user space. */
869                         if (send_sendso_input_hook != NULL) {
870                                 error = send_sendso_input_hook(m, ifp,
871                                     SND_IN, ip6len);
872                                 if (error == 0)
873                                         return (IPPROTO_DONE);
874                                 nd6_na_input(m, off, icmp6len);
875                         } else
876                                 nd6_na_input(m, off, icmp6len);
877                         m = NULL;
878                         goto freeit;
879                 }
880                 if (send_sendso_input_hook != NULL) {
881                         error = send_sendso_input_hook(n, ifp,
882                             SND_IN, ip6len);
883                         if (error == 0)
884                                 goto freeit;
885                         nd6_na_input(n, off, icmp6len);
886                 } else
887                         nd6_na_input(n, off, icmp6len);
888                 /* m stays. */
889                 break;
890
891         case ND_REDIRECT:
892                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
893                 if (code != 0)
894                         goto badcode;
895                 if (icmp6len < sizeof(struct nd_redirect))
896                         goto badlen;
897                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
898                         if (send_sendso_input_hook != NULL) {
899                                 error = send_sendso_input_hook(m, ifp,
900                                     SND_IN, ip6len);
901                                 if (error == 0)
902                                         return (IPPROTO_DONE);
903                             icmp6_redirect_input(m, off);
904                         } else
905                                 icmp6_redirect_input(m, off);
906                         m = NULL;
907                         goto freeit;
908                 }
909                 if (send_sendso_input_hook != NULL) {
910                         error = send_sendso_input_hook(n, ifp,
911                             SND_IN, ip6len);
912                         if (error == 0)
913                                 goto freeit;
914                         icmp6_redirect_input(n, off);
915                 } else
916                         icmp6_redirect_input(n, off);
917                 /* m stays. */
918                 break;
919
920         case ICMP6_ROUTER_RENUMBERING:
921                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
922                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
923                         goto badcode;
924                 if (icmp6len < sizeof(struct icmp6_router_renum))
925                         goto badlen;
926                 break;
927
928         default:
929                 nd6log((LOG_DEBUG,
930                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
931                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
932                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
933                     ifp ? ifp->if_index : 0));
934                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
935                         /* ICMPv6 error: MUST deliver it by spec... */
936                         code = PRC_NCMDS;
937                         /* deliver */
938                 } else {
939                         /* ICMPv6 informational: MUST not deliver */
940                         break;
941                 }
942         deliver:
943                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
944                         /* In this case, m should've been freed. */
945                         return (IPPROTO_DONE);
946                 }
947                 break;
948
949         badcode:
950                 ICMP6STAT_INC(icp6s_badcode);
951                 break;
952
953         badlen:
954                 ICMP6STAT_INC(icp6s_badlen);
955                 break;
956         }
957
958         /* deliver the packet to appropriate sockets */
959         icmp6_rip6_input(&m, *offp);
960
961         return IPPROTO_DONE;
962
963  freeit:
964         m_freem(m);
965         return IPPROTO_DONE;
966 }
967
968 static int
969 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
970 {
971         struct mbuf *m = *mp;
972         struct icmp6_hdr *icmp6;
973         struct ip6_hdr *eip6;
974         u_int32_t notifymtu;
975         struct sockaddr_in6 icmp6src, icmp6dst;
976
977         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
978                 ICMP6STAT_INC(icp6s_tooshort);
979                 goto freeit;
980         }
981 #ifndef PULLDOWN_TEST
982         IP6_EXTHDR_CHECK(m, off,
983             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
984         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
985 #else
986         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
987             sizeof(*icmp6) + sizeof(struct ip6_hdr));
988         if (icmp6 == NULL) {
989                 ICMP6STAT_INC(icp6s_tooshort);
990                 return (-1);
991         }
992 #endif
993         eip6 = (struct ip6_hdr *)(icmp6 + 1);
994
995         /* Detect the upper level protocol */
996         {
997                 void (*ctlfunc)(int, struct sockaddr *, void *);
998                 u_int8_t nxt = eip6->ip6_nxt;
999                 int eoff = off + sizeof(struct icmp6_hdr) +
1000                     sizeof(struct ip6_hdr);
1001                 struct ip6ctlparam ip6cp;
1002                 struct in6_addr *finaldst = NULL;
1003                 int icmp6type = icmp6->icmp6_type;
1004                 struct ip6_frag *fh;
1005                 struct ip6_rthdr *rth;
1006                 struct ip6_rthdr0 *rth0;
1007                 int rthlen;
1008
1009                 while (1) { /* XXX: should avoid infinite loop explicitly? */
1010                         struct ip6_ext *eh;
1011
1012                         switch (nxt) {
1013                         case IPPROTO_HOPOPTS:
1014                         case IPPROTO_DSTOPTS:
1015                         case IPPROTO_AH:
1016 #ifndef PULLDOWN_TEST
1017                                 IP6_EXTHDR_CHECK(m, 0,
1018                                     eoff + sizeof(struct ip6_ext), -1);
1019                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1020 #else
1021                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1022                                     eoff, sizeof(*eh));
1023                                 if (eh == NULL) {
1024                                         ICMP6STAT_INC(icp6s_tooshort);
1025                                         return (-1);
1026                                 }
1027 #endif
1028
1029                                 if (nxt == IPPROTO_AH)
1030                                         eoff += (eh->ip6e_len + 2) << 2;
1031                                 else
1032                                         eoff += (eh->ip6e_len + 1) << 3;
1033                                 nxt = eh->ip6e_nxt;
1034                                 break;
1035                         case IPPROTO_ROUTING:
1036                                 /*
1037                                  * When the erroneous packet contains a
1038                                  * routing header, we should examine the
1039                                  * header to determine the final destination.
1040                                  * Otherwise, we can't properly update
1041                                  * information that depends on the final
1042                                  * destination (e.g. path MTU).
1043                                  */
1044 #ifndef PULLDOWN_TEST
1045                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1046                                 rth = (struct ip6_rthdr *)
1047                                     (mtod(m, caddr_t) + eoff);
1048 #else
1049                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1050                                     eoff, sizeof(*rth));
1051                                 if (rth == NULL) {
1052                                         ICMP6STAT_INC(icp6s_tooshort);
1053                                         return (-1);
1054                                 }
1055 #endif
1056                                 rthlen = (rth->ip6r_len + 1) << 3;
1057                                 /*
1058                                  * XXX: currently there is no
1059                                  * officially defined type other
1060                                  * than type-0.
1061                                  * Note that if the segment left field
1062                                  * is 0, all intermediate hops must
1063                                  * have been passed.
1064                                  */
1065                                 if (rth->ip6r_segleft &&
1066                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1067                                         int hops;
1068
1069 #ifndef PULLDOWN_TEST
1070                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1071                                         rth0 = (struct ip6_rthdr0 *)
1072                                             (mtod(m, caddr_t) + eoff);
1073 #else
1074                                         IP6_EXTHDR_GET(rth0,
1075                                             struct ip6_rthdr0 *, m,
1076                                             eoff, rthlen);
1077                                         if (rth0 == NULL) {
1078                                                 ICMP6STAT_INC(icp6s_tooshort);
1079                                                 return (-1);
1080                                         }
1081 #endif
1082                                         /* just ignore a bogus header */
1083                                         if ((rth0->ip6r0_len % 2) == 0 &&
1084                                             (hops = rth0->ip6r0_len/2))
1085                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1086                                 }
1087                                 eoff += rthlen;
1088                                 nxt = rth->ip6r_nxt;
1089                                 break;
1090                         case IPPROTO_FRAGMENT:
1091 #ifndef PULLDOWN_TEST
1092                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1093                                     sizeof(struct ip6_frag), -1);
1094                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1095                                     eoff);
1096 #else
1097                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1098                                     eoff, sizeof(*fh));
1099                                 if (fh == NULL) {
1100                                         ICMP6STAT_INC(icp6s_tooshort);
1101                                         return (-1);
1102                                 }
1103 #endif
1104                                 /*
1105                                  * Data after a fragment header is meaningless
1106                                  * unless it is the first fragment, but
1107                                  * we'll go to the notify label for path MTU
1108                                  * discovery.
