]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/netinet6/ip6_output.c
always copy ip6_pktopt. remove needcopy and needfree
[FreeBSD/FreeBSD.git] / sys / netinet6 / ip6_output.c
1 /*      $FreeBSD$       */
2 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei 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 /*-
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *      The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
62  */
63
64 #include "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68
69 #include <sys/param.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/proc.h>
73 #include <sys/errno.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/systm.h>
78 #include <sys/kernel.h>
79
80 #include <net/if.h>
81 #include <net/netisr.h>
82 #include <net/route.h>
83 #include <net/pfil.h>
84
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet6/in6_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet/icmp6.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet/in_pcb.h>
92 #include <netinet/tcp_var.h>
93 #include <netinet6/nd6.h>
94
95 #ifdef IPSEC
96 #include <netinet6/ipsec.h>
97 #ifdef INET6
98 #include <netinet6/ipsec6.h>
99 #endif
100 #include <netkey/key.h>
101 #endif /* IPSEC */
102
103 #ifdef FAST_IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #endif /* FAST_IPSEC */
108
109 #include <netinet6/ip6_fw.h>
110
111 #include <net/net_osdep.h>
112
113 #include <netinet6/ip6protosw.h>
114
115 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
116
117 struct ip6_exthdrs {
118         struct mbuf *ip6e_ip6;
119         struct mbuf *ip6e_hbh;
120         struct mbuf *ip6e_dest1;
121         struct mbuf *ip6e_rthdr;
122         struct mbuf *ip6e_dest2;
123 };
124
125 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
126                            int, int));
127 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
128         struct socket *, struct sockopt *));
129 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
130 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int,
131         int, int, int));
132
133 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
134 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
135 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
136 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
137         struct ip6_frag **));
138 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
139 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
140 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
141         struct ifnet *, struct in6_addr *, u_long *, int *));
142 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int));
143
144
145 /*
146  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
147  * header (with pri, len, nxt, hlim, src, dst).
148  * This function may modify ver and hlim only.
149  * The mbuf chain containing the packet will be freed.
150  * The mbuf opt, if present, will not be freed.
151  *
152  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
153  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
154  * which is rt_rmx.rmx_mtu.
155  */
156 int
157 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
158         struct mbuf *m0;
159         struct ip6_pktopts *opt;
160         struct route_in6 *ro;
161         int flags;
162         struct ip6_moptions *im6o;
163         struct ifnet **ifpp;            /* XXX: just for statistics */
164         struct inpcb *inp;
165 {
166         struct ip6_hdr *ip6, *mhip6;
167         struct ifnet *ifp, *origifp;
168         struct mbuf *m = m0;
169         int hlen, tlen, len, off;
170         struct route_in6 ip6route;
171         struct sockaddr_in6 *dst;
172         struct in6_addr odst;
173         int error = 0;
174         struct in6_ifaddr *ia = NULL;
175         u_long mtu;
176         int alwaysfrag, dontfrag;
177         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
178         struct ip6_exthdrs exthdrs;
179         struct in6_addr finaldst;
180         struct route_in6 *ro_pmtu = NULL;
181         int hdrsplit = 0;
182         int needipsec = 0;
183 #if defined(IPSEC) || defined(FAST_IPSEC)
184         int needipsectun = 0;
185         struct secpolicy *sp = NULL;
186 #endif /*IPSEC || FAST_IPSEC*/
187
188         ip6 = mtod(m, struct ip6_hdr *);
189         finaldst = ip6->ip6_dst;
190
191 #define MAKE_EXTHDR(hp, mp)                                             \
192     do {                                                                \
193         if (hp) {                                                       \
194                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
195                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
196                     ((eh)->ip6e_len + 1) << 3);                         \
197                 if (error)                                              \
198                         goto freehdrs;                                  \
199         }                                                               \
200     } while (/*CONSTCOND*/ 0)
201
202         bzero(&exthdrs, sizeof(exthdrs));
203
204         if (opt) {
205                 /* Hop-by-Hop options header */
206                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
207                 /* Destination options header(1st part) */
208                 if (opt->ip6po_rthdr) {
209                         /*
210                          * Destination options header(1st part)
211                          * This only makes sence with a routing header.
212                          * See Section 9.2 of RFC 3542.
213                          * Disabling this part just for MIP6 convenience is
214                          * a bad idea.  We need to think carefully about a
215                          * way to make the advanced API coexist with MIP6
216                          * options, which might automatically be inserted in
217                          * the kernel.
218                          */
219                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
220                 }
221                 /* Routing header */
222                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
223                 /* Destination options header(2nd part) */
224                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
225         }
226
227 #ifdef IPSEC
228         /* get a security policy for this packet */
229         if (inp == NULL)
230                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
231         else
232                 sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
233
234         if (sp == NULL) {
235                 ipsec6stat.out_inval++;
236                 goto freehdrs;
237         }
238
239         error = 0;
240
241         /* check policy */
242         switch (sp->policy) {
243         case IPSEC_POLICY_DISCARD:
244                 /*
245                  * This packet is just discarded.
246                  */
247                 ipsec6stat.out_polvio++;
248                 goto freehdrs;
249
250         case IPSEC_POLICY_BYPASS:
251         case IPSEC_POLICY_NONE:
252                 /* no need to do IPsec. */
253                 needipsec = 0;
254                 break;
255
256         case IPSEC_POLICY_IPSEC:
257                 if (sp->req == NULL) {
258                         /* acquire a policy */
259                         error = key_spdacquire(sp);
260                         goto freehdrs;
261                 }
262                 needipsec = 1;
263                 break;
264
265         case IPSEC_POLICY_ENTRUST:
266         default:
267                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
268         }
269 #endif /* IPSEC */
270 #ifdef FAST_IPSEC
271         /* get a security policy for this packet */
272         if (inp == NULL)
273                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
274         else
275                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
276
277         if (sp == NULL) {
278                 newipsecstat.ips_out_inval++;
279                 goto freehdrs;
280         }
281
282         error = 0;
283
284         /* check policy */
285         switch (sp->policy) {
286         case IPSEC_POLICY_DISCARD:
287                 /*
288                  * This packet is just discarded.
289                  */
290                 newipsecstat.ips_out_polvio++;
291                 goto freehdrs;
292
293         case IPSEC_POLICY_BYPASS:
294         case IPSEC_POLICY_NONE:
295                 /* no need to do IPsec. */
296                 needipsec = 0;
297                 break;
298
299         case IPSEC_POLICY_IPSEC:
300                 if (sp->req == NULL) {
301                         /* acquire a policy */
302                         error = key_spdacquire(sp);
303                         goto freehdrs;
304                 }
305                 needipsec = 1;
306                 break;
307
308         case IPSEC_POLICY_ENTRUST:
309         default:
310                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
311         }
312 #endif /* FAST_IPSEC */
313
314         /*
315          * Calculate the total length of the extension header chain.
316          * Keep the length of the unfragmentable part for fragmentation.
317          */
318         optlen = 0;
319         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
320         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
321         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
322         unfragpartlen = optlen + sizeof(struct ip6_hdr);
323         /* NOTE: we don't add AH/ESP length here. do that later. */
324         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
325
326         /*
327          * If we need IPsec, or there is at least one extension header,
328          * separate IP6 header from the payload.
329          */
330         if ((needipsec || optlen) && !hdrsplit) {
331                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
332                         m = NULL;
333                         goto freehdrs;
334                 }
335                 m = exthdrs.ip6e_ip6;
336                 hdrsplit++;
337         }
338
339         /* adjust pointer */
340         ip6 = mtod(m, struct ip6_hdr *);
341
342         /* adjust mbuf packet header length */
343         m->m_pkthdr.len += optlen;
344         plen = m->m_pkthdr.len - sizeof(*ip6);
345
346         /* If this is a jumbo payload, insert a jumbo payload option. */
347         if (plen > IPV6_MAXPACKET) {
348                 if (!hdrsplit) {
349                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
350                                 m = NULL;
351                                 goto freehdrs;
352                         }
353                         m = exthdrs.ip6e_ip6;
354                         hdrsplit++;
355                 }
356                 /* adjust pointer */
357                 ip6 = mtod(m, struct ip6_hdr *);
358                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
359                         goto freehdrs;
360                 ip6->ip6_plen = 0;
361         } else
362                 ip6->ip6_plen = htons(plen);
363
364         /*
365          * Concatenate headers and fill in next header fields.
366          * Here we have, on "m"
367          *      IPv6 payload
368          * and we insert headers accordingly.  Finally, we should be getting:
369          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
370          *
371          * during the header composing process, "m" points to IPv6 header.
372          * "mprev" points to an extension header prior to esp.
373          */
374         {
375                 u_char *nexthdrp = &ip6->ip6_nxt;
376                 struct mbuf *mprev = m;
377
378                 /*
379                  * we treat dest2 specially.  this makes IPsec processing
380                  * much easier.  the goal here is to make mprev point the
381                  * mbuf prior to dest2.
382                  *
383                  * result: IPv6 dest2 payload
384                  * m and mprev will point to IPv6 header.
385                  */
386                 if (exthdrs.ip6e_dest2) {
387                         if (!hdrsplit)
388                                 panic("assumption failed: hdr not split");
389                         exthdrs.ip6e_dest2->m_next = m->m_next;
390                         m->m_next = exthdrs.ip6e_dest2;
391                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
392                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
393                 }
394
395 #define MAKE_CHAIN(m, mp, p, i)\
396     do {\
397         if (m) {\
398                 if (!hdrsplit) \
399                         panic("assumption failed: hdr not split"); \
400                 *mtod((m), u_char *) = *(p);\
401                 *(p) = (i);\
402                 p = mtod((m), u_char *);\
403                 (m)->m_next = (mp)->m_next;\
404                 (mp)->m_next = (m);\
405                 (mp) = (m);\
406         }\
407     } while (/*CONSTCOND*/ 0)
408                 /*
409                  * result: IPv6 hbh dest1 rthdr dest2 payload
410                  * m will point to IPv6 header.  mprev will point to the
411                  * extension header prior to dest2 (rthdr in the above case).
