if_xname support Part 2/2: Convert remaining netif devices and implement full
[dragonfly.git] / sys / netinet6 / ip6_forward.c
1 /*      $FreeBSD: src/sys/netinet6/ip6_forward.c,v 1.4.2.7 2003/01/24 05:11:35 sam Exp $        */
2 /*      $DragonFly: src/sys/netinet6/ip6_forward.c,v 1.6 2003/12/02 08:00:22 asmodai Exp $      */
3 /*      $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $    */
4
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33
34 #include "opt_ip6fw.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_pfil_hooks.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.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
52 #include <net/if.h>
53 #include <net/route.h>
54 #ifdef PFIL_HOOKS
55 #include <net/pfil.h>
56 #endif
57
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/ip.h>
62 #include <netinet/ip_var.h>
63 #include <netinet6/in6_var.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet/icmp6.h>
67 #include <netinet6/nd6.h>
68
69 #include <netinet/in_pcb.h>
70
71 #ifdef IPSEC
72 #include <netinet6/ipsec.h>
73 #ifdef INET6
74 #include <netinet6/ipsec6.h>
75 #endif
76 #include <netproto/key/key.h>
77 #endif /* IPSEC */
78
79 #ifdef FAST_IPSEC
80 #include "ipsec.h"
81 #include "ipsec6.h"
82 #include <netipsec/key.h>
83 #define IPSEC
84 #endif /* FAST_IPSEC */
85
86 #include <net/ip6fw/ip6_fw.h>
87
88 #include <net/net_osdep.h>
89
90 struct  route_in6 ip6_forward_rt;
91
92 /*
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
96  * of codes and types.
97  *
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.
102  *
103  */
104
105 void
106 ip6_forward(m, srcrt)
107         struct mbuf *m;
108         int srcrt;
109 {
110         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
111         struct sockaddr_in6 *dst;
112         struct rtentry *rt;
113         int error, type = 0, code = 0;
114         struct mbuf *mcopy = NULL;
115         struct ifnet *origifp;  /* maybe unnecessary */
116 #ifdef IPSEC
117         struct secpolicy *sp = NULL;
118 #endif
119
120 #ifdef IPSEC
121         /*
122          * Check AH/ESP integrity.
123          */
124         /*
125          * Don't increment ip6s_cantforward because this is the check
126          * before forwarding packet actually.
127          */
128         if (ipsec6_in_reject(m, NULL)) {
129 #if !defined(FAST_IPSEC)
130                 ipsec6stat.in_polvio++;
131 #endif
132                 m_freem(m);
133                 return;
134         }
135 #endif /* IPSEC */
136
137         /*
138          * Do not forward packets to multicast destination (should be handled
139          * by ip6_mforward().
140          * Do not forward packets with unspecified source.  It was discussed
141          * in July 2000, on ipngwg mailing list.
142          */
143         if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
144             IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
145             IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
146                 ip6stat.ip6s_cantforward++;
147                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
148                 if (ip6_log_time + ip6_log_interval < time_second) {
149                         ip6_log_time = time_second;
150                         log(LOG_DEBUG,
151                             "cannot forward "
152                             "from %s to %s nxt %d received on %s\n",
153                             ip6_sprintf(&ip6->ip6_src),
154                             ip6_sprintf(&ip6->ip6_dst),
155                             ip6->ip6_nxt,
156                             if_name(m->m_pkthdr.rcvif));
157                 }
158                 m_freem(m);
159                 return;
160         }
161
162         if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
163                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
164                 icmp6_error(m, ICMP6_TIME_EXCEEDED,
165                                 ICMP6_TIME_EXCEED_TRANSIT, 0);
166                 return;
167         }
168         ip6->ip6_hlim -= IPV6_HLIMDEC;
169
170         /*
171          * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
172          * size of IPv6 + ICMPv6 headers) bytes of the packet in case
173          * we need to generate an ICMP6 message to the src.
174          * Thanks to M_EXT, in most cases copy will not occur.
175          *
176          * It is important to save it before IPsec processing as IPsec
177          * processing may modify the mbuf.
