2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 * $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
33 * Copyright (c) 1982, 1986, 1988, 1993
34 * The Regents of the University of California. All rights reserved.
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
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.
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
60 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
70 #include <sys/param.h>
71 #include <sys/domain.h>
73 #include <sys/kernel.h>
75 #include <sys/malloc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
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>
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>
111 #include <netipsec/ipsec.h>
112 #include <netipsec/key.h>
115 extern struct domain inet6domain;
117 VNET_DEFINE(struct icmp6stat, icmp6stat);
119 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
120 VNET_DECLARE(struct inpcbhead, ripcb);
121 VNET_DECLARE(int, icmp6errppslim);
122 static VNET_DEFINE(int, icmp6errpps_count) = 0;
123 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
124 VNET_DECLARE(int, icmp6_nodeinfo);
126 #define V_ripcbinfo VNET(ripcbinfo)
127 #define V_ripcb VNET(ripcb)
128 #define V_icmp6errppslim VNET(icmp6errppslim)
129 #define V_icmp6errpps_count VNET(icmp6errpps_count)
130 #define V_icmp6errppslim_last VNET(icmp6errppslim_last)
131 #define V_icmp6_nodeinfo VNET(icmp6_nodeinfo)
133 static void icmp6_errcount(int, int);
134 static int icmp6_rip6_input(struct mbuf **, int);
135 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
136 static const char *icmp6_redirect_diag(struct in6_addr *,
137 struct in6_addr *, struct in6_addr *);
138 static struct mbuf *ni6_input(struct mbuf *, int);
139 static struct mbuf *ni6_nametodns(const char *, int, int);
140 static int ni6_dnsmatch(const char *, int, const char *, int);
141 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
142 struct ifnet **, struct in6_addr *);
143 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
144 struct ifnet *, int);
145 static int icmp6_notify_error(struct mbuf **, int, int, int);
148 * Kernel module interface for updating icmp6stat. The argument is an index
149 * into icmp6stat treated as an array of u_quad_t. While this encodes the
150 * general layout of icmp6stat into the caller, it doesn't encode its
151 * location, so that future changes to add, for example, per-CPU stats
152 * support won't cause binary compatibility problems for kernel modules.
155 kmod_icmp6stat_inc(int statnum)
158 (*((u_quad_t *)&V_icmp6stat + statnum))++;
162 icmp6_errcount(int type, int code)
165 case ICMP6_DST_UNREACH:
167 case ICMP6_DST_UNREACH_NOROUTE:
168 ICMP6STAT_INC(icp6s_odst_unreach_noroute);
170 case ICMP6_DST_UNREACH_ADMIN:
171 ICMP6STAT_INC(icp6s_odst_unreach_admin);
173 case ICMP6_DST_UNREACH_BEYONDSCOPE:
174 ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
176 case ICMP6_DST_UNREACH_ADDR:
177 ICMP6STAT_INC(icp6s_odst_unreach_addr);
179 case ICMP6_DST_UNREACH_NOPORT:
180 ICMP6STAT_INC(icp6s_odst_unreach_noport);
184 case ICMP6_PACKET_TOO_BIG:
185 ICMP6STAT_INC(icp6s_opacket_too_big);
187 case ICMP6_TIME_EXCEEDED:
189 case ICMP6_TIME_EXCEED_TRANSIT:
190 ICMP6STAT_INC(icp6s_otime_exceed_transit);
192 case ICMP6_TIME_EXCEED_REASSEMBLY:
193 ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
197 case ICMP6_PARAM_PROB:
199 case ICMP6_PARAMPROB_HEADER:
200 ICMP6STAT_INC(icp6s_oparamprob_header);
202 case ICMP6_PARAMPROB_NEXTHEADER:
203 ICMP6STAT_INC(icp6s_oparamprob_nextheader);
205 case ICMP6_PARAMPROB_OPTION:
206 ICMP6STAT_INC(icp6s_oparamprob_option);
211 ICMP6STAT_INC(icp6s_oredirect);
214 ICMP6STAT_INC(icp6s_ounknown);
218 * A wrapper function for icmp6_error() necessary when the erroneous packet
219 * may not contain enough scope zone information.
222 icmp6_error2(struct mbuf *m, int type, int code, int param,
230 #ifndef PULLDOWN_TEST
231 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
233 if (m->m_len < sizeof(struct ip6_hdr)) {
234 m = m_pullup(m, sizeof(struct ip6_hdr));
240 ip6 = mtod(m, struct ip6_hdr *);
242 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
244 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
247 icmp6_error(m, type, code, param);
251 * Generate an error packet of type error in response to bad IP6 packet.
254 icmp6_error(struct mbuf *m, int type, int code, int param)
256 struct ip6_hdr *oip6, *nip6;
257 struct icmp6_hdr *icmp6;
262 ICMP6STAT_INC(icp6s_error);
264 /* count per-type-code statistics */
265 icmp6_errcount(type, code);
267 #ifdef M_DECRYPTED /*not openbsd*/
268 if (m->m_flags & M_DECRYPTED) {
269 ICMP6STAT_INC(icp6s_canterror);
274 #ifndef PULLDOWN_TEST
275 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
277 if (m->m_len < sizeof(struct ip6_hdr)) {
278 m = m_pullup(m, sizeof(struct ip6_hdr));
283 oip6 = mtod(m, struct ip6_hdr *);
286 * If the destination address of the erroneous packet is a multicast
287 * address, or the packet was sent using link-layer multicast,
288 * we should basically suppress sending an error (RFC 2463, Section
290 * We have two exceptions (the item e.2 in that section):
291 * - the Packet Too Big message can be sent for path MTU discovery.
292 * - the Parameter Problem Message that can be allowed an icmp6 error
293 * in the option type field. This check has been done in
294 * ip6_unknown_opt(), so we can just check the type and code.
296 if ((m->m_flags & (M_BCAST|M_MCAST) ||
297 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
298 (type != ICMP6_PACKET_TOO_BIG &&
299 (type != ICMP6_PARAM_PROB ||
300 code != ICMP6_PARAMPROB_OPTION)))
304 * RFC 2463, 2.4 (e.5): source address check.
305 * XXX: the case of anycast source?
307 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
308 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
312 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
316 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
317 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
318 struct icmp6_hdr *icp;
320 #ifndef PULLDOWN_TEST
321 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
322 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
324 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
327 ICMP6STAT_INC(icp6s_tooshort);
331 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
332 icp->icmp6_type == ND_REDIRECT) {
335 * Special case: for redirect (which is
336 * informational) we must not send icmp6 error.
338 ICMP6STAT_INC(icp6s_canterror);
341 /* ICMPv6 informational - send the error */
344 /* non-ICMPv6 - send the error */
347 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
349 /* Finally, do rate limitation check. */
350 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
351 ICMP6STAT_INC(icp6s_toofreq);
356 * OK, ICMP6 can be generated.
359 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
360 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
362 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
363 M_PREPEND(m, preplen, M_DONTWAIT); /* FIB is also copied over. */
364 if (m && m->m_len < preplen)
365 m = m_pullup(m, preplen);
367 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
371 nip6 = mtod(m, struct ip6_hdr *);
372 nip6->ip6_src = oip6->ip6_src;
373 nip6->ip6_dst = oip6->ip6_dst;
375 in6_clearscope(&oip6->ip6_src);
376 in6_clearscope(&oip6->ip6_dst);
378 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
379 icmp6->icmp6_type = type;
380 icmp6->icmp6_code = code;
381 icmp6->icmp6_pptr = htonl((u_int32_t)param);
384 * icmp6_reflect() is designed to be in the input path.
385 * icmp6_error() can be called from both input and output path,
386 * and if we are in output path rcvif could contain bogus value.
387 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
388 * information in ip header (nip6).
390 m->m_pkthdr.rcvif = NULL;
392 ICMP6STAT_INC(icp6s_outhist[type]);
393 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
399 * If we can't tell whether or not we can generate ICMP6, free it.
