proc->thread stage 6: kernel threads now create processless LWKT threads.
[dragonfly.git] / sys / netinet6 / icmp6.c
1 /*      $FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $     */
2 /*      $DragonFly: src/sys/netinet6/icmp6.c,v 1.2 2003/06/17 04:28:51 dillon Exp $     */
3 /*      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */
4
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *      The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *      This product includes software developed by the University of
49  *      California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
67  */
68
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/in6_ifattach.h>
100 #include <netinet6/ip6protosw.h>
101
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netkey/key.h>
105 #endif
106
107 #ifdef FAST_IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/key.h>
110 #define IPSEC
111 #endif
112
113 #include <net/net_osdep.h>
114
115 #ifdef HAVE_NRL_INPCB
116 /* inpcb members */
117 #define in6pcb          inpcb
118 #define in6p_laddr      inp_laddr6
119 #define in6p_faddr      inp_faddr6
120 #define in6p_icmp6filt  inp_icmp6filt
121 #define in6p_route      inp_route
122 #define in6p_socket     inp_socket
123 #define in6p_flags      inp_flags
124 #define in6p_moptions   inp_moptions6
125 #define in6p_outputopts inp_outputopts6
126 #define in6p_ip6        inp_ipv6
127 #define in6p_flowinfo   inp_flowinfo
128 #define in6p_sp         inp_sp
129 #define in6p_next       inp_next
130 #define in6p_prev       inp_prev
131 /* macro names */
132 #define sotoin6pcb      sotoinpcb
133 /* function names */
134 #define in6_pcbdetach   in_pcbdetach
135 #define in6_rtchange    in_rtchange
136
137 /*
138  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
139  * others...
140  */
141 #define in6p_ip6_nxt    inp_ipv6.ip6_nxt
142 #endif
143
144 extern struct domain inet6domain;
145 extern struct ip6protosw inet6sw[];
146 extern u_char ip6_protox[];
147
148 struct icmp6stat icmp6stat;
149
150 extern struct inpcbhead ripcb;
151 extern int icmp6errppslim;
152 static int icmp6errpps_count = 0;
153 static struct timeval icmp6errppslim_last;
154 extern int icmp6_nodeinfo;
155
156 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
157 static int icmp6_rip6_input __P((struct mbuf **, int));
158 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
159 static const char *icmp6_redirect_diag __P((struct in6_addr *,
160         struct in6_addr *, struct in6_addr *));
161 #ifndef HAVE_PPSRATECHECK
162 static int ppsratecheck __P((struct timeval *, int *, int));
163 #endif
164 static struct mbuf *ni6_input __P((struct mbuf *, int));
165 static struct mbuf *ni6_nametodns __P((const char *, int, int));
166 static int ni6_dnsmatch __P((const char *, int, const char *, int));
167 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
168                           struct ifnet **, char *));
169 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
170                                 struct ifnet *, int));
171 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
172
173 #ifdef COMPAT_RFC1885
174 static struct route_in6 icmp6_reflect_rt;
175 #endif
176
177
178 void
179 icmp6_init()
180 {
181         mld6_init();
182 }
183
184 static void
185 icmp6_errcount(stat, type, code)
186         struct icmp6errstat *stat;
187         int type, code;
188 {
189         switch (type) {
190         case ICMP6_DST_UNREACH:
191                 switch (code) {
192                 case ICMP6_DST_UNREACH_NOROUTE:
193                         stat->icp6errs_dst_unreach_noroute++;
194                         return;
195                 case ICMP6_DST_UNREACH_ADMIN:
196                         stat->icp6errs_dst_unreach_admin++;
197                         return;
198                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
199                         stat->icp6errs_dst_unreach_beyondscope++;
200                         return;
201                 case ICMP6_DST_UNREACH_ADDR:
202                         stat->icp6errs_dst_unreach_addr++;
203                         return;
204                 case ICMP6_DST_UNREACH_NOPORT:
205                         stat->icp6errs_dst_unreach_noport++;
206                         return;
207                 }
208                 break;
209         case ICMP6_PACKET_TOO_BIG:
210                 stat->icp6errs_packet_too_big++;
211                 return;
212         case ICMP6_TIME_EXCEEDED:
213                 switch (code) {
214                 case ICMP6_TIME_EXCEED_TRANSIT:
215                         stat->icp6errs_time_exceed_transit++;
216                         return;
217                 case ICMP6_TIME_EXCEED_REASSEMBLY:
218                         stat->icp6errs_time_exceed_reassembly++;
219                         return;
220                 }
221                 break;
222         case ICMP6_PARAM_PROB:
223                 switch (code) {
224                 case ICMP6_PARAMPROB_HEADER:
225                         stat->icp6errs_paramprob_header++;
226                         return;
227                 case ICMP6_PARAMPROB_NEXTHEADER:
228                         stat->icp6errs_paramprob_nextheader++;
229                         return;
230                 case ICMP6_PARAMPROB_OPTION:
231                         stat->icp6errs_paramprob_option++;
232                         return;
233                 }
234                 break;
235         case ND_REDIRECT:
236                 stat->icp6errs_redirect++;
237                 return;
238         }
239         stat->icp6errs_unknown++;
240 }
241
242 /*
243  * Generate an error packet of type error in response to bad IP6 packet.
244  */
245 void
246 icmp6_error(m, type, code, param)
247         struct mbuf *m;
248         int type, code, param;
249 {
250         struct ip6_hdr *oip6, *nip6;
251         struct icmp6_hdr *icmp6;
252         u_int preplen;
253         int off;
254         int nxt;
255
256         icmp6stat.icp6s_error++;
257
258         /* count per-type-code statistics */
259         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
260
261 #ifdef M_DECRYPTED      /*not openbsd*/
262         if (m->m_flags & M_DECRYPTED) {
263                 icmp6stat.icp6s_canterror++;
264                 goto freeit;
265         }
266 #endif
267
268 #ifndef PULLDOWN_TEST
269         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
270 #else
271         if (m->m_len < sizeof(struct ip6_hdr)) {
272                 m = m_pullup(m, sizeof(struct ip6_hdr));
273                 if (m == NULL)
274                         return;
275         }
276 #endif
277         oip6 = mtod(m, struct ip6_hdr *);
278
279         /*
280          * Multicast destination check. For unrecognized option errors,
281          * this check has already done in ip6_unknown_opt(), so we can
282          * check only for other errors.
283          */
284         if ((m->m_flags & (M_BCAST|M_MCAST) ||
285              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
286             (type != ICMP6_PACKET_TOO_BIG &&
287              (type != ICMP6_PARAM_PROB ||
288               code != ICMP6_PARAMPROB_OPTION)))
289                 goto freeit;
290
291         /* Source address check. XXX: the case of anycast source? */
292         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
293             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
294                 goto freeit;
295
296         /*
297          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
298          * don't do it.
299          */
300         nxt = -1;
301         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
302         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
303                 struct icmp6_hdr *icp;
304
305 #ifndef PULLDOWN_TEST
306                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
307                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
308 #else
309                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
310                         sizeof(*icp));
311                 if (icp == NULL) {
312                         icmp6stat.icp6s_tooshort++;
313                         return;
314                 }
315 #endif
316                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
317                     icp->icmp6_type == ND_REDIRECT) {
318                         /*
319                          * ICMPv6 error
320                          * Special case: for redirect (which is
321                          * informational) we must not send icmp6 error.
322                          */
323                         icmp6stat.icp6s_canterror++;
324                         goto freeit;
325                 } else {
326                         /* ICMPv6 informational - send the error */
327                 }
328         } else {
329                 /* non-ICMPv6 - send the error */
330         }
331
332         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
333
334         /* Finally, do rate limitation check. */
335         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
336                 icmp6stat.icp6s_toofreq++;
337                 goto freeit;
338         }
339
340         /*
341          * OK, ICMP6 can be generated.
342          */
343
344         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
345                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
346
347         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
348         M_PREPEND(m, preplen, M_DONTWAIT);
349         if (m && m->m_len < preplen)
350                 m = m_pullup(m, preplen);
351         if (m == NULL) {
352                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
353                 return;
354         }
355
356         nip6 = mtod(m, struct ip6_hdr *);
357         nip6->ip6_src  = oip6->ip6_src;
358         nip6->ip6_dst  = oip6->ip6_dst;
359
360         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
361                 oip6->ip6_src.s6_addr16[1] = 0;
362         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
363                 oip6->ip6_dst.s6_addr16[1] = 0;
364
365         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
366         icmp6->icmp6_type = type;
367         icmp6->icmp6_code = code;
368         icmp6->icmp6_pptr = htonl((u_int32_t)param);
369
370         /*
371          * icmp6_reflect() is designed to be in the input path.
372          * icmp6_error() can be called from both input and outut path,
373          * and if we are in output path rcvif could contain bogus value.
374          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
375          * information in ip header (nip6).
376          */
377         m->m_pkthdr.rcvif = NULL;
378
379         icmp6stat.icp6s_outhist[type]++;
380         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
381
382         return;
383
384   freeit:
385         /*
386          * If we can't tell wheter or not we can generate ICMP6, free it.
387          */
388         m_freem(m);
389 }
390
391 /*
392  * Process a received ICMP6 message.
393  */
394 int
395 icmp6_input(mp, offp, proto)
396         struct mbuf **mp;
397         int *offp, proto;
398 {
399         struct mbuf *m = *mp, *n;
400         struct ip6_hdr *ip6, *nip6;
401         struct icmp6_hdr *icmp6, *nicmp6;
402         int off = *offp;
403         int icmp6len = m->m_pkthdr.len - *offp;
404         int code, sum, noff;
405
406 #ifndef PULLDOWN_TEST
407         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
408         /* m might change if M_LOOP.  So, call mtod after this */
409 #endif
410
411         /*
412          * Locate icmp6 structure in mbuf, and check
413          * that not corrupted and of at least minimum length
414          */
415
416         ip6 = mtod(m, struct ip6_hdr *);
417         if (icmp6len < sizeof(struct icmp6_hdr)) {
418                 icmp6stat.icp6s_tooshort++;
419                 goto freeit;
420         }
421
422         /*
423          * calculate the checksum
424          */
425 #ifndef PULLDOWN_TEST
426         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
427 #else
428         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
429         if (icmp6 == NULL) {
430                 icmp6stat.icp6s_tooshort++;
431                 return IPPROTO_DONE;
432         }
433 #endif
434         code = icmp6->icmp6_code;
435
436         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
437                 nd6log((LOG_ERR,
438                     "ICMP6 checksum error(%d|%x) %s\n",
439                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
440                 icmp6stat.icp6s_checksum++;
441                 goto freeit;
442         }
443
444         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
445                 /*
446                  * Deliver very specific ICMP6 type only.