1109                                  */
1110                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1111                                         goto notify;
1112
1113                                 eoff += sizeof(struct ip6_frag);
1114                                 nxt = fh->ip6f_nxt;
1115                                 break;
1116                         default:
1117                                 /*
1118                                  * This case includes ESP and the No Next
1119                                  * Header.  In such cases going to the notify
1120                                  * label does not have any meaning
1121                                  * (i.e. ctlfunc will be NULL), but we go
1122                                  * anyway since we might have to update
1123                                  * path MTU information.
1124                                  */
1125                                 goto notify;
1126                         }
1127                 }
1128           notify:
1129 #ifndef PULLDOWN_TEST
1130                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1131 #else
1132                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1133                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1134                 if (icmp6 == NULL) {
1135                         ICMP6STAT_INC(icp6s_tooshort);
1136                         return (-1);
1137                 }
1138 #endif
1139
1140                 /*
1141                  * retrieve parameters from the inner IPv6 header, and convert
1142                  * them into sockaddr structures.
1143                  * XXX: there is no guarantee that the source or destination
1144                  * addresses of the inner packet are in the same scope as
1145                  * the addresses of the icmp packet.  But there is no other
1146                  * way to determine the zone.
1147                  */
1148                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1149
1150                 bzero(&icmp6dst, sizeof(icmp6dst));
1151                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1152                 icmp6dst.sin6_family = AF_INET6;
1153                 if (finaldst == NULL)
1154                         icmp6dst.sin6_addr = eip6->ip6_dst;
1155                 else
1156                         icmp6dst.sin6_addr = *finaldst;
1157                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1158                         goto freeit;
1159                 bzero(&icmp6src, sizeof(icmp6src));
1160                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1161                 icmp6src.sin6_family = AF_INET6;
1162                 icmp6src.sin6_addr = eip6->ip6_src;
1163                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1164                         goto freeit;
1165                 icmp6src.sin6_flowinfo =
1166                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1167
1168                 if (finaldst == NULL)
1169                         finaldst = &eip6->ip6_dst;
1170                 ip6cp.ip6c_m = m;
1171                 ip6cp.ip6c_icmp6 = icmp6;
1172                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1173                 ip6cp.ip6c_off = eoff;
1174                 ip6cp.ip6c_finaldst = finaldst;
1175                 ip6cp.ip6c_src = &icmp6src;
1176                 ip6cp.ip6c_nxt = nxt;
1177
1178                 m_addr_changed(m);
1179
1180                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1181                         notifymtu = ntohl(icmp6->icmp6_mtu);
1182                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1183                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1184                 }
1185
1186                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1187                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1188                 if (ctlfunc) {
1189                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1190                             &ip6cp);
1191                 }
1192         }
1193         *mp = m;
1194         return (0);
1195
1196   freeit:
1197         m_freem(m);
1198         return (-1);
1199 }
1200
1201 void
1202 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1203 {
1204         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1205         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1206         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1207         u_int mtu = ntohl(icmp6->icmp6_mtu);
1208         struct in_conninfo inc;
1209
1210 #if 0
1211         /*
1212          * RFC2460 section 5, last paragraph.
1213          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1214          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1215          * due to packet translator in the middle.
1216          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1217          * special handling.
1218          */
1219         if (mtu < IPV6_MMTU)
1220                 return;
1221 #endif
1222
1223         /*
1224          * we reject ICMPv6 too big with abnormally small value.
1225          * XXX what is the good definition of "abnormally small"?
1226          */
1227         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1228                 return;
1229
1230         if (!validated)
1231                 return;
1232
1233         /*
1234          * In case the suggested mtu is less than IPV6_MMTU, we
1235          * only need to remember that it was for above mentioned
1236          * "alwaysfrag" case.
1237          * Try to be as close to the spec as possible.
1238          */
1239         if (mtu < IPV6_MMTU)
1240                 mtu = IPV6_MMTU - 8;
1241
1242         bzero(&inc, sizeof(inc));
1243         inc.inc_fibnum = M_GETFIB(m);
1244         inc.inc_flags |= INC_ISIPV6;
1245         inc.inc6_faddr = *dst;
1246         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1247                 return;
1248
1249         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1250                 tcp_hc_updatemtu(&inc, mtu);
1251                 ICMP6STAT_INC(icp6s_pmtuchg);
1252         }
1253 }
1254
1255 /*
1256  * Process a Node Information Query packet, based on
1257  * draft-ietf-ipngwg-icmp-name-lookups-07.
1258  *
1259  * Spec incompatibilities:
1260  * - IPv6 Subject address handling
1261  * - IPv4 Subject address handling support missing
1262  * - Proxy reply (answer even if it's not for me)
1263  * - joins NI group address at in6_ifattach() time only, does not cope
1264  *   with hostname changes by sethostname(3)
1265  */
1266 static struct mbuf *
1267 ni6_input(struct mbuf *m, int off)
1268 {
1269         struct icmp6_nodeinfo *ni6, *nni6;
1270         struct mbuf *n = NULL;
1271         struct prison *pr;
1272         u_int16_t qtype;
1273         int subjlen;
1274         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1275         struct ni_reply_fqdn *fqdn;
1276         int addrs;              /* for NI_QTYPE_NODEADDR */
1277         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1278         struct in6_addr in6_subj; /* subject address */
1279         struct ip6_hdr *ip6;
1280         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1281         char *subj = NULL;
1282         struct in6_ifaddr *ia6 = NULL;
1283
1284         ip6 = mtod(m, struct ip6_hdr *);
1285 #ifndef PULLDOWN_TEST
1286         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1287 #else
1288         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1289         if (ni6 == NULL) {
1290                 /* m is already reclaimed */
1291                 return (NULL);
1292         }
1293 #endif
1294
1295         /*
1296          * Validate IPv6 source address.
1297          * The default configuration MUST be to refuse answering queries from
1298          * global-scope addresses according to RFC4602.
1299          * Notes:
1300          *  - it's not very clear what "refuse" means; this implementation
1301          *    simply drops it.
1302          *  - it's not very easy to identify global-scope (unicast) addresses
1303          *    since there are many prefixes for them.  It should be safer
1304          *    and in practice sufficient to check "all" but loopback and
1305          *    link-local (note that site-local unicast was deprecated and
1306          *    ULA is defined as global scope-wise)
1307          */
1308         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1309             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1310             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1311                 goto bad;
1312
1313         /*
1314          * Validate IPv6 destination address.
1315          *
1316          * The Responder must discard the Query without further processing
1317          * unless it is one of the Responder's unicast or anycast addresses, or
1318          * a link-local scope multicast address which the Responder has joined.
1319          * [RFC4602, Section 5.]
1320          */
1321         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1322                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1323                         goto bad;
1324                 /* else it's a link-local multicast, fine */
1325         } else {                /* unicast or anycast */
1326                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1327                         goto bad; /* XXX impossible */
1328
1329                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1330                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1331                         ifa_free(&ia6->ia_ifa);
1332                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1333                                 "a temporary address in %s:%d",
1334                                __FILE__, __LINE__));
1335                         goto bad;
1336                 }
1337                 ifa_free(&ia6->ia_ifa);
1338         }
1339
1340         /* validate query Subject field. */
1341         qtype = ntohs(ni6->ni_qtype);
1342         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1343         switch (qtype) {
1344         case NI_QTYPE_NOOP:
1345         case NI_QTYPE_SUPTYPES:
1346                 /* 07 draft */
1347                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1348                         break;
1349                 /* FALLTHROUGH */
1350         case NI_QTYPE_FQDN:
1351         case NI_QTYPE_NODEADDR:
1352         case NI_QTYPE_IPV4ADDR:
1353                 switch (ni6->ni_code) {
1354                 case ICMP6_NI_SUBJ_IPV6:
1355 #if ICMP6_NI_SUBJ_IPV6 != 0
1356                 case 0:
1357 #endif
1358                         /*
1359                          * backward compatibility - try to accept 03 draft
1360                          * format, where no Subject is present.