405 * Process a received ICMP6 message.
408 icmp6_input(struct mbuf **mp, int *offp, int proto)
410 struct mbuf *m = *mp, *n;
412 struct ip6_hdr *ip6, *nip6;
413 struct icmp6_hdr *icmp6, *nicmp6;
415 int icmp6len = m->m_pkthdr.len - *offp;
417 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
420 ifp = m->m_pkthdr.rcvif;
422 #ifndef PULLDOWN_TEST
423 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
424 /* m might change if M_LOOP. So, call mtod after this */
428 * Locate icmp6 structure in mbuf, and check
429 * that not corrupted and of at least minimum length
432 ip6 = mtod(m, struct ip6_hdr *);
433 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
434 if (icmp6len < sizeof(struct icmp6_hdr)) {
435 ICMP6STAT_INC(icp6s_tooshort);
440 * Check multicast group membership.
441 * Note: SSM filters are not applied for ICMPv6 traffic.
443 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
444 struct in6_multi *inm;
446 inm = in6m_lookup(ifp, &ip6->ip6_dst);
448 IP6STAT_INC(ip6s_notmember);
449 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
455 * calculate the checksum
457 #ifndef PULLDOWN_TEST
458 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
460 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
462 ICMP6STAT_INC(icp6s_tooshort);
466 code = icmp6->icmp6_code;
468 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
470 "ICMP6 checksum error(%d|%x) %s\n",
471 icmp6->icmp6_type, sum,
472 ip6_sprintf(ip6bufs, &ip6->ip6_src)));
473 ICMP6STAT_INC(icp6s_checksum);
477 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
479 * Deliver very specific ICMP6 type only.
480 * This is important to deliver TOOBIG. Otherwise PMTUD
483 switch (icmp6->icmp6_type) {
484 case ICMP6_DST_UNREACH:
485 case ICMP6_PACKET_TOO_BIG:
486 case ICMP6_TIME_EXCEEDED:
493 ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
494 icmp6_ifstat_inc(ifp, ifs6_in_msg);
495 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
496 icmp6_ifstat_inc(ifp, ifs6_in_error);
498 switch (icmp6->icmp6_type) {
499 case ICMP6_DST_UNREACH:
500 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
502 case ICMP6_DST_UNREACH_NOROUTE:
503 code = PRC_UNREACH_NET;
505 case ICMP6_DST_UNREACH_ADMIN:
506 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
507 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
509 case ICMP6_DST_UNREACH_ADDR:
512 case ICMP6_DST_UNREACH_BEYONDSCOPE:
513 /* I mean "source address was incorrect." */
514 code = PRC_PARAMPROB;
516 case ICMP6_DST_UNREACH_NOPORT:
517 code = PRC_UNREACH_PORT;
525 case ICMP6_PACKET_TOO_BIG:
526 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
528 /* validation is made in icmp6_mtudisc_update */
533 * Updating the path MTU will be done after examining
534 * intermediate extension headers.
539 case ICMP6_TIME_EXCEEDED:
540 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
542 case ICMP6_TIME_EXCEED_TRANSIT:
543 code = PRC_TIMXCEED_INTRANS;
545 case ICMP6_TIME_EXCEED_REASSEMBLY:
546 code = PRC_TIMXCEED_REASS;
554 case ICMP6_PARAM_PROB:
555 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
557 case ICMP6_PARAMPROB_NEXTHEADER:
558 code = PRC_UNREACH_PROTOCOL;
560 case ICMP6_PARAMPROB_HEADER:
561 case ICMP6_PARAMPROB_OPTION:
562 code = PRC_PARAMPROB;
570 case ICMP6_ECHO_REQUEST:
571 icmp6_ifstat_inc(ifp, ifs6_in_echo);
574 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
578 if ((n->m_flags & M_EXT) != 0
579 || n->m_len < off + sizeof(struct icmp6_hdr)) {
581 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
584 MGETHDR(n, M_DONTWAIT, n0->m_type);
585 n0len = n0->m_pkthdr.len; /* save for use below */
587 M_MOVE_PKTHDR(n, n0); /* FIB copied. */
588 if (n && maxlen >= MHLEN) {
589 MCLGET(n, M_DONTWAIT);
590 if ((n->m_flags & M_EXT) == 0) {
601 * Copy IPv6 and ICMPv6 only.
603 nip6 = mtod(n, struct ip6_hdr *);
604 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
605 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
606 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
607 noff = sizeof(struct ip6_hdr);
608 /* new mbuf contains only ipv6+icmpv6 headers */
609 n->m_len = noff + sizeof(struct icmp6_hdr);
611 * Adjust mbuf. ip6_plen will be adjusted in
614 m_adj(n0, off + sizeof(struct icmp6_hdr));
615 /* recalculate complete packet size */
616 n->m_pkthdr.len = n0len + (noff - off);
619 nip6 = mtod(n, struct ip6_hdr *);
620 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
624 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
625 nicmp6->icmp6_code = 0;
627 ICMP6STAT_INC(icp6s_reflect);
628 ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
629 icmp6_reflect(n, noff);
633 case ICMP6_ECHO_REPLY:
634 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
639 case MLD_LISTENER_QUERY:
640 case MLD_LISTENER_REPORT:
641 case MLD_LISTENER_DONE:
642 case MLDV2_LISTENER_REPORT:
644 * Drop MLD traffic which is not link-local, has a hop limit
645 * of greater than 1 hop, or which does not have the
646 * IPv6 HBH Router Alert option.
647 * As IPv6 HBH options are stripped in ip6_input() we must
648 * check an mbuf header flag.
649 * XXX Should we also sanity check that these messages
650 * were directed to a link-local multicast prefix?
652 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
654 if (mld_input(m, off, icmp6len) != 0)
655 return (IPPROTO_DONE);
659 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
661 enum { WRU, FQDN } mode;
663 if (!V_icmp6_nodeinfo)
666 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
668 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
674 #ifndef PULLDOWN_TEST
675 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
678 n = m_copy(m, 0, M_COPYALL);
680 n = ni6_input(n, off);
681 /* XXX meaningless if n == NULL */
682 noff = sizeof(struct ip6_hdr);
686 int maxlen, maxhlen, hlen;
689 * XXX: this combination of flags is pointless,
690 * but should we keep this for compatibility?
692 if ((V_icmp6_nodeinfo & 5) != 5)
697 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
698 if (maxlen >= MCLBYTES) {
702 MGETHDR(n, M_DONTWAIT, m->m_type);
703 if (n && maxlen > MHLEN) {
704 MCLGET(n, M_DONTWAIT);
705 if ((n->m_flags & M_EXT) == 0) {
710 if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
712 * Previous code did a blind M_COPY_PKTHDR
713 * and said "just for rcvif". If true, then
714 * we could tolerate the dup failing (due to
715 * the deep copy of the tag chain). For now
716 * be conservative and just fail.
725 n->m_pkthdr.rcvif = NULL;
727 maxhlen = M_TRAILINGSPACE(n) - maxlen;
728 pr = curthread->td_ucred->cr_prison;
729 mtx_lock(&pr->pr_mtx);
730 hlen = strlen(pr->pr_hostname);
734 * Copy IPv6 and ICMPv6 only.
