]> CyberLeo.Net >> Repos - FreeBSD/releng/10.1.git/blob - sys/netinet6/icmp6.c
Document r273098, options for displaying mkimg(1) internals
[FreeBSD/releng/10.1.git] / sys / netinet6 / icmp6.c
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *      The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
61  */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/icmp6.h>
99 #include <netinet/tcp_var.h>
100
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/send.h>
109
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/key.h>
113 #endif
114
115 extern struct domain inet6domain;
116
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 VNET_DECLARE(struct inpcbhead, ripcb);
126 VNET_DECLARE(int, icmp6errppslim);
127 static VNET_DEFINE(int, icmp6errpps_count) = 0;
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130
131 #define V_ripcbinfo                     VNET(ripcbinfo)
132 #define V_ripcb                         VNET(ripcb)
133 #define V_icmp6errppslim                VNET(icmp6errppslim)
134 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
135 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
136 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
137
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
141 static const char *icmp6_redirect_diag(struct in6_addr *,
142         struct in6_addr *, struct in6_addr *);
143 static struct mbuf *ni6_input(struct mbuf *, int);
144 static struct mbuf *ni6_nametodns(const char *, int, int);
145 static int ni6_dnsmatch(const char *, int, const char *, int);
146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
147                           struct ifnet **, struct in6_addr *);
148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
149                                 struct ifnet *, int);
150 static int icmp6_notify_error(struct mbuf **, int, int, int);
151
152 /*
153  * Kernel module interface for updating icmp6stat.  The argument is an index
154  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
155  * general layout of icmp6stat into the caller, it doesn't encode its
156  * location, so that future changes to add, for example, per-CPU stats
157  * support won't cause binary compatibility problems for kernel modules.
158  */
159 void
160 kmod_icmp6stat_inc(int statnum)
161 {
162
163         counter_u64_add(VNET(icmp6stat)[statnum], 1);
164 }
165
166 static void
167 icmp6_errcount(int type, int code)
168 {
169         switch (type) {
170         case ICMP6_DST_UNREACH:
171                 switch (code) {
172                 case ICMP6_DST_UNREACH_NOROUTE:
173                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
174                         return;
175                 case ICMP6_DST_UNREACH_ADMIN:
176                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
177                         return;
178                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
179                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
180                         return;
181                 case ICMP6_DST_UNREACH_ADDR:
182                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
183                         return;
184                 case ICMP6_DST_UNREACH_NOPORT:
185                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
186                         return;
187                 }
188                 break;
189         case ICMP6_PACKET_TOO_BIG:
190                 ICMP6STAT_INC(icp6s_opacket_too_big);
191                 return;
192         case ICMP6_TIME_EXCEEDED:
193                 switch (code) {
194                 case ICMP6_TIME_EXCEED_TRANSIT:
195                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
196                         return;
197                 case ICMP6_TIME_EXCEED_REASSEMBLY:
198                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
199                         return;
200                 }
201                 break;
202         case ICMP6_PARAM_PROB:
203                 switch (code) {
204                 case ICMP6_PARAMPROB_HEADER:
205                         ICMP6STAT_INC(icp6s_oparamprob_header);
206                         return;
207                 case ICMP6_PARAMPROB_NEXTHEADER:
208                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
209                         return;
210                 case ICMP6_PARAMPROB_OPTION:
211                         ICMP6STAT_INC(icp6s_oparamprob_option);
212                         return;
213                 }
214                 break;
215         case ND_REDIRECT:
216                 ICMP6STAT_INC(icp6s_oredirect);
217                 return;
218         }
219         ICMP6STAT_INC(icp6s_ounknown);
220 }
221
222 /*
223  * A wrapper function for icmp6_error() necessary when the erroneous packet
224  * may not contain enough scope zone information.
225  */
226 void
227 icmp6_error2(struct mbuf *m, int type, int code, int param,
228     struct ifnet *ifp)
229 {
230         struct ip6_hdr *ip6;
231
232         if (ifp == NULL)
233                 return;
234
235 #ifndef PULLDOWN_TEST
236         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
237 #else
238         if (m->m_len < sizeof(struct ip6_hdr)) {
239                 m = m_pullup(m, sizeof(struct ip6_hdr));
240                 if (m == NULL)
241                         return;
242         }
243 #endif
244
245         ip6 = mtod(m, struct ip6_hdr *);
246
247         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
248                 return;
249         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
250                 return;
251
252         icmp6_error(m, type, code, param);
253 }
254
255 /*
256  * Generate an error packet of type error in response to bad IP6 packet.
257  */
258 void
259 icmp6_error(struct mbuf *m, int type, int code, int param)
260 {
261         struct ip6_hdr *oip6, *nip6;
262         struct icmp6_hdr *icmp6;
263         u_int preplen;
264         int off;
265         int nxt;
266
267         ICMP6STAT_INC(icp6s_error);
268
269         /* count per-type-code statistics */
270         icmp6_errcount(type, code);
271
272 #ifdef M_DECRYPTED      /*not openbsd*/
273         if (m->m_flags & M_DECRYPTED) {
274                 ICMP6STAT_INC(icp6s_canterror);
275                 goto freeit;
276         }
277 #endif
278
279 #ifndef PULLDOWN_TEST
280         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
281 #else
282         if (m->m_len < sizeof(struct ip6_hdr)) {
283                 m = m_pullup(m, sizeof(struct ip6_hdr));
284                 if (m == NULL)
285                         return;
286         }
287 #endif
288         oip6 = mtod(m, struct ip6_hdr *);
289
290         /*
291          * If the destination address of the erroneous packet is a multicast
292          * address, or the packet was sent using link-layer multicast,
293          * we should basically suppress sending an error (RFC 2463, Section
294          * 2.4).
295          * We have two exceptions (the item e.2 in that section):
296          * - the Packet Too Big message can be sent for path MTU discovery.
297          * - the Parameter Problem Message that can be allowed an icmp6 error
298          *   in the option type field.  This check has been done in
299          *   ip6_unknown_opt(), so we can just check the type and code.
300          */
301         if ((m->m_flags & (M_BCAST|M_MCAST) ||
302              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
303             (type != ICMP6_PACKET_TOO_BIG &&
304              (type != ICMP6_PARAM_PROB ||
305               code != ICMP6_PARAMPROB_OPTION)))
306                 goto freeit;
307
308         /*
309          * RFC 2463, 2.4 (e.5): source address check.
310          * XXX: the case of anycast source?
311          */
312         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
313             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
314                 goto freeit;
315
316         /*
317          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
318          * don't do it.
319          */
320         nxt = -1;
321         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
322         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
323                 struct icmp6_hdr *icp;
324
325 #ifndef PULLDOWN_TEST
326                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
327                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
328 #else
329                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
330                         sizeof(*icp));
331                 if (icp == NULL) {
332                         ICMP6STAT_INC(icp6s_tooshort);
333                         return;
334                 }
335 #endif
336                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
337                     icp->icmp6_type == ND_REDIRECT) {
338                         /*
339                          * ICMPv6 error
340                          * Special case: for redirect (which is
341                          * informational) we must not send icmp6 error.
342                          */
343                         ICMP6STAT_INC(icp6s_canterror);
344                         goto freeit;
345                 } else {
346                         /* ICMPv6 informational - send the error */
347                 }
348         } else {
349                 /* non-ICMPv6 - send the error */
350         }
351
352         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
353
354         /* Finally, do rate limitation check. */
355         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
356                 ICMP6STAT_INC(icp6s_toofreq);
357                 goto freeit;
358         }
359
360         /*
361          * OK, ICMP6 can be generated.
362          */
363
364         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
365                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
366
367         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
368         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
369         if (m && m->m_len < preplen)
370                 m = m_pullup(m, preplen);
371         if (m == NULL) {
372                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
373                 return;
374         }
375
376         nip6 = mtod(m, struct ip6_hdr *);
377         nip6->ip6_src  = oip6->ip6_src;
378         nip6->ip6_dst  = oip6->ip6_dst;
379
380         in6_clearscope(&oip6->ip6_src);
381         in6_clearscope(&oip6->ip6_dst);
382
383         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
384         icmp6->icmp6_type = type;
385         icmp6->icmp6_code = code;
386         icmp6->icmp6_pptr = htonl((u_int32_t)param);
387
388         /*
389          * icmp6_reflect() is designed to be in the input path.
390          * icmp6_error() can be called from both input and output path,
391          * and if we are in output path rcvif could contain bogus value.
392          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
393          * information in ip header (nip6).
394          */
395         m->m_pkthdr.rcvif = NULL;
396
397         ICMP6STAT_INC(icp6s_outhist[type]);
398         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
399
400         return;
401
402   freeit:
403         /*
404          * If we can't tell whether or not we can generate ICMP6, free it.
405          */
406         m_freem(m);
407 }
408
409 /*
410  * Process a received ICMP6 message.
411  */
412 int
413 icmp6_input(struct mbuf **mp, int *offp, int proto)
414 {
415         struct mbuf *m = *mp, *n;
416         struct ifnet *ifp;
417         struct ip6_hdr *ip6, *nip6;
418         struct icmp6_hdr *icmp6, *nicmp6;
419         int off = *offp;
420         int icmp6len = m->m_pkthdr.len - *offp;
421         int code, sum, noff;
422         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
423         int ip6len, error;
424
425         ifp = m->m_pkthdr.rcvif;
426
427 #ifndef PULLDOWN_TEST
428         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
429         /* m might change if M_LOOP.  So, call mtod after this */
430 #endif
431
432         /*
433          * Locate icmp6 structure in mbuf, and check
434          * that not corrupted and of at least minimum length
435          */
436
437         ip6 = mtod(m, struct ip6_hdr *);
438         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
439         if (icmp6len < sizeof(struct icmp6_hdr)) {
440                 ICMP6STAT_INC(icp6s_tooshort);
441                 goto freeit;
442         }
443
444         /*
445          * Check multicast group membership.
446          * Note: SSM filters are not applied for ICMPv6 traffic.
447          */
448         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
449                 struct in6_multi        *inm;
450
451                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
452                 if (inm == NULL) {
453                         IP6STAT_INC(ip6s_notmember);
454                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
455                         goto freeit;
456                 }
457         }
458
459         /*
460          * calculate the checksum
461          */
462 #ifndef PULLDOWN_TEST
463         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
464 #else
465         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
466         if (icmp6 == NULL) {
467                 ICMP6STAT_INC(icp6s_tooshort);
468                 return IPPROTO_DONE;
469         }
470 #endif
471         code = icmp6->icmp6_code;
472
473         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
474                 nd6log((LOG_ERR,
475                     "ICMP6 checksum error(%d|%x) %s\n",
476                     icmp6->icmp6_type, sum,
477                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
478                 ICMP6STAT_INC(icp6s_checksum);
479                 goto freeit;
480         }
481
482         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
483                 /*
484                  * Deliver very specific ICMP6 type only.