1361                          */
1362                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1363                             subjlen == 0) {
1364                                 oldfqdn++;
1365                                 break;
1366                         }
1367 #if ICMP6_NI_SUBJ_IPV6 != 0
1368                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1369                                 goto bad;
1370 #endif
1371
1372                         if (subjlen != sizeof(struct in6_addr))
1373                                 goto bad;
1374
1375                         /*
1376                          * Validate Subject address.
1377                          *
1378                          * Not sure what exactly "address belongs to the node"
1379                          * means in the spec, is it just unicast, or what?
1380                          *
1381                          * At this moment we consider Subject address as
1382                          * "belong to the node" if the Subject address equals
1383                          * to the IPv6 destination address; validation for
1384                          * IPv6 destination address should have done enough
1385                          * check for us.
1386                          *
1387                          * We do not do proxy at this moment.
1388                          */
1389                         /* m_pulldown instead of copy? */
1390                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1391                             subjlen, (caddr_t)&in6_subj);
1392                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1393                                 goto bad;
1394
1395                         subj = (char *)&in6_subj;
1396                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1397                                 break;
1398
1399                         /*
1400                          * XXX if we are to allow other cases, we should really
1401                          * be careful about scope here.
1402                          * basically, we should disallow queries toward IPv6
1403                          * destination X with subject Y,
1404                          * if scope(X) > scope(Y).
1405                          * if we allow scope(X) > scope(Y), it will result in
1406                          * information leakage across scope boundary.
1407                          */
1408                         goto bad;
1409
1410                 case ICMP6_NI_SUBJ_FQDN:
1411                         /*
1412                          * Validate Subject name with gethostname(3).
1413                          *
1414                          * The behavior may need some debate, since:
1415                          * - we are not sure if the node has FQDN as
1416                          *   hostname (returned by gethostname(3)).
1417                          * - the code does wildcard match for truncated names.
1418                          *   however, we are not sure if we want to perform
1419                          *   wildcard match, if gethostname(3) side has
1420                          *   truncated hostname.
1421                          */
1422                         pr = curthread->td_ucred->cr_prison;
1423                         mtx_lock(&pr->pr_mtx);
1424                         n = ni6_nametodns(pr->pr_hostname,
1425                             strlen(pr->pr_hostname), 0);
1426                         mtx_unlock(&pr->pr_mtx);
1427                         if (!n || n->m_next || n->m_len == 0)
1428                                 goto bad;
1429                         IP6_EXTHDR_GET(subj, char *, m,
1430                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1431                         if (subj == NULL)
1432                                 goto bad;
1433                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1434                             n->m_len)) {
1435                                 goto bad;
1436                         }
1437                         m_freem(n);
1438                         n = NULL;
1439                         break;
1440
1441                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1442                 default:
1443                         goto bad;
1444                 }
1445                 break;
1446         }
1447
1448         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1449         switch (qtype) {
1450         case NI_QTYPE_FQDN:
1451                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1452                         goto bad;
1453                 break;
1454         case NI_QTYPE_NODEADDR:
1455         case NI_QTYPE_IPV4ADDR:
1456                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1457                         goto bad;
1458                 break;
1459         }
1460
1461         /* guess reply length */
1462         switch (qtype) {
1463         case NI_QTYPE_NOOP:
1464                 break;          /* no reply data */
1465         case NI_QTYPE_SUPTYPES:
1466                 replylen += sizeof(u_int32_t);
1467                 break;
1468         case NI_QTYPE_FQDN:
1469                 /* XXX will append an mbuf */
1470                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1471                 break;
1472         case NI_QTYPE_NODEADDR:
1473                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1474                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1475                     sizeof(u_int32_t))) > MCLBYTES)
1476                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1477                 break;
1478         case NI_QTYPE_IPV4ADDR:
1479                 /* unsupported - should respond with unknown Qtype? */
1480                 break;
1481         default:
1482                 /*
1483                  * XXX: We must return a reply with the ICMP6 code
1484                  * `unknown Qtype' in this case.  However we regard the case
1485                  * as an FQDN query for backward compatibility.
1486                  * Older versions set a random value to this field,
1487                  * so it rarely varies in the defined qtypes.
1488                  * But the mechanism is not reliable...
1489                  * maybe we should obsolete older versions.
1490                  */
1491                 qtype = NI_QTYPE_FQDN;
1492                 /* XXX will append an mbuf */
1493                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1494                 oldfqdn++;
1495                 break;
1496         }
1497
1498         /* allocate an mbuf to reply. */
1499         MGETHDR(n, M_DONTWAIT, m->m_type);
1500         if (n == NULL) {
1501                 m_freem(m);
1502                 return (NULL);
1503         }
1504         M_MOVE_PKTHDR(n, m); /* just for recvif and FIB */
1505         if (replylen > MHLEN) {
1506                 if (replylen > MCLBYTES) {
1507                         /*
1508                          * XXX: should we try to allocate more? But MCLBYTES
1509                          * is probably much larger than IPV6_MMTU...
1510                          */
1511                         goto bad;
1512                 }
1513                 MCLGET(n, M_DONTWAIT);
1514                 if ((n->m_flags & M_EXT) == 0) {
1515                         goto bad;
1516                 }
1517         }
1518         n->m_pkthdr.len = n->m_len = replylen;
1519
1520         /* copy mbuf header and IPv6 + Node Information base headers */
1521         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1522         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1523         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1524
1525         /* qtype dependent procedure */
1526         switch (qtype) {
1527         case NI_QTYPE_NOOP:
1528                 nni6->ni_code = ICMP6_NI_SUCCESS;
1529                 nni6->ni_flags = 0;
1530                 break;
1531         case NI_QTYPE_SUPTYPES:
1532         {
1533                 u_int32_t v;
1534                 nni6->ni_code = ICMP6_NI_SUCCESS;
1535                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1536                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1537                 v = (u_int32_t)htonl(0x0000000f);
1538                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1539                 break;
1540         }
1541         case NI_QTYPE_FQDN:
1542                 nni6->ni_code = ICMP6_NI_SUCCESS;
1543                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1544                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1545                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1546                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1547                 /*
1548                  * XXX do we really have FQDN in hostname?
1549                  */
1550                 pr = curthread->td_ucred->cr_prison;
1551                 mtx_lock(&pr->pr_mtx);
1552                 n->m_next = ni6_nametodns(pr->pr_hostname,
1553                     strlen(pr->pr_hostname), oldfqdn);
1554                 mtx_unlock(&pr->pr_mtx);
1555                 if (n->m_next == NULL)
1556                         goto bad;
1557                 /* XXX we assume that n->m_next is not a chain */
1558                 if (n->m_next->m_next != NULL)
1559                         goto bad;
1560                 n->m_pkthdr.len += n->m_next->m_len;
1561                 break;
1562         case NI_QTYPE_NODEADDR:
1563         {
1564                 int lenlim, copied;
1565
1566                 nni6->ni_code = ICMP6_NI_SUCCESS;
1567                 n->m_pkthdr.len = n->m_len =
1568                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1569                 lenlim = M_TRAILINGSPACE(n);
1570                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1571                 /* XXX: reset mbuf length */
1572                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1573                     sizeof(struct icmp6_nodeinfo) + copied;
1574                 break;
1575         }
1576         default:
1577                 break;          /* XXX impossible! */
1578         }
1579
1580         nni6->ni_type = ICMP6_NI_REPLY;
1581         m_freem(m);
1582         return (n);
1583
1584   bad:
1585         m_freem(m);
1586         if (n)
1587                 m_freem(n);
1588         return (NULL);
1589 }
1590
1591 /*
1592  * make a mbuf with DNS-encoded string.  no compression support.
1593  *
1594  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1595  * treated as truncated name (two \0 at the end).  this is a wild guess.