412                  */
413                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
414                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
415                     IPPROTO_DSTOPTS);
416                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
417                     IPPROTO_ROUTING);
418
419 #if defined(IPSEC) || defined(FAST_IPSEC)
420                 if (!needipsec)
421                         goto skip_ipsec2;
422
423                 /*
424                  * pointers after IPsec headers are not valid any more.
425                  * other pointers need a great care too.
426                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
427                  */
428                 exthdrs.ip6e_dest2 = NULL;
429
430             {
431                 struct ip6_rthdr *rh = NULL;
432                 int segleft_org = 0;
433                 struct ipsec_output_state state;
434
435                 if (exthdrs.ip6e_rthdr) {
436                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
437                         segleft_org = rh->ip6r_segleft;
438                         rh->ip6r_segleft = 0;
439                 }
440
441                 bzero(&state, sizeof(state));
442                 state.m = m;
443                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
444                     &needipsectun);
445                 m = state.m;
446                 if (error) {
447                         /* mbuf is already reclaimed in ipsec6_output_trans. */
448                         m = NULL;
449                         switch (error) {
450                         case EHOSTUNREACH:
451                         case ENETUNREACH:
452                         case EMSGSIZE:
453                         case ENOBUFS:
454                         case ENOMEM:
455                                 break;
456                         default:
457                                 printf("ip6_output (ipsec): error code %d\n", error);
458                                 /* FALLTHROUGH */
459                         case ENOENT:
460                                 /* don't show these error codes to the user */
461                                 error = 0;
462                                 break;
463                         }
464                         goto bad;
465                 }
466                 if (exthdrs.ip6e_rthdr) {
467                         /* ah6_output doesn't modify mbuf chain */
468                         rh->ip6r_segleft = segleft_org;
469                 }
470             }
471 skip_ipsec2:;
472 #endif
473         }
474
475         /*
476          * If there is a routing header, replace the destination address field
477          * with the first hop of the routing header.
478          */
479         if (exthdrs.ip6e_rthdr) {
480                 struct ip6_rthdr *rh =
481                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
482                                                   struct ip6_rthdr *));
483                 struct ip6_rthdr0 *rh0;
484                 struct in6_addr *addrs;
485
486                 switch (rh->ip6r_type) {
487                 case IPV6_RTHDR_TYPE_0:
488                          rh0 = (struct ip6_rthdr0 *)rh;
489                          addrs = (struct in6_addr *)(rh0 + 1);
490
491                          ip6->ip6_dst = *addrs;
492                          bcopy((caddr_t)(addrs + 1), (caddr_t)addrs,
493                                sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
494                                  );
495                          *(addrs + rh0->ip6r0_segleft - 1) = finaldst;
496                          break;
497                 default:        /* is it possible? */
498                          error = EINVAL;
499                          goto bad;
500                 }
501         }
502
503         /* Source address validation */
504         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
505             (flags & IPV6_DADOUTPUT) == 0) {
506                 error = EOPNOTSUPP;
507                 ip6stat.ip6s_badscope++;
508                 goto bad;
509         }
510         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
511                 error = EOPNOTSUPP;
512                 ip6stat.ip6s_badscope++;
513                 goto bad;
514         }
515
516         ip6stat.ip6s_localout++;
517
518         /*
519          * Route packet.
520          */
521         if (ro == 0) {
522                 ro = &ip6route;
523                 bzero((caddr_t)ro, sizeof(*ro));
524         }
525         ro_pmtu = ro;
526         if (opt && opt->ip6po_rthdr)
527                 ro = &opt->ip6po_route;
528         dst = (struct sockaddr_in6 *)&ro->ro_dst;
529
530 again:
531         /*
532          * If there is a cached route,
533          * check that it is to the same destination
534          * and is still up. If not, free it and try again.
535          */
536         if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
537                          dst->sin6_family != AF_INET6 ||
538                          !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
539                 RTFREE(ro->ro_rt);
540                 ro->ro_rt = (struct rtentry *)0;
541         }
542         if (ro->ro_rt == 0) {
543                 bzero(dst, sizeof(*dst));
544                 dst->sin6_family = AF_INET6;
545                 dst->sin6_len = sizeof(struct sockaddr_in6);
546                 dst->sin6_addr = ip6->ip6_dst;
547         }
548
549         /*
550          * if specified, try to fill in the traffic class field.
551          * do not override if a non-zero value is already set.
552          * we check the diffserv field and the ecn field separately.
553          */
554         if (opt && opt->ip6po_tclass >= 0) {
555                 int mask = 0;
556
557                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
558                         mask |= 0xfc;
559                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
560                         mask |= 0x03;
561                 if (mask != 0)
562                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
563         }
564
565         /* fill in or override the hop limit field, if necessary. */
566         if (opt && opt->ip6po_hlim != -1)
567                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
568         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
569                 if (im6o != NULL)
570                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
571                 else
572                         ip6->ip6_hlim = ip6_defmcasthlim;
573         }
574
575 #if defined(IPSEC) || defined(FAST_IPSEC)
576         if (needipsec && needipsectun) {
577                 struct ipsec_output_state state;
578
579                 /*
580                  * All the extension headers will become inaccessible
581                  * (since they can be encrypted).
582                  * Don't panic, we need no more updates to extension headers
583                  * on inner IPv6 packet (since they are now encapsulated).
584                  *
585                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
586                  */
587                 bzero(&exthdrs, sizeof(exthdrs));
588                 exthdrs.ip6e_ip6 = m;
589
590                 bzero(&state, sizeof(state));
591                 state.m = m;
592                 state.ro = (struct route *)ro;
593                 state.dst = (struct sockaddr *)dst;
594
595                 error = ipsec6_output_tunnel(&state, sp, flags);
596
597                 m = state.m;
598                 ro = (struct route_in6 *)state.ro;
599                 dst = (struct sockaddr_in6 *)state.dst;
600                 if (error) {
601                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
602                         m0 = m = NULL;
603                         m = NULL;
604                         switch (error) {
605                         case EHOSTUNREACH:
606                         case ENETUNREACH:
607                         case EMSGSIZE:
608                         case ENOBUFS:
609                         case ENOMEM:
610                                 break;
611                         default:
612                                 printf("ip6_output (ipsec): error code %d\n", error);
613                                 /* FALLTHROUGH */
614                         case ENOENT:
615                                 /* don't show these error codes to the user */
616                                 error = 0;
617                                 break;
618                         }
619                         goto bad;
620                 }
621
622                 exthdrs.ip6e_ip6 = m;
623         }
624 #endif /* IPSEC */
625
626         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
627                 /* Unicast */
628
629 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
630 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
631                 /* xxx
632                  * interface selection comes here
633                  * if an interface is specified from an upper layer,
634                  * ifp must point it.
635                  */
636                 if (ro->ro_rt == 0) {
637                         /*
638                          * non-bsdi always clone routes, if parent is
639                          * PRF_CLONING.
640                          */
641                         rtalloc((struct route *)ro);
642                 }
643                 if (ro->ro_rt == 0) {
644                         ip6stat.ip6s_noroute++;
645                         error = EHOSTUNREACH;
646                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
647                         goto bad;
648                 }
649                 /* XXX rt not locked */
650                 ia = ifatoia6(ro->ro_rt->rt_ifa);
651                 ifp = ro->ro_rt->rt_ifp;
652                 ro->ro_rt->rt_rmx.rmx_pksent++;
653                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
654                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
655                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
656
657                 in6_ifstat_inc(ifp, ifs6_out_request);
658
659                 /*
660                  * Check if the outgoing interface conflicts with
661                  * the interface specified by ifi6_ifindex (if specified).
662                  * Note that loopback interface is always okay.
663                  * (this may happen when we are sending a packet to one of
664                  *  our own addresses.)
665                  */
666                 if (opt && opt->ip6po_pktinfo
667                  && opt->ip6po_pktinfo->ipi6_ifindex) {
668                         if (!(ifp->if_flags & IFF_LOOPBACK)
669                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
670                                 ip6stat.ip6s_noroute++;
671                                 in6_ifstat_inc(ifp, ifs6_out_discard);
672                                 error = EHOSTUNREACH;
673                                 goto bad;
674                         }
675                 }
676
677                 if (opt && opt->ip6po_hlim != -1)
678                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
679         } else {
680                 /* Multicast */
681                 struct  in6_multi *in6m;
682
683                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
684
685                 /*
686                  * See if the caller provided any multicast options
687                  */
688                 ifp = NULL;
689                 if (im6o != NULL) {
690                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
691                         if (im6o->im6o_multicast_ifp != NULL)
692                                 ifp = im6o->im6o_multicast_ifp;
693                 } else
694                         ip6->ip6_hlim = ip6_defmcasthlim;
695
696                 /*
697                  * See if the caller provided the outgoing interface
698                  * as an ancillary data.
699                  * Boundary check for ifindex is assumed to be already done.
700                  */
701                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
702                         ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
703
704                 /*
705                  * If the destination is a node-local scope multicast,
706                  * the packet should be loop-backed only.
707                  */
708                 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
709                         /*
710                          * If the outgoing interface is already specified,
711                          * it should be a loopback interface.