485                  * This is important to deliver TOOBIG.  Otherwise PMTUD
486                  * will not work.
487                  */
488                 switch (icmp6->icmp6_type) {
489                 case ICMP6_DST_UNREACH:
490                 case ICMP6_PACKET_TOO_BIG:
491                 case ICMP6_TIME_EXCEEDED:
492                         break;
493                 default:
494                         goto freeit;
495                 }
496         }
497
498         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
499         icmp6_ifstat_inc(ifp, ifs6_in_msg);
500         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
501                 icmp6_ifstat_inc(ifp, ifs6_in_error);
502
503         switch (icmp6->icmp6_type) {
504         case ICMP6_DST_UNREACH:
505                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
506                 switch (code) {
507                 case ICMP6_DST_UNREACH_NOROUTE:
508                         code = PRC_UNREACH_NET;
509                         break;
510                 case ICMP6_DST_UNREACH_ADMIN:
511                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
512                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
513                         break;
514                 case ICMP6_DST_UNREACH_ADDR:
515                         code = PRC_HOSTDEAD;
516                         break;
517                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
518                         /* I mean "source address was incorrect." */
519                         code = PRC_PARAMPROB;
520                         break;
521                 case ICMP6_DST_UNREACH_NOPORT:
522                         code = PRC_UNREACH_PORT;
523                         break;
524                 default:
525                         goto badcode;
526                 }
527                 goto deliver;
528                 break;
529
530         case ICMP6_PACKET_TOO_BIG:
531                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
532
533                 /* validation is made in icmp6_mtudisc_update */
534
535                 code = PRC_MSGSIZE;
536
537                 /*
538                  * Updating the path MTU will be done after examining
539                  * intermediate extension headers.
540                  */
541                 goto deliver;
542                 break;
543
544         case ICMP6_TIME_EXCEEDED:
545                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
546                 switch (code) {
547                 case ICMP6_TIME_EXCEED_TRANSIT:
548                         code = PRC_TIMXCEED_INTRANS;
549                         break;
550                 case ICMP6_TIME_EXCEED_REASSEMBLY:
551                         code = PRC_TIMXCEED_REASS;
552                         break;
553                 default:
554                         goto badcode;
555                 }
556                 goto deliver;
557                 break;
558
559         case ICMP6_PARAM_PROB:
560                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
561                 switch (code) {
562                 case ICMP6_PARAMPROB_NEXTHEADER:
563                         code = PRC_UNREACH_PROTOCOL;
564                         break;
565                 case ICMP6_PARAMPROB_HEADER:
566                 case ICMP6_PARAMPROB_OPTION:
567                         code = PRC_PARAMPROB;
568                         break;
569                 default:
570                         goto badcode;
571                 }
572                 goto deliver;
573                 break;
574
575         case ICMP6_ECHO_REQUEST:
576                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
577                 if (code != 0)
578                         goto badcode;
579                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
580                         /* Give up remote */
581                         break;
582                 }
583                 if ((n->m_flags & M_EXT) != 0
584                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
585                         struct mbuf *n0 = n;
586                         int n0len;
587
588                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
589                         n = m_gethdr(M_NOWAIT, n0->m_type);
590                         if (n == NULL) {
591                                 /* Give up remote */
592                                 m_freem(n0);
593                                 break;
594                         }
595
596                         m_move_pkthdr(n, n0);   /* FIB copied. */
597                         n0len = n0->m_pkthdr.len;       /* save for use below */
598                         /*
599                          * Copy IPv6 and ICMPv6 only.
600                          */
601                         nip6 = mtod(n, struct ip6_hdr *);
602                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
603                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
604                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
605                         noff = sizeof(struct ip6_hdr);
606                         /* new mbuf contains only ipv6+icmpv6 headers */
607                         n->m_len = noff + sizeof(struct icmp6_hdr);
608                         /*
609                          * Adjust mbuf.  ip6_plen will be adjusted in
610                          * ip6_output().
611                          */
612                         m_adj(n0, off + sizeof(struct icmp6_hdr));
613                         /* recalculate complete packet size */
614                         n->m_pkthdr.len = n0len + (noff - off);
615                         n->m_next = n0;
616                 } else {
617                         nip6 = mtod(n, struct ip6_hdr *);
618                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
619                             sizeof(*nicmp6));
620                         noff = off;
621                 }
622                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
623                 nicmp6->icmp6_code = 0;
624                 if (n) {
625                         ICMP6STAT_INC(icp6s_reflect);
626                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
627                         icmp6_reflect(n, noff);
628                 }
629                 break;
630
631         case ICMP6_ECHO_REPLY:
632                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
633                 if (code != 0)
634                         goto badcode;
635                 break;
636
637         case MLD_LISTENER_QUERY:
638         case MLD_LISTENER_REPORT:
639         case MLD_LISTENER_DONE:
640         case MLDV2_LISTENER_REPORT:
641                 /*
642                  * Drop MLD traffic which is not link-local, has a hop limit
643                  * of greater than 1 hop, or which does not have the
644                  * IPv6 HBH Router Alert option.
645                  * As IPv6 HBH options are stripped in ip6_input() we must
646                  * check an mbuf header flag.
647                  * XXX Should we also sanity check that these messages
648                  * were directed to a link-local multicast prefix?
649                  */
650                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
651                         goto freeit;
652                 if (mld_input(m, off, icmp6len) != 0)
653                         return (IPPROTO_DONE);
654                 /* m stays. */
655                 break;
656
657         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
658             {
659                 enum { WRU, FQDN } mode;
660
661                 if (!V_icmp6_nodeinfo)
662                         break;
663
664                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
665                         mode = WRU;
666                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
667                         mode = FQDN;
668                 else
669                         goto badlen;
670
671                 if (mode == FQDN) {
672 #ifndef PULLDOWN_TEST
673                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
674                             IPPROTO_DONE);
675 #endif
676                         n = m_copy(m, 0, M_COPYALL);
677                         if (n)
678                                 n = ni6_input(n, off);
679                         /* XXX meaningless if n == NULL */
680                         noff = sizeof(struct ip6_hdr);
681                 } else {
682                         struct prison *pr;
683                         u_char *p;
684                         int maxhlen, hlen;
685
686                         /*
687                          * XXX: this combination of flags is pointless,
688                          * but should we keep this for compatibility?
689                          */
690                         if ((V_icmp6_nodeinfo & 5) != 5)
691                                 break;
692
693                         if (code != 0)
694                                 goto badcode;
695
696                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
697                         n = m_gethdr(M_NOWAIT, m->m_type);
698                         if (n == NULL) {
699                                 /* Give up remote */
700                                 break;
701                         }
702                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
703                                 /*
704                                  * Previous code did a blind M_COPY_PKTHDR
705                                  * and said "just for rcvif".  If true, then
706                                  * we could tolerate the dup failing (due to
707                                  * the deep copy of the tag chain).  For now
708                                  * be conservative and just fail.
709                                  */
710                                 m_free(n);
711                                 n = NULL;
712                         }
713                         maxhlen = M_TRAILINGSPACE(n) -
714                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
715                         pr = curthread->td_ucred->cr_prison;
716                         mtx_lock(&pr->pr_mtx);
717                         hlen = strlen(pr->pr_hostname);
718                         if (maxhlen > hlen)
719                                 maxhlen = hlen;
720                         /*
721                          * Copy IPv6 and ICMPv6 only.
722                          */
723                         nip6 = mtod(n, struct ip6_hdr *);
724                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
725                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
726                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
727                         p = (u_char *)(nicmp6 + 1);
728                         bzero(p, 4);
729                         /* meaningless TTL */
730                         bcopy(pr->pr_hostname, p + 4, maxhlen);
731                         mtx_unlock(&pr->pr_mtx);
732                         noff = sizeof(struct ip6_hdr);
733                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
734                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
735                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
736                         nicmp6->icmp6_code = 0;
737                 }
738                 if (n) {
739                         ICMP6STAT_INC(icp6s_reflect);
740                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
741                         icmp6_reflect(n, noff);
742                 }
743                 break;
744             }
745
746         case ICMP6_WRUREPLY:
747                 if (code != 0)
748                         goto badcode;
749                 break;
750
751         case ND_ROUTER_SOLICIT:
752                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
753                 if (code != 0)
754                         goto badcode;
755                 if (icmp6len < sizeof(struct nd_router_solicit))
756                         goto badlen;
757                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
758                         /* give up local */
759
760                         /* Send incoming SeND packet to user space. */
761                         if (send_sendso_input_hook != NULL) {
762                                 IP6_EXTHDR_CHECK(m, off,
763                                     icmp6len, IPPROTO_DONE);
764                                 error = send_sendso_input_hook(m, ifp,
765                                     SND_IN, ip6len);
766                                 /* -1 == no app on SEND socket */
767                                 if (error == 0)
768                                         return (IPPROTO_DONE);
769                                 nd6_rs_input(m, off, icmp6len);
770                         } else
771                                 nd6_rs_input(m, off, icmp6len);
772                         m = NULL;
773                         goto freeit;
774                 }
775                 if (send_sendso_input_hook != NULL) {
776                         IP6_EXTHDR_CHECK(n, off,
777                             icmp6len, IPPROTO_DONE);
778                         error = send_sendso_input_hook(n, ifp,
779                             SND_IN, ip6len);
780                         if (error == 0)
781                                 goto freeit;
782                         /* -1 == no app on SEND socket */
783                         nd6_rs_input(n, off, icmp6len);
784                 } else
785                         nd6_rs_input(n, off, icmp6len);
786                 /* m stays. */
787                 break;
788
789         case ND_ROUTER_ADVERT:
790                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
791                 if (code != 0)
792                         goto badcode;
793                 if (icmp6len < sizeof(struct nd_router_advert))
794                         goto badlen;
795                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
796
797                         /* Send incoming SeND-protected/ND packet to user space. */
798                         if (send_sendso_input_hook != NULL) {
799                                 error = send_sendso_input_hook(m, ifp,
800                                     SND_IN, ip6len);
801                                 if (error == 0)
802                                         return (IPPROTO_DONE);
803                                 nd6_ra_input(m, off, icmp6len);
804                         } else
805                                 nd6_ra_input(m, off, icmp6len);
806                         m = NULL;
807                         goto freeit;
808                 }
809                 if (send_sendso_input_hook != NULL) {
810                         error = send_sendso_input_hook(n, ifp,
811                             SND_IN, ip6len);
812                         if (error == 0)
813                                 goto freeit;
814                         nd6_ra_input(n, off, icmp6len);
815                 } else
816                         nd6_ra_input(n, off, icmp6len);