1596  *
1597  * old - return pascal string if non-zero
1598  */
1599 static struct mbuf *
1600 ni6_nametodns(const char *name, int namelen, int old)
1601 {
1602         struct mbuf *m;
1603         char *cp, *ep;
1604         const char *p, *q;
1605         int i, len, nterm;
1606
1607         if (old)
1608                 len = namelen + 1;
1609         else
1610                 len = MCLBYTES;
1611
1612         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1613         MGET(m, M_DONTWAIT, MT_DATA);
1614         if (m && len > MLEN) {
1615                 MCLGET(m, M_DONTWAIT);
1616                 if ((m->m_flags & M_EXT) == 0)
1617                         goto fail;
1618         }
1619         if (!m)
1620                 goto fail;
1621         m->m_next = NULL;
1622
1623         if (old) {
1624                 m->m_len = len;
1625                 *mtod(m, char *) = namelen;
1626                 bcopy(name, mtod(m, char *) + 1, namelen);
1627                 return m;
1628         } else {
1629                 m->m_len = 0;
1630                 cp = mtod(m, char *);
1631                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1632
1633                 /* if not certain about my name, return empty buffer */
1634                 if (namelen == 0)
1635                         return m;
1636
1637                 /*
1638                  * guess if it looks like shortened hostname, or FQDN.
1639                  * shortened hostname needs two trailing "\0".
1640                  */
1641                 i = 0;
1642                 for (p = name; p < name + namelen; p++) {
1643                         if (*p && *p == '.')
1644                                 i++;
1645                 }
1646                 if (i < 2)
1647                         nterm = 2;
1648                 else
1649                         nterm = 1;
1650
1651                 p = name;
1652                 while (cp < ep && p < name + namelen) {
1653                         i = 0;
1654                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1655                                 i++;
1656                         /* result does not fit into mbuf */
1657                         if (cp + i + 1 >= ep)
1658                                 goto fail;
1659                         /*
1660                          * DNS label length restriction, RFC1035 page 8.
1661                          * "i == 0" case is included here to avoid returning
1662                          * 0-length label on "foo..bar".
1663                          */
1664                         if (i <= 0 || i >= 64)
1665                                 goto fail;
1666                         *cp++ = i;
1667                         bcopy(p, cp, i);
1668                         cp += i;
1669                         p = q;
1670                         if (p < name + namelen && *p == '.')
1671                                 p++;
1672                 }
1673                 /* termination */
1674                 if (cp + nterm >= ep)
1675                         goto fail;
1676                 while (nterm-- > 0)
1677                         *cp++ = '\0';
1678                 m->m_len = cp - mtod(m, char *);
1679                 return m;
1680         }
1681
1682         panic("should not reach here");
1683         /* NOTREACHED */
1684
1685  fail:
1686         if (m)
1687                 m_freem(m);
1688         return NULL;
1689 }
1690
1691 /*
1692  * check if two DNS-encoded string matches.  takes care of truncated
1693  * form (with \0\0 at the end).  no compression support.
1694  * XXX upper/lowercase match (see RFC2065)
1695  */
1696 static int
1697 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1698 {
1699         const char *a0, *b0;
1700         int l;
1701
1702         /* simplest case - need validation? */
1703         if (alen == blen && bcmp(a, b, alen) == 0)
1704                 return 1;
1705
1706         a0 = a;
1707         b0 = b;
1708
1709         /* termination is mandatory */
1710         if (alen < 2 || blen < 2)
1711                 return 0;
1712         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1713                 return 0;
1714         alen--;
1715         blen--;
1716
1717         while (a - a0 < alen && b - b0 < blen) {
1718                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1719                         return 0;
1720
1721                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1722                         return 0;
1723                 /* we don't support compression yet */
1724                 if (a[0] >= 64 || b[0] >= 64)
1725                         return 0;
1726
1727                 /* truncated case */
1728                 if (a[0] == 0 && a - a0 == alen - 1)
1729                         return 1;
1730                 if (b[0] == 0 && b - b0 == blen - 1)
1731                         return 1;
1732                 if (a[0] == 0 || b[0] == 0)
1733                         return 0;
1734
1735                 if (a[0] != b[0])
1736                         return 0;
1737                 l = a[0];
1738                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1739                         return 0;
1740                 if (bcmp(a + 1, b + 1, l) != 0)
1741                         return 0;
1742
1743                 a += 1 + l;
1744                 b += 1 + l;
1745         }
1746
1747         if (a - a0 == alen && b - b0 == blen)
1748                 return 1;
1749         else
1750                 return 0;
1751 }
1752
1753 /*
1754  * calculate the number of addresses to be returned in the node info reply.
1755  */
1756 static int
1757 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1758     struct in6_addr *subj)
1759 {
1760         struct ifnet *ifp;
1761         struct in6_ifaddr *ifa6;
1762         struct ifaddr *ifa;
1763         int addrs = 0, addrsofif, iffound = 0;
1764         int niflags = ni6->ni_flags;
1765
1766         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1767                 switch (ni6->ni_code) {
1768                 case ICMP6_NI_SUBJ_IPV6:
1769                         if (subj == NULL) /* must be impossible... */
1770                                 return (0);
1771                         break;
1772                 default:
1773                         /*
1774                          * XXX: we only support IPv6 subject address for
1775                          * this Qtype.
1776                          */
1777                         return (0);
1778                 }
1779         }
1780
1781         IFNET_RLOCK_NOSLEEP();
1782         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1783                 addrsofif = 0;
1784                 IF_ADDR_RLOCK(ifp);
1785                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1786                         if (ifa->ifa_addr->sa_family != AF_INET6)
1787                                 continue;
1788                         ifa6 = (struct in6_ifaddr *)ifa;
1789
1790                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1791                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1792                                 iffound = 1;
1793
1794                         /*
1795                          * IPv4-mapped addresses can only be returned by a
1796                          * Node Information proxy, since they represent
1797                          * addresses of IPv4-only nodes, which perforce do
1798                          * not implement this protocol.
1799                          * [icmp-name-lookups-07, Section 5.4]
1800                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1801                          * this function at this moment.
1802                          */
1803
1804                         /* What do we have to do about ::1? */
1805                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1806                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1807                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1808                                         continue;
1809                                 break;
1810                         case IPV6_ADDR_SCOPE_SITELOCAL:
1811                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1812                                         continue;
1813                                 break;
1814                         case IPV6_ADDR_SCOPE_GLOBAL:
1815                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1816                                         continue;
1817                                 break;
1818                         default:
1819                                 continue;
1820                         }
1821
1822                         /*
1823                          * check if anycast is okay.
1824                          * XXX: just experimental.  not in the spec.
1825                          */
1826                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1827                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1828                                 continue; /* we need only unicast addresses */
1829                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1830                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1831                                 continue;
1832                         }
1833                         addrsofif++; /* count the address */
1834                 }
1835                 IF_ADDR_RUNLOCK(ifp);
1836                 if (iffound) {
1837                         *ifpp = ifp;
1838                         IFNET_RUNLOCK_NOSLEEP();
1839                         return (addrsofif);
1840                 }
1841
1842                 addrs += addrsofif;
1843         }
1844         IFNET_RUNLOCK_NOSLEEP();
1845
1846         return (addrs);
1847 }
1848
1849 static int
1850 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1851     struct ifnet *ifp0, int resid)
1852 {
1853         struct ifnet *ifp;
1854         struct in6_ifaddr *ifa6;
1855         struct ifaddr *ifa;
1856         struct ifnet *ifp_dep = NULL;
1857         int copied = 0, allow_deprecated = 0;
1858         u_char *cp = (u_char *)(nni6 + 1);
1859         int niflags = ni6->ni_flags;
1860         u_int32_t ltime;
1861
1862         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1863                 return (0);     /* needless to copy */
1864
1865         IFNET_RLOCK_NOSLEEP();
1866         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1867   again:
1868
1869         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1870                 IF_ADDR_RLOCK(ifp);
1871                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1872                         if (ifa->ifa_addr->sa_family != AF_INET6)
1873                                 continue;
1874                         ifa6 = (struct in6_ifaddr *)ifa;
1875
1876                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1877                             allow_deprecated == 0) {
1878                                 /*
1879                                  * prefererred address should be put before
1880                                  * deprecated addresses.
