1 /* $NetBSD: ip6_forward.c,v 1.102 2020/08/28 06:32:24 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 1.102 2020/08/28 06:32:24 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/errno.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/syslog.h>
48 #include <sys/percpu.h>
49
50 #include <net/if.h>
51 #include <net/route.h>
52 #include <net/pfil.h>
53
54 #include <netinet/in.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/ip6_private.h>
60 #include <netinet6/scope6_var.h>
61 #include <netinet/icmp6.h>
62 #include <netinet6/nd6.h>
63
64 #ifdef IPSEC
65 #include <netipsec/ipsec.h>
66 #include <netipsec/ipsec6.h>
67 #include <netipsec/key.h>
68 #endif
69
70 extern percpu_t *ip6_forward_rt_percpu;
71
72 extern pfil_head_t *inet6_pfil_hook; /* XXX */
73
74 static void __printflike(4, 5)
ip6_cantforward(const struct ip6_hdr * ip6,const struct ifnet * srcifp,const struct ifnet * dstifp,const char * fmt,...)75 ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp,
76 const struct ifnet *dstifp, const char *fmt, ...)
77 {
78 char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
79 char reason[256];
80 va_list ap;
81 uint64_t *ip6s;
82
83 /* update statistics */
84 ip6s = IP6_STAT_GETREF();
85 ip6s[IP6_STAT_CANTFORWARD]++;
86 if (dstifp)
87 ip6s[IP6_STAT_BADSCOPE]++;
88 IP6_STAT_PUTREF();
89
90 if (dstifp)
91 in6_ifstat_inc(dstifp, ifs6_in_discard);
92
93 if (ip6_log_time + ip6_log_interval >= time_uptime)
94 return;
95 ip6_log_time = time_uptime;
96
97 va_start(ap, fmt);
98 vsnprintf(reason, sizeof(reason), fmt, ap);
99 va_end(ap);
100
101 log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n",
102 IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?",
103 IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?",
104 ip6->ip6_nxt, reason);
105 }
106
107 /*
108 * Forward a packet. If some error occurs return the sender
109 * an icmp packet. Note we can't always generate a meaningful
110 * icmp message because icmp doesn't have a large enough repertoire
111 * of codes and types.
112 *
113 * If not forwarding, just drop the packet. This could be confusing
114 * if ipforwarding was zero but some routing protocol was advancing
115 * us as a gateway to somewhere. However, we must let the routing
116 * protocol deal with that.
117 */
118 void
ip6_forward(struct mbuf * m,int srcrt,struct ifnet * rcvif)119 ip6_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
120 {
121 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
122 const struct sockaddr_in6 *dst;
123 struct rtentry *rt = NULL;
124 int error = 0, type = 0, code = 0;
125 struct mbuf *mcopy = NULL;
126 struct ifnet *origifp; /* maybe unnecessary */
127 uint32_t inzone, outzone;
128 struct in6_addr src_in6, dst_in6;
129 struct route *ro = NULL;
130 #ifdef IPSEC
131 int needipsec = 0;
132 struct secpolicy *sp = NULL;
133 #endif
134
135 /*
136 * Clear any in-bound checksum flags for this packet.
137 */
138 m->m_pkthdr.csum_flags = 0;
139
140 /*
141 * Do not forward packets to multicast destination (should be handled
142 * by ip6_mforward()). Do not forward packets with unspecified source.
143 * It was discussed in July 2000, on ipngwg mailing list.
144 */
145 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
146 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
147 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
148 ip6_cantforward(ip6, rcvif, NULL,
149 ((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" :
150 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" :
151 "unspec/src");
152 goto drop;
153 }
154
155 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
156 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
157 icmp6_error(m, ICMP6_TIME_EXCEEDED,
158 ICMP6_TIME_EXCEED_TRANSIT, 0);
159 goto out;
160 }
161 ip6->ip6_hlim -= IPV6_HLIMDEC;
162
163 /*
164 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
165 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
166 * we need to generate an ICMP6 message to the src.
167 * Thanks to M_EXT, in most cases copy will not occur.
168 *
169 * It is important to save it before IPsec processing as IPsec
170 * processing may modify the mbuf.
171 */
172 mcopy = m_copym(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN),
173 M_DONTWAIT);
174
175 #ifdef IPSEC
176 if (ipsec_used) {
177 /* Check the security policy (SP) for the packet */
178 sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error);
179 if (error != 0) {
180 /*
181 * Hack: -EINVAL is used to signal that a packet
182 * should be silently discarded. This is typically
183 * because we asked key management for an SA and
184 * it was delayed (e.g. kicked up to IKE).
