]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/netinet6/icmp6.c
Dtrace: enablings on defunct providers prevent providers from unregistering
[FreeBSD/FreeBSD.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(struct icmp6errstat *, 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(struct icmp6errstat *stat, int type, int code)
163 {
164         switch (type) {
165         case ICMP6_DST_UNREACH:
166                 switch (code) {
167                 case ICMP6_DST_UNREACH_NOROUTE:
168                         stat->icp6errs_dst_unreach_noroute++;
169                         return;
170                 case ICMP6_DST_UNREACH_ADMIN:
171                         stat->icp6errs_dst_unreach_admin++;
172                         return;
173                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
174                         stat->icp6errs_dst_unreach_beyondscope++;
175                         return;
176                 case ICMP6_DST_UNREACH_ADDR:
177                         stat->icp6errs_dst_unreach_addr++;
178                         return;
179                 case ICMP6_DST_UNREACH_NOPORT:
180                         stat->icp6errs_dst_unreach_noport++;
181                         return;
182                 }
183                 break;
184         case ICMP6_PACKET_TOO_BIG:
185                 stat->icp6errs_packet_too_big++;
186                 return;
187         case ICMP6_TIME_EXCEEDED:
188                 switch (code) {
189                 case ICMP6_TIME_EXCEED_TRANSIT:
190                         stat->icp6errs_time_exceed_transit++;
191                         return;
192                 case ICMP6_TIME_EXCEED_REASSEMBLY:
193                         stat->icp6errs_time_exceed_reassembly++;
194                         return;
195                 }
196                 break;
197         case ICMP6_PARAM_PROB:
198                 switch (code) {
199                 case ICMP6_PARAMPROB_HEADER:
200                         stat->icp6errs_paramprob_header++;
201                         return;
202                 case ICMP6_PARAMPROB_NEXTHEADER:
203                         stat->icp6errs_paramprob_nextheader++;
204                         return;
205                 case ICMP6_PARAMPROB_OPTION:
206                         stat->icp6errs_paramprob_option++;
207                         return;
208                 }
209                 break;
210         case ND_REDIRECT:
211                 stat->icp6errs_redirect++;
212                 return;
213         }
214         stat->icp6errs_unknown++;
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(&V_icmp6stat.icp6s_outerrhist, 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_NOWAIT);        /* FIB is also copied over. */
364         if (m && m->m_len < preplen)
365                 m = m_pullup(m, preplen);
366         if (m == NULL) {
367                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
368                 return;
369         }
370
371         nip6 = mtod(m, struct ip6_hdr *);
372         nip6->ip6_src  = oip6->ip6_src;
373         nip6->ip6_dst  = oip6->ip6_dst;
374
375         in6_clearscope(&oip6->ip6_src);
376         in6_clearscope(&oip6->ip6_dst);
377
378         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
379         icmp6->icmp6_type = type;
380         icmp6->icmp6_code = code;
381         icmp6->icmp6_pptr = htonl((u_int32_t)param);
382
383         /*
384          * icmp6_reflect() is designed to be in the input path.
385          * icmp6_error() can be called from both input and output path,
386          * and if we are in output path rcvif could contain bogus value.
387          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
388          * information in ip header (nip6).
389          */
390         m->m_pkthdr.rcvif = NULL;
391
392         ICMP6STAT_INC(icp6s_outhist[type]);
393         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
394
395         return;
396
397   freeit:
398         /*
399          * If we can't tell whether or not we can generate ICMP6, free it.
400          */
401         m_freem(m);
402 }
403
404 /*
405  * Process a received ICMP6 message.
406  */
407 int
408 icmp6_input(struct mbuf **mp, int *offp, int proto)
409 {
410         struct mbuf *m = *mp, *n;
411         struct ifnet *ifp;
412         struct ip6_hdr *ip6, *nip6;
413         struct icmp6_hdr *icmp6, *nicmp6;
414         int off = *offp;
415         int icmp6len = m->m_pkthdr.len - *offp;
416         int code, sum, noff;
417         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
418         int ip6len, error;
419
420         ifp = m->m_pkthdr.rcvif;
421
422 #ifndef PULLDOWN_TEST
423         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
424         /* m might change if M_LOOP.  So, call mtod after this */
425 #endif
426
427         /*
428          * Locate icmp6 structure in mbuf, and check
429          * that not corrupted and of at least minimum length
430          */
431
432         ip6 = mtod(m, struct ip6_hdr *);
433         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
434         if (icmp6len < sizeof(struct icmp6_hdr)) {
435                 ICMP6STAT_INC(icp6s_tooshort);
436                 goto freeit;
437         }
438
439         /*
440          * Check multicast group membership.
441          * Note: SSM filters are not applied for ICMPv6 traffic.
442          */
443         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
444                 struct in6_multi        *inm;
445
446                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
447                 if (inm == NULL) {
448                         IP6STAT_INC(ip6s_notmember);
449                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
450                         goto freeit;
451                 }
452         }
453
454         /*
455          * calculate the checksum
456          */
457 #ifndef PULLDOWN_TEST
458         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
459 #else
460         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
461         if (icmp6 == NULL) {
462                 ICMP6STAT_INC(icp6s_tooshort);
463                 return IPPROTO_DONE;
464         }
465 #endif
466         code = icmp6->icmp6_code;
467
468         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
469                 nd6log((LOG_ERR,
470                     "ICMP6 checksum error(%d|%x) %s\n",
471                     icmp6->icmp6_type, sum,
472                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
473                 ICMP6STAT_INC(icp6s_checksum);
474                 goto freeit;
475         }
476
477         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
478                 /*
479                  * Deliver very specific ICMP6 type only.
480                  * This is important to deliver TOOBIG.  Otherwise PMTUD
481                  * will not work.
482                  */
483                 switch (icmp6->icmp6_type) {
484                 case ICMP6_DST_UNREACH:
485                 case ICMP6_PACKET_TOO_BIG:
486                 case ICMP6_TIME_EXCEEDED:
487                         break;
488                 default:
489                         goto freeit;
490                 }
491         }
492
493         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
494         icmp6_ifstat_inc(ifp, ifs6_in_msg);
495         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
496                 icmp6_ifstat_inc(ifp, ifs6_in_error);
497
498         switch (icmp6->icmp6_type) {
499         case ICMP6_DST_UNREACH:
500                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
501                 switch (code) {
502                 case ICMP6_DST_UNREACH_NOROUTE:
503                         code = PRC_UNREACH_NET;
504                         break;
505                 case ICMP6_DST_UNREACH_ADMIN:
506                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
507                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
508                         break;
509                 case ICMP6_DST_UNREACH_ADDR:
510                         code = PRC_HOSTDEAD;
511                         break;
512                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
513                         /* I mean "source address was incorrect." */
514                         code = PRC_PARAMPROB;
515                         break;
516                 case ICMP6_DST_UNREACH_NOPORT:
517                         code = PRC_UNREACH_PORT;
518                         break;
519                 default:
520                         goto badcode;
521                 }
522                 goto deliver;
523                 break;
524
525         case ICMP6_PACKET_TOO_BIG:
526                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
527
528                 /* validation is made in icmp6_mtudisc_update */
529
530                 code = PRC_MSGSIZE;
531
532                 /*
533                  * Updating the path MTU will be done after examining
534                  * intermediate extension headers.
535                  */
536                 goto deliver;
537                 break;
538
539         case ICMP6_TIME_EXCEEDED:
540                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
541                 switch (code) {
542                 case ICMP6_TIME_EXCEED_TRANSIT:
543                         code = PRC_TIMXCEED_INTRANS;
544                         break;
545                 case ICMP6_TIME_EXCEED_REASSEMBLY:
546                         code = PRC_TIMXCEED_REASS;
547                         break;
548                 default:
549                         goto badcode;
550                 }
551                 goto deliver;
552                 break;
553
554         case ICMP6_PARAM_PROB:
555                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
556                 switch (code) {
557                 case ICMP6_PARAMPROB_NEXTHEADER:
558                         code = PRC_UNREACH_PROTOCOL;
559                         break;
560                 case ICMP6_PARAMPROB_HEADER:
561                 case ICMP6_PARAMPROB_OPTION:
562                         code = PRC_PARAMPROB;
563                         break;
564                 default:
565                         goto badcode;
566                 }
567                 goto deliver;
568                 break;
569
570         case ICMP6_ECHO_REQUEST:
571                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
572                 if (code != 0)
573                         goto badcode;
574                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
575                         /* Give up remote */
576                         break;
577                 }
578                 if ((n->m_flags & M_EXT) != 0
579                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
580                         struct mbuf *n0 = n;
581                         int n0len;
582
583                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
584                         n = m_gethdr(M_NOWAIT, n0->m_type);
585                         if (n == NULL) {
586                                 /* Give up remote */
587                                 m_freem(n0);
588                                 break;
589                         }
590
591                         m_move_pkthdr(n, n0);   /* FIB copied. */
592                         n0len = n0->m_pkthdr.len;       /* save for use below */
593                         /*
594                          * Copy IPv6 and ICMPv6 only.
595                          */
596                         nip6 = mtod(n, struct ip6_hdr *);
597                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
598                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
599                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
600                         noff = sizeof(struct ip6_hdr);
601                         /* new mbuf contains only ipv6+icmpv6 headers */
602                         n->m_len = noff + sizeof(struct icmp6_hdr);
603                         /*
604                          * Adjust mbuf.  ip6_plen will be adjusted in
605                          * ip6_output().
606                          */
607                         m_adj(n0, off + sizeof(struct icmp6_hdr));
608                         /* recalculate complete packet size */
609                         n->m_pkthdr.len = n0len + (noff - off);
610                         n->m_next = n0;
611                 } else {
612                         nip6 = mtod(n, struct ip6_hdr *);
613                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
614                             sizeof(*nicmp6));
615                         noff = off;
616                 }
617                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
618                 nicmp6->icmp6_code = 0;
619                 if (n) {
620                         ICMP6STAT_INC(icp6s_reflect);
621                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
622                         icmp6_reflect(n, noff);
623                 }
624                 break;
625
626         case ICMP6_ECHO_REPLY:
627                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
628                 if (code != 0)
629                         goto badcode;
630                 break;
631
632         case MLD_LISTENER_QUERY:
633         case MLD_LISTENER_REPORT:
634         case MLD_LISTENER_DONE:
635         case MLDV2_LISTENER_REPORT:
636                 /*
637                  * Drop MLD traffic which is not link-local, has a hop limit
638                  * of greater than 1 hop, or which does not have the
639                  * IPv6 HBH Router Alert option.