447                  * This is important to deilver TOOBIG.  Otherwise PMTUD
448                  * will not work.
449                  */
450                 switch (icmp6->icmp6_type) {
451                 case ICMP6_DST_UNREACH:
452                 case ICMP6_PACKET_TOO_BIG:
453                 case ICMP6_TIME_EXCEEDED:
454                         break;
455                 default:
456                         goto freeit;
457                 }
458         }
459
460         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
461         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
462         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
463                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
464
465         switch (icmp6->icmp6_type) {
466         case ICMP6_DST_UNREACH:
467                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
468                 switch (code) {
469                 case ICMP6_DST_UNREACH_NOROUTE:
470                         code = PRC_UNREACH_NET;
471                         break;
472                 case ICMP6_DST_UNREACH_ADMIN:
473                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
474                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
475                         break;
476                 case ICMP6_DST_UNREACH_ADDR:
477                         code = PRC_HOSTDEAD;
478                         break;
479 #ifdef COMPAT_RFC1885
480                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
481                         code = PRC_UNREACH_SRCFAIL;
482                         break;
483 #else
484                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
485                         /* I mean "source address was incorrect." */
486                         code = PRC_PARAMPROB;
487                         break;
488 #endif
489                 case ICMP6_DST_UNREACH_NOPORT:
490                         code = PRC_UNREACH_PORT;
491                         break;
492                 default:
493                         goto badcode;
494                 }
495                 goto deliver;
496                 break;
497
498         case ICMP6_PACKET_TOO_BIG:
499                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
500                 if (code != 0)
501                         goto badcode;
502
503                 code = PRC_MSGSIZE;
504
505                 /*
506                  * Updating the path MTU will be done after examining
507                  * intermediate extension headers.
508                  */
509                 goto deliver;
510                 break;
511
512         case ICMP6_TIME_EXCEEDED:
513                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
514                 switch (code) {
515                 case ICMP6_TIME_EXCEED_TRANSIT:
516                 case ICMP6_TIME_EXCEED_REASSEMBLY:
517                         code += PRC_TIMXCEED_INTRANS;
518                         break;
519                 default:
520                         goto badcode;
521                 }
522                 goto deliver;
523                 break;
524
525         case ICMP6_PARAM_PROB:
526                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
527                 switch (code) {
528                 case ICMP6_PARAMPROB_NEXTHEADER:
529                         code = PRC_UNREACH_PROTOCOL;
530                         break;
531                 case ICMP6_PARAMPROB_HEADER:
532                 case ICMP6_PARAMPROB_OPTION:
533                         code = PRC_PARAMPROB;
534                         break;
535                 default:
536                         goto badcode;
537                 }
538                 goto deliver;
539                 break;
540
541         case ICMP6_ECHO_REQUEST:
542                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
543                 if (code != 0)
544                         goto badcode;
545                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
546                         /* Give up remote */
547                         break;
548                 }
549                 if ((n->m_flags & M_EXT) != 0
550                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
551                         struct mbuf *n0 = n;
552                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
553                         int n0len;
554
555                         /*
556                          * Prepare an internal mbuf. m_pullup() doesn't
557                          * always copy the length we specified.
558                          */
559                         if (maxlen >= MCLBYTES) {
560                                 /* Give up remote */
561                                 m_freem(n0);
562                                 break;
563                         }
564                         MGETHDR(n, M_DONTWAIT, n0->m_type);
565                         n0len = n0->m_pkthdr.len;       /* save for use below */
566                         if (n)
567                                 M_MOVE_PKTHDR(n, n0);
568                         if (n && maxlen >= MHLEN) {
569                                 MCLGET(n, M_DONTWAIT);
570                                 if ((n->m_flags & M_EXT) == 0) {
571                                         m_free(n);
572                                         n = NULL;
573                                 }
574                         }
575                         if (n == NULL) {
576                                 /* Give up remote */
577                                 m_freem(n0);
578                                 break;
579                         }
580                         /*
581                          * Copy IPv6 and ICMPv6 only.
582                          */
583                         nip6 = mtod(n, struct ip6_hdr *);
584                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
585                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
586                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
587                         noff = sizeof(struct ip6_hdr);
588                         /* new mbuf contains only ipv6+icmpv6 headers */
589                         n->m_len = noff + sizeof(struct icmp6_hdr);
590                         /*
591                          * Adjust mbuf. ip6_plen will be adjusted in
592                          * ip6_output().
593                          */
594                         m_adj(n0, off + sizeof(struct icmp6_hdr));
595                         /* recalculate complete packet size */
596                         n->m_pkthdr.len = n0len + (noff - off);
597                         n->m_next = n0;
598                 } else {
599                         nip6 = mtod(n, struct ip6_hdr *);
600                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
601                         noff = off;
602                 }
603                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
604                 nicmp6->icmp6_code = 0;
605                 if (n) {
606                         icmp6stat.icp6s_reflect++;
607                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
608                         icmp6_reflect(n, noff);
609                 }
610                 break;
611
612         case ICMP6_ECHO_REPLY:
613                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
614                 if (code != 0)
615                         goto badcode;
616                 break;
617
618         case MLD_LISTENER_QUERY:
619         case MLD_LISTENER_REPORT:
620                 if (icmp6len < sizeof(struct mld_hdr))
621                         goto badlen;
622                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
623                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
624                 else
625                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
626                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
627                         /* give up local */
628                         mld6_input(m, off);
629                         m = NULL;
630                         goto freeit;
631                 }
632                 mld6_input(n, off);
633                 /* m stays. */
634                 break;
635
636         case MLD_LISTENER_DONE:
637                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
638                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
639                         goto badlen;
640                 break;          /* nothing to be done in kernel */
641
642         case MLD_MTRACE_RESP:
643         case MLD_MTRACE:
644                 /* XXX: these two are experimental.  not officially defind. */
645                 /* XXX: per-interface statistics? */
646                 break;          /* just pass it to applications */
647
648         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
649             {
650                 enum { WRU, FQDN } mode;
651
652                 if (!icmp6_nodeinfo)
653                         break;
654
655                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
656                         mode = WRU;
657                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
658                         mode = FQDN;
659                 else
660                         goto badlen;
661
662 #define hostnamelen     strlen(hostname)
663                 if (mode == FQDN) {
664 #ifndef PULLDOWN_TEST
665                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
666                                          IPPROTO_DONE);
667 #endif
668                         n = m_copy(m, 0, M_COPYALL);
669                         if (n)
670                                 n = ni6_input(n, off);
671                         /* XXX meaningless if n == NULL */
672                         noff = sizeof(struct ip6_hdr);
673                 } else {
674                         u_char *p;
675                         int maxlen, maxhlen;
676
677                         if ((icmp6_nodeinfo & 5) != 5) 
678                                 break;
679
680                         if (code != 0)
681                                 goto badcode;
682                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
683                         if (maxlen >= MCLBYTES) {
684                                 /* Give up remote */
685                                 break;
686                         }
687                         MGETHDR(n, M_DONTWAIT, m->m_type);
688                         if (n && maxlen > MHLEN) {
689                                 MCLGET(n, M_DONTWAIT);
690                                 if ((n->m_flags & M_EXT) == 0) {
691                                         m_free(n);
692                                         n = NULL;
693                                 }
694                         }
695                         if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
696                                 /*
697                                  * Previous code did a blind M_COPY_PKTHDR
698                                  * and said "just for rcvif".  If true, then
699                                  * we could tolerate the dup failing (due to
700                                  * the deep copy of the tag chain).  For now
701                                  * be conservative and just fail.
