2 /* $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 #include "opt_ip6fw.h"
35 #include "opt_inet6.h"
36 #include "opt_ipsec.h"
37 #include "opt_ipstealth.h"
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/malloc.h>
43 #include <sys/domain.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/errno.h>
48 #include <sys/kernel.h>
49 #include <sys/syslog.h>
52 #include <net/route.h>
55 #include <netinet/in.h>
56 #include <netinet/in_var.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #include <netinet/ip_var.h>
60 #include <netinet6/in6_var.h>
61 #include <netinet/ip6.h>
62 #include <netinet6/ip6_var.h>
63 #include <netinet6/scope6_var.h>
64 #include <netinet/icmp6.h>
65 #include <netinet6/nd6.h>
67 #include <netinet/in_pcb.h>
70 #include <netinet6/ipsec.h>
72 #include <netinet6/ipsec6.h>
74 #include <netkey/key.h>
78 #include <netipsec/ipsec.h>
79 #include <netipsec/ipsec6.h>
80 #include <netipsec/key.h>
82 #endif /* FAST_IPSEC */
84 #include <netinet6/ip6_fw.h>
86 #include <net/net_osdep.h>
88 #include <netinet6/ip6protosw.h>
90 struct route_in6 ip6_forward_rt;
93 * Forward a packet. If some error occurs return the sender
94 * an icmp packet. Note we can't always generate a meaningful
95 * icmp message because icmp doesn't have a large enough repertoire
98 * If not forwarding, just drop the packet. This could be confusing
99 * if ipforwarding was zero but some routing protocol was advancing
100 * us as a gateway to somewhere. However, we must let the routing
101 * protocol deal with that.
106 ip6_forward(m, srcrt)
110 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
111 struct sockaddr_in6 *dst = NULL;
112 struct rtentry *rt = NULL;
113 int error, type = 0, code = 0;
114 struct mbuf *mcopy = NULL;
115 struct ifnet *origifp; /* maybe unnecessary */
116 u_int32_t inzone, outzone;
117 struct in6_addr src_in6, dst_in6;
119 struct secpolicy *sp = NULL;
125 * Check AH/ESP integrity.
128 * Don't increment ip6s_cantforward because this is the check
129 * before forwarding packet actually.
131 if (ipsec6_in_reject(m, NULL)) {
132 #if !defined(FAST_IPSEC)
133 ipsec6stat.in_polvio++;
141 * Do not forward packets to multicast destination (should be handled
143 * Do not forward packets with unspecified source. It was discussed
144 * in July 2000, on the ipngwg mailing list.
146 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
147 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
148 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
149 ip6stat.ip6s_cantforward++;
150 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
151 if (ip6_log_time + ip6_log_interval < time_second) {
152 ip6_log_time = time_second;
155 "from %s to %s nxt %d received on %s\n",
156 ip6_sprintf(&ip6->ip6_src),
157 ip6_sprintf(&ip6->ip6_dst),
159 if_name(m->m_pkthdr.rcvif));
168 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
169 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
170 icmp6_error(m, ICMP6_TIME_EXCEEDED,
171 ICMP6_TIME_EXCEED_TRANSIT, 0);
174 ip6->ip6_hlim -= IPV6_HLIMDEC;
181 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
182 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
183 * we need to generate an ICMP6 message to the src.
184 * Thanks to M_EXT, in most cases copy will not occur.
186 * It is important to save it before IPsec processing as IPsec
187 * processing may modify the mbuf.
189 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
192 /* get a security policy for this packet */
193 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
194 IP_FORWARDING, &error);
196 ipsec6stat.out_inval++;
197 ip6stat.ip6s_cantforward++;
200 /* XXX: what icmp ? */
212 switch (sp->policy) {
213 case IPSEC_POLICY_DISCARD:
215 * This packet is just discarded.
217 ipsec6stat.out_polvio++;
218 ip6stat.ip6s_cantforward++;
222 /* XXX: what icmp ? */
230 case IPSEC_POLICY_BYPASS:
231 case IPSEC_POLICY_NONE:
232 /* no need to do IPsec. */
236 case IPSEC_POLICY_IPSEC:
237 if (sp->req == NULL) {
238 /* XXX should be panic ? */
239 printf("ip6_forward: No IPsec request specified.\n");
240 ip6stat.ip6s_cantforward++;
244 /* XXX: what icmp ? */
255 case IPSEC_POLICY_ENTRUST:
257 /* should be panic ?? */
258 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
264 struct ipsecrequest *isr = NULL;
265 struct ipsec_output_state state;
268 * when the kernel forwards a packet, it is not proper to apply
269 * IPsec transport mode to the packet is not proper. this check
271 * at present, if there is even a transport mode SA request in the
272 * security policy, the kernel does not apply IPsec to the packet.