640                  * As IPv6 HBH options are stripped in ip6_input() we must
641                  * check an mbuf header flag.
642                  * XXX Should we also sanity check that these messages
643                  * were directed to a link-local multicast prefix?
644                  */
645                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
646                         goto freeit;
647                 if (mld_input(m, off, icmp6len) != 0)
648                         return (IPPROTO_DONE);
649                 /* m stays. */
650                 break;
651
652         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
653             {
654                 enum { WRU, FQDN } mode;
655
656                 if (!V_icmp6_nodeinfo)
657                         break;
658
659                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
660                         mode = WRU;
661                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
662                         mode = FQDN;
663                 else
664                         goto badlen;
665
666                 if (mode == FQDN) {
667 #ifndef PULLDOWN_TEST
668                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
669                             IPPROTO_DONE);
670 #endif
671                         n = m_copy(m, 0, M_COPYALL);
672                         if (n)
673                                 n = ni6_input(n, off);
674                         /* XXX meaningless if n == NULL */
675                         noff = sizeof(struct ip6_hdr);
676                 } else {
677                         struct prison *pr;
678                         u_char *p;
679                         int maxhlen, hlen;
680
681                         /*
682                          * XXX: this combination of flags is pointless,
683                          * but should we keep this for compatibility?
684                          */
685                         if ((V_icmp6_nodeinfo & 5) != 5)
686                                 break;
687
688                         if (code != 0)
689                                 goto badcode;
690
691                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
692                         n = m_gethdr(M_NOWAIT, m->m_type);
693                         if (n == NULL) {
694                                 /* Give up remote */
695                                 break;
696                         }
697                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
698                                 /*
699                                  * Previous code did a blind M_COPY_PKTHDR
700                                  * and said "just for rcvif".  If true, then
701                                  * we could tolerate the dup failing (due to
702                                  * the deep copy of the tag chain).  For now
703                                  * be conservative and just fail.
704                                  */
705                                 m_free(n);
706                                 n = NULL;
707                         }
708                         maxhlen = M_TRAILINGSPACE(n) -
709                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
710                         pr = curthread->td_ucred->cr_prison;
711                         mtx_lock(&pr->pr_mtx);
712                         hlen = strlen(pr->pr_hostname);
713                         if (maxhlen > hlen)
714                                 maxhlen = hlen;
715                         /*
716                          * Copy IPv6 and ICMPv6 only.
717                          */
718                         nip6 = mtod(n, struct ip6_hdr *);
719                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
720                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
721                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
722                         p = (u_char *)(nicmp6 + 1);
723                         bzero(p, 4);
724                         /* meaningless TTL */
725                         bcopy(pr->pr_hostname, p + 4, maxhlen);
726                         mtx_unlock(&pr->pr_mtx);
727                         noff = sizeof(struct ip6_hdr);
728                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
729                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
730                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
731                         nicmp6->icmp6_code = 0;
732                 }
733                 if (n) {
734                         ICMP6STAT_INC(icp6s_reflect);
735                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
736                         icmp6_reflect(n, noff);
737                 }
738                 break;
739             }
740
741         case ICMP6_WRUREPLY:
742                 if (code != 0)
743                         goto badcode;
744                 break;
745
746         case ND_ROUTER_SOLICIT:
747                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
748                 if (code != 0)
749                         goto badcode;
750                 if (icmp6len < sizeof(struct nd_router_solicit))
751                         goto badlen;
752                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
753                         /* give up local */
754
755                         /* Send incoming SeND packet to user space. */
756                         if (send_sendso_input_hook != NULL) {
757                                 IP6_EXTHDR_CHECK(m, off,
758                                     icmp6len, IPPROTO_DONE);
759                                 error = send_sendso_input_hook(m, ifp,
760                                     SND_IN, ip6len);
761                                 /* -1 == no app on SEND socket */
762                                 if (error == 0)
763                                         return (IPPROTO_DONE);
764                                 nd6_rs_input(m, off, icmp6len);
765                         } else
766                                 nd6_rs_input(m, off, icmp6len);
767                         m = NULL;
768                         goto freeit;
769                 }
770                 if (send_sendso_input_hook != NULL) {
771                         IP6_EXTHDR_CHECK(n, off,
772                             icmp6len, IPPROTO_DONE);
773                         error = send_sendso_input_hook(n, ifp,
774                             SND_IN, ip6len);
775                         if (error == 0)
776                                 goto freeit;
777                         /* -1 == no app on SEND socket */
778                         nd6_rs_input(n, off, icmp6len);
779                 } else
780                         nd6_rs_input(n, off, icmp6len);
781                 /* m stays. */
782                 break;
783
784         case ND_ROUTER_ADVERT:
785                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
786                 if (code != 0)
787                         goto badcode;
788                 if (icmp6len < sizeof(struct nd_router_advert))
789                         goto badlen;
790                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
791
792                         /* Send incoming SeND-protected/ND packet to user space. */
793                         if (send_sendso_input_hook != NULL) {
794                                 error = send_sendso_input_hook(m, ifp,
795                                     SND_IN, ip6len);
796                                 if (error == 0)
797                                         return (IPPROTO_DONE);
798                                 nd6_ra_input(m, off, icmp6len);
799                         } else
800                                 nd6_ra_input(m, off, icmp6len);
801                         m = NULL;
802                         goto freeit;
803                 }
804                 if (send_sendso_input_hook != NULL) {
805                         error = send_sendso_input_hook(n, ifp,
806                             SND_IN, ip6len);
807                         if (error == 0)
808                                 goto freeit;
809                         nd6_ra_input(n, off, icmp6len);
810                 } else
811                         nd6_ra_input(n, off, icmp6len);
812                 /* m stays. */
813                 break;
814
815         case ND_NEIGHBOR_SOLICIT:
816                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
817                 if (code != 0)
818                         goto badcode;
819                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
820                         goto badlen;
821                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
822                         if (send_sendso_input_hook != NULL) {
823                                 error = send_sendso_input_hook(m, ifp,
824                                     SND_IN, ip6len);
825                                 if (error == 0)
826                                         return (IPPROTO_DONE);
827                                 nd6_ns_input(m, off, icmp6len);
828                         } else
829                                 nd6_ns_input(m, off, icmp6len);
830                         m = NULL;
831                         goto freeit;
832                 }
833                 if (send_sendso_input_hook != NULL) {
834                         error = send_sendso_input_hook(n, ifp,
835                             SND_IN, ip6len);
836                         if (error == 0)
837                                 goto freeit;
838                         nd6_ns_input(n, off, icmp6len);
839                 } else
840                         nd6_ns_input(n, off, icmp6len);
841                 /* m stays. */
842                 break;
843
844         case ND_NEIGHBOR_ADVERT:
845                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
846                 if (code != 0)
847                         goto badcode;
848                 if (icmp6len < sizeof(struct nd_neighbor_advert))
849                         goto badlen;
850                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
851
852                         /* Send incoming SeND-protected/ND packet to user space. */
853                         if (send_sendso_input_hook != NULL) {
854                                 error = send_sendso_input_hook(m, ifp,
855                                     SND_IN, ip6len);
856                                 if (error == 0)
857                                         return (IPPROTO_DONE);
858                                 nd6_na_input(m, off, icmp6len);
859                         } else
860                                 nd6_na_input(m, off, icmp6len);
861                         m = NULL;
862                         goto freeit;
863                 }
864                 if (send_sendso_input_hook != NULL) {
865                         error = send_sendso_input_hook(n, ifp,
866                             SND_IN, ip6len);
867                         if (error == 0)
868                                 goto freeit;
869                         nd6_na_input(n, off, icmp6len);
870                 } else
871                         nd6_na_input(n, off, icmp6len);
872                 /* m stays. */
873                 break;
874
875         case ND_REDIRECT:
876                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
877                 if (code != 0)
878                         goto badcode;
879                 if (icmp6len < sizeof(struct nd_redirect))
880                         goto badlen;
881                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
882                         if (send_sendso_input_hook != NULL) {
883                                 error = send_sendso_input_hook(m, ifp,
884                                     SND_IN, ip6len);
885                                 if (error == 0)
886                                         return (IPPROTO_DONE);
887                             icmp6_redirect_input(m, off);
888                         } else
889                                 icmp6_redirect_input(m, off);
890                         m = NULL;
891                         goto freeit;
892                 }
893                 if (send_sendso_input_hook != NULL) {
894                         error = send_sendso_input_hook(n, ifp,
895                             SND_IN, ip6len);
896                         if (error == 0)
897                                 goto freeit;
898                         icmp6_redirect_input(n, off);
899                 } else
900                         icmp6_redirect_input(n, off);
901                 /* m stays. */
902                 break;
903
904         case ICMP6_ROUTER_RENUMBERING:
905                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
906                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
907                         goto badcode;
908                 if (icmp6len < sizeof(struct icmp6_router_renum))
909                         goto badlen;
910                 break;
911
912         default:
913                 nd6log((LOG_DEBUG,
914                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
915                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
916                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
917                     ifp ? ifp->if_index : 0));
918                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
919                         /* ICMPv6 error: MUST deliver it by spec... */
920                         code = PRC_NCMDS;
921                         /* deliver */
922                 } else {
923                         /* ICMPv6 informational: MUST not deliver */
924                         break;
925                 }
926         deliver:
927                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
928                         /* In this case, m should've been freed. */
929                         return (IPPROTO_DONE);
930                 }
931                 break;
932
933         badcode:
934                 ICMP6STAT_INC(icp6s_badcode);
935                 break;
936
937         badlen:
938                 ICMP6STAT_INC(icp6s_badlen);
939                 break;
940         }
941
942         /* deliver the packet to appropriate sockets */
943         icmp6_rip6_input(&m, *offp);
944
945         return IPPROTO_DONE;
946
947  freeit:
948         m_freem(m);
949         return IPPROTO_DONE;
950 }
951
952 static int
953 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
954 {
955         struct mbuf *m = *mp;
956         struct icmp6_hdr *icmp6;
957         struct ip6_hdr *eip6;
958         u_int32_t notifymtu;
959         struct sockaddr_in6 icmp6src, icmp6dst;
960
961         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
962                 ICMP6STAT_INC(icp6s_tooshort);
963                 goto freeit;
964         }
965 #ifndef PULLDOWN_TEST
966         IP6_EXTHDR_CHECK(m, off,
967             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
968         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
969 #else
970         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
971             sizeof(*icmp6) + sizeof(struct ip6_hdr));
972         if (icmp6 == NULL) {
973                 ICMP6STAT_INC(icp6s_tooshort);
974                 return (-1);
975         }
976 #endif
977         eip6 = (struct ip6_hdr *)(icmp6 + 1);
978
979         /* Detect the upper level protocol */
980         {
981                 void (*ctlfunc)(int, struct sockaddr *, void *);
982                 u_int8_t nxt = eip6->ip6_nxt;
983                 int eoff = off + sizeof(struct icmp6_hdr) +
984                     sizeof(struct ip6_hdr);
985                 struct ip6ctlparam ip6cp;
986                 struct in6_addr *finaldst = NULL;
987                 int icmp6type = icmp6->icmp6_type;
988                 struct ip6_frag *fh;
989                 struct ip6_rthdr *rth;
990                 struct ip6_rthdr0 *rth0;
991                 int rthlen;
992
993                 while (1) { /* XXX: should avoid infinite loop explicitly? */
994                         struct ip6_ext *eh;
995
996                         switch (nxt) {
997                         case IPPROTO_HOPOPTS:
998                         case IPPROTO_DSTOPTS:
999                         case IPPROTO_AH:
1000 #ifndef PULLDOWN_TEST
1001                                 IP6_EXTHDR_CHECK(m, 0,
1002                                     eoff + sizeof(struct ip6_ext), -1);
1003                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1004 #else
1005                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1006                                     eoff, sizeof(*eh));
1007                                 if (eh == NULL) {
1008                                         ICMP6STAT_INC(icp6s_tooshort);
1009                                         return (-1);
1010                                 }
1011 #endif
1012
1013                                 if (nxt == IPPROTO_AH)
1014                                         eoff += (eh->ip6e_len + 2) << 2;
1015                                 else
1016                                         eoff += (eh->ip6e_len + 1) << 3;
1017                                 nxt = eh->ip6e_nxt;
1018                                 break;
1019                         case IPPROTO_ROUTING:
1020                                 /*
1021                                  * When the erroneous packet contains a
1022                                  * routing header, we should examine the
1023                                  * header to determine the final destination.