712                          */
713                         if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
714                                 ip6stat.ip6s_badscope++;
715                                 error = ENETUNREACH; /* XXX: better error? */
716                                 /* XXX correct ifp? */
717                                 in6_ifstat_inc(ifp, ifs6_out_discard);
718                                 goto bad;
719                         } else {
720                                 ifp = &loif[0];
721                         }
722                 }
723
724                 if (opt && opt->ip6po_hlim != -1)
725                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
726
727                 /*
728                  * If caller did not provide an interface lookup a
729                  * default in the routing table.  This is either a
730                  * default for the speicfied group (i.e. a host
731                  * route), or a multicast default (a route for the
732                  * ``net'' ff00::/8).
733                  */
734                 if (ifp == NULL) {
735                         if (ro->ro_rt == 0)
736                                 ro->ro_rt = rtalloc1((struct sockaddr *)
737                                                 &ro->ro_dst, 0, 0UL);
738                         else
739                                 RT_LOCK(ro->ro_rt);
740                         if (ro->ro_rt == 0) {
741                                 ip6stat.ip6s_noroute++;
742                                 error = EHOSTUNREACH;
743                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
744                                 goto bad;
745                         }
746                         ia = ifatoia6(ro->ro_rt->rt_ifa);
747                         ifp = ro->ro_rt->rt_ifp;
748                         ro->ro_rt->rt_rmx.rmx_pksent++;
749                         RT_UNLOCK(ro->ro_rt);
750                 }
751
752                 if ((flags & IPV6_FORWARDING) == 0)
753                         in6_ifstat_inc(ifp, ifs6_out_request);
754                 in6_ifstat_inc(ifp, ifs6_out_mcast);
755
756                 /*
757                  * Confirm that the outgoing interface supports multicast.
758                  */
759                 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
760                         ip6stat.ip6s_noroute++;
761                         in6_ifstat_inc(ifp, ifs6_out_discard);
762                         error = ENETUNREACH;
763                         goto bad;
764                 }
765                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
766                 if (in6m != NULL &&
767                    (im6o == NULL || im6o->im6o_multicast_loop)) {
768                         /*
769                          * If we belong to the destination multicast group
770                          * on the outgoing interface, and the caller did not
771                          * forbid loopback, loop back a copy.
772                          */
773                         ip6_mloopback(ifp, m, dst);
774                 } else {
775                         /*
776                          * If we are acting as a multicast router, perform
777                          * multicast forwarding as if the packet had just
778                          * arrived on the interface to which we are about
779                          * to send.  The multicast forwarding function
780                          * recursively calls this function, using the
781                          * IPV6_FORWARDING flag to prevent infinite recursion.
782                          *
783                          * Multicasts that are looped back by ip6_mloopback(),
784                          * above, will be forwarded by the ip6_input() routine,
785                          * if necessary.
786                          */
787                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
788                                 if (ip6_mforward(ip6, ifp, m) != 0) {
789                                         m_freem(m);
790                                         goto done;
791                                 }
792                         }
793                 }
794                 /*
795                  * Multicasts with a hoplimit of zero may be looped back,
796                  * above, but must not be transmitted on a network.
797                  * Also, multicasts addressed to the loopback interface
798                  * are not sent -- the above call to ip6_mloopback() will
799                  * loop back a copy if this host actually belongs to the
800                  * destination group on the loopback interface.
801                  */
802                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
803                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
804                         m_freem(m);
805                         goto done;
806                 }
807         }
808
809         /*
810          * Fill the outgoing inteface to tell the upper layer
811          * to increment per-interface statistics.
812          */
813         if (ifpp)
814                 *ifpp = ifp;
815
816         /* Determine path MTU. */
817         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
818             &alwaysfrag)) != 0)
819                 goto bad;
820
821         /*
822          * The caller of this function may specify to use the minimum MTU
823          * in some cases.
824          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
825          * setting.  The logic is a bit complicated; by default, unicast
826          * packets will follow path MTU while multicast packets will be sent at
827          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
828          * including unicast ones will be sent at the minimum MTU.  Multicast
829          * packets will always be sent at the minimum MTU unless
830          * IP6PO_MINMTU_DISABLE is explicitly specified.
831          * See RFC 3542 for more details.
832          */
833         if (mtu > IPV6_MMTU) {
834                 if ((flags & IPV6_MINMTU))
835                         mtu = IPV6_MMTU;
836                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
837                         mtu = IPV6_MMTU;
838                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
839                          (opt == NULL ||
840                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
841                         mtu = IPV6_MMTU;
842                 }
843         }
844
845         /* Fake scoped addresses */
846         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
847                 /*
848                  * If source or destination address is a scoped address, and
849                  * the packet is going to be sent to a loopback interface,
850                  * we should keep the original interface.
851                  */
852
853                 /*
854                  * XXX: this is a very experimental and temporary solution.
855                  * We eventually have sockaddr_in6 and use the sin6_scope_id
856                  * field of the structure here.
857                  * We rely on the consistency between two scope zone ids
858                  * of source and destination, which should already be assured.
859                  * Larger scopes than link will be supported in the future. 
860                  */
861                 origifp = NULL;
862                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
863                         origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
864                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
865                         origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
866                 /*
867                  * XXX: origifp can be NULL even in those two cases above.
868                  * For example, if we remove the (only) link-local address
869                  * from the loopback interface, and try to send a link-local
870                  * address without link-id information.  Then the source
871                  * address is ::1, and the destination address is the
872                  * link-local address with its s6_addr16[1] being zero.
873                  * What is worse, if the packet goes to the loopback interface
874                  * by a default rejected route, the null pointer would be
875                  * passed to looutput, and the kernel would hang.
876                  * The following last resort would prevent such disaster.
877                  */
878                 if (origifp == NULL)
879                         origifp = ifp;
880         }
881         else
882                 origifp = ifp;
883         /*
884          * clear embedded scope identifiers if necessary.
885          * in6_clearscope will touch the addresses only when necessary.
886          */
887         in6_clearscope(&ip6->ip6_src);
888         in6_clearscope(&ip6->ip6_dst);
889
890         /*
891          * Check with the firewall...
892          */
893         if (ip6_fw_enable && ip6_fw_chk_ptr) {
894                 u_short port = 0;
895                 m->m_pkthdr.rcvif = NULL;       /* XXX */
896                 /* If ipfw says divert, we have to just drop packet */
897                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
898                         m_freem(m);
899                         goto done;
900                 }
901                 if (!m) {
902                         error = EACCES;
903                         goto done;
904                 }
905         }
906
907         /*
908          * If the outgoing packet contains a hop-by-hop options header,
909          * it must be examined and processed even by the source node.
910          * (RFC 2460, section 4.)
911          */
912         if (exthdrs.ip6e_hbh) {
913                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
914                 u_int32_t dummy; /* XXX unused */
915                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
916
917 #ifdef DIAGNOSTIC
918                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
919                         panic("ip6e_hbh is not continuous");
920 #endif
921                 /*
922                  *  XXX: if we have to send an ICMPv6 error to the sender,
923                  *       we need the M_LOOP flag since icmp6_error() expects
924                  *       the IPv6 and the hop-by-hop options header are
925                  *       continuous unless the flag is set.
926                  */
927                 m->m_flags |= M_LOOP;
928                 m->m_pkthdr.rcvif = ifp;
929                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
930                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
931                     &dummy, &plen) < 0) {
932                         /* m was already freed at this point */
933                         error = EINVAL;/* better error? */
934                         goto done;
935                 }
936                 m->m_flags &= ~M_LOOP; /* XXX */
937                 m->m_pkthdr.rcvif = NULL;
938         }
939
940         /* Jump over all PFIL processing if hooks are not active. */
941         if (inet6_pfil_hook.ph_busy_count == -1)
942                 goto passout;
943
944         odst = ip6->ip6_dst;
945         /* Run through list of hooks for output packets. */
946         error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
947         if (error != 0 || m == NULL)
948                 goto done;
949         ip6 = mtod(m, struct ip6_hdr *);
950
951         /* See if destination IP address was changed by packet filter. */
952         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
953                 m->m_flags |= M_SKIP_FIREWALL;
954                 /* If destination is now ourself drop to ip6_input(). */
955                 if (in6_localaddr(&ip6->ip6_dst)) {
956                         if (m->m_pkthdr.rcvif == NULL)
957                                 m->m_pkthdr.rcvif = loif;
958                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
959                                 m->m_pkthdr.csum_flags |=
960                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
961                                 m->m_pkthdr.csum_data = 0xffff;
962                         }
963                         m->m_pkthdr.csum_flags |=
964                             CSUM_IP_CHECKED | CSUM_IP_VALID;
965                         error = netisr_queue(NETISR_IPV6, m);
966                         goto done;
967                 } else
968                         goto again;     /* Redo the routing table lookup. */
969         }
970
971         /* XXX: IPFIREWALL_FORWARD */
972
973 passout:
974         /*
975          * Send the packet to the outgoing interface.
976          * If necessary, do IPv6 fragmentation before sending.
