]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/netinet6/icmp6.c
Fix panic with RADIX_MPATH, when RTFREE_LOCKED() called for already
[FreeBSD/FreeBSD.git] / sys / netinet6 / icmp6.c
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *      The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
61  */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_dl.h>
90 #include <net/if_llatbl.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 #include <net/vnet.h>
94
95 #include <netinet/in.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet/icmp6.h>
100 #include <netinet/tcp_var.h>
101
102 #include <netinet6/in6_ifattach.h>
103 #include <netinet6/in6_pcb.h>
104 #include <netinet6/ip6protosw.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet6/scope6_var.h>
107 #include <netinet6/mld6_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/send.h>
110
111 #ifdef IPSEC
112 #include <netipsec/ipsec.h>
113 #include <netipsec/key.h>
114 #endif
115
116 extern struct domain inet6domain;
117
118 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
119 VNET_PCPUSTAT_SYSINIT(icmp6stat);
120
121 #ifdef VIMAGE
122 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
123 #endif /* VIMAGE */
124
125 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
126 VNET_DECLARE(struct inpcbhead, ripcb);
127 VNET_DECLARE(int, icmp6errppslim);
128 static VNET_DEFINE(int, icmp6errpps_count) = 0;
129 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
130 VNET_DECLARE(int, icmp6_nodeinfo);
131
132 #define V_ripcbinfo                     VNET(ripcbinfo)
133 #define V_ripcb                         VNET(ripcb)
134 #define V_icmp6errppslim                VNET(icmp6errppslim)
135 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
136 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
137 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
138
139 static void icmp6_errcount(int, int);
140 static int icmp6_rip6_input(struct mbuf **, int);
141 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
142 static const char *icmp6_redirect_diag(struct in6_addr *,
143         struct in6_addr *, struct in6_addr *);
144 static struct mbuf *ni6_input(struct mbuf *, int);
145 static struct mbuf *ni6_nametodns(const char *, int, int);
146 static int ni6_dnsmatch(const char *, int, const char *, int);
147 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
148                           struct ifnet **, struct in6_addr *);
149 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
150                                 struct ifnet *, int);
151 static int icmp6_notify_error(struct mbuf **, int, int, int);
152
153 /*
154  * Kernel module interface for updating icmp6stat.  The argument is an index
155  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
156  * general layout of icmp6stat into the caller, it doesn't encode its
157  * location, so that future changes to add, for example, per-CPU stats
158  * support won't cause binary compatibility problems for kernel modules.
159  */
160 void
161 kmod_icmp6stat_inc(int statnum)
162 {
163
164         counter_u64_add(VNET(icmp6stat)[statnum], 1);
165 }
166
167 static void
168 icmp6_errcount(int type, int code)
169 {
170         switch (type) {
171         case ICMP6_DST_UNREACH:
172                 switch (code) {
173                 case ICMP6_DST_UNREACH_NOROUTE:
174                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
175                         return;
176                 case ICMP6_DST_UNREACH_ADMIN:
177                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
178                         return;
179                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
180                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
181                         return;
182                 case ICMP6_DST_UNREACH_ADDR:
183                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
184                         return;
185                 case ICMP6_DST_UNREACH_NOPORT:
186                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
187                         return;
188                 }
189                 break;
190         case ICMP6_PACKET_TOO_BIG:
191                 ICMP6STAT_INC(icp6s_opacket_too_big);
192                 return;
193         case ICMP6_TIME_EXCEEDED:
194                 switch (code) {
195                 case ICMP6_TIME_EXCEED_TRANSIT:
196                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
197                         return;
198                 case ICMP6_TIME_EXCEED_REASSEMBLY:
199                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
200                         return;
201                 }
202                 break;
203         case ICMP6_PARAM_PROB:
204                 switch (code) {
205                 case ICMP6_PARAMPROB_HEADER:
206                         ICMP6STAT_INC(icp6s_oparamprob_header);
207                         return;
208                 case ICMP6_PARAMPROB_NEXTHEADER:
209                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
210                         return;
211                 case ICMP6_PARAMPROB_OPTION:
212                         ICMP6STAT_INC(icp6s_oparamprob_option);
213                         return;
214                 }
215                 break;
216         case ND_REDIRECT:
217                 ICMP6STAT_INC(icp6s_oredirect);
218                 return;
219         }
220         ICMP6STAT_INC(icp6s_ounknown);
221 }
222
223 /*
224  * A wrapper function for icmp6_error() necessary when the erroneous packet
225  * may not contain enough scope zone information.
226  */
227 void
228 icmp6_error2(struct mbuf *m, int type, int code, int param,
229     struct ifnet *ifp)
230 {
231         struct ip6_hdr *ip6;
232
233         if (ifp == NULL)
234                 return;
235
236 #ifndef PULLDOWN_TEST
237         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
238 #else
239         if (m->m_len < sizeof(struct ip6_hdr)) {
240                 m = m_pullup(m, sizeof(struct ip6_hdr));
241                 if (m == NULL)
242                         return;
243         }
244 #endif
245
246         ip6 = mtod(m, struct ip6_hdr *);
247
248         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
249                 return;
250         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
251                 return;
252
253         icmp6_error(m, type, code, param);
254 }
255
256 /*
257  * Generate an error packet of type error in response to bad IP6 packet.
258  */
259 void
260 icmp6_error(struct mbuf *m, int type, int code, int param)
261 {
262         struct ip6_hdr *oip6, *nip6;
263         struct icmp6_hdr *icmp6;
264         u_int preplen;
265         int off;
266         int nxt;
267
268         ICMP6STAT_INC(icp6s_error);
269
270         /* count per-type-code statistics */
271         icmp6_errcount(type, code);
272
273 #ifdef M_DECRYPTED      /*not openbsd*/
274         if (m->m_flags & M_DECRYPTED) {
275                 ICMP6STAT_INC(icp6s_canterror);
276                 goto freeit;
277         }
278 #endif
279
280 #ifndef PULLDOWN_TEST
281         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
282 #else
283         if (m->m_len < sizeof(struct ip6_hdr)) {
284                 m = m_pullup(m, sizeof(struct ip6_hdr));
285                 if (m == NULL)
286                         return;
287         }
288 #endif
289         oip6 = mtod(m, struct ip6_hdr *);
290
291         /*
292          * If the destination address of the erroneous packet is a multicast
293          * address, or the packet was sent using link-layer multicast,
294          * we should basically suppress sending an error (RFC 2463, Section
295          * 2.4).
296          * We have two exceptions (the item e.2 in that section):
297          * - the Packet Too Big message can be sent for path MTU discovery.
298          * - the Parameter Problem Message that can be allowed an icmp6 error
299          *   in the option type field.  This check has been done in
300          *   ip6_unknown_opt(), so we can just check the type and code.
301          */
302         if ((m->m_flags & (M_BCAST|M_MCAST) ||
303              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
304             (type != ICMP6_PACKET_TOO_BIG &&
305              (type != ICMP6_PARAM_PROB ||
306               code != ICMP6_PARAMPROB_OPTION)))
307                 goto freeit;
308
309         /*
310          * RFC 2463, 2.4 (e.5): source address check.
311          * XXX: the case of anycast source?
312          */
313         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
314             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
315                 goto freeit;
316
317         /*
318          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
319          * don't do it.
320          */
321         nxt = -1;
322         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
323         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
324                 struct icmp6_hdr *icp;
325
326 #ifndef PULLDOWN_TEST
327                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
328                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
329 #else
330                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
331                         sizeof(*icp));
332                 if (icp == NULL) {
333                         ICMP6STAT_INC(icp6s_tooshort);
334                         return;
335                 }
336 #endif
337                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
338                     icp->icmp6_type == ND_REDIRECT) {
339                         /*
340                          * ICMPv6 error
341                          * Special case: for redirect (which is
342                          * informational) we must not send icmp6 error.
343                          */
344                         ICMP6STAT_INC(icp6s_canterror);
345                         goto freeit;
346                 } else {
347                         /* ICMPv6 informational - send the error */
348                 }
349         } else {
350                 /* non-ICMPv6 - send the error */
351         }
352
353         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
354
355         /* Finally, do rate limitation check. */
356         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
357                 ICMP6STAT_INC(icp6s_toofreq);
358                 goto freeit;
359         }
360
361         /*
362          * OK, ICMP6 can be generated.
363          */
364
365         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
366                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
367
368         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
369         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
370         if (m && m->m_len < preplen)
371                 m = m_pullup(m, preplen);
372         if (m == NULL) {
373                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
374                 return;
375         }
376
377         nip6 = mtod(m, struct ip6_hdr *);
378         nip6->ip6_src  = oip6->ip6_src;
379         nip6->ip6_dst  = oip6->ip6_dst;
380
381         in6_clearscope(&oip6->ip6_src);
382         in6_clearscope(&oip6->ip6_dst);
383
384         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
385         icmp6->icmp6_type = type;
386         icmp6->icmp6_code = code;
387         icmp6->icmp6_pptr = htonl((u_int32_t)param);
388
389         /*
390          * icmp6_reflect() is designed to be in the input path.
391          * icmp6_error() can be called from both input and output path,
392          * and if we are in output path rcvif could contain bogus value.
393          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
394          * information in ip header (nip6).
395          */
396         m->m_pkthdr.rcvif = NULL;
397
398         ICMP6STAT_INC(icp6s_outhist[type]);
399         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
400
401         return;
402
403   freeit:
404         /*
405          * If we can't tell whether or not we can generate ICMP6, free it.
406          */
407         m_freem(m);
408 }
409
410 /*
411  * Process a received ICMP6 message.
412  */
413 int
414 icmp6_input(struct mbuf **mp, int *offp, int proto)
415 {
416         struct mbuf *m = *mp, *n;
417         struct ifnet *ifp;
418         struct ip6_hdr *ip6, *nip6;
419         struct icmp6_hdr *icmp6, *nicmp6;
420         int off = *offp;
421         int icmp6len = m->m_pkthdr.len - *offp;
422         int code, sum, noff;
423         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
424         int ip6len, error;
425
426         ifp = m->m_pkthdr.rcvif;
427
428 #ifndef PULLDOWN_TEST
429         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
430         /* m might change if M_LOOP.  So, call mtod after this */
431 #endif
432
433         /*
434          * Locate icmp6 structure in mbuf, and check
435          * that not corrupted and of at least minimum length
436          */
437
438         ip6 = mtod(m, struct ip6_hdr *);
439         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
440         if (icmp6len < sizeof(struct icmp6_hdr)) {
441                 ICMP6STAT_INC(icp6s_tooshort);
442                 goto freeit;
443         }
444
445         /*
446          * Check multicast group membership.
447          * Note: SSM filters are not applied for ICMPv6 traffic.