817                 /* m stays. */
818                 break;
819
820         case ND_NEIGHBOR_SOLICIT:
821                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
822                 if (code != 0)
823                         goto badcode;
824                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
825                         goto badlen;
826                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
827                         if (send_sendso_input_hook != NULL) {
828                                 error = send_sendso_input_hook(m, ifp,
829                                     SND_IN, ip6len);
830                                 if (error == 0)
831                                         return (IPPROTO_DONE);
832                                 nd6_ns_input(m, off, icmp6len);
833                         } else
834                                 nd6_ns_input(m, off, icmp6len);
835                         m = NULL;
836                         goto freeit;
837                 }
838                 if (send_sendso_input_hook != NULL) {
839                         error = send_sendso_input_hook(n, ifp,
840                             SND_IN, ip6len);
841                         if (error == 0)
842                                 goto freeit;
843                         nd6_ns_input(n, off, icmp6len);
844                 } else
845                         nd6_ns_input(n, off, icmp6len);
846                 /* m stays. */
847                 break;
848
849         case ND_NEIGHBOR_ADVERT:
850                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
851                 if (code != 0)
852                         goto badcode;
853                 if (icmp6len < sizeof(struct nd_neighbor_advert))
854                         goto badlen;
855                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
856
857                         /* Send incoming SeND-protected/ND packet to user space. */
858                         if (send_sendso_input_hook != NULL) {
859                                 error = send_sendso_input_hook(m, ifp,
860                                     SND_IN, ip6len);
861                                 if (error == 0)
862                                         return (IPPROTO_DONE);
863                                 nd6_na_input(m, off, icmp6len);
864                         } else
865                                 nd6_na_input(m, off, icmp6len);
866                         m = NULL;
867                         goto freeit;
868                 }
869                 if (send_sendso_input_hook != NULL) {
870                         error = send_sendso_input_hook(n, ifp,
871                             SND_IN, ip6len);
872                         if (error == 0)
873                                 goto freeit;
874                         nd6_na_input(n, off, icmp6len);
875                 } else
876                         nd6_na_input(n, off, icmp6len);
877                 /* m stays. */
878                 break;
879
880         case ND_REDIRECT:
881                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
882                 if (code != 0)
883                         goto badcode;
884                 if (icmp6len < sizeof(struct nd_redirect))
885                         goto badlen;
886                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
887                         if (send_sendso_input_hook != NULL) {
888                                 error = send_sendso_input_hook(m, ifp,
889                                     SND_IN, ip6len);
890                                 if (error == 0)
891                                         return (IPPROTO_DONE);
892                             icmp6_redirect_input(m, off);
893                         } else
894                                 icmp6_redirect_input(m, off);
895                         m = NULL;
896                         goto freeit;
897                 }
898                 if (send_sendso_input_hook != NULL) {
899                         error = send_sendso_input_hook(n, ifp,
900                             SND_IN, ip6len);
901                         if (error == 0)
902                                 goto freeit;
903                         icmp6_redirect_input(n, off);
904                 } else
905                         icmp6_redirect_input(n, off);
906                 /* m stays. */
907                 break;
908
909         case ICMP6_ROUTER_RENUMBERING:
910                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
911                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
912                         goto badcode;
913                 if (icmp6len < sizeof(struct icmp6_router_renum))
914                         goto badlen;
915                 break;
916
917         default:
918                 nd6log((LOG_DEBUG,
919                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
920                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
921                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
922                     ifp ? ifp->if_index : 0));
923                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
924                         /* ICMPv6 error: MUST deliver it by spec... */
925                         code = PRC_NCMDS;
926                         /* deliver */
927                 } else {
928                         /* ICMPv6 informational: MUST not deliver */
929                         break;
930                 }
931         deliver:
932                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
933                         /* In this case, m should've been freed. */
934                         return (IPPROTO_DONE);
935                 }
936                 break;
937
938         badcode:
939                 ICMP6STAT_INC(icp6s_badcode);
940                 break;
941
942         badlen:
943                 ICMP6STAT_INC(icp6s_badlen);
944                 break;
945         }
946
947         /* deliver the packet to appropriate sockets */
948         icmp6_rip6_input(&m, *offp);
949
950         return IPPROTO_DONE;
951
952  freeit:
953         m_freem(m);
954         return IPPROTO_DONE;
955 }
956
957 static int
958 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
959 {
960         struct mbuf *m = *mp;
961         struct icmp6_hdr *icmp6;
962         struct ip6_hdr *eip6;
963         u_int32_t notifymtu;
964         struct sockaddr_in6 icmp6src, icmp6dst;
965
966         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
967                 ICMP6STAT_INC(icp6s_tooshort);
968                 goto freeit;
969         }
970 #ifndef PULLDOWN_TEST
971         IP6_EXTHDR_CHECK(m, off,
972             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
973         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
974 #else
975         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
976             sizeof(*icmp6) + sizeof(struct ip6_hdr));
977         if (icmp6 == NULL) {
978                 ICMP6STAT_INC(icp6s_tooshort);
979                 return (-1);
980         }
981 #endif
982         eip6 = (struct ip6_hdr *)(icmp6 + 1);
983
984         /* Detect the upper level protocol */
985         {
986                 void (*ctlfunc)(int, struct sockaddr *, void *);
987                 u_int8_t nxt = eip6->ip6_nxt;
988                 int eoff = off + sizeof(struct icmp6_hdr) +
989                     sizeof(struct ip6_hdr);
990                 struct ip6ctlparam ip6cp;
991                 struct in6_addr *finaldst = NULL;
992                 int icmp6type = icmp6->icmp6_type;
993                 struct ip6_frag *fh;
994                 struct ip6_rthdr *rth;
995                 struct ip6_rthdr0 *rth0;
996                 int rthlen;
997
998                 while (1) { /* XXX: should avoid infinite loop explicitly? */
999                         struct ip6_ext *eh;
1000
1001                         switch (nxt) {
1002                         case IPPROTO_HOPOPTS:
1003                         case IPPROTO_DSTOPTS:
1004                         case IPPROTO_AH:
1005 #ifndef PULLDOWN_TEST
1006                                 IP6_EXTHDR_CHECK(m, 0,
1007                                     eoff + sizeof(struct ip6_ext), -1);
1008                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1009 #else
1010                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1011                                     eoff, sizeof(*eh));
1012                                 if (eh == NULL) {
1013                                         ICMP6STAT_INC(icp6s_tooshort);
1014                                         return (-1);
1015                                 }
1016 #endif
1017
1018                                 if (nxt == IPPROTO_AH)
1019                                         eoff += (eh->ip6e_len + 2) << 2;
1020                                 else
1021                                         eoff += (eh->ip6e_len + 1) << 3;
1022                                 nxt = eh->ip6e_nxt;
1023                                 break;
1024                         case IPPROTO_ROUTING:
1025                                 /*
1026                                  * When the erroneous packet contains a
1027                                  * routing header, we should examine the
1028                                  * header to determine the final destination.
1029                                  * Otherwise, we can't properly update
1030                                  * information that depends on the final
1031                                  * destination (e.g. path MTU).
1032                                  */
1033 #ifndef PULLDOWN_TEST
1034                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1035                                 rth = (struct ip6_rthdr *)
1036                                     (mtod(m, caddr_t) + eoff);
1037 #else
1038                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1039                                     eoff, sizeof(*rth));
1040                                 if (rth == NULL) {
1041                                         ICMP6STAT_INC(icp6s_tooshort);
1042                                         return (-1);
1043                                 }
1044 #endif
1045                                 rthlen = (rth->ip6r_len + 1) << 3;
1046                                 /*
1047                                  * XXX: currently there is no
1048                                  * officially defined type other
1049                                  * than type-0.
1050                                  * Note that if the segment left field
1051                                  * is 0, all intermediate hops must
1052                                  * have been passed.
1053                                  */
1054                                 if (rth->ip6r_segleft &&
1055                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1056                                         int hops;
1057
1058 #ifndef PULLDOWN_TEST
1059                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1060                                         rth0 = (struct ip6_rthdr0 *)
1061                                             (mtod(m, caddr_t) + eoff);
1062 #else
1063                                         IP6_EXTHDR_GET(rth0,
1064                                             struct ip6_rthdr0 *, m,
1065                                             eoff, rthlen);
1066                                         if (rth0 == NULL) {
1067                                                 ICMP6STAT_INC(icp6s_tooshort);
1068                                                 return (-1);
1069                                         }
1070 #endif
1071                                         /* just ignore a bogus header */
1072                                         if ((rth0->ip6r0_len % 2) == 0 &&
1073                                             (hops = rth0->ip6r0_len/2))
1074                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1075                                 }
1076                                 eoff += rthlen;
1077                                 nxt = rth->ip6r_nxt;
1078                                 break;
1079                         case IPPROTO_FRAGMENT:
1080 #ifndef PULLDOWN_TEST
1081                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1082                                     sizeof(struct ip6_frag), -1);
1083                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1084                                     eoff);
1085 #else
1086                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1087                                     eoff, sizeof(*fh));
1088                                 if (fh == NULL) {
1089                                         ICMP6STAT_INC(icp6s_tooshort);
1090                                         return (-1);
1091                                 }
1092 #endif
1093                                 /*
1094                                  * Data after a fragment header is meaningless
1095                                  * unless it is the first fragment, but
1096                                  * we'll go to the notify label for path MTU
1097                                  * discovery.
1098                                  */
1099                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1100                                         goto notify;
1101
1102                                 eoff += sizeof(struct ip6_frag);
1103                                 nxt = fh->ip6f_nxt;
1104                                 break;
1105                         default:
1106                                 /*
1107                                  * This case includes ESP and the No Next
1108                                  * Header.  In such cases going to the notify
1109                                  * label does not have any meaning
1110                                  * (i.e. ctlfunc will be NULL), but we go
1111                                  * anyway since we might have to update
1112                                  * path MTU information.
1113                                  */
1114                                 goto notify;
1115                         }
1116                 }
1117           notify:
1118 #ifndef PULLDOWN_TEST
1119                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1120 #else
1121                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1122                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1123                 if (icmp6 == NULL) {
1124                         ICMP6STAT_INC(icp6s_tooshort);
1125                         return (-1);
1126                 }
1127 #endif
1128
1129                 /*
1130                  * retrieve parameters from the inner IPv6 header, and convert
1131                  * them into sockaddr structures.
1132                  * XXX: there is no guarantee that the source or destination
1133                  * addresses of the inner packet are in the same scope as
1134                  * the addresses of the icmp packet.  But there is no other
1135                  * way to determine the zone.
