]> CyberLeo.Net >> Repos - FreeBSD/stable/10.git/blob - sys/netinet6/icmp6.c
MFC r368207,368607:
[FreeBSD/stable/10.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 & (ICMP6_NODEINFO_FQDNOK |
683                             ICMP6_NODEINFO_TMPADDROK)) !=
684                             (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
685                                 break;
686
687                         if (code != 0)
688                                 goto badcode;
689
690                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
691                         n = m_gethdr(M_NOWAIT, m->m_type);
692                         if (n == NULL) {
693                                 /* Give up remote */
694                                 break;
695                         }
696                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
697                                 /*
698                                  * Previous code did a blind M_COPY_PKTHDR
699                                  * and said "just for rcvif".  If true, then
700                                  * we could tolerate the dup failing (due to
701                                  * the deep copy of the tag chain).  For now
702                                  * be conservative and just fail.
703                                  */
704                                 m_free(n);
705                                 n = NULL;
706                         }
707                         maxhlen = M_TRAILINGSPACE(n) -
708                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
709                         pr = curthread->td_ucred->cr_prison;
710                         mtx_lock(&pr->pr_mtx);
711                         hlen = strlen(pr->pr_hostname);
712                         if (maxhlen > hlen)
713                                 maxhlen = hlen;
714                         /*
715                          * Copy IPv6 and ICMPv6 only.
716                          */
717                         nip6 = mtod(n, struct ip6_hdr *);
718                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
719                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
720                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
721                         p = (u_char *)(nicmp6 + 1);
722                         bzero(p, 4);
723                         /* meaningless TTL */
724                         bcopy(pr->pr_hostname, p + 4, maxhlen);
725                         mtx_unlock(&pr->pr_mtx);
726                         noff = sizeof(struct ip6_hdr);
727                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
728                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
729                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
730                         nicmp6->icmp6_code = 0;
731                 }
732                 if (n) {
733                         ICMP6STAT_INC(icp6s_reflect);
734                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
735                         icmp6_reflect(n, noff);
736                 }
737                 break;
738             }
739
740         case ICMP6_WRUREPLY:
741                 if (code != 0)
742                         goto badcode;
743                 break;
744
745         case ND_ROUTER_SOLICIT:
746                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
747                 if (code != 0)
748                         goto badcode;
749                 if (icmp6len < sizeof(struct nd_router_solicit))
750                         goto badlen;
751                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
752                         /* give up local */
753
754                         /* Send incoming SeND packet to user space. */
755                         if (send_sendso_input_hook != NULL) {
756                                 IP6_EXTHDR_CHECK(m, off,
757                                     icmp6len, IPPROTO_DONE);
758                                 error = send_sendso_input_hook(m, ifp,
759                                     SND_IN, ip6len);
760                                 /* -1 == no app on SEND socket */
761                                 if (error == 0)
762                                         return (IPPROTO_DONE);
763                                 nd6_rs_input(m, off, icmp6len);
764                         } else
765                                 nd6_rs_input(m, off, icmp6len);
766                         m = NULL;
767                         goto freeit;
768                 }
769                 if (send_sendso_input_hook != NULL) {
770                         IP6_EXTHDR_CHECK(n, off,
771                             icmp6len, IPPROTO_DONE);
772                         error = send_sendso_input_hook(n, ifp,
773                             SND_IN, ip6len);
774                         if (error == 0)
775                                 goto freeit;
776                         /* -1 == no app on SEND socket */
777                         nd6_rs_input(n, off, icmp6len);
778                 } else
779                         nd6_rs_input(n, off, icmp6len);
780                 /* m stays. */
781                 break;
782
783         case ND_ROUTER_ADVERT:
784                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
785                 if (code != 0)
786                         goto badcode;
787                 if (icmp6len < sizeof(struct nd_router_advert))
788                         goto badlen;
789                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
790
791                         /* Send incoming SeND-protected/ND packet to user space. */
792                         if (send_sendso_input_hook != NULL) {
793                                 error = send_sendso_input_hook(m, ifp,
794                                     SND_IN, ip6len);
795                                 if (error == 0)
796                                         return (IPPROTO_DONE);
797                                 nd6_ra_input(m, off, icmp6len);
798                         } else
799                                 nd6_ra_input(m, off, icmp6len);
800                         m = NULL;
801                         goto freeit;
802                 }
803                 if (send_sendso_input_hook != NULL) {
804                         error = send_sendso_input_hook(n, ifp,
805                             SND_IN, ip6len);
806                         if (error == 0)
807                                 goto freeit;
808                         nd6_ra_input(n, off, icmp6len);
809                 } else
810                         nd6_ra_input(n, off, icmp6len);
811                 /* m stays. */
812                 break;
813
814         case ND_NEIGHBOR_SOLICIT:
815                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
816                 if (code != 0)
817                         goto badcode;
818                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
819                         goto badlen;
820                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
821                         if (send_sendso_input_hook != NULL) {
822                                 error = send_sendso_input_hook(m, ifp,
823                                     SND_IN, ip6len);
824                                 if (error == 0)
825                                         return (IPPROTO_DONE);
826                                 nd6_ns_input(m, off, icmp6len);
827                         } else
828                                 nd6_ns_input(m, off, icmp6len);
829                         m = NULL;
830                         goto freeit;
831                 }
832                 if (send_sendso_input_hook != NULL) {
833                         error = send_sendso_input_hook(n, ifp,
834                             SND_IN, ip6len);
835                         if (error == 0)
836                                 goto freeit;
837                         nd6_ns_input(n, off, icmp6len);
838                 } else
839                         nd6_ns_input(n, off, icmp6len);
840                 /* m stays. */
841                 break;
842
843         case ND_NEIGHBOR_ADVERT:
844                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
845                 if (code != 0)
846                         goto badcode;
847                 if (icmp6len < sizeof(struct nd_neighbor_advert))
848                         goto badlen;
849                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
850
851                         /* Send incoming SeND-protected/ND packet to user space. */
852                         if (send_sendso_input_hook != NULL) {
853                                 error = send_sendso_input_hook(m, ifp,
854                                     SND_IN, ip6len);
855                                 if (error == 0)
856                                         return (IPPROTO_DONE);
857                                 nd6_na_input(m, off, icmp6len);
858                         } else
859                                 nd6_na_input(m, off, icmp6len);
860                         m = NULL;
861                         goto freeit;
862                 }
863                 if (send_sendso_input_hook != NULL) {
864                         error = send_sendso_input_hook(n, ifp,
865                             SND_IN, ip6len);
866                         if (error == 0)
867                                 goto freeit;
868                         nd6_na_input(n, off, icmp6len);
869                 } else
870                         nd6_na_input(n, off, icmp6len);
871                 /* m stays. */
872                 break;
873
874         case ND_REDIRECT:
875                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
876                 if (code != 0)
877                         goto badcode;
878                 if (icmp6len < sizeof(struct nd_redirect))
879                         goto badlen;
880                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
881                         if (send_sendso_input_hook != NULL) {
882                                 error = send_sendso_input_hook(m, ifp,
883                                     SND_IN, ip6len);
884                                 if (error == 0)
885                                         return (IPPROTO_DONE);
886                             icmp6_redirect_input(m, off);
887                         } else
888                                 icmp6_redirect_input(m, off);
889                         m = NULL;
890                         goto freeit;
891                 }
892                 if (send_sendso_input_hook != NULL) {
893                         error = send_sendso_input_hook(n, ifp,
894                             SND_IN, ip6len);
895                         if (error == 0)
896                                 goto freeit;
897                         icmp6_redirect_input(n, off);
898                 } else
899                         icmp6_redirect_input(n, off);
900                 /* m stays. */
901                 break;
902
903         case ICMP6_ROUTER_RENUMBERING:
904                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
905                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
906                         goto badcode;
907                 if (icmp6len < sizeof(struct icmp6_router_renum))
908                         goto badlen;
909                 break;
910
911         default:
912                 nd6log((LOG_DEBUG,
913                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
914                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
915                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
916                     ifp ? ifp->if_index : 0));
917                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
918                         /* ICMPv6 error: MUST deliver it by spec... */
919                         code = PRC_NCMDS;
920                         /* deliver */
921                 } else {
922                         /* ICMPv6 informational: MUST not deliver */
923                         break;
924                 }
925         deliver:
926                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
927                         /* In this case, m should've been freed. */
928                         return (IPPROTO_DONE);
929                 }
930                 break;
931
932         badcode:
933                 ICMP6STAT_INC(icp6s_badcode);
934                 break;
935
936         badlen:
