1/* $NetBSD: ip6_forward.c,v 1.81 2016/08/31 09:14:47 ozaki-r Exp $ */
2/* $KAME: ip6_forward.c,v 1.109 2002/09/11 08:10:17 sakane Exp $ */
3
4/*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__KERNEL_RCSID(0, "$NetBSD: ip6_forward.c,v 1.81 2016/08/31 09:14:47 ozaki-r Exp $");
35
36#ifdef _KERNEL_OPT
37#include "opt_gateway.h"
38#include "opt_ipsec.h"
39#endif
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/mbuf.h>
44#include <sys/domain.h>
45#include <sys/protosw.h>
46#include <sys/socket.h>
47#include <sys/errno.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/syslog.h>
51#include <sys/percpu.h>
52
53#include <net/if.h>
54#include <net/route.h>
55#include <net/pfil.h>
56
57#include <netinet/in.h>
58#include <netinet/in_var.h>
59#include <netinet/ip_var.h>
60#include <netinet/ip6.h>
61#include <netinet6/ip6_var.h>
62#include <netinet6/ip6_private.h>
63#include <netinet6/scope6_var.h>
64#include <netinet/icmp6.h>
65#include <netinet6/nd6.h>
66
67#ifdef IPSEC
68#include <netipsec/ipsec.h>
69#include <netipsec/ipsec6.h>
70#include <netipsec/key.h>
71#include <netipsec/xform.h>
72#endif /* IPSEC */
73
74#include <net/net_osdep.h>
75
76extern percpu_t *ip6_forward_rt_percpu;
77
78extern pfil_head_t *inet6_pfil_hook; /* XXX */
79
80static void __printflike(4, 5)
81ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp,
82 const struct ifnet *dstifp, const char *fmt, ...)
83{
84 char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
85 char reason[256];
86 va_list ap;
87 uint64_t *ip6s;
88
89 /* update statistics */
90 ip6s = IP6_STAT_GETREF();
91 ip6s[IP6_STAT_CANTFORWARD]++;
92 if (dstifp)
93 ip6s[IP6_STAT_BADSCOPE]++;
94 IP6_STAT_PUTREF();
95
96 if (dstifp)
97 in6_ifstat_inc(dstifp, ifs6_in_discard);
98
99 if (ip6_log_time + ip6_log_interval >= time_uptime)
100 return;
101 ip6_log_time = time_uptime;
102
103 va_start(ap, fmt);
104 vsnprintf(reason, sizeof(reason), fmt, ap);
105 va_end(ap);
106
107 log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n",
108 IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?",
109 IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?",
110 ip6->ip6_nxt, reason);
111}
112
113/*
114 * Forward a packet. If some error occurs return the sender
115 * an icmp packet. Note we can't always generate a meaningful
116 * icmp message because icmp doesn't have a large enough repertoire
117 * of codes and types.
118 *
119 * If not forwarding, just drop the packet. This could be confusing
120 * if ipforwarding was zero but some routing protocol was advancing
121 * us as a gateway to somewhere. However, we must let the routing
122 * protocol deal with that.
123 *
124 */
125
126void
127ip6_forward(struct mbuf *m, int srcrt)
128{
129 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
130 const struct sockaddr_in6 *dst;
131 struct rtentry *rt;
132 int error = 0, type = 0, code = 0;
133 struct mbuf *mcopy = NULL;
134 struct ifnet *origifp; /* maybe unnecessary */
135 uint32_t inzone, outzone;
136 struct in6_addr src_in6, dst_in6;
137 struct ifnet *rcvif = NULL;
138 struct psref psref;
139 struct route *ro;
140#ifdef IPSEC
141 int needipsec = 0;
142 struct secpolicy *sp = NULL;
143#endif
144
145 /*
146 * Clear any in-bound checksum flags for this packet.
147 */
148 m->m_pkthdr.csum_flags = 0;
149
150 rcvif = m_get_rcvif_psref(m, &psref);
151 if (__predict_false(rcvif == NULL))
152 goto drop;
153
154 /*
155 * Do not forward packets to multicast destination (should be handled
156 * by ip6_mforward().
157 * Do not forward packets with unspecified source. It was discussed
158 * in July 2000, on ipngwg mailing list.
159 */
160 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
161 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
162 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
163 ip6_cantforward(ip6, rcvif, NULL,
164 ((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" :
165 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" :
166 "unspec/src");
167 goto drop;
168 }
169
170 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
171 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
172 icmp6_error(m, ICMP6_TIME_EXCEEDED,
173 ICMP6_TIME_EXCEED_TRANSIT, 0);
174 goto out;
175 }
176 ip6->ip6_hlim -= IPV6_HLIMDEC;
177
178 /*
179 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
180 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
181 * we need to generate an ICMP6 message to the src.
182 * Thanks to M_EXT, in most cases copy will not occur.