736 nip6 = mtod(n, struct ip6_hdr *);
737 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
738 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
739 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
740 p = (u_char *)(nicmp6 + 1);
742 /* meaningless TTL */
743 bcopy(pr->pr_hostname, p + 4, maxhlen);
744 mtx_unlock(&pr->pr_mtx);
745 noff = sizeof(struct ip6_hdr);
746 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
747 sizeof(struct icmp6_hdr) + 4 + maxhlen;
748 nicmp6->icmp6_type = ICMP6_WRUREPLY;
749 nicmp6->icmp6_code = 0;
752 ICMP6STAT_INC(icp6s_reflect);
753 ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
754 icmp6_reflect(n, noff);
764 case ND_ROUTER_SOLICIT:
765 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
768 if (icmp6len < sizeof(struct nd_router_solicit))
770 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
773 /* Send incoming SeND packet to user space. */
774 if (send_sendso_input_hook != NULL) {
775 IP6_EXTHDR_CHECK(m, off,
776 icmp6len, IPPROTO_DONE);
777 error = send_sendso_input_hook(m, ifp,
779 /* -1 == no app on SEND socket */
781 return (IPPROTO_DONE);
782 nd6_rs_input(m, off, icmp6len);
784 nd6_rs_input(m, off, icmp6len);
788 if (send_sendso_input_hook != NULL) {
789 IP6_EXTHDR_CHECK(n, off,
790 icmp6len, IPPROTO_DONE);
791 error = send_sendso_input_hook(n, ifp,
795 /* -1 == no app on SEND socket */
796 nd6_rs_input(n, off, icmp6len);
798 nd6_rs_input(n, off, icmp6len);
802 case ND_ROUTER_ADVERT:
803 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
806 if (icmp6len < sizeof(struct nd_router_advert))
808 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
810 /* Send incoming SeND-protected/ND packet to user space. */
811 if (send_sendso_input_hook != NULL) {
812 error = send_sendso_input_hook(m, ifp,
815 return (IPPROTO_DONE);
816 nd6_ra_input(m, off, icmp6len);
818 nd6_ra_input(m, off, icmp6len);
822 if (send_sendso_input_hook != NULL) {
823 error = send_sendso_input_hook(n, ifp,
827 nd6_ra_input(n, off, icmp6len);
829 nd6_ra_input(n, off, icmp6len);
833 case ND_NEIGHBOR_SOLICIT:
834 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
837 if (icmp6len < sizeof(struct nd_neighbor_solicit))
839 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
840 if (send_sendso_input_hook != NULL) {
841 error = send_sendso_input_hook(m, ifp,
844 return (IPPROTO_DONE);
845 nd6_ns_input(m, off, icmp6len);
847 nd6_ns_input(m, off, icmp6len);
851 if (send_sendso_input_hook != NULL) {
852 error = send_sendso_input_hook(n, ifp,
856 nd6_ns_input(n, off, icmp6len);
858 nd6_ns_input(n, off, icmp6len);
862 case ND_NEIGHBOR_ADVERT:
863 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
866 if (icmp6len < sizeof(struct nd_neighbor_advert))
868 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
870 /* Send incoming SeND-protected/ND packet to user space. */
871 if (send_sendso_input_hook != NULL) {
872 error = send_sendso_input_hook(m, ifp,
875 return (IPPROTO_DONE);
876 nd6_na_input(m, off, icmp6len);
878 nd6_na_input(m, off, icmp6len);
882 if (send_sendso_input_hook != NULL) {
883 error = send_sendso_input_hook(n, ifp,
887 nd6_na_input(n, off, icmp6len);
889 nd6_na_input(n, off, icmp6len);
894 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
897 if (icmp6len < sizeof(struct nd_redirect))
899 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
900 if (send_sendso_input_hook != NULL) {
901 error = send_sendso_input_hook(m, ifp,
904 return (IPPROTO_DONE);
905 icmp6_redirect_input(m, off);
907 icmp6_redirect_input(m, off);
911 if (send_sendso_input_hook != NULL) {
912 error = send_sendso_input_hook(n, ifp,
916 icmp6_redirect_input(n, off);
918 icmp6_redirect_input(n, off);
922 case ICMP6_ROUTER_RENUMBERING:
923 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
924 code != ICMP6_ROUTER_RENUMBERING_RESULT)
926 if (icmp6len < sizeof(struct icmp6_router_renum))
932 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
933 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
934 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
935 ifp ? ifp->if_index : 0));
936 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
937 /* ICMPv6 error: MUST deliver it by spec... */
941 /* ICMPv6 informational: MUST not deliver */
945 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
946 /* In this case, m should've been freed. */
947 return (IPPROTO_DONE);
952 ICMP6STAT_INC(icp6s_badcode);
956 ICMP6STAT_INC(icp6s_badlen);
960 /* deliver the packet to appropriate sockets */
961 icmp6_rip6_input(&m, *offp);
971 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
973 struct mbuf *m = *mp;
974 struct icmp6_hdr *icmp6;
975 struct ip6_hdr *eip6;
977 struct sockaddr_in6 icmp6src, icmp6dst;
979 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
980 ICMP6STAT_INC(icp6s_tooshort);
983 #ifndef PULLDOWN_TEST
984 IP6_EXTHDR_CHECK(m, off,
985 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
986 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
988 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
989 sizeof(*icmp6) + sizeof(struct ip6_hdr));
991 ICMP6STAT_INC(icp6s_tooshort);
995 eip6 = (struct ip6_hdr *)(icmp6 + 1);
997 /* Detect the upper level protocol */
999 void (*ctlfunc)(int, struct sockaddr *, void *);
1000 u_int8_t nxt = eip6->ip6_nxt;
1001 int eoff = off + sizeof(struct icmp6_hdr) +
1002 sizeof(struct ip6_hdr);
1003 struct ip6ctlparam ip6cp;
1004 struct in6_addr *finaldst = NULL;
1005 int icmp6type = icmp6->icmp6_type;
1006 struct ip6_frag *fh;
1007 struct ip6_rthdr *rth;
1008 struct ip6_rthdr0 *rth0;
1011 while (1) { /* XXX: should avoid infinite loop explicitly? */
1015 case IPPROTO_HOPOPTS:
1016 case IPPROTO_DSTOPTS:
1018 #ifndef PULLDOWN_TEST
1019 IP6_EXTHDR_CHECK(m, 0,
1020 eoff + sizeof(struct ip6_ext), -1);
1021 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1023 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1026 ICMP6STAT_INC(icp6s_tooshort);
1031 if (nxt == IPPROTO_AH)
1032 eoff += (eh->ip6e_len + 2) << 2;
1034 eoff += (eh->ip6e_len + 1) << 3;
1037 case IPPROTO_ROUTING:
1039 * When the erroneous packet contains a
1040 * routing header, we should examine the
1041 * header to determine the final destination.
1042 * Otherwise, we can't properly update
1043 * information that depends on the final
1044 * destination (e.g. path MTU).
1046 #ifndef PULLDOWN_TEST
1047 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1048 rth = (struct ip6_rthdr *)
1049 (mtod(m, caddr_t) + eoff);
1051 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1052 eoff, sizeof(*rth));
1054 ICMP6STAT_INC(icp6s_tooshort);
1058 rthlen = (rth->ip6r_len + 1) << 3;
1060 * XXX: currently there is no
1061 * officially defined type other
1063 * Note that if the segment left field
1064 * is 0, all intermediate hops must
1067 if (rth->ip6r_segleft &&
1068 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1071 #ifndef PULLDOWN_TEST
1072 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1073 rth0 = (struct ip6_rthdr0 *)
1074 (mtod(m, caddr_t) + eoff);
1076 IP6_EXTHDR_GET(rth0,
1077 struct ip6_rthdr0 *, m,
1080 ICMP6STAT_INC(icp6s_tooshort);
1084 /* just ignore a bogus header */
1085 if ((rth0->ip6r0_len % 2) == 0 &&
1086 (hops = rth0->ip6r0_len/2))
1087 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1090 nxt = rth->ip6r_nxt;
1092 case IPPROTO_FRAGMENT:
1093 #ifndef PULLDOWN_TEST
1094 IP6_EXTHDR_CHECK(m, 0, eoff +
1095 sizeof(struct ip6_frag), -1);
1096 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1099 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1102 ICMP6STAT_INC(icp6s_tooshort);
1107 * Data after a fragment header is meaningless
1108 * unless it is the first fragment, but
1109 * we'll go to the notify label for path MTU
1112 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1115 eoff += sizeof(struct ip6_frag);
1120 * This case includes ESP and the No Next
1121 * Header. In such cases going to the notify
1122 * label does not have any meaning
1123 * (i.e. ctlfunc will be NULL), but we go
1124 * anyway since we might have to update
1125 * path MTU information.