937                 ICMP6STAT_INC(icp6s_badlen);
938                 break;
939         }
940
941         /* deliver the packet to appropriate sockets */
942         icmp6_rip6_input(&m, *offp);
943
944         return IPPROTO_DONE;
945
946  freeit:
947         m_freem(m);
948         return IPPROTO_DONE;
949 }
950
951 static int
952 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
953 {
954         struct mbuf *m = *mp;
955         struct icmp6_hdr *icmp6;
956         struct ip6_hdr *eip6;
957         u_int32_t notifymtu;
958         struct sockaddr_in6 icmp6src, icmp6dst;
959
960         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
961                 ICMP6STAT_INC(icp6s_tooshort);
962                 goto freeit;
963         }
964 #ifndef PULLDOWN_TEST
965         IP6_EXTHDR_CHECK(m, off,
966             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
967         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
968 #else
969         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
970             sizeof(*icmp6) + sizeof(struct ip6_hdr));
971         if (icmp6 == NULL) {
972                 ICMP6STAT_INC(icp6s_tooshort);
973                 return (-1);
974         }
975 #endif
976         eip6 = (struct ip6_hdr *)(icmp6 + 1);
977
978         /* Detect the upper level protocol */
979         {
980                 void (*ctlfunc)(int, struct sockaddr *, void *);
981                 u_int8_t nxt = eip6->ip6_nxt;
982                 int eoff = off + sizeof(struct icmp6_hdr) +
983                     sizeof(struct ip6_hdr);
984                 struct ip6ctlparam ip6cp;
985                 struct in6_addr *finaldst = NULL;
986                 int icmp6type = icmp6->icmp6_type;
987                 struct ip6_frag *fh;
988                 struct ip6_rthdr *rth;
989                 struct ip6_rthdr0 *rth0;
990                 int rthlen;
991
992                 while (1) { /* XXX: should avoid infinite loop explicitly? */
993                         struct ip6_ext *eh;
994
995                         switch (nxt) {
996                         case IPPROTO_HOPOPTS:
997                         case IPPROTO_DSTOPTS:
998                         case IPPROTO_AH:
999 #ifndef PULLDOWN_TEST
1000                                 IP6_EXTHDR_CHECK(m, 0,
1001                                     eoff + sizeof(struct ip6_ext), -1);
1002                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1003 #else
1004                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1005                                     eoff, sizeof(*eh));
1006                                 if (eh == NULL) {
1007                                         ICMP6STAT_INC(icp6s_tooshort);
1008                                         return (-1);
1009                                 }
1010 #endif
1011
1012                                 if (nxt == IPPROTO_AH)
1013                                         eoff += (eh->ip6e_len + 2) << 2;
1014                                 else
1015                                         eoff += (eh->ip6e_len + 1) << 3;
1016                                 nxt = eh->ip6e_nxt;
1017                                 break;
1018                         case IPPROTO_ROUTING:
1019                                 /*
1020                                  * When the erroneous packet contains a
1021                                  * routing header, we should examine the
1022                                  * header to determine the final destination.
1023                                  * Otherwise, we can't properly update
1024                                  * information that depends on the final
1025                                  * destination (e.g. path MTU).
1026                                  */
1027 #ifndef PULLDOWN_TEST
1028                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1029                                 rth = (struct ip6_rthdr *)
1030                                     (mtod(m, caddr_t) + eoff);
1031 #else
1032                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1033                                     eoff, sizeof(*rth));
1034                                 if (rth == NULL) {
1035                                         ICMP6STAT_INC(icp6s_tooshort);
1036                                         return (-1);
1037                                 }
1038 #endif
1039                                 rthlen = (rth->ip6r_len + 1) << 3;
1040                                 /*
1041                                  * XXX: currently there is no
1042                                  * officially defined type other
1043                                  * than type-0.
1044                                  * Note that if the segment left field
1045                                  * is 0, all intermediate hops must
1046                                  * have been passed.
1047                                  */
1048                                 if (rth->ip6r_segleft &&
1049                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1050                                         int hops;
1051
1052 #ifndef PULLDOWN_TEST
1053                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1054                                         rth0 = (struct ip6_rthdr0 *)
1055                                             (mtod(m, caddr_t) + eoff);
1056 #else
1057                                         IP6_EXTHDR_GET(rth0,
1058                                             struct ip6_rthdr0 *, m,
1059                                             eoff, rthlen);
1060                                         if (rth0 == NULL) {
1061                                                 ICMP6STAT_INC(icp6s_tooshort);
1062                                                 return (-1);
1063                                         }
1064 #endif
1065                                         /* just ignore a bogus header */
1066                                         if ((rth0->ip6r0_len % 2) == 0 &&
1067                                             (hops = rth0->ip6r0_len/2))
1068                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1069                                 }
1070                                 eoff += rthlen;
1071                                 nxt = rth->ip6r_nxt;
1072                                 break;
1073                         case IPPROTO_FRAGMENT:
1074 #ifndef PULLDOWN_TEST
1075                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1076                                     sizeof(struct ip6_frag), -1);
1077                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1078                                     eoff);
1079 #else
1080                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1081                                     eoff, sizeof(*fh));
1082                                 if (fh == NULL) {
1083                                         ICMP6STAT_INC(icp6s_tooshort);
1084                                         return (-1);
1085                                 }
1086 #endif
1087                                 /*
1088                                  * Data after a fragment header is meaningless
1089                                  * unless it is the first fragment, but
1090                                  * we'll go to the notify label for path MTU
1091                                  * discovery.
1092                                  */
1093                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1094                                         goto notify;
1095
1096                                 eoff += sizeof(struct ip6_frag);
1097                                 nxt = fh->ip6f_nxt;
1098                                 break;
1099                         default:
1100                                 /*
1101                                  * This case includes ESP and the No Next
1102                                  * Header.  In such cases going to the notify
1103                                  * label does not have any meaning
1104                                  * (i.e. ctlfunc will be NULL), but we go
1105                                  * anyway since we might have to update
1106                                  * path MTU information.
1107                                  */
1108                                 goto notify;
1109                         }
1110                 }
1111           notify:
1112 #ifndef PULLDOWN_TEST
1113                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1114 #else
1115                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1116                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1117                 if (icmp6 == NULL) {
1118                         ICMP6STAT_INC(icp6s_tooshort);
1119                         return (-1);
1120                 }
1121 #endif
1122
1123                 /*
1124                  * retrieve parameters from the inner IPv6 header, and convert
1125                  * them into sockaddr structures.
1126                  * XXX: there is no guarantee that the source or destination
1127                  * addresses of the inner packet are in the same scope as
1128                  * the addresses of the icmp packet.  But there is no other
1129                  * way to determine the zone.
1130                  */
1131                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1132
1133                 bzero(&icmp6dst, sizeof(icmp6dst));
1134                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1135                 icmp6dst.sin6_family = AF_INET6;
1136                 if (finaldst == NULL)
1137                         icmp6dst.sin6_addr = eip6->ip6_dst;
1138                 else
1139                         icmp6dst.sin6_addr = *finaldst;
1140                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1141                         goto freeit;
1142                 bzero(&icmp6src, sizeof(icmp6src));
1143                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1144                 icmp6src.sin6_family = AF_INET6;
1145                 icmp6src.sin6_addr = eip6->ip6_src;
1146                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1147                         goto freeit;
1148                 icmp6src.sin6_flowinfo =
1149                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1150
1151                 if (finaldst == NULL)
1152                         finaldst = &eip6->ip6_dst;
1153                 ip6cp.ip6c_m = m;
1154                 ip6cp.ip6c_icmp6 = icmp6;
1155                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1156                 ip6cp.ip6c_off = eoff;
1157                 ip6cp.ip6c_finaldst = finaldst;
1158                 ip6cp.ip6c_src = &icmp6src;
1159                 ip6cp.ip6c_nxt = nxt;
1160
1161                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1162                         notifymtu = ntohl(icmp6->icmp6_mtu);
1163                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1164                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1165                 }
1166
1167                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1168                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1169                 if (ctlfunc) {
1170                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1171                             &ip6cp);
1172                 }
1173         }
1174         *mp = m;
1175         return (0);
1176
1177   freeit:
1178         m_freem(m);
1179         return (-1);
1180 }
1181
1182 void
1183 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1184 {
1185         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1186         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1187         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1188         u_int mtu = ntohl(icmp6->icmp6_mtu);
1189         struct in_conninfo inc;
1190
1191 #if 0
1192         /*
1193          * RFC2460 section 5, last paragraph.
1194          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1195          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1196          * due to packet translator in the middle.
1197          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1198          * special handling.
1199          */
1200         if (mtu < IPV6_MMTU)
1201                 return;
1202 #endif
1203
1204         /*
1205          * we reject ICMPv6 too big with abnormally small value.
1206          * XXX what is the good definition of "abnormally small"?
