]> CyberLeo.Net >> Repos - FreeBSD/releng/10.2.git/blob - sys/netinet6/icmp6.c
- Copy stable/10@285827 to releng/10.2 in preparation for 10.2-RC1
[FreeBSD/releng/10.2.git] / sys / netinet6 / icmp6.c
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *      The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
61  */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68
69 #include <sys/param.h>
70 #include <sys/domain.h>
71 #include <sys/jail.h>
72 #include <sys/kernel.h>
73 #include <sys/lock.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/protosw.h>
78 #include <sys/signalvar.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sx.h>
82 #include <sys/syslog.h>
83 #include <sys/systm.h>
84 #include <sys/time.h>
85
86 #include <net/if.h>
87 #include <net/if_dl.h>
88 #include <net/if_llatbl.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/vnet.h>
92
93 #include <netinet/in.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip6.h>
97 #include <netinet/icmp6.h>
98 #include <netinet/tcp_var.h>
99
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/in6_pcb.h>
102 #include <netinet6/ip6protosw.h>
103 #include <netinet6/ip6_var.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/mld6_var.h>
106 #include <netinet6/nd6.h>
107 #include <netinet6/send.h>
108
109 extern struct domain inet6domain;
110
111 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
112 VNET_PCPUSTAT_SYSINIT(icmp6stat);
113
114 #ifdef VIMAGE
115 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
116 #endif /* VIMAGE */
117
118 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
119 VNET_DECLARE(struct inpcbhead, ripcb);
120 VNET_DECLARE(int, icmp6errppslim);
121 static VNET_DEFINE(int, icmp6errpps_count) = 0;
122 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
123 VNET_DECLARE(int, icmp6_nodeinfo);
124
125 #define V_ripcbinfo                     VNET(ripcbinfo)
126 #define V_ripcb                         VNET(ripcb)
127 #define V_icmp6errppslim                VNET(icmp6errppslim)
128 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
129 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
130 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
131
132 static void icmp6_errcount(int, int);
133 static int icmp6_rip6_input(struct mbuf **, int);
134 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
135 static const char *icmp6_redirect_diag(struct in6_addr *,
136         struct in6_addr *, struct in6_addr *);
137 static struct mbuf *ni6_input(struct mbuf *, int);
138 static struct mbuf *ni6_nametodns(const char *, int, int);
139 static int ni6_dnsmatch(const char *, int, const char *, int);
140 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
141                           struct ifnet **, struct in6_addr *);
142 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
143                                 struct ifnet *, int);
144 static int icmp6_notify_error(struct mbuf **, int, int, int);
145
146 /*
147  * Kernel module interface for updating icmp6stat.  The argument is an index
148  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
149  * general layout of icmp6stat into the caller, it doesn't encode its
150  * location, so that future changes to add, for example, per-CPU stats
151  * support won't cause binary compatibility problems for kernel modules.
152  */
153 void
154 kmod_icmp6stat_inc(int statnum)
155 {
156
157         counter_u64_add(VNET(icmp6stat)[statnum], 1);
158 }
159
160 static void
161 icmp6_errcount(int type, int code)
162 {
163         switch (type) {
164         case ICMP6_DST_UNREACH:
165                 switch (code) {
166                 case ICMP6_DST_UNREACH_NOROUTE:
167                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
168                         return;
169                 case ICMP6_DST_UNREACH_ADMIN:
170                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
171                         return;
172                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
173                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
174                         return;
175                 case ICMP6_DST_UNREACH_ADDR:
176                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
177                         return;
178                 case ICMP6_DST_UNREACH_NOPORT:
179                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
180                         return;
181                 }
182                 break;
183         case ICMP6_PACKET_TOO_BIG:
184                 ICMP6STAT_INC(icp6s_opacket_too_big);
185                 return;
186         case ICMP6_TIME_EXCEEDED:
187                 switch (code) {
188                 case ICMP6_TIME_EXCEED_TRANSIT:
189                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
190                         return;
191                 case ICMP6_TIME_EXCEED_REASSEMBLY:
192                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
193                         return;
194                 }
195                 break;
196         case ICMP6_PARAM_PROB:
197                 switch (code) {
198                 case ICMP6_PARAMPROB_HEADER:
199                         ICMP6STAT_INC(icp6s_oparamprob_header);
200                         return;
201                 case ICMP6_PARAMPROB_NEXTHEADER:
202                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
203                         return;
204                 case ICMP6_PARAMPROB_OPTION:
205                         ICMP6STAT_INC(icp6s_oparamprob_option);
206                         return;
207                 }
208                 break;
209         case ND_REDIRECT:
210                 ICMP6STAT_INC(icp6s_oredirect);
211                 return;
212         }
213         ICMP6STAT_INC(icp6s_ounknown);
214 }
215
216 /*
217  * A wrapper function for icmp6_error() necessary when the erroneous packet
218  * may not contain enough scope zone information.
219  */
220 void
221 icmp6_error2(struct mbuf *m, int type, int code, int param,
222     struct ifnet *ifp)
223 {
224         struct ip6_hdr *ip6;
225
226         if (ifp == NULL)
227                 return;
228
229 #ifndef PULLDOWN_TEST
230         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
231 #else
232         if (m->m_len < sizeof(struct ip6_hdr)) {
233                 m = m_pullup(m, sizeof(struct ip6_hdr));
234                 if (m == NULL)
235                         return;
236         }
237 #endif
238
239         ip6 = mtod(m, struct ip6_hdr *);
240
241         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
242                 return;
243         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
244                 return;
245
246         icmp6_error(m, type, code, param);
247 }
248
249 /*
250  * Generate an error packet of type error in response to bad IP6 packet.
251  */
252 void
253 icmp6_error(struct mbuf *m, int type, int code, int param)
254 {
255         struct ip6_hdr *oip6, *nip6;
256         struct icmp6_hdr *icmp6;
257         u_int preplen;
258         int off;
259         int nxt;
260
261         ICMP6STAT_INC(icp6s_error);
262
263         /* count per-type-code statistics */
264         icmp6_errcount(type, code);
265
266 #ifdef M_DECRYPTED      /*not openbsd*/
267         if (m->m_flags & M_DECRYPTED) {
268                 ICMP6STAT_INC(icp6s_canterror);
269                 goto freeit;
270         }
271 #endif
272
273 #ifndef PULLDOWN_TEST
274         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
275 #else
276         if (m->m_len < sizeof(struct ip6_hdr)) {
277                 m = m_pullup(m, sizeof(struct ip6_hdr));
278                 if (m == NULL)
279                         return;
280         }
281 #endif
282         oip6 = mtod(m, struct ip6_hdr *);
283
284         /*
285          * If the destination address of the erroneous packet is a multicast
286          * address, or the packet was sent using link-layer multicast,
287          * we should basically suppress sending an error (RFC 2463, Section
288          * 2.4).
289          * We have two exceptions (the item e.2 in that section):
290          * - the Packet Too Big message can be sent for path MTU discovery.
291          * - the Parameter Problem Message that can be allowed an icmp6 error
292          *   in the option type field.  This check has been done in
293          *   ip6_unknown_opt(), so we can just check the type and code.
294          */
295         if ((m->m_flags & (M_BCAST|M_MCAST) ||
296              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
297             (type != ICMP6_PACKET_TOO_BIG &&
298              (type != ICMP6_PARAM_PROB ||
299               code != ICMP6_PARAMPROB_OPTION)))
300                 goto freeit;
301
302         /*
303          * RFC 2463, 2.4 (e.5): source address check.
304          * XXX: the case of anycast source?
305          */
306         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
307             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
308                 goto freeit;
309
310         /*
311          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
312          * don't do it.
313          */
314         nxt = -1;
315         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
316         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
317                 struct icmp6_hdr *icp;
318
319 #ifndef PULLDOWN_TEST
320                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
321                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
322 #else
323                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
324                         sizeof(*icp));
325                 if (icp == NULL) {
326                         ICMP6STAT_INC(icp6s_tooshort);
327                         return;
328                 }
329 #endif
330                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
331                     icp->icmp6_type == ND_REDIRECT) {
332                         /*
333                          * ICMPv6 error
334                          * Special case: for redirect (which is
335                          * informational) we must not send icmp6 error.
336                          */
337                         ICMP6STAT_INC(icp6s_canterror);
338                         goto freeit;
339                 } else {
340                         /* ICMPv6 informational - send the error */
341                 }
342         } else {
343                 /* non-ICMPv6 - send the error */
344         }
345
346         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
347
348         /* Finally, do rate limitation check. */
349         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
350                 ICMP6STAT_INC(icp6s_toofreq);
351                 goto freeit;
352         }
353
354         /*
355          * OK, ICMP6 can be generated.
356          */
357
358         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
359                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
360
361         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
362         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
363         if (m == NULL) {
364                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
365                 return;
366         }
367
368         nip6 = mtod(m, struct ip6_hdr *);
369         nip6->ip6_src  = oip6->ip6_src;
370         nip6->ip6_dst  = oip6->ip6_dst;
371
372         in6_clearscope(&oip6->ip6_src);
373         in6_clearscope(&oip6->ip6_dst);
374
375         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
376         icmp6->icmp6_type = type;
377         icmp6->icmp6_code = code;
378         icmp6->icmp6_pptr = htonl((u_int32_t)param);
379
380         /*
381          * icmp6_reflect() is designed to be in the input path.
382          * icmp6_error() can be called from both input and output path,
383          * and if we are in output path rcvif could contain bogus value.
384          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
385          * information in ip header (nip6).
386          */
387         m->m_pkthdr.rcvif = NULL;
388
389         ICMP6STAT_INC(icp6s_outhist[type]);
390         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
391
392         return;
393
394   freeit:
395         /*
396          * If we can't tell whether or not we can generate ICMP6, free it.
397          */
398         m_freem(m);
399 }
400
401 /*
402  * Process a received ICMP6 message.
403  */
404 int
405 icmp6_input(struct mbuf **mp, int *offp, int proto)
406 {
407         struct mbuf *m = *mp, *n;
408         struct ifnet *ifp;
409         struct ip6_hdr *ip6, *nip6;
410         struct icmp6_hdr *icmp6, *nicmp6;
411         int off = *offp;
412         int icmp6len = m->m_pkthdr.len - *offp;
413         int code, sum, noff;
414         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
415         int ip6len, error;
416
417         ifp = m->m_pkthdr.rcvif;
418
419 #ifndef PULLDOWN_TEST
420         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
421         /* m might change if M_LOOP.  So, call mtod after this */
422 #endif
423
424         /*
425          * Locate icmp6 structure in mbuf, and check
426          * that not corrupted and of at least minimum length
427          */
428
429         ip6 = mtod(m, struct ip6_hdr *);
430         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
431         if (icmp6len < sizeof(struct icmp6_hdr)) {
432                 ICMP6STAT_INC(icp6s_tooshort);
433                 goto freeit;
434         }
435
436         /*
437          * Check multicast group membership.