977          *
978          * the logic here is rather complex:
979          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
980          * 1-a: send as is if tlen <= path mtu
981          * 1-b: fragment if tlen > path mtu
982          *
983          * 2: if user asks us not to fragment (dontfrag == 1)
984          * 2-a: send as is if tlen <= interface mtu
985          * 2-b: error if tlen > interface mtu
986          *
987          * 3: if we always need to attach fragment header (alwaysfrag == 1)
988          *      always fragment
989          *
990          * 4: if dontfrag == 1 && alwaysfrag == 1
991          *      error, as we cannot handle this conflicting request
992          */
993         tlen = m->m_pkthdr.len;
994
995         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
996                 dontfrag = 1;
997         else
998                 dontfrag = 0;
999         if (dontfrag && alwaysfrag) {   /* case 4 */
1000                 /* conflicting request - can't transmit */
1001                 error = EMSGSIZE;
1002                 goto bad;
1003         }
1004         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
1005                 /*
1006                  * Even if the DONTFRAG option is specified, we cannot send the
1007                  * packet when the data length is larger than the MTU of the
1008                  * outgoing interface.
1009                  * Notify the error by sending IPV6_PATHMTU ancillary data as
1010                  * well as returning an error code (the latter is not described
1011                  * in the API spec.)
1012                  */
1013                 u_int32_t mtu32;
1014                 struct ip6ctlparam ip6cp;
1015
1016                 mtu32 = (u_int32_t)mtu;
1017                 bzero(&ip6cp, sizeof(ip6cp));
1018                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1019                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1020                     (void *)&ip6cp);
1021
1022                 error = EMSGSIZE;
1023                 goto bad;
1024         }
1025
1026         /*
1027          * transmit packet without fragmentation
1028          */
1029         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1030                 struct in6_ifaddr *ia6;
1031
1032                 ip6 = mtod(m, struct ip6_hdr *);
1033                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1034                 if (ia6) {
1035                         /* Record statistics for this interface address. */
1036                         ia6->ia_ifa.if_opackets++;
1037                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1038                 }
1039 #ifdef IPSEC
1040                 /* clean ipsec history once it goes out of the node */
1041                 ipsec_delaux(m);
1042 #endif
1043                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1044                 goto done;
1045         }
1046
1047         /*
1048          * try to fragment the packet.  case 1-b and 3
1049          */
1050         if (mtu < IPV6_MMTU) {
1051                 /* path MTU cannot be less than IPV6_MMTU */
1052                 error = EMSGSIZE;
1053                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1054                 goto bad;
1055         } else if (ip6->ip6_plen == 0) {
1056                 /* jumbo payload cannot be fragmented */
1057                 error = EMSGSIZE;
1058                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1059                 goto bad;
1060         } else {
1061                 struct mbuf **mnext, *m_frgpart;
1062                 struct ip6_frag *ip6f;
1063                 u_int32_t id = htonl(ip6_randomid());
1064                 u_char nextproto;
1065 #if 0
1066                 struct ip6ctlparam ip6cp;
1067                 u_int32_t mtu32;
1068 #endif
1069                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1070
1071                 /*
1072                  * Too large for the destination or interface;
1073                  * fragment if possible.
1074                  * Must be able to put at least 8 bytes per fragment.
1075                  */
1076                 hlen = unfragpartlen;
1077                 if (mtu > IPV6_MAXPACKET)
1078                         mtu = IPV6_MAXPACKET;
1079
1080 #if 0
1081                 /*
1082                  * It is believed this code is a leftover from the
1083                  * development of the IPV6_RECVPATHMTU sockopt and 
1084                  * associated work to implement RFC3542.
1085                  * It's not entirely clear what the intent of the API
1086                  * is at this point, so disable this code for now.
1087                  * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1088                  * will send notifications if the application requests.
1089                  */
1090
1091                 /* Notify a proper path MTU to applications. */
1092                 mtu32 = (u_int32_t)mtu;
1093                 bzero(&ip6cp, sizeof(ip6cp));
1094                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1095                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1096                     (void *)&ip6cp);
1097 #endif
1098
1099                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1100                 if (len < 8) {
1101                         error = EMSGSIZE;
1102                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1103                         goto bad;
1104                 }
1105
1106                 /*
1107                  * Verify that we have any chance at all of being able to queue
1108                  *      the packet or packet fragments
1109                  */
1110                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1111                     < tlen  /* - hlen */)) {
1112                         error = ENOBUFS;
1113                         ip6stat.ip6s_odropped++;
1114                         goto bad;
1115                 }
1116
1117                 mnext = &m->m_nextpkt;
1118
1119                 /*
1120                  * Change the next header field of the last header in the
1121                  * unfragmentable part.
1122                  */
1123                 if (exthdrs.ip6e_rthdr) {
1124                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1125                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1126                 } else if (exthdrs.ip6e_dest1) {
1127                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1128                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1129                 } else if (exthdrs.ip6e_hbh) {
1130                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1131                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1132                 } else {
1133                         nextproto = ip6->ip6_nxt;
1134                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1135                 }
1136
1137                 /*
1138                  * Loop through length of segment after first fragment,
1139                  * make new header and copy data of each part and link onto
1140                  * chain.
1141                  */
1142                 m0 = m;
1143                 for (off = hlen; off < tlen; off += len) {
1144                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1145                         if (!m) {
1146                                 error = ENOBUFS;
1147                                 ip6stat.ip6s_odropped++;
1148                                 goto sendorfree;
1149                         }
1150                         m->m_pkthdr.rcvif = NULL;
1151                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1152                         *mnext = m;
1153                         mnext = &m->m_nextpkt;
1154                         m->m_data += max_linkhdr;
1155                         mhip6 = mtod(m, struct ip6_hdr *);
1156                         *mhip6 = *ip6;
1157                         m->m_len = sizeof(*mhip6);
1158                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1159                         if (error) {
1160                                 ip6stat.ip6s_odropped++;
1161                                 goto sendorfree;
1162                         }
1163                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1164                         if (off + len >= tlen)
1165                                 len = tlen - off;
1166                         else
1167                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1168                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1169                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1170                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1171                                 error = ENOBUFS;
1172                                 ip6stat.ip6s_odropped++;
1173                                 goto sendorfree;
1174                         }
1175                         m_cat(m, m_frgpart);
1176                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1177                         m->m_pkthdr.rcvif = NULL;
1178                         ip6f->ip6f_reserved = 0;
1179                         ip6f->ip6f_ident = id;
1180                         ip6f->ip6f_nxt = nextproto;
1181                         ip6stat.ip6s_ofragments++;
1182                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1183                 }
1184
1185                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1186         }
1187
1188         /*
1189          * Remove leading garbages.
1190          */
1191 sendorfree:
1192         m = m0->m_nextpkt;
1193         m0->m_nextpkt = 0;
1194         m_freem(m0);
1195         for (m0 = m; m; m = m0) {
1196                 m0 = m->m_nextpkt;
1197                 m->m_nextpkt = 0;
1198                 if (error == 0) {
1199                         /* Record statistics for this interface address. */
1200                         if (ia) {
1201                                 ia->ia_ifa.if_opackets++;
1202                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1203                         }
1204 #ifdef IPSEC
1205                         /* clean ipsec history once it goes out of the node */
1206                         ipsec_delaux(m);
1207 #endif
1208                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1209                 } else
1210                         m_freem(m);
1211         }
1212
1213         if (error == 0)
1214                 ip6stat.ip6s_fragmented++;
1215
1216 done:
1217         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1218                 RTFREE(ro->ro_rt);
1219         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1220                 RTFREE(ro_pmtu->ro_rt);
1221         }
1222
1223 #ifdef IPSEC
1224         if (sp != NULL)
1225                 key_freesp(sp);
1226 #endif /* IPSEC */
1227 #ifdef FAST_IPSEC
1228         if (sp != NULL)
1229                 KEY_FREESP(&sp);
1230 #endif /* FAST_IPSEC */
1231
1232         return (error);
1233
1234 freehdrs:
1235         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1236         m_freem(exthdrs.ip6e_dest1);
1237         m_freem(exthdrs.ip6e_rthdr);
1238         m_freem(exthdrs.ip6e_dest2);
1239         /* FALLTHROUGH */
1240 bad:
1241         m_freem(m);
1242         goto done;
1243 }
1244
1245 static int
1246 ip6_copyexthdr(mp, hdr, hlen)
1247         struct mbuf **mp;
1248         caddr_t hdr;
1249         int hlen;
1250 {
1251         struct mbuf *m;
1252
1253         if (hlen > MCLBYTES)
1254                 return (ENOBUFS); /* XXX */
1255
1256         MGET(m, M_DONTWAIT, MT_DATA);
1257         if (!m)
1258                 return (ENOBUFS);
1259
1260         if (hlen > MLEN) {
1261                 MCLGET(m, M_DONTWAIT);
1262                 if ((m->m_flags & M_EXT) == 0) {
1263                         m_free(m);
1264                         return (ENOBUFS);
1265                 }
1266         }
1267         m->m_len = hlen;
1268         if (hdr)
1269                 bcopy(hdr, mtod(m, caddr_t), hlen);
1270
1271         *mp = m;
1272         return (0);
1273 }
1274
1275 /*
1276  * Insert jumbo payload option.
1277  */
1278 static int
1279 ip6_insert_jumboopt(exthdrs, plen)
1280         struct ip6_exthdrs *exthdrs;
1281         u_int32_t plen;
1282 {
1283         struct mbuf *mopt;
1284         u_char *optbuf;
1285         u_int32_t v;
1286
1287 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1288
1289         /*
1290          * If there is no hop-by-hop options header, allocate new one.
1291          * If there is one but it doesn't have enough space to store the
1292          * jumbo payload option, allocate a cluster to store the whole options.
1293          * Otherwise, use it to store the options.