185 */
186 if (error == -EINVAL)
187 error = 0;
188 m_freem(m);
189 IP6_STATINC(IP6_STAT_IPSECDROP_OUT);
190 goto freecopy;
191 }
192 }
193 #endif
194
195 ro = rtcache_percpu_getref(ip6_forward_rt_percpu);
196 if (srcrt) {
197 union {
198 struct sockaddr dst;
199 struct sockaddr_in6 dst6;
200 } u;
201
202 sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
203 rt = rtcache_lookup(ro, &u.dst);
204 if (rt == NULL) {
205 IP6_STATINC(IP6_STAT_NOROUTE);
206 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
207 if (mcopy) {
208 icmp6_error(mcopy, ICMP6_DST_UNREACH,
209 ICMP6_DST_UNREACH_NOROUTE, 0);
210 }
211 goto drop;
212 }
213 } else if ((rt = rtcache_validate(ro)) == NULL &&
214 (rt = rtcache_update(ro, 1)) == NULL) {
215 /*
216 * rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to
217 * ip6->ip6_dst
218 */
219 IP6_STATINC(IP6_STAT_NOROUTE);
220 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
221 if (mcopy) {
222 icmp6_error(mcopy, ICMP6_DST_UNREACH,
223 ICMP6_DST_UNREACH_NOROUTE, 0);
224 }
225 goto drop;
226 }
227 dst = satocsin6(rtcache_getdst(ro));
228
229 /*
230 * Source scope check: if a packet can't be delivered to its
231 * destination for the reason that the destination is beyond the scope
232 * of the source address, discard the packet and return an icmp6
233 * destination unreachable error with Code 2 (beyond scope of source
234 * address). We use a local copy of ip6_src, since in6_setscope()
235 * will possibly modify its first argument.
236 * [draft-ietf-ipngwg-icmp-v3-07, Section 3.1]
237 */
238 src_in6 = ip6->ip6_src;
239 inzone = outzone = ~0;
240 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 ||
241 in6_setscope(&src_in6, rcvif, &inzone) != 0 ||
242 inzone != outzone) {
243 ip6_cantforward(ip6, rcvif, rt->rt_ifp,
244 "src[%s] inzone %d outzone %d",
245 in6_getscopename(&ip6->ip6_src), inzone, outzone);
246 if (mcopy)
247 icmp6_error(mcopy, ICMP6_DST_UNREACH,
248 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
249 goto drop;
250 }
251
252 #ifdef IPSEC
253 /*
254 * If we need to encapsulate the packet, do it here.
255 * ipsec6_process_packet will send the packet using ip6_output.
256 */
257 if (needipsec) {
258 int s = splsoftnet();
259 error = ipsec6_process_packet(m, sp->req, 0);
260 splx(s);
261 /* m is freed */
262 if (mcopy)
263 goto freecopy;
264 goto out;
265 }
266 #endif
267
268 /*
269 * Destination scope check: if a packet is going to break the scope
270 * zone of packet's destination address, discard it. This case should
271 * usually be prevented by appropriately-configured routing table, but
272 * we need an explicit check because we may mistakenly forward the
273 * packet to a different zone by (e.g.) a default route.
274 */
275 dst_in6 = ip6->ip6_dst;
276 inzone = outzone = ~0;
277 if (in6_setscope(&dst_in6, rcvif, &inzone) != 0 ||
278 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
279 inzone != outzone) {
280 ip6_cantforward(ip6, rcvif, rt->rt_ifp,
281 "dst[%s] inzone %d outzone %d",
282 in6_getscopename(&ip6->ip6_dst), inzone, outzone);
283 if (mcopy)
284 icmp6_error(mcopy, ICMP6_DST_UNREACH,
285 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
286 goto drop;
287 }
288
289 if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
290 IP6_STATINC(IP6_STAT_TOOBIG);
291 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
292 if (mcopy)
293 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0,
294 rt->rt_ifp->if_mtu);
295 goto drop;
296 }
297
298 if (rt->rt_flags & RTF_GATEWAY)
299 dst = (struct sockaddr_in6 *)rt->rt_gateway;
300
301 /*
302 * If we are to forward the packet using the same interface
303 * as one we got the packet from, perhaps we should send a redirect
304 * to sender to shortcut a hop.
305 * Only send redirect if source is sending directly to us,
306 * and if packet was not source routed (or has any options).
307 * Also, don't send redirect if forwarding using a route
308 * modified by a redirect.
309 */
310 if (rt->rt_ifp == rcvif && !srcrt && ip6_sendredirects &&
311 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
312 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
313 nd6_is_addr_neighbor(satocsin6(rtcache_getdst(ro)),
314 rt->rt_ifp)) {
315 /*
316 * If the incoming interface is equal to the outgoing
317 * one, the link attached to the interface is
318 * point-to-point, and the IPv6 destination is
319 * regarded as on-link on the link, then it will be
320 * highly probable that the destination address does
321 * not exist on the link and that the packet is going
322 * to loop. Thus, we immediately drop the packet and
323 * send an ICMPv6 error message.