1207          */
1208         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1209                 return;
1210
1211         if (!validated)
1212                 return;
1213
1214         /*
1215          * In case the suggested mtu is less than IPV6_MMTU, we
1216          * only need to remember that it was for above mentioned
1217          * "alwaysfrag" case.
1218          * Try to be as close to the spec as possible.
1219          */
1220         if (mtu < IPV6_MMTU)
1221                 mtu = IPV6_MMTU - 8;
1222
1223         bzero(&inc, sizeof(inc));
1224         inc.inc_fibnum = M_GETFIB(m);
1225         inc.inc_flags |= INC_ISIPV6;
1226         inc.inc6_faddr = *dst;
1227         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1228                 return;
1229
1230         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1231                 tcp_hc_updatemtu(&inc, mtu);
1232                 ICMP6STAT_INC(icp6s_pmtuchg);
1233         }
1234 }
1235
1236 /*
1237  * Process a Node Information Query packet, based on
1238  * draft-ietf-ipngwg-icmp-name-lookups-07.
1239  *
1240  * Spec incompatibilities:
1241  * - IPv6 Subject address handling
1242  * - IPv4 Subject address handling support missing
1243  * - Proxy reply (answer even if it's not for me)
1244  * - joins NI group address at in6_ifattach() time only, does not cope
1245  *   with hostname changes by sethostname(3)
1246  */
1247 static struct mbuf *
1248 ni6_input(struct mbuf *m, int off)
1249 {
1250         struct icmp6_nodeinfo *ni6, *nni6;
1251         struct mbuf *n = NULL;
1252         struct prison *pr;
1253         u_int16_t qtype;
1254         int subjlen;
1255         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1256         struct ni_reply_fqdn *fqdn;
1257         int addrs;              /* for NI_QTYPE_NODEADDR */
1258         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1259         struct in6_addr in6_subj; /* subject address */
1260         struct ip6_hdr *ip6;
1261         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1262         char *subj = NULL;
1263         struct in6_ifaddr *ia6 = NULL;
1264
1265         ip6 = mtod(m, struct ip6_hdr *);
1266 #ifndef PULLDOWN_TEST
1267         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1268 #else
1269         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1270         if (ni6 == NULL) {
1271                 /* m is already reclaimed */
1272                 return (NULL);
1273         }
1274 #endif
1275
1276         /*
1277          * Validate IPv6 source address.
1278          * The default configuration MUST be to refuse answering queries from
1279          * global-scope addresses according to RFC4602.
1280          * Notes:
1281          *  - it's not very clear what "refuse" means; this implementation
1282          *    simply drops it.
1283          *  - it's not very easy to identify global-scope (unicast) addresses
1284          *    since there are many prefixes for them.  It should be safer
1285          *    and in practice sufficient to check "all" but loopback and
1286          *    link-local (note that site-local unicast was deprecated and
1287          *    ULA is defined as global scope-wise)
1288          */
1289         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1290             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1291             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1292                 goto bad;
1293
1294         /*
1295          * Validate IPv6 destination address.
1296          *
1297          * The Responder must discard the Query without further processing
1298          * unless it is one of the Responder's unicast or anycast addresses, or
1299          * a link-local scope multicast address which the Responder has joined.
1300          * [RFC4602, Section 5.]
1301          */
1302         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1303                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1304                         goto bad;
1305                 /* else it's a link-local multicast, fine */
1306         } else {                /* unicast or anycast */
1307                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1308                         goto bad; /* XXX impossible */
1309
1310                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1311                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1312                         ifa_free(&ia6->ia_ifa);
1313                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1314                                 "a temporary address in %s:%d",
1315                                __FILE__, __LINE__));
1316                         goto bad;
1317                 }
1318                 ifa_free(&ia6->ia_ifa);
1319         }
1320
1321         /* validate query Subject field. */
1322         qtype = ntohs(ni6->ni_qtype);
1323         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1324         switch (qtype) {
1325         case NI_QTYPE_NOOP:
1326         case NI_QTYPE_SUPTYPES:
1327                 /* 07 draft */
1328                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1329                         break;
1330                 /* FALLTHROUGH */
1331         case NI_QTYPE_FQDN:
1332         case NI_QTYPE_NODEADDR:
1333         case NI_QTYPE_IPV4ADDR:
1334                 switch (ni6->ni_code) {
1335                 case ICMP6_NI_SUBJ_IPV6:
1336 #if ICMP6_NI_SUBJ_IPV6 != 0
1337                 case 0:
1338 #endif
1339                         /*
1340                          * backward compatibility - try to accept 03 draft
1341                          * format, where no Subject is present.
1342                          */
1343                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1344                             subjlen == 0) {
1345                                 oldfqdn++;
1346                                 break;
1347                         }
1348 #if ICMP6_NI_SUBJ_IPV6 != 0
1349                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1350                                 goto bad;
1351 #endif
1352
1353                         if (subjlen != sizeof(struct in6_addr))
1354                                 goto bad;
1355
1356                         /*
1357                          * Validate Subject address.
1358                          *
1359                          * Not sure what exactly "address belongs to the node"
1360                          * means in the spec, is it just unicast, or what?
1361                          *
1362                          * At this moment we consider Subject address as
1363                          * "belong to the node" if the Subject address equals
1364                          * to the IPv6 destination address; validation for
1365                          * IPv6 destination address should have done enough
1366                          * check for us.
1367                          *
1368                          * We do not do proxy at this moment.
1369                          */
1370                         /* m_pulldown instead of copy? */
1371                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1372                             subjlen, (caddr_t)&in6_subj);
1373                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1374                                 goto bad;
1375
1376                         subj = (char *)&in6_subj;
1377                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1378                                 break;
1379
1380                         /*
1381                          * XXX if we are to allow other cases, we should really
1382                          * be careful about scope here.
1383                          * basically, we should disallow queries toward IPv6
1384                          * destination X with subject Y,
1385                          * if scope(X) > scope(Y).
1386                          * if we allow scope(X) > scope(Y), it will result in
1387                          * information leakage across scope boundary.
1388                          */
1389                         goto bad;
1390
1391                 case ICMP6_NI_SUBJ_FQDN:
1392                         /*
1393                          * Validate Subject name with gethostname(3).
1394                          *
1395                          * The behavior may need some debate, since:
1396                          * - we are not sure if the node has FQDN as
1397                          *   hostname (returned by gethostname(3)).
1398                          * - the code does wildcard match for truncated names.
1399                          *   however, we are not sure if we want to perform
1400                          *   wildcard match, if gethostname(3) side has
1401                          *   truncated hostname.
1402                          */
1403                         pr = curthread->td_ucred->cr_prison;
1404                         mtx_lock(&pr->pr_mtx);
1405                         n = ni6_nametodns(pr->pr_hostname,
1406                             strlen(pr->pr_hostname), 0);
1407                         mtx_unlock(&pr->pr_mtx);
1408                         if (!n || n->m_next || n->m_len == 0)
1409                                 goto bad;
1410                         IP6_EXTHDR_GET(subj, char *, m,
1411                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1412                         if (subj == NULL)
1413                                 goto bad;
1414                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1415                             n->m_len)) {
1416                                 goto bad;
1417                         }
1418                         m_freem(n);
1419                         n = NULL;
1420                         break;
1421
1422                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1423                 default:
1424                         goto bad;
1425                 }
1426                 break;
1427         }
1428
1429         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1430         switch (qtype) {
1431         case NI_QTYPE_FQDN:
1432                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1433                         goto bad;
1434                 break;
1435         case NI_QTYPE_NODEADDR:
1436         case NI_QTYPE_IPV4ADDR:
1437                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1438                         goto bad;
1439                 break;
1440         }
1441
1442         /* guess reply length */
1443         switch (qtype) {
1444         case NI_QTYPE_NOOP:
1445                 break;          /* no reply data */
1446         case NI_QTYPE_SUPTYPES:
1447                 replylen += sizeof(u_int32_t);
1448                 break;
1449         case NI_QTYPE_FQDN:
1450                 /* XXX will append an mbuf */
1451                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1452                 break;
1453         case NI_QTYPE_NODEADDR:
1454                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1455                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1456                     sizeof(u_int32_t))) > MCLBYTES)
1457                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1458                 break;
1459         case NI_QTYPE_IPV4ADDR:
1460                 /* unsupported - should respond with unknown Qtype? */
1461                 break;
1462         default:
1463                 /*
1464                  * XXX: We must return a reply with the ICMP6 code
1465                  * `unknown Qtype' in this case.  However we regard the case
1466                  * as an FQDN query for backward compatibility.
1467                  * Older versions set a random value to this field,
1468                  * so it rarely varies in the defined qtypes.
1469                  * But the mechanism is not reliable...
1470                  * maybe we should obsolete older versions.