448          */
449         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
450                 struct in6_multi        *inm;
451
452                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
453                 if (inm == NULL) {
454                         IP6STAT_INC(ip6s_notmember);
455                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
456                         goto freeit;
457                 }
458         }
459
460         /*
461          * calculate the checksum
462          */
463 #ifndef PULLDOWN_TEST
464         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
465 #else
466         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
467         if (icmp6 == NULL) {
468                 ICMP6STAT_INC(icp6s_tooshort);
469                 return IPPROTO_DONE;
470         }
471 #endif
472         code = icmp6->icmp6_code;
473
474         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
475                 nd6log((LOG_ERR,
476                     "ICMP6 checksum error(%d|%x) %s\n",
477                     icmp6->icmp6_type, sum,
478                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
479                 ICMP6STAT_INC(icp6s_checksum);
480                 goto freeit;
481         }
482
483         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
484                 /*
485                  * Deliver very specific ICMP6 type only.
486                  * This is important to deliver TOOBIG.  Otherwise PMTUD
487                  * will not work.
488                  */
489                 switch (icmp6->icmp6_type) {
490                 case ICMP6_DST_UNREACH:
491                 case ICMP6_PACKET_TOO_BIG:
492                 case ICMP6_TIME_EXCEEDED:
493                         break;
494                 default:
495                         goto freeit;
496                 }
497         }
498
499         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
500         icmp6_ifstat_inc(ifp, ifs6_in_msg);
501         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
502                 icmp6_ifstat_inc(ifp, ifs6_in_error);
503
504         switch (icmp6->icmp6_type) {
505         case ICMP6_DST_UNREACH:
506                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
507                 switch (code) {
508                 case ICMP6_DST_UNREACH_NOROUTE:
509                         code = PRC_UNREACH_NET;
510                         break;
511                 case ICMP6_DST_UNREACH_ADMIN:
512                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
513                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
514                         break;
515                 case ICMP6_DST_UNREACH_ADDR:
516                         code = PRC_HOSTDEAD;
517                         break;
518                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
519                         /* I mean "source address was incorrect." */
520                         code = PRC_PARAMPROB;
521                         break;
522                 case ICMP6_DST_UNREACH_NOPORT:
523                         code = PRC_UNREACH_PORT;
524                         break;
525                 default:
526                         goto badcode;
527                 }
528                 goto deliver;
529                 break;
530
531         case ICMP6_PACKET_TOO_BIG:
532                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
533
534                 /* validation is made in icmp6_mtudisc_update */
535
536                 code = PRC_MSGSIZE;
537
538                 /*
539                  * Updating the path MTU will be done after examining
540                  * intermediate extension headers.
541                  */
542                 goto deliver;
543                 break;
544
545         case ICMP6_TIME_EXCEEDED:
546                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
547                 switch (code) {
548                 case ICMP6_TIME_EXCEED_TRANSIT:
549                         code = PRC_TIMXCEED_INTRANS;
550                         break;
551                 case ICMP6_TIME_EXCEED_REASSEMBLY:
552                         code = PRC_TIMXCEED_REASS;
553                         break;
554                 default:
555                         goto badcode;
556                 }
557                 goto deliver;
558                 break;
559
560         case ICMP6_PARAM_PROB:
561                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
562                 switch (code) {
563                 case ICMP6_PARAMPROB_NEXTHEADER:
564                         code = PRC_UNREACH_PROTOCOL;
565                         break;
566                 case ICMP6_PARAMPROB_HEADER:
567                 case ICMP6_PARAMPROB_OPTION:
568                         code = PRC_PARAMPROB;
569                         break;
570                 default:
571                         goto badcode;
572                 }
573                 goto deliver;
574                 break;
575
576         case ICMP6_ECHO_REQUEST:
577                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
578                 if (code != 0)
579                         goto badcode;
580                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
581                         /* Give up remote */
582                         break;
583                 }
584                 if ((n->m_flags & M_EXT) != 0
585                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
586                         struct mbuf *n0 = n;
587                         int n0len;
588
589                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
590                         n = m_gethdr(M_NOWAIT, n0->m_type);
591                         if (n == NULL) {
592                                 /* Give up remote */
593                                 m_freem(n0);
594                                 break;
595                         }
596
597                         m_move_pkthdr(n, n0);   /* FIB copied. */
598                         n0len = n0->m_pkthdr.len;       /* save for use below */
599                         /*
600                          * Copy IPv6 and ICMPv6 only.
601                          */
602                         nip6 = mtod(n, struct ip6_hdr *);
603                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
604                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
605                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
606                         noff = sizeof(struct ip6_hdr);
607                         /* new mbuf contains only ipv6+icmpv6 headers */
608                         n->m_len = noff + sizeof(struct icmp6_hdr);
609                         /*
610                          * Adjust mbuf.  ip6_plen will be adjusted in
611                          * ip6_output().
612                          */
613                         m_adj(n0, off + sizeof(struct icmp6_hdr));
614                         /* recalculate complete packet size */
615                         n->m_pkthdr.len = n0len + (noff - off);
616                         n->m_next = n0;
617                 } else {
618                         nip6 = mtod(n, struct ip6_hdr *);
619                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
620                             sizeof(*nicmp6));
621                         noff = off;
622                 }
623                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
624                 nicmp6->icmp6_code = 0;
625                 if (n) {
626                         ICMP6STAT_INC(icp6s_reflect);
627                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
628                         icmp6_reflect(n, noff);
629                 }
630                 break;
631
632         case ICMP6_ECHO_REPLY:
633                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
634                 if (code != 0)
635                         goto badcode;
636                 break;
637
638         case MLD_LISTENER_QUERY:
639         case MLD_LISTENER_REPORT:
640         case MLD_LISTENER_DONE:
641         case MLDV2_LISTENER_REPORT:
642                 /*
643                  * Drop MLD traffic which is not link-local, has a hop limit
644                  * of greater than 1 hop, or which does not have the
645                  * IPv6 HBH Router Alert option.
646                  * As IPv6 HBH options are stripped in ip6_input() we must
647                  * check an mbuf header flag.
648                  * XXX Should we also sanity check that these messages
649                  * were directed to a link-local multicast prefix?
650                  */
651                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
652                         goto freeit;
653                 if (mld_input(m, off, icmp6len) != 0)
654                         return (IPPROTO_DONE);
655                 /* m stays. */
656                 break;
657
658         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
659             {
660                 enum { WRU, FQDN } mode;
661
662                 if (!V_icmp6_nodeinfo)
663                         break;
664
665                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
666                         mode = WRU;
667                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
668                         mode = FQDN;
669                 else
670                         goto badlen;
671
672                 if (mode == FQDN) {
673 #ifndef PULLDOWN_TEST
674                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
675                             IPPROTO_DONE);
676 #endif
677                         n = m_copy(m, 0, M_COPYALL);
678                         if (n)
679                                 n = ni6_input(n, off);
680                         /* XXX meaningless if n == NULL */
681                         noff = sizeof(struct ip6_hdr);
682                 } else {
683                         struct prison *pr;
684                         u_char *p;
685                         int maxhlen, hlen;
686
687                         /*
688                          * XXX: this combination of flags is pointless,
689                          * but should we keep this for compatibility?
690                          */
691                         if ((V_icmp6_nodeinfo & 5) != 5)
692                                 break;
693
694                         if (code != 0)
695                                 goto badcode;
696
697                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
698                         n = m_gethdr(M_NOWAIT, m->m_type);
699                         if (n == NULL) {
700                                 /* Give up remote */
701                                 break;
702                         }
703                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
704                                 /*
705                                  * Previous code did a blind M_COPY_PKTHDR
706                                  * and said "just for rcvif".  If true, then
707                                  * we could tolerate the dup failing (due to
708                                  * the deep copy of the tag chain).  For now
709                                  * be conservative and just fail.
710                                  */
711                                 m_free(n);
712                                 n = NULL;
713                         }
714                         maxhlen = M_TRAILINGSPACE(n) -
715                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
716                         pr = curthread->td_ucred->cr_prison;
717                         mtx_lock(&pr->pr_mtx);
718                         hlen = strlen(pr->pr_hostname);
719                         if (maxhlen > hlen)
720                                 maxhlen = hlen;
721                         /*
722                          * Copy IPv6 and ICMPv6 only.