324 * For other routing loops, we dare to let the packet
325 * go to the loop, so that a remote diagnosing host
326 * can detect the loop by traceroute.
327 * type/code is based on suggestion by Rich Draves.
328 * not sure if it is the best pick.
329 */
330 icmp6_error(mcopy, ICMP6_DST_UNREACH,
331 ICMP6_DST_UNREACH_ADDR, 0);
332 goto drop;
333 }
334 type = ND_REDIRECT;
335 }
336
337 /*
338 * Fake scoped addresses. Note that even link-local source or
339 * destination can appear, if the originating node just sends the
340 * packet to us (without address resolution for the destination).
341 * Since both icmp6_error and icmp6_redirect_output fill the embedded
342 * link identifiers, we can do this stuff after making a copy for
343 * returning an error.
344 */
345 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
346 /*
347 * See corresponding comments in ip6_output.
348 * XXX: but is it possible that ip6_forward() sends a packet
349 * to a loopback interface? I don't think so, and thus
350 * I bark here. (jinmei@kame.net)
351 * XXX: it is common to route invalid packets to loopback.
352 * also, the codepath will be visited on use of ::1 in
353 * rthdr. (itojun)
354 */
355 #if 1
356 if (0)
357 #else
358 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
359 #endif
360 {
361 char ip6bufs[INET6_ADDRSTRLEN];
362 char ip6bufd[INET6_ADDRSTRLEN];
363
364 printf("ip6_forward: outgoing interface is loopback. "
365 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
366 IN6_PRINT(ip6bufs, &ip6->ip6_src),
367 IN6_PRINT(ip6bufd, &ip6->ip6_dst),
368 ip6->ip6_nxt, if_name(rcvif),
369 if_name(rt->rt_ifp));
370 }
371
372 /* we can just use rcvif in forwarding. */
373 origifp = rcvif;
374 } else {
375 origifp = rt->rt_ifp;
376 }
377
378 /*
379 * clear embedded scope identifiers if necessary.
380 * in6_clearscope will touch the addresses only when necessary.
381 */
382 in6_clearscope(&ip6->ip6_src);
383 in6_clearscope(&ip6->ip6_dst);
384
385 /*
386 * Run through list of hooks for output packets.
387 */
388 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp,
389 PFIL_OUT)) != 0) {
390 IP6_STATINC(IP6_STAT_PFILDROP_OUT);
391 goto senderr;
392 }
393 if (m == NULL) {
394 IP6_STATINC(IP6_STAT_PFILDROP_OUT);
395 goto freecopy;
396 }
397 ip6 = mtod(m, struct ip6_hdr *);
398
399 error = ip6_if_output(rt->rt_ifp, origifp, m, dst, rt);
400 if (error) {
401 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
402 IP6_STATINC(IP6_STAT_CANTFORWARD);
403 } else {
404 IP6_STATINC(IP6_STAT_FORWARD);
405 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
406 if (type)
407 IP6_STATINC(IP6_STAT_REDIRECTSENT);
408 else {
409 #ifdef GATEWAY
410 /* Need to release rt here */
411 rtcache_unref(rt, ro);
412 rt = NULL;
413 if (mcopy->m_flags & M_CANFASTFWD)
414 ip6flow_create(ro, mcopy);
415 #endif
416 if (mcopy)
417 goto freecopy;
418 }
419 }
420
421 senderr:
422 if (mcopy == NULL)
423 goto out;
424 switch (error) {
425 case 0:
426 if (type == ND_REDIRECT) {
427 icmp6_redirect_output(mcopy, rt);
428 goto out;
429 }
430 goto freecopy;
431
432 case EMSGSIZE:
433 /* xxx MTU is constant in PPP? */
434 goto freecopy;
435
436 case ENOBUFS:
437 /* Tell source to slow down like source quench in IP? */
438 goto freecopy;
439
440 case ENETUNREACH: /* shouldn't happen, checked above */
441 case EHOSTUNREACH:
442 case ENETDOWN:
443 case EHOSTDOWN:
444 default:
445 type = ICMP6_DST_UNREACH;
446 code = ICMP6_DST_UNREACH_ADDR;
447 break;
448 }
449 icmp6_error(mcopy, type, code, 0);
450 goto out;
451
452 freecopy:
453 m_freem(mcopy);
454 goto out;
455
456 drop:
457 m_freem(m);
458
459 out:
460 #ifdef IPSEC
461 if (sp != NULL)
462 KEY_SP_UNREF(&sp);
463 #endif
464 rtcache_unref(rt, ro);
465 if (ro != NULL)
466 rtcache_percpu_putref(ip6_forward_rt_percpu);
467 return;
468 }
469