438          * Note: SSM filters are not applied for ICMPv6 traffic.
439          */
440         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
441                 struct in6_multi        *inm;
442
443                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
444                 if (inm == NULL) {
445                         IP6STAT_INC(ip6s_notmember);
446                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
447                         goto freeit;
448                 }
449         }
450
451         /*
452          * calculate the checksum
453          */
454 #ifndef PULLDOWN_TEST
455         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
456 #else
457         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
458         if (icmp6 == NULL) {
459                 ICMP6STAT_INC(icp6s_tooshort);
460                 return IPPROTO_DONE;
461         }
462 #endif
463         code = icmp6->icmp6_code;
464
465         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
466                 nd6log((LOG_ERR,
467                     "ICMP6 checksum error(%d|%x) %s\n",
468                     icmp6->icmp6_type, sum,
469                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
470                 ICMP6STAT_INC(icp6s_checksum);
471                 goto freeit;
472         }
473
474         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
475                 /*
476                  * Deliver very specific ICMP6 type only.
477                  * This is important to deliver TOOBIG.  Otherwise PMTUD
478                  * will not work.
479                  */
480                 switch (icmp6->icmp6_type) {
481                 case ICMP6_DST_UNREACH:
482                 case ICMP6_PACKET_TOO_BIG:
483                 case ICMP6_TIME_EXCEEDED:
484                         break;
485                 default:
486                         goto freeit;
487                 }
488         }
489
490         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
491         icmp6_ifstat_inc(ifp, ifs6_in_msg);
492         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
493                 icmp6_ifstat_inc(ifp, ifs6_in_error);
494
495         switch (icmp6->icmp6_type) {
496         case ICMP6_DST_UNREACH:
497                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
498                 switch (code) {
499                 case ICMP6_DST_UNREACH_NOROUTE:
500                         code = PRC_UNREACH_NET;
501                         break;
502                 case ICMP6_DST_UNREACH_ADMIN:
503                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
504                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
505                         break;
506                 case ICMP6_DST_UNREACH_ADDR:
507                         code = PRC_HOSTDEAD;
508                         break;
509                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
510                         /* I mean "source address was incorrect." */
511                         code = PRC_PARAMPROB;
512                         break;
513                 case ICMP6_DST_UNREACH_NOPORT:
514                         code = PRC_UNREACH_PORT;
515                         break;
516                 default:
517                         goto badcode;
518                 }
519                 goto deliver;
520                 break;
521
522         case ICMP6_PACKET_TOO_BIG:
523                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
524
525                 /* validation is made in icmp6_mtudisc_update */
526
527                 code = PRC_MSGSIZE;
528
529                 /*
530                  * Updating the path MTU will be done after examining
531                  * intermediate extension headers.
532                  */
533                 goto deliver;
534                 break;
535
536         case ICMP6_TIME_EXCEEDED:
537                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
538                 switch (code) {
539                 case ICMP6_TIME_EXCEED_TRANSIT:
540                         code = PRC_TIMXCEED_INTRANS;
541                         break;
542                 case ICMP6_TIME_EXCEED_REASSEMBLY:
543                         code = PRC_TIMXCEED_REASS;
544                         break;
545                 default:
546                         goto badcode;
547                 }
548                 goto deliver;
549                 break;
550
551         case ICMP6_PARAM_PROB:
552                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
553                 switch (code) {
554                 case ICMP6_PARAMPROB_NEXTHEADER:
555                         code = PRC_UNREACH_PROTOCOL;
556                         break;
557                 case ICMP6_PARAMPROB_HEADER:
558                 case ICMP6_PARAMPROB_OPTION:
559                         code = PRC_PARAMPROB;
560                         break;
561                 default:
562                         goto badcode;
563                 }
564                 goto deliver;
565                 break;
566
567         case ICMP6_ECHO_REQUEST:
568                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
569                 if (code != 0)
570                         goto badcode;
571                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
572                         /* Give up remote */
573                         break;
574                 }
575                 if ((n->m_flags & M_EXT) != 0
576                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
577                         struct mbuf *n0 = n;
578                         int n0len;
579
580                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
581                         n = m_gethdr(M_NOWAIT, n0->m_type);
582                         if (n == NULL) {
583                                 /* Give up remote */
584                                 m_freem(n0);
585                                 break;
586                         }
587
588                         m_move_pkthdr(n, n0);   /* FIB copied. */
589                         n0len = n0->m_pkthdr.len;       /* save for use below */
590                         /*
591                          * Copy IPv6 and ICMPv6 only.
592                          */
593                         nip6 = mtod(n, struct ip6_hdr *);
594                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
595                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
596                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
597                         noff = sizeof(struct ip6_hdr);
598                         /* new mbuf contains only ipv6+icmpv6 headers */
599                         n->m_len = noff + sizeof(struct icmp6_hdr);
600                         /*
601                          * Adjust mbuf.  ip6_plen will be adjusted in
602                          * ip6_output().
603                          */
604                         m_adj(n0, off + sizeof(struct icmp6_hdr));
605                         /* recalculate complete packet size */
606                         n->m_pkthdr.len = n0len + (noff - off);
607                         n->m_next = n0;
608                 } else {
609                         nip6 = mtod(n, struct ip6_hdr *);
610                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
611                             sizeof(*nicmp6));
612                         noff = off;
613                 }
614                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
615                 nicmp6->icmp6_code = 0;
616                 if (n) {
617                         ICMP6STAT_INC(icp6s_reflect);
618                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
619                         icmp6_reflect(n, noff);
620                 }
621                 break;
622
623         case ICMP6_ECHO_REPLY:
624                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
625                 if (code != 0)
626                         goto badcode;
627                 break;
628
629         case MLD_LISTENER_QUERY:
630         case MLD_LISTENER_REPORT:
631         case MLD_LISTENER_DONE:
632         case MLDV2_LISTENER_REPORT:
633                 /*
634                  * Drop MLD traffic which is not link-local, has a hop limit
635                  * of greater than 1 hop, or which does not have the
636                  * IPv6 HBH Router Alert option.
637                  * As IPv6 HBH options are stripped in ip6_input() we must
638                  * check an mbuf header flag.
639                  * XXX Should we also sanity check that these messages
640                  * were directed to a link-local multicast prefix?
641                  */
642                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
643                         goto freeit;
644                 if (mld_input(m, off, icmp6len) != 0)
645                         return (IPPROTO_DONE);
646                 /* m stays. */
647                 break;
648
649         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
650             {
651                 enum { WRU, FQDN } mode;
652
653                 if (!V_icmp6_nodeinfo)
654                         break;
655
656                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
657                         mode = WRU;
658                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
659                         mode = FQDN;
660                 else
661                         goto badlen;
662
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                         struct prison *pr;
675                         u_char *p;
676                         int maxhlen, hlen;
677
678                         /*
679                          * XXX: this combination of flags is pointless,
680                          * but should we keep this for compatibility?
681                          */
682                         if ((V_icmp6_nodeinfo & 5) != 5)
683                                 break;
684
685                         if (code != 0)
686                                 goto badcode;
687
688                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
689                         n = m_gethdr(M_NOWAIT, m->m_type);
690                         if (n == NULL) {
691                                 /* Give up remote */
692                                 break;
693                         }
694                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
695                                 /*
696                                  * Previous code did a blind M_COPY_PKTHDR
697                                  * and said "just for rcvif".  If true, then
698                                  * we could tolerate the dup failing (due to
699                                  * the deep copy of the tag chain).  For now
700                                  * be conservative and just fail.
701                                  */
702                                 m_free(n);
703                                 n = NULL;
704                         }
705                         maxhlen = M_TRAILINGSPACE(n) -
706                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
707                         pr = curthread->td_ucred->cr_prison;
708                         mtx_lock(&pr->pr_mtx);
709                         hlen = strlen(pr->pr_hostname);
710                         if (maxhlen > hlen)
711                                 maxhlen = hlen;
712                         /*
713                          * Copy IPv6 and ICMPv6 only.