723                          */
724                         nip6 = mtod(n, struct ip6_hdr *);
725                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
726                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
727                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
728                         p = (u_char *)(nicmp6 + 1);
729                         bzero(p, 4);
730                         /* meaningless TTL */
731                         bcopy(pr->pr_hostname, p + 4, maxhlen);
732                         mtx_unlock(&pr->pr_mtx);
733                         noff = sizeof(struct ip6_hdr);
734                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
735                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
736                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
737                         nicmp6->icmp6_code = 0;
738                 }
739                 if (n) {
740                         ICMP6STAT_INC(icp6s_reflect);
741                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
742                         icmp6_reflect(n, noff);
743                 }
744                 break;
745             }
746
747         case ICMP6_WRUREPLY:
748                 if (code != 0)
749                         goto badcode;
750                 break;
751
752         case ND_ROUTER_SOLICIT:
753                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
754                 if (code != 0)
755                         goto badcode;
756                 if (icmp6len < sizeof(struct nd_router_solicit))
757                         goto badlen;
758                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
759                         /* give up local */
760
761                         /* Send incoming SeND packet to user space. */
762                         if (send_sendso_input_hook != NULL) {
763                                 IP6_EXTHDR_CHECK(m, off,
764                                     icmp6len, IPPROTO_DONE);
765                                 error = send_sendso_input_hook(m, ifp,
766                                     SND_IN, ip6len);
767                                 /* -1 == no app on SEND socket */
768                                 if (error == 0)
769                                         return (IPPROTO_DONE);
770                                 nd6_rs_input(m, off, icmp6len);
771                         } else
772                                 nd6_rs_input(m, off, icmp6len);
773                         m = NULL;
774                         goto freeit;
775                 }
776                 if (send_sendso_input_hook != NULL) {
777                         IP6_EXTHDR_CHECK(n, off,
778                             icmp6len, IPPROTO_DONE);
779                         error = send_sendso_input_hook(n, ifp,
780                             SND_IN, ip6len);
781                         if (error == 0)
782                                 goto freeit;
783                         /* -1 == no app on SEND socket */
784                         nd6_rs_input(n, off, icmp6len);
785                 } else
786                         nd6_rs_input(n, off, icmp6len);
787                 /* m stays. */
788                 break;
789
790         case ND_ROUTER_ADVERT:
791                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
792                 if (code != 0)
793                         goto badcode;
794                 if (icmp6len < sizeof(struct nd_router_advert))
795                         goto badlen;
796                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
797
798                         /* Send incoming SeND-protected/ND packet to user space. */
799                         if (send_sendso_input_hook != NULL) {
800                                 error = send_sendso_input_hook(m, ifp,
801                                     SND_IN, ip6len);
802                                 if (error == 0)
803                                         return (IPPROTO_DONE);
804                                 nd6_ra_input(m, off, icmp6len);
805                         } else
806                                 nd6_ra_input(m, off, icmp6len);
807                         m = NULL;
808                         goto freeit;
809                 }
810                 if (send_sendso_input_hook != NULL) {
811                         error = send_sendso_input_hook(n, ifp,
812                             SND_IN, ip6len);
813                         if (error == 0)
814                                 goto freeit;
815                         nd6_ra_input(n, off, icmp6len);
816                 } else
817                         nd6_ra_input(n, off, icmp6len);
818                 /* m stays. */
819                 break;
820
821         case ND_NEIGHBOR_SOLICIT:
822                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
823                 if (code != 0)
824                         goto badcode;
825                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
826                         goto badlen;
827                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
828                         if (send_sendso_input_hook != NULL) {
829                                 error = send_sendso_input_hook(m, ifp,
830                                     SND_IN, ip6len);
831                                 if (error == 0)
832                                         return (IPPROTO_DONE);
833                                 nd6_ns_input(m, off, icmp6len);
834                         } else
835                                 nd6_ns_input(m, off, icmp6len);
836                         m = NULL;
837                         goto freeit;
838                 }
839                 if (send_sendso_input_hook != NULL) {
840                         error = send_sendso_input_hook(n, ifp,
841                             SND_IN, ip6len);
842                         if (error == 0)
843                                 goto freeit;
844                         nd6_ns_input(n, off, icmp6len);
845                 } else
846                         nd6_ns_input(n, off, icmp6len);
847                 /* m stays. */
848                 break;
849
850         case ND_NEIGHBOR_ADVERT:
851                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
852                 if (code != 0)
853                         goto badcode;
854                 if (icmp6len < sizeof(struct nd_neighbor_advert))
855                         goto badlen;
856                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
857
858                         /* Send incoming SeND-protected/ND packet to user space. */
859                         if (send_sendso_input_hook != NULL) {
860                                 error = send_sendso_input_hook(m, ifp,
861                                     SND_IN, ip6len);
862                                 if (error == 0)
863                                         return (IPPROTO_DONE);
864                                 nd6_na_input(m, off, icmp6len);
865                         } else
866                                 nd6_na_input(m, off, icmp6len);
867                         m = NULL;
868                         goto freeit;
869                 }
870                 if (send_sendso_input_hook != NULL) {
871                         error = send_sendso_input_hook(n, ifp,
872                             SND_IN, ip6len);
873                         if (error == 0)
874                                 goto freeit;
875                         nd6_na_input(n, off, icmp6len);
876                 } else
877                         nd6_na_input(n, off, icmp6len);
878                 /* m stays. */
879                 break;
880
881         case ND_REDIRECT:
882                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
883                 if (code != 0)
884                         goto badcode;
885                 if (icmp6len < sizeof(struct nd_redirect))
886                         goto badlen;
887                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
888                         if (send_sendso_input_hook != NULL) {
889                                 error = send_sendso_input_hook(m, ifp,
890                                     SND_IN, ip6len);
891                                 if (error == 0)
892                                         return (IPPROTO_DONE);
893                             icmp6_redirect_input(m, off);
894                         } else
895                                 icmp6_redirect_input(m, off);
896                         m = NULL;
897                         goto freeit;
898                 }
899                 if (send_sendso_input_hook != NULL) {
900                         error = send_sendso_input_hook(n, ifp,
901                             SND_IN, ip6len);
902                         if (error == 0)
903                                 goto freeit;
904                         icmp6_redirect_input(n, off);
905                 } else
906                         icmp6_redirect_input(n, off);
907                 /* m stays. */
908                 break;
909
910         case ICMP6_ROUTER_RENUMBERING:
911                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
912                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
913                         goto badcode;
914                 if (icmp6len < sizeof(struct icmp6_router_renum))
915                         goto badlen;
916                 break;
917
918         default:
919                 nd6log((LOG_DEBUG,
920                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
921                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
922                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
923                     ifp ? ifp->if_index : 0));
924                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
925                         /* ICMPv6 error: MUST deliver it by spec... */
926                         code = PRC_NCMDS;
927                         /* deliver */
928                 } else {
929                         /* ICMPv6 informational: MUST not deliver */
930                         break;
931                 }
932         deliver:
933                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
934                         /* In this case, m should've been freed. */
935                         return (IPPROTO_DONE);
936                 }
937                 break;
938
939         badcode:
940                 ICMP6STAT_INC(icp6s_badcode);
941                 break;
942
943         badlen:
944                 ICMP6STAT_INC(icp6s_badlen);
945                 break;
946         }
947
948         /* deliver the packet to appropriate sockets */
949         icmp6_rip6_input(&m, *offp);
950
951         return IPPROTO_DONE;
952
953  freeit:
954         m_freem(m);
955         return IPPROTO_DONE;
956 }
957
958 static int
959 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
960 {
961         struct mbuf *m = *mp;
962         struct icmp6_hdr *icmp6;
963         struct ip6_hdr *eip6;
964         u_int32_t notifymtu;
965         struct sockaddr_in6 icmp6src, icmp6dst;
966
967         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
968                 ICMP6STAT_INC(icp6s_tooshort);
969                 goto freeit;
970         }
971 #ifndef PULLDOWN_TEST
972         IP6_EXTHDR_CHECK(m, off,
973             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
974         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
975 #else
976         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
977             sizeof(*icmp6) + sizeof(struct ip6_hdr));
978         if (icmp6 == NULL) {
979                 ICMP6STAT_INC(icp6s_tooshort);
980                 return (-1);
981         }
982 #endif
983         eip6 = (struct ip6_hdr *)(icmp6 + 1);
984
985         /* Detect the upper level protocol */
986         {
987                 void (*ctlfunc)(int, struct sockaddr *, void *);
988                 u_int8_t nxt = eip6->ip6_nxt;
989                 int eoff = off + sizeof(struct icmp6_hdr) +
990                     sizeof(struct ip6_hdr);
991                 struct ip6ctlparam ip6cp;
992                 struct in6_addr *finaldst = NULL;
993                 int icmp6type = icmp6->icmp6_type;
994                 struct ip6_frag *fh;
995                 struct ip6_rthdr *rth;
996                 struct ip6_rthdr0 *rth0;
997                 int rthlen;
998
999                 while (1) { /* XXX: should avoid infinite loop explicitly? */
1000                         struct ip6_ext *eh;
1001
1002                         switch (nxt) {
1003                         case IPPROTO_HOPOPTS:
1004                         case IPPROTO_DSTOPTS:
1005                         case IPPROTO_AH:
1006 #ifndef PULLDOWN_TEST
1007                                 IP6_EXTHDR_CHECK(m, 0,
1008                                     eoff + sizeof(struct ip6_ext), -1);
1009                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1010 #else
1011                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1012                                     eoff, sizeof(*eh));
1013                                 if (eh == NULL) {
1014                                         ICMP6STAT_INC(icp6s_tooshort);
1015                                         return (-1);
1016                                 }
1017 #endif
1018
1019                                 if (nxt == IPPROTO_AH)
1020                                         eoff += (eh->ip6e_len + 2) << 2;
1021                                 else
1022                                         eoff += (eh->ip6e_len + 1) << 3;
1023                                 nxt = eh->ip6e_nxt;
1024                                 break;
1025                         case IPPROTO_ROUTING:
1026                                 /*
1027                                  * When the erroneous packet contains a
1028                                  * routing header, we should examine the
1029                                  * header to determine the final destination.
1030                                  * Otherwise, we can't properly update
1031                                  * information that depends on the final
1032                                  * destination (e.g. path MTU).
1033                                  */
1034 #ifndef PULLDOWN_TEST
1035                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1036                                 rth = (struct ip6_rthdr *)
1037                                     (mtod(m, caddr_t) + eoff);
1038 #else
1039                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1040                                     eoff, sizeof(*rth));
1041                                 if (rth == NULL) {
1042                                         ICMP6STAT_INC(icp6s_tooshort);
1043                                         return (-1);
1044                                 }
1045 #endif
1046                                 rthlen = (rth->ip6r_len + 1) << 3;
1047                                 /*
1048                                  * XXX: currently there is no
1049                                  * officially defined type other
1050                                  * than type-0.
1051                                  * Note that if the segment left field
1052                                  * is 0, all intermediate hops must
1053                                  * have been passed.
1054                                  */
1055                                 if (rth->ip6r_segleft &&
1056                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1057                                         int hops;
1058
1059 #ifndef PULLDOWN_TEST
1060                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1061                                         rth0 = (struct ip6_rthdr0 *)
1062                                             (mtod(m, caddr_t) + eoff);
1063 #else
1064                                         IP6_EXTHDR_GET(rth0,
1065                                             struct ip6_rthdr0 *, m,
1066                                             eoff, rthlen);
1067                                         if (rth0 == NULL) {
1068                                                 ICMP6STAT_INC(icp6s_tooshort);
1069                                                 return (-1);
1070                                         }
1071 #endif
1072                                         /* just ignore a bogus header */
1073                                         if ((rth0->ip6r0_len % 2) == 0 &&
1074                                             (hops = rth0->ip6r0_len/2))
1075                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1076                                 }
1077                                 eoff += rthlen;
1078                                 nxt = rth->ip6r_nxt;
1079                                 break;
1080                         case IPPROTO_FRAGMENT:
1081 #ifndef PULLDOWN_TEST
1082                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1083                                     sizeof(struct ip6_frag), -1);
1084                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1085                                     eoff);
1086 #else
1087                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1088                                     eoff, sizeof(*fh));
1089                                 if (fh == NULL) {
1090                                         ICMP6STAT_INC(icp6s_tooshort);
1091                                         return (-1);
1092                                 }
1093 #endif
1094                                 /*
1095                                  * Data after a fragment header is meaningless
1096                                  * unless it is the first fragment, but
1097                                  * we'll go to the notify label for path MTU
1098                                  * discovery.
1099                                  */
1100                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1101                                         goto notify;
1102
1103                                 eoff += sizeof(struct ip6_frag);
1104                                 nxt = fh->ip6f_nxt;
1105                                 break;
1106                         default:
1107                                 /*
1108                                  * This case includes ESP and the No Next
1109                                  * Header.  In such cases going to the notify
1110                                  * label does not have any meaning
1111                                  * (i.e. ctlfunc will be NULL), but we go
1112                                  * anyway since we might have to update
1113                                  * path MTU information.
1114                                  */
1115                                 goto notify;
1116                         }
1117                 }
1118           notify:
1119 #ifndef PULLDOWN_TEST
1120                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1121 #else
1122                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1123                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1124                 if (icmp6 == NULL) {
1125                         ICMP6STAT_INC(icp6s_tooshort);
1126                         return (-1);
1127                 }
1128 #endif
1129
1130                 /*
1131                  * retrieve parameters from the inner IPv6 header, and convert
1132                  * them into sockaddr structures.
1133                  * XXX: there is no guarantee that the source or destination
1134                  * addresses of the inner packet are in the same scope as
1135                  * the addresses of the icmp packet.  But there is no other
1136                  * way to determine the zone.