1136                  */
1137                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1138
1139                 bzero(&icmp6dst, sizeof(icmp6dst));
1140                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1141                 icmp6dst.sin6_family = AF_INET6;
1142                 if (finaldst == NULL)
1143                         icmp6dst.sin6_addr = eip6->ip6_dst;
1144                 else
1145                         icmp6dst.sin6_addr = *finaldst;
1146                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1147                         goto freeit;
1148                 bzero(&icmp6src, sizeof(icmp6src));
1149                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1150                 icmp6src.sin6_family = AF_INET6;
1151                 icmp6src.sin6_addr = eip6->ip6_src;
1152                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1153                         goto freeit;
1154                 icmp6src.sin6_flowinfo =
1155                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1156
1157                 if (finaldst == NULL)
1158                         finaldst = &eip6->ip6_dst;
1159                 ip6cp.ip6c_m = m;
1160                 ip6cp.ip6c_icmp6 = icmp6;
1161                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1162                 ip6cp.ip6c_off = eoff;
1163                 ip6cp.ip6c_finaldst = finaldst;
1164                 ip6cp.ip6c_src = &icmp6src;
1165                 ip6cp.ip6c_nxt = nxt;
1166
1167                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1168                         notifymtu = ntohl(icmp6->icmp6_mtu);
1169                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1170                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1171                 }
1172
1173                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1174                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1175                 if (ctlfunc) {
1176                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1177                             &ip6cp);
1178                 }
1179         }
1180         *mp = m;
1181         return (0);
1182
1183   freeit:
1184         m_freem(m);
1185         return (-1);
1186 }
1187
1188 void
1189 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1190 {
1191         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1192         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1193         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1194         u_int mtu = ntohl(icmp6->icmp6_mtu);
1195         struct in_conninfo inc;
1196
1197 #if 0
1198         /*
1199          * RFC2460 section 5, last paragraph.
1200          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1201          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1202          * due to packet translator in the middle.
1203          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1204          * special handling.
1205          */
1206         if (mtu < IPV6_MMTU)
1207                 return;
1208 #endif
1209
1210         /*
1211          * we reject ICMPv6 too big with abnormally small value.
1212          * XXX what is the good definition of "abnormally small"?
1213          */
1214         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1215                 return;
1216
1217         if (!validated)
1218                 return;
1219
1220         /*
1221          * In case the suggested mtu is less than IPV6_MMTU, we
1222          * only need to remember that it was for above mentioned
1223          * "alwaysfrag" case.
1224          * Try to be as close to the spec as possible.
1225          */
1226         if (mtu < IPV6_MMTU)
1227                 mtu = IPV6_MMTU - 8;
1228
1229         bzero(&inc, sizeof(inc));
1230         inc.inc_fibnum = M_GETFIB(m);
1231         inc.inc_flags |= INC_ISIPV6;
1232         inc.inc6_faddr = *dst;
1233         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1234                 return;
1235
1236         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1237                 tcp_hc_updatemtu(&inc, mtu);
1238                 ICMP6STAT_INC(icp6s_pmtuchg);
1239         }
1240 }
1241
1242 /*
1243  * Process a Node Information Query packet, based on
1244  * draft-ietf-ipngwg-icmp-name-lookups-07.
1245  *
1246  * Spec incompatibilities:
1247  * - IPv6 Subject address handling
1248  * - IPv4 Subject address handling support missing
1249  * - Proxy reply (answer even if it's not for me)
1250  * - joins NI group address at in6_ifattach() time only, does not cope
1251  *   with hostname changes by sethostname(3)
1252  */
1253 static struct mbuf *
1254 ni6_input(struct mbuf *m, int off)
1255 {
1256         struct icmp6_nodeinfo *ni6, *nni6;
1257         struct mbuf *n = NULL;
1258         struct prison *pr;
1259         u_int16_t qtype;
1260         int subjlen;
1261         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1262         struct ni_reply_fqdn *fqdn;
1263         int addrs;              /* for NI_QTYPE_NODEADDR */
1264         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1265         struct in6_addr in6_subj; /* subject address */
1266         struct ip6_hdr *ip6;
1267         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1268         char *subj = NULL;
1269         struct in6_ifaddr *ia6 = NULL;
1270
1271         ip6 = mtod(m, struct ip6_hdr *);
1272 #ifndef PULLDOWN_TEST
1273         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1274 #else
1275         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1276         if (ni6 == NULL) {
1277                 /* m is already reclaimed */
1278                 return (NULL);
1279         }
1280 #endif
1281
1282         /*
1283          * Validate IPv6 source address.
1284          * The default configuration MUST be to refuse answering queries from
1285          * global-scope addresses according to RFC4602.
1286          * Notes:
1287          *  - it's not very clear what "refuse" means; this implementation
1288          *    simply drops it.
1289          *  - it's not very easy to identify global-scope (unicast) addresses
1290          *    since there are many prefixes for them.  It should be safer
1291          *    and in practice sufficient to check "all" but loopback and
1292          *    link-local (note that site-local unicast was deprecated and
1293          *    ULA is defined as global scope-wise)
1294          */
1295         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1296             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1297             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1298                 goto bad;
1299
1300         /*
1301          * Validate IPv6 destination address.
1302          *
1303          * The Responder must discard the Query without further processing
1304          * unless it is one of the Responder's unicast or anycast addresses, or
1305          * a link-local scope multicast address which the Responder has joined.
1306          * [RFC4602, Section 5.]
1307          */
1308         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1309                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1310                         goto bad;
1311                 /* else it's a link-local multicast, fine */
1312         } else {                /* unicast or anycast */
1313                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1314                         goto bad; /* XXX impossible */
1315
1316                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1317                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1318                         ifa_free(&ia6->ia_ifa);
1319                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1320                                 "a temporary address in %s:%d",
1321                                __FILE__, __LINE__));
1322                         goto bad;
1323                 }
1324                 ifa_free(&ia6->ia_ifa);
1325         }
1326
1327         /* validate query Subject field. */
1328         qtype = ntohs(ni6->ni_qtype);
1329         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1330         switch (qtype) {
1331         case NI_QTYPE_NOOP:
1332         case NI_QTYPE_SUPTYPES:
1333                 /* 07 draft */
1334                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1335                         break;
1336                 /* FALLTHROUGH */
1337         case NI_QTYPE_FQDN:
1338         case NI_QTYPE_NODEADDR:
1339         case NI_QTYPE_IPV4ADDR:
1340                 switch (ni6->ni_code) {
1341                 case ICMP6_NI_SUBJ_IPV6:
1342 #if ICMP6_NI_SUBJ_IPV6 != 0
1343                 case 0:
1344 #endif
1345                         /*
1346                          * backward compatibility - try to accept 03 draft
1347                          * format, where no Subject is present.
1348                          */
1349                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1350                             subjlen == 0) {
1351                                 oldfqdn++;
1352                                 break;
1353                         }
1354 #if ICMP6_NI_SUBJ_IPV6 != 0
1355                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1356                                 goto bad;
1357 #endif
1358
1359                         if (subjlen != sizeof(struct in6_addr))
1360                                 goto bad;
1361
1362                         /*
1363                          * Validate Subject address.
1364                          *
1365                          * Not sure what exactly "address belongs to the node"
1366                          * means in the spec, is it just unicast, or what?
1367                          *
1368                          * At this moment we consider Subject address as
1369                          * "belong to the node" if the Subject address equals
1370                          * to the IPv6 destination address; validation for
1371                          * IPv6 destination address should have done enough
1372                          * check for us.
1373                          *
1374                          * We do not do proxy at this moment.
1375                          */
1376                         /* m_pulldown instead of copy? */
1377                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1378                             subjlen, (caddr_t)&in6_subj);
1379                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1380                                 goto bad;
1381
1382                         subj = (char *)&in6_subj;
1383                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1384                                 break;
1385
1386                         /*
1387                          * XXX if we are to allow other cases, we should really
1388                          * be careful about scope here.
1389                          * basically, we should disallow queries toward IPv6
1390                          * destination X with subject Y,
1391                          * if scope(X) > scope(Y).
1392                          * if we allow scope(X) > scope(Y), it will result in
1393                          * information leakage across scope boundary.
1394                          */
1395                         goto bad;
1396
1397                 case ICMP6_NI_SUBJ_FQDN:
1398                         /*
1399                          * Validate Subject name with gethostname(3).
1400                          *
1401                          * The behavior may need some debate, since:
1402                          * - we are not sure if the node has FQDN as
1403                          *   hostname (returned by gethostname(3)).
1404                          * - the code does wildcard match for truncated names.
1405                          *   however, we are not sure if we want to perform
1406                          *   wildcard match, if gethostname(3) side has
1407                          *   truncated hostname.
1408                          */
1409                         pr = curthread->td_ucred->cr_prison;
1410                         mtx_lock(&pr->pr_mtx);
1411                         n = ni6_nametodns(pr->pr_hostname,
1412                             strlen(pr->pr_hostname), 0);
1413                         mtx_unlock(&pr->pr_mtx);
1414                         if (!n || n->m_next || n->m_len == 0)
1415                                 goto bad;
1416                         IP6_EXTHDR_GET(subj, char *, m,
1417                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1418                         if (subj == NULL)
1419                                 goto bad;
1420                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1421                             n->m_len)) {
1422                                 goto bad;
1423                         }
1424                         m_freem(n);
1425                         n = NULL;
1426                         break;
1427
1428                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1429                 default:
1430                         goto bad;
1431                 }
1432                 break;
1433         }
1434
1435         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1436         switch (qtype) {
1437         case NI_QTYPE_FQDN:
1438                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1439                         goto bad;
1440                 break;
1441         case NI_QTYPE_NODEADDR:
1442         case NI_QTYPE_IPV4ADDR:
1443                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1444                         goto bad;
1445                 break;
1446         }
1447
1448         /* guess reply length */
1449         switch (qtype) {
1450         case NI_QTYPE_NOOP:
1451                 break;          /* no reply data */
1452         case NI_QTYPE_SUPTYPES:
1453                 replylen += sizeof(u_int32_t);
1454                 break;
1455         case NI_QTYPE_FQDN:
1456                 /* XXX will append an mbuf */
1457                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1458                 break;
1459         case NI_QTYPE_NODEADDR:
1460                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1461                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1462                     sizeof(u_int32_t))) > MCLBYTES)
1463                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1464                 break;
1465         case NI_QTYPE_IPV4ADDR:
1466                 /* unsupported - should respond with unknown Qtype? */
1467                 break;
1468         default:
1469                 /*
1470                  * XXX: We must return a reply with the ICMP6 code
1471                  * `unknown Qtype' in this case.  However we regard the case
1472                  * as an FQDN query for backward compatibility.
1473                  * Older versions set a random value to this field,
1474                  * so it rarely varies in the defined qtypes.
1475                  * But the mechanism is not reliable...
1476                  * maybe we should obsolete older versions.
1477                  */
1478                 qtype = NI_QTYPE_FQDN;
1479                 /* XXX will append an mbuf */
1480                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1481                 oldfqdn++;
1482                 break;
1483         }
1484
1485         /* Allocate an mbuf to reply. */
1486         if (replylen > MCLBYTES) {
1487                 /*
1488                  * XXX: should we try to allocate more? But MCLBYTES
1489                  * is probably much larger than IPV6_MMTU...