714                          */
715                         nip6 = mtod(n, struct ip6_hdr *);
716                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
717                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
718                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
719                         p = (u_char *)(nicmp6 + 1);
720                         bzero(p, 4);
721                         /* meaningless TTL */
722                         bcopy(pr->pr_hostname, p + 4, maxhlen);
723                         mtx_unlock(&pr->pr_mtx);
724                         noff = sizeof(struct ip6_hdr);
725                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
726                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
727                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
728                         nicmp6->icmp6_code = 0;
729                 }
730                 if (n) {
731                         ICMP6STAT_INC(icp6s_reflect);
732                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
733                         icmp6_reflect(n, noff);
734                 }
735                 break;
736             }
737
738         case ICMP6_WRUREPLY:
739                 if (code != 0)
740                         goto badcode;
741                 break;
742
743         case ND_ROUTER_SOLICIT:
744                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
745                 if (code != 0)
746                         goto badcode;
747                 if (icmp6len < sizeof(struct nd_router_solicit))
748                         goto badlen;
749                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
750                         /* give up local */
751
752                         /* Send incoming SeND packet to user space. */
753                         if (send_sendso_input_hook != NULL) {
754                                 IP6_EXTHDR_CHECK(m, off,
755                                     icmp6len, IPPROTO_DONE);
756                                 error = send_sendso_input_hook(m, ifp,
757                                     SND_IN, ip6len);
758                                 /* -1 == no app on SEND socket */
759                                 if (error == 0)
760                                         return (IPPROTO_DONE);
761                                 nd6_rs_input(m, off, icmp6len);
762                         } else
763                                 nd6_rs_input(m, off, icmp6len);
764                         m = NULL;
765                         goto freeit;
766                 }
767                 if (send_sendso_input_hook != NULL) {
768                         IP6_EXTHDR_CHECK(n, off,
769                             icmp6len, IPPROTO_DONE);
770                         error = send_sendso_input_hook(n, ifp,
771                             SND_IN, ip6len);
772                         if (error == 0)
773                                 goto freeit;
774                         /* -1 == no app on SEND socket */
775                         nd6_rs_input(n, off, icmp6len);
776                 } else
777                         nd6_rs_input(n, off, icmp6len);
778                 /* m stays. */
779                 break;
780
781         case ND_ROUTER_ADVERT:
782                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
783                 if (code != 0)
784                         goto badcode;
785                 if (icmp6len < sizeof(struct nd_router_advert))
786                         goto badlen;
787                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
788
789                         /* Send incoming SeND-protected/ND packet to user space. */
790                         if (send_sendso_input_hook != NULL) {
791                                 error = send_sendso_input_hook(m, ifp,
792                                     SND_IN, ip6len);
793                                 if (error == 0)
794                                         return (IPPROTO_DONE);
795                                 nd6_ra_input(m, off, icmp6len);
796                         } else
797                                 nd6_ra_input(m, off, icmp6len);
798                         m = NULL;
799                         goto freeit;
800                 }
801                 if (send_sendso_input_hook != NULL) {
802                         error = send_sendso_input_hook(n, ifp,
803                             SND_IN, ip6len);
804                         if (error == 0)
805                                 goto freeit;
806                         nd6_ra_input(n, off, icmp6len);
807                 } else
808                         nd6_ra_input(n, off, icmp6len);
809                 /* m stays. */
810                 break;
811
812         case ND_NEIGHBOR_SOLICIT:
813                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
814                 if (code != 0)
815                         goto badcode;
816                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
817                         goto badlen;
818                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
819                         if (send_sendso_input_hook != NULL) {
820                                 error = send_sendso_input_hook(m, ifp,
821                                     SND_IN, ip6len);
822                                 if (error == 0)
823                                         return (IPPROTO_DONE);
824                                 nd6_ns_input(m, off, icmp6len);
825                         } else
826                                 nd6_ns_input(m, off, icmp6len);
827                         m = NULL;
828                         goto freeit;
829                 }
830                 if (send_sendso_input_hook != NULL) {
831                         error = send_sendso_input_hook(n, ifp,
832                             SND_IN, ip6len);
833                         if (error == 0)
834                                 goto freeit;
835                         nd6_ns_input(n, off, icmp6len);
836                 } else
837                         nd6_ns_input(n, off, icmp6len);
838                 /* m stays. */
839                 break;
840
841         case ND_NEIGHBOR_ADVERT:
842                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
843                 if (code != 0)
844                         goto badcode;
845                 if (icmp6len < sizeof(struct nd_neighbor_advert))
846                         goto badlen;
847                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
848
849                         /* Send incoming SeND-protected/ND packet to user space. */
850                         if (send_sendso_input_hook != NULL) {
851                                 error = send_sendso_input_hook(m, ifp,
852                                     SND_IN, ip6len);
853                                 if (error == 0)
854                                         return (IPPROTO_DONE);
855                                 nd6_na_input(m, off, icmp6len);
856                         } else
857                                 nd6_na_input(m, off, icmp6len);
858                         m = NULL;
859                         goto freeit;
860                 }
861                 if (send_sendso_input_hook != NULL) {
862                         error = send_sendso_input_hook(n, ifp,
863                             SND_IN, ip6len);
864                         if (error == 0)
865                                 goto freeit;
866                         nd6_na_input(n, off, icmp6len);
867                 } else
868                         nd6_na_input(n, off, icmp6len);
869                 /* m stays. */
870                 break;
871
872         case ND_REDIRECT:
873                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
874                 if (code != 0)
875                         goto badcode;
876                 if (icmp6len < sizeof(struct nd_redirect))
877                         goto badlen;
878                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
879                         if (send_sendso_input_hook != NULL) {
880                                 error = send_sendso_input_hook(m, ifp,
881                                     SND_IN, ip6len);
882                                 if (error == 0)
883                                         return (IPPROTO_DONE);
884                             icmp6_redirect_input(m, off);
885                         } else
886                                 icmp6_redirect_input(m, off);
887                         m = NULL;
888                         goto freeit;
889                 }
890                 if (send_sendso_input_hook != NULL) {
891                         error = send_sendso_input_hook(n, ifp,
892                             SND_IN, ip6len);
893                         if (error == 0)
894                                 goto freeit;
895                         icmp6_redirect_input(n, off);
896                 } else
897                         icmp6_redirect_input(n, off);
898                 /* m stays. */
899                 break;
900
901         case ICMP6_ROUTER_RENUMBERING:
902                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
903                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
904                         goto badcode;
905                 if (icmp6len < sizeof(struct icmp6_router_renum))
906                         goto badlen;
907                 break;
908
909         default:
910                 nd6log((LOG_DEBUG,
911                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
912                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
913                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
914                     ifp ? ifp->if_index : 0));
915                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
916                         /* ICMPv6 error: MUST deliver it by spec... */
917                         code = PRC_NCMDS;
918                         /* deliver */
919                 } else {
920                         /* ICMPv6 informational: MUST not deliver */
921                         break;
922                 }
923         deliver:
924                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
925                         /* In this case, m should've been freed. */
926                         return (IPPROTO_DONE);
927                 }
928                 break;
929
930         badcode:
931                 ICMP6STAT_INC(icp6s_badcode);
932                 break;
933
934         badlen:
935                 ICMP6STAT_INC(icp6s_badlen);
936                 break;
937         }
938
939         /* deliver the packet to appropriate sockets */
940         icmp6_rip6_input(&m, *offp);
941
942         return IPPROTO_DONE;
943
944  freeit:
945         m_freem(m);
946         return IPPROTO_DONE;
947 }
948
949 static int
950 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
951 {
952         struct mbuf *m = *mp;
953         struct icmp6_hdr *icmp6;
954         struct ip6_hdr *eip6;
955         u_int32_t notifymtu;
956         struct sockaddr_in6 icmp6src, icmp6dst;
957
958         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
959                 ICMP6STAT_INC(icp6s_tooshort);
960                 goto freeit;
961         }
962 #ifndef PULLDOWN_TEST
963         IP6_EXTHDR_CHECK(m, off,
964             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
965         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
966 #else
967         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
968             sizeof(*icmp6) + sizeof(struct ip6_hdr));
969         if (icmp6 == NULL) {
970                 ICMP6STAT_INC(icp6s_tooshort);
971                 return (-1);
972         }
973 #endif
974         eip6 = (struct ip6_hdr *)(icmp6 + 1);
975
976         /* Detect the upper level protocol */
977         {
978                 void (*ctlfunc)(int, struct sockaddr *, void *);
979                 u_int8_t nxt = eip6->ip6_nxt;
980                 int eoff = off + sizeof(struct icmp6_hdr) +
981                     sizeof(struct ip6_hdr);
982                 struct ip6ctlparam ip6cp;
983                 struct in6_addr *finaldst = NULL;
984                 int icmp6type = icmp6->icmp6_type;
985                 struct ip6_frag *fh;
986                 struct ip6_rthdr *rth;
987                 struct ip6_rthdr0 *rth0;
988                 int rthlen;
989
990                 while (1) { /* XXX: should avoid infinite loop explicitly? */
991                         struct ip6_ext *eh;
992
993                         switch (nxt) {
994                         case IPPROTO_HOPOPTS:
995                         case IPPROTO_DSTOPTS:
996                         case IPPROTO_AH:
997 #ifndef PULLDOWN_TEST
998                                 IP6_EXTHDR_CHECK(m, 0,
999                                     eoff + sizeof(struct ip6_ext), -1);
1000                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1001 #else
1002                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1003                                     eoff, sizeof(*eh));
1004                                 if (eh == NULL) {
1005                                         ICMP6STAT_INC(icp6s_tooshort);
1006                                         return (-1);
1007                                 }
1008 #endif
1009
1010                                 if (nxt == IPPROTO_AH)
1011                                         eoff += (eh->ip6e_len + 2) << 2;
1012                                 else
1013                                         eoff += (eh->ip6e_len + 1) << 3;
1014                                 nxt = eh->ip6e_nxt;
1015                                 break;
1016                         case IPPROTO_ROUTING:
1017                                 /*
1018                                  * When the erroneous packet contains a
1019                                  * routing header, we should examine the
1020                                  * header to determine the final destination.
1021                                  * Otherwise, we can't properly update
1022                                  * information that depends on the final
1023                                  * destination (e.g. path MTU).
1024                                  */
1025 #ifndef PULLDOWN_TEST
1026                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1027                                 rth = (struct ip6_rthdr *)
1028                                     (mtod(m, caddr_t) + eoff);
1029 #else
1030                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1031                                     eoff, sizeof(*rth));
1032                                 if (rth == NULL) {
1033                                         ICMP6STAT_INC(icp6s_tooshort);
1034                                         return (-1);
1035                                 }
1036 #endif
1037                                 rthlen = (rth->ip6r_len + 1) << 3;
1038                                 /*
1039                                  * XXX: currently there is no
1040                                  * officially defined type other
1041                                  * than type-0.
1042                                  * Note that if the segment left field
1043                                  * is 0, all intermediate hops must
1044                                  * have been passed.
1045                                  */
1046                                 if (rth->ip6r_segleft &&
1047                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1048                                         int hops;
1049
1050 #ifndef PULLDOWN_TEST
1051                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1052                                         rth0 = (struct ip6_rthdr0 *)
1053                                             (mtod(m, caddr_t) + eoff);
1054 #else
1055                                         IP6_EXTHDR_GET(rth0,
1056                                             struct ip6_rthdr0 *, m,
1057                                             eoff, rthlen);
1058                                         if (rth0 == NULL) {
1059                                                 ICMP6STAT_INC(icp6s_tooshort);
1060                                                 return (-1);
1061                                         }
1062 #endif
1063                                         /* just ignore a bogus header */
1064                                         if ((rth0->ip6r0_len % 2) == 0 &&
1065                                             (hops = rth0->ip6r0_len/2))
1066                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1067                                 }
1068                                 eoff += rthlen;
1069                                 nxt = rth->ip6r_nxt;
1070                                 break;
1071                         case IPPROTO_FRAGMENT:
1072 #ifndef PULLDOWN_TEST
1073                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1074                                     sizeof(struct ip6_frag), -1);
1075                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1076                                     eoff);
1077 #else
1078                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1079                                     eoff, sizeof(*fh));
1080                                 if (fh == NULL) {
1081                                         ICMP6STAT_INC(icp6s_tooshort);
1082                                         return (-1);
1083                                 }
1084 #endif
1085                                 /*
1086                                  * Data after a fragment header is meaningless
1087                                  * unless it is the first fragment, but
1088                                  * we'll go to the notify label for path MTU
1089                                  * discovery.
1090                                  */
1091                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1092                                         goto notify;
1093
1094                                 eoff += sizeof(struct ip6_frag);
1095                                 nxt = fh->ip6f_nxt;
1096                                 break;
1097                         default:
1098                                 /*
1099                                  * This case includes ESP and the No Next
1100                                  * Header.  In such cases going to the notify
1101                                  * label does not have any meaning
1102                                  * (i.e. ctlfunc will be NULL), but we go
1103                                  * anyway since we might have to update
1104                                  * path MTU information.
1105                                  */
1106                                 goto notify;
1107                         }
1108                 }
1109           notify:
1110 #ifndef PULLDOWN_TEST
1111                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1112 #else
1113                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1114                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1115                 if (icmp6 == NULL) {
1116                         ICMP6STAT_INC(icp6s_tooshort);
1117                         return (-1);
1118                 }
1119 #endif
1120
1121                 /*
1122                  * retrieve parameters from the inner IPv6 header, and convert
1123                  * them into sockaddr structures.
1124                  * XXX: there is no guarantee that the source or destination
1125                  * addresses of the inner packet are in the same scope as
1126                  * the addresses of the icmp packet.  But there is no other
1127                  * way to determine the zone.