1137                  */
1138                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1139
1140                 bzero(&icmp6dst, sizeof(icmp6dst));
1141                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1142                 icmp6dst.sin6_family = AF_INET6;
1143                 if (finaldst == NULL)
1144                         icmp6dst.sin6_addr = eip6->ip6_dst;
1145                 else
1146                         icmp6dst.sin6_addr = *finaldst;
1147                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1148                         goto freeit;
1149                 bzero(&icmp6src, sizeof(icmp6src));
1150                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1151                 icmp6src.sin6_family = AF_INET6;
1152                 icmp6src.sin6_addr = eip6->ip6_src;
1153                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1154                         goto freeit;
1155                 icmp6src.sin6_flowinfo =
1156                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1157
1158                 if (finaldst == NULL)
1159                         finaldst = &eip6->ip6_dst;
1160                 ip6cp.ip6c_m = m;
1161                 ip6cp.ip6c_icmp6 = icmp6;
1162                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1163                 ip6cp.ip6c_off = eoff;
1164                 ip6cp.ip6c_finaldst = finaldst;
1165                 ip6cp.ip6c_src = &icmp6src;
1166                 ip6cp.ip6c_nxt = nxt;
1167
1168                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1169                         notifymtu = ntohl(icmp6->icmp6_mtu);
1170                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1171                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1172                 }
1173
1174                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1175                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1176                 if (ctlfunc) {
1177                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1178                             &ip6cp);
1179                 }
1180         }
1181         *mp = m;
1182         return (0);
1183
1184   freeit:
1185         m_freem(m);
1186         return (-1);
1187 }
1188
1189 void
1190 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1191 {
1192         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1193         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1194         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1195         u_int mtu = ntohl(icmp6->icmp6_mtu);
1196         struct in_conninfo inc;
1197
1198 #if 0
1199         /*
1200          * RFC2460 section 5, last paragraph.
1201          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1202          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1203          * due to packet translator in the middle.
1204          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1205          * special handling.
1206          */
1207         if (mtu < IPV6_MMTU)
1208                 return;
1209 #endif
1210
1211         /*
1212          * we reject ICMPv6 too big with abnormally small value.
1213          * XXX what is the good definition of "abnormally small"?
1214          */
1215         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1216                 return;
1217
1218         if (!validated)
1219                 return;
1220
1221         /*
1222          * In case the suggested mtu is less than IPV6_MMTU, we
1223          * only need to remember that it was for above mentioned
1224          * "alwaysfrag" case.
1225          * Try to be as close to the spec as possible.
1226          */
1227         if (mtu < IPV6_MMTU)
1228                 mtu = IPV6_MMTU - 8;
1229
1230         bzero(&inc, sizeof(inc));
1231         inc.inc_fibnum = M_GETFIB(m);
1232         inc.inc_flags |= INC_ISIPV6;
1233         inc.inc6_faddr = *dst;
1234         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1235                 return;
1236
1237         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1238                 tcp_hc_updatemtu(&inc, mtu);
1239                 ICMP6STAT_INC(icp6s_pmtuchg);
1240         }
1241 }
1242
1243 /*
1244  * Process a Node Information Query packet, based on
1245  * draft-ietf-ipngwg-icmp-name-lookups-07.
1246  *
1247  * Spec incompatibilities:
1248  * - IPv6 Subject address handling
1249  * - IPv4 Subject address handling support missing
1250  * - Proxy reply (answer even if it's not for me)
1251  * - joins NI group address at in6_ifattach() time only, does not cope
1252  *   with hostname changes by sethostname(3)
1253  */
1254 static struct mbuf *
1255 ni6_input(struct mbuf *m, int off)
1256 {
1257         struct icmp6_nodeinfo *ni6, *nni6;
1258         struct mbuf *n = NULL;
1259         struct prison *pr;
1260         u_int16_t qtype;
1261         int subjlen;
1262         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1263         struct ni_reply_fqdn *fqdn;
1264         int addrs;              /* for NI_QTYPE_NODEADDR */
1265         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1266         struct in6_addr in6_subj; /* subject address */
1267         struct ip6_hdr *ip6;
1268         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1269         char *subj = NULL;
1270         struct in6_ifaddr *ia6 = NULL;
1271
1272         ip6 = mtod(m, struct ip6_hdr *);
1273 #ifndef PULLDOWN_TEST
1274         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1275 #else
1276         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1277         if (ni6 == NULL) {
1278                 /* m is already reclaimed */
1279                 return (NULL);
1280         }
1281 #endif
1282
1283         /*
1284          * Validate IPv6 source address.
1285          * The default configuration MUST be to refuse answering queries from
1286          * global-scope addresses according to RFC4602.
1287          * Notes:
1288          *  - it's not very clear what "refuse" means; this implementation
1289          *    simply drops it.
1290          *  - it's not very easy to identify global-scope (unicast) addresses
1291          *    since there are many prefixes for them.  It should be safer
1292          *    and in practice sufficient to check "all" but loopback and
1293          *    link-local (note that site-local unicast was deprecated and
1294          *    ULA is defined as global scope-wise)
1295          */
1296         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1297             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1298             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1299                 goto bad;
1300
1301         /*
1302          * Validate IPv6 destination address.
1303          *
1304          * The Responder must discard the Query without further processing
1305          * unless it is one of the Responder's unicast or anycast addresses, or
1306          * a link-local scope multicast address which the Responder has joined.
1307          * [RFC4602, Section 5.]
1308          */
1309         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1310                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1311                         goto bad;
1312                 /* else it's a link-local multicast, fine */
1313         } else {                /* unicast or anycast */
1314                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1315                         goto bad; /* XXX impossible */
1316
1317                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1318                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1319                         ifa_free(&ia6->ia_ifa);
1320                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1321                                 "a temporary address in %s:%d",
1322                                __FILE__, __LINE__));
1323                         goto bad;
1324                 }
1325                 ifa_free(&ia6->ia_ifa);
1326         }
1327
1328         /* validate query Subject field. */
1329         qtype = ntohs(ni6->ni_qtype);
1330         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1331         switch (qtype) {
1332         case NI_QTYPE_NOOP:
1333         case NI_QTYPE_SUPTYPES:
1334                 /* 07 draft */
1335                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1336                         break;
1337                 /* FALLTHROUGH */
1338         case NI_QTYPE_FQDN:
1339         case NI_QTYPE_NODEADDR:
1340         case NI_QTYPE_IPV4ADDR:
1341                 switch (ni6->ni_code) {
1342                 case ICMP6_NI_SUBJ_IPV6:
1343 #if ICMP6_NI_SUBJ_IPV6 != 0
1344                 case 0:
1345 #endif
1346                         /*
1347                          * backward compatibility - try to accept 03 draft
1348                          * format, where no Subject is present.
1349                          */
1350                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1351                             subjlen == 0) {
1352                                 oldfqdn++;
1353                                 break;
1354                         }
1355 #if ICMP6_NI_SUBJ_IPV6 != 0
1356                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1357                                 goto bad;
1358 #endif
1359
1360                         if (subjlen != sizeof(struct in6_addr))
1361                                 goto bad;
1362
1363                         /*
1364                          * Validate Subject address.
1365                          *
1366                          * Not sure what exactly "address belongs to the node"
1367                          * means in the spec, is it just unicast, or what?
1368                          *
1369                          * At this moment we consider Subject address as
1370                          * "belong to the node" if the Subject address equals
1371                          * to the IPv6 destination address; validation for
1372                          * IPv6 destination address should have done enough
1373                          * check for us.
1374                          *
1375                          * We do not do proxy at this moment.
1376                          */
1377                         /* m_pulldown instead of copy? */
1378                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1379                             subjlen, (caddr_t)&in6_subj);
1380                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1381                                 goto bad;
1382
1383                         subj = (char *)&in6_subj;
1384                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1385                                 break;
1386
1387                         /*
1388                          * XXX if we are to allow other cases, we should really
1389                          * be careful about scope here.
1390                          * basically, we should disallow queries toward IPv6
1391                          * destination X with subject Y,
1392                          * if scope(X) > scope(Y).
1393                          * if we allow scope(X) > scope(Y), it will result in
1394                          * information leakage across scope boundary.
1395                          */
1396                         goto bad;
1397
1398                 case ICMP6_NI_SUBJ_FQDN:
1399                         /*
1400                          * Validate Subject name with gethostname(3).
1401                          *
1402                          * The behavior may need some debate, since:
1403                          * - we are not sure if the node has FQDN as
1404                          *   hostname (returned by gethostname(3)).
1405                          * - the code does wildcard match for truncated names.
1406                          *   however, we are not sure if we want to perform
1407                          *   wildcard match, if gethostname(3) side has
1408                          *   truncated hostname.
1409                          */
1410                         pr = curthread->td_ucred->cr_prison;
1411                         mtx_lock(&pr->pr_mtx);
1412                         n = ni6_nametodns(pr->pr_hostname,
1413                             strlen(pr->pr_hostname), 0);
1414                         mtx_unlock(&pr->pr_mtx);
1415                         if (!n || n->m_next || n->m_len == 0)
1416                                 goto bad;
1417                         IP6_EXTHDR_GET(subj, char *, m,
1418                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1419                         if (subj == NULL)
1420                                 goto bad;
1421                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1422                             n->m_len)) {
1423                                 goto bad;
1424                         }
1425                         m_freem(n);
1426                         n = NULL;
1427                         break;
1428
1429                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1430                 default:
1431                         goto bad;
1432                 }
1433                 break;
1434         }
1435
1436         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1437         switch (qtype) {
1438         case NI_QTYPE_FQDN:
1439                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1440                         goto bad;
1441                 break;
1442         case NI_QTYPE_NODEADDR:
1443         case NI_QTYPE_IPV4ADDR:
1444                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1445                         goto bad;
1446                 break;
1447         }
1448
1449         /* guess reply length */
1450         switch (qtype) {
1451         case NI_QTYPE_NOOP:
1452                 break;          /* no reply data */
1453         case NI_QTYPE_SUPTYPES:
1454                 replylen += sizeof(u_int32_t);
1455                 break;
1456         case NI_QTYPE_FQDN:
1457                 /* XXX will append an mbuf */
1458                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1459                 break;
1460         case NI_QTYPE_NODEADDR:
1461                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1462                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1463                     sizeof(u_int32_t))) > MCLBYTES)
1464                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1465                 break;
1466         case NI_QTYPE_IPV4ADDR:
1467                 /* unsupported - should respond with unknown Qtype? */
1468                 break;
1469         default:
1470                 /*
1471                  * XXX: We must return a reply with the ICMP6 code
1472                  * `unknown Qtype' in this case.  However we regard the case
1473                  * as an FQDN query for backward compatibility.
1474                  * Older versions set a random value to this field,
1475                  * so it rarely varies in the defined qtypes.
1476                  * But the mechanism is not reliable...
1477                  * maybe we should obsolete older versions.