1490                  */
1491                 goto bad;
1492         }
1493         if (replylen > MHLEN)
1494                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1495         else
1496                 n = m_gethdr(M_NOWAIT, m->m_type);
1497         if (n == NULL) {
1498                 m_freem(m);
1499                 return (NULL);
1500         }
1501         m_move_pkthdr(n, m); /* just for recvif and FIB */
1502         n->m_pkthdr.len = n->m_len = replylen;
1503
1504         /* copy mbuf header and IPv6 + Node Information base headers */
1505         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1506         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1507         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1508
1509         /* qtype dependent procedure */
1510         switch (qtype) {
1511         case NI_QTYPE_NOOP:
1512                 nni6->ni_code = ICMP6_NI_SUCCESS;
1513                 nni6->ni_flags = 0;
1514                 break;
1515         case NI_QTYPE_SUPTYPES:
1516         {
1517                 u_int32_t v;
1518                 nni6->ni_code = ICMP6_NI_SUCCESS;
1519                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1520                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1521                 v = (u_int32_t)htonl(0x0000000f);
1522                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1523                 break;
1524         }
1525         case NI_QTYPE_FQDN:
1526                 nni6->ni_code = ICMP6_NI_SUCCESS;
1527                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1528                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1529                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1530                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1531                 /*
1532                  * XXX do we really have FQDN in hostname?
1533                  */
1534                 pr = curthread->td_ucred->cr_prison;
1535                 mtx_lock(&pr->pr_mtx);
1536                 n->m_next = ni6_nametodns(pr->pr_hostname,
1537                     strlen(pr->pr_hostname), oldfqdn);
1538                 mtx_unlock(&pr->pr_mtx);
1539                 if (n->m_next == NULL)
1540                         goto bad;
1541                 /* XXX we assume that n->m_next is not a chain */
1542                 if (n->m_next->m_next != NULL)
1543                         goto bad;
1544                 n->m_pkthdr.len += n->m_next->m_len;
1545                 break;
1546         case NI_QTYPE_NODEADDR:
1547         {
1548                 int lenlim, copied;
1549
1550                 nni6->ni_code = ICMP6_NI_SUCCESS;
1551                 n->m_pkthdr.len = n->m_len =
1552                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1553                 lenlim = M_TRAILINGSPACE(n);
1554                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1555                 /* XXX: reset mbuf length */
1556                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1557                     sizeof(struct icmp6_nodeinfo) + copied;
1558                 break;
1559         }
1560         default:
1561                 break;          /* XXX impossible! */
1562         }
1563
1564         nni6->ni_type = ICMP6_NI_REPLY;
1565         m_freem(m);
1566         return (n);
1567
1568   bad:
1569         m_freem(m);
1570         if (n)
1571                 m_freem(n);
1572         return (NULL);
1573 }
1574
1575 /*
1576  * make a mbuf with DNS-encoded string.  no compression support.
1577  *
1578  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1579  * treated as truncated name (two \0 at the end).  this is a wild guess.
1580  *
1581  * old - return pascal string if non-zero
1582  */
1583 static struct mbuf *
1584 ni6_nametodns(const char *name, int namelen, int old)
1585 {
1586         struct mbuf *m;
1587         char *cp, *ep;
1588         const char *p, *q;
1589         int i, len, nterm;
1590
1591         if (old)
1592                 len = namelen + 1;
1593         else
1594                 len = MCLBYTES;
1595
1596         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1597         if (len > MLEN)
1598                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1599         else
1600                 m = m_get(M_NOWAIT, MT_DATA);
1601         if (m == NULL)
1602                 goto fail;
1603
1604         if (old) {
1605                 m->m_len = len;
1606                 *mtod(m, char *) = namelen;
1607                 bcopy(name, mtod(m, char *) + 1, namelen);
1608                 return m;
1609         } else {
1610                 m->m_len = 0;
1611                 cp = mtod(m, char *);
1612                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1613
1614                 /* if not certain about my name, return empty buffer */
1615                 if (namelen == 0)
1616                         return m;
1617
1618                 /*
1619                  * guess if it looks like shortened hostname, or FQDN.
1620                  * shortened hostname needs two trailing "\0".
1621                  */
1622                 i = 0;
1623                 for (p = name; p < name + namelen; p++) {
1624                         if (*p && *p == '.')
1625                                 i++;
1626                 }
1627                 if (i < 2)
1628                         nterm = 2;
1629                 else
1630                         nterm = 1;
1631
1632                 p = name;
1633                 while (cp < ep && p < name + namelen) {
1634                         i = 0;
1635                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1636                                 i++;
1637                         /* result does not fit into mbuf */
1638                         if (cp + i + 1 >= ep)
1639                                 goto fail;
1640                         /*
1641                          * DNS label length restriction, RFC1035 page 8.
1642                          * "i == 0" case is included here to avoid returning
1643                          * 0-length label on "foo..bar".
1644                          */
1645                         if (i <= 0 || i >= 64)
1646                                 goto fail;
1647                         *cp++ = i;
1648                         bcopy(p, cp, i);
1649                         cp += i;
1650                         p = q;
1651                         if (p < name + namelen && *p == '.')
1652                                 p++;
1653                 }
1654                 /* termination */
1655                 if (cp + nterm >= ep)
1656                         goto fail;
1657                 while (nterm-- > 0)
1658                         *cp++ = '\0';
1659                 m->m_len = cp - mtod(m, char *);
1660                 return m;
1661         }
1662
1663         panic("should not reach here");
1664         /* NOTREACHED */
1665
1666  fail:
1667         if (m)
1668                 m_freem(m);
1669         return NULL;
1670 }
1671
1672 /*
1673  * check if two DNS-encoded string matches.  takes care of truncated
1674  * form (with \0\0 at the end).  no compression support.
1675  * XXX upper/lowercase match (see RFC2065)
1676  */
1677 static int
1678 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1679 {
1680         const char *a0, *b0;
1681         int l;
1682
1683         /* simplest case - need validation? */
1684         if (alen == blen && bcmp(a, b, alen) == 0)
1685                 return 1;
1686
1687         a0 = a;
1688         b0 = b;
1689
1690         /* termination is mandatory */
1691         if (alen < 2 || blen < 2)
1692                 return 0;
1693         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1694                 return 0;
1695         alen--;
1696         blen--;
1697
1698         while (a - a0 < alen && b - b0 < blen) {
1699                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1700                         return 0;
1701
1702                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1703                         return 0;
1704                 /* we don't support compression yet */
1705                 if (a[0] >= 64 || b[0] >= 64)
1706                         return 0;
1707
1708                 /* truncated case */
1709                 if (a[0] == 0 && a - a0 == alen - 1)
1710                         return 1;
1711                 if (b[0] == 0 && b - b0 == blen - 1)
1712                         return 1;
1713                 if (a[0] == 0 || b[0] == 0)
1714                         return 0;
1715
1716                 if (a[0] != b[0])
1717                         return 0;
1718                 l = a[0];
1719                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1720                         return 0;
1721                 if (bcmp(a + 1, b + 1, l) != 0)
1722                         return 0;
1723
1724                 a += 1 + l;
1725                 b += 1 + l;
1726         }
1727
1728         if (a - a0 == alen && b - b0 == blen)
1729                 return 1;
1730         else
1731                 return 0;
1732 }
1733
1734 /*
1735  * calculate the number of addresses to be returned in the node info reply.
1736  */
1737 static int
1738 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1739     struct in6_addr *subj)
1740 {
1741         struct ifnet *ifp;
1742         struct in6_ifaddr *ifa6;
1743         struct ifaddr *ifa;
1744         int addrs = 0, addrsofif, iffound = 0;
1745         int niflags = ni6->ni_flags;
1746
1747         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1748                 switch (ni6->ni_code) {
1749                 case ICMP6_NI_SUBJ_IPV6:
1750                         if (subj == NULL) /* must be impossible... */
1751                                 return (0);
1752                         break;
1753                 default:
1754                         /*
1755                          * XXX: we only support IPv6 subject address for
1756                          * this Qtype.
1757                          */
1758                         return (0);
1759                 }
1760         }
1761
1762         IFNET_RLOCK_NOSLEEP();
1763         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1764                 addrsofif = 0;
1765                 IF_ADDR_RLOCK(ifp);
1766                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1767                         if (ifa->ifa_addr->sa_family != AF_INET6)
1768                                 continue;
1769                         ifa6 = (struct in6_ifaddr *)ifa;
1770
1771                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1772                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1773                                 iffound = 1;
1774
1775                         /*
1776                          * IPv4-mapped addresses can only be returned by a
1777                          * Node Information proxy, since they represent
1778                          * addresses of IPv4-only nodes, which perforce do
1779                          * not implement this protocol.
1780                          * [icmp-name-lookups-07, Section 5.4]
1781                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1782                          * this function at this moment.
1783                          */
1784
1785                         /* What do we have to do about ::1? */
1786                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1787                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1788                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1789                                         continue;
1790                                 break;
1791                         case IPV6_ADDR_SCOPE_SITELOCAL:
1792                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1793                                         continue;
1794                                 break;
1795                         case IPV6_ADDR_SCOPE_GLOBAL:
1796                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1797                                         continue;
1798                                 break;
1799                         default:
1800                                 continue;
1801                         }
1802
1803                         /*
1804                          * check if anycast is okay.
1805                          * XXX: just experimental.  not in the spec.
1806                          */
1807                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1808                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1809                                 continue; /* we need only unicast addresses */
1810                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1811                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1812                                 continue;
1813                         }
1814                         addrsofif++; /* count the address */
1815                 }
1816                 IF_ADDR_RUNLOCK(ifp);
1817                 if (iffound) {
1818                         *ifpp = ifp;
1819                         IFNET_RUNLOCK_NOSLEEP();
1820                         return (addrsofif);
1821                 }
1822
1823                 addrs += addrsofif;
1824         }
1825         IFNET_RUNLOCK_NOSLEEP();
1826
1827         return (addrs);
1828 }
1829
1830 static int
1831 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1832     struct ifnet *ifp0, int resid)
1833 {
1834         struct ifnet *ifp;
1835         struct in6_ifaddr *ifa6;
1836         struct ifaddr *ifa;
1837         struct ifnet *ifp_dep = NULL;
1838         int copied = 0, allow_deprecated = 0;
1839         u_char *cp = (u_char *)(nni6 + 1);
1840         int niflags = ni6->ni_flags;
1841         u_int32_t ltime;
1842
1843         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1844                 return (0);     /* needless to copy */
1845
1846         IFNET_RLOCK_NOSLEEP();
1847         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1848   again:
1849
1850         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1851                 IF_ADDR_RLOCK(ifp);
1852                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1853                         if (ifa->ifa_addr->sa_family != AF_INET6)
1854                                 continue;
1855                         ifa6 = (struct in6_ifaddr *)ifa;
1856
1857                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1858                             allow_deprecated == 0) {
1859                                 /*
1860                                  * prefererred address should be put before
1861                                  * deprecated addresses.