1024                                  * Otherwise, we can't properly update
1025                                  * information that depends on the final
1026                                  * destination (e.g. path MTU).
1027                                  */
1028 #ifndef PULLDOWN_TEST
1029                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1030                                 rth = (struct ip6_rthdr *)
1031                                     (mtod(m, caddr_t) + eoff);
1032 #else
1033                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1034                                     eoff, sizeof(*rth));
1035                                 if (rth == NULL) {
1036                                         ICMP6STAT_INC(icp6s_tooshort);
1037                                         return (-1);
1038                                 }
1039 #endif
1040                                 rthlen = (rth->ip6r_len + 1) << 3;
1041                                 /*
1042                                  * XXX: currently there is no
1043                                  * officially defined type other
1044                                  * than type-0.
1045                                  * Note that if the segment left field
1046                                  * is 0, all intermediate hops must
1047                                  * have been passed.
1048                                  */
1049                                 if (rth->ip6r_segleft &&
1050                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1051                                         int hops;
1052
1053 #ifndef PULLDOWN_TEST
1054                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1055                                         rth0 = (struct ip6_rthdr0 *)
1056                                             (mtod(m, caddr_t) + eoff);
1057 #else
1058                                         IP6_EXTHDR_GET(rth0,
1059                                             struct ip6_rthdr0 *, m,
1060                                             eoff, rthlen);
1061                                         if (rth0 == NULL) {
1062                                                 ICMP6STAT_INC(icp6s_tooshort);
1063                                                 return (-1);
1064                                         }
1065 #endif
1066                                         /* just ignore a bogus header */
1067                                         if ((rth0->ip6r0_len % 2) == 0 &&
1068                                             (hops = rth0->ip6r0_len/2))
1069                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1070                                 }
1071                                 eoff += rthlen;
1072                                 nxt = rth->ip6r_nxt;
1073                                 break;
1074                         case IPPROTO_FRAGMENT:
1075 #ifndef PULLDOWN_TEST
1076                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1077                                     sizeof(struct ip6_frag), -1);
1078                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1079                                     eoff);
1080 #else
1081                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1082                                     eoff, sizeof(*fh));
1083                                 if (fh == NULL) {
1084                                         ICMP6STAT_INC(icp6s_tooshort);
1085                                         return (-1);
1086                                 }
1087 #endif
1088                                 /*
1089                                  * Data after a fragment header is meaningless
1090                                  * unless it is the first fragment, but
1091                                  * we'll go to the notify label for path MTU
1092                                  * discovery.
1093                                  */
1094                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1095                                         goto notify;
1096
1097                                 eoff += sizeof(struct ip6_frag);
1098                                 nxt = fh->ip6f_nxt;
1099                                 break;
1100                         default:
1101                                 /*
1102                                  * This case includes ESP and the No Next
1103                                  * Header.  In such cases going to the notify
1104                                  * label does not have any meaning
1105                                  * (i.e. ctlfunc will be NULL), but we go
1106                                  * anyway since we might have to update
1107                                  * path MTU information.
1108                                  */
1109                                 goto notify;
1110                         }
1111                 }
1112           notify:
1113 #ifndef PULLDOWN_TEST
1114                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1115 #else
1116                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1117                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1118                 if (icmp6 == NULL) {
1119                         ICMP6STAT_INC(icp6s_tooshort);
1120                         return (-1);
1121                 }
1122 #endif
1123
1124                 /*
1125                  * retrieve parameters from the inner IPv6 header, and convert
1126                  * them into sockaddr structures.
1127                  * XXX: there is no guarantee that the source or destination
1128                  * addresses of the inner packet are in the same scope as
1129                  * the addresses of the icmp packet.  But there is no other
1130                  * way to determine the zone.
1131                  */
1132                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1133
1134                 bzero(&icmp6dst, sizeof(icmp6dst));
1135                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1136                 icmp6dst.sin6_family = AF_INET6;
1137                 if (finaldst == NULL)
1138                         icmp6dst.sin6_addr = eip6->ip6_dst;
1139                 else
1140                         icmp6dst.sin6_addr = *finaldst;
1141                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1142                         goto freeit;
1143                 bzero(&icmp6src, sizeof(icmp6src));
1144                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1145                 icmp6src.sin6_family = AF_INET6;
1146                 icmp6src.sin6_addr = eip6->ip6_src;
1147                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1148                         goto freeit;
1149                 icmp6src.sin6_flowinfo =
1150                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1151
1152                 if (finaldst == NULL)
1153                         finaldst = &eip6->ip6_dst;
1154                 ip6cp.ip6c_m = m;
1155                 ip6cp.ip6c_icmp6 = icmp6;
1156                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1157                 ip6cp.ip6c_off = eoff;
1158                 ip6cp.ip6c_finaldst = finaldst;
1159                 ip6cp.ip6c_src = &icmp6src;
1160                 ip6cp.ip6c_nxt = nxt;
1161
1162                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1163                         notifymtu = ntohl(icmp6->icmp6_mtu);
1164                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1165                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1166                 }
1167
1168                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1169                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1170                 if (ctlfunc) {
1171                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1172                             &ip6cp);
1173                 }
1174         }
1175         *mp = m;
1176         return (0);
1177
1178   freeit:
1179         m_freem(m);
1180         return (-1);
1181 }
1182
1183 void
1184 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1185 {
1186         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1187         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1188         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1189         u_int mtu = ntohl(icmp6->icmp6_mtu);
1190         struct in_conninfo inc;
1191
1192 #if 0
1193         /*
1194          * RFC2460 section 5, last paragraph.
1195          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1196          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1197          * due to packet translator in the middle.
1198          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1199          * special handling.
1200          */
1201         if (mtu < IPV6_MMTU)
1202                 return;
1203 #endif
1204
1205         /*
1206          * we reject ICMPv6 too big with abnormally small value.
1207          * XXX what is the good definition of "abnormally small"?
1208          */
1209         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1210                 return;
1211
1212         if (!validated)
1213                 return;
1214
1215         /*
1216          * In case the suggested mtu is less than IPV6_MMTU, we
1217          * only need to remember that it was for above mentioned
1218          * "alwaysfrag" case.
1219          * Try to be as close to the spec as possible.
1220          */
1221         if (mtu < IPV6_MMTU)
1222                 mtu = IPV6_MMTU - 8;
1223
1224         bzero(&inc, sizeof(inc));
1225         inc.inc_flags |= INC_ISIPV6;
1226         inc.inc6_faddr = *dst;
1227         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1228                 return;
1229
1230         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1231                 tcp_hc_updatemtu(&inc, mtu);
1232                 ICMP6STAT_INC(icp6s_pmtuchg);
1233         }
1234 }
1235
1236 /*
1237  * Process a Node Information Query packet, based on
1238  * draft-ietf-ipngwg-icmp-name-lookups-07.
1239  *
1240  * Spec incompatibilities:
1241  * - IPv6 Subject address handling
1242  * - IPv4 Subject address handling support missing
1243  * - Proxy reply (answer even if it's not for me)
1244  * - joins NI group address at in6_ifattach() time only, does not cope
1245  *   with hostname changes by sethostname(3)
1246  */
1247 static struct mbuf *
1248 ni6_input(struct mbuf *m, int off)
1249 {
1250         struct icmp6_nodeinfo *ni6, *nni6;
1251         struct mbuf *n = NULL;
1252         struct prison *pr;
1253         u_int16_t qtype;
1254         int subjlen;
1255         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1256         struct ni_reply_fqdn *fqdn;
1257         int addrs;              /* for NI_QTYPE_NODEADDR */
1258         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1259         struct in6_addr in6_subj; /* subject address */
1260         struct ip6_hdr *ip6;
1261         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1262         char *subj = NULL;
1263         struct in6_ifaddr *ia6 = NULL;
1264
1265         ip6 = mtod(m, struct ip6_hdr *);
1266 #ifndef PULLDOWN_TEST
1267         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1268 #else
1269         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1270         if (ni6 == NULL) {
1271                 /* m is already reclaimed */
1272                 return (NULL);
1273         }
1274 #endif
1275
1276         /*
1277          * Validate IPv6 source address.