702                                  */
703                                 m_free(n);
704                                 n = NULL;
705                         }
706                         if (n == NULL) {
707                                 /* Give up remote */
708                                 break;
709                         }
710                         n->m_pkthdr.rcvif = NULL;
711                         n->m_len = 0;
712                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
713                         if (maxhlen > hostnamelen)
714                                 maxhlen = hostnamelen;
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                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
725                         noff = sizeof(struct ip6_hdr);
726                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
727                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
728                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
729                         nicmp6->icmp6_code = 0;
730                 }
731 #undef hostnamelen
732                 if (n) {
733                         icmp6stat.icp6s_reflect++;
734                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
735                         icmp6_reflect(n, noff);
736                 }
737                 break;
738             }
739
740         case ICMP6_WRUREPLY:
741                 if (code != 0)
742                         goto badcode;
743                 break;
744
745         case ND_ROUTER_SOLICIT:
746                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
747                 if (code != 0)
748                         goto badcode;
749                 if (icmp6len < sizeof(struct nd_router_solicit))
750                         goto badlen;
751                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
752                         /* give up local */
753                         nd6_rs_input(m, off, icmp6len);
754                         m = NULL;
755                         goto freeit;
756                 }
757                 nd6_rs_input(n, off, icmp6len);
758                 /* m stays. */
759                 break;
760
761         case ND_ROUTER_ADVERT:
762                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
763                 if (code != 0)
764                         goto badcode;
765                 if (icmp6len < sizeof(struct nd_router_advert))
766                         goto badlen;
767                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
768                         /* give up local */
769                         nd6_ra_input(m, off, icmp6len);
770                         m = NULL;
771                         goto freeit;
772                 }
773                 nd6_ra_input(n, off, icmp6len);
774                 /* m stays. */
775                 break;
776
777         case ND_NEIGHBOR_SOLICIT:
778                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
779                 if (code != 0)
780                         goto badcode;
781                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
782                         goto badlen;
783                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
784                         /* give up local */
785                         nd6_ns_input(m, off, icmp6len);
786                         m = NULL;
787                         goto freeit;
788                 }
789                 nd6_ns_input(n, off, icmp6len);
790                 /* m stays. */
791                 break;
792
793         case ND_NEIGHBOR_ADVERT:
794                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
795                 if (code != 0)
796                         goto badcode;
797                 if (icmp6len < sizeof(struct nd_neighbor_advert))
798                         goto badlen;
799                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800                         /* give up local */
801                         nd6_na_input(m, off, icmp6len);
802                         m = NULL;
803                         goto freeit;
804                 }
805                 nd6_na_input(n, off, icmp6len);
806                 /* m stays. */
807                 break;
808
809         case ND_REDIRECT:
810                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
811                 if (code != 0)
812                         goto badcode;
813                 if (icmp6len < sizeof(struct nd_redirect))
814                         goto badlen;
815                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
816                         /* give up local */
817                         icmp6_redirect_input(m, off);
818                         m = NULL;
819                         goto freeit;
820                 }
821                 icmp6_redirect_input(n, off);
822                 /* m stays. */
823                 break;
824
825         case ICMP6_ROUTER_RENUMBERING:
826                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
827                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
828                         goto badcode;
829                 if (icmp6len < sizeof(struct icmp6_router_renum))
830                         goto badlen;
831                 break;
832
833         default:
834                 nd6log((LOG_DEBUG,
835                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
836                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
837                     ip6_sprintf(&ip6->ip6_dst),
838                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
839                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
840                         /* ICMPv6 error: MUST deliver it by spec... */
841                         code = PRC_NCMDS;
842                         /* deliver */
843                 } else {
844                         /* ICMPv6 informational: MUST not deliver */
845                         break;
846                 }
847         deliver:
848                 if (icmp6_notify_error(m, off, icmp6len, code)) {
849                         /* In this case, m should've been freed. */
850                         return(IPPROTO_DONE);
851                 }
852                 break;
853
854         badcode:
855                 icmp6stat.icp6s_badcode++;
856                 break;
857
858         badlen:
859                 icmp6stat.icp6s_badlen++;
860                 break;
861         }
862
863         /* deliver the packet to appropriate sockets */
864         icmp6_rip6_input(&m, *offp);
865
866         return IPPROTO_DONE;
867
868  freeit:
869         m_freem(m);
870         return IPPROTO_DONE;
871 }
872
873 static int
874 icmp6_notify_error(m, off, icmp6len, code)
875         struct mbuf *m;
876         int off, icmp6len;
877 {
878         struct icmp6_hdr *icmp6;
879         struct ip6_hdr *eip6;
880         u_int32_t notifymtu;
881         struct sockaddr_in6 icmp6src, icmp6dst;
882
883         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
884                 icmp6stat.icp6s_tooshort++;
885                 goto freeit;
886         }
887 #ifndef PULLDOWN_TEST
888         IP6_EXTHDR_CHECK(m, off,
889                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
890                          -1);
891         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
892 #else
893         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
894                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
895         if (icmp6 == NULL) {
896                 icmp6stat.icp6s_tooshort++;
897                 return(-1);
898         }
899 #endif
900         eip6 = (struct ip6_hdr *)(icmp6 + 1);
901
902         /* Detect the upper level protocol */
903         {
904                 void (*ctlfunc) __P((int, struct sockaddr *, void *));
905                 u_int8_t nxt = eip6->ip6_nxt;
906                 int eoff = off + sizeof(struct icmp6_hdr) +
907                         sizeof(struct ip6_hdr);
908                 struct ip6ctlparam ip6cp;
909                 struct in6_addr *finaldst = NULL;
910                 int icmp6type = icmp6->icmp6_type;
911                 struct ip6_frag *fh;
912                 struct ip6_rthdr *rth;
913                 struct ip6_rthdr0 *rth0;
914                 int rthlen;
915
916                 while (1) { /* XXX: should avoid infinite loop explicitly? */
917                         struct ip6_ext *eh;
918
919                         switch (nxt) {
920                         case IPPROTO_HOPOPTS:
921                         case IPPROTO_DSTOPTS:
922                         case IPPROTO_AH:
923 #ifndef PULLDOWN_TEST
924                                 IP6_EXTHDR_CHECK(m, 0, eoff +
925                                                  sizeof(struct ip6_ext),
926                                                  -1);
927                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
928                                                         + eoff);
929 #else
930                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
931                                                eoff, sizeof(*eh));
932                                 if (eh == NULL) {
933                                         icmp6stat.icp6s_tooshort++;
934                                         return(-1);
935                                 }
936 #endif
937                                 
938                                 if (nxt == IPPROTO_AH)
939                                         eoff += (eh->ip6e_len + 2) << 2;
940                                 else
941                                         eoff += (eh->ip6e_len + 1) << 3;
942                                 nxt = eh->ip6e_nxt;
943                                 break;
944                         case IPPROTO_ROUTING:
945                                 /*
946                                  * When the erroneous packet contains a
947                                  * routing header, we should examine the
948                                  * header to determine the final destination.
949                                  * Otherwise, we can't properly update
950                                  * information that depends on the final
951                                  * destination (e.g. path MTU).
952                                  */
953 #ifndef PULLDOWN_TEST
954                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
955                                                  -1);
956                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
957                                                            + eoff);
958 #else
959                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
960                                                eoff, sizeof(*rth));
961                                 if (rth == NULL) {
962                                         icmp6stat.icp6s_tooshort++;
963                                         return(-1);
964                                 }
965 #endif
966                                 rthlen = (rth->ip6r_len + 1) << 3;
967                                 /*
968                                  * XXX: currently there is no
969                                  * officially defined type other
970                                  * than type-0.
971                                  * Note that if the segment left field
972                                  * is 0, all intermediate hops must
973                                  * have been passed.
974                                  */
975                                 if (rth->ip6r_segleft &&
976                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
977                                         int hops;
978
979 #ifndef PULLDOWN_TEST
980                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
981                                                          -1);
982                                         rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
983 #else
984                                         IP6_EXTHDR_GET(rth0,
985                                                        struct ip6_rthdr0 *, m,
986                                                        eoff, rthlen);
987                                         if (rth0 == NULL) {
988                                                 icmp6stat.icp6s_tooshort++;
989                                                 return(-1);
990                                         }
991 #endif
992                                         /* just ignore a bogus header */
993                                         if ((rth0->ip6r0_len % 2) == 0 &&
994                                             (hops = rth0->ip6r0_len/2))
995                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
996                                 }
997                                 eoff += rthlen;
998                                 nxt = rth->ip6r_nxt;
999                                 break;
1000                         case IPPROTO_FRAGMENT:
1001 #ifndef PULLDOWN_TEST
1002                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1003                                                  sizeof(struct ip6_frag),
1004                                                  -1);
1005                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
1006                                                          + eoff);
1007 #else
1008                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1009                                                eoff, sizeof(*fh));
1010                                 if (fh == NULL) {
1011                                         icmp6stat.icp6s_tooshort++;
1012                                         return(-1);
1013                                 }
1014 #endif
1015                                 /*
1016                                  * Data after a fragment header is meaningless
1017                                  * unless it is the first fragment, but
1018                                  * we'll go to the notify label for path MTU
1019                                  * discovery.
1020                                  */
1021                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1022                                         goto notify;
1023
1024                                 eoff += sizeof(struct ip6_frag);
1025                                 nxt = fh->ip6f_nxt;
1026                                 break;
1027                         default:
1028                                 /*
1029                                  * This case includes ESP and the No Next
1030                                  * Header.  In such cases going to the notify
1031                                  * label does not have any meaning
1032                                  * (i.e. ctlfunc will be NULL), but we go
1033                                  * anyway since we might have to update
1034                                  * path MTU information.
1035                                  */
1036                                 goto notify;
1037                         }
1038                 }
1039           notify:
1040 #ifndef PULLDOWN_TEST
1041                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1042 #else
1043                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1044                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
1045                 if (icmp6 == NULL) {
1046                         icmp6stat.icp6s_tooshort++;
1047                         return(-1);
1048                 }
1049 #endif
1050
1051                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1052                 bzero(&icmp6dst, sizeof(icmp6dst));
1053                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1054                 icmp6dst.sin6_family = AF_INET6;
1055                 if (finaldst == NULL)
1056                         icmp6dst.sin6_addr = eip6->ip6_dst;
1057                 else
1058                         icmp6dst.sin6_addr = *finaldst;
1059                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1060                                                           &icmp6dst.sin6_addr);
1061 #ifndef SCOPEDROUTING
1062                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1063                                    NULL, NULL)) {
1064                         /* should be impossbile */
1065                         nd6log((LOG_DEBUG,
1066                             "icmp6_notify_error: in6_embedscope failed\n"));
1067                         goto freeit;
1068                 }
1069 #endif
1070
1071                 /*
1072                  * retrieve parameters from the inner IPv6 header, and convert
1073                  * them into sockaddr structures.
1074                  */
1075                 bzero(&icmp6src, sizeof(icmp6src));
1076                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1077                 icmp6src.sin6_family = AF_INET6;
1078                 icmp6src.sin6_addr = eip6->ip6_src;
1079                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1080                                                           &icmp6src.sin6_addr);
1081 #ifndef SCOPEDROUTING
1082                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1083                                    NULL, NULL)) {
1084                         /* should be impossbile */
1085                         nd6log((LOG_DEBUG,
1086                             "icmp6_notify_error: in6_embedscope failed\n"));
1087                         goto freeit;
1088                 }
1089 #endif
1090                 icmp6src.sin6_flowinfo =
1091                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1092
1093                 if (finaldst == NULL)
1094                         finaldst = &eip6->ip6_dst;
1095                 ip6cp.ip6c_m = m;
1096                 ip6cp.ip6c_icmp6 = icmp6;
1097                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1098                 ip6cp.ip6c_off = eoff;
1099                 ip6cp.ip6c_finaldst = finaldst;
1100                 ip6cp.ip6c_src = &icmp6src;
1101                 ip6cp.ip6c_nxt = nxt;
1102
1103                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1104                         notifymtu = ntohl(icmp6->icmp6_mtu);
1105                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1106                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1107                 }
1108
1109                 ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1110                         (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1111                 if (ctlfunc) {
1112                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1113                                           &ip6cp);
1114                 }
1115         }
1116         return(0);
1117
1118   freeit:
1119         m_freem(m);
1120         return(-1);
1121 }
1122
1123 void
1124 icmp6_mtudisc_update(ip6cp, validated)
1125         struct ip6ctlparam *ip6cp;
1126         int validated;
1127 {
1128         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1129         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1130         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1131         u_int mtu = ntohl(icmp6->icmp6_mtu);
1132         struct rtentry *rt = NULL;
1133         struct sockaddr_in6 sin6;
1134
1135         if (!validated)
1136                 return;
1137
1138         bzero(&sin6, sizeof(sin6));
1139         sin6.sin6_family = PF_INET6;
1140         sin6.sin6_len = sizeof(struct sockaddr_in6);
1141         sin6.sin6_addr = *dst;
1142         /* XXX normally, this won't happen */
1143         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1144                 sin6.sin6_addr.s6_addr16[1] =
1145                     htons(m->m_pkthdr.rcvif->if_index);
1146         }
1147         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1148         rt = rtalloc1((struct sockaddr *)&sin6, 0,
1149                       RTF_CLONING | RTF_PRCLONING);
1150
1151         if (rt && (rt->rt_flags & RTF_HOST)
1152             && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1153                 if (mtu < IPV6_MMTU) {
1154                                 /* xxx */
1155                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1156                 } else if (mtu < rt->rt_ifp->if_mtu &&
1157                            rt->rt_rmx.rmx_mtu > mtu) {
1158                         icmp6stat.icp6s_pmtuchg++;
1159                         rt->rt_rmx.rmx_mtu = mtu;
1160                 }
1161         }
1162         if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1163                 RTFREE(rt);
1164         }
1165 }
1166
1167 /*
1168  * Process a Node Information Query packet, based on
1169  * draft-ietf-ipngwg-icmp-name-lookups-07.