178          */
179         mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
180
181 #ifdef IPSEC
182         /* get a security policy for this packet */
183         sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
184             &error);
185         if (sp == NULL) {
186                 ipsec6stat.out_inval++;
187                 ip6stat.ip6s_cantforward++;
188                 if (mcopy) {
189 #if 0
190                         /* XXX: what icmp ? */
191 #else
192                         m_freem(mcopy);
193 #endif
194                 }
195                 m_freem(m);
196                 return;
197         }
198
199         error = 0;
200
201         /* check policy */
202         switch (sp->policy) {
203         case IPSEC_POLICY_DISCARD:
204                 /*
205                  * This packet is just discarded.
206                  */
207                 ipsec6stat.out_polvio++;
208                 ip6stat.ip6s_cantforward++;
209                 key_freesp(sp);
210                 if (mcopy) {
211 #if 0
212                         /* XXX: what icmp ? */
213 #else
214                         m_freem(mcopy);
215 #endif
216                 }
217                 m_freem(m);
218                 return;
219
220         case IPSEC_POLICY_BYPASS:
221         case IPSEC_POLICY_NONE:
222                 /* no need to do IPsec. */
223                 key_freesp(sp);
224                 goto skip_ipsec;
225
226         case IPSEC_POLICY_IPSEC:
227                 if (sp->req == NULL) {
228                         /* XXX should be panic ? */
229                         printf("ip6_forward: No IPsec request specified.\n");
230                         ip6stat.ip6s_cantforward++;
231                         key_freesp(sp);
232                         if (mcopy) {
233 #if 0
234                                 /* XXX: what icmp ? */
235 #else
236                                 m_freem(mcopy);
237 #endif
238                         }
239                         m_freem(m);
240                         return;
241                 }
242                 /* do IPsec */
243                 break;
244
245         case IPSEC_POLICY_ENTRUST:
246         default:
247                 /* should be panic ?? */
248                 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
249                 key_freesp(sp);
250                 goto skip_ipsec;
251         }
252
253     {
254         struct ipsec_output_state state;
255
256         /*
257          * All the extension headers will become inaccessible
258          * (since they can be encrypted).
259          * Don't panic, we need no more updates to extension headers
260          * on inner IPv6 packet (since they are now encapsulated).
261          *
262          * IPv6 [ESP|AH] IPv6 [extension headers] payload
263          */
264         bzero(&state, sizeof(state));
265         state.m = m;
266         state.ro = NULL;        /* update at ipsec6_output_tunnel() */
267         state.dst = NULL;       /* update at ipsec6_output_tunnel() */
268
269         error = ipsec6_output_tunnel(&state, sp, 0);
270
271         m = state.m;
272         key_freesp(sp);
273
274         if (error) {
275                 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
276                 switch (error) {
277                 case EHOSTUNREACH:
278                 case ENETUNREACH:
279                 case EMSGSIZE:
280                 case ENOBUFS:
281                 case ENOMEM:
282                         break;
283                 default:
284                         printf("ip6_output (ipsec): error code %d\n", error);
285                         /* fall through */
286                 case ENOENT:
287                         /* don't show these error codes to the user */
288                         break;
289                 }
290                 ip6stat.ip6s_cantforward++;
291                 if (mcopy) {
292 #if 0
293                         /* XXX: what icmp ? */
294 #else
295                         m_freem(mcopy);
296 #endif
297                 }
298                 m_freem(m);
299                 return;
300         }
301     }
302     skip_ipsec:
303 #endif /* IPSEC */
304
305         dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
306         if (!srcrt) {
307                 /*
308                  * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
309                  */
310                 if (ip6_forward_rt.ro_rt == 0 ||
311                     (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
312                         if (ip6_forward_rt.ro_rt) {
313                                 RTFREE(ip6_forward_rt.ro_rt);
314                                 ip6_forward_rt.ro_rt = 0;
315                         }
316                         /* this probably fails but give it a try again */
317                         rtalloc_ign((struct route *)&ip6_forward_rt,
318                                     RTF_PRCLONING);
319                 }
320
321                 if (ip6_forward_rt.ro_rt == 0) {
322                         ip6stat.ip6s_noroute++;
323                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
324                         if (mcopy) {
325                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
326                                             ICMP6_DST_UNREACH_NOROUTE, 0);
327                         }
328                         m_freem(m);
329                         return;
330                 }
331         } else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
332                  !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
333                 if (ip6_forward_rt.ro_rt) {