1881                                  */
1882
1883                                 /* record the interface for later search */
1884                                 if (ifp_dep == NULL)
1885                                         ifp_dep = ifp;
1886
1887                                 continue;
1888                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1889                             allow_deprecated != 0)
1890                                 continue; /* we now collect deprecated addrs */
1891
1892                         /* What do we have to do about ::1? */
1893                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1894                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1895                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1896                                         continue;
1897                                 break;
1898                         case IPV6_ADDR_SCOPE_SITELOCAL:
1899                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1900                                         continue;
1901                                 break;
1902                         case IPV6_ADDR_SCOPE_GLOBAL:
1903                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1904                                         continue;
1905                                 break;
1906                         default:
1907                                 continue;
1908                         }
1909
1910                         /*
1911                          * check if anycast is okay.
1912                          * XXX: just experimental.  not in the spec.
1913                          */
1914                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1915                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1916                                 continue;
1917                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1918                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1919                                 continue;
1920                         }
1921
1922                         /* now we can copy the address */
1923                         if (resid < sizeof(struct in6_addr) +
1924                             sizeof(u_int32_t)) {
1925                                 IF_ADDR_RUNLOCK(ifp);
1926                                 /*
1927                                  * We give up much more copy.
1928                                  * Set the truncate flag and return.
1929                                  */
1930                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1931                                 IFNET_RUNLOCK_NOSLEEP();
1932                                 return (copied);
1933                         }
1934
1935                         /*
1936                          * Set the TTL of the address.
1937                          * The TTL value should be one of the following
1938                          * according to the specification:
1939                          *
1940                          * 1. The remaining lifetime of a DHCP lease on the
1941                          *    address, or
1942                          * 2. The remaining Valid Lifetime of a prefix from
1943                          *    which the address was derived through Stateless
1944                          *    Autoconfiguration.
1945                          *
1946                          * Note that we currently do not support stateful
1947                          * address configuration by DHCPv6, so the former
1948                          * case can't happen.
1949                          */
1950                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1951                                 ltime = ND6_INFINITE_LIFETIME;
1952                         else {
1953                                 if (ifa6->ia6_lifetime.ia6t_expire >
1954                                     time_second)
1955                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1956                                 else
1957                                         ltime = 0;
1958                         }
1959
1960                         bcopy(&ltime, cp, sizeof(u_int32_t));
1961                         cp += sizeof(u_int32_t);
1962
1963                         /* copy the address itself */
1964                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1965                             sizeof(struct in6_addr));
1966                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1967                         cp += sizeof(struct in6_addr);
1968
1969                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1970                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1971                 }
1972                 IF_ADDR_RUNLOCK(ifp);
1973                 if (ifp0)       /* we need search only on the specified IF */
1974                         break;
1975         }
1976
1977         if (allow_deprecated == 0 && ifp_dep != NULL) {
1978                 ifp = ifp_dep;
1979                 allow_deprecated = 1;
1980
1981                 goto again;
1982         }
1983
1984         IFNET_RUNLOCK_NOSLEEP();
1985
1986         return (copied);
1987 }
1988
1989 /*
1990  * XXX almost dup'ed code with rip6_input.
1991  */
1992 static int
1993 icmp6_rip6_input(struct mbuf **mp, int off)
1994 {
1995         struct mbuf *m = *mp;
1996         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1997         struct inpcb *in6p;
1998         struct inpcb *last = NULL;
1999         struct sockaddr_in6 fromsa;
2000         struct icmp6_hdr *icmp6;
2001         struct mbuf *opts = NULL;
2002
2003 #ifndef PULLDOWN_TEST
2004         /* this is assumed to be safe. */
2005         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
2006 #else
2007         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
2008         if (icmp6 == NULL) {
2009                 /* m is already reclaimed */
2010                 return (IPPROTO_DONE);
2011         }
2012 #endif
2013
2014         /*
2015          * XXX: the address may have embedded scope zone ID, which should be
2016          * hidden from applications.
2017          */
2018         bzero(&fromsa, sizeof(fromsa));
2019         fromsa.sin6_family = AF_INET6;
2020         fromsa.sin6_len = sizeof(struct sockaddr_in6);
2021         fromsa.sin6_addr = ip6->ip6_src;
2022         if (sa6_recoverscope(&fromsa)) {
2023                 m_freem(m);
2024                 return (IPPROTO_DONE);
2025         }
2026
2027         INP_INFO_RLOCK(&V_ripcbinfo);
2028         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2029                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2030                         continue;
2031                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2032                         continue;
2033                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2034                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2035                         continue;
2036                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2037                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2038                         continue;
2039                 INP_RLOCK(in6p);
2040                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2041                     in6p->in6p_icmp6filt)) {
2042                         INP_RUNLOCK(in6p);
2043                         continue;
2044                 }
2045                 if (last != NULL) {
2046                         struct  mbuf *n = NULL;
2047
2048                         /*
2049                          * Recent network drivers tend to allocate a single
2050                          * mbuf cluster, rather than to make a couple of
2051                          * mbufs without clusters.  Also, since the IPv6 code
2052                          * path tries to avoid m_pullup(), it is highly
2053                          * probable that we still have an mbuf cluster here
2054                          * even though the necessary length can be stored in an
2055                          * mbuf's internal buffer.
2056                          * Meanwhile, the default size of the receive socket
2057                          * buffer for raw sockets is not so large.  This means
2058                          * the possibility of packet loss is relatively higher
2059                          * than before.  To avoid this scenario, we copy the
2060                          * received data to a separate mbuf that does not use
2061                          * a cluster, if possible.
2062                          * XXX: it is better to copy the data after stripping
2063                          * intermediate headers.
2064                          */
2065                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2066                             m->m_len <= MHLEN) {
2067                                 MGET(n, M_DONTWAIT, m->m_type);
2068                                 if (n != NULL) {
2069                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2070                                                 bcopy(m->m_data, n->m_data,
2071                                                       m->m_len);
2072                                                 n->m_len = m->m_len;
2073                                         } else {
2074                                                 m_free(n);
2075                                                 n = NULL;
2076                                         }
2077                                 }
2078                         }
2079                         if (n != NULL ||
2080                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2081                                 if (last->inp_flags & INP_CONTROLOPTS)
2082                                         ip6_savecontrol(last, n, &opts);
2083                                 /* strip intermediate headers */
2084                                 m_adj(n, off);
2085                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2086                                 if (sbappendaddr_locked(
2087                                     &last->inp_socket->so_rcv,
2088                                     (struct sockaddr *)&fromsa, n, opts)
2089                                     == 0) {
2090                                         /* should notify about lost packet */
2091                                         m_freem(n);
2092                                         if (opts) {
2093                                                 m_freem(opts);
2094                                         }
2095                                         SOCKBUF_UNLOCK(
2096                                             &last->inp_socket->so_rcv);
2097                                 } else
2098                                         sorwakeup_locked(last->inp_socket);
2099                                 opts = NULL;
2100                         }
2101                         INP_RUNLOCK(last);
2102                 }
2103                 last = in6p;
2104         }
2105         INP_INFO_RUNLOCK(&V_ripcbinfo);
2106         if (last != NULL) {
2107                 if (last->inp_flags & INP_CONTROLOPTS)
2108                         ip6_savecontrol(last, m, &opts);
2109                 /* strip intermediate headers */
2110                 m_adj(m, off);
2111
2112                 /* avoid using mbuf clusters if possible (see above) */
2113                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2114                     m->m_len <= MHLEN) {
2115                         struct mbuf *n;
2116
2117                         MGET(n, M_DONTWAIT, m->m_type);
2118                         if (n != NULL) {
2119                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2120                                         bcopy(m->m_data, n->m_data, m->m_len);
2121                                         n->m_len = m->m_len;
2122
2123                                         m_freem(m);
2124                                         m = n;
2125                                 } else {
2126                                         m_freem(n);
2127                                         n = NULL;
2128                                 }
2129                         }
2130                 }
2131                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2132                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2133                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2134                         m_freem(m);
2135                         if (opts)
2136                                 m_freem(opts);
2137                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2138                 } else
2139                         sorwakeup_locked(last->inp_socket);
2140                 INP_RUNLOCK(last);
2141         } else {
2142                 m_freem(m);
2143                 IP6STAT_DEC(ip6s_delivered);
2144         }
2145         return IPPROTO_DONE;
2146 }
2147
2148 /*
2149  * Reflect the ip6 packet back to the source.
2150  * OFF points to the icmp6 header, counted from the top of the mbuf.