273 * this check is not enough because the following case is valid.
274 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
276 for (isr = sp->req; isr; isr = isr->next) {
277 if (isr->saidx.mode == IPSEC_MODE_ANY)
279 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
284 * if there's no need for tunnel mode IPsec, skip.
291 * All the extension headers will become inaccessible
292 * (since they can be encrypted).
293 * Don't panic, we need no more updates to extension headers
294 * on inner IPv6 packet (since they are now encapsulated).
296 * IPv6 [ESP|AH] IPv6 [extension headers] payload
298 bzero(&state, sizeof(state));
300 state.ro = NULL; /* update at ipsec6_output_tunnel() */
301 state.dst = NULL; /* update at ipsec6_output_tunnel() */
303 error = ipsec6_output_tunnel(&state, sp, 0);
309 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
318 printf("ip6_output (ipsec): error code %d\n", error);
321 /* don't show these error codes to the user */
324 ip6stat.ip6s_cantforward++;
327 /* XXX: what icmp ? */
336 if (ip6 != mtod(m, struct ip6_hdr *)) {
338 * now tunnel mode headers are added. we are originating
339 * packet instead of forwarding the packet.
341 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL,
347 dst = (struct sockaddr_in6 *)state.dst;
348 rt = state.ro ? state.ro->ro_rt : NULL;
349 if (dst != NULL && rt != NULL)
360 dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
362 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */
363 if (ip6_forward_rt.ro_rt == 0 ||
364 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
365 if (ip6_forward_rt.ro_rt) {
366 RTFREE(ip6_forward_rt.ro_rt);
367 ip6_forward_rt.ro_rt = 0;
370 /* this probably fails but give it a try again */
371 rtalloc((struct route *)&ip6_forward_rt);
374 if (ip6_forward_rt.ro_rt == 0) {
375 ip6stat.ip6s_noroute++;
376 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
378 icmp6_error(mcopy, ICMP6_DST_UNREACH,
379 ICMP6_DST_UNREACH_NOROUTE, 0);
384 } else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
385 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
386 if (ip6_forward_rt.ro_rt) {
387 RTFREE(ip6_forward_rt.ro_rt);
388 ip6_forward_rt.ro_rt = 0;
390 bzero(dst, sizeof(*dst));
391 dst->sin6_len = sizeof(struct sockaddr_in6);
392 dst->sin6_family = AF_INET6;
393 dst->sin6_addr = ip6->ip6_dst;
395 rtalloc((struct route *)&ip6_forward_rt);
396 if (ip6_forward_rt.ro_rt == 0) {
397 ip6stat.ip6s_noroute++;
398 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
400 icmp6_error(mcopy, ICMP6_DST_UNREACH,
401 ICMP6_DST_UNREACH_NOROUTE, 0);
407 rt = ip6_forward_rt.ro_rt;
413 * Source scope check: if a packet can't be delivered to its
414 * destination for the reason that the destination is beyond the scope
415 * of the source address, discard the packet and return an icmp6
416 * destination unreachable error with Code 2 (beyond scope of source
417 * address). We use a local copy of ip6_src, since in6_setscope()
418 * will possibly modify its first argument.
419 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
421 src_in6 = ip6->ip6_src;
422 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
423 /* XXX: this should not happen */
424 ip6stat.ip6s_cantforward++;
425 ip6stat.ip6s_badscope++;
429 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
430 ip6stat.ip6s_cantforward++;
431 ip6stat.ip6s_badscope++;
435 if (inzone != outzone
440 ip6stat.ip6s_cantforward++;
441 ip6stat.ip6s_badscope++;
442 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
444 if (ip6_log_time + ip6_log_interval < time_second) {
445 ip6_log_time = time_second;
448 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
449 ip6_sprintf(&ip6->ip6_src),
450 ip6_sprintf(&ip6->ip6_dst),
452 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
455 icmp6_error(mcopy, ICMP6_DST_UNREACH,
456 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
462 * Destination scope check: if a packet is going to break the scope
463 * zone of packet's destination address, discard it. This case should
464 * usually be prevented by appropriately-configured routing table, but
465 * we need an explicit check because we may mistakenly forward the
466 * packet to a different zone by (e.g.) a default route.