1131 #ifndef PULLDOWN_TEST
1132 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1134 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1135 sizeof(*icmp6) + sizeof(struct ip6_hdr));
1136 if (icmp6 == NULL) {
1137 ICMP6STAT_INC(icp6s_tooshort);
1143 * retrieve parameters from the inner IPv6 header, and convert
1144 * them into sockaddr structures.
1145 * XXX: there is no guarantee that the source or destination
1146 * addresses of the inner packet are in the same scope as
1147 * the addresses of the icmp packet. But there is no other
1148 * way to determine the zone.
1150 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1152 bzero(&icmp6dst, sizeof(icmp6dst));
1153 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1154 icmp6dst.sin6_family = AF_INET6;
1155 if (finaldst == NULL)
1156 icmp6dst.sin6_addr = eip6->ip6_dst;
1158 icmp6dst.sin6_addr = *finaldst;
1159 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1161 bzero(&icmp6src, sizeof(icmp6src));
1162 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1163 icmp6src.sin6_family = AF_INET6;
1164 icmp6src.sin6_addr = eip6->ip6_src;
1165 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1167 icmp6src.sin6_flowinfo =
1168 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1170 if (finaldst == NULL)
1171 finaldst = &eip6->ip6_dst;
1173 ip6cp.ip6c_icmp6 = icmp6;
1174 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1175 ip6cp.ip6c_off = eoff;
1176 ip6cp.ip6c_finaldst = finaldst;
1177 ip6cp.ip6c_src = &icmp6src;
1178 ip6cp.ip6c_nxt = nxt;
1182 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1183 notifymtu = ntohl(icmp6->icmp6_mtu);
1184 ip6cp.ip6c_cmdarg = (void *)¬ifymtu;
1185 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/
1188 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1189 (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1191 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1204 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1206 struct in6_addr *dst = ip6cp->ip6c_finaldst;
1207 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1208 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1209 u_int mtu = ntohl(icmp6->icmp6_mtu);
1210 struct in_conninfo inc;
1214 * RFC2460 section 5, last paragraph.
1215 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1216 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1217 * due to packet translator in the middle.
1218 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1221 if (mtu < IPV6_MMTU)
1226 * we reject ICMPv6 too big with abnormally small value.
1227 * XXX what is the good definition of "abnormally small"?
1229 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1236 * In case the suggested mtu is less than IPV6_MMTU, we
1237 * only need to remember that it was for above mentioned
1238 * "alwaysfrag" case.
1239 * Try to be as close to the spec as possible.
1241 if (mtu < IPV6_MMTU)
1242 mtu = IPV6_MMTU - 8;
1244 bzero(&inc, sizeof(inc));
1245 inc.inc_fibnum = M_GETFIB(m);
1246 inc.inc_flags |= INC_ISIPV6;
1247 inc.inc6_faddr = *dst;
1248 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1251 if (mtu < tcp_maxmtu6(&inc, NULL)) {
1252 tcp_hc_updatemtu(&inc, mtu);
1253 ICMP6STAT_INC(icp6s_pmtuchg);
1258 * Process a Node Information Query packet, based on
1259 * draft-ietf-ipngwg-icmp-name-lookups-07.
1261 * Spec incompatibilities:
1262 * - IPv6 Subject address handling
1263 * - IPv4 Subject address handling support missing
1264 * - Proxy reply (answer even if it's not for me)
1265 * - joins NI group address at in6_ifattach() time only, does not cope
1266 * with hostname changes by sethostname(3)
1268 static struct mbuf *
1269 ni6_input(struct mbuf *m, int off)
1271 struct icmp6_nodeinfo *ni6, *nni6;
1272 struct mbuf *n = NULL;
1276 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1277 struct ni_reply_fqdn *fqdn;
1278 int addrs; /* for NI_QTYPE_NODEADDR */
1279 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1280 struct in6_addr in6_subj; /* subject address */
1281 struct ip6_hdr *ip6;
1282 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1284 struct in6_ifaddr *ia6 = NULL;
1286 ip6 = mtod(m, struct ip6_hdr *);
1287 #ifndef PULLDOWN_TEST
1288 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1290 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1292 /* m is already reclaimed */
1298 * Validate IPv6 source address.
1299 * The default configuration MUST be to refuse answering queries from
1300 * global-scope addresses according to RFC4602.
1302 * - it's not very clear what "refuse" means; this implementation
1304 * - it's not very easy to identify global-scope (unicast) addresses
1305 * since there are many prefixes for them. It should be safer
1306 * and in practice sufficient to check "all" but loopback and
1307 * link-local (note that site-local unicast was deprecated and
1308 * ULA is defined as global scope-wise)
1310 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1311 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1312 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1316 * Validate IPv6 destination address.
1318 * The Responder must discard the Query without further processing
1319 * unless it is one of the Responder's unicast or anycast addresses, or
1320 * a link-local scope multicast address which the Responder has joined.
1321 * [RFC4602, Section 5.]
1323 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1324 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1326 /* else it's a link-local multicast, fine */
1327 } else { /* unicast or anycast */
1328 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1329 goto bad; /* XXX impossible */
1331 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1332 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1333 ifa_free(&ia6->ia_ifa);
1334 nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1335 "a temporary address in %s:%d",
1336 __FILE__, __LINE__));
1339 ifa_free(&ia6->ia_ifa);
1342 /* validate query Subject field. */
1343 qtype = ntohs(ni6->ni_qtype);
1344 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1347 case NI_QTYPE_SUPTYPES:
1349 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1353 case NI_QTYPE_NODEADDR:
1354 case NI_QTYPE_IPV4ADDR:
1355 switch (ni6->ni_code) {
1356 case ICMP6_NI_SUBJ_IPV6:
1357 #if ICMP6_NI_SUBJ_IPV6 != 0
1361 * backward compatibility - try to accept 03 draft
1362 * format, where no Subject is present.
1364 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1369 #if ICMP6_NI_SUBJ_IPV6 != 0
1370 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1374 if (subjlen != sizeof(struct in6_addr))
1378 * Validate Subject address.
1380 * Not sure what exactly "address belongs to the node"
1381 * means in the spec, is it just unicast, or what?
1383 * At this moment we consider Subject address as
1384 * "belong to the node" if the Subject address equals
1385 * to the IPv6 destination address; validation for
1386 * IPv6 destination address should have done enough
1389 * We do not do proxy at this moment.
1391 /* m_pulldown instead of copy? */
1392 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1393 subjlen, (caddr_t)&in6_subj);
1394 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1397 subj = (char *)&in6_subj;
1398 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1402 * XXX if we are to allow other cases, we should really
1403 * be careful about scope here.
1404 * basically, we should disallow queries toward IPv6
1405 * destination X with subject Y,
1406 * if scope(X) > scope(Y).
1407 * if we allow scope(X) > scope(Y), it will result in
1408 * information leakage across scope boundary.
1412 case ICMP6_NI_SUBJ_FQDN:
1414 * Validate Subject name with gethostname(3).
1416 * The behavior may need some debate, since:
1417 * - we are not sure if the node has FQDN as
1418 * hostname (returned by gethostname(3)).
1419 * - the code does wildcard match for truncated names.
1420 * however, we are not sure if we want to perform
1421 * wildcard match, if gethostname(3) side has
1422 * truncated hostname.