1128                  */
1129                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1130
1131                 bzero(&icmp6dst, sizeof(icmp6dst));
1132                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1133                 icmp6dst.sin6_family = AF_INET6;
1134                 if (finaldst == NULL)
1135                         icmp6dst.sin6_addr = eip6->ip6_dst;
1136                 else
1137                         icmp6dst.sin6_addr = *finaldst;
1138                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1139                         goto freeit;
1140                 bzero(&icmp6src, sizeof(icmp6src));
1141                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1142                 icmp6src.sin6_family = AF_INET6;
1143                 icmp6src.sin6_addr = eip6->ip6_src;
1144                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1145                         goto freeit;
1146                 icmp6src.sin6_flowinfo =
1147                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1148
1149                 if (finaldst == NULL)
1150                         finaldst = &eip6->ip6_dst;
1151                 ip6cp.ip6c_m = m;
1152                 ip6cp.ip6c_icmp6 = icmp6;
1153                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1154                 ip6cp.ip6c_off = eoff;
1155                 ip6cp.ip6c_finaldst = finaldst;
1156                 ip6cp.ip6c_src = &icmp6src;
1157                 ip6cp.ip6c_nxt = nxt;
1158
1159                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1160                         notifymtu = ntohl(icmp6->icmp6_mtu);
1161                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1162                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1163                 }
1164
1165                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1166                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1167                 if (ctlfunc) {
1168                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1169                             &ip6cp);
1170                 }
1171         }
1172         *mp = m;
1173         return (0);
1174
1175   freeit:
1176         m_freem(m);
1177         return (-1);
1178 }
1179
1180 void
1181 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1182 {
1183         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1184         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1185         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1186         u_int mtu = ntohl(icmp6->icmp6_mtu);
1187         struct in_conninfo inc;
1188
1189 #if 0
1190         /*
1191          * RFC2460 section 5, last paragraph.
1192          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1193          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1194          * due to packet translator in the middle.
1195          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1196          * special handling.
1197          */
1198         if (mtu < IPV6_MMTU)
1199                 return;
1200 #endif
1201
1202         /*
1203          * we reject ICMPv6 too big with abnormally small value.
1204          * XXX what is the good definition of "abnormally small"?
1205          */
1206         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1207                 return;
1208
1209         if (!validated)
1210                 return;
1211
1212         /*
1213          * In case the suggested mtu is less than IPV6_MMTU, we
1214          * only need to remember that it was for above mentioned
1215          * "alwaysfrag" case.
1216          * Try to be as close to the spec as possible.
1217          */
1218         if (mtu < IPV6_MMTU)
1219                 mtu = IPV6_MMTU - 8;
1220
1221         bzero(&inc, sizeof(inc));
1222         inc.inc_fibnum = M_GETFIB(m);
1223         inc.inc_flags |= INC_ISIPV6;
1224         inc.inc6_faddr = *dst;
1225         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1226                 return;
1227
1228         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1229                 tcp_hc_updatemtu(&inc, mtu);
1230                 ICMP6STAT_INC(icp6s_pmtuchg);
1231         }
1232 }
1233
1234 /*
1235  * Process a Node Information Query packet, based on
1236  * draft-ietf-ipngwg-icmp-name-lookups-07.
1237  *
1238  * Spec incompatibilities:
1239  * - IPv6 Subject address handling
1240  * - IPv4 Subject address handling support missing
1241  * - Proxy reply (answer even if it's not for me)
1242  * - joins NI group address at in6_ifattach() time only, does not cope
1243  *   with hostname changes by sethostname(3)
1244  */
1245 static struct mbuf *
1246 ni6_input(struct mbuf *m, int off)
1247 {
1248         struct icmp6_nodeinfo *ni6, *nni6;
1249         struct mbuf *n = NULL;
1250         struct prison *pr;
1251         u_int16_t qtype;
1252         int subjlen;
1253         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1254         struct ni_reply_fqdn *fqdn;
1255         int addrs;              /* for NI_QTYPE_NODEADDR */
1256         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1257         struct in6_addr in6_subj; /* subject address */
1258         struct ip6_hdr *ip6;
1259         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1260         char *subj = NULL;
1261         struct in6_ifaddr *ia6 = NULL;
1262
1263         ip6 = mtod(m, struct ip6_hdr *);
1264 #ifndef PULLDOWN_TEST
1265         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1266 #else
1267         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1268         if (ni6 == NULL) {
1269                 /* m is already reclaimed */
1270                 return (NULL);
1271         }
1272 #endif
1273
1274         /*
1275          * Validate IPv6 source address.
1276          * The default configuration MUST be to refuse answering queries from
1277          * global-scope addresses according to RFC4602.
1278          * Notes:
1279          *  - it's not very clear what "refuse" means; this implementation
1280          *    simply drops it.
1281          *  - it's not very easy to identify global-scope (unicast) addresses
1282          *    since there are many prefixes for them.  It should be safer
1283          *    and in practice sufficient to check "all" but loopback and
1284          *    link-local (note that site-local unicast was deprecated and
1285          *    ULA is defined as global scope-wise)
1286          */
1287         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1288             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1289             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1290                 goto bad;
1291
1292         /*
1293          * Validate IPv6 destination address.
1294          *
1295          * The Responder must discard the Query without further processing
1296          * unless it is one of the Responder's unicast or anycast addresses, or
1297          * a link-local scope multicast address which the Responder has joined.
1298          * [RFC4602, Section 5.]
1299          */
1300         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1301                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1302                         goto bad;
1303                 /* else it's a link-local multicast, fine */
1304         } else {                /* unicast or anycast */
1305                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1306                         goto bad; /* XXX impossible */
1307
1308                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1309                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1310                         ifa_free(&ia6->ia_ifa);
1311                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1312                                 "a temporary address in %s:%d",
1313                                __FILE__, __LINE__));
1314                         goto bad;
1315                 }
1316                 ifa_free(&ia6->ia_ifa);
1317         }
1318
1319         /* validate query Subject field. */
1320         qtype = ntohs(ni6->ni_qtype);
1321         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1322         switch (qtype) {
1323         case NI_QTYPE_NOOP:
1324         case NI_QTYPE_SUPTYPES:
1325                 /* 07 draft */
1326                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1327                         break;
1328                 /* FALLTHROUGH */
1329         case NI_QTYPE_FQDN:
1330         case NI_QTYPE_NODEADDR:
1331         case NI_QTYPE_IPV4ADDR:
1332                 switch (ni6->ni_code) {
1333                 case ICMP6_NI_SUBJ_IPV6:
1334 #if ICMP6_NI_SUBJ_IPV6 != 0
1335                 case 0:
1336 #endif
1337                         /*
1338                          * backward compatibility - try to accept 03 draft
1339                          * format, where no Subject is present.
1340                          */
1341                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1342                             subjlen == 0) {
1343                                 oldfqdn++;
1344                                 break;
1345                         }
1346 #if ICMP6_NI_SUBJ_IPV6 != 0
1347                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1348                                 goto bad;
1349 #endif
1350
1351                         if (subjlen != sizeof(struct in6_addr))
1352                                 goto bad;
1353
1354                         /*
1355                          * Validate Subject address.
1356                          *
1357                          * Not sure what exactly "address belongs to the node"
1358                          * means in the spec, is it just unicast, or what?
1359                          *
1360                          * At this moment we consider Subject address as
1361                          * "belong to the node" if the Subject address equals
1362                          * to the IPv6 destination address; validation for
1363                          * IPv6 destination address should have done enough
1364                          * check for us.
1365                          *
1366                          * We do not do proxy at this moment.
1367                          */
1368                         /* m_pulldown instead of copy? */
1369                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1370                             subjlen, (caddr_t)&in6_subj);
1371                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1372                                 goto bad;
1373
1374                         subj = (char *)&in6_subj;
1375                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1376                                 break;
1377
1378                         /*
1379                          * XXX if we are to allow other cases, we should really
1380                          * be careful about scope here.
1381                          * basically, we should disallow queries toward IPv6
1382                          * destination X with subject Y,
1383                          * if scope(X) > scope(Y).
1384                          * if we allow scope(X) > scope(Y), it will result in
1385                          * information leakage across scope boundary.
1386                          */
1387                         goto bad;
1388
1389                 case ICMP6_NI_SUBJ_FQDN:
1390                         /*
1391                          * Validate Subject name with gethostname(3).
1392                          *
1393                          * The behavior may need some debate, since:
1394                          * - we are not sure if the node has FQDN as
1395                          *   hostname (returned by gethostname(3)).
1396                          * - the code does wildcard match for truncated names.
1397                          *   however, we are not sure if we want to perform
1398                          *   wildcard match, if gethostname(3) side has
1399                          *   truncated hostname.
1400                          */
1401                         pr = curthread->td_ucred->cr_prison;
1402                         mtx_lock(&pr->pr_mtx);
1403                         n = ni6_nametodns(pr->pr_hostname,
1404                             strlen(pr->pr_hostname), 0);
1405                         mtx_unlock(&pr->pr_mtx);
1406                         if (!n || n->m_next || n->m_len == 0)
1407                                 goto bad;
1408                         IP6_EXTHDR_GET(subj, char *, m,
1409                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1410                         if (subj == NULL)
1411                                 goto bad;
1412                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1413                             n->m_len)) {
1414                                 goto bad;
1415                         }
1416                         m_freem(n);
1417                         n = NULL;
1418                         break;
1419
1420                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1421                 default:
1422                         goto bad;
1423                 }
1424                 break;
1425         }
1426
1427         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1428         switch (qtype) {
1429         case NI_QTYPE_FQDN:
1430                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1431                         goto bad;
1432                 break;
1433         case NI_QTYPE_NODEADDR:
1434         case NI_QTYPE_IPV4ADDR:
1435                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1436                         goto bad;
1437                 break;
1438         }
1439
1440         /* guess reply length */
1441         switch (qtype) {
1442         case NI_QTYPE_NOOP:
1443                 break;          /* no reply data */
1444         case NI_QTYPE_SUPTYPES:
1445                 replylen += sizeof(u_int32_t);
1446                 break;
1447         case NI_QTYPE_FQDN:
1448                 /* XXX will append an mbuf */
1449                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1450                 break;
1451         case NI_QTYPE_NODEADDR:
1452                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1453                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1454                     sizeof(u_int32_t))) > MCLBYTES)
1455                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1456                 break;
1457         case NI_QTYPE_IPV4ADDR:
1458                 /* unsupported - should respond with unknown Qtype? */
1459                 break;
1460         default:
1461                 /*
1462                  * XXX: We must return a reply with the ICMP6 code
1463                  * `unknown Qtype' in this case.  However we regard the case
1464                  * as an FQDN query for backward compatibility.
1465                  * Older versions set a random value to this field,
1466                  * so it rarely varies in the defined qtypes.
1467                  * But the mechanism is not reliable...
1468                  * maybe we should obsolete older versions.