334                         RTFREE(ip6_forward_rt.ro_rt);
335                         ip6_forward_rt.ro_rt = 0;
336                 }
337                 bzero(dst, sizeof(*dst));
338                 dst->sin6_len = sizeof(struct sockaddr_in6);
339                 dst->sin6_family = AF_INET6;
340                 dst->sin6_addr = ip6->ip6_dst;
341
342                 rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
343                 if (ip6_forward_rt.ro_rt == 0) {
344                         ip6stat.ip6s_noroute++;
345                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
346                         if (mcopy) {
347                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
348                                             ICMP6_DST_UNREACH_NOROUTE, 0);
349                         }
350                         m_freem(m);
351                         return;
352                 }
353         }
354         rt = ip6_forward_rt.ro_rt;
355
356         /*
357          * Scope check: if a packet can't be delivered to its destination
358          * for the reason that the destination is beyond the scope of the
359          * source address, discard the packet and return an icmp6 destination
360          * unreachable error with Code 2 (beyond scope of source address).
361          * [draft-ietf-ipngwg-icmp-v3-02.txt, Section 3.1]
362          */
363         if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
364             in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
365                 ip6stat.ip6s_cantforward++;
366                 ip6stat.ip6s_badscope++;
367                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
368
369                 if (ip6_log_time + ip6_log_interval < time_second) {
370                         ip6_log_time = time_second;
371                         log(LOG_DEBUG,
372                             "cannot forward "
373                             "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
374                             ip6_sprintf(&ip6->ip6_src),
375                             ip6_sprintf(&ip6->ip6_dst),
376                             ip6->ip6_nxt,
377                             if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
378                 }
379                 if (mcopy)
380                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
381                                     ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
382                 m_freem(m);
383                 return;
384         }
385
386         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
387                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
388                 if (mcopy) {
389                         u_long mtu;
390 #ifdef IPSEC
391                         struct secpolicy *sp;
392                         int ipsecerror;
393                         size_t ipsechdrsiz;
394 #endif
395
396                         mtu = rt->rt_ifp->if_mtu;
397 #ifdef IPSEC
398                         /*
399                          * When we do IPsec tunnel ingress, we need to play
400                          * with if_mtu value (decrement IPsec header size
401                          * from mtu value).  The code is much simpler than v4
402                          * case, as we have the outgoing interface for
403                          * encapsulated packet as "rt->rt_ifp".
404                          */
405                         sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
406                                 IP_FORWARDING, &ipsecerror);
407                         if (sp) {
408                                 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
409                                         IPSEC_DIR_OUTBOUND, NULL);
410                                 if (ipsechdrsiz < mtu)
411                                         mtu -= ipsechdrsiz;
412                         }
413
414                         /*
415                          * if mtu becomes less than minimum MTU,
416                          * tell minimum MTU (and I'll need to fragment it).
417                          */
418                         if (mtu < IPV6_MMTU)
419                                 mtu = IPV6_MMTU;
420 #endif
421                         icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
422                 }
423                 m_freem(m);
424                 return;
425         }
426
427         if (rt->rt_flags & RTF_GATEWAY)
428                 dst = (struct sockaddr_in6 *)rt->rt_gateway;
429
430         /*
431          * If we are to forward the packet using the same interface
432          * as one we got the packet from, perhaps we should send a redirect
433          * to sender to shortcut a hop.
434          * Only send redirect if source is sending directly to us,
435          * and if packet was not source routed (or has any options).
436          * Also, don't send redirect if forwarding using a route
437          * modified by a redirect.
438          */
439         if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
440             (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
441                 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
442                         /*
443                          * If the incoming interface is equal to the outgoing
444                          * one, and the link attached to the interface is
445                          * point-to-point, then it will be highly probable
446                          * that a routing loop occurs. Thus, we immediately
447                          * drop the packet and send an ICMPv6 error message.