1424 pr = curthread->td_ucred->cr_prison;
1425 mtx_lock(&pr->pr_mtx);
1426 n = ni6_nametodns(pr->pr_hostname,
1427 strlen(pr->pr_hostname), 0);
1428 mtx_unlock(&pr->pr_mtx);
1429 if (!n || n->m_next || n->m_len == 0)
1431 IP6_EXTHDR_GET(subj, char *, m,
1432 off + sizeof(struct icmp6_nodeinfo), subjlen);
1435 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1443 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1450 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1453 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1456 case NI_QTYPE_NODEADDR:
1457 case NI_QTYPE_IPV4ADDR:
1458 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1463 /* guess reply length */
1466 break; /* no reply data */
1467 case NI_QTYPE_SUPTYPES:
1468 replylen += sizeof(u_int32_t);
1471 /* XXX will append an mbuf */
1472 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1474 case NI_QTYPE_NODEADDR:
1475 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1476 if ((replylen += addrs * (sizeof(struct in6_addr) +
1477 sizeof(u_int32_t))) > MCLBYTES)
1478 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1480 case NI_QTYPE_IPV4ADDR:
1481 /* unsupported - should respond with unknown Qtype? */
1485 * XXX: We must return a reply with the ICMP6 code
1486 * `unknown Qtype' in this case. However we regard the case
1487 * as an FQDN query for backward compatibility.
1488 * Older versions set a random value to this field,
1489 * so it rarely varies in the defined qtypes.
1490 * But the mechanism is not reliable...
1491 * maybe we should obsolete older versions.
1493 qtype = NI_QTYPE_FQDN;
1494 /* XXX will append an mbuf */
1495 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1500 /* allocate an mbuf to reply. */
1501 MGETHDR(n, M_DONTWAIT, m->m_type);
1506 M_MOVE_PKTHDR(n, m); /* just for recvif and FIB */
1507 if (replylen > MHLEN) {
1508 if (replylen > MCLBYTES) {
1510 * XXX: should we try to allocate more? But MCLBYTES
1511 * is probably much larger than IPV6_MMTU...
1515 MCLGET(n, M_DONTWAIT);
1516 if ((n->m_flags & M_EXT) == 0) {
1520 n->m_pkthdr.len = n->m_len = replylen;
1522 /* copy mbuf header and IPv6 + Node Information base headers */
1523 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1524 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1525 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1527 /* qtype dependent procedure */
1530 nni6->ni_code = ICMP6_NI_SUCCESS;
1533 case NI_QTYPE_SUPTYPES:
1536 nni6->ni_code = ICMP6_NI_SUCCESS;
1537 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1538 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1539 v = (u_int32_t)htonl(0x0000000f);
1540 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1544 nni6->ni_code = ICMP6_NI_SUCCESS;
1545 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1546 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1547 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1548 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1550 * XXX do we really have FQDN in hostname?
1552 pr = curthread->td_ucred->cr_prison;
1553 mtx_lock(&pr->pr_mtx);
1554 n->m_next = ni6_nametodns(pr->pr_hostname,
1555 strlen(pr->pr_hostname), oldfqdn);
1556 mtx_unlock(&pr->pr_mtx);
1557 if (n->m_next == NULL)
1559 /* XXX we assume that n->m_next is not a chain */
1560 if (n->m_next->m_next != NULL)
1562 n->m_pkthdr.len += n->m_next->m_len;
1564 case NI_QTYPE_NODEADDR:
1568 nni6->ni_code = ICMP6_NI_SUCCESS;
1569 n->m_pkthdr.len = n->m_len =
1570 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1571 lenlim = M_TRAILINGSPACE(n);
1572 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1573 /* XXX: reset mbuf length */
1574 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1575 sizeof(struct icmp6_nodeinfo) + copied;
1579 break; /* XXX impossible! */
1582 nni6->ni_type = ICMP6_NI_REPLY;
1594 * make a mbuf with DNS-encoded string. no compression support.
1596 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1597 * treated as truncated name (two \0 at the end). this is a wild guess.
1599 * old - return pascal string if non-zero
1601 static struct mbuf *
1602 ni6_nametodns(const char *name, int namelen, int old)
1614 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1615 MGET(m, M_DONTWAIT, MT_DATA);
1616 if (m && len > MLEN) {
1617 MCLGET(m, M_DONTWAIT);
1618 if ((m->m_flags & M_EXT) == 0)
1627 *mtod(m, char *) = namelen;
1628 bcopy(name, mtod(m, char *) + 1, namelen);
1632 cp = mtod(m, char *);
1633 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1635 /* if not certain about my name, return empty buffer */
1640 * guess if it looks like shortened hostname, or FQDN.
1641 * shortened hostname needs two trailing "\0".
1644 for (p = name; p < name + namelen; p++) {
1645 if (*p && *p == '.')
1654 while (cp < ep && p < name + namelen) {
1656 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1658 /* result does not fit into mbuf */
1659 if (cp + i + 1 >= ep)
1662 * DNS label length restriction, RFC1035 page 8.
1663 * "i == 0" case is included here to avoid returning
1664 * 0-length label on "foo..bar".
1666 if (i <= 0 || i >= 64)
1672 if (p < name + namelen && *p == '.')
1676 if (cp + nterm >= ep)
1680 m->m_len = cp - mtod(m, char *);
1684 panic("should not reach here");
1694 * check if two DNS-encoded string matches. takes care of truncated
1695 * form (with \0\0 at the end). no compression support.
1696 * XXX upper/lowercase match (see RFC2065)
1699 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1701 const char *a0, *b0;
1704 /* simplest case - need validation? */
1705 if (alen == blen && bcmp(a, b, alen) == 0)
1711 /* termination is mandatory */
1712 if (alen < 2 || blen < 2)
1714 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1719 while (a - a0 < alen && b - b0 < blen) {
1720 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1723 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1725 /* we don't support compression yet */
1726 if (a[0] >= 64 || b[0] >= 64)
1729 /* truncated case */
1730 if (a[0] == 0 && a - a0 == alen - 1)
1732 if (b[0] == 0 && b - b0 == blen - 1)
1734 if (a[0] == 0 || b[0] == 0)
1740 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1742 if (bcmp(a + 1, b + 1, l) != 0)
1749 if (a - a0 == alen && b - b0 == blen)
1756 * calculate the number of addresses to be returned in the node info reply.
1759 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1760 struct in6_addr *subj)
1763 struct in6_ifaddr *ifa6;
1765 int addrs = 0, addrsofif, iffound = 0;
1766 int niflags = ni6->ni_flags;
1768 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1769 switch (ni6->ni_code) {
1770 case ICMP6_NI_SUBJ_IPV6:
1771 if (subj == NULL) /* must be impossible... */
1776 * XXX: we only support IPv6 subject address for
1783 IFNET_RLOCK_NOSLEEP();
1784 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1787 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1788 if (ifa->ifa_addr->sa_family != AF_INET6)
1790 ifa6 = (struct in6_ifaddr *)ifa;
1792 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1793 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1797 * IPv4-mapped addresses can only be returned by a
1798 * Node Information proxy, since they represent
1799 * addresses of IPv4-only nodes, which perforce do
1800 * not implement this protocol.
1801 * [icmp-name-lookups-07, Section 5.4]
1802 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1803 * this function at this moment.
1806 /* What do we have to do about ::1? */
1807 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1808 case IPV6_ADDR_SCOPE_LINKLOCAL:
1809 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1812 case IPV6_ADDR_SCOPE_SITELOCAL:
1813 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1816 case IPV6_ADDR_SCOPE_GLOBAL:
1817 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1825 * check if anycast is okay.
1826 * XXX: just experimental. not in the spec.
1828 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1829 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1830 continue; /* we need only unicast addresses */
1831 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1832 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1835 addrsofif++; /* count the address */
1837 IF_ADDR_RUNLOCK(ifp);
1840 IFNET_RUNLOCK_NOSLEEP();
1846 IFNET_RUNLOCK_NOSLEEP();
1852 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1853 struct ifnet *ifp0, int resid)
1856 struct in6_ifaddr *ifa6;
1858 struct ifnet *ifp_dep = NULL;
1859 int copied = 0, allow_deprecated = 0;
1860 u_char *cp = (u_char *)(nni6 + 1);
1861 int niflags = ni6->ni_flags;
1864 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1865 return (0); /* needless to copy */
1867 IFNET_RLOCK_NOSLEEP();
1868 ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1871 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1873 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1874 if (ifa->ifa_addr->sa_family != AF_INET6)
1876 ifa6 = (struct in6_ifaddr *)ifa;
1878 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1879 allow_deprecated == 0) {
1881 * prefererred address should be put before
1882 * deprecated addresses.