1469                  */
1470                 qtype = NI_QTYPE_FQDN;
1471                 /* XXX will append an mbuf */
1472                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1473                 oldfqdn++;
1474                 break;
1475         }
1476
1477         /* Allocate an mbuf to reply. */
1478         if (replylen > MCLBYTES) {
1479                 /*
1480                  * XXX: should we try to allocate more? But MCLBYTES
1481                  * is probably much larger than IPV6_MMTU...
1482                  */
1483                 goto bad;
1484         }
1485         if (replylen > MHLEN)
1486                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1487         else
1488                 n = m_gethdr(M_NOWAIT, m->m_type);
1489         if (n == NULL) {
1490                 m_freem(m);
1491                 return (NULL);
1492         }
1493         m_move_pkthdr(n, m); /* just for recvif and FIB */
1494         n->m_pkthdr.len = n->m_len = replylen;
1495
1496         /* copy mbuf header and IPv6 + Node Information base headers */
1497         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1498         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1499         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1500
1501         /* qtype dependent procedure */
1502         switch (qtype) {
1503         case NI_QTYPE_NOOP:
1504                 nni6->ni_code = ICMP6_NI_SUCCESS;
1505                 nni6->ni_flags = 0;
1506                 break;
1507         case NI_QTYPE_SUPTYPES:
1508         {
1509                 u_int32_t v;
1510                 nni6->ni_code = ICMP6_NI_SUCCESS;
1511                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1512                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1513                 v = (u_int32_t)htonl(0x0000000f);
1514                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1515                 break;
1516         }
1517         case NI_QTYPE_FQDN:
1518                 nni6->ni_code = ICMP6_NI_SUCCESS;
1519                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1520                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1521                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1522                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1523                 /*
1524                  * XXX do we really have FQDN in hostname?
1525                  */
1526                 pr = curthread->td_ucred->cr_prison;
1527                 mtx_lock(&pr->pr_mtx);
1528                 n->m_next = ni6_nametodns(pr->pr_hostname,
1529                     strlen(pr->pr_hostname), oldfqdn);
1530                 mtx_unlock(&pr->pr_mtx);
1531                 if (n->m_next == NULL)
1532                         goto bad;
1533                 /* XXX we assume that n->m_next is not a chain */
1534                 if (n->m_next->m_next != NULL)
1535                         goto bad;
1536                 n->m_pkthdr.len += n->m_next->m_len;
1537                 break;
1538         case NI_QTYPE_NODEADDR:
1539         {
1540                 int lenlim, copied;
1541
1542                 nni6->ni_code = ICMP6_NI_SUCCESS;
1543                 n->m_pkthdr.len = n->m_len =
1544                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1545                 lenlim = M_TRAILINGSPACE(n);
1546                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1547                 /* XXX: reset mbuf length */
1548                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1549                     sizeof(struct icmp6_nodeinfo) + copied;
1550                 break;
1551         }
1552         default:
1553                 break;          /* XXX impossible! */
1554         }
1555
1556         nni6->ni_type = ICMP6_NI_REPLY;
1557         m_freem(m);
1558         return (n);
1559
1560   bad:
1561         m_freem(m);
1562         if (n)
1563                 m_freem(n);
1564         return (NULL);
1565 }
1566
1567 /*
1568  * make a mbuf with DNS-encoded string.  no compression support.
1569  *
1570  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1571  * treated as truncated name (two \0 at the end).  this is a wild guess.
1572  *
1573  * old - return pascal string if non-zero
1574  */
1575 static struct mbuf *
1576 ni6_nametodns(const char *name, int namelen, int old)
1577 {
1578         struct mbuf *m;
1579         char *cp, *ep;
1580         const char *p, *q;
1581         int i, len, nterm;
1582
1583         if (old)
1584                 len = namelen + 1;
1585         else
1586                 len = MCLBYTES;
1587
1588         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1589         if (len > MLEN)
1590                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1591         else
1592                 m = m_get(M_NOWAIT, MT_DATA);
1593         if (m == NULL)
1594                 goto fail;
1595
1596         if (old) {
1597                 m->m_len = len;
1598                 *mtod(m, char *) = namelen;
1599                 bcopy(name, mtod(m, char *) + 1, namelen);
1600                 return m;
1601         } else {
1602                 m->m_len = 0;
1603                 cp = mtod(m, char *);
1604                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1605
1606                 /* if not certain about my name, return empty buffer */
1607                 if (namelen == 0)
1608                         return m;
1609
1610                 /*
1611                  * guess if it looks like shortened hostname, or FQDN.
1612                  * shortened hostname needs two trailing "\0".
1613                  */
1614                 i = 0;
1615                 for (p = name; p < name + namelen; p++) {
1616                         if (*p && *p == '.')
1617                                 i++;
1618                 }
1619                 if (i < 2)
1620                         nterm = 2;
1621                 else
1622                         nterm = 1;
1623
1624                 p = name;
1625                 while (cp < ep && p < name + namelen) {
1626                         i = 0;
1627                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1628                                 i++;
1629                         /* result does not fit into mbuf */
1630                         if (cp + i + 1 >= ep)
1631                                 goto fail;
1632                         /*
1633                          * DNS label length restriction, RFC1035 page 8.
1634                          * "i == 0" case is included here to avoid returning
1635                          * 0-length label on "foo..bar".
1636                          */
1637                         if (i <= 0 || i >= 64)
1638                                 goto fail;
1639                         *cp++ = i;
1640                         bcopy(p, cp, i);
1641                         cp += i;
1642                         p = q;
1643                         if (p < name + namelen && *p == '.')
1644                                 p++;
1645                 }
1646                 /* termination */
1647                 if (cp + nterm >= ep)
1648                         goto fail;
1649                 while (nterm-- > 0)
1650                         *cp++ = '\0';
1651                 m->m_len = cp - mtod(m, char *);
1652                 return m;
1653         }
1654
1655         panic("should not reach here");
1656         /* NOTREACHED */
1657
1658  fail:
1659         if (m)
1660                 m_freem(m);
1661         return NULL;
1662 }
1663
1664 /*
1665  * check if two DNS-encoded string matches.  takes care of truncated
1666  * form (with \0\0 at the end).  no compression support.
1667  * XXX upper/lowercase match (see RFC2065)
1668  */
1669 static int
1670 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1671 {
1672         const char *a0, *b0;
1673         int l;
1674
1675         /* simplest case - need validation? */
1676         if (alen == blen && bcmp(a, b, alen) == 0)
1677                 return 1;
1678
1679         a0 = a;
1680         b0 = b;
1681
1682         /* termination is mandatory */
1683         if (alen < 2 || blen < 2)
1684                 return 0;
1685         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1686                 return 0;
1687         alen--;
1688         blen--;
1689
1690         while (a - a0 < alen && b - b0 < blen) {
1691                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1692                         return 0;
1693
1694                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1695                         return 0;
1696                 /* we don't support compression yet */
1697                 if (a[0] >= 64 || b[0] >= 64)
1698                         return 0;
1699
1700                 /* truncated case */
1701                 if (a[0] == 0 && a - a0 == alen - 1)
1702                         return 1;
1703                 if (b[0] == 0 && b - b0 == blen - 1)
1704                         return 1;
1705                 if (a[0] == 0 || b[0] == 0)
1706                         return 0;
1707
1708                 if (a[0] != b[0])
1709                         return 0;
1710                 l = a[0];
1711                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1712                         return 0;
1713                 if (bcmp(a + 1, b + 1, l) != 0)
1714                         return 0;
1715
1716                 a += 1 + l;
1717                 b += 1 + l;
1718         }
1719
1720         if (a - a0 == alen && b - b0 == blen)
1721                 return 1;
1722         else
1723                 return 0;
1724 }
1725
1726 /*
1727  * calculate the number of addresses to be returned in the node info reply.
1728  */
1729 static int
1730 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1731     struct in6_addr *subj)
1732 {
1733         struct ifnet *ifp;
1734         struct in6_ifaddr *ifa6;
1735         struct ifaddr *ifa;
1736         int addrs = 0, addrsofif, iffound = 0;
1737         int niflags = ni6->ni_flags;
1738
1739         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1740                 switch (ni6->ni_code) {
1741                 case ICMP6_NI_SUBJ_IPV6:
1742                         if (subj == NULL) /* must be impossible... */
1743                                 return (0);
1744                         break;
1745                 default:
1746                         /*
1747                          * XXX: we only support IPv6 subject address for
1748                          * this Qtype.
1749                          */
1750                         return (0);
1751                 }
1752         }
1753
1754         IFNET_RLOCK_NOSLEEP();
1755         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1756                 addrsofif = 0;
1757                 IF_ADDR_RLOCK(ifp);
1758                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1759                         if (ifa->ifa_addr->sa_family != AF_INET6)
1760                                 continue;
1761                         ifa6 = (struct in6_ifaddr *)ifa;
1762
1763                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1764                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1765                                 iffound = 1;
1766
1767                         /*
1768                          * IPv4-mapped addresses can only be returned by a
1769                          * Node Information proxy, since they represent
1770                          * addresses of IPv4-only nodes, which perforce do
1771                          * not implement this protocol.
1772                          * [icmp-name-lookups-07, Section 5.4]
1773                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1774                          * this function at this moment.
1775                          */
1776
1777                         /* What do we have to do about ::1? */
1778                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1779                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1780                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1781                                         continue;
1782                                 break;
1783                         case IPV6_ADDR_SCOPE_SITELOCAL:
1784                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1785                                         continue;
1786                                 break;
1787                         case IPV6_ADDR_SCOPE_GLOBAL:
1788                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1789                                         continue;
1790                                 break;
1791                         default:
1792                                 continue;
1793                         }
1794
1795                         /*
1796                          * check if anycast is okay.
1797                          * XXX: just experimental.  not in the spec.
1798                          */
1799                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1800                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1801                                 continue; /* we need only unicast addresses */
1802                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1803                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1804                                 continue;
1805                         }
1806                         addrsofif++; /* count the address */
1807                 }
1808                 IF_ADDR_RUNLOCK(ifp);
1809                 if (iffound) {
1810                         *ifpp = ifp;
1811                         IFNET_RUNLOCK_NOSLEEP();
1812                         return (addrsofif);
1813                 }
1814
1815                 addrs += addrsofif;
1816         }
1817         IFNET_RUNLOCK_NOSLEEP();
1818
1819         return (addrs);
1820 }
1821
1822 static int
1823 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1824     struct ifnet *ifp0, int resid)
1825 {
1826         struct ifnet *ifp;
1827         struct in6_ifaddr *ifa6;
1828         struct ifaddr *ifa;
1829         struct ifnet *ifp_dep = NULL;
1830         int copied = 0, allow_deprecated = 0;
1831         u_char *cp = (u_char *)(nni6 + 1);
1832         int niflags = ni6->ni_flags;
1833         u_int32_t ltime;
1834
1835         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1836                 return (0);     /* needless to copy */
1837
1838         IFNET_RLOCK_NOSLEEP();
1839         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1840   again:
1841
1842         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1843                 IF_ADDR_RLOCK(ifp);
1844                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1845                         if (ifa->ifa_addr->sa_family != AF_INET6)
1846                                 continue;
1847                         ifa6 = (struct in6_ifaddr *)ifa;
1848
1849                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1850                             allow_deprecated == 0) {
1851                                 /*
1852                                  * prefererred address should be put before
1853                                  * deprecated addresses.