1471                  */
1472                 qtype = NI_QTYPE_FQDN;
1473                 /* XXX will append an mbuf */
1474                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1475                 oldfqdn++;
1476                 break;
1477         }
1478
1479         /* Allocate an mbuf to reply. */
1480         if (replylen > MCLBYTES) {
1481                 /*
1482                  * XXX: should we try to allocate more? But MCLBYTES
1483                  * is probably much larger than IPV6_MMTU...
1484                  */
1485                 goto bad;
1486         }
1487         if (replylen > MHLEN)
1488                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1489         else
1490                 n = m_gethdr(M_NOWAIT, m->m_type);
1491         if (n == NULL) {
1492                 m_freem(m);
1493                 return (NULL);
1494         }
1495         m_move_pkthdr(n, m); /* just for recvif and FIB */
1496         n->m_pkthdr.len = n->m_len = replylen;
1497
1498         /* copy mbuf header and IPv6 + Node Information base headers */
1499         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1500         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1501         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1502
1503         /* qtype dependent procedure */
1504         switch (qtype) {
1505         case NI_QTYPE_NOOP:
1506                 nni6->ni_code = ICMP6_NI_SUCCESS;
1507                 nni6->ni_flags = 0;
1508                 break;
1509         case NI_QTYPE_SUPTYPES:
1510         {
1511                 u_int32_t v;
1512                 nni6->ni_code = ICMP6_NI_SUCCESS;
1513                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1514                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1515                 v = (u_int32_t)htonl(0x0000000f);
1516                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1517                 break;
1518         }
1519         case NI_QTYPE_FQDN:
1520                 nni6->ni_code = ICMP6_NI_SUCCESS;
1521                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1522                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1523                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1524                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1525                 /*
1526                  * XXX do we really have FQDN in hostname?
1527                  */
1528                 pr = curthread->td_ucred->cr_prison;
1529                 mtx_lock(&pr->pr_mtx);
1530                 n->m_next = ni6_nametodns(pr->pr_hostname,
1531                     strlen(pr->pr_hostname), oldfqdn);
1532                 mtx_unlock(&pr->pr_mtx);
1533                 if (n->m_next == NULL)
1534                         goto bad;
1535                 /* XXX we assume that n->m_next is not a chain */
1536                 if (n->m_next->m_next != NULL)
1537                         goto bad;
1538                 n->m_pkthdr.len += n->m_next->m_len;
1539                 break;
1540         case NI_QTYPE_NODEADDR:
1541         {
1542                 int lenlim, copied;
1543
1544                 nni6->ni_code = ICMP6_NI_SUCCESS;
1545                 n->m_pkthdr.len = n->m_len =
1546                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1547                 lenlim = M_TRAILINGSPACE(n);
1548                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1549                 /* XXX: reset mbuf length */
1550                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1551                     sizeof(struct icmp6_nodeinfo) + copied;
1552                 break;
1553         }
1554         default:
1555                 break;          /* XXX impossible! */
1556         }
1557
1558         nni6->ni_type = ICMP6_NI_REPLY;
1559         m_freem(m);
1560         return (n);
1561
1562   bad:
1563         m_freem(m);
1564         if (n)
1565                 m_freem(n);
1566         return (NULL);
1567 }
1568
1569 /*
1570  * make a mbuf with DNS-encoded string.  no compression support.
1571  *
1572  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1573  * treated as truncated name (two \0 at the end).  this is a wild guess.
1574  *
1575  * old - return pascal string if non-zero
1576  */
1577 static struct mbuf *
1578 ni6_nametodns(const char *name, int namelen, int old)
1579 {
1580         struct mbuf *m;
1581         char *cp, *ep;
1582         const char *p, *q;
1583         int i, len, nterm;
1584
1585         if (old)
1586                 len = namelen + 1;
1587         else
1588                 len = MCLBYTES;
1589
1590         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1591         if (len > MLEN)
1592                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1593         else
1594                 m = m_get(M_NOWAIT, MT_DATA);
1595         if (m == NULL)
1596                 goto fail;
1597
1598         if (old) {
1599                 m->m_len = len;
1600                 *mtod(m, char *) = namelen;
1601                 bcopy(name, mtod(m, char *) + 1, namelen);
1602                 return m;
1603         } else {
1604                 m->m_len = 0;
1605                 cp = mtod(m, char *);
1606                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1607
1608                 /* if not certain about my name, return empty buffer */
1609                 if (namelen == 0)
1610                         return m;
1611
1612                 /*
1613                  * guess if it looks like shortened hostname, or FQDN.
1614                  * shortened hostname needs two trailing "\0".
1615                  */
1616                 i = 0;
1617                 for (p = name; p < name + namelen; p++) {
1618                         if (*p && *p == '.')
1619                                 i++;
1620                 }
1621                 if (i < 2)
1622                         nterm = 2;
1623                 else
1624                         nterm = 1;
1625
1626                 p = name;
1627                 while (cp < ep && p < name + namelen) {
1628                         i = 0;
1629                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1630                                 i++;
1631                         /* result does not fit into mbuf */
1632                         if (cp + i + 1 >= ep)
1633                                 goto fail;
1634                         /*
1635                          * DNS label length restriction, RFC1035 page 8.
1636                          * "i == 0" case is included here to avoid returning
1637                          * 0-length label on "foo..bar".
1638                          */
1639                         if (i <= 0 || i >= 64)
1640                                 goto fail;
1641                         *cp++ = i;
1642                         bcopy(p, cp, i);
1643                         cp += i;
1644                         p = q;
1645                         if (p < name + namelen && *p == '.')
1646                                 p++;
1647                 }
1648                 /* termination */
1649                 if (cp + nterm >= ep)
1650                         goto fail;
1651                 while (nterm-- > 0)
1652                         *cp++ = '\0';
1653                 m->m_len = cp - mtod(m, char *);
1654                 return m;
1655         }
1656
1657         panic("should not reach here");
1658         /* NOTREACHED */
1659
1660  fail:
1661         if (m)
1662                 m_freem(m);
1663         return NULL;
1664 }
1665
1666 /*
1667  * check if two DNS-encoded string matches.  takes care of truncated
1668  * form (with \0\0 at the end).  no compression support.
1669  * XXX upper/lowercase match (see RFC2065)
1670  */
1671 static int
1672 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1673 {
1674         const char *a0, *b0;
1675         int l;
1676
1677         /* simplest case - need validation? */
1678         if (alen == blen && bcmp(a, b, alen) == 0)
1679                 return 1;
1680
1681         a0 = a;
1682         b0 = b;
1683
1684         /* termination is mandatory */
1685         if (alen < 2 || blen < 2)
1686                 return 0;
1687         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1688                 return 0;
1689         alen--;
1690         blen--;
1691
1692         while (a - a0 < alen && b - b0 < blen) {
1693                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1694                         return 0;
1695
1696                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1697                         return 0;
1698                 /* we don't support compression yet */
1699                 if (a[0] >= 64 || b[0] >= 64)
1700                         return 0;
1701
1702                 /* truncated case */
1703                 if (a[0] == 0 && a - a0 == alen - 1)
1704                         return 1;
1705                 if (b[0] == 0 && b - b0 == blen - 1)
1706                         return 1;
1707                 if (a[0] == 0 || b[0] == 0)
1708                         return 0;
1709
1710                 if (a[0] != b[0])
1711                         return 0;
1712                 l = a[0];
1713                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1714                         return 0;
1715                 if (bcmp(a + 1, b + 1, l) != 0)
1716                         return 0;
1717
1718                 a += 1 + l;
1719                 b += 1 + l;
1720         }
1721
1722         if (a - a0 == alen && b - b0 == blen)
1723                 return 1;
1724         else
1725                 return 0;
1726 }
1727
1728 /*
1729  * calculate the number of addresses to be returned in the node info reply.
1730  */
1731 static int
1732 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1733     struct in6_addr *subj)
1734 {
1735         struct ifnet *ifp;
1736         struct in6_ifaddr *ifa6;
1737         struct ifaddr *ifa;
1738         int addrs = 0, addrsofif, iffound = 0;
1739         int niflags = ni6->ni_flags;
1740
1741         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1742                 switch (ni6->ni_code) {
1743                 case ICMP6_NI_SUBJ_IPV6:
1744                         if (subj == NULL) /* must be impossible... */
1745                                 return (0);
1746                         break;
1747                 default:
1748                         /*
1749                          * XXX: we only support IPv6 subject address for
1750                          * this Qtype.
1751                          */
1752                         return (0);
1753                 }
1754         }
1755
1756         IFNET_RLOCK_NOSLEEP();
1757         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1758                 addrsofif = 0;
1759                 IF_ADDR_RLOCK(ifp);
1760                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1761                         if (ifa->ifa_addr->sa_family != AF_INET6)
1762                                 continue;
1763                         ifa6 = (struct in6_ifaddr *)ifa;
1764
1765                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1766                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1767                                 iffound = 1;
1768
1769                         /*
1770                          * IPv4-mapped addresses can only be returned by a
1771                          * Node Information proxy, since they represent
1772                          * addresses of IPv4-only nodes, which perforce do
1773                          * not implement this protocol.