1885 /* record the interface for later search */
1886 if (ifp_dep == NULL)
1890 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1891 allow_deprecated != 0)
1892 continue; /* we now collect deprecated addrs */
1894 /* What do we have to do about ::1? */
1895 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1896 case IPV6_ADDR_SCOPE_LINKLOCAL:
1897 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1900 case IPV6_ADDR_SCOPE_SITELOCAL:
1901 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1904 case IPV6_ADDR_SCOPE_GLOBAL:
1905 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1913 * check if anycast is okay.
1914 * XXX: just experimental. not in the spec.
1916 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1917 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1919 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1920 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1924 /* now we can copy the address */
1925 if (resid < sizeof(struct in6_addr) +
1926 sizeof(u_int32_t)) {
1927 IF_ADDR_RUNLOCK(ifp);
1929 * We give up much more copy.
1930 * Set the truncate flag and return.
1932 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1933 IFNET_RUNLOCK_NOSLEEP();
1938 * Set the TTL of the address.
1939 * The TTL value should be one of the following
1940 * according to the specification:
1942 * 1. The remaining lifetime of a DHCP lease on the
1944 * 2. The remaining Valid Lifetime of a prefix from
1945 * which the address was derived through Stateless
1946 * Autoconfiguration.
1948 * Note that we currently do not support stateful
1949 * address configuration by DHCPv6, so the former
1950 * case can't happen.
1952 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1953 ltime = ND6_INFINITE_LIFETIME;
1955 if (ifa6->ia6_lifetime.ia6t_expire >
1957 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1962 bcopy(<ime, cp, sizeof(u_int32_t));
1963 cp += sizeof(u_int32_t);
1965 /* copy the address itself */
1966 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1967 sizeof(struct in6_addr));
1968 in6_clearscope((struct in6_addr *)cp); /* XXX */
1969 cp += sizeof(struct in6_addr);
1971 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1972 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1974 IF_ADDR_RUNLOCK(ifp);
1975 if (ifp0) /* we need search only on the specified IF */
1979 if (allow_deprecated == 0 && ifp_dep != NULL) {
1981 allow_deprecated = 1;
1986 IFNET_RUNLOCK_NOSLEEP();
1992 * XXX almost dup'ed code with rip6_input.
1995 icmp6_rip6_input(struct mbuf **mp, int off)
1997 struct mbuf *m = *mp;
1998 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2000 struct inpcb *last = NULL;
2001 struct sockaddr_in6 fromsa;
2002 struct icmp6_hdr *icmp6;
2003 struct mbuf *opts = NULL;
2005 #ifndef PULLDOWN_TEST
2006 /* this is assumed to be safe. */
2007 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
2009 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
2010 if (icmp6 == NULL) {
2011 /* m is already reclaimed */
2012 return (IPPROTO_DONE);
2017 * XXX: the address may have embedded scope zone ID, which should be
2018 * hidden from applications.
2020 bzero(&fromsa, sizeof(fromsa));
2021 fromsa.sin6_family = AF_INET6;
2022 fromsa.sin6_len = sizeof(struct sockaddr_in6);
2023 fromsa.sin6_addr = ip6->ip6_src;
2024 if (sa6_recoverscope(&fromsa)) {
2026 return (IPPROTO_DONE);
2029 INP_INFO_RLOCK(&V_ripcbinfo);
2030 LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2031 if ((in6p->inp_vflag & INP_IPV6) == 0)
2033 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2035 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2036 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2038 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2039 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2042 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2043 in6p->in6p_icmp6filt)) {
2048 struct mbuf *n = NULL;
2051 * Recent network drivers tend to allocate a single
2052 * mbuf cluster, rather than to make a couple of
2053 * mbufs without clusters. Also, since the IPv6 code
2054 * path tries to avoid m_pullup(), it is highly
2055 * probable that we still have an mbuf cluster here
2056 * even though the necessary length can be stored in an
2057 * mbuf's internal buffer.
2058 * Meanwhile, the default size of the receive socket
2059 * buffer for raw sockets is not so large. This means
2060 * the possibility of packet loss is relatively higher
2061 * than before. To avoid this scenario, we copy the
2062 * received data to a separate mbuf that does not use
2063 * a cluster, if possible.
2064 * XXX: it is better to copy the data after stripping
2065 * intermediate headers.
2067 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2068 m->m_len <= MHLEN) {
2069 MGET(n, M_DONTWAIT, m->m_type);
2071 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2072 bcopy(m->m_data, n->m_data,
2074 n->m_len = m->m_len;
2082 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2083 if (last->inp_flags & INP_CONTROLOPTS)
2084 ip6_savecontrol(last, n, &opts);
2085 /* strip intermediate headers */
2087 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2088 if (sbappendaddr_locked(
2089 &last->inp_socket->so_rcv,
2090 (struct sockaddr *)&fromsa, n, opts)
2092 /* should notify about lost packet */
2098 &last->inp_socket->so_rcv);
2100 sorwakeup_locked(last->inp_socket);
2107 INP_INFO_RUNLOCK(&V_ripcbinfo);
2109 if (last->inp_flags & INP_CONTROLOPTS)
2110 ip6_savecontrol(last, m, &opts);
2111 /* strip intermediate headers */
2114 /* avoid using mbuf clusters if possible (see above) */
2115 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2116 m->m_len <= MHLEN) {
2119 MGET(n, M_DONTWAIT, m->m_type);
2121 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2122 bcopy(m->m_data, n->m_data, m->m_len);
2123 n->m_len = m->m_len;
2133 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2134 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2135 (struct sockaddr *)&fromsa, m, opts) == 0) {
2139 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2141 sorwakeup_locked(last->inp_socket);
2145 IP6STAT_DEC(ip6s_delivered);
2147 return IPPROTO_DONE;
2151 * Reflect the ip6 packet back to the source.
2152 * OFF points to the icmp6 header, counted from the top of the mbuf.
2155 icmp6_reflect(struct mbuf *m, size_t off)
2157 struct ip6_hdr *ip6;
2158 struct icmp6_hdr *icmp6;
2159 struct in6_ifaddr *ia = NULL;
2162 struct ifnet *outif = NULL;
2163 struct in6_addr origdst, src, *srcp = NULL;
2165 /* too short to reflect */
2166 if (off < sizeof(struct ip6_hdr)) {
2168 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2169 (u_long)off, (u_long)sizeof(struct ip6_hdr),
2170 __FILE__, __LINE__));
2175 * If there are extra headers between IPv6 and ICMPv6, strip
2176 * off that header first.
2179 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2180 panic("assumption failed in icmp6_reflect");
2182 if (off > sizeof(struct ip6_hdr)) {
2184 struct ip6_hdr nip6;
2186 l = off - sizeof(struct ip6_hdr);
2187 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2189 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2191 if ((m = m_pullup(m, l)) == NULL)
2194 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2195 } else /* off == sizeof(struct ip6_hdr) */ {
2197 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2199 if ((m = m_pullup(m, l)) == NULL)
2203 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2204 ip6 = mtod(m, struct ip6_hdr *);
2205 ip6->ip6_nxt = IPPROTO_ICMPV6;
2206 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2207 type = icmp6->icmp6_type; /* keep type for statistics */
2208 code = icmp6->icmp6_code; /* ditto. */
2210 origdst = ip6->ip6_dst;
2212 * ip6_input() drops a packet if its src is multicast.
2213 * So, the src is never multicast.