1278          * The default configuration MUST be to refuse answering queries from
1279          * global-scope addresses according to RFC4602.
1280          * Notes:
1281          *  - it's not very clear what "refuse" means; this implementation
1282          *    simply drops it.
1283          *  - it's not very easy to identify global-scope (unicast) addresses
1284          *    since there are many prefixes for them.  It should be safer
1285          *    and in practice sufficient to check "all" but loopback and
1286          *    link-local (note that site-local unicast was deprecated and
1287          *    ULA is defined as global scope-wise)
1288          */
1289         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1290             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1291             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1292                 goto bad;
1293
1294         /*
1295          * Validate IPv6 destination address.
1296          *
1297          * The Responder must discard the Query without further processing
1298          * unless it is one of the Responder's unicast or anycast addresses, or
1299          * a link-local scope multicast address which the Responder has joined.
1300          * [RFC4602, Section 5.]
1301          */
1302         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1303                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1304                         goto bad;
1305                 /* else it's a link-local multicast, fine */
1306         } else {                /* unicast or anycast */
1307                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1308                         goto bad; /* XXX impossible */
1309
1310                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1311                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1312                         ifa_free(&ia6->ia_ifa);
1313                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1314                                 "a temporary address in %s:%d",
1315                                __FILE__, __LINE__));
1316                         goto bad;
1317                 }
1318                 ifa_free(&ia6->ia_ifa);
1319         }
1320
1321         /* validate query Subject field. */
1322         qtype = ntohs(ni6->ni_qtype);
1323         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1324         switch (qtype) {
1325         case NI_QTYPE_NOOP:
1326         case NI_QTYPE_SUPTYPES:
1327                 /* 07 draft */
1328                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1329                         break;
1330                 /* FALLTHROUGH */
1331         case NI_QTYPE_FQDN:
1332         case NI_QTYPE_NODEADDR:
1333         case NI_QTYPE_IPV4ADDR:
1334                 switch (ni6->ni_code) {
1335                 case ICMP6_NI_SUBJ_IPV6:
1336 #if ICMP6_NI_SUBJ_IPV6 != 0
1337                 case 0:
1338 #endif
1339                         /*
1340                          * backward compatibility - try to accept 03 draft
1341                          * format, where no Subject is present.
1342                          */
1343                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1344                             subjlen == 0) {
1345                                 oldfqdn++;
1346                                 break;
1347                         }
1348 #if ICMP6_NI_SUBJ_IPV6 != 0
1349                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1350                                 goto bad;
1351 #endif
1352
1353                         if (subjlen != sizeof(struct in6_addr))
1354                                 goto bad;
1355
1356                         /*
1357                          * Validate Subject address.
1358                          *
1359                          * Not sure what exactly "address belongs to the node"
1360                          * means in the spec, is it just unicast, or what?
1361                          *
1362                          * At this moment we consider Subject address as
1363                          * "belong to the node" if the Subject address equals
1364                          * to the IPv6 destination address; validation for
1365                          * IPv6 destination address should have done enough
1366                          * check for us.
1367                          *
1368                          * We do not do proxy at this moment.
1369                          */
1370                         /* m_pulldown instead of copy? */
1371                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1372                             subjlen, (caddr_t)&in6_subj);
1373                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1374                                 goto bad;
1375
1376                         subj = (char *)&in6_subj;
1377                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1378                                 break;
1379
1380                         /*
1381                          * XXX if we are to allow other cases, we should really
1382                          * be careful about scope here.
1383                          * basically, we should disallow queries toward IPv6
1384                          * destination X with subject Y,
1385                          * if scope(X) > scope(Y).
1386                          * if we allow scope(X) > scope(Y), it will result in
1387                          * information leakage across scope boundary.
1388                          */
1389                         goto bad;
1390
1391                 case ICMP6_NI_SUBJ_FQDN:
1392                         /*
1393                          * Validate Subject name with gethostname(3).
1394                          *
1395                          * The behavior may need some debate, since:
1396                          * - we are not sure if the node has FQDN as
1397                          *   hostname (returned by gethostname(3)).
1398                          * - the code does wildcard match for truncated names.
1399                          *   however, we are not sure if we want to perform
1400                          *   wildcard match, if gethostname(3) side has
1401                          *   truncated hostname.
1402                          */
1403                         pr = curthread->td_ucred->cr_prison;
1404                         mtx_lock(&pr->pr_mtx);
1405                         n = ni6_nametodns(pr->pr_hostname,
1406                             strlen(pr->pr_hostname), 0);
1407                         mtx_unlock(&pr->pr_mtx);
1408                         if (!n || n->m_next || n->m_len == 0)
1409                                 goto bad;
1410                         IP6_EXTHDR_GET(subj, char *, m,
1411                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1412                         if (subj == NULL)
1413                                 goto bad;
1414                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1415                             n->m_len)) {
1416                                 goto bad;
1417                         }
1418                         m_freem(n);
1419                         n = NULL;
1420                         break;
1421
1422                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1423                 default:
1424                         goto bad;
1425                 }
1426                 break;
1427         }
1428
1429         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1430         switch (qtype) {
1431         case NI_QTYPE_FQDN:
1432                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1433                         goto bad;
1434                 break;
1435         case NI_QTYPE_NODEADDR:
1436         case NI_QTYPE_IPV4ADDR:
1437                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1438                         goto bad;
1439                 break;
1440         }
1441
1442         /* guess reply length */
1443         switch (qtype) {
1444         case NI_QTYPE_NOOP:
1445                 break;          /* no reply data */
1446         case NI_QTYPE_SUPTYPES:
1447                 replylen += sizeof(u_int32_t);
1448                 break;
1449         case NI_QTYPE_FQDN:
1450                 /* XXX will append an mbuf */
1451                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1452                 break;
1453         case NI_QTYPE_NODEADDR:
1454                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1455                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1456                     sizeof(u_int32_t))) > MCLBYTES)
1457                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1458                 break;
1459         case NI_QTYPE_IPV4ADDR:
1460                 /* unsupported - should respond with unknown Qtype? */
1461                 break;
1462         default:
1463                 /*
1464                  * XXX: We must return a reply with the ICMP6 code
1465                  * `unknown Qtype' in this case.  However we regard the case
1466                  * as an FQDN query for backward compatibility.
1467                  * Older versions set a random value to this field,
1468                  * so it rarely varies in the defined qtypes.
1469                  * But the mechanism is not reliable...
1470                  * maybe we should obsolete older versions.
1471                  */
1472                 qtype = NI_QTYPE_FQDN;
1473                 /* XXX will append an mbuf */
1474                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1475                 oldfqdn++;
1476                 break;
1477         }
1478
1479         /* Allocate an mbuf to reply. */
1480         if (replylen > MCLBYTES) {
1481                 /*
1482                  * XXX: should we try to allocate more? But MCLBYTES
1483                  * is probably much larger than IPV6_MMTU...
1484                  */
1485                 goto bad;
1486         }
1487         if (replylen > MHLEN)
1488                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1489         else
1490                 n = m_gethdr(M_NOWAIT, m->m_type);
1491         if (n == NULL) {
1492                 m_freem(m);
1493                 return (NULL);
1494         }
1495         m_move_pkthdr(n, m); /* just for recvif and FIB */
1496         n->m_pkthdr.len = n->m_len = replylen;
1497
1498         /* copy mbuf header and IPv6 + Node Information base headers */
1499         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1500         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1501         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1502
1503         /* qtype dependent procedure */
1504         switch (qtype) {
1505         case NI_QTYPE_NOOP:
1506                 nni6->ni_code = ICMP6_NI_SUCCESS;
1507                 nni6->ni_flags = 0;
1508                 break;
1509         case NI_QTYPE_SUPTYPES:
1510         {
1511                 u_int32_t v;
1512                 nni6->ni_code = ICMP6_NI_SUCCESS;
1513                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1514                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1515                 v = (u_int32_t)htonl(0x0000000f);
1516                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1517                 break;
1518         }
1519         case NI_QTYPE_FQDN:
1520                 nni6->ni_code = ICMP6_NI_SUCCESS;
1521                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1522                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1523                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1524                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1525                 /*
1526                  * XXX do we really have FQDN in hostname?
1527                  */
1528                 pr = curthread->td_ucred->cr_prison;
1529                 mtx_lock(&pr->pr_mtx);
1530                 n->m_next = ni6_nametodns(pr->pr_hostname,
1531                     strlen(pr->pr_hostname), oldfqdn);
1532                 mtx_unlock(&pr->pr_mtx);
1533                 if (n->m_next == NULL)
1534                         goto bad;
1535                 /* XXX we assume that n->m_next is not a chain */
1536                 if (n->m_next->m_next != NULL)
1537                         goto bad;
1538                 n->m_pkthdr.len += n->m_next->m_len;
1539                 break;
1540         case NI_QTYPE_NODEADDR:
1541         {
1542                 int lenlim, copied;
1543
1544                 nni6->ni_code = ICMP6_NI_SUCCESS;
1545                 n->m_pkthdr.len = n->m_len =
1546                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1547                 lenlim = M_TRAILINGSPACE(n);
1548                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1549                 /* XXX: reset mbuf length */
1550                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1551                     sizeof(struct icmp6_nodeinfo) + copied;
1552                 break;
1553         }
1554         default:
1555                 break;          /* XXX impossible! */
1556         }
1557
1558         nni6->ni_type = ICMP6_NI_REPLY;
1559         m_freem(m);
1560         return (n);
1561
1562   bad:
1563         m_freem(m);
1564         if (n)
1565                 m_freem(n);
1566         return (NULL);
1567 }
1568
1569 /*
1570  * make a mbuf with DNS-encoded string.  no compression support.