183 *
184 * It is important to save it before IPsec processing as IPsec
185 * processing may modify the mbuf.
186 */
187 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
188
189#ifdef IPSEC
190 if (ipsec_used) {
191 /* Check the security policy (SP) for the packet */
192
193 sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error);
194 if (error != 0) {
195 /*
196 * Hack: -EINVAL is used to signal that a packet
197 * should be silently discarded. This is typically
198 * because we asked key management for an SA and
199 * it was delayed (e.g. kicked up to IKE).
200 */
201 if (error == -EINVAL)
202 error = 0;
203 goto freecopy;
204 }
205 }
206#endif /* IPSEC */
207
208 ro = percpu_getref(ip6_forward_rt_percpu);
209 if (srcrt) {
210 union {
211 struct sockaddr dst;
212 struct sockaddr_in6 dst6;
213 } u;
214
215 sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
216 if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
217 IP6_STATINC(IP6_STAT_NOROUTE);
218 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
219 if (mcopy) {
220 icmp6_error(mcopy, ICMP6_DST_UNREACH,
221 ICMP6_DST_UNREACH_NOROUTE, 0);
222 }
223 percpu_putref(ip6_forward_rt_percpu);
224 goto drop;
225 }
226 } else if ((rt = rtcache_validate(ro)) == NULL &&
227 (rt = rtcache_update(ro, 1)) == NULL) {
228 /*
229 * rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to
230 * ip6->ip6_dst
231 */
232 IP6_STATINC(IP6_STAT_NOROUTE);
233 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
234 if (mcopy) {
235 icmp6_error(mcopy, ICMP6_DST_UNREACH,
236 ICMP6_DST_UNREACH_NOROUTE, 0);
237 }
238 percpu_putref(ip6_forward_rt_percpu);
239 goto drop;
240 }
241 dst = satocsin6(rtcache_getdst(ro));
242 percpu_putref(ip6_forward_rt_percpu);
243
244 /*
245 * Source scope check: if a packet can't be delivered to its
246 * destination for the reason that the destination is beyond the scope
247 * of the source address, discard the packet and return an icmp6
248 * destination unreachable error with Code 2 (beyond scope of source
249 * address). We use a local copy of ip6_src, since in6_setscope()
250 * will possibly modify its first argument.
251 * [draft-ietf-ipngwg-icmp-v3-07, Section 3.1]
252 */
253 src_in6 = ip6->ip6_src;
254 inzone = outzone = ~0;
255 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 ||
256 in6_setscope(&src_in6, rcvif, &inzone) != 0 ||
257 inzone != outzone) {
258 ip6_cantforward(ip6, rcvif, rt->rt_ifp,
259 "src[%s] inzone %d outzone %d",
260 in6_getscopename(&ip6->ip6_src), inzone, outzone);
261 if (mcopy)
262 icmp6_error(mcopy, ICMP6_DST_UNREACH,
263 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
264 goto drop;
265 }
266
267#ifdef IPSEC
268 /*
269 * If we need to encapsulate the packet, do it here
270 * ipsec6_proces_packet will send the packet using ip6_output
271 */
272 if (needipsec) {
273 int s = splsoftnet();
274 error = ipsec6_process_packet(m, sp->req);
275 splx(s);
276 if (mcopy)
277 goto freecopy;
278 }
279#endif
280
281 /*
282 * Destination scope check: if a packet is going to break the scope
283 * zone of packet's destination address, discard it. This case should
284 * usually be prevented by appropriately-configured routing table, but
285 * we need an explicit check because we may mistakenly forward the
286 * packet to a different zone by (e.g.) a default route.
287 */
288 dst_in6 = ip6->ip6_dst;
289 inzone = outzone = ~0;
290 if (in6_setscope(&dst_in6, rcvif, &inzone) != 0 ||
291 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
292 inzone != outzone) {
293 ip6_cantforward(ip6, rcvif, rt->rt_ifp,
294 "dst[%s] inzone %d outzone %d",
295 in6_getscopename(&ip6->ip6_dst), inzone, outzone);
296 if (mcopy)
297 icmp6_error(mcopy, ICMP6_DST_UNREACH,
298 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
299 goto drop;
300 }
301
302 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
303 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
304 if (mcopy) {
305 u_long mtu;
306
307 mtu = IN6_LINKMTU(rt->rt_ifp);
308 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
309 }
310 goto drop;
311 }
312
313 if (rt->rt_flags & RTF_GATEWAY)
314 dst = (struct sockaddr_in6 *)rt->rt_gateway;
315
316 /*
317 * If we are to forward the packet using the same interface
318 * as one we got the packet from, perhaps we should send a redirect
319 * to sender to shortcut a hop.
320 * Only send redirect if source is sending directly to us,
321 * and if packet was not source routed (or has any options).
322 * Also, don't send redirect if forwarding using a route
323 * modified by a redirect.