1478                  */
1479                 qtype = NI_QTYPE_FQDN;
1480                 /* XXX will append an mbuf */
1481                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1482                 oldfqdn++;
1483                 break;
1484         }
1485
1486         /* Allocate an mbuf to reply. */
1487         if (replylen > MCLBYTES) {
1488                 /*
1489                  * XXX: should we try to allocate more? But MCLBYTES
1490                  * is probably much larger than IPV6_MMTU...
1491                  */
1492                 goto bad;
1493         }
1494         if (replylen > MHLEN)
1495                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1496         else
1497                 n = m_gethdr(M_NOWAIT, m->m_type);
1498         if (n == NULL) {
1499                 m_freem(m);
1500                 return (NULL);
1501         }
1502         m_move_pkthdr(n, m); /* just for recvif and FIB */
1503         n->m_pkthdr.len = n->m_len = replylen;
1504
1505         /* copy mbuf header and IPv6 + Node Information base headers */
1506         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1507         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1508         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1509
1510         /* qtype dependent procedure */
1511         switch (qtype) {
1512         case NI_QTYPE_NOOP:
1513                 nni6->ni_code = ICMP6_NI_SUCCESS;
1514                 nni6->ni_flags = 0;
1515                 break;
1516         case NI_QTYPE_SUPTYPES:
1517         {
1518                 u_int32_t v;
1519                 nni6->ni_code = ICMP6_NI_SUCCESS;
1520                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1521                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1522                 v = (u_int32_t)htonl(0x0000000f);
1523                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1524                 break;
1525         }
1526         case NI_QTYPE_FQDN:
1527                 nni6->ni_code = ICMP6_NI_SUCCESS;
1528                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1529                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1530                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1531                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1532                 /*
1533                  * XXX do we really have FQDN in hostname?
1534                  */
1535                 pr = curthread->td_ucred->cr_prison;
1536                 mtx_lock(&pr->pr_mtx);
1537                 n->m_next = ni6_nametodns(pr->pr_hostname,
1538                     strlen(pr->pr_hostname), oldfqdn);
1539                 mtx_unlock(&pr->pr_mtx);
1540                 if (n->m_next == NULL)
1541                         goto bad;
1542                 /* XXX we assume that n->m_next is not a chain */
1543                 if (n->m_next->m_next != NULL)
1544                         goto bad;
1545                 n->m_pkthdr.len += n->m_next->m_len;
1546                 break;
1547         case NI_QTYPE_NODEADDR:
1548         {
1549                 int lenlim, copied;
1550
1551                 nni6->ni_code = ICMP6_NI_SUCCESS;
1552                 n->m_pkthdr.len = n->m_len =
1553                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1554                 lenlim = M_TRAILINGSPACE(n);
1555                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1556                 /* XXX: reset mbuf length */
1557                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1558                     sizeof(struct icmp6_nodeinfo) + copied;
1559                 break;
1560         }
1561         default:
1562                 break;          /* XXX impossible! */
1563         }
1564
1565         nni6->ni_type = ICMP6_NI_REPLY;
1566         m_freem(m);
1567         return (n);
1568
1569   bad:
1570         m_freem(m);
1571         if (n)
1572                 m_freem(n);
1573         return (NULL);
1574 }
1575
1576 /*
1577  * make a mbuf with DNS-encoded string.  no compression support.
1578  *
1579  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1580  * treated as truncated name (two \0 at the end).  this is a wild guess.
1581  *
1582  * old - return pascal string if non-zero
1583  */
1584 static struct mbuf *
1585 ni6_nametodns(const char *name, int namelen, int old)
1586 {
1587         struct mbuf *m;
1588         char *cp, *ep;
1589         const char *p, *q;
1590         int i, len, nterm;
1591
1592         if (old)
1593                 len = namelen + 1;
1594         else
1595                 len = MCLBYTES;
1596
1597         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1598         if (len > MLEN)
1599                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1600         else
1601                 m = m_get(M_NOWAIT, MT_DATA);
1602         if (m == NULL)
1603                 goto fail;
1604
1605         if (old) {
1606                 m->m_len = len;
1607                 *mtod(m, char *) = namelen;
1608                 bcopy(name, mtod(m, char *) + 1, namelen);
1609                 return m;
1610         } else {
1611                 m->m_len = 0;
1612                 cp = mtod(m, char *);
1613                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1614
1615                 /* if not certain about my name, return empty buffer */
1616                 if (namelen == 0)
1617                         return m;
1618
1619                 /*
1620                  * guess if it looks like shortened hostname, or FQDN.
1621                  * shortened hostname needs two trailing "\0".
1622                  */
1623                 i = 0;
1624                 for (p = name; p < name + namelen; p++) {
1625                         if (*p && *p == '.')
1626                                 i++;
1627                 }
1628                 if (i < 2)
1629                         nterm = 2;
1630                 else
1631                         nterm = 1;
1632
1633                 p = name;
1634                 while (cp < ep && p < name + namelen) {
1635                         i = 0;
1636                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1637                                 i++;
1638                         /* result does not fit into mbuf */
1639                         if (cp + i + 1 >= ep)
1640                                 goto fail;
1641                         /*
1642                          * DNS label length restriction, RFC1035 page 8.
1643                          * "i == 0" case is included here to avoid returning
1644                          * 0-length label on "foo..bar".
1645                          */
1646                         if (i <= 0 || i >= 64)
1647                                 goto fail;
1648                         *cp++ = i;
1649                         bcopy(p, cp, i);
1650                         cp += i;
1651                         p = q;
1652                         if (p < name + namelen && *p == '.')
1653                                 p++;
1654                 }
1655                 /* termination */
1656                 if (cp + nterm >= ep)
1657                         goto fail;
1658                 while (nterm-- > 0)
1659                         *cp++ = '\0';
1660                 m->m_len = cp - mtod(m, char *);
1661                 return m;
1662         }
1663
1664         panic("should not reach here");
1665         /* NOTREACHED */
1666
1667  fail:
1668         if (m)
1669                 m_freem(m);
1670         return NULL;
1671 }
1672
1673 /*
1674  * check if two DNS-encoded string matches.  takes care of truncated
1675  * form (with \0\0 at the end).  no compression support.
1676  * XXX upper/lowercase match (see RFC2065)
1677  */
1678 static int
1679 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1680 {
1681         const char *a0, *b0;
1682         int l;
1683
1684         /* simplest case - need validation? */
1685         if (alen == blen && bcmp(a, b, alen) == 0)
1686                 return 1;
1687
1688         a0 = a;
1689         b0 = b;
1690
1691         /* termination is mandatory */
1692         if (alen < 2 || blen < 2)
1693                 return 0;
1694         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1695                 return 0;
1696         alen--;
1697         blen--;
1698
1699         while (a - a0 < alen && b - b0 < blen) {
1700                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1701                         return 0;
1702
1703                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1704                         return 0;
1705                 /* we don't support compression yet */
1706                 if (a[0] >= 64 || b[0] >= 64)
1707                         return 0;
1708
1709                 /* truncated case */
1710                 if (a[0] == 0 && a - a0 == alen - 1)
1711                         return 1;
1712                 if (b[0] == 0 && b - b0 == blen - 1)
1713                         return 1;
1714                 if (a[0] == 0 || b[0] == 0)
1715                         return 0;
1716
1717                 if (a[0] != b[0])
1718                         return 0;
1719                 l = a[0];
1720                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1721                         return 0;
1722                 if (bcmp(a + 1, b + 1, l) != 0)
1723                         return 0;
1724
1725                 a += 1 + l;
1726                 b += 1 + l;
1727         }
1728
1729         if (a - a0 == alen && b - b0 == blen)
1730                 return 1;
1731         else
1732                 return 0;
1733 }
1734
1735 /*
1736  * calculate the number of addresses to be returned in the node info reply.
1737  */
1738 static int
1739 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1740     struct in6_addr *subj)
1741 {
1742         struct ifnet *ifp;
1743         struct in6_ifaddr *ifa6;
1744         struct ifaddr *ifa;
1745         int addrs = 0, addrsofif, iffound = 0;
1746         int niflags = ni6->ni_flags;
1747
1748         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1749                 switch (ni6->ni_code) {
1750                 case ICMP6_NI_SUBJ_IPV6:
1751                         if (subj == NULL) /* must be impossible... */
1752                                 return (0);
1753                         break;
1754                 default:
1755                         /*
1756                          * XXX: we only support IPv6 subject address for
1757                          * this Qtype.
1758                          */
1759                         return (0);
1760                 }
1761         }
1762
1763         IFNET_RLOCK_NOSLEEP();
1764         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1765                 addrsofif = 0;
1766                 IF_ADDR_RLOCK(ifp);
1767                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1768                         if (ifa->ifa_addr->sa_family != AF_INET6)
1769                                 continue;
1770                         ifa6 = (struct in6_ifaddr *)ifa;
1771
1772                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1773                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1774                                 iffound = 1;
1775
1776                         /*
1777                          * IPv4-mapped addresses can only be returned by a
1778                          * Node Information proxy, since they represent
1779                          * addresses of IPv4-only nodes, which perforce do
1780                          * not implement this protocol.
1781                          * [icmp-name-lookups-07, Section 5.4]
1782                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1783                          * this function at this moment.
1784                          */
1785
1786                         /* What do we have to do about ::1? */
1787                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1788                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1789                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1790                                         continue;
1791                                 break;
1792                         case IPV6_ADDR_SCOPE_SITELOCAL:
1793                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1794                                         continue;
1795                                 break;
1796                         case IPV6_ADDR_SCOPE_GLOBAL:
1797                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1798                                         continue;
1799                                 break;
1800                         default:
1801                                 continue;
1802                         }
1803
1804                         /*
1805                          * check if anycast is okay.
1806                          * XXX: just experimental.  not in the spec.
1807                          */
1808                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1809                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1810                                 continue; /* we need only unicast addresses */
1811                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1812                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1813                                 continue;
1814                         }
1815                         addrsofif++; /* count the address */
1816                 }
1817                 IF_ADDR_RUNLOCK(ifp);
1818                 if (iffound) {
1819                         *ifpp = ifp;
1820                         IFNET_RUNLOCK_NOSLEEP();
1821                         return (addrsofif);
1822                 }
1823
1824                 addrs += addrsofif;
1825         }
1826         IFNET_RUNLOCK_NOSLEEP();
1827
1828         return (addrs);
1829 }
1830
1831 static int
1832 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1833     struct ifnet *ifp0, int resid)
1834 {
1835         struct ifnet *ifp;
1836         struct in6_ifaddr *ifa6;
1837         struct ifaddr *ifa;
1838         struct ifnet *ifp_dep = NULL;
1839         int copied = 0, allow_deprecated = 0;
1840         u_char *cp = (u_char *)(nni6 + 1);
1841         int niflags = ni6->ni_flags;
1842         u_int32_t ltime;
1843
1844         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1845                 return (0);     /* needless to copy */
1846
1847         IFNET_RLOCK_NOSLEEP();
1848         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1849   again:
1850
1851         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1852                 IF_ADDR_RLOCK(ifp);
1853                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1854                         if (ifa->ifa_addr->sa_family != AF_INET6)
1855                                 continue;
1856                         ifa6 = (struct in6_ifaddr *)ifa;
1857
1858                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1859                             allow_deprecated == 0) {
1860                                 /*
1861                                  * prefererred address should be put before
1862                                  * deprecated addresses.