1571  *
1572  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1573  * treated as truncated name (two \0 at the end).  this is a wild guess.
1574  *
1575  * old - return pascal string if non-zero
1576  */
1577 static struct mbuf *
1578 ni6_nametodns(const char *name, int namelen, int old)
1579 {
1580         struct mbuf *m;
1581         char *cp, *ep;
1582         const char *p, *q;
1583         int i, len, nterm;
1584
1585         if (old)
1586                 len = namelen + 1;
1587         else
1588                 len = MCLBYTES;
1589
1590         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1591         if (len > MLEN)
1592                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1593         else
1594                 m = m_get(M_NOWAIT, MT_DATA);
1595         if (m == NULL)
1596                 goto fail;
1597
1598         if (old) {
1599                 m->m_len = len;
1600                 *mtod(m, char *) = namelen;
1601                 bcopy(name, mtod(m, char *) + 1, namelen);
1602                 return m;
1603         } else {
1604                 m->m_len = 0;
1605                 cp = mtod(m, char *);
1606                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1607
1608                 /* if not certain about my name, return empty buffer */
1609                 if (namelen == 0)
1610                         return m;
1611
1612                 /*
1613                  * guess if it looks like shortened hostname, or FQDN.
1614                  * shortened hostname needs two trailing "\0".
1615                  */
1616                 i = 0;
1617                 for (p = name; p < name + namelen; p++) {
1618                         if (*p && *p == '.')
1619                                 i++;
1620                 }
1621                 if (i < 2)
1622                         nterm = 2;
1623                 else
1624                         nterm = 1;
1625
1626                 p = name;
1627                 while (cp < ep && p < name + namelen) {
1628                         i = 0;
1629                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1630                                 i++;
1631                         /* result does not fit into mbuf */
1632                         if (cp + i + 1 >= ep)
1633                                 goto fail;
1634                         /*
1635                          * DNS label length restriction, RFC1035 page 8.
1636                          * "i == 0" case is included here to avoid returning
1637                          * 0-length label on "foo..bar".
1638                          */
1639                         if (i <= 0 || i >= 64)
1640                                 goto fail;
1641                         *cp++ = i;
1642                         bcopy(p, cp, i);
1643                         cp += i;
1644                         p = q;
1645                         if (p < name + namelen && *p == '.')
1646                                 p++;
1647                 }
1648                 /* termination */
1649                 if (cp + nterm >= ep)
1650                         goto fail;
1651                 while (nterm-- > 0)
1652                         *cp++ = '\0';
1653                 m->m_len = cp - mtod(m, char *);
1654                 return m;
1655         }
1656
1657         panic("should not reach here");
1658         /* NOTREACHED */
1659
1660  fail:
1661         if (m)
1662                 m_freem(m);
1663         return NULL;
1664 }
1665
1666 /*
1667  * check if two DNS-encoded string matches.  takes care of truncated
1668  * form (with \0\0 at the end).  no compression support.
1669  * XXX upper/lowercase match (see RFC2065)
1670  */
1671 static int
1672 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1673 {
1674         const char *a0, *b0;
1675         int l;
1676
1677         /* simplest case - need validation? */
1678         if (alen == blen && bcmp(a, b, alen) == 0)
1679                 return 1;
1680
1681         a0 = a;
1682         b0 = b;
1683
1684         /* termination is mandatory */
1685         if (alen < 2 || blen < 2)
1686                 return 0;
1687         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1688                 return 0;
1689         alen--;
1690         blen--;
1691
1692         while (a - a0 < alen && b - b0 < blen) {
1693                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1694                         return 0;
1695
1696                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1697                         return 0;
1698                 /* we don't support compression yet */
1699                 if (a[0] >= 64 || b[0] >= 64)
1700                         return 0;
1701
1702                 /* truncated case */
1703                 if (a[0] == 0 && a - a0 == alen - 1)
1704                         return 1;
1705                 if (b[0] == 0 && b - b0 == blen - 1)
1706                         return 1;
1707                 if (a[0] == 0 || b[0] == 0)
1708                         return 0;
1709
1710                 if (a[0] != b[0])
1711                         return 0;
1712                 l = a[0];
1713                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1714                         return 0;
1715                 if (bcmp(a + 1, b + 1, l) != 0)
1716                         return 0;
1717
1718                 a += 1 + l;
1719                 b += 1 + l;
1720         }
1721
1722         if (a - a0 == alen && b - b0 == blen)
1723                 return 1;
1724         else
1725                 return 0;
1726 }
1727
1728 /*
1729  * calculate the number of addresses to be returned in the node info reply.
1730  */
1731 static int
1732 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1733     struct in6_addr *subj)
1734 {
1735         struct ifnet *ifp;
1736         struct in6_ifaddr *ifa6;
1737         struct ifaddr *ifa;
1738         int addrs = 0, addrsofif, iffound = 0;
1739         int niflags = ni6->ni_flags;
1740
1741         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1742                 switch (ni6->ni_code) {
1743                 case ICMP6_NI_SUBJ_IPV6:
1744                         if (subj == NULL) /* must be impossible... */
1745                                 return (0);
1746                         break;
1747                 default:
1748                         /*
1749                          * XXX: we only support IPv6 subject address for
1750                          * this Qtype.
1751                          */
1752                         return (0);
1753                 }
1754         }
1755
1756         IFNET_RLOCK_NOSLEEP();
1757         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1758                 addrsofif = 0;
1759                 IF_ADDR_RLOCK(ifp);
1760                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1761                         if (ifa->ifa_addr->sa_family != AF_INET6)
1762                                 continue;
1763                         ifa6 = (struct in6_ifaddr *)ifa;
1764
1765                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1766                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1767                                 iffound = 1;
1768
1769                         /*
1770                          * IPv4-mapped addresses can only be returned by a
1771                          * Node Information proxy, since they represent
1772                          * addresses of IPv4-only nodes, which perforce do
1773                          * not implement this protocol.
1774                          * [icmp-name-lookups-07, Section 5.4]
1775                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1776                          * this function at this moment.
1777                          */
1778
1779                         /* What do we have to do about ::1? */
1780                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1781                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1782                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1783                                         continue;
1784                                 break;
1785                         case IPV6_ADDR_SCOPE_SITELOCAL:
1786                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1787                                         continue;
1788                                 break;
1789                         case IPV6_ADDR_SCOPE_GLOBAL:
1790                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1791                                         continue;
1792                                 break;
1793                         default:
1794                                 continue;
1795                         }
1796
1797                         /*
1798                          * check if anycast is okay.
1799                          * XXX: just experimental.  not in the spec.
1800                          */
1801                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1802                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1803                                 continue; /* we need only unicast addresses */
1804                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1805                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1806                                 continue;
1807                         }
1808                         addrsofif++; /* count the address */
1809                 }
1810                 IF_ADDR_RUNLOCK(ifp);
1811                 if (iffound) {
1812                         *ifpp = ifp;
1813                         IFNET_RUNLOCK_NOSLEEP();
1814                         return (addrsofif);
1815                 }
1816
1817                 addrs += addrsofif;
1818         }
1819         IFNET_RUNLOCK_NOSLEEP();
1820
1821         return (addrs);
1822 }
1823
1824 static int
1825 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1826     struct ifnet *ifp0, int resid)
1827 {
1828         struct ifnet *ifp;
1829         struct in6_ifaddr *ifa6;
1830         struct ifaddr *ifa;
1831         struct ifnet *ifp_dep = NULL;
1832         int copied = 0, allow_deprecated = 0;
1833         u_char *cp = (u_char *)(nni6 + 1);
1834         int niflags = ni6->ni_flags;
1835         u_int32_t ltime;
1836
1837         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1838                 return (0);     /* needless to copy */
1839
1840         IFNET_RLOCK_NOSLEEP();
1841         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1842   again:
1843
1844         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1845                 IF_ADDR_RLOCK(ifp);
1846                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1847                         if (ifa->ifa_addr->sa_family != AF_INET6)
1848                                 continue;
1849                         ifa6 = (struct in6_ifaddr *)ifa;
1850
1851                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1852                             allow_deprecated == 0) {
1853                                 /*
1854                                  * prefererred address should be put before
1855                                  * deprecated addresses.
1856                                  */
1857
1858                                 /* record the interface for later search */
1859                                 if (ifp_dep == NULL)
1860                                         ifp_dep = ifp;
1861
1862                                 continue;
1863                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1864                             allow_deprecated != 0)
1865                                 continue; /* we now collect deprecated addrs */
1866
1867                         /* What do we have to do about ::1? */
1868                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1869                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1870                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1871                                         continue;
1872                                 break;
1873                         case IPV6_ADDR_SCOPE_SITELOCAL:
1874                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1875                                         continue;
1876                                 break;
1877                         case IPV6_ADDR_SCOPE_GLOBAL:
1878                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1879                                         continue;
1880                                 break;
1881                         default:
1882                                 continue;
1883                         }
1884
1885                         /*
1886                          * check if anycast is okay.
1887                          * XXX: just experimental.  not in the spec.
1888                          */
1889                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1890                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1891                                 continue;
1892                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1893                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1894                                 continue;
1895                         }
1896
1897                         /* now we can copy the address */
1898                         if (resid < sizeof(struct in6_addr) +
1899                             sizeof(u_int32_t)) {
1900                                 IF_ADDR_RUNLOCK(ifp);
1901                                 /*
1902                                  * We give up much more copy.
1903                                  * Set the truncate flag and return.
1904                                  */
1905                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1906                                 IFNET_RUNLOCK_NOSLEEP();
1907                                 return (copied);
1908                         }
1909
1910                         /*
1911                          * Set the TTL of the address.
1912                          * The TTL value should be one of the following
1913                          * according to the specification:
1914                          *
1915                          * 1. The remaining lifetime of a DHCP lease on the
1916                          *    address, or
1917                          * 2. The remaining Valid Lifetime of a prefix from
1918                          *    which the address was derived through Stateless
1919                          *    Autoconfiguration.
1920                          *
1921                          * Note that we currently do not support stateful
1922                          * address configuration by DHCPv6, so the former
1923                          * case can't happen.