1294          */
1295         if (exthdrs->ip6e_hbh == 0) {
1296                 MGET(mopt, M_DONTWAIT, MT_DATA);
1297                 if (mopt == 0)
1298                         return (ENOBUFS);
1299                 mopt->m_len = JUMBOOPTLEN;
1300                 optbuf = mtod(mopt, u_char *);
1301                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1302                 exthdrs->ip6e_hbh = mopt;
1303         } else {
1304                 struct ip6_hbh *hbh;
1305
1306                 mopt = exthdrs->ip6e_hbh;
1307                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1308                         /*
1309                          * XXX assumption:
1310                          * - exthdrs->ip6e_hbh is not referenced from places
1311                          *   other than exthdrs.
1312                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1313                          */
1314                         int oldoptlen = mopt->m_len;
1315                         struct mbuf *n;
1316
1317                         /*
1318                          * XXX: give up if the whole (new) hbh header does
1319                          * not fit even in an mbuf cluster.
1320                          */
1321                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1322                                 return (ENOBUFS);
1323
1324                         /*
1325                          * As a consequence, we must always prepare a cluster
1326                          * at this point.
1327                          */
1328                         MGET(n, M_DONTWAIT, MT_DATA);
1329                         if (n) {
1330                                 MCLGET(n, M_DONTWAIT);
1331                                 if ((n->m_flags & M_EXT) == 0) {
1332                                         m_freem(n);
1333                                         n = NULL;
1334                                 }
1335                         }
1336                         if (!n)
1337                                 return (ENOBUFS);
1338                         n->m_len = oldoptlen + JUMBOOPTLEN;
1339                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1340                             oldoptlen);
1341                         optbuf = mtod(n, caddr_t) + oldoptlen;
1342                         m_freem(mopt);
1343                         mopt = exthdrs->ip6e_hbh = n;
1344                 } else {
1345                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1346                         mopt->m_len += JUMBOOPTLEN;
1347                 }
1348                 optbuf[0] = IP6OPT_PADN;
1349                 optbuf[1] = 1;
1350
1351                 /*
1352                  * Adjust the header length according to the pad and
1353                  * the jumbo payload option.
1354                  */
1355                 hbh = mtod(mopt, struct ip6_hbh *);
1356                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1357         }
1358
1359         /* fill in the option. */
1360         optbuf[2] = IP6OPT_JUMBO;
1361         optbuf[3] = 4;
1362         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1363         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1364
1365         /* finally, adjust the packet header length */
1366         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1367
1368         return (0);
1369 #undef JUMBOOPTLEN
1370 }
1371
1372 /*
1373  * Insert fragment header and copy unfragmentable header portions.
1374  */
1375 static int
1376 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1377         struct mbuf *m0, *m;
1378         int hlen;
1379         struct ip6_frag **frghdrp;
1380 {
1381         struct mbuf *n, *mlast;
1382
1383         if (hlen > sizeof(struct ip6_hdr)) {
1384                 n = m_copym(m0, sizeof(struct ip6_hdr),
1385                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1386                 if (n == 0)
1387                         return (ENOBUFS);
1388                 m->m_next = n;
1389         } else
1390                 n = m;
1391
1392         /* Search for the last mbuf of unfragmentable part. */
1393         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1394                 ;
1395
1396         if ((mlast->m_flags & M_EXT) == 0 &&
1397             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1398                 /* use the trailing space of the last mbuf for the fragment hdr */
1399                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1400                     mlast->m_len);
1401                 mlast->m_len += sizeof(struct ip6_frag);
1402                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1403         } else {
1404                 /* allocate a new mbuf for the fragment header */
1405                 struct mbuf *mfrg;
1406
1407                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1408                 if (mfrg == 0)
1409                         return (ENOBUFS);
1410                 mfrg->m_len = sizeof(struct ip6_frag);
1411                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1412                 mlast->m_next = mfrg;
1413         }
1414
1415         return (0);
1416 }
1417
1418 static int
1419 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1420         struct route_in6 *ro_pmtu, *ro;
1421         struct ifnet *ifp;
1422         struct in6_addr *dst;
1423         u_long *mtup;
1424         int *alwaysfragp;
1425 {
1426         u_int32_t mtu = 0;
1427         int alwaysfrag = 0;
1428         int error = 0;
1429
1430         if (ro_pmtu != ro) {
1431                 /* The first hop and the final destination may differ. */
1432                 struct sockaddr_in6 *sa6_dst =
1433                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1434                 if (ro_pmtu->ro_rt &&
1435                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1436                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1437                         RTFREE(ro_pmtu->ro_rt);
1438                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1439                 }
1440                 if (ro_pmtu->ro_rt == NULL) {
1441                         bzero(sa6_dst, sizeof(*sa6_dst));
1442                         sa6_dst->sin6_family = AF_INET6;
1443                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1444                         sa6_dst->sin6_addr = *dst;
1445
1446                         rtalloc((struct route *)ro_pmtu);
1447                 }
1448         }
1449         if (ro_pmtu->ro_rt) {
1450                 u_int32_t ifmtu;
1451                 struct in_conninfo inc;
1452
1453                 bzero(&inc, sizeof(inc));
1454                 inc.inc_flags = 1; /* IPv6 */
1455                 inc.inc6_faddr = *dst;
1456
1457                 if (ifp == NULL)
1458                         ifp = ro_pmtu->ro_rt->rt_ifp;
1459                 ifmtu = IN6_LINKMTU(ifp);
1460                 mtu = tcp_hc_getmtu(&inc);
1461                 if (mtu)
1462                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1463                 else
1464                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1465                 if (mtu == 0)
1466                         mtu = ifmtu;
1467                 else if (mtu < IPV6_MMTU) {
1468                         /*
1469                          * RFC2460 section 5, last paragraph:
1470                          * if we record ICMPv6 too big message with
1471                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1472                          * or smaller, with framgent header attached.
1473                          * (fragment header is needed regardless from the
1474                          * packet size, for translators to identify packets)
1475                          */
1476                         alwaysfrag = 1;
1477                         mtu = IPV6_MMTU;
1478                 } else if (mtu > ifmtu) {
1479                         /*
1480                          * The MTU on the route is larger than the MTU on
1481                          * the interface!  This shouldn't happen, unless the
1482                          * MTU of the interface has been changed after the
1483                          * interface was brought up.  Change the MTU in the
1484                          * route to match the interface MTU (as long as the
1485                          * field isn't locked).
1486                          */
1487                         mtu = ifmtu;
1488                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1489                 }
1490         } else if (ifp) {
1491                 mtu = IN6_LINKMTU(ifp);
1492         } else
1493                 error = EHOSTUNREACH; /* XXX */
1494
1495         *mtup = mtu;
1496         if (alwaysfragp)
1497                 *alwaysfragp = alwaysfrag;
1498         return (error);
1499 }
1500
1501 /*
1502  * IP6 socket option processing.
1503  */
1504 int
1505 ip6_ctloutput(so, sopt)
1506         struct socket *so;
1507         struct sockopt *sopt;
1508 {
1509         int privileged, optdatalen, uproto;
1510         void *optdata;
1511         struct inpcb *in6p = sotoinpcb(so);
1512         int error, optval;
1513         int level, op, optname;
1514         int optlen;
1515         struct thread *td;
1516
1517         if (sopt) {
1518                 level = sopt->sopt_level;
1519                 op = sopt->sopt_dir;
1520                 optname = sopt->sopt_name;
1521                 optlen = sopt->sopt_valsize;
1522                 td = sopt->sopt_td;
1523         } else {
1524                 panic("ip6_ctloutput: arg soopt is NULL");
1525         }
1526         error = optval = 0;
1527
1528         privileged = (td == 0 || suser(td)) ? 0 : 1;
1529         uproto = (int)so->so_proto->pr_protocol;
1530
1531         if (level == IPPROTO_IPV6) {
1532                 switch (op) {
1533
1534                 case SOPT_SET:
1535                         switch (optname) {
1536                         case IPV6_2292PKTOPTIONS:
1537 #ifdef IPV6_PKTOPTIONS
1538                         case IPV6_PKTOPTIONS:
1539 #endif
1540                         {
1541                                 struct mbuf *m;
1542
1543                                 error = soopt_getm(sopt, &m); /* XXX */
1544                                 if (error != 0)
1545                                         break;
1546                                 error = soopt_mcopyin(sopt, m); /* XXX */
1547                                 if (error != 0)
1548                                         break;
1549                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1550                                                     m, so, sopt);
1551                                 m_freem(m); /* XXX */
1552                                 break;
1553                         }
1554
1555                         /*
1556                          * Use of some Hop-by-Hop options or some
1557                          * Destination options, might require special
1558                          * privilege.  That is, normal applications
1559                          * (without special privilege) might be forbidden
1560                          * from setting certain options in outgoing packets,
1561                          * and might never see certain options in received
1562                          * packets. [RFC 2292 Section 6]
1563                          * KAME specific note:
1564                          *  KAME prevents non-privileged users from sending or
1565                          *  receiving ANY hbh/dst options in order to avoid
1566                          *  overhead of parsing options in the kernel.