468 dst_in6 = ip6->ip6_dst;
469 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
470 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
472 ip6stat.ip6s_cantforward++;
473 ip6stat.ip6s_badscope++;
478 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
479 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
483 struct secpolicy *sp;
488 mtu = IN6_LINKMTU(rt->rt_ifp);
491 * When we do IPsec tunnel ingress, we need to play
492 * with the link value (decrement IPsec header size
493 * from mtu value). The code is much simpler than v4
494 * case, as we have the outgoing interface for
495 * encapsulated packet as "rt->rt_ifp".
497 sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
498 IP_FORWARDING, &ipsecerror);
500 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
501 IPSEC_DIR_OUTBOUND, NULL);
502 if (ipsechdrsiz < mtu)
507 * if mtu becomes less than minimum MTU,
508 * tell minimum MTU (and I'll need to fragment it).
513 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
519 if (rt->rt_flags & RTF_GATEWAY)
520 dst = (struct sockaddr_in6 *)rt->rt_gateway;
523 * If we are to forward the packet using the same interface
524 * as one we got the packet from, perhaps we should send a redirect
525 * to sender to shortcut a hop.
526 * Only send redirect if source is sending directly to us,
527 * and if packet was not source routed (or has any options).
528 * Also, don't send redirect if forwarding using a route
529 * modified by a redirect.
531 if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
535 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
536 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
538 * If the incoming interface is equal to the outgoing
539 * one, and the link attached to the interface is
540 * point-to-point, then it will be highly probable
541 * that a routing loop occurs. Thus, we immediately
542 * drop the packet and send an ICMPv6 error message.
544 * type/code is based on suggestion by Rich Draves.
545 * not sure if it is the best pick.
547 icmp6_error(mcopy, ICMP6_DST_UNREACH,
548 ICMP6_DST_UNREACH_ADDR, 0);
556 * Check with the firewall...
558 if (ip6_fw_enable && ip6_fw_chk_ptr) {
560 /* If ipfw says divert, we have to just drop packet */
561 if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
570 * Fake scoped addresses. Note that even link-local source or
571 * destinaion can appear, if the originating node just sends the
572 * packet to us (without address resolution for the destination).
573 * Since both icmp6_error and icmp6_redirect_output fill the embedded
574 * link identifiers, we can do this stuff after making a copy for
575 * returning an error.
577 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
579 * See corresponding comments in ip6_output.
580 * XXX: but is it possible that ip6_forward() sends a packet
581 * to a loopback interface? I don't think so, and thus
582 * I bark here. (jinmei@kame.net)
583 * XXX: it is common to route invalid packets to loopback.
584 * also, the codepath will be visited on use of ::1 in
590 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
593 printf("ip6_forward: outgoing interface is loopback. "
594 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
595 ip6_sprintf(&ip6->ip6_src),
596 ip6_sprintf(&ip6->ip6_dst),
597 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
598 if_name(rt->rt_ifp));
601 /* we can just use rcvif in forwarding. */
602 origifp = m->m_pkthdr.rcvif;
605 origifp = rt->rt_ifp;
607 * clear embedded scope identifiers if necessary.
608 * in6_clearscope will touch the addresses only when necessary.
610 in6_clearscope(&ip6->ip6_src);
611 in6_clearscope(&ip6->ip6_dst);
613 /* Jump over all PFIL processing if hooks are not active. */
614 if (!PFIL_HOOKED(&inet6_pfil_hook))
617 /* Run through list of hooks for output packets. */
618 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
623 ip6 = mtod(m, struct ip6_hdr *);
626 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
628 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
629 ip6stat.ip6s_cantforward++;
631 ip6stat.ip6s_forward++;
632 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
634 ip6stat.ip6s_redirectsent++;
646 if (type == ND_REDIRECT) {
647 icmp6_redirect_output(mcopy, rt);
653 /* xxx MTU is constant in PPP? */
657 /* Tell source to slow down like source quench in IP? */
660 case ENETUNREACH: /* shouldn't happen, checked above */
665 type = ICMP6_DST_UNREACH;
666 code = ICMP6_DST_UNREACH_ADDR;
669 icmp6_error(mcopy, type, code, 0);