1924                          */
1925                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1926                                 ltime = ND6_INFINITE_LIFETIME;
1927                         else {
1928                                 if (ifa6->ia6_lifetime.ia6t_expire >
1929                                     time_second)
1930                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1931                                 else
1932                                         ltime = 0;
1933                         }
1934
1935                         bcopy(&ltime, cp, sizeof(u_int32_t));
1936                         cp += sizeof(u_int32_t);
1937
1938                         /* copy the address itself */
1939                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1940                             sizeof(struct in6_addr));
1941                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1942                         cp += sizeof(struct in6_addr);
1943
1944                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1945                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1946                 }
1947                 IF_ADDR_RUNLOCK(ifp);
1948                 if (ifp0)       /* we need search only on the specified IF */
1949                         break;
1950         }
1951
1952         if (allow_deprecated == 0 && ifp_dep != NULL) {
1953                 ifp = ifp_dep;
1954                 allow_deprecated = 1;
1955
1956                 goto again;
1957         }
1958
1959         IFNET_RUNLOCK_NOSLEEP();
1960
1961         return (copied);
1962 }
1963
1964 /*
1965  * XXX almost dup'ed code with rip6_input.
1966  */
1967 static int
1968 icmp6_rip6_input(struct mbuf **mp, int off)
1969 {
1970         struct mbuf *m = *mp;
1971         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1972         struct inpcb *in6p;
1973         struct inpcb *last = NULL;
1974         struct sockaddr_in6 fromsa;
1975         struct icmp6_hdr *icmp6;
1976         struct mbuf *opts = NULL;
1977
1978 #ifndef PULLDOWN_TEST
1979         /* this is assumed to be safe. */
1980         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1981 #else
1982         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1983         if (icmp6 == NULL) {
1984                 /* m is already reclaimed */
1985                 return (IPPROTO_DONE);
1986         }
1987 #endif
1988
1989         /*
1990          * XXX: the address may have embedded scope zone ID, which should be
1991          * hidden from applications.
1992          */
1993         bzero(&fromsa, sizeof(fromsa));
1994         fromsa.sin6_family = AF_INET6;
1995         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1996         fromsa.sin6_addr = ip6->ip6_src;
1997         if (sa6_recoverscope(&fromsa)) {
1998                 m_freem(m);
1999                 return (IPPROTO_DONE);
2000         }
2001
2002         INP_INFO_RLOCK(&V_ripcbinfo);
2003         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2004                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2005                         continue;
2006                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2007                         continue;
2008                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2009                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2010                         continue;
2011                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2012                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2013                         continue;
2014                 INP_RLOCK(in6p);
2015                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2016                     in6p->in6p_icmp6filt)) {
2017                         INP_RUNLOCK(in6p);
2018                         continue;
2019                 }
2020                 if (last != NULL) {
2021                         struct  mbuf *n = NULL;
2022
2023                         /*
2024                          * Recent network drivers tend to allocate a single
2025                          * mbuf cluster, rather than to make a couple of
2026                          * mbufs without clusters.  Also, since the IPv6 code
2027                          * path tries to avoid m_pullup(), it is highly
2028                          * probable that we still have an mbuf cluster here
2029                          * even though the necessary length can be stored in an
2030                          * mbuf's internal buffer.
2031                          * Meanwhile, the default size of the receive socket
2032                          * buffer for raw sockets is not so large.  This means
2033                          * the possibility of packet loss is relatively higher
2034                          * than before.  To avoid this scenario, we copy the
2035                          * received data to a separate mbuf that does not use
2036                          * a cluster, if possible.
2037                          * XXX: it is better to copy the data after stripping
2038                          * intermediate headers.
2039                          */
2040                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2041                             m->m_len <= MHLEN) {
2042                                 n = m_get(M_NOWAIT, m->m_type);
2043                                 if (n != NULL) {
2044                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2045                                                 bcopy(m->m_data, n->m_data,
2046                                                       m->m_len);
2047                                                 n->m_len = m->m_len;
2048                                         } else {
2049                                                 m_free(n);
2050                                                 n = NULL;
2051                                         }
2052                                 }
2053                         }
2054                         if (n != NULL ||
2055                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2056                                 if (last->inp_flags & INP_CONTROLOPTS)
2057                                         ip6_savecontrol(last, n, &opts);
2058                                 /* strip intermediate headers */
2059                                 m_adj(n, off);
2060                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2061                                 if (sbappendaddr_locked(
2062                                     &last->inp_socket->so_rcv,
2063                                     (struct sockaddr *)&fromsa, n, opts)
2064                                     == 0) {
2065                                         /* should notify about lost packet */
2066                                         m_freem(n);
2067                                         if (opts) {
2068                                                 m_freem(opts);
2069                                         }
2070                                         SOCKBUF_UNLOCK(
2071                                             &last->inp_socket->so_rcv);
2072                                 } else
2073                                         sorwakeup_locked(last->inp_socket);
2074                                 opts = NULL;
2075                         }
2076                         INP_RUNLOCK(last);
2077                 }
2078                 last = in6p;
2079         }
2080         INP_INFO_RUNLOCK(&V_ripcbinfo);
2081         if (last != NULL) {
2082                 if (last->inp_flags & INP_CONTROLOPTS)
2083                         ip6_savecontrol(last, m, &opts);
2084                 /* strip intermediate headers */
2085                 m_adj(m, off);
2086
2087                 /* avoid using mbuf clusters if possible (see above) */
2088                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2089                     m->m_len <= MHLEN) {
2090                         struct mbuf *n;
2091
2092                         n = m_get(M_NOWAIT, m->m_type);
2093                         if (n != NULL) {
2094                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2095                                         bcopy(m->m_data, n->m_data, m->m_len);
2096                                         n->m_len = m->m_len;
2097
2098                                         m_freem(m);
2099                                         m = n;
2100                                 } else {
2101                                         m_freem(n);
2102                                         n = NULL;
2103                                 }
2104                         }
2105                 }
2106                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2107                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2108                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2109                         m_freem(m);
2110                         if (opts)
2111                                 m_freem(opts);
2112                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2113                 } else
2114                         sorwakeup_locked(last->inp_socket);
2115                 INP_RUNLOCK(last);
2116         } else {
2117                 m_freem(m);
2118                 IP6STAT_DEC(ip6s_delivered);
2119         }
2120         return IPPROTO_DONE;
2121 }
2122
2123 /*
2124  * Reflect the ip6 packet back to the source.
2125  * OFF points to the icmp6 header, counted from the top of the mbuf.
2126  */
2127 void
2128 icmp6_reflect(struct mbuf *m, size_t off)
2129 {
2130         struct ip6_hdr *ip6;
2131         struct icmp6_hdr *icmp6;
2132         struct in6_ifaddr *ia = NULL;
2133         int plen;
2134         int type, code;
2135         struct ifnet *outif = NULL;
2136         struct in6_addr origdst, src, *srcp = NULL;
2137
2138         /* too short to reflect */
2139         if (off < sizeof(struct ip6_hdr)) {
2140                 nd6log((LOG_DEBUG,
2141                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2142                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2143                     __FILE__, __LINE__));
2144                 goto bad;
2145         }
2146
2147         /*
2148          * If there are extra headers between IPv6 and ICMPv6, strip
2149          * off that header first.
2150          */
2151 #ifdef DIAGNOSTIC
2152         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2153                 panic("assumption failed in icmp6_reflect");
2154 #endif
2155         if (off > sizeof(struct ip6_hdr)) {
2156                 size_t l;
2157                 struct ip6_hdr nip6;
2158
2159                 l = off - sizeof(struct ip6_hdr);
2160                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2161                 m_adj(m, l);
2162                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2163                 if (m->m_len < l) {
2164                         if ((m = m_pullup(m, l)) == NULL)
2165                                 return;
2166                 }
2167                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2168         } else /* off == sizeof(struct ip6_hdr) */ {
2169                 size_t l;
2170                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2171                 if (m->m_len < l) {
2172                         if ((m = m_pullup(m, l)) == NULL)
2173                                 return;
2174                 }
2175         }
2176         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2177         ip6 = mtod(m, struct ip6_hdr *);
2178         ip6->ip6_nxt = IPPROTO_ICMPV6;
2179         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2180         type = icmp6->icmp6_type; /* keep type for statistics */
2181         code = icmp6->icmp6_code; /* ditto. */
2182
2183         origdst = ip6->ip6_dst;
2184         /*
2185          * ip6_input() drops a packet if its src is multicast.
2186          * So, the src is never multicast.
2187          */
2188         ip6->ip6_dst = ip6->ip6_src;
2189
2190         /*
2191          * If the incoming packet was addressed directly to us (i.e. unicast),
2192          * use dst as the src for the reply.
2193          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2194          * (for example) when we encounter an error while forwarding procedure
2195          * destined to a duplicated address of ours.
2196          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2197          * procedure of an outgoing packet of our own, in which case we need
2198          * to search in the ifaddr list.
2199          */
2200         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2201                 if ((ia = ip6_getdstifaddr(m))) {
2202                         if (!(ia->ia6_flags &
2203                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2204                                 srcp = &ia->ia_addr.sin6_addr;
2205                 } else {
2206                         struct sockaddr_in6 d;
2207
2208                         bzero(&d, sizeof(d));
2209                         d.sin6_family = AF_INET6;
2210                         d.sin6_len = sizeof(d);
2211                         d.sin6_addr = origdst;
2212                         ia = (struct in6_ifaddr *)
2213                             ifa_ifwithaddr((struct sockaddr *)&d);
2214                         if (ia &&
2215                             !(ia->ia6_flags &
2216                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2217                                 srcp = &ia->ia_addr.sin6_addr;
2218                         }
2219                 }
2220         }
2221
2222         if (srcp == NULL) {
2223                 int e;
2224                 struct sockaddr_in6 sin6;
2225                 struct route_in6 ro;
2226
2227                 /*
2228                  * This case matches to multicasts, our anycast, or unicasts
2229                  * that we do not own.  Select a source address based on the
2230                  * source address of the erroneous packet.