1774                          * [icmp-name-lookups-07, Section 5.4]
1775                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1776                          * this function at this moment.
1777                          */
1778
1779                         /* What do we have to do about ::1? */
1780                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1781                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1782                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1783                                         continue;
1784                                 break;
1785                         case IPV6_ADDR_SCOPE_SITELOCAL:
1786                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1787                                         continue;
1788                                 break;
1789                         case IPV6_ADDR_SCOPE_GLOBAL:
1790                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1791                                         continue;
1792                                 break;
1793                         default:
1794                                 continue;
1795                         }
1796
1797                         /*
1798                          * check if anycast is okay.
1799                          * XXX: just experimental.  not in the spec.
1800                          */
1801                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1802                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1803                                 continue; /* we need only unicast addresses */
1804                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1805                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1806                                 continue;
1807                         }
1808                         addrsofif++; /* count the address */
1809                 }
1810                 IF_ADDR_RUNLOCK(ifp);
1811                 if (iffound) {
1812                         *ifpp = ifp;
1813                         IFNET_RUNLOCK_NOSLEEP();
1814                         return (addrsofif);
1815                 }
1816
1817                 addrs += addrsofif;
1818         }
1819         IFNET_RUNLOCK_NOSLEEP();
1820
1821         return (addrs);
1822 }
1823
1824 static int
1825 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1826     struct ifnet *ifp0, int resid)
1827 {
1828         struct ifnet *ifp;
1829         struct in6_ifaddr *ifa6;
1830         struct ifaddr *ifa;
1831         struct ifnet *ifp_dep = NULL;
1832         int copied = 0, allow_deprecated = 0;
1833         u_char *cp = (u_char *)(nni6 + 1);
1834         int niflags = ni6->ni_flags;
1835         u_int32_t ltime;
1836
1837         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1838                 return (0);     /* needless to copy */
1839
1840         IFNET_RLOCK_NOSLEEP();
1841         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1842   again:
1843
1844         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1845                 IF_ADDR_RLOCK(ifp);
1846                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1847                         if (ifa->ifa_addr->sa_family != AF_INET6)
1848                                 continue;
1849                         ifa6 = (struct in6_ifaddr *)ifa;
1850
1851                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1852                             allow_deprecated == 0) {
1853                                 /*
1854                                  * prefererred address should be put before
1855                                  * deprecated addresses.
1856                                  */
1857
1858                                 /* record the interface for later search */
1859                                 if (ifp_dep == NULL)
1860                                         ifp_dep = ifp;
1861
1862                                 continue;
1863                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1864                             allow_deprecated != 0)
1865                                 continue; /* we now collect deprecated addrs */
1866
1867                         /* What do we have to do about ::1? */
1868                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1869                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1870                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1871                                         continue;
1872                                 break;
1873                         case IPV6_ADDR_SCOPE_SITELOCAL:
1874                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1875                                         continue;
1876                                 break;
1877                         case IPV6_ADDR_SCOPE_GLOBAL:
1878                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1879                                         continue;
1880                                 break;
1881                         default:
1882                                 continue;
1883                         }
1884
1885                         /*
1886                          * check if anycast is okay.
1887                          * XXX: just experimental.  not in the spec.
1888                          */
1889                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1890                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1891                                 continue;
1892                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1893                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1894                                 continue;
1895                         }
1896
1897                         /* now we can copy the address */
1898                         if (resid < sizeof(struct in6_addr) +
1899                             sizeof(u_int32_t)) {
1900                                 IF_ADDR_RUNLOCK(ifp);
1901                                 /*
1902                                  * We give up much more copy.
1903                                  * Set the truncate flag and return.
1904                                  */
1905                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1906                                 IFNET_RUNLOCK_NOSLEEP();
1907                                 return (copied);
1908                         }
1909
1910                         /*
1911                          * Set the TTL of the address.
1912                          * The TTL value should be one of the following
1913                          * according to the specification:
1914                          *
1915                          * 1. The remaining lifetime of a DHCP lease on the
1916                          *    address, or
1917                          * 2. The remaining Valid Lifetime of a prefix from
1918                          *    which the address was derived through Stateless
1919                          *    Autoconfiguration.
1920                          *
1921                          * Note that we currently do not support stateful
1922                          * address configuration by DHCPv6, so the former
1923                          * case can't happen.
1924                          */
1925                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1926                                 ltime = ND6_INFINITE_LIFETIME;
1927                         else {
1928                                 if (ifa6->ia6_lifetime.ia6t_expire >
1929                                     time_uptime)
1930                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1931                                 else
1932                                         ltime = 0;
1933                         }
1934
1935                         bcopy(&ltime, cp, sizeof(u_int32_t));
1936                         cp += sizeof(u_int32_t);
1937
1938                         /* copy the address itself */
1939                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1940                             sizeof(struct in6_addr));
1941                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1942                         cp += sizeof(struct in6_addr);
1943
1944                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1945                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1946                 }
1947                 IF_ADDR_RUNLOCK(ifp);
1948                 if (ifp0)       /* we need search only on the specified IF */
1949                         break;
1950         }
1951
1952         if (allow_deprecated == 0 && ifp_dep != NULL) {
1953                 ifp = ifp_dep;
1954                 allow_deprecated = 1;
1955
1956                 goto again;
1957         }
1958
1959         IFNET_RUNLOCK_NOSLEEP();
1960
1961         return (copied);
1962 }
1963
1964 /*
1965  * XXX almost dup'ed code with rip6_input.
1966  */
1967 static int
1968 icmp6_rip6_input(struct mbuf **mp, int off)
1969 {
1970         struct mbuf *m = *mp;
1971         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1972         struct inpcb *in6p;
1973         struct inpcb *last = NULL;
1974         struct sockaddr_in6 fromsa;
1975         struct icmp6_hdr *icmp6;
1976         struct mbuf *opts = NULL;
1977
1978 #ifndef PULLDOWN_TEST
1979         /* this is assumed to be safe. */
1980         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1981 #else
1982         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1983         if (icmp6 == NULL) {
1984                 /* m is already reclaimed */
1985                 return (IPPROTO_DONE);
1986         }
1987 #endif
1988
1989         /*
1990          * XXX: the address may have embedded scope zone ID, which should be
1991          * hidden from applications.
1992          */
1993         bzero(&fromsa, sizeof(fromsa));
1994         fromsa.sin6_family = AF_INET6;
1995         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1996         fromsa.sin6_addr = ip6->ip6_src;
1997         if (sa6_recoverscope(&fromsa)) {
1998                 m_freem(m);
1999                 return (IPPROTO_DONE);
2000         }
2001
2002         INP_INFO_RLOCK(&V_ripcbinfo);
2003         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2004                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2005                         continue;
2006                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2007                         continue;
2008                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2009                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2010                         continue;
2011                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2012                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2013                         continue;
2014                 INP_RLOCK(in6p);
2015                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2016                     in6p->in6p_icmp6filt)) {
2017                         INP_RUNLOCK(in6p);
2018                         continue;
2019                 }
2020                 if (last != NULL) {
2021                         struct  mbuf *n = NULL;
2022
2023                         /*
2024                          * Recent network drivers tend to allocate a single
2025                          * mbuf cluster, rather than to make a couple of
2026                          * mbufs without clusters.  Also, since the IPv6 code
2027                          * path tries to avoid m_pullup(), it is highly
2028                          * probable that we still have an mbuf cluster here
2029                          * even though the necessary length can be stored in an
2030                          * mbuf's internal buffer.
2031                          * Meanwhile, the default size of the receive socket
2032                          * buffer for raw sockets is not so large.  This means
2033                          * the possibility of packet loss is relatively higher
2034                          * than before.  To avoid this scenario, we copy the
2035                          * received data to a separate mbuf that does not use
2036                          * a cluster, if possible.
2037                          * XXX: it is better to copy the data after stripping
2038                          * intermediate headers.