2151  */
2152 void
2153 icmp6_reflect(struct mbuf *m, size_t off)
2154 {
2155         struct ip6_hdr *ip6;
2156         struct icmp6_hdr *icmp6;
2157         struct in6_ifaddr *ia = NULL;
2158         int plen;
2159         int type, code;
2160         struct ifnet *outif = NULL;
2161         struct in6_addr origdst, src, *srcp = NULL;
2162
2163         /* too short to reflect */
2164         if (off < sizeof(struct ip6_hdr)) {
2165                 nd6log((LOG_DEBUG,
2166                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2167                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2168                     __FILE__, __LINE__));
2169                 goto bad;
2170         }
2171
2172         /*
2173          * If there are extra headers between IPv6 and ICMPv6, strip
2174          * off that header first.
2175          */
2176 #ifdef DIAGNOSTIC
2177         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2178                 panic("assumption failed in icmp6_reflect");
2179 #endif
2180         if (off > sizeof(struct ip6_hdr)) {
2181                 size_t l;
2182                 struct ip6_hdr nip6;
2183
2184                 l = off - sizeof(struct ip6_hdr);
2185                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2186                 m_adj(m, l);
2187                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2188                 if (m->m_len < l) {
2189                         if ((m = m_pullup(m, l)) == NULL)
2190                                 return;
2191                 }
2192                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2193         } else /* off == sizeof(struct ip6_hdr) */ {
2194                 size_t l;
2195                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2196                 if (m->m_len < l) {
2197                         if ((m = m_pullup(m, l)) == NULL)
2198                                 return;
2199                 }
2200         }
2201         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2202         ip6 = mtod(m, struct ip6_hdr *);
2203         ip6->ip6_nxt = IPPROTO_ICMPV6;
2204         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2205         type = icmp6->icmp6_type; /* keep type for statistics */
2206         code = icmp6->icmp6_code; /* ditto. */
2207
2208         origdst = ip6->ip6_dst;
2209         /*
2210          * ip6_input() drops a packet if its src is multicast.
2211          * So, the src is never multicast.
2212          */
2213         ip6->ip6_dst = ip6->ip6_src;
2214
2215         /*
2216          * If the incoming packet was addressed directly to us (i.e. unicast),
2217          * use dst as the src for the reply.
2218          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2219          * (for example) when we encounter an error while forwarding procedure
2220          * destined to a duplicated address of ours.
2221          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2222          * procedure of an outgoing packet of our own, in which case we need
2223          * to search in the ifaddr list.
2224          */
2225         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2226                 if ((ia = ip6_getdstifaddr(m))) {
2227                         if (!(ia->ia6_flags &
2228                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2229                                 srcp = &ia->ia_addr.sin6_addr;
2230                 } else {
2231                         struct sockaddr_in6 d;
2232
2233                         bzero(&d, sizeof(d));
2234                         d.sin6_family = AF_INET6;
2235                         d.sin6_len = sizeof(d);
2236                         d.sin6_addr = origdst;
2237                         ia = (struct in6_ifaddr *)
2238                             ifa_ifwithaddr((struct sockaddr *)&d);
2239                         if (ia &&
2240                             !(ia->ia6_flags &
2241                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2242                                 srcp = &ia->ia_addr.sin6_addr;
2243                         }
2244                 }
2245         }
2246
2247         if (srcp == NULL) {
2248                 int e;
2249                 struct sockaddr_in6 sin6;
2250                 struct route_in6 ro;
2251
2252                 /*
2253                  * This case matches to multicasts, our anycast, or unicasts
2254                  * that we do not own.  Select a source address based on the
2255                  * source address of the erroneous packet.
2256                  */
2257                 bzero(&sin6, sizeof(sin6));
2258                 sin6.sin6_family = AF_INET6;
2259                 sin6.sin6_len = sizeof(sin6);
2260                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2261
2262                 bzero(&ro, sizeof(ro));
2263                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2264                 if (ro.ro_rt)
2265                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2266                 if (e) {
2267                         char ip6buf[INET6_ADDRSTRLEN];
2268                         nd6log((LOG_DEBUG,
2269                             "icmp6_reflect: source can't be determined: "
2270                             "dst=%s, error=%d\n",
2271                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2272                         goto bad;
2273                 }
2274                 srcp = &src;
2275         }
2276
2277         ip6->ip6_src = *srcp;
2278         ip6->ip6_flow = 0;
2279         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2280         ip6->ip6_vfc |= IPV6_VERSION;
2281         ip6->ip6_nxt = IPPROTO_ICMPV6;
2282         if (outif)
2283                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2284         else if (m->m_pkthdr.rcvif) {
2285                 /* XXX: This may not be the outgoing interface */
2286                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2287         } else
2288                 ip6->ip6_hlim = V_ip6_defhlim;
2289
2290         icmp6->icmp6_cksum = 0;
2291         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2292             sizeof(struct ip6_hdr), plen);
2293
2294         /*
2295          * XXX option handling
2296          */
2297
2298         m->m_flags &= ~(M_BCAST|M_MCAST);
2299
2300         m_addr_changed(m);
2301
2302         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2303         if (outif)
2304                 icmp6_ifoutstat_inc(outif, type, code);
2305
2306         if (ia != NULL)
2307                 ifa_free(&ia->ia_ifa);
2308         return;
2309
2310  bad:
2311         if (ia != NULL)
2312                 ifa_free(&ia->ia_ifa);
2313         m_freem(m);
2314         return;
2315 }
2316
2317 void
2318 icmp6_fasttimo(void)
2319 {
2320
2321         mld_fasttimo();
2322 }
2323
2324 void
2325 icmp6_slowtimo(void)
2326 {
2327
2328         mld_slowtimo();
2329 }
2330
2331 static const char *
2332 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2333     struct in6_addr *tgt6)
2334 {
2335         static char buf[1024];
2336         char ip6bufs[INET6_ADDRSTRLEN];
2337         char ip6bufd[INET6_ADDRSTRLEN];
2338         char ip6buft[INET6_ADDRSTRLEN];
2339         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2340             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2341             ip6_sprintf(ip6buft, tgt6));
2342         return buf;
2343 }
2344
2345 void
2346 icmp6_redirect_input(struct mbuf *m, int off)
2347 {
2348         struct ifnet *ifp;
2349         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2350         struct nd_redirect *nd_rd;
2351         int icmp6len = ntohs(ip6->ip6_plen);
2352         char *lladdr = NULL;
2353         int lladdrlen = 0;
2354         struct rtentry *rt = NULL;
2355         int is_router;
2356         int is_onlink;
2357         struct in6_addr src6 = ip6->ip6_src;
2358         struct in6_addr redtgt6;
2359         struct in6_addr reddst6;
2360         union nd_opts ndopts;
2361         char ip6buf[INET6_ADDRSTRLEN];
2362
2363         M_ASSERTPKTHDR(m);
2364         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2365
2366         ifp = m->m_pkthdr.rcvif;
2367
2368         /* XXX if we are router, we don't update route by icmp6 redirect */
2369         if (V_ip6_forwarding)
2370                 goto freeit;
2371         if (!V_icmp6_rediraccept)
2372                 goto freeit;
2373
2374 #ifndef PULLDOWN_TEST
2375         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2376         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2377 #else
2378         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2379         if (nd_rd == NULL) {
2380                 ICMP6STAT_INC(icp6s_tooshort);
2381                 return;
2382         }
2383 #endif
2384         redtgt6 = nd_rd->nd_rd_target;
2385         reddst6 = nd_rd->nd_rd_dst;
2386
2387         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2388             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2389                 goto freeit;
2390         }
2391
2392         /* validation */
2393         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2394                 nd6log((LOG_ERR,
2395                     "ICMP6 redirect sent from %s rejected; "
2396                     "must be from linklocal\n",
2397                     ip6_sprintf(ip6buf, &src6)));
2398                 goto bad;
2399         }
2400         if (ip6->ip6_hlim != 255) {
2401                 nd6log((LOG_ERR,
2402                     "ICMP6 redirect sent from %s rejected; "
2403                     "hlim=%d (must be 255)\n",
2404                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2405                 goto bad;
2406         }
2407     {
2408         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2409         struct sockaddr_in6 sin6;
2410         struct in6_addr *gw6;
2411
2412         bzero(&sin6, sizeof(sin6));
2413         sin6.sin6_family = AF_INET6;
2414         sin6.sin6_len = sizeof(struct sockaddr_in6);
2415         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2416         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2417         if (rt) {
2418                 if (rt->rt_gateway == NULL ||
2419                     rt->rt_gateway->sa_family != AF_INET6) {