1567                          */
1568                         case IPV6_RECVHOPOPTS:
1569                         case IPV6_RECVDSTOPTS:
1570                         case IPV6_RECVRTHDRDSTOPTS:
1571                                 if (!privileged) {
1572                                         error = EPERM;
1573                                         break;
1574                                 }
1575                                 /* FALLTHROUGH */
1576                         case IPV6_UNICAST_HOPS:
1577                         case IPV6_HOPLIMIT:
1578                         case IPV6_FAITH:
1579
1580                         case IPV6_RECVPKTINFO:
1581                         case IPV6_RECVHOPLIMIT:
1582                         case IPV6_RECVRTHDR:
1583                         case IPV6_RECVPATHMTU:
1584                         case IPV6_RECVTCLASS:
1585                         case IPV6_V6ONLY:
1586                         case IPV6_AUTOFLOWLABEL:
1587                                 if (optlen != sizeof(int)) {
1588                                         error = EINVAL;
1589                                         break;
1590                                 }
1591                                 error = sooptcopyin(sopt, &optval,
1592                                         sizeof optval, sizeof optval);
1593                                 if (error)
1594                                         break;
1595                                 switch (optname) {
1596
1597                                 case IPV6_UNICAST_HOPS:
1598                                         if (optval < -1 || optval >= 256)
1599                                                 error = EINVAL;
1600                                         else {
1601                                                 /* -1 = kernel default */
1602                                                 in6p->in6p_hops = optval;
1603                                                 if ((in6p->in6p_vflag &
1604                                                      INP_IPV4) != 0)
1605                                                         in6p->inp_ip_ttl = optval;
1606                                         }
1607                                         break;
1608 #define OPTSET(bit) \
1609 do { \
1610         if (optval) \
1611                 in6p->in6p_flags |= (bit); \
1612         else \
1613                 in6p->in6p_flags &= ~(bit); \
1614 } while (/*CONSTCOND*/ 0)
1615 #define OPTSET2292(bit) \
1616 do { \
1617         in6p->in6p_flags |= IN6P_RFC2292; \
1618         if (optval) \
1619                 in6p->in6p_flags |= (bit); \
1620         else \
1621                 in6p->in6p_flags &= ~(bit); \
1622 } while (/*CONSTCOND*/ 0)
1623 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1624
1625                                 case IPV6_RECVPKTINFO:
1626                                         /* cannot mix with RFC2292 */
1627                                         if (OPTBIT(IN6P_RFC2292)) {
1628                                                 error = EINVAL;
1629                                                 break;
1630                                         }
1631                                         OPTSET(IN6P_PKTINFO);
1632                                         break;
1633
1634                                 case IPV6_HOPLIMIT:
1635                                 {
1636                                         struct ip6_pktopts **optp;
1637
1638                                         /* cannot mix with RFC2292 */
1639                                         if (OPTBIT(IN6P_RFC2292)) {
1640                                                 error = EINVAL;
1641                                                 break;
1642                                         }
1643                                         optp = &in6p->in6p_outputopts;
1644                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1645                                                            (u_char *)&optval,
1646                                                            sizeof(optval),
1647                                                            optp,
1648                                                            privileged, uproto);
1649                                         break;
1650                                 }
1651
1652                                 case IPV6_RECVHOPLIMIT:
1653                                         /* cannot mix with RFC2292 */
1654                                         if (OPTBIT(IN6P_RFC2292)) {
1655                                                 error = EINVAL;
1656                                                 break;
1657                                         }
1658                                         OPTSET(IN6P_HOPLIMIT);
1659                                         break;
1660
1661                                 case IPV6_RECVHOPOPTS:
1662                                         /* cannot mix with RFC2292 */
1663                                         if (OPTBIT(IN6P_RFC2292)) {
1664                                                 error = EINVAL;
1665                                                 break;
1666                                         }
1667                                         OPTSET(IN6P_HOPOPTS);
1668                                         break;
1669
1670                                 case IPV6_RECVDSTOPTS:
1671                                         /* cannot mix with RFC2292 */
1672                                         if (OPTBIT(IN6P_RFC2292)) {
1673                                                 error = EINVAL;
1674                                                 break;
1675                                         }
1676                                         OPTSET(IN6P_DSTOPTS);
1677                                         break;
1678
1679                                 case IPV6_RECVRTHDRDSTOPTS:
1680                                         /* cannot mix with RFC2292 */
1681                                         if (OPTBIT(IN6P_RFC2292)) {
1682                                                 error = EINVAL;
1683                                                 break;
1684                                         }
1685                                         OPTSET(IN6P_RTHDRDSTOPTS);
1686                                         break;
1687
1688                                 case IPV6_RECVRTHDR:
1689                                         /* cannot mix with RFC2292 */
1690                                         if (OPTBIT(IN6P_RFC2292)) {
1691                                                 error = EINVAL;
1692                                                 break;
1693                                         }
1694                                         OPTSET(IN6P_RTHDR);
1695                                         break;
1696
1697                                 case IPV6_FAITH:
1698                                         OPTSET(IN6P_FAITH);
1699                                         break;
1700
1701                                 case IPV6_RECVPATHMTU:
1702                                         /*
1703                                          * We ignore this option for TCP
1704                                          * sockets.
1705                                          * (RFC3542 leaves this case
1706                                          * unspecified.)
1707                                          */
1708                                         if (uproto != IPPROTO_TCP)
1709                                                 OPTSET(IN6P_MTU);
1710                                         break;
1711
1712                                 case IPV6_V6ONLY:
1713                                         /*
1714                                          * make setsockopt(IPV6_V6ONLY)
1715                                          * available only prior to bind(2).
1716                                          * see ipng mailing list, Jun 22 2001.
1717                                          */
1718                                         if (in6p->in6p_lport ||
1719                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1720                                                 error = EINVAL;
1721                                                 break;
1722                                         }
1723                                         OPTSET(IN6P_IPV6_V6ONLY);
1724                                         if (optval)
1725                                                 in6p->in6p_vflag &= ~INP_IPV4;
1726                                         else
1727                                                 in6p->in6p_vflag |= INP_IPV4;
1728                                         break;
1729                                 case IPV6_RECVTCLASS:
1730                                         /* cannot mix with RFC2292 XXX */
1731                                         if (OPTBIT(IN6P_RFC2292)) {
1732                                                 error = EINVAL;
1733                                                 break;
1734                                         }
1735                                         OPTSET(IN6P_TCLASS);
1736                                         break;
1737                                 case IPV6_AUTOFLOWLABEL:
1738                                         OPTSET(IN6P_AUTOFLOWLABEL);
1739                                         break;
1740
1741                                 }
1742                                 break;
1743
1744                         case IPV6_TCLASS:
1745                         case IPV6_DONTFRAG:
1746                         case IPV6_USE_MIN_MTU:
1747                         case IPV6_PREFER_TEMPADDR:
1748                                 if (optlen != sizeof(optval)) {
1749                                         error = EINVAL;
1750                                         break;
1751                                 }
1752                                 error = sooptcopyin(sopt, &optval,
1753                                         sizeof optval, sizeof optval);
1754                                 if (error)
1755                                         break;
1756                                 {
1757                                         struct ip6_pktopts **optp;
1758                                         optp = &in6p->in6p_outputopts;
1759                                         error = ip6_pcbopt(optname,
1760                                                            (u_char *)&optval,
1761                                                            sizeof(optval),
1762                                                            optp,
1763                                                            privileged, uproto);
1764                                         break;
1765                                 }
1766
1767                         case IPV6_2292PKTINFO:
1768                         case IPV6_2292HOPLIMIT:
1769                         case IPV6_2292HOPOPTS:
1770                         case IPV6_2292DSTOPTS:
1771                         case IPV6_2292RTHDR:
1772                                 /* RFC 2292 */
1773                                 if (optlen != sizeof(int)) {
1774                                         error = EINVAL;
1775                                         break;
1776                                 }
1777                                 error = sooptcopyin(sopt, &optval,
1778                                         sizeof optval, sizeof optval);
1779                                 if (error)
1780                                         break;
1781                                 switch (optname) {
1782                                 case IPV6_2292PKTINFO:
1783                                         OPTSET2292(IN6P_PKTINFO);
1784                                         break;
1785                                 case IPV6_2292HOPLIMIT:
1786                                         OPTSET2292(IN6P_HOPLIMIT);
1787                                         break;
1788                                 case IPV6_2292HOPOPTS:
1789                                         /*
1790                                          * Check super-user privilege.
1791                                          * See comments for IPV6_RECVHOPOPTS.