1862                                  */
1863
1864                                 /* record the interface for later search */
1865                                 if (ifp_dep == NULL)
1866                                         ifp_dep = ifp;
1867
1868                                 continue;
1869                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1870                             allow_deprecated != 0)
1871                                 continue; /* we now collect deprecated addrs */
1872
1873                         /* What do we have to do about ::1? */
1874                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1875                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1876                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1877                                         continue;
1878                                 break;
1879                         case IPV6_ADDR_SCOPE_SITELOCAL:
1880                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1881                                         continue;
1882                                 break;
1883                         case IPV6_ADDR_SCOPE_GLOBAL:
1884                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1885                                         continue;
1886                                 break;
1887                         default:
1888                                 continue;
1889                         }
1890
1891                         /*
1892                          * check if anycast is okay.
1893                          * XXX: just experimental.  not in the spec.
1894                          */
1895                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1896                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1897                                 continue;
1898                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1899                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1900                                 continue;
1901                         }
1902
1903                         /* now we can copy the address */
1904                         if (resid < sizeof(struct in6_addr) +
1905                             sizeof(u_int32_t)) {
1906                                 IF_ADDR_RUNLOCK(ifp);
1907                                 /*
1908                                  * We give up much more copy.
1909                                  * Set the truncate flag and return.
1910                                  */
1911                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1912                                 IFNET_RUNLOCK_NOSLEEP();
1913                                 return (copied);
1914                         }
1915
1916                         /*
1917                          * Set the TTL of the address.
1918                          * The TTL value should be one of the following
1919                          * according to the specification:
1920                          *
1921                          * 1. The remaining lifetime of a DHCP lease on the
1922                          *    address, or
1923                          * 2. The remaining Valid Lifetime of a prefix from
1924                          *    which the address was derived through Stateless
1925                          *    Autoconfiguration.
1926                          *
1927                          * Note that we currently do not support stateful
1928                          * address configuration by DHCPv6, so the former
1929                          * case can't happen.
1930                          */
1931                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1932                                 ltime = ND6_INFINITE_LIFETIME;
1933                         else {
1934                                 if (ifa6->ia6_lifetime.ia6t_expire >
1935                                     time_uptime)
1936                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1937                                 else
1938                                         ltime = 0;
1939                         }
1940
1941                         bcopy(&ltime, cp, sizeof(u_int32_t));
1942                         cp += sizeof(u_int32_t);
1943
1944                         /* copy the address itself */
1945                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1946                             sizeof(struct in6_addr));
1947                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1948                         cp += sizeof(struct in6_addr);
1949
1950                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1951                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1952                 }
1953                 IF_ADDR_RUNLOCK(ifp);
1954                 if (ifp0)       /* we need search only on the specified IF */
1955                         break;
1956         }
1957
1958         if (allow_deprecated == 0 && ifp_dep != NULL) {
1959                 ifp = ifp_dep;
1960                 allow_deprecated = 1;
1961
1962                 goto again;
1963         }
1964
1965         IFNET_RUNLOCK_NOSLEEP();
1966
1967         return (copied);
1968 }
1969
1970 /*
1971  * XXX almost dup'ed code with rip6_input.
1972  */
1973 static int
1974 icmp6_rip6_input(struct mbuf **mp, int off)
1975 {
1976         struct mbuf *m = *mp;
1977         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1978         struct inpcb *in6p;
1979         struct inpcb *last = NULL;
1980         struct sockaddr_in6 fromsa;
1981         struct icmp6_hdr *icmp6;
1982         struct mbuf *opts = NULL;
1983
1984 #ifndef PULLDOWN_TEST
1985         /* this is assumed to be safe. */
1986         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1987 #else
1988         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1989         if (icmp6 == NULL) {
1990                 /* m is already reclaimed */
1991                 return (IPPROTO_DONE);
1992         }
1993 #endif
1994
1995         /*
1996          * XXX: the address may have embedded scope zone ID, which should be
1997          * hidden from applications.
1998          */
1999         bzero(&fromsa, sizeof(fromsa));
2000         fromsa.sin6_family = AF_INET6;
2001         fromsa.sin6_len = sizeof(struct sockaddr_in6);
2002         fromsa.sin6_addr = ip6->ip6_src;
2003         if (sa6_recoverscope(&fromsa)) {
2004                 m_freem(m);
2005                 return (IPPROTO_DONE);
2006         }
2007
2008         INP_INFO_RLOCK(&V_ripcbinfo);
2009         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2010                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2011                         continue;
2012                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2013                         continue;
2014                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2015                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2016                         continue;
2017                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2018                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2019                         continue;
2020                 INP_RLOCK(in6p);
2021                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2022                     in6p->in6p_icmp6filt)) {
2023                         INP_RUNLOCK(in6p);
2024                         continue;
2025                 }
2026                 if (last != NULL) {
2027                         struct  mbuf *n = NULL;
2028
2029                         /*
2030                          * Recent network drivers tend to allocate a single
2031                          * mbuf cluster, rather than to make a couple of
2032                          * mbufs without clusters.  Also, since the IPv6 code
2033                          * path tries to avoid m_pullup(), it is highly
2034                          * probable that we still have an mbuf cluster here
2035                          * even though the necessary length can be stored in an
2036                          * mbuf's internal buffer.
2037                          * Meanwhile, the default size of the receive socket
2038                          * buffer for raw sockets is not so large.  This means
2039                          * the possibility of packet loss is relatively higher
2040                          * than before.  To avoid this scenario, we copy the
2041                          * received data to a separate mbuf that does not use
2042                          * a cluster, if possible.
2043                          * XXX: it is better to copy the data after stripping
2044                          * intermediate headers.
2045                          */
2046                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2047                             m->m_len <= MHLEN) {
2048                                 n = m_get(M_NOWAIT, m->m_type);
2049                                 if (n != NULL) {
2050                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2051                                                 bcopy(m->m_data, n->m_data,
2052                                                       m->m_len);
2053                                                 n->m_len = m->m_len;
2054                                         } else {
2055                                                 m_free(n);
2056                                                 n = NULL;
2057                                         }
2058                                 }
2059                         }
2060                         if (n != NULL ||
2061                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2062                                 if (last->inp_flags & INP_CONTROLOPTS)
2063                                         ip6_savecontrol(last, n, &opts);
2064                                 /* strip intermediate headers */
2065                                 m_adj(n, off);
2066                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2067                                 if (sbappendaddr_locked(
2068                                     &last->inp_socket->so_rcv,
2069                                     (struct sockaddr *)&fromsa, n, opts)
2070                                     == 0) {
2071                                         /* should notify about lost packet */
2072                                         m_freem(n);
2073                                         if (opts) {
2074                                                 m_freem(opts);
2075                                         }
2076                                         SOCKBUF_UNLOCK(
2077                                             &last->inp_socket->so_rcv);
2078                                 } else
2079                                         sorwakeup_locked(last->inp_socket);
2080                                 opts = NULL;
2081                         }
2082                         INP_RUNLOCK(last);
2083                 }
2084                 last = in6p;
2085         }
2086         INP_INFO_RUNLOCK(&V_ripcbinfo);
2087         if (last != NULL) {
2088                 if (last->inp_flags & INP_CONTROLOPTS)
2089                         ip6_savecontrol(last, m, &opts);
2090                 /* strip intermediate headers */
2091                 m_adj(m, off);
2092
2093                 /* avoid using mbuf clusters if possible (see above) */
2094                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2095                     m->m_len <= MHLEN) {
2096                         struct mbuf *n;
2097
2098                         n = m_get(M_NOWAIT, m->m_type);
2099                         if (n != NULL) {
2100                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2101                                         bcopy(m->m_data, n->m_data, m->m_len);
2102                                         n->m_len = m->m_len;
2103
2104                                         m_freem(m);
2105                                         m = n;
2106                                 } else {
2107                                         m_freem(n);
2108                                         n = NULL;
2109                                 }
2110                         }
2111                 }
2112                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2113                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2114                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2115                         m_freem(m);
2116                         if (opts)
2117                                 m_freem(opts);
2118                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2119                 } else
2120                         sorwakeup_locked(last->inp_socket);
2121                 INP_RUNLOCK(last);
2122         } else {
2123                 m_freem(m);
2124                 IP6STAT_DEC(ip6s_delivered);
2125         }
2126         return IPPROTO_DONE;
2127 }
2128
2129 /*
2130  * Reflect the ip6 packet back to the source.
2131  * OFF points to the icmp6 header, counted from the top of the mbuf.
2132  */
2133 void
2134 icmp6_reflect(struct mbuf *m, size_t off)
2135 {
2136         struct ip6_hdr *ip6;
2137         struct icmp6_hdr *icmp6;
2138         struct in6_ifaddr *ia = NULL;
2139         int plen;
2140         int type, code;
2141         struct ifnet *outif = NULL;
2142         struct in6_addr origdst, src, *srcp = NULL;
2143
2144         /* too short to reflect */
2145         if (off < sizeof(struct ip6_hdr)) {
2146                 nd6log((LOG_DEBUG,
2147                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2148                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2149                     __FILE__, __LINE__));
2150                 goto bad;
2151         }
2152
2153         /*
2154          * If there are extra headers between IPv6 and ICMPv6, strip
2155          * off that header first.
2156          */
2157 #ifdef DIAGNOSTIC
2158         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2159                 panic("assumption failed in icmp6_reflect");
2160 #endif
2161         if (off > sizeof(struct ip6_hdr)) {
2162                 size_t l;
2163                 struct ip6_hdr nip6;
2164
2165                 l = off - sizeof(struct ip6_hdr);
2166                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2167                 m_adj(m, l);
2168                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2169                 if (m->m_len < l) {
2170                         if ((m = m_pullup(m, l)) == NULL)
2171                                 return;
2172                 }
2173                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2174         } else /* off == sizeof(struct ip6_hdr) */ {
2175                 size_t l;
2176                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2177                 if (m->m_len < l) {
2178                         if ((m = m_pullup(m, l)) == NULL)
2179                                 return;
2180                 }
2181         }
2182         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2183         ip6 = mtod(m, struct ip6_hdr *);
2184         ip6->ip6_nxt = IPPROTO_ICMPV6;
2185         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2186         type = icmp6->icmp6_type; /* keep type for statistics */
2187         code = icmp6->icmp6_code; /* ditto. */
2188
2189         origdst = ip6->ip6_dst;
2190         /*
2191          * ip6_input() drops a packet if its src is multicast.
2192          * So, the src is never multicast.
2193          */
2194         ip6->ip6_dst = ip6->ip6_src;
2195
2196         /*
2197          * If the incoming packet was addressed directly to us (i.e. unicast),
2198          * use dst as the src for the reply.
2199          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2200          * (for example) when we encounter an error while forwarding procedure
2201          * destined to a duplicated address of ours.