1863                                  */
1864
1865                                 /* record the interface for later search */
1866                                 if (ifp_dep == NULL)
1867                                         ifp_dep = ifp;
1868
1869                                 continue;
1870                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1871                             allow_deprecated != 0)
1872                                 continue; /* we now collect deprecated addrs */
1873
1874                         /* What do we have to do about ::1? */
1875                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1876                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1877                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1878                                         continue;
1879                                 break;
1880                         case IPV6_ADDR_SCOPE_SITELOCAL:
1881                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1882                                         continue;
1883                                 break;
1884                         case IPV6_ADDR_SCOPE_GLOBAL:
1885                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1886                                         continue;
1887                                 break;
1888                         default:
1889                                 continue;
1890                         }
1891
1892                         /*
1893                          * check if anycast is okay.
1894                          * XXX: just experimental.  not in the spec.
1895                          */
1896                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1897                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1898                                 continue;
1899                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1900                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1901                                 continue;
1902                         }
1903
1904                         /* now we can copy the address */
1905                         if (resid < sizeof(struct in6_addr) +
1906                             sizeof(u_int32_t)) {
1907                                 IF_ADDR_RUNLOCK(ifp);
1908                                 /*
1909                                  * We give up much more copy.
1910                                  * Set the truncate flag and return.
1911                                  */
1912                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1913                                 IFNET_RUNLOCK_NOSLEEP();
1914                                 return (copied);
1915                         }
1916
1917                         /*
1918                          * Set the TTL of the address.
1919                          * The TTL value should be one of the following
1920                          * according to the specification:
1921                          *
1922                          * 1. The remaining lifetime of a DHCP lease on the
1923                          *    address, or
1924                          * 2. The remaining Valid Lifetime of a prefix from
1925                          *    which the address was derived through Stateless
1926                          *    Autoconfiguration.
1927                          *
1928                          * Note that we currently do not support stateful
1929                          * address configuration by DHCPv6, so the former
1930                          * case can't happen.
1931                          */
1932                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1933                                 ltime = ND6_INFINITE_LIFETIME;
1934                         else {
1935                                 if (ifa6->ia6_lifetime.ia6t_expire >
1936                                     time_uptime)
1937                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1938                                 else
1939                                         ltime = 0;
1940                         }
1941
1942                         bcopy(&ltime, cp, sizeof(u_int32_t));
1943                         cp += sizeof(u_int32_t);
1944
1945                         /* copy the address itself */
1946                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1947                             sizeof(struct in6_addr));
1948                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1949                         cp += sizeof(struct in6_addr);
1950
1951                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1952                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1953                 }
1954                 IF_ADDR_RUNLOCK(ifp);
1955                 if (ifp0)       /* we need search only on the specified IF */
1956                         break;
1957         }
1958
1959         if (allow_deprecated == 0 && ifp_dep != NULL) {
1960                 ifp = ifp_dep;
1961                 allow_deprecated = 1;
1962
1963                 goto again;
1964         }
1965
1966         IFNET_RUNLOCK_NOSLEEP();
1967
1968         return (copied);
1969 }
1970
1971 /*
1972  * XXX almost dup'ed code with rip6_input.
1973  */
1974 static int
1975 icmp6_rip6_input(struct mbuf **mp, int off)
1976 {
1977         struct mbuf *m = *mp;
1978         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1979         struct inpcb *in6p;
1980         struct inpcb *last = NULL;
1981         struct sockaddr_in6 fromsa;
1982         struct icmp6_hdr *icmp6;
1983         struct mbuf *opts = NULL;
1984
1985 #ifndef PULLDOWN_TEST
1986         /* this is assumed to be safe. */
1987         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1988 #else
1989         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1990         if (icmp6 == NULL) {
1991                 /* m is already reclaimed */
1992                 return (IPPROTO_DONE);
1993         }
1994 #endif
1995
1996         /*
1997          * XXX: the address may have embedded scope zone ID, which should be
1998          * hidden from applications.
1999          */
2000         bzero(&fromsa, sizeof(fromsa));
2001         fromsa.sin6_family = AF_INET6;
2002         fromsa.sin6_len = sizeof(struct sockaddr_in6);
2003         fromsa.sin6_addr = ip6->ip6_src;
2004         if (sa6_recoverscope(&fromsa)) {
2005                 m_freem(m);
2006                 return (IPPROTO_DONE);
2007         }
2008
2009         INP_INFO_RLOCK(&V_ripcbinfo);
2010         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2011                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2012                         continue;
2013                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2014                         continue;
2015                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2016                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2017                         continue;
2018                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2019                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2020                         continue;
2021                 INP_RLOCK(in6p);
2022                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2023                     in6p->in6p_icmp6filt)) {
2024                         INP_RUNLOCK(in6p);
2025                         continue;
2026                 }
2027                 if (last != NULL) {
2028                         struct  mbuf *n = NULL;
2029
2030                         /*
2031                          * Recent network drivers tend to allocate a single
2032                          * mbuf cluster, rather than to make a couple of
2033                          * mbufs without clusters.  Also, since the IPv6 code
2034                          * path tries to avoid m_pullup(), it is highly
2035                          * probable that we still have an mbuf cluster here
2036                          * even though the necessary length can be stored in an
2037                          * mbuf's internal buffer.
2038                          * Meanwhile, the default size of the receive socket
2039                          * buffer for raw sockets is not so large.  This means
2040                          * the possibility of packet loss is relatively higher
2041                          * than before.  To avoid this scenario, we copy the
2042                          * received data to a separate mbuf that does not use
2043                          * a cluster, if possible.
2044                          * XXX: it is better to copy the data after stripping
2045                          * intermediate headers.
2046                          */
2047                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2048                             m->m_len <= MHLEN) {
2049                                 n = m_get(M_NOWAIT, m->m_type);
2050                                 if (n != NULL) {
2051                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2052                                                 bcopy(m->m_data, n->m_data,
2053                                                       m->m_len);
2054                                                 n->m_len = m->m_len;
2055                                         } else {
2056                                                 m_free(n);
2057                                                 n = NULL;
2058                                         }
2059                                 }
2060                         }
2061                         if (n != NULL ||
2062                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2063                                 if (last->inp_flags & INP_CONTROLOPTS)
2064                                         ip6_savecontrol(last, n, &opts);
2065                                 /* strip intermediate headers */
2066                                 m_adj(n, off);
2067                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2068                                 if (sbappendaddr_locked(
2069                                     &last->inp_socket->so_rcv,
2070                                     (struct sockaddr *)&fromsa, n, opts)
2071                                     == 0) {
2072                                         /* should notify about lost packet */
2073                                         m_freem(n);
2074                                         if (opts) {
2075                                                 m_freem(opts);
2076                                         }
2077                                         SOCKBUF_UNLOCK(
2078                                             &last->inp_socket->so_rcv);
2079                                 } else
2080                                         sorwakeup_locked(last->inp_socket);
2081                                 opts = NULL;
2082                         }
2083                         INP_RUNLOCK(last);
2084                 }
2085                 last = in6p;
2086         }
2087         INP_INFO_RUNLOCK(&V_ripcbinfo);
2088         if (last != NULL) {
2089                 if (last->inp_flags & INP_CONTROLOPTS)
2090                         ip6_savecontrol(last, m, &opts);
2091                 /* strip intermediate headers */
2092                 m_adj(m, off);
2093
2094                 /* avoid using mbuf clusters if possible (see above) */
2095                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2096                     m->m_len <= MHLEN) {
2097                         struct mbuf *n;
2098
2099                         n = m_get(M_NOWAIT, m->m_type);
2100                         if (n != NULL) {
2101                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2102                                         bcopy(m->m_data, n->m_data, m->m_len);
2103                                         n->m_len = m->m_len;
2104
2105                                         m_freem(m);
2106                                         m = n;
2107                                 } else {
2108                                         m_freem(n);
2109                                         n = NULL;
2110                                 }
2111                         }
2112                 }
2113                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2114                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2115                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2116                         m_freem(m);
2117                         if (opts)
2118                                 m_freem(opts);
2119                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2120                 } else
2121                         sorwakeup_locked(last->inp_socket);
2122                 INP_RUNLOCK(last);
2123         } else {
2124                 m_freem(m);
2125                 IP6STAT_DEC(ip6s_delivered);
2126         }
2127         return IPPROTO_DONE;
2128 }
2129
2130 /*
2131  * Reflect the ip6 packet back to the source.
2132  * OFF points to the icmp6 header, counted from the top of the mbuf.
2133  */
2134 void
2135 icmp6_reflect(struct mbuf *m, size_t off)
2136 {
2137         struct ip6_hdr *ip6;
2138         struct icmp6_hdr *icmp6;
2139         struct in6_ifaddr *ia = NULL;
2140         int plen;
2141         int type, code;
2142         struct ifnet *outif = NULL;
2143         struct in6_addr origdst, src, *srcp = NULL;
2144
2145         /* too short to reflect */
2146         if (off < sizeof(struct ip6_hdr)) {
2147                 nd6log((LOG_DEBUG,
2148                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2149                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2150                     __FILE__, __LINE__));
2151                 goto bad;
2152         }
2153
2154         /*
2155          * If there are extra headers between IPv6 and ICMPv6, strip
2156          * off that header first.
2157          */
2158 #ifdef DIAGNOSTIC
2159         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2160                 panic("assumption failed in icmp6_reflect");
2161 #endif
2162         if (off > sizeof(struct ip6_hdr)) {
2163                 size_t l;
2164                 struct ip6_hdr nip6;
2165
2166                 l = off - sizeof(struct ip6_hdr);
2167                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2168                 m_adj(m, l);
2169                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2170                 if (m->m_len < l) {
2171                         if ((m = m_pullup(m, l)) == NULL)
2172                                 return;
2173                 }
2174                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2175         } else /* off == sizeof(struct ip6_hdr) */ {
2176                 size_t l;
2177                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2178                 if (m->m_len < l) {
2179                         if ((m = m_pullup(m, l)) == NULL)
2180                                 return;
2181                 }
2182         }
2183         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2184         ip6 = mtod(m, struct ip6_hdr *);
2185         ip6->ip6_nxt = IPPROTO_ICMPV6;
2186         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2187         type = icmp6->icmp6_type; /* keep type for statistics */
2188         code = icmp6->icmp6_code; /* ditto. */
2189
2190         origdst = ip6->ip6_dst;
2191         /*
2192          * ip6_input() drops a packet if its src is multicast.
2193          * So, the src is never multicast.
2194          */
2195         ip6->ip6_dst = ip6->ip6_src;
2196
2197         /*
2198          * If the incoming packet was addressed directly to us (i.e. unicast),
2199          * use dst as the src for the reply.