1854                                  */
1855
1856                                 /* record the interface for later search */
1857                                 if (ifp_dep == NULL)
1858                                         ifp_dep = ifp;
1859
1860                                 continue;
1861                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1862                             allow_deprecated != 0)
1863                                 continue; /* we now collect deprecated addrs */
1864
1865                         /* What do we have to do about ::1? */
1866                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1867                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1868                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1869                                         continue;
1870                                 break;
1871                         case IPV6_ADDR_SCOPE_SITELOCAL:
1872                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1873                                         continue;
1874                                 break;
1875                         case IPV6_ADDR_SCOPE_GLOBAL:
1876                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1877                                         continue;
1878                                 break;
1879                         default:
1880                                 continue;
1881                         }
1882
1883                         /*
1884                          * check if anycast is okay.
1885                          * XXX: just experimental.  not in the spec.
1886                          */
1887                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1888                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1889                                 continue;
1890                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1891                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1892                                 continue;
1893                         }
1894
1895                         /* now we can copy the address */
1896                         if (resid < sizeof(struct in6_addr) +
1897                             sizeof(u_int32_t)) {
1898                                 IF_ADDR_RUNLOCK(ifp);
1899                                 /*
1900                                  * We give up much more copy.
1901                                  * Set the truncate flag and return.
1902                                  */
1903                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1904                                 IFNET_RUNLOCK_NOSLEEP();
1905                                 return (copied);
1906                         }
1907
1908                         /*
1909                          * Set the TTL of the address.
1910                          * The TTL value should be one of the following
1911                          * according to the specification:
1912                          *
1913                          * 1. The remaining lifetime of a DHCP lease on the
1914                          *    address, or
1915                          * 2. The remaining Valid Lifetime of a prefix from
1916                          *    which the address was derived through Stateless
1917                          *    Autoconfiguration.
1918                          *
1919                          * Note that we currently do not support stateful
1920                          * address configuration by DHCPv6, so the former
1921                          * case can't happen.
1922                          */
1923                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1924                                 ltime = ND6_INFINITE_LIFETIME;
1925                         else {
1926                                 if (ifa6->ia6_lifetime.ia6t_expire >
1927                                     time_uptime)
1928                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1929                                 else
1930                                         ltime = 0;
1931                         }
1932
1933                         bcopy(&ltime, cp, sizeof(u_int32_t));
1934                         cp += sizeof(u_int32_t);
1935
1936                         /* copy the address itself */
1937                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1938                             sizeof(struct in6_addr));
1939                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1940                         cp += sizeof(struct in6_addr);
1941
1942                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1943                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1944                 }
1945                 IF_ADDR_RUNLOCK(ifp);
1946                 if (ifp0)       /* we need search only on the specified IF */
1947                         break;
1948         }
1949
1950         if (allow_deprecated == 0 && ifp_dep != NULL) {
1951                 ifp = ifp_dep;
1952                 allow_deprecated = 1;
1953
1954                 goto again;
1955         }
1956
1957         IFNET_RUNLOCK_NOSLEEP();
1958
1959         return (copied);
1960 }
1961
1962 /*
1963  * XXX almost dup'ed code with rip6_input.
1964  */
1965 static int
1966 icmp6_rip6_input(struct mbuf **mp, int off)
1967 {
1968         struct mbuf *m = *mp;
1969         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1970         struct inpcb *in6p;
1971         struct inpcb *last = NULL;
1972         struct sockaddr_in6 fromsa;
1973         struct icmp6_hdr *icmp6;
1974         struct mbuf *opts = NULL;
1975
1976 #ifndef PULLDOWN_TEST
1977         /* this is assumed to be safe. */
1978         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1979 #else
1980         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1981         if (icmp6 == NULL) {
1982                 /* m is already reclaimed */
1983                 return (IPPROTO_DONE);
1984         }
1985 #endif
1986
1987         /*
1988          * XXX: the address may have embedded scope zone ID, which should be
1989          * hidden from applications.
1990          */
1991         bzero(&fromsa, sizeof(fromsa));
1992         fromsa.sin6_family = AF_INET6;
1993         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1994         fromsa.sin6_addr = ip6->ip6_src;
1995         if (sa6_recoverscope(&fromsa)) {
1996                 m_freem(m);
1997                 return (IPPROTO_DONE);
1998         }
1999
2000         INP_INFO_RLOCK(&V_ripcbinfo);
2001         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2002                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2003                         continue;
2004                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2005                         continue;
2006                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2007                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2008                         continue;
2009                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2010                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2011                         continue;
2012                 INP_RLOCK(in6p);
2013                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2014                     in6p->in6p_icmp6filt)) {
2015                         INP_RUNLOCK(in6p);
2016                         continue;
2017                 }
2018                 if (last != NULL) {
2019                         struct  mbuf *n = NULL;
2020
2021                         /*
2022                          * Recent network drivers tend to allocate a single
2023                          * mbuf cluster, rather than to make a couple of
2024                          * mbufs without clusters.  Also, since the IPv6 code
2025                          * path tries to avoid m_pullup(), it is highly
2026                          * probable that we still have an mbuf cluster here
2027                          * even though the necessary length can be stored in an
2028                          * mbuf's internal buffer.
2029                          * Meanwhile, the default size of the receive socket
2030                          * buffer for raw sockets is not so large.  This means
2031                          * the possibility of packet loss is relatively higher
2032                          * than before.  To avoid this scenario, we copy the
2033                          * received data to a separate mbuf that does not use
2034                          * a cluster, if possible.
2035                          * XXX: it is better to copy the data after stripping
2036                          * intermediate headers.
2037                          */
2038                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2039                             m->m_len <= MHLEN) {
2040                                 n = m_get(M_NOWAIT, m->m_type);
2041                                 if (n != NULL) {
2042                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2043                                                 bcopy(m->m_data, n->m_data,
2044                                                       m->m_len);
2045                                                 n->m_len = m->m_len;
2046                                         } else {
2047                                                 m_free(n);
2048                                                 n = NULL;
2049                                         }
2050                                 }
2051                         }
2052                         if (n != NULL ||
2053                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2054                                 if (last->inp_flags & INP_CONTROLOPTS)
2055                                         ip6_savecontrol(last, n, &opts);
2056                                 /* strip intermediate headers */
2057                                 m_adj(n, off);
2058                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2059                                 if (sbappendaddr_locked(
2060                                     &last->inp_socket->so_rcv,
2061                                     (struct sockaddr *)&fromsa, n, opts)
2062                                     == 0) {
2063                                         /* should notify about lost packet */
2064                                         m_freem(n);
2065                                         if (opts) {
2066                                                 m_freem(opts);
2067                                         }
2068                                         SOCKBUF_UNLOCK(
2069                                             &last->inp_socket->so_rcv);
2070                                 } else
2071                                         sorwakeup_locked(last->inp_socket);
2072                                 opts = NULL;
2073                         }
2074                         INP_RUNLOCK(last);
2075                 }
2076                 last = in6p;
2077         }
2078         INP_INFO_RUNLOCK(&V_ripcbinfo);
2079         if (last != NULL) {
2080                 if (last->inp_flags & INP_CONTROLOPTS)
2081                         ip6_savecontrol(last, m, &opts);
2082                 /* strip intermediate headers */
2083                 m_adj(m, off);
2084
2085                 /* avoid using mbuf clusters if possible (see above) */
2086                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2087                     m->m_len <= MHLEN) {
2088                         struct mbuf *n;
2089
2090                         n = m_get(M_NOWAIT, m->m_type);
2091                         if (n != NULL) {
2092                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2093                                         bcopy(m->m_data, n->m_data, m->m_len);
2094                                         n->m_len = m->m_len;
2095
2096                                         m_freem(m);
2097                                         m = n;
2098                                 } else {
2099                                         m_freem(n);
2100                                         n = NULL;
2101                                 }
2102                         }
2103                 }
2104                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2105                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2106                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2107                         m_freem(m);
2108                         if (opts)
2109                                 m_freem(opts);
2110                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2111                 } else
2112                         sorwakeup_locked(last->inp_socket);
2113                 INP_RUNLOCK(last);
2114         } else {
2115                 m_freem(m);
2116                 IP6STAT_DEC(ip6s_delivered);
2117         }
2118         return IPPROTO_DONE;
2119 }
2120
2121 /*
2122  * Reflect the ip6 packet back to the source.
2123  * OFF points to the icmp6 header, counted from the top of the mbuf.
2124  */
2125 void
2126 icmp6_reflect(struct mbuf *m, size_t off)
2127 {
2128         struct ip6_hdr *ip6;
2129         struct icmp6_hdr *icmp6;
2130         struct in6_ifaddr *ia = NULL;
2131         int plen;
2132         int type, code;
2133         struct ifnet *outif = NULL;
2134         struct in6_addr origdst, src, *srcp = NULL;
2135
2136         /* too short to reflect */
2137         if (off < sizeof(struct ip6_hdr)) {
2138                 nd6log((LOG_DEBUG,
2139                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2140                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2141                     __FILE__, __LINE__));
2142                 goto bad;
2143         }
2144
2145         /*
2146          * If there are extra headers between IPv6 and ICMPv6, strip
2147          * off that header first.
2148          */
2149 #ifdef DIAGNOSTIC
2150         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2151                 panic("assumption failed in icmp6_reflect");
2152 #endif
2153         if (off > sizeof(struct ip6_hdr)) {
2154                 size_t l;
2155                 struct ip6_hdr nip6;
2156
2157                 l = off - sizeof(struct ip6_hdr);
2158                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2159                 m_adj(m, l);
2160                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2161                 if (m->m_len < l) {
2162                         if ((m = m_pullup(m, l)) == NULL)
2163                                 return;
2164                 }
2165                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2166         } else /* off == sizeof(struct ip6_hdr) */ {
2167                 size_t l;
2168                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2169                 if (m->m_len < l) {
2170                         if ((m = m_pullup(m, l)) == NULL)
2171                                 return;
2172                 }
2173         }
2174         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2175         ip6 = mtod(m, struct ip6_hdr *);
2176         ip6->ip6_nxt = IPPROTO_ICMPV6;
2177         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2178         type = icmp6->icmp6_type; /* keep type for statistics */
2179         code = icmp6->icmp6_code; /* ditto. */
2180
2181         origdst = ip6->ip6_dst;
2182         /*
2183          * ip6_input() drops a packet if its src is multicast.
2184          * So, the src is never multicast.