2039                          */
2040                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2041                             m->m_len <= MHLEN) {
2042                                 n = m_get(M_NOWAIT, m->m_type);
2043                                 if (n != NULL) {
2044                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2045                                                 bcopy(m->m_data, n->m_data,
2046                                                       m->m_len);
2047                                                 n->m_len = m->m_len;
2048                                         } else {
2049                                                 m_free(n);
2050                                                 n = NULL;
2051                                         }
2052                                 }
2053                         }
2054                         if (n != NULL ||
2055                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2056                                 if (last->inp_flags & INP_CONTROLOPTS)
2057                                         ip6_savecontrol(last, n, &opts);
2058                                 /* strip intermediate headers */
2059                                 m_adj(n, off);
2060                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2061                                 if (sbappendaddr_locked(
2062                                     &last->inp_socket->so_rcv,
2063                                     (struct sockaddr *)&fromsa, n, opts)
2064                                     == 0) {
2065                                         /* should notify about lost packet */
2066                                         m_freem(n);
2067                                         if (opts) {
2068                                                 m_freem(opts);
2069                                         }
2070                                         SOCKBUF_UNLOCK(
2071                                             &last->inp_socket->so_rcv);
2072                                 } else
2073                                         sorwakeup_locked(last->inp_socket);
2074                                 opts = NULL;
2075                         }
2076                         INP_RUNLOCK(last);
2077                 }
2078                 last = in6p;
2079         }
2080         INP_INFO_RUNLOCK(&V_ripcbinfo);
2081         if (last != NULL) {
2082                 if (last->inp_flags & INP_CONTROLOPTS)
2083                         ip6_savecontrol(last, m, &opts);
2084                 /* strip intermediate headers */
2085                 m_adj(m, off);
2086
2087                 /* avoid using mbuf clusters if possible (see above) */
2088                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2089                     m->m_len <= MHLEN) {
2090                         struct mbuf *n;
2091
2092                         n = m_get(M_NOWAIT, m->m_type);
2093                         if (n != NULL) {
2094                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2095                                         bcopy(m->m_data, n->m_data, m->m_len);
2096                                         n->m_len = m->m_len;
2097
2098                                         m_freem(m);
2099                                         m = n;
2100                                 } else {
2101                                         m_freem(n);
2102                                         n = NULL;
2103                                 }
2104                         }
2105                 }
2106                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2107                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2108                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2109                         m_freem(m);
2110                         if (opts)
2111                                 m_freem(opts);
2112                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2113                 } else
2114                         sorwakeup_locked(last->inp_socket);
2115                 INP_RUNLOCK(last);
2116         } else {
2117                 m_freem(m);
2118                 IP6STAT_DEC(ip6s_delivered);
2119         }
2120         return IPPROTO_DONE;
2121 }
2122
2123 /*
2124  * Reflect the ip6 packet back to the source.
2125  * OFF points to the icmp6 header, counted from the top of the mbuf.
2126  */
2127 void
2128 icmp6_reflect(struct mbuf *m, size_t off)
2129 {
2130         struct ip6_hdr *ip6;
2131         struct icmp6_hdr *icmp6;
2132         struct in6_ifaddr *ia = NULL;
2133         int plen;
2134         int type, code;
2135         struct ifnet *outif = NULL;
2136         struct in6_addr origdst, src, *srcp = NULL;
2137
2138         /* too short to reflect */
2139         if (off < sizeof(struct ip6_hdr)) {
2140                 nd6log((LOG_DEBUG,
2141                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2142                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2143                     __FILE__, __LINE__));
2144                 goto bad;
2145         }
2146
2147         /*
2148          * If there are extra headers between IPv6 and ICMPv6, strip
2149          * off that header first.
2150          */
2151 #ifdef DIAGNOSTIC
2152         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2153                 panic("assumption failed in icmp6_reflect");
2154 #endif
2155         if (off > sizeof(struct ip6_hdr)) {
2156                 size_t l;
2157                 struct ip6_hdr nip6;
2158
2159                 l = off - sizeof(struct ip6_hdr);
2160                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2161                 m_adj(m, l);
2162                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2163                 if (m->m_len < l) {
2164                         if ((m = m_pullup(m, l)) == NULL)
2165                                 return;
2166                 }
2167                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2168         } else /* off == sizeof(struct ip6_hdr) */ {
2169                 size_t l;
2170                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2171                 if (m->m_len < l) {
2172                         if ((m = m_pullup(m, l)) == NULL)
2173                                 return;
2174                 }
2175         }
2176         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2177         ip6 = mtod(m, struct ip6_hdr *);
2178         ip6->ip6_nxt = IPPROTO_ICMPV6;
2179         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2180         type = icmp6->icmp6_type; /* keep type for statistics */
2181         code = icmp6->icmp6_code; /* ditto. */
2182
2183         origdst = ip6->ip6_dst;
2184         /*
2185          * ip6_input() drops a packet if its src is multicast.
2186          * So, the src is never multicast.
2187          */
2188         ip6->ip6_dst = ip6->ip6_src;
2189
2190         /*
2191          * If the incoming packet was addressed directly to us (i.e. unicast),
2192          * use dst as the src for the reply.
2193          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2194          * (for example) when we encounter an error while forwarding procedure
2195          * destined to a duplicated address of ours.
2196          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2197          * procedure of an outgoing packet of our own, in which case we need
2198          * to search in the ifaddr list.
2199          */
2200         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2201                 if ((ia = ip6_getdstifaddr(m))) {
2202                         if (!(ia->ia6_flags &
2203                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2204                                 srcp = &ia->ia_addr.sin6_addr;
2205                 } else {
2206                         struct sockaddr_in6 d;
2207
2208                         bzero(&d, sizeof(d));
2209                         d.sin6_family = AF_INET6;
2210                         d.sin6_len = sizeof(d);
2211                         d.sin6_addr = origdst;
2212                         ia = (struct in6_ifaddr *)
2213                             ifa_ifwithaddr((struct sockaddr *)&d);
2214                         if (ia &&
2215                             !(ia->ia6_flags &
2216                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2217                                 srcp = &ia->ia_addr.sin6_addr;
2218                         }
2219                 }
2220         }
2221
2222         if (srcp == NULL) {
2223                 int e;
2224                 struct sockaddr_in6 sin6;
2225                 struct route_in6 ro;
2226
2227                 /*
2228                  * This case matches to multicasts, our anycast, or unicasts
2229                  * that we do not own.  Select a source address based on the
2230                  * source address of the erroneous packet.
2231                  */
2232                 bzero(&sin6, sizeof(sin6));
2233                 sin6.sin6_family = AF_INET6;
2234                 sin6.sin6_len = sizeof(sin6);
2235                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2236
2237                 bzero(&ro, sizeof(ro));
2238                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2239                 if (ro.ro_rt)
2240                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2241                 if (e) {
2242                         char ip6buf[INET6_ADDRSTRLEN];
2243                         nd6log((LOG_DEBUG,
2244                             "icmp6_reflect: source can't be determined: "
2245                             "dst=%s, error=%d\n",
2246                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2247                         goto bad;
2248                 }
2249                 srcp = &src;
2250         }
2251
2252         ip6->ip6_src = *srcp;
2253         ip6->ip6_flow = 0;
2254         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2255         ip6->ip6_vfc |= IPV6_VERSION;
2256         ip6->ip6_nxt = IPPROTO_ICMPV6;
2257         if (outif)
2258                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2259         else if (m->m_pkthdr.rcvif) {
2260                 /* XXX: This may not be the outgoing interface */
2261                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2262         } else
2263                 ip6->ip6_hlim = V_ip6_defhlim;
2264
2265         icmp6->icmp6_cksum = 0;
2266         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2267             sizeof(struct ip6_hdr), plen);
2268
2269         /*
2270          * XXX option handling
2271          */
2272
2273         m->m_flags &= ~(M_BCAST|M_MCAST);
2274
2275         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2276         if (outif)
2277                 icmp6_ifoutstat_inc(outif, type, code);
2278
2279         if (ia != NULL)
2280                 ifa_free(&ia->ia_ifa);
2281         return;
2282
2283  bad:
2284         if (ia != NULL)
2285                 ifa_free(&ia->ia_ifa);
2286         m_freem(m);
2287         return;
2288 }
2289
2290 void
2291 icmp6_fasttimo(void)
2292 {
2293
2294         mld_fasttimo();
2295 }
2296
2297 void
2298 icmp6_slowtimo(void)
2299 {
2300
2301         mld_slowtimo();
2302 }
2303
2304 static const char *
2305 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2306     struct in6_addr *tgt6)
2307 {
2308         static char buf[1024];
2309         char ip6bufs[INET6_ADDRSTRLEN];
2310         char ip6bufd[INET6_ADDRSTRLEN];
2311         char ip6buft[INET6_ADDRSTRLEN];
2312         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2313             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2314             ip6_sprintf(ip6buft, tgt6));
2315         return buf;
2316 }
2317
2318 void
2319 icmp6_redirect_input(struct mbuf *m, int off)
2320 {
2321         struct ifnet *ifp;
2322         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2323         struct nd_redirect *nd_rd;
2324         int icmp6len = ntohs(ip6->ip6_plen);
2325         char *lladdr = NULL;
2326         int lladdrlen = 0;
2327         struct rtentry *rt = NULL;
2328         int is_router;
2329         int is_onlink;
2330         struct in6_addr src6 = ip6->ip6_src;
2331         struct in6_addr redtgt6;
2332         struct in6_addr reddst6;