2420                         RTFREE_LOCKED(rt);
2421                         nd6log((LOG_ERR,
2422                             "ICMP6 redirect rejected; no route "
2423                             "with inet6 gateway found for redirect dst: %s\n",
2424                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2425                         goto bad;
2426                 }
2427
2428                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2429                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2430                         RTFREE_LOCKED(rt);
2431                         nd6log((LOG_ERR,
2432                             "ICMP6 redirect rejected; "
2433                             "not equal to gw-for-src=%s (must be same): "
2434                             "%s\n",
2435                             ip6_sprintf(ip6buf, gw6),
2436                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2437                         goto bad;
2438                 }
2439         } else {
2440                 nd6log((LOG_ERR,
2441                     "ICMP6 redirect rejected; "
2442                     "no route found for redirect dst: %s\n",
2443                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2444                 goto bad;
2445         }
2446         RTFREE_LOCKED(rt);
2447         rt = NULL;
2448     }
2449         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2450                 nd6log((LOG_ERR,
2451                     "ICMP6 redirect rejected; "
2452                     "redirect dst must be unicast: %s\n",
2453                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2454                 goto bad;
2455         }
2456
2457         is_router = is_onlink = 0;
2458         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2459                 is_router = 1;  /* router case */
2460         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2461                 is_onlink = 1;  /* on-link destination case */
2462         if (!is_router && !is_onlink) {
2463                 nd6log((LOG_ERR,
2464                     "ICMP6 redirect rejected; "
2465                     "neither router case nor onlink case: %s\n",
2466                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2467                 goto bad;
2468         }
2469         /* validation passed */
2470
2471         icmp6len -= sizeof(*nd_rd);
2472         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2473         if (nd6_options(&ndopts) < 0) {
2474                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2475                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2476                 /* nd6_options have incremented stats */
2477                 goto freeit;
2478         }
2479
2480         if (ndopts.nd_opts_tgt_lladdr) {
2481                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2482                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2483         }
2484
2485         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2486                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2487                     "(if %d, icmp6 packet %d): %s\n",
2488                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2489                     ifp->if_addrlen, lladdrlen - 2,
2490                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2491                 goto bad;
2492         }
2493
2494         /* RFC 2461 8.3 */
2495         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2496             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2497
2498         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2499                 /* perform rtredirect */
2500                 struct sockaddr_in6 sdst;
2501                 struct sockaddr_in6 sgw;
2502                 struct sockaddr_in6 ssrc;
2503                 u_int fibnum;
2504
2505                 bzero(&sdst, sizeof(sdst));
2506                 bzero(&sgw, sizeof(sgw));
2507                 bzero(&ssrc, sizeof(ssrc));
2508                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2509                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2510                         sizeof(struct sockaddr_in6);
2511                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2512                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2513                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2514                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2515                         in6_rtredirect((struct sockaddr *)&sdst,
2516                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2517                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2518                             fibnum);
2519         }
2520         /* finally update cached route in each socket via pfctlinput */
2521     {
2522         struct sockaddr_in6 sdst;
2523
2524         bzero(&sdst, sizeof(sdst));
2525         sdst.sin6_family = AF_INET6;
2526         sdst.sin6_len = sizeof(struct sockaddr_in6);
2527         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2528         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2529 #ifdef IPSEC
2530         key_sa_routechange((struct sockaddr *)&sdst);
2531 #endif /* IPSEC */
2532     }
2533
2534  freeit:
2535         m_freem(m);
2536         return;
2537
2538  bad:
2539         ICMP6STAT_INC(icp6s_badredirect);
2540         m_freem(m);
2541 }
2542
2543 void
2544 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2545 {
2546         struct ifnet *ifp;      /* my outgoing interface */
2547         struct in6_addr *ifp_ll6;
2548         struct in6_addr *router_ll6;
2549         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2550         struct mbuf *m = NULL;  /* newly allocated one */
2551         struct m_tag *mtag;
2552         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2553         struct nd_redirect *nd_rd;
2554         struct llentry *ln = NULL;
2555         size_t maxlen;
2556         u_char *p;
2557         struct ifnet *outif = NULL;
2558         struct sockaddr_in6 src_sa;
2559
2560         icmp6_errcount(ND_REDIRECT, 0);
2561
2562         /* if we are not router, we don't send icmp6 redirect */
2563         if (!V_ip6_forwarding)
2564                 goto fail;
2565
2566         /* sanity check */
2567         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2568                 goto fail;
2569
2570         /*
2571          * Address check:
2572          *  the source address must identify a neighbor, and
2573          *  the destination address must not be a multicast address
2574          *  [RFC 2461, sec 8.2]
2575          */
2576         sip6 = mtod(m0, struct ip6_hdr *);
2577         bzero(&src_sa, sizeof(src_sa));
2578         src_sa.sin6_family = AF_INET6;
2579         src_sa.sin6_len = sizeof(src_sa);
2580         src_sa.sin6_addr = sip6->ip6_src;
2581         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2582                 goto fail;
2583         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2584                 goto fail;      /* what should we do here? */
2585
2586         /* rate limit */
2587         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2588                 goto fail;
2589
2590         /*
2591          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2592          * we almost always ask for an mbuf cluster for simplicity.
2593          * (MHLEN < IPV6_MMTU is almost always true)
2594          */
2595 #if IPV6_MMTU >= MCLBYTES
2596 # error assumption failed about IPV6_MMTU and MCLBYTES
2597 #endif
2598         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2599         if (m && IPV6_MMTU >= MHLEN)
2600                 MCLGET(m, M_DONTWAIT);
2601         if (!m)
2602                 goto fail;
2603         M_SETFIB(m, rt->rt_fibnum);
2604         m->m_pkthdr.rcvif = NULL;
2605         m->m_len = 0;
2606         maxlen = M_TRAILINGSPACE(m);
2607         maxlen = min(IPV6_MMTU, maxlen);
2608         /* just for safety */
2609         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2610             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2611                 goto fail;
2612         }
2613
2614         {
2615                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2616                 struct in6_ifaddr *ia;
2617                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2618                                                  IN6_IFF_NOTREADY|
2619                                                  IN6_IFF_ANYCAST)) == NULL)
2620                         goto fail;
2621                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2622                 /* XXXRW: reference released prematurely. */
2623                 ifa_free(&ia->ia_ifa);
2624         }
2625
2626         /* get ip6 linklocal address for the router. */
2627         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2628                 struct sockaddr_in6 *sin6;
2629                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2630                 router_ll6 = &sin6->sin6_addr;
2631                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2632                         router_ll6 = (struct in6_addr *)NULL;
2633         } else
2634                 router_ll6 = (struct in6_addr *)NULL;
2635
2636         /* ip6 */
2637         ip6 = mtod(m, struct ip6_hdr *);
2638         ip6->ip6_flow = 0;
2639         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2640         ip6->ip6_vfc |= IPV6_VERSION;
2641         /* ip6->ip6_plen will be set later */
2642         ip6->ip6_nxt = IPPROTO_ICMPV6;
2643         ip6->ip6_hlim = 255;
2644         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2645         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2646         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2647
2648         /* ND Redirect */
2649         nd_rd = (struct nd_redirect *)(ip6 + 1);
2650         nd_rd->nd_rd_type = ND_REDIRECT;
2651         nd_rd->nd_rd_code = 0;
2652         nd_rd->nd_rd_reserved = 0;
2653         if (rt->rt_flags & RTF_GATEWAY) {
2654                 /*
2655                  * nd_rd->nd_rd_target must be a link-local address in
2656                  * better router cases.