2202          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2203          * procedure of an outgoing packet of our own, in which case we need
2204          * to search in the ifaddr list.
2205          */
2206         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2207                 if ((ia = ip6_getdstifaddr(m))) {
2208                         if (!(ia->ia6_flags &
2209                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2210                                 srcp = &ia->ia_addr.sin6_addr;
2211                 } else {
2212                         struct sockaddr_in6 d;
2213
2214                         bzero(&d, sizeof(d));
2215                         d.sin6_family = AF_INET6;
2216                         d.sin6_len = sizeof(d);
2217                         d.sin6_addr = origdst;
2218                         ia = (struct in6_ifaddr *)
2219                             ifa_ifwithaddr((struct sockaddr *)&d);
2220                         if (ia &&
2221                             !(ia->ia6_flags &
2222                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2223                                 srcp = &ia->ia_addr.sin6_addr;
2224                         }
2225                 }
2226         }
2227
2228         if (srcp == NULL) {
2229                 int e;
2230                 struct sockaddr_in6 sin6;
2231                 struct route_in6 ro;
2232
2233                 /*
2234                  * This case matches to multicasts, our anycast, or unicasts
2235                  * that we do not own.  Select a source address based on the
2236                  * source address of the erroneous packet.
2237                  */
2238                 bzero(&sin6, sizeof(sin6));
2239                 sin6.sin6_family = AF_INET6;
2240                 sin6.sin6_len = sizeof(sin6);
2241                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2242
2243                 bzero(&ro, sizeof(ro));
2244                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2245                 if (ro.ro_rt)
2246                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2247                 if (e) {
2248                         char ip6buf[INET6_ADDRSTRLEN];
2249                         nd6log((LOG_DEBUG,
2250                             "icmp6_reflect: source can't be determined: "
2251                             "dst=%s, error=%d\n",
2252                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2253                         goto bad;
2254                 }
2255                 srcp = &src;
2256         }
2257
2258         ip6->ip6_src = *srcp;
2259         ip6->ip6_flow = 0;
2260         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2261         ip6->ip6_vfc |= IPV6_VERSION;
2262         ip6->ip6_nxt = IPPROTO_ICMPV6;
2263         if (outif)
2264                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2265         else if (m->m_pkthdr.rcvif) {
2266                 /* XXX: This may not be the outgoing interface */
2267                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2268         } else
2269                 ip6->ip6_hlim = V_ip6_defhlim;
2270
2271         icmp6->icmp6_cksum = 0;
2272         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2273             sizeof(struct ip6_hdr), plen);
2274
2275         /*
2276          * XXX option handling
2277          */
2278
2279         m->m_flags &= ~(M_BCAST|M_MCAST);
2280
2281         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2282         if (outif)
2283                 icmp6_ifoutstat_inc(outif, type, code);
2284
2285         if (ia != NULL)
2286                 ifa_free(&ia->ia_ifa);
2287         return;
2288
2289  bad:
2290         if (ia != NULL)
2291                 ifa_free(&ia->ia_ifa);
2292         m_freem(m);
2293         return;
2294 }
2295
2296 void
2297 icmp6_fasttimo(void)
2298 {
2299
2300         mld_fasttimo();
2301 }
2302
2303 void
2304 icmp6_slowtimo(void)
2305 {
2306
2307         mld_slowtimo();
2308 }
2309
2310 static const char *
2311 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2312     struct in6_addr *tgt6)
2313 {
2314         static char buf[1024];
2315         char ip6bufs[INET6_ADDRSTRLEN];
2316         char ip6bufd[INET6_ADDRSTRLEN];
2317         char ip6buft[INET6_ADDRSTRLEN];
2318         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2319             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2320             ip6_sprintf(ip6buft, tgt6));
2321         return buf;
2322 }
2323
2324 void
2325 icmp6_redirect_input(struct mbuf *m, int off)
2326 {
2327         struct ifnet *ifp;
2328         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2329         struct nd_redirect *nd_rd;
2330         int icmp6len = ntohs(ip6->ip6_plen);
2331         char *lladdr = NULL;
2332         int lladdrlen = 0;
2333         struct rtentry *rt = NULL;
2334         int is_router;
2335         int is_onlink;
2336         struct in6_addr src6 = ip6->ip6_src;
2337         struct in6_addr redtgt6;
2338         struct in6_addr reddst6;
2339         union nd_opts ndopts;
2340         char ip6buf[INET6_ADDRSTRLEN];
2341
2342         M_ASSERTPKTHDR(m);
2343         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2344
2345         ifp = m->m_pkthdr.rcvif;
2346
2347         /* XXX if we are router, we don't update route by icmp6 redirect */
2348         if (V_ip6_forwarding)
2349                 goto freeit;
2350         if (!V_icmp6_rediraccept)
2351                 goto freeit;
2352
2353 #ifndef PULLDOWN_TEST
2354         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2355         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2356 #else
2357         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2358         if (nd_rd == NULL) {
2359                 ICMP6STAT_INC(icp6s_tooshort);
2360                 return;
2361         }
2362 #endif
2363         redtgt6 = nd_rd->nd_rd_target;
2364         reddst6 = nd_rd->nd_rd_dst;
2365
2366         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2367             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2368                 goto freeit;
2369         }
2370
2371         /* validation */
2372         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2373                 nd6log((LOG_ERR,
2374                     "ICMP6 redirect sent from %s rejected; "
2375                     "must be from linklocal\n",
2376                     ip6_sprintf(ip6buf, &src6)));
2377                 goto bad;
2378         }
2379         if (ip6->ip6_hlim != 255) {
2380                 nd6log((LOG_ERR,
2381                     "ICMP6 redirect sent from %s rejected; "
2382                     "hlim=%d (must be 255)\n",
2383                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2384                 goto bad;
2385         }
2386     {
2387         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2388         struct sockaddr_in6 sin6;
2389         struct in6_addr *gw6;
2390
2391         bzero(&sin6, sizeof(sin6));
2392         sin6.sin6_family = AF_INET6;
2393         sin6.sin6_len = sizeof(struct sockaddr_in6);
2394         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2395         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2396         if (rt) {
2397                 if (rt->rt_gateway == NULL ||
2398                     rt->rt_gateway->sa_family != AF_INET6) {
2399                         RTFREE_LOCKED(rt);
2400                         nd6log((LOG_ERR,
2401                             "ICMP6 redirect rejected; no route "
2402                             "with inet6 gateway found for redirect dst: %s\n",
2403                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2404                         goto bad;
2405                 }
2406
2407                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2408                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2409                         RTFREE_LOCKED(rt);
2410                         nd6log((LOG_ERR,
2411                             "ICMP6 redirect rejected; "
2412                             "not equal to gw-for-src=%s (must be same): "
2413                             "%s\n",
2414                             ip6_sprintf(ip6buf, gw6),
2415                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2416                         goto bad;
2417                 }
2418         } else {
2419                 nd6log((LOG_ERR,
2420                     "ICMP6 redirect rejected; "
2421                     "no route found for redirect dst: %s\n",
2422                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2423                 goto bad;
2424         }
2425         RTFREE_LOCKED(rt);
2426         rt = NULL;
2427     }
2428         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2429                 nd6log((LOG_ERR,
2430                     "ICMP6 redirect rejected; "
2431                     "redirect dst must be unicast: %s\n",
2432                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2433                 goto bad;
2434         }
2435
2436         is_router = is_onlink = 0;
2437         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2438                 is_router = 1;  /* router case */
2439         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2440                 is_onlink = 1;  /* on-link destination case */
2441         if (!is_router && !is_onlink) {
2442                 nd6log((LOG_ERR,
2443                     "ICMP6 redirect rejected; "
2444                     "neither router case nor onlink case: %s\n",
2445                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2446                 goto bad;
2447         }
2448         /* validation passed */
2449
2450         icmp6len -= sizeof(*nd_rd);
2451         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2452         if (nd6_options(&ndopts) < 0) {
2453                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2454                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2455                 /* nd6_options have incremented stats */
2456                 goto freeit;
2457         }
2458
2459         if (ndopts.nd_opts_tgt_lladdr) {
2460                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2461                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2462         }
2463
2464         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2465                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2466                     "(if %d, icmp6 packet %d): %s\n",
2467                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2468                     ifp->if_addrlen, lladdrlen - 2,
2469                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2470                 goto bad;
2471         }
2472
2473         /* RFC 2461 8.3 */
2474         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2475             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2476
2477         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2478                 /* perform rtredirect */
2479                 struct sockaddr_in6 sdst;
2480                 struct sockaddr_in6 sgw;
2481                 struct sockaddr_in6 ssrc;
2482                 u_int fibnum;
2483
2484                 bzero(&sdst, sizeof(sdst));
2485                 bzero(&sgw, sizeof(sgw));
2486                 bzero(&ssrc, sizeof(ssrc));
2487                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2488                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2489                         sizeof(struct sockaddr_in6);
2490                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2491                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2492                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2493                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2494                         in6_rtredirect((struct sockaddr *)&sdst,
2495                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2496                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2497                             fibnum);
2498         }
2499         /* finally update cached route in each socket via pfctlinput */
2500     {
2501         struct sockaddr_in6 sdst;
2502
2503         bzero(&sdst, sizeof(sdst));
2504         sdst.sin6_family = AF_INET6;
2505         sdst.sin6_len = sizeof(struct sockaddr_in6);
2506         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2507         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2508 #ifdef IPSEC
2509         key_sa_routechange((struct sockaddr *)&sdst);
2510 #endif /* IPSEC */
2511     }
2512
2513  freeit:
2514         m_freem(m);
2515         return;
2516
2517  bad:
2518         ICMP6STAT_INC(icp6s_badredirect);
2519         m_freem(m);
2520 }
2521
2522 void
2523 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2524 {
2525         struct ifnet *ifp;      /* my outgoing interface */
2526         struct in6_addr *ifp_ll6;
2527         struct in6_addr *router_ll6;
2528         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2529         struct mbuf *m = NULL;  /* newly allocated one */
2530         struct m_tag *mtag;
2531         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2532         struct nd_redirect *nd_rd;
2533         struct llentry *ln = NULL;
2534         size_t maxlen;
2535         u_char *p;
2536         struct ifnet *outif = NULL;
2537         struct sockaddr_in6 src_sa;
2538
2539         icmp6_errcount(ND_REDIRECT, 0);
2540
2541         /* if we are not router, we don't send icmp6 redirect */
2542         if (!V_ip6_forwarding)
2543                 goto fail;
2544
2545         /* sanity check */
2546         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2547                 goto fail;
2548
2549         /*
2550          * Address check:
2551          *  the source address must identify a neighbor, and
2552          *  the destination address must not be a multicast address
2553          *  [RFC 2461, sec 8.2]
2554          */
2555         sip6 = mtod(m0, struct ip6_hdr *);
2556         bzero(&src_sa, sizeof(src_sa));
2557         src_sa.sin6_family = AF_INET6;
2558         src_sa.sin6_len = sizeof(src_sa);
2559         src_sa.sin6_addr = sip6->ip6_src;
2560         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2561                 goto fail;
2562         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2563                 goto fail;      /* what should we do here? */
2564
2565         /* rate limit */
2566         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2567                 goto fail;
2568
2569         /*
2570          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2571          * we almost always ask for an mbuf cluster for simplicity.