2333         union nd_opts ndopts;
2334         char ip6buf[INET6_ADDRSTRLEN];
2335
2336         M_ASSERTPKTHDR(m);
2337         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2338
2339         ifp = m->m_pkthdr.rcvif;
2340
2341         /* XXX if we are router, we don't update route by icmp6 redirect */
2342         if (V_ip6_forwarding)
2343                 goto freeit;
2344         if (!V_icmp6_rediraccept)
2345                 goto freeit;
2346
2347 #ifndef PULLDOWN_TEST
2348         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2349         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2350 #else
2351         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2352         if (nd_rd == NULL) {
2353                 ICMP6STAT_INC(icp6s_tooshort);
2354                 return;
2355         }
2356 #endif
2357         redtgt6 = nd_rd->nd_rd_target;
2358         reddst6 = nd_rd->nd_rd_dst;
2359
2360         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2361             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2362                 goto freeit;
2363         }
2364
2365         /* validation */
2366         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2367                 nd6log((LOG_ERR,
2368                     "ICMP6 redirect sent from %s rejected; "
2369                     "must be from linklocal\n",
2370                     ip6_sprintf(ip6buf, &src6)));
2371                 goto bad;
2372         }
2373         if (ip6->ip6_hlim != 255) {
2374                 nd6log((LOG_ERR,
2375                     "ICMP6 redirect sent from %s rejected; "
2376                     "hlim=%d (must be 255)\n",
2377                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2378                 goto bad;
2379         }
2380     {
2381         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2382         struct sockaddr_in6 sin6;
2383         struct in6_addr *gw6;
2384
2385         bzero(&sin6, sizeof(sin6));
2386         sin6.sin6_family = AF_INET6;
2387         sin6.sin6_len = sizeof(struct sockaddr_in6);
2388         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2389         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2390         if (rt) {
2391                 if (rt->rt_gateway == NULL ||
2392                     rt->rt_gateway->sa_family != AF_INET6) {
2393                         RTFREE_LOCKED(rt);
2394                         nd6log((LOG_ERR,
2395                             "ICMP6 redirect rejected; no route "
2396                             "with inet6 gateway found for redirect dst: %s\n",
2397                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2398                         goto bad;
2399                 }
2400
2401                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2402                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2403                         RTFREE_LOCKED(rt);
2404                         nd6log((LOG_ERR,
2405                             "ICMP6 redirect rejected; "
2406                             "not equal to gw-for-src=%s (must be same): "
2407                             "%s\n",
2408                             ip6_sprintf(ip6buf, gw6),
2409                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2410                         goto bad;
2411                 }
2412         } else {
2413                 nd6log((LOG_ERR,
2414                     "ICMP6 redirect rejected; "
2415                     "no route found for redirect dst: %s\n",
2416                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2417                 goto bad;
2418         }
2419         RTFREE_LOCKED(rt);
2420         rt = NULL;
2421     }
2422         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2423                 nd6log((LOG_ERR,
2424                     "ICMP6 redirect rejected; "
2425                     "redirect dst must be unicast: %s\n",
2426                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2427                 goto bad;
2428         }
2429
2430         is_router = is_onlink = 0;
2431         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2432                 is_router = 1;  /* router case */
2433         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2434                 is_onlink = 1;  /* on-link destination case */
2435         if (!is_router && !is_onlink) {
2436                 nd6log((LOG_ERR,
2437                     "ICMP6 redirect rejected; "
2438                     "neither router case nor onlink case: %s\n",
2439                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2440                 goto bad;
2441         }
2442         /* validation passed */
2443
2444         icmp6len -= sizeof(*nd_rd);
2445         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2446         if (nd6_options(&ndopts) < 0) {
2447                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2448                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2449                 /* nd6_options have incremented stats */
2450                 goto freeit;
2451         }
2452
2453         if (ndopts.nd_opts_tgt_lladdr) {
2454                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2455                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2456         }
2457
2458         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2459                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2460                     "(if %d, icmp6 packet %d): %s\n",
2461                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2462                     ifp->if_addrlen, lladdrlen - 2,
2463                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2464                 goto bad;
2465         }
2466
2467         /* RFC 2461 8.3 */
2468         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2469             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2470
2471         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2472                 /* perform rtredirect */
2473                 struct sockaddr_in6 sdst;
2474                 struct sockaddr_in6 sgw;
2475                 struct sockaddr_in6 ssrc;
2476                 u_int fibnum;
2477
2478                 bzero(&sdst, sizeof(sdst));
2479                 bzero(&sgw, sizeof(sgw));
2480                 bzero(&ssrc, sizeof(ssrc));
2481                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2482                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2483                         sizeof(struct sockaddr_in6);
2484                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2485                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2486                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2487                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2488                         in6_rtredirect((struct sockaddr *)&sdst,
2489                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2490                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2491                             fibnum);
2492         }
2493         /* finally update cached route in each socket via pfctlinput */
2494     {
2495         struct sockaddr_in6 sdst;
2496
2497         bzero(&sdst, sizeof(sdst));
2498         sdst.sin6_family = AF_INET6;
2499         sdst.sin6_len = sizeof(struct sockaddr_in6);
2500         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2501         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2502     }
2503
2504  freeit:
2505         m_freem(m);
2506         return;
2507
2508  bad:
2509         ICMP6STAT_INC(icp6s_badredirect);
2510         m_freem(m);
2511 }
2512
2513 void
2514 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2515 {
2516         struct ifnet *ifp;      /* my outgoing interface */
2517         struct in6_addr *ifp_ll6;
2518         struct in6_addr *router_ll6;
2519         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2520         struct mbuf *m = NULL;  /* newly allocated one */
2521         struct m_tag *mtag;
2522         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2523         struct nd_redirect *nd_rd;
2524         struct llentry *ln = NULL;
2525         size_t maxlen;
2526         u_char *p;
2527         struct ifnet *outif = NULL;
2528         struct sockaddr_in6 src_sa;
2529
2530         icmp6_errcount(ND_REDIRECT, 0);
2531
2532         /* if we are not router, we don't send icmp6 redirect */
2533         if (!V_ip6_forwarding)
2534                 goto fail;
2535
2536         /* sanity check */
2537         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2538                 goto fail;
2539
2540         /*
2541          * Address check:
2542          *  the source address must identify a neighbor, and
2543          *  the destination address must not be a multicast address
2544          *  [RFC 2461, sec 8.2]
2545          */
2546         sip6 = mtod(m0, struct ip6_hdr *);
2547         bzero(&src_sa, sizeof(src_sa));
2548         src_sa.sin6_family = AF_INET6;
2549         src_sa.sin6_len = sizeof(src_sa);
2550         src_sa.sin6_addr = sip6->ip6_src;
2551         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2552                 goto fail;
2553         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2554                 goto fail;      /* what should we do here? */
2555
2556         /* rate limit */
2557         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2558                 goto fail;
2559
2560         /*
2561          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2562          * we almost always ask for an mbuf cluster for simplicity.
2563          * (MHLEN < IPV6_MMTU is almost always true)
2564          */
2565 #if IPV6_MMTU >= MCLBYTES
2566 # error assumption failed about IPV6_MMTU and MCLBYTES
2567 #endif
2568         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2569         if (m == NULL)
2570                 goto fail;
2571         M_SETFIB(m, rt->rt_fibnum);
2572         maxlen = M_TRAILINGSPACE(m);
2573         maxlen = min(IPV6_MMTU, maxlen);
2574         /* just for safety */
2575         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2576             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2577                 goto fail;
2578         }
2579
2580         {
2581                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2582                 struct in6_ifaddr *ia;
2583                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2584                                                  IN6_IFF_NOTREADY|
2585                                                  IN6_IFF_ANYCAST)) == NULL)
2586                         goto fail;
2587                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2588                 /* XXXRW: reference released prematurely. */
2589                 ifa_free(&ia->ia_ifa);
2590         }
2591
2592         /* get ip6 linklocal address for the router. */
2593         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2594                 struct sockaddr_in6 *sin6;
2595                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2596                 router_ll6 = &sin6->sin6_addr;
2597                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2598                         router_ll6 = (struct in6_addr *)NULL;
2599         } else
2600                 router_ll6 = (struct in6_addr *)NULL;
2601
2602         /* ip6 */
2603         ip6 = mtod(m, struct ip6_hdr *);
2604         ip6->ip6_flow = 0;
2605         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2606         ip6->ip6_vfc |= IPV6_VERSION;
2607         /* ip6->ip6_plen will be set later */
2608         ip6->ip6_nxt = IPPROTO_ICMPV6;
2609         ip6->ip6_hlim = 255;
2610         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2611         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2612         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2613
2614         /* ND Redirect */
2615         nd_rd = (struct nd_redirect *)(ip6 + 1);
2616         nd_rd->nd_rd_type = ND_REDIRECT;
2617         nd_rd->nd_rd_code = 0;
2618         nd_rd->nd_rd_reserved = 0;
2619         if (rt->rt_flags & RTF_GATEWAY) {
2620                 /*
2621                  * nd_rd->nd_rd_target must be a link-local address in
2622                  * better router cases.