2231                  */
2232                 bzero(&sin6, sizeof(sin6));
2233                 sin6.sin6_family = AF_INET6;
2234                 sin6.sin6_len = sizeof(sin6);
2235                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2236
2237                 bzero(&ro, sizeof(ro));
2238                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2239                 if (ro.ro_rt)
2240                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2241                 if (e) {
2242                         char ip6buf[INET6_ADDRSTRLEN];
2243                         nd6log((LOG_DEBUG,
2244                             "icmp6_reflect: source can't be determined: "
2245                             "dst=%s, error=%d\n",
2246                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2247                         goto bad;
2248                 }
2249                 srcp = &src;
2250         }
2251
2252         ip6->ip6_src = *srcp;
2253         ip6->ip6_flow = 0;
2254         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2255         ip6->ip6_vfc |= IPV6_VERSION;
2256         ip6->ip6_nxt = IPPROTO_ICMPV6;
2257         if (outif)
2258                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2259         else if (m->m_pkthdr.rcvif) {
2260                 /* XXX: This may not be the outgoing interface */
2261                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2262         } else
2263                 ip6->ip6_hlim = V_ip6_defhlim;
2264
2265         icmp6->icmp6_cksum = 0;
2266         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2267             sizeof(struct ip6_hdr), plen);
2268
2269         /*
2270          * XXX option handling
2271          */
2272
2273         m->m_flags &= ~(M_BCAST|M_MCAST);
2274
2275         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2276         if (outif)
2277                 icmp6_ifoutstat_inc(outif, type, code);
2278
2279         if (ia != NULL)
2280                 ifa_free(&ia->ia_ifa);
2281         return;
2282
2283  bad:
2284         if (ia != NULL)
2285                 ifa_free(&ia->ia_ifa);
2286         m_freem(m);
2287         return;
2288 }
2289
2290 void
2291 icmp6_fasttimo(void)
2292 {
2293
2294         mld_fasttimo();
2295 }
2296
2297 void
2298 icmp6_slowtimo(void)
2299 {
2300
2301         mld_slowtimo();
2302 }
2303
2304 static const char *
2305 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2306     struct in6_addr *tgt6)
2307 {
2308         static char buf[1024];
2309         char ip6bufs[INET6_ADDRSTRLEN];
2310         char ip6bufd[INET6_ADDRSTRLEN];
2311         char ip6buft[INET6_ADDRSTRLEN];
2312         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2313             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2314             ip6_sprintf(ip6buft, tgt6));
2315         return buf;
2316 }
2317
2318 void
2319 icmp6_redirect_input(struct mbuf *m, int off)
2320 {
2321         struct ifnet *ifp;
2322         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2323         struct nd_redirect *nd_rd;
2324         int icmp6len = ntohs(ip6->ip6_plen);
2325         char *lladdr = NULL;
2326         int lladdrlen = 0;
2327         struct rtentry *rt = NULL;
2328         int is_router;
2329         int is_onlink;
2330         struct in6_addr src6 = ip6->ip6_src;
2331         struct in6_addr redtgt6;
2332         struct in6_addr reddst6;
2333         union nd_opts ndopts;
2334         char ip6buf[INET6_ADDRSTRLEN];
2335
2336         M_ASSERTPKTHDR(m);
2337         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2338
2339         ifp = m->m_pkthdr.rcvif;
2340
2341         /* XXX if we are router, we don't update route by icmp6 redirect */
2342         if (V_ip6_forwarding)
2343                 goto freeit;
2344         if (!V_icmp6_rediraccept)
2345                 goto freeit;
2346
2347 #ifndef PULLDOWN_TEST
2348         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2349         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2350 #else
2351         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2352         if (nd_rd == NULL) {
2353                 ICMP6STAT_INC(icp6s_tooshort);
2354                 return;
2355         }
2356 #endif
2357         redtgt6 = nd_rd->nd_rd_target;
2358         reddst6 = nd_rd->nd_rd_dst;
2359
2360         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2361             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2362                 goto freeit;
2363         }
2364
2365         /* validation */
2366         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2367                 nd6log((LOG_ERR,
2368                     "ICMP6 redirect sent from %s rejected; "
2369                     "must be from linklocal\n",
2370                     ip6_sprintf(ip6buf, &src6)));
2371                 goto bad;
2372         }
2373         if (ip6->ip6_hlim != 255) {
2374                 nd6log((LOG_ERR,
2375                     "ICMP6 redirect sent from %s rejected; "
2376                     "hlim=%d (must be 255)\n",
2377                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2378                 goto bad;
2379         }
2380     {
2381         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2382         struct sockaddr_in6 sin6;
2383         struct in6_addr *gw6;
2384
2385         bzero(&sin6, sizeof(sin6));
2386         sin6.sin6_family = AF_INET6;
2387         sin6.sin6_len = sizeof(struct sockaddr_in6);
2388         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2389         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2390         if (rt) {
2391                 if (rt->rt_gateway == NULL ||
2392                     rt->rt_gateway->sa_family != AF_INET6) {
2393                         RTFREE_LOCKED(rt);
2394                         nd6log((LOG_ERR,
2395                             "ICMP6 redirect rejected; no route "
2396                             "with inet6 gateway found for redirect dst: %s\n",
2397                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2398                         goto bad;
2399                 }
2400
2401                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2402                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2403                         RTFREE_LOCKED(rt);
2404                         nd6log((LOG_ERR,
2405                             "ICMP6 redirect rejected; "
2406                             "not equal to gw-for-src=%s (must be same): "
2407                             "%s\n",
2408                             ip6_sprintf(ip6buf, gw6),
2409                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2410                         goto bad;
2411                 }
2412         } else {
2413                 nd6log((LOG_ERR,
2414                     "ICMP6 redirect rejected; "
2415                     "no route found for redirect dst: %s\n",
2416                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2417                 goto bad;
2418         }
2419         RTFREE_LOCKED(rt);
2420         rt = NULL;
2421     }
2422         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2423                 nd6log((LOG_ERR,
2424                     "ICMP6 redirect rejected; "
2425                     "redirect dst must be unicast: %s\n",
2426                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2427                 goto bad;
2428         }
2429
2430         is_router = is_onlink = 0;
2431         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2432                 is_router = 1;  /* router case */
2433         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2434                 is_onlink = 1;  /* on-link destination case */
2435         if (!is_router && !is_onlink) {
2436                 nd6log((LOG_ERR,
2437                     "ICMP6 redirect rejected; "
2438                     "neither router case nor onlink case: %s\n",
2439                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2440                 goto bad;
2441         }
2442         /* validation passed */
2443
2444         icmp6len -= sizeof(*nd_rd);
2445         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2446         if (nd6_options(&ndopts) < 0) {
2447                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2448                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2449                 /* nd6_options have incremented stats */
2450                 goto freeit;
2451         }
2452
2453         if (ndopts.nd_opts_tgt_lladdr) {
2454                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2455                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2456         }
2457
2458         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2459                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2460                     "(if %d, icmp6 packet %d): %s\n",
2461                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2462                     ifp->if_addrlen, lladdrlen - 2,
2463                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2464                 goto bad;
2465         }
2466
2467         /* RFC 2461 8.3 */
2468         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2469             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2470
2471         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2472                 /* perform rtredirect */
2473                 struct sockaddr_in6 sdst;
2474                 struct sockaddr_in6 sgw;
2475                 struct sockaddr_in6 ssrc;
2476                 u_int fibnum;
2477
2478                 bzero(&sdst, sizeof(sdst));
2479                 bzero(&sgw, sizeof(sgw));
2480                 bzero(&ssrc, sizeof(ssrc));
2481                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2482                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2483                         sizeof(struct sockaddr_in6);
2484                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2485                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2486                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2487                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2488                         in6_rtredirect((struct sockaddr *)&sdst,
2489                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2490                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2491                             fibnum);
2492         }
2493         /* finally update cached route in each socket via pfctlinput */
2494     {
2495         struct sockaddr_in6 sdst;
2496
2497         bzero(&sdst, sizeof(sdst));
2498         sdst.sin6_family = AF_INET6;
2499         sdst.sin6_len = sizeof(struct sockaddr_in6);
2500         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2501         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2502 #ifdef IPSEC
2503         key_sa_routechange((struct sockaddr *)&sdst);
2504 #endif /* IPSEC */
2505     }
2506
2507  freeit:
2508         m_freem(m);
2509         return;
2510
2511  bad:
2512         ICMP6STAT_INC(icp6s_badredirect);
2513         m_freem(m);
2514 }
2515
2516 void
2517 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2518 {
2519         struct ifnet *ifp;      /* my outgoing interface */
2520         struct in6_addr *ifp_ll6;
2521         struct in6_addr *router_ll6;
2522         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2523         struct mbuf *m = NULL;  /* newly allocated one */
2524         struct m_tag *mtag;
2525         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2526         struct nd_redirect *nd_rd;
2527         struct llentry *ln = NULL;
2528         size_t maxlen;
2529         u_char *p;
2530         struct ifnet *outif = NULL;
2531         struct sockaddr_in6 src_sa;
2532
2533         icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2534
2535         /* if we are not router, we don't send icmp6 redirect */
2536         if (!V_ip6_forwarding)
2537                 goto fail;
2538
2539         /* sanity check */
2540         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2541                 goto fail;
2542
2543         /*
2544          * Address check:
2545          *  the source address must identify a neighbor, and
2546          *  the destination address must not be a multicast address
2547          *  [RFC 2461, sec 8.2]
2548          */
2549         sip6 = mtod(m0, struct ip6_hdr *);
2550         bzero(&src_sa, sizeof(src_sa));
2551         src_sa.sin6_family = AF_INET6;
2552         src_sa.sin6_len = sizeof(src_sa);
2553         src_sa.sin6_addr = sip6->ip6_src;
2554         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2555                 goto fail;
2556         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2557                 goto fail;      /* what should we do here? */
2558
2559         /* rate limit */
2560         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2561                 goto fail;
2562
2563         /*
2564          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2565          * we almost always ask for an mbuf cluster for simplicity.