2200          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2201          * (for example) when we encounter an error while forwarding procedure
2202          * destined to a duplicated address of ours.
2203          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2204          * procedure of an outgoing packet of our own, in which case we need
2205          * to search in the ifaddr list.
2206          */
2207         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2208                 if ((ia = ip6_getdstifaddr(m))) {
2209                         if (!(ia->ia6_flags &
2210                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2211                                 srcp = &ia->ia_addr.sin6_addr;
2212                 } else {
2213                         struct sockaddr_in6 d;
2214
2215                         bzero(&d, sizeof(d));
2216                         d.sin6_family = AF_INET6;
2217                         d.sin6_len = sizeof(d);
2218                         d.sin6_addr = origdst;
2219                         ia = (struct in6_ifaddr *)
2220                             ifa_ifwithaddr((struct sockaddr *)&d);
2221                         if (ia &&
2222                             !(ia->ia6_flags &
2223                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2224                                 srcp = &ia->ia_addr.sin6_addr;
2225                         }
2226                 }
2227         }
2228
2229         if (srcp == NULL) {
2230                 int e;
2231                 struct sockaddr_in6 sin6;
2232                 struct route_in6 ro;
2233
2234                 /*
2235                  * This case matches to multicasts, our anycast, or unicasts
2236                  * that we do not own.  Select a source address based on the
2237                  * source address of the erroneous packet.
2238                  */
2239                 bzero(&sin6, sizeof(sin6));
2240                 sin6.sin6_family = AF_INET6;
2241                 sin6.sin6_len = sizeof(sin6);
2242                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2243
2244                 bzero(&ro, sizeof(ro));
2245                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2246                 if (ro.ro_rt)
2247                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2248                 if (e) {
2249                         char ip6buf[INET6_ADDRSTRLEN];
2250                         nd6log((LOG_DEBUG,
2251                             "icmp6_reflect: source can't be determined: "
2252                             "dst=%s, error=%d\n",
2253                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2254                         goto bad;
2255                 }
2256                 srcp = &src;
2257         }
2258
2259         ip6->ip6_src = *srcp;
2260         ip6->ip6_flow = 0;
2261         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2262         ip6->ip6_vfc |= IPV6_VERSION;
2263         ip6->ip6_nxt = IPPROTO_ICMPV6;
2264         if (outif)
2265                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2266         else if (m->m_pkthdr.rcvif) {
2267                 /* XXX: This may not be the outgoing interface */
2268                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2269         } else
2270                 ip6->ip6_hlim = V_ip6_defhlim;
2271
2272         icmp6->icmp6_cksum = 0;
2273         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2274             sizeof(struct ip6_hdr), plen);
2275
2276         /*
2277          * XXX option handling
2278          */
2279
2280         m->m_flags &= ~(M_BCAST|M_MCAST);
2281
2282         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2283         if (outif)
2284                 icmp6_ifoutstat_inc(outif, type, code);
2285
2286         if (ia != NULL)
2287                 ifa_free(&ia->ia_ifa);
2288         return;
2289
2290  bad:
2291         if (ia != NULL)
2292                 ifa_free(&ia->ia_ifa);
2293         m_freem(m);
2294         return;
2295 }
2296
2297 void
2298 icmp6_fasttimo(void)
2299 {
2300
2301         mld_fasttimo();
2302 }
2303
2304 void
2305 icmp6_slowtimo(void)
2306 {
2307
2308         mld_slowtimo();
2309 }
2310
2311 static const char *
2312 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2313     struct in6_addr *tgt6)
2314 {
2315         static char buf[1024];
2316         char ip6bufs[INET6_ADDRSTRLEN];
2317         char ip6bufd[INET6_ADDRSTRLEN];
2318         char ip6buft[INET6_ADDRSTRLEN];
2319         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2320             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2321             ip6_sprintf(ip6buft, tgt6));
2322         return buf;
2323 }
2324
2325 void
2326 icmp6_redirect_input(struct mbuf *m, int off)
2327 {
2328         struct ifnet *ifp;
2329         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2330         struct nd_redirect *nd_rd;
2331         int icmp6len = ntohs(ip6->ip6_plen);
2332         char *lladdr = NULL;
2333         int lladdrlen = 0;
2334         struct rtentry *rt = NULL;
2335         int is_router;
2336         int is_onlink;
2337         struct in6_addr src6 = ip6->ip6_src;
2338         struct in6_addr redtgt6;
2339         struct in6_addr reddst6;
2340         union nd_opts ndopts;
2341         char ip6buf[INET6_ADDRSTRLEN];
2342
2343         M_ASSERTPKTHDR(m);
2344         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2345
2346         ifp = m->m_pkthdr.rcvif;
2347
2348         /* XXX if we are router, we don't update route by icmp6 redirect */
2349         if (V_ip6_forwarding)
2350                 goto freeit;
2351         if (!V_icmp6_rediraccept)
2352                 goto freeit;
2353
2354 #ifndef PULLDOWN_TEST
2355         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2356         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2357 #else
2358         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2359         if (nd_rd == NULL) {
2360                 ICMP6STAT_INC(icp6s_tooshort);
2361                 return;
2362         }
2363 #endif
2364         redtgt6 = nd_rd->nd_rd_target;
2365         reddst6 = nd_rd->nd_rd_dst;
2366
2367         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2368             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2369                 goto freeit;
2370         }
2371
2372         /* validation */
2373         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2374                 nd6log((LOG_ERR,
2375                     "ICMP6 redirect sent from %s rejected; "
2376                     "must be from linklocal\n",
2377                     ip6_sprintf(ip6buf, &src6)));
2378                 goto bad;
2379         }
2380         if (ip6->ip6_hlim != 255) {
2381                 nd6log((LOG_ERR,
2382                     "ICMP6 redirect sent from %s rejected; "
2383                     "hlim=%d (must be 255)\n",
2384                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2385                 goto bad;
2386         }
2387     {
2388         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2389         struct sockaddr_in6 sin6;
2390         struct in6_addr *gw6;
2391
2392         bzero(&sin6, sizeof(sin6));
2393         sin6.sin6_family = AF_INET6;
2394         sin6.sin6_len = sizeof(struct sockaddr_in6);
2395         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2396         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2397         if (rt) {
2398                 if (rt->rt_gateway == NULL ||
2399                     rt->rt_gateway->sa_family != AF_INET6) {
2400                         RTFREE_LOCKED(rt);
2401                         nd6log((LOG_ERR,
2402                             "ICMP6 redirect rejected; no route "
2403                             "with inet6 gateway found for redirect dst: %s\n",
2404                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2405                         goto bad;
2406                 }
2407
2408                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2409                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2410                         RTFREE_LOCKED(rt);
2411                         nd6log((LOG_ERR,
2412                             "ICMP6 redirect rejected; "
2413                             "not equal to gw-for-src=%s (must be same): "
2414                             "%s\n",
2415                             ip6_sprintf(ip6buf, gw6),
2416                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2417                         goto bad;
2418                 }
2419         } else {
2420                 nd6log((LOG_ERR,
2421                     "ICMP6 redirect rejected; "
2422                     "no route found for redirect dst: %s\n",
2423                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2424                 goto bad;
2425         }
2426         RTFREE_LOCKED(rt);
2427         rt = NULL;
2428     }
2429         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2430                 nd6log((LOG_ERR,
2431                     "ICMP6 redirect rejected; "
2432                     "redirect dst must be unicast: %s\n",
2433                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2434                 goto bad;
2435         }
2436
2437         is_router = is_onlink = 0;
2438         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2439                 is_router = 1;  /* router case */
2440         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2441                 is_onlink = 1;  /* on-link destination case */
2442         if (!is_router && !is_onlink) {
2443                 nd6log((LOG_ERR,
2444                     "ICMP6 redirect rejected; "
2445                     "neither router case nor onlink case: %s\n",
2446                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2447                 goto bad;
2448         }
2449         /* validation passed */
2450
2451         icmp6len -= sizeof(*nd_rd);
2452         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2453         if (nd6_options(&ndopts) < 0) {
2454                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2455                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2456                 /* nd6_options have incremented stats */
2457                 goto freeit;
2458         }
2459
2460         if (ndopts.nd_opts_tgt_lladdr) {
2461                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2462                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2463         }
2464
2465         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2466                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2467                     "(if %d, icmp6 packet %d): %s\n",
2468                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2469                     ifp->if_addrlen, lladdrlen - 2,
2470                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2471                 goto bad;
2472         }
2473
2474         /* RFC 2461 8.3 */
2475         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2476             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2477
2478         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2479                 /* perform rtredirect */
2480                 struct sockaddr_in6 sdst;
2481                 struct sockaddr_in6 sgw;
2482                 struct sockaddr_in6 ssrc;
2483                 u_int fibnum;
2484
2485                 bzero(&sdst, sizeof(sdst));
2486                 bzero(&sgw, sizeof(sgw));
2487                 bzero(&ssrc, sizeof(ssrc));
2488                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2489                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2490                         sizeof(struct sockaddr_in6);
2491                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2492                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2493                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2494                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2495                         in6_rtredirect((struct sockaddr *)&sdst,
2496                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2497                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2498                             fibnum);
2499         }
2500         /* finally update cached route in each socket via pfctlinput */
2501     {
2502         struct sockaddr_in6 sdst;
2503
2504         bzero(&sdst, sizeof(sdst));
2505         sdst.sin6_family = AF_INET6;
2506         sdst.sin6_len = sizeof(struct sockaddr_in6);
2507         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2508         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2509 #ifdef IPSEC
2510         key_sa_routechange((struct sockaddr *)&sdst);
2511 #endif /* IPSEC */
2512     }
2513
2514  freeit:
2515         m_freem(m);
2516         return;
2517
2518  bad:
2519         ICMP6STAT_INC(icp6s_badredirect);
2520         m_freem(m);
2521 }
2522
2523 void
2524 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2525 {
2526         struct ifnet *ifp;      /* my outgoing interface */
2527         struct in6_addr *ifp_ll6;
2528         struct in6_addr *router_ll6;
2529         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2530         struct mbuf *m = NULL;  /* newly allocated one */
2531         struct m_tag *mtag;
2532         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2533         struct nd_redirect *nd_rd;
2534         struct llentry *ln = NULL;
2535         size_t maxlen;
2536         u_char *p;
2537         struct ifnet *outif = NULL;
2538         struct sockaddr_in6 src_sa;
2539
2540         icmp6_errcount(ND_REDIRECT, 0);
2541
2542         /* if we are not router, we don't send icmp6 redirect */
2543         if (!V_ip6_forwarding)
2544                 goto fail;
2545
2546         /* sanity check */
2547         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2548                 goto fail;
2549
2550         /*
2551          * Address check:
2552          *  the source address must identify a neighbor, and
2553          *  the destination address must not be a multicast address
2554          *  [RFC 2461, sec 8.2]
2555          */
2556         sip6 = mtod(m0, struct ip6_hdr *);
2557         bzero(&src_sa, sizeof(src_sa));
2558         src_sa.sin6_family = AF_INET6;
2559         src_sa.sin6_len = sizeof(src_sa);
2560         src_sa.sin6_addr = sip6->ip6_src;
2561         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2562                 goto fail;
2563         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2564                 goto fail;      /* what should we do here? */
2565
2566         /* rate limit */
2567         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2568                 goto fail;
2569
2570         /*
2571          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2572          * we almost always ask for an mbuf cluster for simplicity.