1792                                          */
1793                                         if (!privileged)
1794                                                 return (EPERM);
1795                                         OPTSET2292(IN6P_HOPOPTS);
1796                                         break;
1797                                 case IPV6_2292DSTOPTS:
1798                                         if (!privileged)
1799                                                 return (EPERM);
1800                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1801                                         break;
1802                                 case IPV6_2292RTHDR:
1803                                         OPTSET2292(IN6P_RTHDR);
1804                                         break;
1805                                 }
1806                                 break;
1807                         case IPV6_PKTINFO:
1808                         case IPV6_HOPOPTS:
1809                         case IPV6_RTHDR:
1810                         case IPV6_DSTOPTS:
1811                         case IPV6_RTHDRDSTOPTS:
1812                         case IPV6_NEXTHOP:
1813                         {
1814                                 /* new advanced API (RFC3542) */
1815                                 u_char *optbuf;
1816                                 int optlen;
1817                                 struct ip6_pktopts **optp;
1818
1819                                 /* cannot mix with RFC2292 */
1820                                 if (OPTBIT(IN6P_RFC2292)) {
1821                                         error = EINVAL;
1822                                         break;
1823                                 }
1824
1825                                 switch (optname) {
1826                                 case IPV6_HOPOPTS:
1827                                 case IPV6_DSTOPTS:
1828                                 case IPV6_RTHDRDSTOPTS:
1829                                 case IPV6_NEXTHOP:
1830                                         if (!privileged)
1831                                                 error = EPERM;
1832                                         break;
1833                                 }
1834                                 if (error)
1835                                         break;
1836
1837                                 switch (optname) {
1838                                 case IPV6_PKTINFO:
1839                                         optlen = sizeof(struct in6_pktinfo);
1840                                         break;
1841                                 case IPV6_NEXTHOP:
1842                                         optlen = SOCK_MAXADDRLEN;
1843                                         break;
1844                                 default:
1845                                         optlen = IPV6_MAXOPTHDR;
1846                                         break;
1847                                 }
1848                                 if (sopt->sopt_valsize > optlen) {
1849                                         error = EINVAL;
1850                                         break;
1851                                 }
1852
1853                                 optlen = sopt->sopt_valsize;
1854                                 optbuf = malloc(optlen, M_TEMP, M_WAITOK);
1855                                 error = sooptcopyin(sopt, optbuf, optlen,
1856                                     optlen);
1857                                 if (error) {
1858                                         free(optbuf, M_TEMP);
1859                                         break;
1860                                 }
1861
1862                                 optp = &in6p->in6p_outputopts;
1863                                 error = ip6_pcbopt(optname,
1864                                                    optbuf, optlen,
1865                                                    optp, privileged, uproto);
1866                                 free(optbuf, M_TEMP);
1867                                 break;
1868                         }
1869 #undef OPTSET
1870
1871                         case IPV6_MULTICAST_IF:
1872                         case IPV6_MULTICAST_HOPS:
1873                         case IPV6_MULTICAST_LOOP:
1874                         case IPV6_JOIN_GROUP:
1875                         case IPV6_LEAVE_GROUP:
1876                             {
1877                                 if (sopt->sopt_valsize > MLEN) {
1878                                         error = EMSGSIZE;
1879                                         break;
1880                                 }
1881                                 /* XXX */
1882                             }
1883                             /* FALLTHROUGH */
1884                             {
1885                                 struct mbuf *m;
1886
1887                                 if (sopt->sopt_valsize > MCLBYTES) {
1888                                         error = EMSGSIZE;
1889                                         break;
1890                                 }
1891                                 /* XXX */
1892                                 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_HEADER);
1893                                 if (m == 0) {
1894                                         error = ENOBUFS;
1895                                         break;
1896                                 }
1897                                 if (sopt->sopt_valsize > MLEN) {
1898                                         MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1899                                         if ((m->m_flags & M_EXT) == 0) {
1900                                                 m_free(m);
1901                                                 error = ENOBUFS;
1902                                                 break;
1903                                         }
1904                                 }
1905                                 m->m_len = sopt->sopt_valsize;
1906                                 error = sooptcopyin(sopt, mtod(m, char *),
1907                                                     m->m_len, m->m_len);
1908                                 if (error) {
1909                                         (void)m_free(m);
1910                                         break;
1911                                 }
1912                                 error = ip6_setmoptions(sopt->sopt_name,
1913                                                         &in6p->in6p_moptions,
1914                                                         m);
1915                                 (void)m_free(m);
1916                             }
1917                                 break;
1918
1919                         case IPV6_PORTRANGE:
1920                                 error = sooptcopyin(sopt, &optval,
1921                                     sizeof optval, sizeof optval);
1922                                 if (error)
1923                                         break;
1924
1925                                 switch (optval) {
1926                                 case IPV6_PORTRANGE_DEFAULT:
1927                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1928                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1929                                         break;
1930
1931                                 case IPV6_PORTRANGE_HIGH:
1932                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1933                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1934                                         break;
1935
1936                                 case IPV6_PORTRANGE_LOW:
1937                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1938                                         in6p->in6p_flags |= IN6P_LOWPORT;
1939                                         break;
1940
1941                                 default:
1942                                         error = EINVAL;
1943                                         break;
1944                                 }
1945                                 break;
1946
1947 #if defined(IPSEC) || defined(FAST_IPSEC)
1948                         case IPV6_IPSEC_POLICY:
1949                             {
1950                                 caddr_t req = NULL;
1951                                 size_t len = 0;
1952                                 struct mbuf *m;
1953
1954                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1955                                         break;
1956                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1957                                         break;
1958                                 if (m) {
1959                                         req = mtod(m, caddr_t);
1960                                         len = m->m_len;
1961                                 }
1962                                 error = ipsec6_set_policy(in6p, optname, req,
1963                                                           len, privileged);
1964                                 m_freem(m);
1965                             }
1966                                 break;
1967 #endif /* KAME IPSEC */
1968
1969                         case IPV6_FW_ADD:
1970                         case IPV6_FW_DEL:
1971                         case IPV6_FW_FLUSH:
1972                         case IPV6_FW_ZERO:
1973                             {
1974                                 struct mbuf *m;
1975                                 struct mbuf **mp = &m;
1976
1977                                 if (ip6_fw_ctl_ptr == NULL)
1978                                         return EINVAL;
1979                                 /* XXX */
1980                                 if ((error = soopt_getm(sopt, &m)) != 0)
1981                                         break;
1982                                 /* XXX */
1983                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
1984                                         break;
1985                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1986                                 m = *mp;
1987                             }
1988                                 break;
1989
1990                         default:
1991                                 error = ENOPROTOOPT;
1992                                 break;
1993                         }
1994                         break;
1995
1996                 case SOPT_GET:
1997                         switch (optname) {
1998
1999                         case IPV6_2292PKTOPTIONS:
2000 #ifdef IPV6_PKTOPTIONS
2001                         case IPV6_PKTOPTIONS:
2002 #endif
2003                                 /*
2004                                  * RFC3542 (effectively) deprecated the
2005                                  * semantics of the 2292-style pktoptions.
2006                                  * Since it was not reliable in nature (i.e.,
2007                                  * applications had to expect the lack of some
2008                                  * information after all), it would make sense
2009                                  * to simplify this part by always returning
2010                                  * empty data.
2011                                  */
2012                                 sopt->sopt_valsize = 0;
2013                                 break;
2014
2015                         case IPV6_RECVHOPOPTS:
2016                         case IPV6_RECVDSTOPTS:
2017                         case IPV6_RECVRTHDRDSTOPTS:
2018                         case IPV6_UNICAST_HOPS:
2019                         case IPV6_RECVPKTINFO:
2020                         case IPV6_RECVHOPLIMIT:
2021                         case IPV6_RECVRTHDR:
2022                         case IPV6_RECVPATHMTU:
2023
2024                         case IPV6_FAITH:
2025                         case IPV6_V6ONLY:
2026                         case IPV6_PORTRANGE:
2027                         case IPV6_RECVTCLASS:
2028                         case IPV6_AUTOFLOWLABEL:
2029                                 switch (optname) {
2030
2031                                 case IPV6_RECVHOPOPTS:
2032                                         optval = OPTBIT(IN6P_HOPOPTS);
2033                                         break;
2034
2035                                 case IPV6_RECVDSTOPTS:
2036                                         optval = OPTBIT(IN6P_DSTOPTS);
2037                                         break;
2038
2039                                 case IPV6_RECVRTHDRDSTOPTS:
2040                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2041                                         break;
2042
2043                                 case IPV6_UNICAST_HOPS:
2044                                         optval = in6p->in6p_hops;
2045                                         break;
2046
2047                                 case IPV6_RECVPKTINFO:
2048                                         optval = OPTBIT(IN6P_PKTINFO);
2049                                         break;
2050
2051                                 case IPV6_RECVHOPLIMIT:
2052                                         optval = OPTBIT(IN6P_HOPLIMIT);
2053                                         break;
2054
2055                                 case IPV6_RECVRTHDR:
2056                                         optval = OPTBIT(IN6P_RTHDR);
2057                                         break;
2058
2059                                 case IPV6_RECVPATHMTU:
2060                                         optval = OPTBIT(IN6P_MTU);
2061                                         break;
2062
2063                                 case IPV6_FAITH:
2064                                         optval = OPTBIT(IN6P_FAITH);
2065                                         break;
2066
2067                                 case IPV6_V6ONLY:
2068                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2069                                         break;
2070
2071                                 case IPV6_PORTRANGE:
2072                                     {
2073                                         int flags;
2074                                         flags = in6p->in6p_flags;
2075                                         if (flags & IN6P_HIGHPORT)
2076                                                 optval = IPV6_PORTRANGE_HIGH;
2077                                         else if (flags & IN6P_LOWPORT)
2078                                                 optval = IPV6_PORTRANGE_LOW;
2079                                         else
2080                                                 optval = 0;
2081                                         break;
2082                                     }
2083                                 case IPV6_RECVTCLASS:
2084                                         optval = OPTBIT(IN6P_TCLASS);
2085                                         break;
2086
2087                                 case IPV6_AUTOFLOWLABEL:
2088                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2089                                         break;
2090                                 }
2091                                 if (error)
2092                                         break;
2093                                 error = sooptcopyout(sopt, &optval,
2094                                         sizeof optval);
2095                                 break;
2096
2097                         case IPV6_PATHMTU:
2098                         {
2099                                 u_long pmtu = 0;
2100                                 struct ip6_mtuinfo mtuinfo;
2101                                 struct route_in6 sro;
2102
2103                                 bzero(&sro, sizeof(sro));
2104
2105                                 if (!(so->so_state & SS_ISCONNECTED))
2106                                         return (ENOTCONN);
2107                                 /*
2108                                  * XXX: we dot not consider the case of source
2109                                  * routing, or optional information to specify
2110                                  * the outgoing interface.