1170  * 
1171  * Spec incompatibilities:
1172  * - IPv6 Subject address handling
1173  * - IPv4 Subject address handling support missing
1174  * - Proxy reply (answer even if it's not for me)
1175  * - joins NI group address at in6_ifattach() time only, does not cope
1176  *   with hostname changes by sethostname(3)
1177  */
1178 #define hostnamelen     strlen(hostname)
1179 static struct mbuf *
1180 ni6_input(m, off)
1181         struct mbuf *m;
1182         int off;
1183 {
1184         struct icmp6_nodeinfo *ni6, *nni6;
1185         struct mbuf *n = NULL;
1186         u_int16_t qtype;
1187         int subjlen;
1188         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1189         struct ni_reply_fqdn *fqdn;
1190         int addrs;              /* for NI_QTYPE_NODEADDR */
1191         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1192         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1193         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1194         struct ip6_hdr *ip6;
1195         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1196         char *subj = NULL;
1197         struct in6_ifaddr *ia6 = NULL;
1198
1199         ip6 = mtod(m, struct ip6_hdr *);
1200 #ifndef PULLDOWN_TEST
1201         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1202 #else
1203         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1204         if (ni6 == NULL) {
1205                 /* m is already reclaimed */
1206                 return NULL;
1207         }
1208 #endif
1209
1210         /*
1211          * Validate IPv6 destination address.
1212          *
1213          * The Responder must discard the Query without further processing
1214          * unless it is one of the Responder's unicast or anycast addresses, or
1215          * a link-local scope multicast address which the Responder has joined.
1216          * [icmp-name-lookups-07, Section 4.]
1217          */
1218         bzero(&sin6, sizeof(sin6));
1219         sin6.sin6_family = AF_INET6;
1220         sin6.sin6_len = sizeof(struct sockaddr_in6);
1221         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1222         /* XXX scopeid */
1223         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1224                 /* unicast/anycast, fine */
1225                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1226                     (icmp6_nodeinfo & 4) == 0) {
1227                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1228                                 "a temporary address in %s:%d",
1229                                __FILE__, __LINE__));
1230                         goto bad;
1231                 }
1232         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1233                 ; /* link-local multicast, fine */
1234         else
1235                 goto bad;
1236
1237         /* validate query Subject field. */
1238         qtype = ntohs(ni6->ni_qtype);
1239         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1240         switch (qtype) {
1241         case NI_QTYPE_NOOP:
1242         case NI_QTYPE_SUPTYPES:
1243                 /* 07 draft */
1244                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1245                         break;
1246                 /* FALLTHROUGH */
1247         case NI_QTYPE_FQDN:
1248         case NI_QTYPE_NODEADDR:
1249                 switch (ni6->ni_code) {
1250                 case ICMP6_NI_SUBJ_IPV6:
1251 #if ICMP6_NI_SUBJ_IPV6 != 0
1252                 case 0:
1253 #endif
1254                         /*
1255                          * backward compatibility - try to accept 03 draft
1256                          * format, where no Subject is present.
1257                          */
1258                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1259                             subjlen == 0) {
1260                                 oldfqdn++;
1261                                 break;
1262                         }
1263 #if ICMP6_NI_SUBJ_IPV6 != 0
1264                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1265                                 goto bad;
1266 #endif
1267
1268                         if (subjlen != sizeof(sin6.sin6_addr))
1269                                 goto bad;
1270
1271                         /*
1272                          * Validate Subject address.
1273                          *
1274                          * Not sure what exactly "address belongs to the node"
1275                          * means in the spec, is it just unicast, or what?
1276                          *
1277                          * At this moment we consider Subject address as
1278                          * "belong to the node" if the Subject address equals
1279                          * to the IPv6 destination address; validation for
1280                          * IPv6 destination address should have done enough
1281                          * check for us.
1282                          *
1283                          * We do not do proxy at this moment.
1284                          */
1285                         /* m_pulldown instead of copy? */
1286                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1287                             subjlen, (caddr_t)&sin6.sin6_addr);
1288                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1289                                                               &sin6.sin6_addr);
1290 #ifndef SCOPEDROUTING
1291                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1292 #endif
1293                         bzero(&sin6_d, sizeof(sin6_d));
1294                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1295                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1296                         sin6_d.sin6_addr = ip6->ip6_dst;
1297                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1298                                                                 &ip6->ip6_dst);
1299 #ifndef SCOPEDROUTING
1300                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1301 #endif
1302                         subj = (char *)&sin6;
1303                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1304                                 break;
1305
1306                         /*
1307                          * XXX if we are to allow other cases, we should really
1308                          * be careful about scope here.
1309                          * basically, we should disallow queries toward IPv6
1310                          * destination X with subject Y, if scope(X) > scope(Y).
1311                          * if we allow scope(X) > scope(Y), it will result in
1312                          * information leakage across scope boundary.
1313                          */
1314                         goto bad;
1315
1316                 case ICMP6_NI_SUBJ_FQDN:
1317                         /*
1318                          * Validate Subject name with gethostname(3).
1319                          *
1320                          * The behavior may need some debate, since:
1321                          * - we are not sure if the node has FQDN as
1322                          *   hostname (returned by gethostname(3)).
1323                          * - the code does wildcard match for truncated names.
1324                          *   however, we are not sure if we want to perform
1325                          *   wildcard match, if gethostname(3) side has
1326                          *   truncated hostname.
1327                          */
1328                         n = ni6_nametodns(hostname, hostnamelen, 0);
1329                         if (!n || n->m_next || n->m_len == 0)
1330                                 goto bad;
1331                         IP6_EXTHDR_GET(subj, char *, m,
1332                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1333                         if (subj == NULL)
1334                                 goto bad;
1335                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1336                                         n->m_len)) {
1337                                 goto bad;
1338                         }
1339                         m_freem(n);
1340                         n = NULL;
1341                         break;
1342
1343                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1344                 default:
1345                         goto bad;
1346                 }
1347                 break;
1348         }
1349
1350         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1351         switch (qtype) {
1352         case NI_QTYPE_FQDN:
1353                 if ((icmp6_nodeinfo & 1) == 0)
1354                         goto bad;
1355                 break;
1356         case NI_QTYPE_NODEADDR:
1357                 if ((icmp6_nodeinfo & 2) == 0)
1358                         goto bad;
1359                 break;
1360         }
1361
1362         /* guess reply length */
1363         switch (qtype) {
1364         case NI_QTYPE_NOOP:
1365                 break;          /* no reply data */
1366         case NI_QTYPE_SUPTYPES:
1367                 replylen += sizeof(u_int32_t);
1368                 break;
1369         case NI_QTYPE_FQDN:
1370                 /* XXX will append an mbuf */
1371                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1372                 break;
1373         case NI_QTYPE_NODEADDR:
1374                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1375                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1376                                           sizeof(u_int32_t))) > MCLBYTES)
1377                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1378                 break;
1379         default:
1380                 /*
1381                  * XXX: We must return a reply with the ICMP6 code
1382                  * `unknown Qtype' in this case. However we regard the case
1383                  * as an FQDN query for backward compatibility.
1384                  * Older versions set a random value to this field,
1385                  * so it rarely varies in the defined qtypes.
1386                  * But the mechanism is not reliable...
1387                  * maybe we should obsolete older versions.
1388                  */
1389                 qtype = NI_QTYPE_FQDN;
1390                 /* XXX will append an mbuf */
1391                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1392                 oldfqdn++;
1393                 break;
1394         }
1395
1396         /* allocate an mbuf to reply. */
1397         MGETHDR(n, M_DONTWAIT, m->m_type);
1398         if (n == NULL) {
1399                 m_freem(m);
1400                 return(NULL);
1401         }
1402         M_MOVE_PKTHDR(n, m); /* just for recvif */
1403         if (replylen > MHLEN) {
1404                 if (replylen > MCLBYTES) {
1405                         /*
1406                          * XXX: should we try to allocate more? But MCLBYTES
1407                          * is probably much larger than IPV6_MMTU...
1408                          */
1409                         goto bad;
1410                 }
1411                 MCLGET(n, M_DONTWAIT);
1412                 if ((n->m_flags & M_EXT) == 0) {
1413                         goto bad;
1414                 }
1415         }
1416         n->m_pkthdr.len = n->m_len = replylen;
1417
1418         /* copy mbuf header and IPv6 + Node Information base headers */
1419         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1420         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1421         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1422
1423         /* qtype dependent procedure */
1424         switch (qtype) {
1425         case NI_QTYPE_NOOP:
1426                 nni6->ni_code = ICMP6_NI_SUCCESS;
1427                 nni6->ni_flags = 0;
1428                 break;
1429         case NI_QTYPE_SUPTYPES:
1430         {
1431                 u_int32_t v;
1432                 nni6->ni_code = ICMP6_NI_SUCCESS;
1433                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1434                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1435                 v = (u_int32_t)htonl(0x0000000f);
1436                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1437                 break;
1438         }
1439         case NI_QTYPE_FQDN:
1440                 nni6->ni_code = ICMP6_NI_SUCCESS;
1441                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1442                                                 sizeof(struct ip6_hdr) +
1443                                                 sizeof(struct icmp6_nodeinfo));
1444                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1445                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1446                 /*
1447                  * XXX do we really have FQDN in variable "hostname"?
1448                  */
1449                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1450                 if (n->m_next == NULL)
1451                         goto bad;
1452                 /* XXX we assume that n->m_next is not a chain */
1453                 if (n->m_next->m_next != NULL)
1454                         goto bad;
1455                 n->m_pkthdr.len += n->m_next->m_len;
1456                 break;
1457         case NI_QTYPE_NODEADDR:
1458         {
1459                 int lenlim, copied;
1460
1461                 nni6->ni_code = ICMP6_NI_SUCCESS;
1462                 n->m_pkthdr.len = n->m_len =
1463                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1464                 lenlim = M_TRAILINGSPACE(n);
1465                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1466                 /* XXX: reset mbuf length */
1467                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1468                         sizeof(struct icmp6_nodeinfo) + copied;
1469                 break;
1470         }
1471         default:
1472                 break;          /* XXX impossible! */
1473         }
1474
1475         nni6->ni_type = ICMP6_NI_REPLY;
1476         m_freem(m);
1477         return(n);
1478
1479   bad:
1480         m_freem(m);
1481         if (n)
1482                 m_freem(n);
1483         return(NULL);
1484 }
1485 #undef hostnamelen
1486
1487 /*
1488  * make a mbuf with DNS-encoded string.  no compression support.