2573          * (MHLEN < IPV6_MMTU is almost always true)
2574          */
2575 #if IPV6_MMTU >= MCLBYTES
2576 # error assumption failed about IPV6_MMTU and MCLBYTES
2577 #endif
2578         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2579         if (m == NULL)
2580                 goto fail;
2581         M_SETFIB(m, rt->rt_fibnum);
2582         maxlen = M_TRAILINGSPACE(m);
2583         maxlen = min(IPV6_MMTU, maxlen);
2584         /* just for safety */
2585         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2586             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2587                 goto fail;
2588         }
2589
2590         {
2591                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2592                 struct in6_ifaddr *ia;
2593                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2594                                                  IN6_IFF_NOTREADY|
2595                                                  IN6_IFF_ANYCAST)) == NULL)
2596                         goto fail;
2597                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2598                 /* XXXRW: reference released prematurely. */
2599                 ifa_free(&ia->ia_ifa);
2600         }
2601
2602         /* get ip6 linklocal address for the router. */
2603         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2604                 struct sockaddr_in6 *sin6;
2605                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2606                 router_ll6 = &sin6->sin6_addr;
2607                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2608                         router_ll6 = (struct in6_addr *)NULL;
2609         } else
2610                 router_ll6 = (struct in6_addr *)NULL;
2611
2612         /* ip6 */
2613         ip6 = mtod(m, struct ip6_hdr *);
2614         ip6->ip6_flow = 0;
2615         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2616         ip6->ip6_vfc |= IPV6_VERSION;
2617         /* ip6->ip6_plen will be set later */
2618         ip6->ip6_nxt = IPPROTO_ICMPV6;
2619         ip6->ip6_hlim = 255;
2620         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2621         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2622         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2623
2624         /* ND Redirect */
2625         nd_rd = (struct nd_redirect *)(ip6 + 1);
2626         nd_rd->nd_rd_type = ND_REDIRECT;
2627         nd_rd->nd_rd_code = 0;
2628         nd_rd->nd_rd_reserved = 0;
2629         if (rt->rt_flags & RTF_GATEWAY) {
2630                 /*
2631                  * nd_rd->nd_rd_target must be a link-local address in
2632                  * better router cases.
2633                  */
2634                 if (!router_ll6)
2635                         goto fail;
2636                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2637                     sizeof(nd_rd->nd_rd_target));
2638                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2639                     sizeof(nd_rd->nd_rd_dst));
2640         } else {
2641                 /* make sure redtgt == reddst */
2642                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2643                     sizeof(nd_rd->nd_rd_target));
2644                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2645                     sizeof(nd_rd->nd_rd_dst));
2646         }
2647
2648         p = (u_char *)(nd_rd + 1);
2649
2650         if (!router_ll6)
2651                 goto nolladdropt;
2652
2653         {
2654                 /* target lladdr option */
2655                 int len;
2656                 struct nd_opt_hdr *nd_opt;
2657                 char *lladdr;
2658
2659                 IF_AFDATA_RLOCK(ifp);
2660                 ln = nd6_lookup(router_ll6, 0, ifp);
2661                 IF_AFDATA_RUNLOCK(ifp);
2662                 if (ln == NULL)
2663                         goto nolladdropt;
2664
2665                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2666                 len = (len + 7) & ~7;   /* round by 8 */
2667                 /* safety check */
2668                 if (len + (p - (u_char *)ip6) > maxlen)                         
2669                         goto nolladdropt;
2670
2671                 if (ln->la_flags & LLE_VALID) {
2672                         nd_opt = (struct nd_opt_hdr *)p;
2673                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2674                         nd_opt->nd_opt_len = len >> 3;
2675                         lladdr = (char *)(nd_opt + 1);
2676                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2677                         p += len;
2678                 }
2679         }
2680 nolladdropt:
2681         if (ln != NULL)
2682                 LLE_RUNLOCK(ln);
2683                 
2684         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2685
2686         /* just to be safe */
2687 #ifdef M_DECRYPTED      /*not openbsd*/
2688         if (m0->m_flags & M_DECRYPTED)
2689                 goto noredhdropt;
2690 #endif
2691         if (p - (u_char *)ip6 > maxlen)
2692                 goto noredhdropt;
2693
2694         {
2695                 /* redirected header option */
2696                 int len;
2697                 struct nd_opt_rd_hdr *nd_opt_rh;
2698
2699                 /*
2700                  * compute the maximum size for icmp6 redirect header option.
2701                  * XXX room for auth header?
2702                  */
2703                 len = maxlen - (p - (u_char *)ip6);
2704                 len &= ~7;
2705
2706                 /* This is just for simplicity. */
2707                 if (m0->m_pkthdr.len != m0->m_len) {
2708                         if (m0->m_next) {
2709                                 m_freem(m0->m_next);
2710                                 m0->m_next = NULL;
2711                         }
2712                         m0->m_pkthdr.len = m0->m_len;
2713                 }
2714
2715                 /*
2716                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2717                  * about padding/truncate rule for the original IP packet.
2718                  * From the discussion on IPv6imp in Feb 1999,
2719                  * the consensus was:
2720                  * - "attach as much as possible" is the goal
2721                  * - pad if not aligned (original size can be guessed by
2722                  *   original ip6 header)
2723                  * Following code adds the padding if it is simple enough,
2724                  * and truncates if not.
2725                  */
2726                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2727                         panic("assumption failed in %s:%d", __FILE__,
2728                             __LINE__);
2729
2730                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2731                         /* not enough room, truncate */
2732                         m0->m_pkthdr.len = m0->m_len = len -
2733                             sizeof(*nd_opt_rh);
2734                 } else {
2735                         /* enough room, pad or truncate */
2736                         size_t extra;
2737
2738                         extra = m0->m_pkthdr.len % 8;
2739                         if (extra) {
2740                                 /* pad if easy enough, truncate if not */
2741                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2742                                         /* pad */
2743                                         m0->m_len += (8 - extra);
2744                                         m0->m_pkthdr.len += (8 - extra);
2745                                 } else {
2746                                         /* truncate */
2747                                         m0->m_pkthdr.len -= extra;
2748                                         m0->m_len -= extra;
2749                                 }
2750                         }
2751                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2752                         m0->m_pkthdr.len = m0->m_len = len -
2753                             sizeof(*nd_opt_rh);
2754                 }
2755
2756                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2757                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2758                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2759                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2760                 p += sizeof(*nd_opt_rh);
2761                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2762
2763                 /* connect m0 to m */
2764                 m_tag_delete_chain(m0, NULL);
2765                 m0->m_flags &= ~M_PKTHDR;
2766                 m->m_next = m0;
2767                 m->m_pkthdr.len = m->m_len + m0->m_len;
2768                 m0 = NULL;
2769         }
2770 noredhdropt:;
2771         if (m0) {
2772                 m_freem(m0);
2773                 m0 = NULL;
2774         }
2775
2776         /* XXX: clear embedded link IDs in the inner header */
2777         in6_clearscope(&sip6->ip6_src);
2778         in6_clearscope(&sip6->ip6_dst);
2779         in6_clearscope(&nd_rd->nd_rd_target);
2780         in6_clearscope(&nd_rd->nd_rd_dst);
2781
2782         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2783
2784         nd_rd->nd_rd_cksum = 0;
2785         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2786             sizeof(*ip6), ntohs(ip6->ip6_plen));
2787
2788         if (send_sendso_input_hook != NULL) {
2789                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2790                         M_NOWAIT);
2791                 if (mtag == NULL)
2792                         goto fail;
2793                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2794                 m_tag_prepend(m, mtag);
2795         }
2796
2797         /* send the packet to outside... */
2798         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2799         if (outif) {
2800                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2801                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2802         }
2803         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2804
2805         return;
2806
2807 fail:
2808         if (m)
2809                 m_freem(m);
2810         if (m0)
2811                 m_freem(m0);
2812 }
2813
2814 /*
2815  * ICMPv6 socket option processing.
2816  */
2817 int
2818 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2819 {
2820         int error = 0;
2821         int optlen;
2822         struct inpcb *inp = sotoinpcb(so);
2823         int level, op, optname;
2824
2825         if (sopt) {
2826                 level = sopt->sopt_level;
2827                 op = sopt->sopt_dir;
2828                 optname = sopt->sopt_name;
2829                 optlen = sopt->sopt_valsize;
2830         } else
2831                 level = op = optname = optlen = 0;
2832
2833         if (level != IPPROTO_ICMPV6) {
2834                 return EINVAL;
2835         }
2836
2837         switch (op) {
2838         case PRCO_SETOPT:
2839                 switch (optname) {
2840                 case ICMP6_FILTER:
2841                     {
2842                         struct icmp6_filter ic6f;
2843
2844                         if (optlen != sizeof(ic6f)) {
2845                                 error = EMSGSIZE;
2846                                 break;
2847                         }
2848                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2849                         if (error == 0) {
2850                                 INP_WLOCK(inp);
2851                                 *inp->in6p_icmp6filt = ic6f;
2852                                 INP_WUNLOCK(inp);
2853                         }
2854                         break;
2855                     }
2856
2857                 default:
2858                         error = ENOPROTOOPT;
2859                         break;
2860                 }
2861                 break;
2862
2863         case PRCO_GETOPT:
2864                 switch (optname) {
2865                 case ICMP6_FILTER:
2866                     {
2867                         struct icmp6_filter ic6f;
2868
2869                         INP_RLOCK(inp);
2870                         ic6f = *inp->in6p_icmp6filt;
2871                         INP_RUNLOCK(inp);
2872                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2873                         break;
2874                     }
2875
2876                 default:
2877                         error = ENOPROTOOPT;
2878                         break;
2879                 }
2880                 break;
2881         }
2882
2883         return (error);
2884 }
2885
2886 /*
2887  * Perform rate limit check.
2888  * Returns 0 if it is okay to send the icmp6 packet.
2889  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2890  * limitation.
2891  *
2892  * XXX per-destination/type check necessary?
2893  *
2894  * dst - not used at this moment
2895  * type - not used at this moment
2896  * code - not used at this moment
2897  */
2898 static int
2899 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2900     const int code)
2901 {
2902         int ret;
2903
2904         ret = 0;        /* okay to send */
2905
2906         /* PPS limit */
2907         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2908             V_icmp6errppslim)) {
2909                 /* The packet is subject to rate limit */
2910                 ret++;
2911         }
2912
2913         return ret;
2914 }