2623                  */
2624                 if (!router_ll6)
2625                         goto fail;
2626                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2627                     sizeof(nd_rd->nd_rd_target));
2628                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2629                     sizeof(nd_rd->nd_rd_dst));
2630         } else {
2631                 /* make sure redtgt == reddst */
2632                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2633                     sizeof(nd_rd->nd_rd_target));
2634                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2635                     sizeof(nd_rd->nd_rd_dst));
2636         }
2637
2638         p = (u_char *)(nd_rd + 1);
2639
2640         if (!router_ll6)
2641                 goto nolladdropt;
2642
2643         {
2644                 /* target lladdr option */
2645                 int len;
2646                 struct nd_opt_hdr *nd_opt;
2647                 char *lladdr;
2648
2649                 IF_AFDATA_RLOCK(ifp);
2650                 ln = nd6_lookup(router_ll6, 0, ifp);
2651                 IF_AFDATA_RUNLOCK(ifp);
2652                 if (ln == NULL)
2653                         goto nolladdropt;
2654
2655                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2656                 len = (len + 7) & ~7;   /* round by 8 */
2657                 /* safety check */
2658                 if (len + (p - (u_char *)ip6) > maxlen)                         
2659                         goto nolladdropt;
2660
2661                 if (ln->la_flags & LLE_VALID) {
2662                         nd_opt = (struct nd_opt_hdr *)p;
2663                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2664                         nd_opt->nd_opt_len = len >> 3;
2665                         lladdr = (char *)(nd_opt + 1);
2666                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2667                         p += len;
2668                 }
2669         }
2670 nolladdropt:
2671         if (ln != NULL)
2672                 LLE_RUNLOCK(ln);
2673                 
2674         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2675
2676         /* just to be safe */
2677 #ifdef M_DECRYPTED      /*not openbsd*/
2678         if (m0->m_flags & M_DECRYPTED)
2679                 goto noredhdropt;
2680 #endif
2681         if (p - (u_char *)ip6 > maxlen)
2682                 goto noredhdropt;
2683
2684         {
2685                 /* redirected header option */
2686                 int len;
2687                 struct nd_opt_rd_hdr *nd_opt_rh;
2688
2689                 /*
2690                  * compute the maximum size for icmp6 redirect header option.
2691                  * XXX room for auth header?
2692                  */
2693                 len = maxlen - (p - (u_char *)ip6);
2694                 len &= ~7;
2695
2696                 /* This is just for simplicity. */
2697                 if (m0->m_pkthdr.len != m0->m_len) {
2698                         if (m0->m_next) {
2699                                 m_freem(m0->m_next);
2700                                 m0->m_next = NULL;
2701                         }
2702                         m0->m_pkthdr.len = m0->m_len;
2703                 }
2704
2705                 /*
2706                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2707                  * about padding/truncate rule for the original IP packet.
2708                  * From the discussion on IPv6imp in Feb 1999,
2709                  * the consensus was:
2710                  * - "attach as much as possible" is the goal
2711                  * - pad if not aligned (original size can be guessed by
2712                  *   original ip6 header)
2713                  * Following code adds the padding if it is simple enough,
2714                  * and truncates if not.
2715                  */
2716                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2717                         panic("assumption failed in %s:%d", __FILE__,
2718                             __LINE__);
2719
2720                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2721                         /* not enough room, truncate */
2722                         m0->m_pkthdr.len = m0->m_len = len -
2723                             sizeof(*nd_opt_rh);
2724                 } else {
2725                         /* enough room, pad or truncate */
2726                         size_t extra;
2727
2728                         extra = m0->m_pkthdr.len % 8;
2729                         if (extra) {
2730                                 /* pad if easy enough, truncate if not */
2731                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2732                                         /* pad */
2733                                         m0->m_len += (8 - extra);
2734                                         m0->m_pkthdr.len += (8 - extra);
2735                                 } else {
2736                                         /* truncate */
2737                                         m0->m_pkthdr.len -= extra;
2738                                         m0->m_len -= extra;
2739                                 }
2740                         }
2741                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2742                         m0->m_pkthdr.len = m0->m_len = len -
2743                             sizeof(*nd_opt_rh);
2744                 }
2745
2746                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2747                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2748                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2749                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2750                 p += sizeof(*nd_opt_rh);
2751                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2752
2753                 /* connect m0 to m */
2754                 m_tag_delete_chain(m0, NULL);
2755                 m0->m_flags &= ~M_PKTHDR;
2756                 m->m_next = m0;
2757                 m->m_pkthdr.len = m->m_len + m0->m_len;
2758                 m0 = NULL;
2759         }
2760 noredhdropt:;
2761         if (m0) {
2762                 m_freem(m0);
2763                 m0 = NULL;
2764         }
2765
2766         /* XXX: clear embedded link IDs in the inner header */
2767         in6_clearscope(&sip6->ip6_src);
2768         in6_clearscope(&sip6->ip6_dst);
2769         in6_clearscope(&nd_rd->nd_rd_target);
2770         in6_clearscope(&nd_rd->nd_rd_dst);
2771
2772         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2773
2774         nd_rd->nd_rd_cksum = 0;
2775         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2776             sizeof(*ip6), ntohs(ip6->ip6_plen));
2777
2778         if (send_sendso_input_hook != NULL) {
2779                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2780                         M_NOWAIT);
2781                 if (mtag == NULL)
2782                         goto fail;
2783                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2784                 m_tag_prepend(m, mtag);
2785         }
2786
2787         /* send the packet to outside... */
2788         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2789         if (outif) {
2790                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2791                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2792         }
2793         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2794
2795         return;
2796
2797 fail:
2798         if (m)
2799                 m_freem(m);
2800         if (m0)
2801                 m_freem(m0);
2802 }
2803
2804 /*
2805  * ICMPv6 socket option processing.
2806  */
2807 int
2808 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2809 {
2810         int error = 0;
2811         int optlen;
2812         struct inpcb *inp = sotoinpcb(so);
2813         int level, op, optname;
2814
2815         if (sopt) {
2816                 level = sopt->sopt_level;
2817                 op = sopt->sopt_dir;
2818                 optname = sopt->sopt_name;
2819                 optlen = sopt->sopt_valsize;
2820         } else
2821                 level = op = optname = optlen = 0;
2822
2823         if (level != IPPROTO_ICMPV6) {
2824                 return EINVAL;
2825         }
2826
2827         switch (op) {
2828         case PRCO_SETOPT:
2829                 switch (optname) {
2830                 case ICMP6_FILTER:
2831                     {
2832                         struct icmp6_filter ic6f;
2833
2834                         if (optlen != sizeof(ic6f)) {
2835                                 error = EMSGSIZE;
2836                                 break;
2837                         }
2838                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2839                         if (error == 0) {
2840                                 INP_WLOCK(inp);
2841                                 *inp->in6p_icmp6filt = ic6f;
2842                                 INP_WUNLOCK(inp);
2843                         }
2844                         break;
2845                     }
2846
2847                 default:
2848                         error = ENOPROTOOPT;
2849                         break;
2850                 }
2851                 break;
2852
2853         case PRCO_GETOPT:
2854                 switch (optname) {
2855                 case ICMP6_FILTER:
2856                     {
2857                         struct icmp6_filter ic6f;
2858
2859                         INP_RLOCK(inp);
2860                         ic6f = *inp->in6p_icmp6filt;
2861                         INP_RUNLOCK(inp);
2862                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2863                         break;
2864                     }
2865
2866                 default:
2867                         error = ENOPROTOOPT;
2868                         break;
2869                 }
2870                 break;
2871         }
2872
2873         return (error);
2874 }
2875
2876 /*
2877  * Perform rate limit check.
2878  * Returns 0 if it is okay to send the icmp6 packet.
2879  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2880  * limitation.
2881  *
2882  * XXX per-destination/type check necessary?
2883  *
2884  * dst - not used at this moment
2885  * type - not used at this moment
2886  * code - not used at this moment
2887  */
2888 static int
2889 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2890     const int code)
2891 {
2892         int ret;
2893
2894         ret = 0;        /* okay to send */
2895
2896         /* PPS limit */
2897         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2898             V_icmp6errppslim)) {
2899                 /* The packet is subject to rate limit */
2900                 ret++;
2901         }
2902
2903         return ret;
2904 }