1489  *
1490  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1491  * treated as truncated name (two \0 at the end).  this is a wild guess.
1492  */
1493 static struct mbuf *
1494 ni6_nametodns(name, namelen, old)
1495         const char *name;
1496         int namelen;
1497         int old;        /* return pascal string if non-zero */
1498 {
1499         struct mbuf *m;
1500         char *cp, *ep;
1501         const char *p, *q;
1502         int i, len, nterm;
1503
1504         if (old)
1505                 len = namelen + 1;
1506         else
1507                 len = MCLBYTES;
1508
1509         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1510         MGET(m, M_DONTWAIT, MT_DATA);
1511         if (m && len > MLEN) {
1512                 MCLGET(m, M_DONTWAIT);
1513                 if ((m->m_flags & M_EXT) == 0)
1514                         goto fail;
1515         }
1516         if (!m)
1517                 goto fail;
1518         m->m_next = NULL;
1519
1520         if (old) {
1521                 m->m_len = len;
1522                 *mtod(m, char *) = namelen;
1523                 bcopy(name, mtod(m, char *) + 1, namelen);
1524                 return m;
1525         } else {
1526                 m->m_len = 0;
1527                 cp = mtod(m, char *);
1528                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1529
1530                 /* if not certain about my name, return empty buffer */
1531                 if (namelen == 0)
1532                         return m;
1533
1534                 /*
1535                  * guess if it looks like shortened hostname, or FQDN.
1536                  * shortened hostname needs two trailing "\0".
1537                  */
1538                 i = 0;
1539                 for (p = name; p < name + namelen; p++) {
1540                         if (*p && *p == '.')
1541                                 i++;
1542                 }
1543                 if (i < 2)
1544                         nterm = 2;
1545                 else
1546                         nterm = 1;
1547
1548                 p = name;
1549                 while (cp < ep && p < name + namelen) {
1550                         i = 0;
1551                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1552                                 i++;
1553                         /* result does not fit into mbuf */
1554                         if (cp + i + 1 >= ep)
1555                                 goto fail;
1556                         /*
1557                          * DNS label length restriction, RFC1035 page 8.
1558                          * "i == 0" case is included here to avoid returning
1559                          * 0-length label on "foo..bar".
1560                          */
1561                         if (i <= 0 || i >= 64)
1562                                 goto fail;
1563                         *cp++ = i;
1564                         bcopy(p, cp, i);
1565                         cp += i;
1566                         p = q;
1567                         if (p < name + namelen && *p == '.')
1568                                 p++;
1569                 }
1570                 /* termination */
1571                 if (cp + nterm >= ep)
1572                         goto fail;
1573                 while (nterm-- > 0)
1574                         *cp++ = '\0';
1575                 m->m_len = cp - mtod(m, char *);
1576                 return m;
1577         }
1578
1579         panic("should not reach here");
1580         /* NOTREACHED */
1581
1582  fail:
1583         if (m)
1584                 m_freem(m);
1585         return NULL;
1586 }
1587
1588 /*
1589  * check if two DNS-encoded string matches.  takes care of truncated
1590  * form (with \0\0 at the end).  no compression support.
1591  * XXX upper/lowercase match (see RFC2065)
1592  */
1593 static int
1594 ni6_dnsmatch(a, alen, b, blen)
1595         const char *a;
1596         int alen;
1597         const char *b;
1598         int blen;
1599 {
1600         const char *a0, *b0;
1601         int l;
1602
1603         /* simplest case - need validation? */
1604         if (alen == blen && bcmp(a, b, alen) == 0)
1605                 return 1;
1606
1607         a0 = a;
1608         b0 = b;
1609
1610         /* termination is mandatory */
1611         if (alen < 2 || blen < 2)
1612                 return 0;
1613         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1614                 return 0;
1615         alen--;
1616         blen--;
1617
1618         while (a - a0 < alen && b - b0 < blen) {
1619                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1620                         return 0;
1621
1622                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1623                         return 0;
1624                 /* we don't support compression yet */
1625                 if (a[0] >= 64 || b[0] >= 64)
1626                         return 0;
1627
1628                 /* truncated case */
1629                 if (a[0] == 0 && a - a0 == alen - 1)
1630                         return 1;
1631                 if (b[0] == 0 && b - b0 == blen - 1)
1632                         return 1;
1633                 if (a[0] == 0 || b[0] == 0)
1634                         return 0;
1635
1636                 if (a[0] != b[0])
1637                         return 0;
1638                 l = a[0];
1639                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1640                         return 0;
1641                 if (bcmp(a + 1, b + 1, l) != 0)
1642                         return 0;
1643
1644                 a += 1 + l;
1645                 b += 1 + l;
1646         }
1647
1648         if (a - a0 == alen && b - b0 == blen)
1649                 return 1;
1650         else
1651                 return 0;
1652 }
1653
1654 /*
1655  * calculate the number of addresses to be returned in the node info reply.
1656  */
1657 static int
1658 ni6_addrs(ni6, m, ifpp, subj)
1659         struct icmp6_nodeinfo *ni6;
1660         struct mbuf *m;
1661         struct ifnet **ifpp;
1662         char *subj;
1663 {
1664         struct ifnet *ifp;
1665         struct in6_ifaddr *ifa6;
1666         struct ifaddr *ifa;
1667         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1668         int addrs = 0, addrsofif, iffound = 0;
1669         int niflags = ni6->ni_flags;
1670
1671         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1672                 switch (ni6->ni_code) {
1673                 case ICMP6_NI_SUBJ_IPV6:
1674                         if (subj == NULL) /* must be impossible... */
1675                                 return(0);
1676                         subj_ip6 = (struct sockaddr_in6 *)subj;
1677                         break;
1678                 default:
1679                         /*
1680                          * XXX: we only support IPv6 subject address for
1681                          * this Qtype.
1682                          */
1683                         return(0);
1684                 }
1685         }
1686
1687         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1688         {
1689                 addrsofif = 0;
1690                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1691                 {
1692                         if (ifa->ifa_addr->sa_family != AF_INET6)
1693                                 continue;
1694                         ifa6 = (struct in6_ifaddr *)ifa;
1695
1696                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1697                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1698                                                &ifa6->ia_addr.sin6_addr))
1699                                 iffound = 1;
1700
1701                         /*
1702                          * IPv4-mapped addresses can only be returned by a
1703                          * Node Information proxy, since they represent
1704                          * addresses of IPv4-only nodes, which perforce do
1705                          * not implement this protocol.
1706                          * [icmp-name-lookups-07, Section 5.4]
1707                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1708                          * this function at this moment.
1709                          */
1710
1711                         /* What do we have to do about ::1? */
1712                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1713                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1714                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1715                                         continue;
1716                                 break;
1717                         case IPV6_ADDR_SCOPE_SITELOCAL:
1718                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1719                                         continue;
1720                                 break;
1721                         case IPV6_ADDR_SCOPE_GLOBAL:
1722                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1723                                         continue;
1724                                 break;
1725                         default:
1726                                 continue;
1727                         }
1728
1729                         /*
1730                          * check if anycast is okay.
1731                          * XXX: just experimental.  not in the spec.
1732                          */
1733                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1734                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1735                                 continue; /* we need only unicast addresses */
1736                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1737                             (icmp6_nodeinfo & 4) == 0) {
1738                                 continue;
1739                         }
1740                         addrsofif++; /* count the address */
1741                 }
1742                 if (iffound) {
1743                         *ifpp = ifp;
1744                         return(addrsofif);
1745                 }
1746
1747                 addrs += addrsofif;
1748         }
1749
1750         return(addrs);
1751 }
1752
1753 static int
1754 ni6_store_addrs(ni6, nni6, ifp0, resid)
1755         struct icmp6_nodeinfo *ni6, *nni6;
1756         struct ifnet *ifp0;
1757         int resid;
1758 {
1759         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1760         struct in6_ifaddr *ifa6;
1761         struct ifaddr *ifa;
1762         struct ifnet *ifp_dep = NULL;
1763         int copied = 0, allow_deprecated = 0;
1764         u_char *cp = (u_char *)(nni6 + 1);
1765         int niflags = ni6->ni_flags;
1766         u_int32_t ltime;
1767
1768         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1769                 return(0);      /* needless to copy */
1770
1771   again:
1772
1773         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1774         {
1775                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1776                      ifa = ifa->ifa_list.tqe_next)
1777                 {
1778                         if (ifa->ifa_addr->sa_family != AF_INET6)
1779                                 continue;
1780                         ifa6 = (struct in6_ifaddr *)ifa;
1781
1782                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1783                             allow_deprecated == 0) {
1784                                 /*
1785                                  * prefererred address should be put before
1786                                  * deprecated addresses.
1787                                  */
1788
1789                                 /* record the interface for later search */
1790                                 if (ifp_dep == NULL)
1791                                         ifp_dep = ifp;
1792
1793                                 continue;
1794                         }
1795                         else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1796                                  allow_deprecated != 0)
1797                                 continue; /* we now collect deprecated addrs */
1798
1799                         /* What do we have to do about ::1? */
1800                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1801                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1802                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1803                                         continue;
1804                                 break;
1805                         case IPV6_ADDR_SCOPE_SITELOCAL:
1806                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1807                                         continue;
1808                                 break;
1809                         case IPV6_ADDR_SCOPE_GLOBAL:
1810                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1811                                         continue;
1812                                 break;
1813                         default:
1814                                 continue;
1815                         }
1816
1817                         /*
1818                          * check if anycast is okay.
1819                          * XXX: just experimental. not in the spec.
1820                          */
1821                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1822                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1823                                 continue;
1824                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1825                             (icmp6_nodeinfo & 4) == 0) {
1826                                 continue;
1827                         }
1828
1829                         /* now we can copy the address */
1830                         if (resid < sizeof(struct in6_addr) +
1831                             sizeof(u_int32_t)) {
1832                                 /*
1833                                  * We give up much more copy.
1834                                  * Set the truncate flag and return.