448                          *
449                          * type/code is based on suggestion by Rich Draves.
450                          * not sure if it is the best pick.
451                          */
452                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
453                                     ICMP6_DST_UNREACH_ADDR, 0);
454                         m_freem(m);
455                         return;
456                 }
457                 type = ND_REDIRECT;
458         }
459
460         /*
461          * Check with the firewall...
462          */
463         if (ip6_fw_enable && ip6_fw_chk_ptr) {
464                 u_short port = 0;
465                 /* If ipfw says divert, we have to just drop packet */
466                 if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
467                         m_freem(m);
468                         goto freecopy;
469                 }
470                 if (!m)
471                         goto freecopy;
472         }
473
474         /*
475          * Fake scoped addresses. Note that even link-local source or
476          * destinaion can appear, if the originating node just sends the
477          * packet to us (without address resolution for the destination).
478          * Since both icmp6_error and icmp6_redirect_output fill the embedded
479          * link identifiers, we can do this stuff after making a copy for
480          * returning an error.
481          */
482         if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
483                 /*
484                  * See corresponding comments in ip6_output.
485                  * XXX: but is it possible that ip6_forward() sends a packet
486                  *      to a loopback interface? I don't think so, and thus
487                  *      I bark here. (jinmei@kame.net)
488                  * XXX: it is common to route invalid packets to loopback.
489                  *      also, the codepath will be visited on use of ::1 in
490                  *      rthdr. (itojun)
491                  */
492 #if 1
493                 if (0)
494 #else
495                 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
496 #endif
497                 {
498                         printf("ip6_forward: outgoing interface is loopback. "
499                                 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
500                                 ip6_sprintf(&ip6->ip6_src),
501                                 ip6_sprintf(&ip6->ip6_dst),
502                                 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
503                                 if_name(rt->rt_ifp));
504                 }
505
506                 /* we can just use rcvif in forwarding. */
507                 origifp = m->m_pkthdr.rcvif;
508         }
509         else
510                 origifp = rt->rt_ifp;
511 #ifndef SCOPEDROUTING
512         /*
513          * clear embedded scope identifiers if necessary.
514          * in6_clearscope will touch the addresses only when necessary.
515          */
516         in6_clearscope(&ip6->ip6_src);
517         in6_clearscope(&ip6->ip6_dst);
518 #endif
519
520 #ifdef PFIL_HOOKS
521         /*
522          * Run through list of hooks for output packets.
523          */
524         error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT);
525         if (error != 0)
526                 goto senderr;
527         if (m == NULL)
528                 goto freecopy;
529         ip6 = mtod(m, struct ip6_hdr *);
530 #endif /* PFIL_HOOKS */
531
532         error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
533         if (error) {
534                 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
535                 ip6stat.ip6s_cantforward++;
536         } else {
537                 ip6stat.ip6s_forward++;
538                 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
539                 if (type)
540                         ip6stat.ip6s_redirectsent++;
541                 else {
542                         if (mcopy)
543                                 goto freecopy;
544                 }
545         }
546 #ifdef PFIL_HOOKS
547 senderr:
548 #endif
549         if (mcopy == NULL)
550                 return;
551         switch (error) {
552         case 0:
553 #if 1
554                 if (type == ND_REDIRECT) {
555                         icmp6_redirect_output(mcopy, rt);
556                         return;
557                 }
558 #endif
559                 goto freecopy;
560
561         case EMSGSIZE:
562                 /* xxx MTU is constant in PPP? */
563                 goto freecopy;
564
565         case ENOBUFS:
566                 /* Tell source to slow down like source quench in IP? */
567                 goto freecopy;
568
569         case ENETUNREACH:       /* shouldn't happen, checked above */
570         case EHOSTUNREACH:
571         case ENETDOWN:
572         case EHOSTDOWN:
573         default:
574                 type = ICMP6_DST_UNREACH;
575                 code = ICMP6_DST_UNREACH_ADDR;
576                 break;
577         }
578         icmp6_error(mcopy, type, code, 0);
579         return;
580
581  freecopy:
582         m_freem(mcopy);
583         return;
584 }