2566          * (MHLEN < IPV6_MMTU is almost always true)
2567          */
2568 #if IPV6_MMTU >= MCLBYTES
2569 # error assumption failed about IPV6_MMTU and MCLBYTES
2570 #endif
2571         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2572         if (m == NULL)
2573                 goto fail;
2574         M_SETFIB(m, rt->rt_fibnum);
2575         maxlen = M_TRAILINGSPACE(m);
2576         maxlen = min(IPV6_MMTU, maxlen);
2577         /* just for safety */
2578         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2579             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2580                 goto fail;
2581         }
2582
2583         {
2584                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2585                 struct in6_ifaddr *ia;
2586                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2587                                                  IN6_IFF_NOTREADY|
2588                                                  IN6_IFF_ANYCAST)) == NULL)
2589                         goto fail;
2590                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2591                 /* XXXRW: reference released prematurely. */
2592                 ifa_free(&ia->ia_ifa);
2593         }
2594
2595         /* get ip6 linklocal address for the router. */
2596         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2597                 struct sockaddr_in6 *sin6;
2598                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2599                 router_ll6 = &sin6->sin6_addr;
2600                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2601                         router_ll6 = (struct in6_addr *)NULL;
2602         } else
2603                 router_ll6 = (struct in6_addr *)NULL;
2604
2605         /* ip6 */
2606         ip6 = mtod(m, struct ip6_hdr *);
2607         ip6->ip6_flow = 0;
2608         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2609         ip6->ip6_vfc |= IPV6_VERSION;
2610         /* ip6->ip6_plen will be set later */
2611         ip6->ip6_nxt = IPPROTO_ICMPV6;
2612         ip6->ip6_hlim = 255;
2613         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2614         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2615         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2616
2617         /* ND Redirect */
2618         nd_rd = (struct nd_redirect *)(ip6 + 1);
2619         nd_rd->nd_rd_type = ND_REDIRECT;
2620         nd_rd->nd_rd_code = 0;
2621         nd_rd->nd_rd_reserved = 0;
2622         if (rt->rt_flags & RTF_GATEWAY) {
2623                 /*
2624                  * nd_rd->nd_rd_target must be a link-local address in
2625                  * better router cases.
2626                  */
2627                 if (!router_ll6)
2628                         goto fail;
2629                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2630                     sizeof(nd_rd->nd_rd_target));
2631                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2632                     sizeof(nd_rd->nd_rd_dst));
2633         } else {
2634                 /* make sure redtgt == reddst */
2635                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2636                     sizeof(nd_rd->nd_rd_target));
2637                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2638                     sizeof(nd_rd->nd_rd_dst));
2639         }
2640
2641         p = (u_char *)(nd_rd + 1);
2642
2643         if (!router_ll6)
2644                 goto nolladdropt;
2645
2646         {
2647                 /* target lladdr option */
2648                 int len;
2649                 struct nd_opt_hdr *nd_opt;
2650                 char *lladdr;
2651
2652                 IF_AFDATA_RLOCK(ifp);
2653                 ln = nd6_lookup(router_ll6, 0, ifp);
2654                 IF_AFDATA_RUNLOCK(ifp);
2655                 if (ln == NULL)
2656                         goto nolladdropt;
2657
2658                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2659                 len = (len + 7) & ~7;   /* round by 8 */
2660                 /* safety check */
2661                 if (len + (p - (u_char *)ip6) > maxlen)                         
2662                         goto nolladdropt;
2663
2664                 if (ln->la_flags & LLE_VALID) {
2665                         nd_opt = (struct nd_opt_hdr *)p;
2666                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2667                         nd_opt->nd_opt_len = len >> 3;
2668                         lladdr = (char *)(nd_opt + 1);
2669                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2670                         p += len;
2671                 }
2672         }
2673 nolladdropt:
2674         if (ln != NULL)
2675                 LLE_RUNLOCK(ln);
2676                 
2677         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2678
2679         /* just to be safe */
2680 #ifdef M_DECRYPTED      /*not openbsd*/
2681         if (m0->m_flags & M_DECRYPTED)
2682                 goto noredhdropt;
2683 #endif
2684         if (p - (u_char *)ip6 > maxlen)
2685                 goto noredhdropt;
2686
2687         {
2688                 /* redirected header option */
2689                 int len;
2690                 struct nd_opt_rd_hdr *nd_opt_rh;
2691
2692                 /*
2693                  * compute the maximum size for icmp6 redirect header option.
2694                  * XXX room for auth header?
2695                  */
2696                 len = maxlen - (p - (u_char *)ip6);
2697                 len &= ~7;
2698
2699                 /* This is just for simplicity. */
2700                 if (m0->m_pkthdr.len != m0->m_len) {
2701                         if (m0->m_next) {
2702                                 m_freem(m0->m_next);
2703                                 m0->m_next = NULL;
2704                         }
2705                         m0->m_pkthdr.len = m0->m_len;
2706                 }
2707
2708                 /*
2709                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2710                  * about padding/truncate rule for the original IP packet.
2711                  * From the discussion on IPv6imp in Feb 1999,
2712                  * the consensus was:
2713                  * - "attach as much as possible" is the goal
2714                  * - pad if not aligned (original size can be guessed by
2715                  *   original ip6 header)
2716                  * Following code adds the padding if it is simple enough,
2717                  * and truncates if not.
2718                  */
2719                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2720                         panic("assumption failed in %s:%d", __FILE__,
2721                             __LINE__);
2722
2723                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2724                         /* not enough room, truncate */
2725                         m0->m_pkthdr.len = m0->m_len = len -
2726                             sizeof(*nd_opt_rh);
2727                 } else {
2728                         /* enough room, pad or truncate */
2729                         size_t extra;
2730
2731                         extra = m0->m_pkthdr.len % 8;
2732                         if (extra) {
2733                                 /* pad if easy enough, truncate if not */
2734                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2735                                         /* pad */
2736                                         m0->m_len += (8 - extra);
2737                                         m0->m_pkthdr.len += (8 - extra);
2738                                 } else {
2739                                         /* truncate */
2740                                         m0->m_pkthdr.len -= extra;
2741                                         m0->m_len -= extra;
2742                                 }
2743                         }
2744                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2745                         m0->m_pkthdr.len = m0->m_len = len -
2746                             sizeof(*nd_opt_rh);
2747                 }
2748
2749                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2750                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2751                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2752                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2753                 p += sizeof(*nd_opt_rh);
2754                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2755
2756                 /* connect m0 to m */
2757                 m_tag_delete_chain(m0, NULL);
2758                 m0->m_flags &= ~M_PKTHDR;
2759                 m->m_next = m0;
2760                 m->m_pkthdr.len = m->m_len + m0->m_len;
2761                 m0 = NULL;
2762         }
2763 noredhdropt:;
2764         if (m0) {
2765                 m_freem(m0);
2766                 m0 = NULL;
2767         }
2768
2769         /* XXX: clear embedded link IDs in the inner header */
2770         in6_clearscope(&sip6->ip6_src);
2771         in6_clearscope(&sip6->ip6_dst);
2772         in6_clearscope(&nd_rd->nd_rd_target);
2773         in6_clearscope(&nd_rd->nd_rd_dst);
2774
2775         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2776
2777         nd_rd->nd_rd_cksum = 0;
2778         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2779             sizeof(*ip6), ntohs(ip6->ip6_plen));
2780
2781         if (send_sendso_input_hook != NULL) {
2782                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2783                         M_NOWAIT);
2784                 if (mtag == NULL)
2785                         goto fail;
2786                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2787                 m_tag_prepend(m, mtag);
2788         }
2789
2790         /* send the packet to outside... */
2791         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2792         if (outif) {
2793                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2794                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2795         }
2796         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2797
2798         return;
2799
2800 fail:
2801         if (m)
2802                 m_freem(m);
2803         if (m0)
2804                 m_freem(m0);
2805 }
2806
2807 /*
2808  * ICMPv6 socket option processing.
2809  */
2810 int
2811 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2812 {
2813         int error = 0;
2814         int optlen;
2815         struct inpcb *inp = sotoinpcb(so);
2816         int level, op, optname;
2817
2818         if (sopt) {
2819                 level = sopt->sopt_level;
2820                 op = sopt->sopt_dir;
2821                 optname = sopt->sopt_name;
2822                 optlen = sopt->sopt_valsize;
2823         } else
2824                 level = op = optname = optlen = 0;
2825
2826         if (level != IPPROTO_ICMPV6) {
2827                 return EINVAL;
2828         }
2829
2830         switch (op) {
2831         case PRCO_SETOPT:
2832                 switch (optname) {
2833                 case ICMP6_FILTER:
2834                     {
2835                         struct icmp6_filter ic6f;
2836
2837                         if (optlen != sizeof(ic6f)) {
2838                                 error = EMSGSIZE;
2839                                 break;
2840                         }
2841                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2842                         if (error == 0) {
2843                                 INP_WLOCK(inp);
2844                                 *inp->in6p_icmp6filt = ic6f;
2845                                 INP_WUNLOCK(inp);
2846                         }
2847                         break;
2848                     }
2849
2850                 default:
2851                         error = ENOPROTOOPT;
2852                         break;
2853                 }
2854                 break;
2855
2856         case PRCO_GETOPT:
2857                 switch (optname) {
2858                 case ICMP6_FILTER:
2859                     {
2860                         struct icmp6_filter ic6f;
2861
2862                         INP_RLOCK(inp);
2863                         ic6f = *inp->in6p_icmp6filt;
2864                         INP_RUNLOCK(inp);
2865                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2866                         break;
2867                     }
2868
2869                 default:
2870                         error = ENOPROTOOPT;
2871                         break;
2872                 }
2873                 break;
2874         }
2875
2876         return (error);
2877 }
2878
2879 /*
2880  * Perform rate limit check.
2881  * Returns 0 if it is okay to send the icmp6 packet.
2882  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2883  * limitation.
2884  *
2885  * XXX per-destination/type check necessary?
2886  *
2887  * dst - not used at this moment
2888  * type - not used at this moment
2889  * code - not used at this moment
2890  */
2891 static int
2892 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2893     const int code)
2894 {
2895         int ret;
2896
2897         ret = 0;        /* okay to send */
2898
2899         /* PPS limit */
2900         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2901             V_icmp6errppslim)) {
2902                 /* The packet is subject to rate limit */
2903                 ret++;
2904         }
2905
2906         return ret;
2907 }