1835                                  */
1836                                 nni6->ni_flags |=
1837                                         NI_NODEADDR_FLAG_TRUNCATE;
1838                                 return(copied);
1839                         }
1840
1841                         /*
1842                          * Set the TTL of the address.
1843                          * The TTL value should be one of the following
1844                          * according to the specification:
1845                          *
1846                          * 1. The remaining lifetime of a DHCP lease on the
1847                          *    address, or
1848                          * 2. The remaining Valid Lifetime of a prefix from
1849                          *    which the address was derived through Stateless
1850                          *    Autoconfiguration.
1851                          *
1852                          * Note that we currently do not support stateful
1853                          * address configuration by DHCPv6, so the former
1854                          * case can't happen.
1855                          */
1856                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1857                                 ltime = ND6_INFINITE_LIFETIME;
1858                         else {
1859                                 if (ifa6->ia6_lifetime.ia6t_expire >
1860                                     time_second)
1861                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1862                                 else
1863                                         ltime = 0;
1864                         }
1865                         
1866                         bcopy(&ltime, cp, sizeof(u_int32_t));
1867                         cp += sizeof(u_int32_t);
1868
1869                         /* copy the address itself */
1870                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1871                               sizeof(struct in6_addr));
1872                         /* XXX: KAME link-local hack; remove ifindex */
1873                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1874                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1875                         cp += sizeof(struct in6_addr);
1876                         
1877                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1878                         copied += (sizeof(struct in6_addr) +
1879                                    sizeof(u_int32_t));
1880                 }
1881                 if (ifp0)       /* we need search only on the specified IF */
1882                         break;
1883         }
1884
1885         if (allow_deprecated == 0 && ifp_dep != NULL) {
1886                 ifp = ifp_dep;
1887                 allow_deprecated = 1;
1888
1889                 goto again;
1890         }
1891
1892         return(copied);
1893 }
1894
1895 /*
1896  * XXX almost dup'ed code with rip6_input.
1897  */
1898 static int
1899 icmp6_rip6_input(mp, off)
1900         struct  mbuf **mp;
1901         int     off;
1902 {
1903         struct mbuf *m = *mp;
1904         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1905         struct in6pcb *in6p;
1906         struct in6pcb *last = NULL;
1907         struct sockaddr_in6 rip6src;
1908         struct icmp6_hdr *icmp6;
1909         struct mbuf *opts = NULL;
1910
1911 #ifndef PULLDOWN_TEST
1912         /* this is assumed to be safe. */
1913         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1914 #else
1915         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1916         if (icmp6 == NULL) {
1917                 /* m is already reclaimed */
1918                 return IPPROTO_DONE;
1919         }
1920 #endif
1921
1922         bzero(&rip6src, sizeof(rip6src));
1923         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1924         rip6src.sin6_family = AF_INET6;
1925         /* KAME hack: recover scopeid */
1926         (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1927
1928         LIST_FOREACH(in6p, &ripcb, inp_list)
1929         {
1930                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1931                         continue;
1932 #ifdef HAVE_NRL_INPCB
1933                 if (!(in6p->in6p_flags & INP_IPV6))
1934                         continue;
1935 #endif
1936                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1937                         continue;
1938                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1939                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1940                         continue;
1941                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1942                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1943                         continue;
1944                 if (in6p->in6p_icmp6filt
1945                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1946                                  in6p->in6p_icmp6filt))
1947                         continue;
1948                 if (last) {
1949                         struct  mbuf *n;
1950                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1951                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1952                                         ip6_savecontrol(last, &opts, ip6, n);
1953                                 /* strip intermediate headers */
1954                                 m_adj(n, off);
1955                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1956                                                  (struct sockaddr *)&rip6src,
1957                                                  n, opts) == 0) {
1958                                         /* should notify about lost packet */
1959                                         m_freem(n);
1960                                         if (opts) {
1961                                                 m_freem(opts);
1962                                         }
1963                                 } else
1964                                         sorwakeup(last->in6p_socket);
1965                                 opts = NULL;
1966                         }
1967                 }
1968                 last = in6p;
1969         }
1970         if (last) {
1971                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1972                         ip6_savecontrol(last, &opts, ip6, m);
1973                 /* strip intermediate headers */
1974                 m_adj(m, off);
1975                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1976                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1977                         m_freem(m);
1978                         if (opts)
1979                                 m_freem(opts);
1980                 } else
1981                         sorwakeup(last->in6p_socket);
1982         } else {
1983                 m_freem(m);
1984                 ip6stat.ip6s_delivered--;
1985         }
1986         return IPPROTO_DONE;
1987 }
1988
1989 /*
1990  * Reflect the ip6 packet back to the source.
1991  * OFF points to the icmp6 header, counted from the top of the mbuf.
1992  */
1993 void
1994 icmp6_reflect(m, off)
1995         struct  mbuf *m;
1996         size_t off;
1997 {
1998         struct ip6_hdr *ip6;
1999         struct icmp6_hdr *icmp6;
2000         struct in6_ifaddr *ia;
2001         struct in6_addr t, *src = 0;
2002         int plen;
2003         int type, code;
2004         struct ifnet *outif = NULL;
2005         struct sockaddr_in6 sa6_src, sa6_dst;
2006 #ifdef COMPAT_RFC1885
2007         int mtu = IPV6_MMTU;
2008         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2009 #endif
2010
2011         /* too short to reflect */
2012         if (off < sizeof(struct ip6_hdr)) {
2013                 nd6log((LOG_DEBUG,
2014                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2015                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2016                     __FILE__, __LINE__));
2017                 goto bad;
2018         }
2019
2020         /*
2021          * If there are extra headers between IPv6 and ICMPv6, strip
2022          * off that header first.
2023          */
2024 #ifdef DIAGNOSTIC
2025         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2026                 panic("assumption failed in icmp6_reflect");
2027 #endif
2028         if (off > sizeof(struct ip6_hdr)) {
2029                 size_t l;
2030                 struct ip6_hdr nip6;
2031
2032                 l = off - sizeof(struct ip6_hdr);
2033                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2034                 m_adj(m, l);
2035                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2036                 if (m->m_len < l) {
2037                         if ((m = m_pullup(m, l)) == NULL)
2038                                 return;
2039                 }
2040                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2041         } else /* off == sizeof(struct ip6_hdr) */ {
2042                 size_t l;
2043                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2044                 if (m->m_len < l) {
2045                         if ((m = m_pullup(m, l)) == NULL)
2046                                 return;
2047                 }
2048         }
2049         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2050         ip6 = mtod(m, struct ip6_hdr *);
2051         ip6->ip6_nxt = IPPROTO_ICMPV6;
2052         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2053         type = icmp6->icmp6_type; /* keep type for statistics */
2054         code = icmp6->icmp6_code; /* ditto. */
2055
2056         t = ip6->ip6_dst;
2057         /*
2058          * ip6_input() drops a packet if its src is multicast.
2059          * So, the src is never multicast.
2060          */
2061         ip6->ip6_dst = ip6->ip6_src;
2062
2063         /*
2064          * XXX: make sure to embed scope zone information, using
2065          * already embedded IDs or the received interface (if any).
2066          * Note that rcvif may be NULL.
2067          * TODO: scoped routing case (XXX).
2068          */
2069         bzero(&sa6_src, sizeof(sa6_src));
2070         sa6_src.sin6_family = AF_INET6;
2071         sa6_src.sin6_len = sizeof(sa6_src);
2072         sa6_src.sin6_addr = ip6->ip6_dst;
2073         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2074         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2075         bzero(&sa6_dst, sizeof(sa6_dst));
2076         sa6_dst.sin6_family = AF_INET6;
2077         sa6_dst.sin6_len = sizeof(sa6_dst);
2078         sa6_dst.sin6_addr = t;
2079         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2080         in6_embedscope(&t, &sa6_dst, NULL, NULL);
2081
2082 #ifdef COMPAT_RFC1885
2083         /*
2084          * xxx guess MTU
2085          * RFC 1885 requires that echo reply should be truncated if it
2086          * does not fit in with (return) path MTU, but the description was
2087          * removed in the new spec.
2088          */
2089         if (icmp6_reflect_rt.ro_rt == 0 ||
2090             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2091                 if (icmp6_reflect_rt.ro_rt) {
2092                         RTFREE(icmp6_reflect_rt.ro_rt);
2093                         icmp6_reflect_rt.ro_rt = 0;
2094                 }
2095                 bzero(sin6, sizeof(*sin6));
2096                 sin6->sin6_family = PF_INET6;
2097                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2098                 sin6->sin6_addr = ip6->ip6_dst;
2099
2100                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2101                             RTF_PRCLONING);
2102         }
2103
2104         if (icmp6_reflect_rt.ro_rt == 0)
2105                 goto bad;
2106
2107         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2108             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2109                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2110
2111         if (mtu < m->m_pkthdr.len) {
2112                 plen -= (m->m_pkthdr.len - mtu);
2113                 m_adj(m, mtu - m->m_pkthdr.len);
2114         }
2115 #endif
2116         /*
2117          * If the incoming packet was addressed directly to us(i.e. unicast),
2118          * use dst as the src for the reply.
2119          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2120          * (for example) when we encounter an error while forwarding procedure
2121          * destined to a duplicated address of ours.
2122          */
2123         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2124                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2125                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2126                         src = &t;
2127                         break;
2128                 }
2129         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2130                 /*
2131                  * This is the case if the dst is our link-local address
2132                  * and the sender is also ourselves.
2133                  */
2134                 src = &t;
2135         }
2136
2137         if (src == 0) {
2138                 int e;
2139                 struct route_in6 ro;
2140
2141                 /*
2142                  * This case matches to multicasts, our anycast, or unicasts
2143                  * that we do not own.  Select a source address based on the
2144                  * source address of the erroneous packet.