2185          */
2186         ip6->ip6_dst = ip6->ip6_src;
2187
2188         /*
2189          * If the incoming packet was addressed directly to us (i.e. unicast),
2190          * use dst as the src for the reply.
2191          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2192          * (for example) when we encounter an error while forwarding procedure
2193          * destined to a duplicated address of ours.
2194          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2195          * procedure of an outgoing packet of our own, in which case we need
2196          * to search in the ifaddr list.
2197          */
2198         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2199                 if ((ia = ip6_getdstifaddr(m))) {
2200                         if (!(ia->ia6_flags &
2201                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2202                                 srcp = &ia->ia_addr.sin6_addr;
2203                 } else {
2204                         struct sockaddr_in6 d;
2205
2206                         bzero(&d, sizeof(d));
2207                         d.sin6_family = AF_INET6;
2208                         d.sin6_len = sizeof(d);
2209                         d.sin6_addr = origdst;
2210                         ia = (struct in6_ifaddr *)
2211                             ifa_ifwithaddr((struct sockaddr *)&d);
2212                         if (ia &&
2213                             !(ia->ia6_flags &
2214                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2215                                 srcp = &ia->ia_addr.sin6_addr;
2216                         }
2217                 }
2218         }
2219
2220         if (srcp == NULL) {
2221                 int e;
2222                 struct sockaddr_in6 sin6;
2223                 struct route_in6 ro;
2224
2225                 /*
2226                  * This case matches to multicasts, our anycast, or unicasts
2227                  * that we do not own.  Select a source address based on the
2228                  * source address of the erroneous packet.
2229                  */
2230                 bzero(&sin6, sizeof(sin6));
2231                 sin6.sin6_family = AF_INET6;
2232                 sin6.sin6_len = sizeof(sin6);
2233                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2234
2235                 bzero(&ro, sizeof(ro));
2236                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2237                 if (ro.ro_rt)
2238                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2239                 if (e) {
2240                         char ip6buf[INET6_ADDRSTRLEN];
2241                         nd6log((LOG_DEBUG,
2242                             "icmp6_reflect: source can't be determined: "
2243                             "dst=%s, error=%d\n",
2244                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2245                         goto bad;
2246                 }
2247                 srcp = &src;
2248         }
2249
2250         ip6->ip6_src = *srcp;
2251         ip6->ip6_flow = 0;
2252         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2253         ip6->ip6_vfc |= IPV6_VERSION;
2254         ip6->ip6_nxt = IPPROTO_ICMPV6;
2255         if (outif)
2256                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2257         else if (m->m_pkthdr.rcvif) {
2258                 /* XXX: This may not be the outgoing interface */
2259                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2260         } else
2261                 ip6->ip6_hlim = V_ip6_defhlim;
2262
2263         icmp6->icmp6_cksum = 0;
2264         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2265             sizeof(struct ip6_hdr), plen);
2266
2267         /*
2268          * XXX option handling
2269          */
2270
2271         m->m_flags &= ~(M_BCAST|M_MCAST);
2272
2273         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2274         if (outif)
2275                 icmp6_ifoutstat_inc(outif, type, code);
2276
2277         if (ia != NULL)
2278                 ifa_free(&ia->ia_ifa);
2279         return;
2280
2281  bad:
2282         if (ia != NULL)
2283                 ifa_free(&ia->ia_ifa);
2284         m_freem(m);
2285         return;
2286 }
2287
2288 void
2289 icmp6_fasttimo(void)
2290 {
2291
2292         mld_fasttimo();
2293 }
2294
2295 void
2296 icmp6_slowtimo(void)
2297 {
2298
2299         mld_slowtimo();
2300 }
2301
2302 static const char *
2303 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2304     struct in6_addr *tgt6)
2305 {
2306         static char buf[1024];
2307         char ip6bufs[INET6_ADDRSTRLEN];
2308         char ip6bufd[INET6_ADDRSTRLEN];
2309         char ip6buft[INET6_ADDRSTRLEN];
2310         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2311             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2312             ip6_sprintf(ip6buft, tgt6));
2313         return buf;
2314 }
2315
2316 void
2317 icmp6_redirect_input(struct mbuf *m, int off)
2318 {
2319         struct ifnet *ifp;
2320         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2321         struct nd_redirect *nd_rd;
2322         int icmp6len = ntohs(ip6->ip6_plen);
2323         char *lladdr = NULL;
2324         int lladdrlen = 0;
2325         struct rtentry *rt = NULL;
2326         int is_router;
2327         int is_onlink;
2328         struct in6_addr src6 = ip6->ip6_src;
2329         struct in6_addr redtgt6;
2330         struct in6_addr reddst6;
2331         union nd_opts ndopts;
2332         char ip6buf[INET6_ADDRSTRLEN];
2333
2334         M_ASSERTPKTHDR(m);
2335         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2336
2337         ifp = m->m_pkthdr.rcvif;
2338
2339         /* XXX if we are router, we don't update route by icmp6 redirect */
2340         if (V_ip6_forwarding)
2341                 goto freeit;
2342         if (!V_icmp6_rediraccept)
2343                 goto freeit;
2344
2345 #ifndef PULLDOWN_TEST
2346         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2347         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2348 #else
2349         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2350         if (nd_rd == NULL) {
2351                 ICMP6STAT_INC(icp6s_tooshort);
2352                 return;
2353         }
2354 #endif
2355         redtgt6 = nd_rd->nd_rd_target;
2356         reddst6 = nd_rd->nd_rd_dst;
2357
2358         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2359             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2360                 goto freeit;
2361         }
2362
2363         /* validation */
2364         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2365                 nd6log((LOG_ERR,
2366                     "ICMP6 redirect sent from %s rejected; "
2367                     "must be from linklocal\n",
2368                     ip6_sprintf(ip6buf, &src6)));
2369                 goto bad;
2370         }
2371         if (ip6->ip6_hlim != 255) {
2372                 nd6log((LOG_ERR,
2373                     "ICMP6 redirect sent from %s rejected; "
2374                     "hlim=%d (must be 255)\n",
2375                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2376                 goto bad;
2377         }
2378     {
2379         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2380         struct sockaddr_in6 sin6;
2381         struct in6_addr *gw6;
2382
2383         bzero(&sin6, sizeof(sin6));
2384         sin6.sin6_family = AF_INET6;
2385         sin6.sin6_len = sizeof(struct sockaddr_in6);
2386         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2387         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2388         if (rt) {
2389                 if (rt->rt_gateway == NULL ||
2390                     rt->rt_gateway->sa_family != AF_INET6) {
2391                         RTFREE_LOCKED(rt);
2392                         nd6log((LOG_ERR,
2393                             "ICMP6 redirect rejected; no route "
2394                             "with inet6 gateway found for redirect dst: %s\n",
2395                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2396                         goto bad;
2397                 }
2398
2399                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2400                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2401                         RTFREE_LOCKED(rt);
2402                         nd6log((LOG_ERR,
2403                             "ICMP6 redirect rejected; "
2404                             "not equal to gw-for-src=%s (must be same): "
2405                             "%s\n",
2406                             ip6_sprintf(ip6buf, gw6),
2407                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2408                         goto bad;
2409                 }
2410         } else {
2411                 nd6log((LOG_ERR,
2412                     "ICMP6 redirect rejected; "
2413                     "no route found for redirect dst: %s\n",
2414                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2415                 goto bad;
2416         }
2417         RTFREE_LOCKED(rt);
2418         rt = NULL;
2419     }
2420         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2421                 nd6log((LOG_ERR,
2422                     "ICMP6 redirect rejected; "
2423                     "redirect dst must be unicast: %s\n",
2424                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2425                 goto bad;
2426         }
2427
2428         is_router = is_onlink = 0;
2429         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2430                 is_router = 1;  /* router case */
2431         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2432                 is_onlink = 1;  /* on-link destination case */
2433         if (!is_router && !is_onlink) {
2434                 nd6log((LOG_ERR,
2435                     "ICMP6 redirect rejected; "
2436                     "neither router case nor onlink case: %s\n",
2437                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2438                 goto bad;
2439         }
2440         /* validation passed */
2441
2442         icmp6len -= sizeof(*nd_rd);
2443         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2444         if (nd6_options(&ndopts) < 0) {
2445                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2446                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2447                 /* nd6_options have incremented stats */
2448                 goto freeit;
2449         }
2450
2451         if (ndopts.nd_opts_tgt_lladdr) {
2452                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2453                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2454         }
2455
2456         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2457                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2458                     "(if %d, icmp6 packet %d): %s\n",
2459                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2460                     ifp->if_addrlen, lladdrlen - 2,
2461                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2462                 goto bad;
2463         }
2464
2465         /* RFC 2461 8.3 */
2466         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2467             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2468
2469         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2470                 /* perform rtredirect */
2471                 struct sockaddr_in6 sdst;
2472                 struct sockaddr_in6 sgw;
2473                 struct sockaddr_in6 ssrc;
2474                 u_int fibnum;
2475
2476                 bzero(&sdst, sizeof(sdst));
2477                 bzero(&sgw, sizeof(sgw));
2478                 bzero(&ssrc, sizeof(ssrc));
2479                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2480                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2481                         sizeof(struct sockaddr_in6);
2482                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2483                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2484                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2485                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2486                         in6_rtredirect((struct sockaddr *)&sdst,
2487                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2488                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2489                             fibnum);
2490         }
2491         /* finally update cached route in each socket via pfctlinput */
2492     {
2493         struct sockaddr_in6 sdst;
2494
2495         bzero(&sdst, sizeof(sdst));
2496         sdst.sin6_family = AF_INET6;
2497         sdst.sin6_len = sizeof(struct sockaddr_in6);
2498         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2499         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2500     }
2501
2502  freeit:
2503         m_freem(m);
2504         return;
2505
2506  bad:
2507         ICMP6STAT_INC(icp6s_badredirect);
2508         m_freem(m);
2509 }
2510
2511 void
2512 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2513 {
2514         struct ifnet *ifp;      /* my outgoing interface */
2515         struct in6_addr *ifp_ll6;
2516         struct in6_addr *router_ll6;
2517         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2518         struct mbuf *m = NULL;  /* newly allocated one */
2519         struct m_tag *mtag;
2520         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2521         struct nd_redirect *nd_rd;
2522         struct llentry *ln = NULL;
2523         size_t maxlen;
2524         u_char *p;
2525         struct ifnet *outif = NULL;
2526         struct sockaddr_in6 src_sa;
2527
2528         icmp6_errcount(ND_REDIRECT, 0);
2529
2530         /* if we are not router, we don't send icmp6 redirect */
2531         if (!V_ip6_forwarding)
2532                 goto fail;
2533
2534         /* sanity check */
2535         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2536                 goto fail;
2537
2538         /*
2539          * Address check:
2540          *  the source address must identify a neighbor, and
2541          *  the destination address must not be a multicast address
2542          *  [RFC 2461, sec 8.2]
2543          */
2544         sip6 = mtod(m0, struct ip6_hdr *);
2545         bzero(&src_sa, sizeof(src_sa));
2546         src_sa.sin6_family = AF_INET6;
2547         src_sa.sin6_len = sizeof(src_sa);
2548         src_sa.sin6_addr = sip6->ip6_src;
2549         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2550                 goto fail;
2551         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2552                 goto fail;      /* what should we do here? */
2553
2554         /* rate limit */
2555         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2556                 goto fail;
2557
2558         /*
2559          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2560          * we almost always ask for an mbuf cluster for simplicity.