2657                  */
2658                 if (!router_ll6)
2659                         goto fail;
2660                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2661                     sizeof(nd_rd->nd_rd_target));
2662                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2663                     sizeof(nd_rd->nd_rd_dst));
2664         } else {
2665                 /* make sure redtgt == reddst */
2666                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2667                     sizeof(nd_rd->nd_rd_target));
2668                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2669                     sizeof(nd_rd->nd_rd_dst));
2670         }
2671
2672         p = (u_char *)(nd_rd + 1);
2673
2674         if (!router_ll6)
2675                 goto nolladdropt;
2676
2677         {
2678                 /* target lladdr option */
2679                 int len;
2680                 struct nd_opt_hdr *nd_opt;
2681                 char *lladdr;
2682
2683                 IF_AFDATA_RLOCK(ifp);
2684                 ln = nd6_lookup(router_ll6, 0, ifp);
2685                 IF_AFDATA_RUNLOCK(ifp);
2686                 if (ln == NULL)
2687                         goto nolladdropt;
2688
2689                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2690                 len = (len + 7) & ~7;   /* round by 8 */
2691                 /* safety check */
2692                 if (len + (p - (u_char *)ip6) > maxlen)                         
2693                         goto nolladdropt;
2694
2695                 if (ln->la_flags & LLE_VALID) {
2696                         nd_opt = (struct nd_opt_hdr *)p;
2697                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2698                         nd_opt->nd_opt_len = len >> 3;
2699                         lladdr = (char *)(nd_opt + 1);
2700                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2701                         p += len;
2702                 }
2703         }
2704 nolladdropt:
2705         if (ln != NULL)
2706                 LLE_RUNLOCK(ln);
2707                 
2708         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2709
2710         /* just to be safe */
2711 #ifdef M_DECRYPTED      /*not openbsd*/
2712         if (m0->m_flags & M_DECRYPTED)
2713                 goto noredhdropt;
2714 #endif
2715         if (p - (u_char *)ip6 > maxlen)
2716                 goto noredhdropt;
2717
2718         {
2719                 /* redirected header option */
2720                 int len;
2721                 struct nd_opt_rd_hdr *nd_opt_rh;
2722
2723                 /*
2724                  * compute the maximum size for icmp6 redirect header option.
2725                  * XXX room for auth header?
2726                  */
2727                 len = maxlen - (p - (u_char *)ip6);
2728                 len &= ~7;
2729
2730                 /* This is just for simplicity. */
2731                 if (m0->m_pkthdr.len != m0->m_len) {
2732                         if (m0->m_next) {
2733                                 m_freem(m0->m_next);
2734                                 m0->m_next = NULL;
2735                         }
2736                         m0->m_pkthdr.len = m0->m_len;
2737                 }
2738
2739                 /*
2740                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2741                  * about padding/truncate rule for the original IP packet.
2742                  * From the discussion on IPv6imp in Feb 1999,
2743                  * the consensus was:
2744                  * - "attach as much as possible" is the goal
2745                  * - pad if not aligned (original size can be guessed by
2746                  *   original ip6 header)
2747                  * Following code adds the padding if it is simple enough,
2748                  * and truncates if not.
2749                  */
2750                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2751                         panic("assumption failed in %s:%d", __FILE__,
2752                             __LINE__);
2753
2754                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2755                         /* not enough room, truncate */
2756                         m0->m_pkthdr.len = m0->m_len = len -
2757                             sizeof(*nd_opt_rh);
2758                 } else {
2759                         /* enough room, pad or truncate */
2760                         size_t extra;
2761
2762                         extra = m0->m_pkthdr.len % 8;
2763                         if (extra) {
2764                                 /* pad if easy enough, truncate if not */
2765                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2766                                         /* pad */
2767                                         m0->m_len += (8 - extra);
2768                                         m0->m_pkthdr.len += (8 - extra);
2769                                 } else {
2770                                         /* truncate */
2771                                         m0->m_pkthdr.len -= extra;
2772                                         m0->m_len -= extra;
2773                                 }
2774                         }
2775                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2776                         m0->m_pkthdr.len = m0->m_len = len -
2777                             sizeof(*nd_opt_rh);
2778                 }
2779
2780                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2781                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2782                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2783                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2784                 p += sizeof(*nd_opt_rh);
2785                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2786
2787                 /* connect m0 to m */
2788                 m_tag_delete_chain(m0, NULL);
2789                 m0->m_flags &= ~M_PKTHDR;
2790                 m->m_next = m0;
2791                 m->m_pkthdr.len = m->m_len + m0->m_len;
2792                 m0 = NULL;
2793         }
2794 noredhdropt:;
2795         if (m0) {
2796                 m_freem(m0);
2797                 m0 = NULL;
2798         }
2799
2800         /* XXX: clear embedded link IDs in the inner header */
2801         in6_clearscope(&sip6->ip6_src);
2802         in6_clearscope(&sip6->ip6_dst);
2803         in6_clearscope(&nd_rd->nd_rd_target);
2804         in6_clearscope(&nd_rd->nd_rd_dst);
2805
2806         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2807
2808         nd_rd->nd_rd_cksum = 0;
2809         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2810             sizeof(*ip6), ntohs(ip6->ip6_plen));
2811
2812         if (send_sendso_input_hook != NULL) {
2813                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2814                         M_NOWAIT);
2815                 if (mtag == NULL)
2816                         goto fail;
2817                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2818                 m_tag_prepend(m, mtag);
2819         }
2820
2821         /* send the packet to outside... */
2822         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2823         if (outif) {
2824                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2825                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2826         }
2827         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2828
2829         return;
2830
2831 fail:
2832         if (m)
2833                 m_freem(m);
2834         if (m0)
2835                 m_freem(m0);
2836 }
2837
2838 /*
2839  * ICMPv6 socket option processing.
2840  */
2841 int
2842 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2843 {
2844         int error = 0;
2845         int optlen;
2846         struct inpcb *inp = sotoinpcb(so);
2847         int level, op, optname;
2848
2849         if (sopt) {
2850                 level = sopt->sopt_level;
2851                 op = sopt->sopt_dir;
2852                 optname = sopt->sopt_name;
2853                 optlen = sopt->sopt_valsize;
2854         } else
2855                 level = op = optname = optlen = 0;
2856
2857         if (level != IPPROTO_ICMPV6) {
2858                 return EINVAL;
2859         }
2860
2861         switch (op) {
2862         case PRCO_SETOPT:
2863                 switch (optname) {
2864                 case ICMP6_FILTER:
2865                     {
2866                         struct icmp6_filter ic6f;
2867
2868                         if (optlen != sizeof(ic6f)) {
2869                                 error = EMSGSIZE;
2870                                 break;
2871                         }
2872                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2873                         if (error == 0) {
2874                                 INP_WLOCK(inp);
2875                                 *inp->in6p_icmp6filt = ic6f;
2876                                 INP_WUNLOCK(inp);
2877                         }
2878                         break;
2879                     }
2880
2881                 default:
2882                         error = ENOPROTOOPT;
2883                         break;
2884                 }
2885                 break;
2886
2887         case PRCO_GETOPT:
2888                 switch (optname) {
2889                 case ICMP6_FILTER:
2890                     {
2891                         struct icmp6_filter ic6f;
2892
2893                         INP_RLOCK(inp);
2894                         ic6f = *inp->in6p_icmp6filt;
2895                         INP_RUNLOCK(inp);
2896                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2897                         break;
2898                     }
2899
2900                 default:
2901                         error = ENOPROTOOPT;
2902                         break;
2903                 }
2904                 break;
2905         }
2906
2907         return (error);
2908 }
2909
2910 /*
2911  * Perform rate limit check.
2912  * Returns 0 if it is okay to send the icmp6 packet.
2913  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2914  * limitation.
2915  *
2916  * XXX per-destination/type check necessary?
2917  *
2918  * dst - not used at this moment
2919  * type - not used at this moment
2920  * code - not used at this moment
2921  */
2922 static int
2923 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2924     const int code)
2925 {
2926         int ret;
2927
2928         ret = 0;        /* okay to send */
2929
2930         /* PPS limit */
2931         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2932             V_icmp6errppslim)) {
2933                 /* The packet is subject to rate limit */
2934                 ret++;
2935         }
2936
2937         return ret;
2938 }