324 */
325 if (rt->rt_ifp == rcvif && !srcrt && ip6_sendredirects &&
326 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
327 ro = percpu_getref(ip6_forward_rt_percpu);
328 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
329 nd6_is_addr_neighbor(satocsin6(rtcache_getdst(ro)),
330 rt->rt_ifp)) {
331 percpu_putref(ip6_forward_rt_percpu);
332 /*
333 * If the incoming interface is equal to the outgoing
334 * one, the link attached to the interface is
335 * point-to-point, and the IPv6 destination is
336 * regarded as on-link on the link, then it will be
337 * highly probable that the destination address does
338 * not exist on the link and that the packet is going
339 * to loop. Thus, we immediately drop the packet and
340 * send an ICMPv6 error message.
341 * For other routing loops, we dare to let the packet
342 * go to the loop, so that a remote diagnosing host
343 * can detect the loop by traceroute.
344 * type/code is based on suggestion by Rich Draves.
345 * not sure if it is the best pick.
346 */
347 icmp6_error(mcopy, ICMP6_DST_UNREACH,
348 ICMP6_DST_UNREACH_ADDR, 0);
349 goto drop;
350 }
351 percpu_putref(ip6_forward_rt_percpu);
352 type = ND_REDIRECT;
353 }
354
355 /*
356 * Fake scoped addresses. Note that even link-local source or
357 * destinaion can appear, if the originating node just sends the
358 * packet to us (without address resolution for the destination).
359 * Since both icmp6_error and icmp6_redirect_output fill the embedded
360 * link identifiers, we can do this stuff after making a copy for
361 * returning an error.
362 */
363 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
364 /*
365 * See corresponding comments in ip6_output.
366 * XXX: but is it possible that ip6_forward() sends a packet
367 * to a loopback interface? I don't think so, and thus
368 * I bark here. (jinmei@kame.net)
369 * XXX: it is common to route invalid packets to loopback.
370 * also, the codepath will be visited on use of ::1 in
371 * rthdr. (itojun)
372 */
373#if 1
374 if (0)
375#else
376 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
377#endif
378 {
379 printf("ip6_forward: outgoing interface is loopback. "
380 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
381 ip6_sprintf(&ip6->ip6_src),
382 ip6_sprintf(&ip6->ip6_dst),
383 ip6->ip6_nxt, if_name(rcvif),
384 if_name(rt->rt_ifp));
385 }
386
387 /* we can just use rcvif in forwarding. */
388 origifp = rcvif;
389 }
390 else
391 origifp = rt->rt_ifp;
392 /*
393 * clear embedded scope identifiers if necessary.
394 * in6_clearscope will touch the addresses only when necessary.
395 */
396 in6_clearscope(&ip6->ip6_src);
397 in6_clearscope(&ip6->ip6_dst);
398
399 /*
400 * Run through list of hooks for output packets.
401 */
402 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp,
403 PFIL_OUT)) != 0)
404 goto senderr;
405 if (m == NULL)
406 goto freecopy;
407 ip6 = mtod(m, struct ip6_hdr *);
408
409 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
410 if (error) {
411 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
412 IP6_STATINC(IP6_STAT_CANTFORWARD);
413 } else {
414 IP6_STATINC(IP6_STAT_FORWARD);
415 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
416 if (type)
417 IP6_STATINC(IP6_STAT_REDIRECTSENT);
418 else {
419#ifdef GATEWAY
420 ro = percpu_getref(ip6_forward_rt_percpu);
421 if (m->m_flags & M_CANFASTFWD)
422 ip6flow_create(ro, m);
423 percpu_putref(ip6_forward_rt_percpu);
424#endif
425 if (mcopy)
426 goto freecopy;
427 }
428 }
429
430 senderr:
431 if (mcopy == NULL)
432 goto out;
433 switch (error) {
434 case 0:
435 if (type == ND_REDIRECT) {
436 icmp6_redirect_output(mcopy, rt);
437 goto out;
438 }
439 goto freecopy;
440
441 case EMSGSIZE:
442 /* xxx MTU is constant in PPP? */
443 goto freecopy;
444
445 case ENOBUFS:
446 /* Tell source to slow down like source quench in IP? */
447 goto freecopy;
448
449 case ENETUNREACH: /* shouldn't happen, checked above */
450 case EHOSTUNREACH:
451 case ENETDOWN:
452 case EHOSTDOWN:
453 default:
454 type = ICMP6_DST_UNREACH;
455 code = ICMP6_DST_UNREACH_ADDR;
456 break;
457 }
458 icmp6_error(mcopy, type, code, 0);
459 goto out;
460
461 freecopy:
462 m_freem(mcopy);
463 goto out;
464 drop:
465 m_freem(m);
466 out:
467 if (rcvif != NULL)
468 m_put_rcvif_psref(rcvif, &psref);
469 return;
470}
471