2572          * (MHLEN < IPV6_MMTU is almost always true)
2573          */
2574 #if IPV6_MMTU >= MCLBYTES
2575 # error assumption failed about IPV6_MMTU and MCLBYTES
2576 #endif
2577         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2578         if (m == NULL)
2579                 goto fail;
2580         M_SETFIB(m, rt->rt_fibnum);
2581         maxlen = M_TRAILINGSPACE(m);
2582         maxlen = min(IPV6_MMTU, maxlen);
2583         /* just for safety */
2584         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2585             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2586                 goto fail;
2587         }
2588
2589         {
2590                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2591                 struct in6_ifaddr *ia;
2592                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2593                                                  IN6_IFF_NOTREADY|
2594                                                  IN6_IFF_ANYCAST)) == NULL)
2595                         goto fail;
2596                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2597                 /* XXXRW: reference released prematurely. */
2598                 ifa_free(&ia->ia_ifa);
2599         }
2600
2601         /* get ip6 linklocal address for the router. */
2602         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2603                 struct sockaddr_in6 *sin6;
2604                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2605                 router_ll6 = &sin6->sin6_addr;
2606                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2607                         router_ll6 = (struct in6_addr *)NULL;
2608         } else
2609                 router_ll6 = (struct in6_addr *)NULL;
2610
2611         /* ip6 */
2612         ip6 = mtod(m, struct ip6_hdr *);
2613         ip6->ip6_flow = 0;
2614         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2615         ip6->ip6_vfc |= IPV6_VERSION;
2616         /* ip6->ip6_plen will be set later */
2617         ip6->ip6_nxt = IPPROTO_ICMPV6;
2618         ip6->ip6_hlim = 255;
2619         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2620         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2621         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2622
2623         /* ND Redirect */
2624         nd_rd = (struct nd_redirect *)(ip6 + 1);
2625         nd_rd->nd_rd_type = ND_REDIRECT;
2626         nd_rd->nd_rd_code = 0;
2627         nd_rd->nd_rd_reserved = 0;
2628         if (rt->rt_flags & RTF_GATEWAY) {
2629                 /*
2630                  * nd_rd->nd_rd_target must be a link-local address in
2631                  * better router cases.
2632                  */
2633                 if (!router_ll6)
2634                         goto fail;
2635                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2636                     sizeof(nd_rd->nd_rd_target));
2637                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2638                     sizeof(nd_rd->nd_rd_dst));
2639         } else {
2640                 /* make sure redtgt == reddst */
2641                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2642                     sizeof(nd_rd->nd_rd_target));
2643                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2644                     sizeof(nd_rd->nd_rd_dst));
2645         }
2646
2647         p = (u_char *)(nd_rd + 1);
2648
2649         if (!router_ll6)
2650                 goto nolladdropt;
2651
2652         {
2653                 /* target lladdr option */
2654                 int len;
2655                 struct nd_opt_hdr *nd_opt;
2656                 char *lladdr;
2657
2658                 IF_AFDATA_RLOCK(ifp);
2659                 ln = nd6_lookup(router_ll6, 0, ifp);
2660                 IF_AFDATA_RUNLOCK(ifp);
2661                 if (ln == NULL)
2662                         goto nolladdropt;
2663
2664                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2665                 len = (len + 7) & ~7;   /* round by 8 */
2666                 /* safety check */
2667                 if (len + (p - (u_char *)ip6) > maxlen)                         
2668                         goto nolladdropt;
2669
2670                 if (ln->la_flags & LLE_VALID) {
2671                         nd_opt = (struct nd_opt_hdr *)p;
2672                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2673                         nd_opt->nd_opt_len = len >> 3;
2674                         lladdr = (char *)(nd_opt + 1);
2675                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2676                         p += len;
2677                 }
2678         }
2679 nolladdropt:
2680         if (ln != NULL)
2681                 LLE_RUNLOCK(ln);
2682                 
2683         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2684
2685         /* just to be safe */
2686 #ifdef M_DECRYPTED      /*not openbsd*/
2687         if (m0->m_flags & M_DECRYPTED)
2688                 goto noredhdropt;
2689 #endif
2690         if (p - (u_char *)ip6 > maxlen)
2691                 goto noredhdropt;
2692
2693         {
2694                 /* redirected header option */
2695                 int len;
2696                 struct nd_opt_rd_hdr *nd_opt_rh;
2697
2698                 /*
2699                  * compute the maximum size for icmp6 redirect header option.
2700                  * XXX room for auth header?
2701                  */
2702                 len = maxlen - (p - (u_char *)ip6);
2703                 len &= ~7;
2704
2705                 /* This is just for simplicity. */
2706                 if (m0->m_pkthdr.len != m0->m_len) {
2707                         if (m0->m_next) {
2708                                 m_freem(m0->m_next);
2709                                 m0->m_next = NULL;
2710                         }
2711                         m0->m_pkthdr.len = m0->m_len;
2712                 }
2713
2714                 /*
2715                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2716                  * about padding/truncate rule for the original IP packet.
2717                  * From the discussion on IPv6imp in Feb 1999,
2718                  * the consensus was:
2719                  * - "attach as much as possible" is the goal
2720                  * - pad if not aligned (original size can be guessed by
2721                  *   original ip6 header)
2722                  * Following code adds the padding if it is simple enough,
2723                  * and truncates if not.
2724                  */
2725                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2726                         panic("assumption failed in %s:%d", __FILE__,
2727                             __LINE__);
2728
2729                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2730                         /* not enough room, truncate */
2731                         m0->m_pkthdr.len = m0->m_len = len -
2732                             sizeof(*nd_opt_rh);
2733                 } else {
2734                         /* enough room, pad or truncate */
2735                         size_t extra;
2736
2737                         extra = m0->m_pkthdr.len % 8;
2738                         if (extra) {
2739                                 /* pad if easy enough, truncate if not */
2740                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2741                                         /* pad */
2742                                         m0->m_len += (8 - extra);
2743                                         m0->m_pkthdr.len += (8 - extra);
2744                                 } else {
2745                                         /* truncate */
2746                                         m0->m_pkthdr.len -= extra;
2747                                         m0->m_len -= extra;
2748                                 }
2749                         }
2750                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2751                         m0->m_pkthdr.len = m0->m_len = len -
2752                             sizeof(*nd_opt_rh);
2753                 }
2754
2755                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2756                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2757                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2758                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2759                 p += sizeof(*nd_opt_rh);
2760                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2761
2762                 /* connect m0 to m */
2763                 m_tag_delete_chain(m0, NULL);
2764                 m0->m_flags &= ~M_PKTHDR;
2765                 m->m_next = m0;
2766                 m->m_pkthdr.len = m->m_len + m0->m_len;
2767                 m0 = NULL;
2768         }
2769 noredhdropt:;
2770         if (m0) {
2771                 m_freem(m0);
2772                 m0 = NULL;
2773         }
2774
2775         /* XXX: clear embedded link IDs in the inner header */
2776         in6_clearscope(&sip6->ip6_src);
2777         in6_clearscope(&sip6->ip6_dst);
2778         in6_clearscope(&nd_rd->nd_rd_target);
2779         in6_clearscope(&nd_rd->nd_rd_dst);
2780
2781         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2782
2783         nd_rd->nd_rd_cksum = 0;
2784         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2785             sizeof(*ip6), ntohs(ip6->ip6_plen));
2786
2787         if (send_sendso_input_hook != NULL) {
2788                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2789                         M_NOWAIT);
2790                 if (mtag == NULL)
2791                         goto fail;
2792                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2793                 m_tag_prepend(m, mtag);
2794         }
2795
2796         /* send the packet to outside... */
2797         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2798         if (outif) {
2799                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2800                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2801         }
2802         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2803
2804         return;
2805
2806 fail:
2807         if (m)
2808                 m_freem(m);
2809         if (m0)
2810                 m_freem(m0);
2811 }
2812
2813 /*
2814  * ICMPv6 socket option processing.
2815  */
2816 int
2817 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2818 {
2819         int error = 0;
2820         int optlen;
2821         struct inpcb *inp = sotoinpcb(so);
2822         int level, op, optname;
2823
2824         if (sopt) {
2825                 level = sopt->sopt_level;
2826                 op = sopt->sopt_dir;
2827                 optname = sopt->sopt_name;
2828                 optlen = sopt->sopt_valsize;
2829         } else
2830                 level = op = optname = optlen = 0;
2831
2832         if (level != IPPROTO_ICMPV6) {
2833                 return EINVAL;
2834         }
2835
2836         switch (op) {
2837         case PRCO_SETOPT:
2838                 switch (optname) {
2839                 case ICMP6_FILTER:
2840                     {
2841                         struct icmp6_filter ic6f;
2842
2843                         if (optlen != sizeof(ic6f)) {
2844                                 error = EMSGSIZE;
2845                                 break;
2846                         }
2847                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2848                         if (error == 0) {
2849                                 INP_WLOCK(inp);
2850                                 *inp->in6p_icmp6filt = ic6f;
2851                                 INP_WUNLOCK(inp);
2852                         }
2853                         break;
2854                     }
2855
2856                 default:
2857                         error = ENOPROTOOPT;
2858                         break;
2859                 }
2860                 break;
2861
2862         case PRCO_GETOPT:
2863                 switch (optname) {
2864                 case ICMP6_FILTER:
2865                     {
2866                         struct icmp6_filter ic6f;
2867
2868                         INP_RLOCK(inp);
2869                         ic6f = *inp->in6p_icmp6filt;
2870                         INP_RUNLOCK(inp);
2871                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2872                         break;
2873                     }
2874
2875                 default:
2876                         error = ENOPROTOOPT;
2877                         break;
2878                 }
2879                 break;
2880         }
2881
2882         return (error);
2883 }
2884
2885 /*
2886  * Perform rate limit check.
2887  * Returns 0 if it is okay to send the icmp6 packet.
2888  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2889  * limitation.
2890  *
2891  * XXX per-destination/type check necessary?
2892  *
2893  * dst - not used at this moment
2894  * type - not used at this moment
2895  * code - not used at this moment
2896  */
2897 static int
2898 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2899     const int code)
2900 {
2901         int ret;
2902
2903         ret = 0;        /* okay to send */
2904
2905         /* PPS limit */
2906         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2907             V_icmp6errppslim)) {
2908                 /* The packet is subject to rate limit */
2909                 ret++;
2910         }
2911
2912         return ret;
2913 }