2145                  */
2146                 bzero(&ro, sizeof(ro));
2147                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2148                 if (ro.ro_rt)
2149                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2150                 if (src == NULL) {
2151                         nd6log((LOG_DEBUG,
2152                             "icmp6_reflect: source can't be determined: "
2153                             "dst=%s, error=%d\n",
2154                             ip6_sprintf(&sa6_src.sin6_addr), e));
2155                         goto bad;
2156                 }
2157         }
2158
2159         ip6->ip6_src = *src;
2160
2161         ip6->ip6_flow = 0;
2162         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2163         ip6->ip6_vfc |= IPV6_VERSION;
2164         ip6->ip6_nxt = IPPROTO_ICMPV6;
2165         if (m->m_pkthdr.rcvif) {
2166                 /* XXX: This may not be the outgoing interface */
2167                 ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2168         } else
2169                 ip6->ip6_hlim = ip6_defhlim;
2170
2171         icmp6->icmp6_cksum = 0;
2172         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2173                                         sizeof(struct ip6_hdr), plen);
2174
2175         /*
2176          * XXX option handling
2177          */
2178
2179         m->m_flags &= ~(M_BCAST|M_MCAST);
2180
2181 #ifdef COMPAT_RFC1885
2182         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2183 #else
2184         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2185 #endif
2186         if (outif)
2187                 icmp6_ifoutstat_inc(outif, type, code);
2188
2189         return;
2190
2191  bad:
2192         m_freem(m);
2193         return;
2194 }
2195
2196 void
2197 icmp6_fasttimo()
2198 {
2199
2200         mld6_fasttimeo();
2201 }
2202
2203 static const char *
2204 icmp6_redirect_diag(src6, dst6, tgt6)
2205         struct in6_addr *src6;
2206         struct in6_addr *dst6;
2207         struct in6_addr *tgt6;
2208 {
2209         static char buf[1024];
2210         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2211                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2212         return buf;
2213 }
2214
2215 void
2216 icmp6_redirect_input(m, off)
2217         struct mbuf *m;
2218         int off;
2219 {
2220         struct ifnet *ifp = m->m_pkthdr.rcvif;
2221         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2222         struct nd_redirect *nd_rd;
2223         int icmp6len = ntohs(ip6->ip6_plen);
2224         char *lladdr = NULL;
2225         int lladdrlen = 0;
2226         u_char *redirhdr = NULL;
2227         int redirhdrlen = 0;
2228         struct rtentry *rt = NULL;
2229         int is_router;
2230         int is_onlink;
2231         struct in6_addr src6 = ip6->ip6_src;
2232         struct in6_addr redtgt6;
2233         struct in6_addr reddst6;
2234         union nd_opts ndopts;
2235
2236         if (!m || !ifp)
2237                 return;
2238
2239         /* XXX if we are router, we don't update route by icmp6 redirect */
2240         if (ip6_forwarding)
2241                 goto freeit;
2242         if (!icmp6_rediraccept)
2243                 goto freeit;
2244
2245 #ifndef PULLDOWN_TEST
2246         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2247         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2248 #else
2249         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2250         if (nd_rd == NULL) {
2251                 icmp6stat.icp6s_tooshort++;
2252                 return;
2253         }
2254 #endif
2255         redtgt6 = nd_rd->nd_rd_target;
2256         reddst6 = nd_rd->nd_rd_dst;
2257
2258         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2259                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2260         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2261                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2262
2263         /* validation */
2264         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2265                 nd6log((LOG_ERR,
2266                         "ICMP6 redirect sent from %s rejected; "
2267                         "must be from linklocal\n", ip6_sprintf(&src6)));
2268                 goto bad;
2269         }
2270         if (ip6->ip6_hlim != 255) {
2271                 nd6log((LOG_ERR,
2272                         "ICMP6 redirect sent from %s rejected; "
2273                         "hlim=%d (must be 255)\n",
2274                         ip6_sprintf(&src6), ip6->ip6_hlim));
2275                 goto bad;
2276         }
2277     {
2278         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2279         struct sockaddr_in6 sin6;
2280         struct in6_addr *gw6;
2281
2282         bzero(&sin6, sizeof(sin6));
2283         sin6.sin6_family = AF_INET6;
2284         sin6.sin6_len = sizeof(struct sockaddr_in6);
2285         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2286         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2287         if (rt) {
2288                 if (rt->rt_gateway == NULL ||
2289                     rt->rt_gateway->sa_family != AF_INET6) {
2290                         nd6log((LOG_ERR,
2291                             "ICMP6 redirect rejected; no route "
2292                             "with inet6 gateway found for redirect dst: %s\n",
2293                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2294                         RTFREE(rt);
2295                         goto bad;
2296                 }
2297
2298                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2299                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2300                         nd6log((LOG_ERR,
2301                                 "ICMP6 redirect rejected; "
2302                                 "not equal to gw-for-src=%s (must be same): "
2303                                 "%s\n",
2304                                 ip6_sprintf(gw6),
2305                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2306                         RTFREE(rt);
2307                         goto bad;
2308                 }
2309         } else {
2310                 nd6log((LOG_ERR,
2311                         "ICMP6 redirect rejected; "
2312                         "no route found for redirect dst: %s\n",
2313                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2314                 goto bad;
2315         }
2316         RTFREE(rt);
2317         rt = NULL;
2318     }
2319         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2320                 nd6log((LOG_ERR,
2321                         "ICMP6 redirect rejected; "
2322                         "redirect dst must be unicast: %s\n",
2323                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2324                 goto bad;
2325         }
2326
2327         is_router = is_onlink = 0;
2328         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2329                 is_router = 1;  /* router case */
2330         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2331                 is_onlink = 1;  /* on-link destination case */
2332         if (!is_router && !is_onlink) {
2333                 nd6log((LOG_ERR,
2334                         "ICMP6 redirect rejected; "
2335                         "neither router case nor onlink case: %s\n",
2336                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2337                 goto bad;
2338         }
2339         /* validation passed */
2340
2341         icmp6len -= sizeof(*nd_rd);
2342         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2343         if (nd6_options(&ndopts) < 0) {
2344                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2345                         "invalid ND option, rejected: %s\n",
2346                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2347                 /* nd6_options have incremented stats */
2348                 goto freeit;
2349         }
2350
2351         if (ndopts.nd_opts_tgt_lladdr) {
2352                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2353                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2354         }
2355
2356         if (ndopts.nd_opts_rh) {
2357                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2358                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2359         }
2360
2361         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2362                 nd6log((LOG_INFO,
2363                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2364                         "(if %d, icmp6 packet %d): %s\n",
2365                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2366                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2367                 goto bad;
2368         }
2369
2370         /* RFC 2461 8.3 */
2371         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2372                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2373
2374         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2375                 /* perform rtredirect */
2376                 struct sockaddr_in6 sdst;
2377                 struct sockaddr_in6 sgw;
2378                 struct sockaddr_in6 ssrc;
2379
2380                 bzero(&sdst, sizeof(sdst));
2381                 bzero(&sgw, sizeof(sgw));
2382                 bzero(&ssrc, sizeof(ssrc));
2383                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2384                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2385                         sizeof(struct sockaddr_in6);
2386                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2387                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2388                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2389                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2390                            (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2391                            (struct sockaddr *)&ssrc,
2392                            (struct rtentry **)NULL);
2393         }
2394         /* finally update cached route in each socket via pfctlinput */
2395     {
2396         struct sockaddr_in6 sdst;
2397
2398         bzero(&sdst, sizeof(sdst));
2399         sdst.sin6_family = AF_INET6;
2400         sdst.sin6_len = sizeof(struct sockaddr_in6);
2401         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2402         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2403 #ifdef IPSEC
2404         key_sa_routechange((struct sockaddr *)&sdst);
2405 #endif
2406     }
2407
2408  freeit:
2409         m_freem(m);
2410         return;
2411
2412  bad:
2413         icmp6stat.icp6s_badredirect++;
2414         m_freem(m);
2415 }
2416
2417 void
2418 icmp6_redirect_output(m0, rt)
2419         struct mbuf *m0;
2420         struct rtentry *rt;
2421 {
2422         struct ifnet *ifp;      /* my outgoing interface */
2423         struct in6_addr *ifp_ll6;
2424         struct in6_addr *router_ll6;
2425         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2426         struct mbuf *m = NULL;  /* newly allocated one */
2427         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2428         struct nd_redirect *nd_rd;
2429         size_t maxlen;
2430         u_char *p;
2431         struct ifnet *outif = NULL;
2432         struct sockaddr_in6 src_sa;
2433
2434         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2435
2436         /* if we are not router, we don't send icmp6 redirect */
2437         if (!ip6_forwarding || ip6_accept_rtadv)
2438                 goto fail;
2439
2440         /* sanity check */
2441         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2442                 goto fail;
2443
2444         /*
2445          * Address check:
2446          *  the source address must identify a neighbor, and
2447          *  the destination address must not be a multicast address
2448          *  [RFC 2461, sec 8.2]
2449          */
2450         sip6 = mtod(m0, struct ip6_hdr *);
2451         bzero(&src_sa, sizeof(src_sa));
2452         src_sa.sin6_family = AF_INET6;
2453         src_sa.sin6_len = sizeof(src_sa);
2454         src_sa.sin6_addr = sip6->ip6_src;
2455         /* we don't currently use sin6_scope_id, but eventually use it */
2456         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2457         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2458                 goto fail;
2459         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2460                 goto fail;      /* what should we do here? */
2461
2462         /* rate limit */
2463         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2464                 goto fail;
2465
2466         /*
2467          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2468          * we almost always ask for an mbuf cluster for simplicity.
2469          * (MHLEN < IPV6_MMTU is almost always true)
2470          */
2471 #if IPV6_MMTU >= MCLBYTES
2472 # error assumption failed about IPV6_MMTU and MCLBYTES
2473 #endif
2474         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2475         if (m && IPV6_MMTU >= MHLEN)
2476                 MCLGET(m, M_DONTWAIT);
2477         if (!m)
2478                 goto fail;
2479         m->m_pkthdr.rcvif = NULL;
2480         m->m_len = 0;
2481         maxlen = M_TRAILINGSPACE(m);
2482         maxlen = min(IPV6_MMTU, maxlen);
2483         /* just for safety */
2484         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2485             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2486                 goto fail;
2487         }
2488
2489         {
2490                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2491                 struct in6_ifaddr *ia;
2492                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2493                                                  IN6_IFF_NOTREADY|
2494                                                  IN6_IFF_ANYCAST)) == NULL)
2495                         goto fail;
2496                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2497         }
2498
2499         /* get ip6 linklocal address for the router. */
2500         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2501                 struct sockaddr_in6 *sin6;
2502                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2503                 router_ll6 = &sin6->sin6_addr;
2504                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2505                         router_ll6 = (struct in6_addr *)NULL;
2506         } else
2507                 router_ll6 = (struct in6_addr *)NULL;
2508
2509         /* ip6 */
2510         ip6 = mtod(m, struct ip6_hdr *);
2511         ip6->ip6_flow = 0;
2512         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2513         ip6->ip6_vfc |= IPV6_VERSION;
2514         /* ip6->ip6_plen will be set later */
2515         ip6->ip6_nxt = IPPROTO_ICMPV6;
2516         ip6->ip6_hlim = 255;
2517         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2518         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2519         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2520
2521         /* ND Redirect */
2522         nd_rd = (struct nd_redirect *)(ip6 + 1);
2523         nd_rd->nd_rd_type = ND_REDIRECT;
2524         nd_rd->nd_rd_code = 0;
2525         nd_rd->nd_rd_reserved = 0;
2526         if (rt->rt_flags & RTF_GATEWAY) {
2527                 /*
2528                  * nd_rd->nd_rd_target must be a link-local address in
2529                  * better router cases.