2215 ip6->ip6_dst = ip6->ip6_src;
2218 * If the incoming packet was addressed directly to us (i.e. unicast),
2219 * use dst as the src for the reply.
2220 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2221 * (for example) when we encounter an error while forwarding procedure
2222 * destined to a duplicated address of ours.
2223 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2224 * procedure of an outgoing packet of our own, in which case we need
2225 * to search in the ifaddr list.
2227 if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2228 if ((ia = ip6_getdstifaddr(m))) {
2229 if (!(ia->ia6_flags &
2230 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2231 srcp = &ia->ia_addr.sin6_addr;
2233 struct sockaddr_in6 d;
2235 bzero(&d, sizeof(d));
2236 d.sin6_family = AF_INET6;
2237 d.sin6_len = sizeof(d);
2238 d.sin6_addr = origdst;
2239 ia = (struct in6_ifaddr *)
2240 ifa_ifwithaddr((struct sockaddr *)&d);
2243 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2244 srcp = &ia->ia_addr.sin6_addr;
2251 struct sockaddr_in6 sin6;
2252 struct route_in6 ro;
2255 * This case matches to multicasts, our anycast, or unicasts
2256 * that we do not own. Select a source address based on the
2257 * source address of the erroneous packet.
2259 bzero(&sin6, sizeof(sin6));
2260 sin6.sin6_family = AF_INET6;
2261 sin6.sin6_len = sizeof(sin6);
2262 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2264 bzero(&ro, sizeof(ro));
2265 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2267 RTFREE(ro.ro_rt); /* XXX: we could use this */
2269 char ip6buf[INET6_ADDRSTRLEN];
2271 "icmp6_reflect: source can't be determined: "
2272 "dst=%s, error=%d\n",
2273 ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2279 ip6->ip6_src = *srcp;
2281 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2282 ip6->ip6_vfc |= IPV6_VERSION;
2283 ip6->ip6_nxt = IPPROTO_ICMPV6;
2285 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2286 else if (m->m_pkthdr.rcvif) {
2287 /* XXX: This may not be the outgoing interface */
2288 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2290 ip6->ip6_hlim = V_ip6_defhlim;
2292 icmp6->icmp6_cksum = 0;
2293 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2294 sizeof(struct ip6_hdr), plen);
2297 * XXX option handling
2300 m->m_flags &= ~(M_BCAST|M_MCAST);
2304 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2306 icmp6_ifoutstat_inc(outif, type, code);
2309 ifa_free(&ia->ia_ifa);
2314 ifa_free(&ia->ia_ifa);
2320 icmp6_fasttimo(void)
2327 icmp6_slowtimo(void)
2334 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2335 struct in6_addr *tgt6)
2337 static char buf[1024];
2338 char ip6bufs[INET6_ADDRSTRLEN];
2339 char ip6bufd[INET6_ADDRSTRLEN];
2340 char ip6buft[INET6_ADDRSTRLEN];
2341 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2342 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2343 ip6_sprintf(ip6buft, tgt6));
2348 icmp6_redirect_input(struct mbuf *m, int off)
2351 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2352 struct nd_redirect *nd_rd;
2353 int icmp6len = ntohs(ip6->ip6_plen);
2354 char *lladdr = NULL;
2356 struct rtentry *rt = NULL;
2359 struct in6_addr src6 = ip6->ip6_src;
2360 struct in6_addr redtgt6;
2361 struct in6_addr reddst6;
2362 union nd_opts ndopts;
2363 char ip6buf[INET6_ADDRSTRLEN];
2366 KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2368 ifp = m->m_pkthdr.rcvif;
2370 /* XXX if we are router, we don't update route by icmp6 redirect */
2371 if (V_ip6_forwarding)
2373 if (!V_icmp6_rediraccept)
2376 #ifndef PULLDOWN_TEST
2377 IP6_EXTHDR_CHECK(m, off, icmp6len,);
2378 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2380 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2381 if (nd_rd == NULL) {
2382 ICMP6STAT_INC(icp6s_tooshort);
2386 redtgt6 = nd_rd->nd_rd_target;
2387 reddst6 = nd_rd->nd_rd_dst;
2389 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2390 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2395 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2397 "ICMP6 redirect sent from %s rejected; "
2398 "must be from linklocal\n",
2399 ip6_sprintf(ip6buf, &src6)));
2402 if (ip6->ip6_hlim != 255) {
2404 "ICMP6 redirect sent from %s rejected; "
2405 "hlim=%d (must be 255)\n",
2406 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2410 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2411 struct sockaddr_in6 sin6;
2412 struct in6_addr *gw6;
2414 bzero(&sin6, sizeof(sin6));
2415 sin6.sin6_family = AF_INET6;
2416 sin6.sin6_len = sizeof(struct sockaddr_in6);
2417 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2418 rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2420 if (rt->rt_gateway == NULL ||
2421 rt->rt_gateway->sa_family != AF_INET6) {
2424 "ICMP6 redirect rejected; no route "
2425 "with inet6 gateway found for redirect dst: %s\n",
2426 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2430 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2431 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2434 "ICMP6 redirect rejected; "
2435 "not equal to gw-for-src=%s (must be same): "
2437 ip6_sprintf(ip6buf, gw6),
2438 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2443 "ICMP6 redirect rejected; "
2444 "no route found for redirect dst: %s\n",
2445 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2451 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2453 "ICMP6 redirect rejected; "
2454 "redirect dst must be unicast: %s\n",
2455 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2459 is_router = is_onlink = 0;
2460 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2461 is_router = 1; /* router case */
2462 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2463 is_onlink = 1; /* on-link destination case */
2464 if (!is_router && !is_onlink) {
2466 "ICMP6 redirect rejected; "
2467 "neither router case nor onlink case: %s\n",
2468 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2471 /* validation passed */
2473 icmp6len -= sizeof(*nd_rd);
2474 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2475 if (nd6_options(&ndopts) < 0) {
2476 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2477 __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2478 /* nd6_options have incremented stats */
2482 if (ndopts.nd_opts_tgt_lladdr) {
2483 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2484 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2487 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2488 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2489 "(if %d, icmp6 packet %d): %s\n",
2490 __func__, ip6_sprintf(ip6buf, &redtgt6),
2491 ifp->if_addrlen, lladdrlen - 2,
2492 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2497 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2498 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2500 if (!is_onlink) { /* better router case. perform rtredirect. */
2501 /* perform rtredirect */
2502 struct sockaddr_in6 sdst;
2503 struct sockaddr_in6 sgw;
2504 struct sockaddr_in6 ssrc;
2507 bzero(&sdst, sizeof(sdst));
2508 bzero(&sgw, sizeof(sgw));
2509 bzero(&ssrc, sizeof(ssrc));
2510 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2511 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2512 sizeof(struct sockaddr_in6);
2513 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2514 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2515 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2516 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2517 in6_rtredirect((struct sockaddr *)&sdst,
2518 (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2519 RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2522 /* finally update cached route in each socket via pfctlinput */
2524 struct sockaddr_in6 sdst;
2526 bzero(&sdst, sizeof(sdst));
2527 sdst.sin6_family = AF_INET6;
2528 sdst.sin6_len = sizeof(struct sockaddr_in6);
2529 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2530 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2532 key_sa_routechange((struct sockaddr *)&sdst);
2541 ICMP6STAT_INC(icp6s_badredirect);
2546 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2548 struct ifnet *ifp; /* my outgoing interface */
2549 struct in6_addr *ifp_ll6;
2550 struct in6_addr *router_ll6;
2551 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2552 struct mbuf *m = NULL; /* newly allocated one */
2554 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2555 struct nd_redirect *nd_rd;
2556 struct llentry *ln = NULL;
2559 struct ifnet *outif = NULL;
2560 struct sockaddr_in6 src_sa;
2562 icmp6_errcount(ND_REDIRECT, 0);
2564 /* if we are not router, we don't send icmp6 redirect */
2565 if (!V_ip6_forwarding)
2569 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2574 * the source address must identify a neighbor, and
2575 * the destination address must not be a multicast address
2576 * [RFC 2461, sec 8.2]
2578 sip6 = mtod(m0, struct ip6_hdr *);
2579 bzero(&src_sa, sizeof(src_sa));
2580 src_sa.sin6_family = AF_INET6;
2581 src_sa.sin6_len = sizeof(src_sa);
2582 src_sa.sin6_addr = sip6->ip6_src;
2583 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2585 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2586 goto fail; /* what should we do here? */
2589 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2593 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2594 * we almost always ask for an mbuf cluster for simplicity.