2561          * (MHLEN < IPV6_MMTU is almost always true)
2562          */
2563 #if IPV6_MMTU >= MCLBYTES
2564 # error assumption failed about IPV6_MMTU and MCLBYTES
2565 #endif
2566         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2567         if (m == NULL)
2568                 goto fail;
2569         M_SETFIB(m, rt->rt_fibnum);
2570         maxlen = M_TRAILINGSPACE(m);
2571         maxlen = min(IPV6_MMTU, maxlen);
2572         /* just for safety */
2573         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2574             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2575                 goto fail;
2576         }
2577
2578         {
2579                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2580                 struct in6_ifaddr *ia;
2581                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2582                                                  IN6_IFF_NOTREADY|
2583                                                  IN6_IFF_ANYCAST)) == NULL)
2584                         goto fail;
2585                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2586                 /* XXXRW: reference released prematurely. */
2587                 ifa_free(&ia->ia_ifa);
2588         }
2589
2590         /* get ip6 linklocal address for the router. */
2591         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2592                 struct sockaddr_in6 *sin6;
2593                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2594                 router_ll6 = &sin6->sin6_addr;
2595                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2596                         router_ll6 = (struct in6_addr *)NULL;
2597         } else
2598                 router_ll6 = (struct in6_addr *)NULL;
2599
2600         /* ip6 */
2601         ip6 = mtod(m, struct ip6_hdr *);
2602         ip6->ip6_flow = 0;
2603         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2604         ip6->ip6_vfc |= IPV6_VERSION;
2605         /* ip6->ip6_plen will be set later */
2606         ip6->ip6_nxt = IPPROTO_ICMPV6;
2607         ip6->ip6_hlim = 255;
2608         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2609         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2610         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2611
2612         /* ND Redirect */
2613         nd_rd = (struct nd_redirect *)(ip6 + 1);
2614         nd_rd->nd_rd_type = ND_REDIRECT;
2615         nd_rd->nd_rd_code = 0;
2616         nd_rd->nd_rd_reserved = 0;
2617         if (rt->rt_flags & RTF_GATEWAY) {
2618                 /*
2619                  * nd_rd->nd_rd_target must be a link-local address in
2620                  * better router cases.
2621                  */
2622                 if (!router_ll6)
2623                         goto fail;
2624                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2625                     sizeof(nd_rd->nd_rd_target));
2626                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2627                     sizeof(nd_rd->nd_rd_dst));
2628         } else {
2629                 /* make sure redtgt == reddst */
2630                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2631                     sizeof(nd_rd->nd_rd_target));
2632                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2633                     sizeof(nd_rd->nd_rd_dst));
2634         }
2635
2636         p = (u_char *)(nd_rd + 1);
2637
2638         if (!router_ll6)
2639                 goto nolladdropt;
2640
2641         {
2642                 /* target lladdr option */
2643                 int len;
2644                 struct nd_opt_hdr *nd_opt;
2645                 char *lladdr;
2646
2647                 IF_AFDATA_RLOCK(ifp);
2648                 ln = nd6_lookup(router_ll6, 0, ifp);
2649                 IF_AFDATA_RUNLOCK(ifp);
2650                 if (ln == NULL)
2651                         goto nolladdropt;
2652
2653                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2654                 len = (len + 7) & ~7;   /* round by 8 */
2655                 /* safety check */
2656                 if (len + (p - (u_char *)ip6) > maxlen)                         
2657                         goto nolladdropt;
2658
2659                 if (ln->la_flags & LLE_VALID) {
2660                         nd_opt = (struct nd_opt_hdr *)p;
2661                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2662                         nd_opt->nd_opt_len = len >> 3;
2663                         lladdr = (char *)(nd_opt + 1);
2664                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2665                         p += len;
2666                 }
2667         }
2668 nolladdropt:
2669         if (ln != NULL)
2670                 LLE_RUNLOCK(ln);
2671                 
2672         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2673
2674         /* just to be safe */
2675 #ifdef M_DECRYPTED      /*not openbsd*/
2676         if (m0->m_flags & M_DECRYPTED)
2677                 goto noredhdropt;
2678 #endif
2679         if (p - (u_char *)ip6 > maxlen)
2680                 goto noredhdropt;
2681
2682         {
2683                 /* redirected header option */
2684                 int len;
2685                 struct nd_opt_rd_hdr *nd_opt_rh;
2686
2687                 /*
2688                  * compute the maximum size for icmp6 redirect header option.
2689                  * XXX room for auth header?
2690                  */
2691                 len = maxlen - (p - (u_char *)ip6);
2692                 len &= ~7;
2693
2694                 /* This is just for simplicity. */
2695                 if (m0->m_pkthdr.len != m0->m_len) {
2696                         if (m0->m_next) {
2697                                 m_freem(m0->m_next);
2698                                 m0->m_next = NULL;
2699                         }
2700                         m0->m_pkthdr.len = m0->m_len;
2701                 }
2702
2703                 /*
2704                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2705                  * about padding/truncate rule for the original IP packet.
2706                  * From the discussion on IPv6imp in Feb 1999,
2707                  * the consensus was:
2708                  * - "attach as much as possible" is the goal
2709                  * - pad if not aligned (original size can be guessed by
2710                  *   original ip6 header)
2711                  * Following code adds the padding if it is simple enough,
2712                  * and truncates if not.
2713                  */
2714                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2715                         panic("assumption failed in %s:%d", __FILE__,
2716                             __LINE__);
2717
2718                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2719                         /* not enough room, truncate */
2720                         m0->m_pkthdr.len = m0->m_len = len -
2721                             sizeof(*nd_opt_rh);
2722                 } else {
2723                         /* enough room, pad or truncate */
2724                         size_t extra;
2725
2726                         extra = m0->m_pkthdr.len % 8;
2727                         if (extra) {
2728                                 /* pad if easy enough, truncate if not */
2729                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2730                                         /* pad */
2731                                         m0->m_len += (8 - extra);
2732                                         m0->m_pkthdr.len += (8 - extra);
2733                                 } else {
2734                                         /* truncate */
2735                                         m0->m_pkthdr.len -= extra;
2736                                         m0->m_len -= extra;
2737                                 }
2738                         }
2739                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2740                         m0->m_pkthdr.len = m0->m_len = len -
2741                             sizeof(*nd_opt_rh);
2742                 }
2743
2744                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2745                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2746                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2747                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2748                 p += sizeof(*nd_opt_rh);
2749                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2750
2751                 /* connect m0 to m */
2752                 m_tag_delete_chain(m0, NULL);
2753                 m0->m_flags &= ~M_PKTHDR;
2754                 m->m_next = m0;
2755                 m->m_pkthdr.len = m->m_len + m0->m_len;
2756                 m0 = NULL;
2757         }
2758 noredhdropt:;
2759         if (m0) {
2760                 m_freem(m0);
2761                 m0 = NULL;
2762         }
2763
2764         /* XXX: clear embedded link IDs in the inner header */
2765         in6_clearscope(&sip6->ip6_src);
2766         in6_clearscope(&sip6->ip6_dst);
2767         in6_clearscope(&nd_rd->nd_rd_target);
2768         in6_clearscope(&nd_rd->nd_rd_dst);
2769
2770         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2771
2772         nd_rd->nd_rd_cksum = 0;
2773         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2774             sizeof(*ip6), ntohs(ip6->ip6_plen));
2775
2776         if (send_sendso_input_hook != NULL) {
2777                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2778                         M_NOWAIT);
2779                 if (mtag == NULL)
2780                         goto fail;
2781                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2782                 m_tag_prepend(m, mtag);
2783         }
2784
2785         /* send the packet to outside... */
2786         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2787         if (outif) {
2788                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2789                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2790         }
2791         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2792
2793         return;
2794
2795 fail:
2796         if (m)
2797                 m_freem(m);
2798         if (m0)
2799                 m_freem(m0);
2800 }
2801
2802 /*
2803  * ICMPv6 socket option processing.
2804  */
2805 int
2806 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2807 {
2808         int error = 0;
2809         int optlen;
2810         struct inpcb *inp = sotoinpcb(so);
2811         int level, op, optname;
2812
2813         if (sopt) {
2814                 level = sopt->sopt_level;
2815                 op = sopt->sopt_dir;
2816                 optname = sopt->sopt_name;
2817                 optlen = sopt->sopt_valsize;
2818         } else
2819                 level = op = optname = optlen = 0;
2820
2821         if (level != IPPROTO_ICMPV6) {
2822                 return EINVAL;
2823         }
2824
2825         switch (op) {
2826         case PRCO_SETOPT:
2827                 switch (optname) {
2828                 case ICMP6_FILTER:
2829                     {
2830                         struct icmp6_filter ic6f;
2831
2832                         if (optlen != sizeof(ic6f)) {
2833                                 error = EMSGSIZE;
2834                                 break;
2835                         }
2836                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2837                         if (error == 0) {
2838                                 INP_WLOCK(inp);
2839                                 *inp->in6p_icmp6filt = ic6f;
2840                                 INP_WUNLOCK(inp);
2841                         }
2842                         break;
2843                     }
2844
2845                 default:
2846                         error = ENOPROTOOPT;
2847                         break;
2848                 }
2849                 break;
2850
2851         case PRCO_GETOPT:
2852                 switch (optname) {
2853                 case ICMP6_FILTER:
2854                     {
2855                         struct icmp6_filter ic6f;
2856
2857                         INP_RLOCK(inp);
2858                         ic6f = *inp->in6p_icmp6filt;
2859                         INP_RUNLOCK(inp);
2860                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2861                         break;
2862                     }
2863
2864                 default:
2865                         error = ENOPROTOOPT;
2866                         break;
2867                 }
2868                 break;
2869         }
2870
2871         return (error);
2872 }
2873
2874 /*
2875  * Perform rate limit check.
2876  * Returns 0 if it is okay to send the icmp6 packet.
2877  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2878  * limitation.
2879  *
2880  * XXX per-destination/type check necessary?
2881  *
2882  * dst - not used at this moment
2883  * type - not used at this moment
2884  * code - not used at this moment
2885  */
2886 static int
2887 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2888     const int code)
2889 {
2890         int ret;
2891
2892         ret = 0;        /* okay to send */
2893
2894         /* PPS limit */
2895         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2896             V_icmp6errppslim)) {
2897                 /* The packet is subject to rate limit */
2898                 ret++;
2899         }
2900
2901         return ret;
2902 }