2530                  */
2531                 if (!router_ll6)
2532                         goto fail;
2533                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2534                       sizeof(nd_rd->nd_rd_target));
2535                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2536                       sizeof(nd_rd->nd_rd_dst));
2537         } else {
2538                 /* make sure redtgt == reddst */
2539                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2540                       sizeof(nd_rd->nd_rd_target));
2541                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2542                       sizeof(nd_rd->nd_rd_dst));
2543         }
2544
2545         p = (u_char *)(nd_rd + 1);
2546
2547         if (!router_ll6)
2548                 goto nolladdropt;
2549
2550     {
2551         /* target lladdr option */
2552         struct rtentry *rt_router = NULL;
2553         int len;
2554         struct sockaddr_dl *sdl;
2555         struct nd_opt_hdr *nd_opt;
2556         char *lladdr;
2557
2558         rt_router = nd6_lookup(router_ll6, 0, ifp);
2559         if (!rt_router)
2560                 goto nolladdropt;
2561         len = sizeof(*nd_opt) + ifp->if_addrlen;
2562         len = (len + 7) & ~7;   /* round by 8 */
2563         /* safety check */
2564         if (len + (p - (u_char *)ip6) > maxlen)
2565                 goto nolladdropt;
2566         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2567             (rt_router->rt_flags & RTF_LLINFO) &&
2568             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2569             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2570             sdl->sdl_alen) {
2571                 nd_opt = (struct nd_opt_hdr *)p;
2572                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2573                 nd_opt->nd_opt_len = len >> 3;
2574                 lladdr = (char *)(nd_opt + 1);
2575                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2576                 p += len;
2577         }
2578     }
2579 nolladdropt:;
2580
2581         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2582
2583         /* just to be safe */
2584 #ifdef M_DECRYPTED      /*not openbsd*/
2585         if (m0->m_flags & M_DECRYPTED)
2586                 goto noredhdropt;
2587 #endif
2588         if (p - (u_char *)ip6 > maxlen)
2589                 goto noredhdropt;
2590
2591     {
2592         /* redirected header option */
2593         int len;
2594         struct nd_opt_rd_hdr *nd_opt_rh;
2595
2596         /*
2597          * compute the maximum size for icmp6 redirect header option.
2598          * XXX room for auth header?
2599          */
2600         len = maxlen - (p - (u_char *)ip6);
2601         len &= ~7;
2602
2603         /* This is just for simplicity. */
2604         if (m0->m_pkthdr.len != m0->m_len) {
2605                 if (m0->m_next) {
2606                         m_freem(m0->m_next);
2607                         m0->m_next = NULL;
2608                 }
2609                 m0->m_pkthdr.len = m0->m_len;
2610         }
2611
2612         /*
2613          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2614          * about padding/truncate rule for the original IP packet.
2615          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2616          * - "attach as much as possible" is the goal
2617          * - pad if not aligned (original size can be guessed by original
2618          *   ip6 header)
2619          * Following code adds the padding if it is simple enough,
2620          * and truncates if not.
2621          */
2622         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2623                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2624
2625         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2626                 /* not enough room, truncate */
2627                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2628         } else {
2629                 /* enough room, pad or truncate */
2630                 size_t extra;
2631
2632                 extra = m0->m_pkthdr.len % 8;
2633                 if (extra) {
2634                         /* pad if easy enough, truncate if not */
2635                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2636                                 /* pad */
2637                                 m0->m_len += (8 - extra);
2638                                 m0->m_pkthdr.len += (8 - extra);
2639                         } else {
2640                                 /* truncate */
2641                                 m0->m_pkthdr.len -= extra;
2642                                 m0->m_len -= extra;
2643                         }
2644                 }
2645                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2646                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2647         }
2648
2649         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2650         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2651         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2652         nd_opt_rh->nd_opt_rh_len = len >> 3;
2653         p += sizeof(*nd_opt_rh);
2654         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2655
2656         /* connect m0 to m */
2657         m->m_next = m0;
2658         m->m_pkthdr.len = m->m_len + m0->m_len;
2659     }
2660 noredhdropt:;
2661
2662         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2663                 sip6->ip6_src.s6_addr16[1] = 0;
2664         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2665                 sip6->ip6_dst.s6_addr16[1] = 0;
2666 #if 0
2667         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2668                 ip6->ip6_src.s6_addr16[1] = 0;
2669         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2670                 ip6->ip6_dst.s6_addr16[1] = 0;
2671 #endif
2672         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2673                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2674         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2675                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2676
2677         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2678
2679         nd_rd->nd_rd_cksum = 0;
2680         nd_rd->nd_rd_cksum
2681                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2682
2683         /* send the packet to outside... */
2684         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2685         if (outif) {
2686                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2687                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2688         }
2689         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2690
2691         return;
2692
2693 fail:
2694         if (m)
2695                 m_freem(m);
2696         if (m0)
2697                 m_freem(m0);
2698 }
2699
2700 #ifdef HAVE_NRL_INPCB
2701 #define sotoin6pcb      sotoinpcb
2702 #define in6pcb          inpcb
2703 #define in6p_icmp6filt  inp_icmp6filt
2704 #endif
2705 /*
2706  * ICMPv6 socket option processing.
2707  */
2708 int
2709 icmp6_ctloutput(so, sopt)
2710         struct socket *so;
2711         struct sockopt *sopt;
2712 {
2713         int error = 0;
2714         int optlen;
2715         struct inpcb *inp = sotoinpcb(so);
2716         int level, op, optname;
2717
2718         if (sopt) {
2719                 level = sopt->sopt_level;
2720                 op = sopt->sopt_dir;
2721                 optname = sopt->sopt_name;
2722                 optlen = sopt->sopt_valsize;
2723         } else
2724                 level = op = optname = optlen = 0;
2725
2726         if (level != IPPROTO_ICMPV6) {
2727                 return EINVAL;
2728         }
2729
2730         switch (op) {
2731         case PRCO_SETOPT:
2732                 switch (optname) {
2733                 case ICMP6_FILTER:
2734                     {
2735                         struct icmp6_filter *p;
2736
2737                         if (optlen != sizeof(*p)) {
2738                                 error = EMSGSIZE;
2739                                 break;
2740                         }
2741                         if (inp->in6p_icmp6filt == NULL) {
2742                                 error = EINVAL;
2743                                 break;
2744                         }
2745                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2746                                 optlen);
2747                         break;
2748                     }
2749
2750                 default:
2751                         error = ENOPROTOOPT;
2752                         break;
2753                 }
2754                 break;
2755
2756         case PRCO_GETOPT:
2757                 switch (optname) {
2758                 case ICMP6_FILTER:
2759                     {
2760                         if (inp->in6p_icmp6filt == NULL) {
2761                                 error = EINVAL;
2762                                 break;
2763                         }
2764                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2765                                 sizeof(struct icmp6_filter));
2766                         break;
2767                     }
2768
2769                 default:
2770                         error = ENOPROTOOPT;
2771                         break;
2772                 }
2773                 break;
2774         }
2775
2776         return(error);
2777 }
2778 #ifdef HAVE_NRL_INPCB
2779 #undef sotoin6pcb
2780 #undef in6pcb
2781 #undef in6p_icmp6filt
2782 #endif
2783
2784 #ifndef HAVE_PPSRATECHECK
2785 #ifndef timersub
2786 #define timersub(tvp, uvp, vvp)                                         \
2787         do {                                                            \
2788                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2789                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2790                 if ((vvp)->tv_usec < 0) {                               \
2791                         (vvp)->tv_sec--;                                \
2792                         (vvp)->tv_usec += 1000000;                      \
2793                 }                                                       \
2794         } while (0)
2795 #endif
2796
2797 /*
2798  * ppsratecheck(): packets (or events) per second limitation.
2799  */
2800 static int
2801 ppsratecheck(lasttime, curpps, maxpps)
2802         struct timeval *lasttime;
2803         int *curpps;
2804         int maxpps;     /* maximum pps allowed */
2805 {
2806         struct timeval tv, delta;
2807         int s, rv;
2808
2809         s = splclock(); 
2810         microtime(&tv);
2811         splx(s);
2812
2813         timersub(&tv, lasttime, &delta);
2814
2815         /*
2816          * Check for 0,0 so that the message will be seen at least once.
2817          * If more than one second has passed since the last update of
2818          * lasttime, reset the counter.
2819          *
2820          * We do increment *curpps even in *curpps < maxpps case, as some may
2821          * try to use *curpps for stat purposes as well.
2822          */
2823         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2824             delta.tv_sec >= 1) {
2825                 *lasttime = tv;
2826                 *curpps = 0;
2827                 rv = 1;
2828         } else if (maxpps < 0)
2829                 rv = 1;
2830         else if (*curpps < maxpps)
2831                 rv = 1;
2832         else
2833                 rv = 0;
2834
2835 #if 1 /* DIAGNOSTIC? */
2836         /* be careful about wrap-around */
2837         if (*curpps + 1 > *curpps)
2838                 *curpps = *curpps + 1;
2839 #else
2840         /*
2841          * assume that there's not too many calls to this function.
2842          * not sure if the assumption holds, as it depends on *caller's*
2843          * behavior, not the behavior of this function.
2844          * IMHO it is wrong to make assumption on the caller's behavior,
2845          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2846          */
2847         *curpps = *curpps + 1;
2848 #endif
2849
2850         return (rv);
2851 }
2852 #endif
2853
2854 /*
2855  * Perform rate limit check.
2856  * Returns 0 if it is okay to send the icmp6 packet.
2857  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2858  * limitation.
2859  *
2860  * XXX per-destination/type check necessary?
2861  */
2862 static int
2863 icmp6_ratelimit(dst, type, code)
2864         const struct in6_addr *dst;     /* not used at this moment */
2865         const int type;                 /* not used at this moment */
2866         const int code;                 /* not used at this moment */
2867 {
2868         int ret;
2869
2870         ret = 0;        /* okay to send */
2871
2872         /* PPS limit */
2873         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2874             icmp6errppslim)) {
2875                 /* The packet is subject to rate limit */
2876                 ret++;
2877         }
2878
2879         return ret;
2880 }