2595 * (MHLEN < IPV6_MMTU is almost always true)
2597 #if IPV6_MMTU >= MCLBYTES
2598 # error assumption failed about IPV6_MMTU and MCLBYTES
2600 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2601 if (m && IPV6_MMTU >= MHLEN)
2602 MCLGET(m, M_DONTWAIT);
2605 M_SETFIB(m, rt->rt_fibnum);
2606 m->m_pkthdr.rcvif = NULL;
2608 maxlen = M_TRAILINGSPACE(m);
2609 maxlen = min(IPV6_MMTU, maxlen);
2610 /* just for safety */
2611 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2612 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2617 /* get ip6 linklocal address for ifp(my outgoing interface). */
2618 struct in6_ifaddr *ia;
2619 if ((ia = in6ifa_ifpforlinklocal(ifp,
2621 IN6_IFF_ANYCAST)) == NULL)
2623 ifp_ll6 = &ia->ia_addr.sin6_addr;
2624 /* XXXRW: reference released prematurely. */
2625 ifa_free(&ia->ia_ifa);
2628 /* get ip6 linklocal address for the router. */
2629 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2630 struct sockaddr_in6 *sin6;
2631 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2632 router_ll6 = &sin6->sin6_addr;
2633 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2634 router_ll6 = (struct in6_addr *)NULL;
2636 router_ll6 = (struct in6_addr *)NULL;
2639 ip6 = mtod(m, struct ip6_hdr *);
2641 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2642 ip6->ip6_vfc |= IPV6_VERSION;
2643 /* ip6->ip6_plen will be set later */
2644 ip6->ip6_nxt = IPPROTO_ICMPV6;
2645 ip6->ip6_hlim = 255;
2646 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2647 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2648 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2651 nd_rd = (struct nd_redirect *)(ip6 + 1);
2652 nd_rd->nd_rd_type = ND_REDIRECT;
2653 nd_rd->nd_rd_code = 0;
2654 nd_rd->nd_rd_reserved = 0;
2655 if (rt->rt_flags & RTF_GATEWAY) {
2657 * nd_rd->nd_rd_target must be a link-local address in
2658 * better router cases.
2662 bcopy(router_ll6, &nd_rd->nd_rd_target,
2663 sizeof(nd_rd->nd_rd_target));
2664 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2665 sizeof(nd_rd->nd_rd_dst));
2667 /* make sure redtgt == reddst */
2668 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2669 sizeof(nd_rd->nd_rd_target));
2670 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2671 sizeof(nd_rd->nd_rd_dst));
2674 p = (u_char *)(nd_rd + 1);
2680 /* target lladdr option */
2682 struct nd_opt_hdr *nd_opt;
2685 IF_AFDATA_RLOCK(ifp);
2686 ln = nd6_lookup(router_ll6, 0, ifp);
2687 IF_AFDATA_RUNLOCK(ifp);
2691 len = sizeof(*nd_opt) + ifp->if_addrlen;
2692 len = (len + 7) & ~7; /* round by 8 */
2694 if (len + (p - (u_char *)ip6) > maxlen)
2697 if (ln->la_flags & LLE_VALID) {
2698 nd_opt = (struct nd_opt_hdr *)p;
2699 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2700 nd_opt->nd_opt_len = len >> 3;
2701 lladdr = (char *)(nd_opt + 1);
2702 bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2710 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2712 /* just to be safe */
2713 #ifdef M_DECRYPTED /*not openbsd*/
2714 if (m0->m_flags & M_DECRYPTED)
2717 if (p - (u_char *)ip6 > maxlen)
2721 /* redirected header option */
2723 struct nd_opt_rd_hdr *nd_opt_rh;
2726 * compute the maximum size for icmp6 redirect header option.
2727 * XXX room for auth header?
2729 len = maxlen - (p - (u_char *)ip6);
2732 /* This is just for simplicity. */
2733 if (m0->m_pkthdr.len != m0->m_len) {
2735 m_freem(m0->m_next);
2738 m0->m_pkthdr.len = m0->m_len;
2742 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2743 * about padding/truncate rule for the original IP packet.
2744 * From the discussion on IPv6imp in Feb 1999,
2745 * the consensus was:
2746 * - "attach as much as possible" is the goal
2747 * - pad if not aligned (original size can be guessed by
2748 * original ip6 header)
2749 * Following code adds the padding if it is simple enough,
2750 * and truncates if not.
2752 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2753 panic("assumption failed in %s:%d", __FILE__,
2756 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2757 /* not enough room, truncate */
2758 m0->m_pkthdr.len = m0->m_len = len -
2761 /* enough room, pad or truncate */
2764 extra = m0->m_pkthdr.len % 8;
2766 /* pad if easy enough, truncate if not */
2767 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2769 m0->m_len += (8 - extra);
2770 m0->m_pkthdr.len += (8 - extra);
2773 m0->m_pkthdr.len -= extra;
2777 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2778 m0->m_pkthdr.len = m0->m_len = len -
2782 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2783 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2784 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2785 nd_opt_rh->nd_opt_rh_len = len >> 3;
2786 p += sizeof(*nd_opt_rh);
2787 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2789 /* connect m0 to m */
2790 m_tag_delete_chain(m0, NULL);
2791 m0->m_flags &= ~M_PKTHDR;
2793 m->m_pkthdr.len = m->m_len + m0->m_len;
2802 /* XXX: clear embedded link IDs in the inner header */
2803 in6_clearscope(&sip6->ip6_src);
2804 in6_clearscope(&sip6->ip6_dst);
2805 in6_clearscope(&nd_rd->nd_rd_target);
2806 in6_clearscope(&nd_rd->nd_rd_dst);
2808 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2810 nd_rd->nd_rd_cksum = 0;
2811 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2812 sizeof(*ip6), ntohs(ip6->ip6_plen));
2814 if (send_sendso_input_hook != NULL) {
2815 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2819 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2820 m_tag_prepend(m, mtag);
2823 /* send the packet to outside... */
2824 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2826 icmp6_ifstat_inc(outif, ifs6_out_msg);
2827 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2829 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2841 * ICMPv6 socket option processing.
2844 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2848 struct inpcb *inp = sotoinpcb(so);
2849 int level, op, optname;
2852 level = sopt->sopt_level;
2853 op = sopt->sopt_dir;
2854 optname = sopt->sopt_name;
2855 optlen = sopt->sopt_valsize;
2857 level = op = optname = optlen = 0;
2859 if (level != IPPROTO_ICMPV6) {
2868 struct icmp6_filter ic6f;
2870 if (optlen != sizeof(ic6f)) {
2874 error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2877 *inp->in6p_icmp6filt = ic6f;
2884 error = ENOPROTOOPT;
2893 struct icmp6_filter ic6f;
2896 ic6f = *inp->in6p_icmp6filt;
2898 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2903 error = ENOPROTOOPT;
2913 * Perform rate limit check.
2914 * Returns 0 if it is okay to send the icmp6 packet.
2915 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2918 * XXX per-destination/type check necessary?
2920 * dst - not used at this moment
2921 * type - not used at this moment
2922 * code - not used at this moment
2925 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2930 ret = 0; /* okay to send */
2933 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2934 V_icmp6errppslim)) {
2935 /* The packet is subject to rate limit */