2111                                  */
2112                                 error = ip6_getpmtu(&sro, NULL, NULL,
2113                                     &in6p->in6p_faddr, &pmtu, NULL);
2114                                 if (sro.ro_rt)
2115                                         RTFREE(sro.ro_rt);
2116                                 if (error)
2117                                         break;
2118                                 if (pmtu > IPV6_MAXPACKET)
2119                                         pmtu = IPV6_MAXPACKET;
2120
2121                                 bzero(&mtuinfo, sizeof(mtuinfo));
2122                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2123                                 optdata = (void *)&mtuinfo;
2124                                 optdatalen = sizeof(mtuinfo);
2125                                 error = sooptcopyout(sopt, optdata,
2126                                     optdatalen);
2127                                 break;
2128                         }
2129
2130                         case IPV6_2292PKTINFO:
2131                         case IPV6_2292HOPLIMIT:
2132                         case IPV6_2292HOPOPTS:
2133                         case IPV6_2292RTHDR:
2134                         case IPV6_2292DSTOPTS:
2135                                 switch (optname) {
2136                                 case IPV6_2292PKTINFO:
2137                                         optval = OPTBIT(IN6P_PKTINFO);
2138                                         break;
2139                                 case IPV6_2292HOPLIMIT:
2140                                         optval = OPTBIT(IN6P_HOPLIMIT);
2141                                         break;
2142                                 case IPV6_2292HOPOPTS:
2143                                         optval = OPTBIT(IN6P_HOPOPTS);
2144                                         break;
2145                                 case IPV6_2292RTHDR:
2146                                         optval = OPTBIT(IN6P_RTHDR);
2147                                         break;
2148                                 case IPV6_2292DSTOPTS:
2149                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2150                                         break;
2151                                 }
2152                                 error = sooptcopyout(sopt, &optval,
2153                                     sizeof optval);
2154                                 break;
2155                         case IPV6_PKTINFO:
2156                         case IPV6_HOPOPTS:
2157                         case IPV6_RTHDR:
2158                         case IPV6_DSTOPTS:
2159                         case IPV6_RTHDRDSTOPTS:
2160                         case IPV6_NEXTHOP:
2161                         case IPV6_TCLASS:
2162                         case IPV6_DONTFRAG:
2163                         case IPV6_USE_MIN_MTU:
2164                         case IPV6_PREFER_TEMPADDR:
2165                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2166                                     optname, sopt);
2167                                 break;
2168
2169                         case IPV6_MULTICAST_IF:
2170                         case IPV6_MULTICAST_HOPS:
2171                         case IPV6_MULTICAST_LOOP:
2172                         case IPV6_JOIN_GROUP:
2173                         case IPV6_LEAVE_GROUP:
2174                             {
2175                                 struct mbuf *m;
2176                                 error = ip6_getmoptions(sopt->sopt_name,
2177                                     in6p->in6p_moptions, &m);
2178                                 if (error == 0)
2179                                         error = sooptcopyout(sopt,
2180                                             mtod(m, char *), m->m_len);
2181                                 m_freem(m);
2182                             }
2183                                 break;
2184
2185 #if defined(IPSEC) || defined(FAST_IPSEC)
2186                         case IPV6_IPSEC_POLICY:
2187                           {
2188                                 caddr_t req = NULL;
2189                                 size_t len = 0;
2190                                 struct mbuf *m = NULL;
2191                                 struct mbuf **mp = &m;
2192                                 size_t ovalsize = sopt->sopt_valsize;
2193                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2194
2195                                 error = soopt_getm(sopt, &m); /* XXX */
2196                                 if (error != 0)
2197                                         break;
2198                                 error = soopt_mcopyin(sopt, m); /* XXX */
2199                                 if (error != 0)
2200                                         break;
2201                                 sopt->sopt_valsize = ovalsize;
2202                                 sopt->sopt_val = oval;
2203                                 if (m) {
2204                                         req = mtod(m, caddr_t);
2205                                         len = m->m_len;
2206                                 }
2207                                 error = ipsec6_get_policy(in6p, req, len, mp);
2208                                 if (error == 0)
2209                                         error = soopt_mcopyout(sopt, m); /* XXX */
2210                                 if (error == 0 && m)
2211                                         m_freem(m);
2212                                 break;
2213                           }
2214 #endif /* KAME IPSEC */
2215
2216                         case IPV6_FW_GET:
2217                           {
2218                                 struct mbuf *m;
2219                                 struct mbuf **mp = &m;
2220
2221                                 if (ip6_fw_ctl_ptr == NULL)
2222                                 {
2223                                         return EINVAL;
2224                                 }
2225                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2226                                 if (error == 0)
2227                                         error = soopt_mcopyout(sopt, m); /* XXX */
2228                                 if (error == 0 && m)
2229                                         m_freem(m);
2230                           }
2231                                 break;
2232
2233                         default:
2234                                 error = ENOPROTOOPT;
2235                                 break;
2236                         }
2237                         break;
2238                 }
2239         } else {                /* level != IPPROTO_IPV6 */
2240                 error = EINVAL;
2241         }
2242         return (error);
2243 }
2244
2245 int
2246 ip6_raw_ctloutput(so, sopt)
2247         struct socket *so;
2248         struct sockopt *sopt;
2249 {
2250         int error = 0, optval, optlen;
2251         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2252         struct in6pcb *in6p = sotoin6pcb(so);
2253         int level, op, optname;
2254
2255         if (sopt) {
2256                 level = sopt->sopt_level;
2257                 op = sopt->sopt_dir;
2258                 optname = sopt->sopt_name;
2259                 optlen = sopt->sopt_valsize;
2260         } else
2261                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2262
2263         if (level != IPPROTO_IPV6) {
2264                 return (EINVAL);
2265         }
2266
2267         switch (optname) {
2268         case IPV6_CHECKSUM:
2269                 /*
2270                  * For ICMPv6 sockets, no modification allowed for checksum
2271                  * offset, permit "no change" values to help existing apps.
2272                  *
2273                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2274                  * for an ICMPv6 socket will fail."
2275                  * The current behavior does not meet RFC3542.
2276                  */
2277                 switch (op) {
2278                 case SOPT_SET:
2279                         if (optlen != sizeof(int)) {
2280                                 error = EINVAL;
2281                                 break;
2282                         }
2283                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2284                                             sizeof(optval));
2285                         if (error)
2286                                 break;
2287                         if ((optval % 2) != 0) {
2288                                 /* the API assumes even offset values */
2289                                 error = EINVAL;
2290                         } else if (so->so_proto->pr_protocol ==
2291                             IPPROTO_ICMPV6) {
2292                                 if (optval != icmp6off)
2293                                         error = EINVAL;
2294                         } else
2295                                 in6p->in6p_cksum = optval;
2296                         break;
2297
2298                 case SOPT_GET:
2299                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2300                                 optval = icmp6off;
2301                         else
2302                                 optval = in6p->in6p_cksum;
2303
2304                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2305                         break;
2306
2307                 default:
2308                         error = EINVAL;
2309                         break;
2310                 }
2311                 break;
2312
2313         default:
2314                 error = ENOPROTOOPT;
2315                 break;
2316         }
2317
2318         return (error);
2319 }
2320
2321 /*
2322  * Set up IP6 options in pcb for insertion in output packets or
2323  * specifying behavior of outgoing packets.
2324  */
2325 static int
2326 ip6_pcbopts(pktopt, m, so, sopt)
2327         struct ip6_pktopts **pktopt;
2328         struct mbuf *m;
2329         struct socket *so;
2330         struct sockopt *sopt;
2331 {
2332         struct ip6_pktopts *opt = *pktopt;
2333         int error = 0;
2334         struct thread *td = sopt->sopt_td;
2335         int priv = 0;
2336
2337         /* turn off any old options. */
2338         if (opt) {
2339 #ifdef DIAGNOSTIC
2340                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2341                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2342                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2343                         printf("ip6_pcbopts: all specified options are cleared.\n");
2344 #endif
2345                 ip6_clearpktopts(opt, -1);
2346         } else
2347                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2348         *pktopt = NULL;
2349
2350         if (!m || m->m_len == 0) {
2351                 /*
2352                  * Only turning off any previous options, regardless of
2353                  * whether the opt is just created or given.
2354                  */
2355                 free(opt, M_IP6OPT);
2356                 return (0);
2357         }
2358
2359         /*  set options specified by user. */
2360         if (td && !suser(td))
2361                 priv = 1;
2362         if ((error = ip6_setpktopts(m, opt, NULL, priv,
2363             so->so_proto->pr_protocol)) != 0) {
2364                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2365                 free(opt, M_IP6OPT);
2366                 return (error);
2367         }
2368         *pktopt = opt;
2369         return (0);
2370 }
2371
2372 /*
2373  * initialize ip6_pktopts.  beware that there are non-zero default values in
2374  * the struct.
2375  */
2376 void
2377 ip6_initpktopts(opt)
2378         struct ip6_pktopts *opt;
2379 {
2380
2381         bzero(opt, sizeof(*opt));
2382         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2383         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2384         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2385         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2386 }
2387
2388 static int
2389 ip6_pcbopt(optname, buf, len, pktopt, priv, uproto)
2390         int optname, len, priv;
2391         u_char *buf;
2392         struct ip6_pktopts **pktopt;
2393         int uproto;
2394 {
2395         struct ip6_pktopts *opt;
2396
2397         if (*pktopt == NULL) {
2398                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2399                     M_WAITOK);
2400                 ip6_initpktopts(*pktopt);
2401         }
2402         opt = *pktopt;
2403
2404         return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2405 }
2406
2407 static int
2408 ip6_getpcbopt(pktopt, optname, sopt)
2409         struct ip6_pktopts *pktopt;
2410         struct sockopt *sopt;
2411         int optname;
2412 {
2413         void *optdata = NULL;
2414         int optdatalen = 0;
2415         struct ip6_ext *ip6e;
2416         int error = 0;
2417         struct in6_pktinfo null_pktinfo;
2418         int deftclass = 0, on;
2419         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2420         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2421
2422         switch (optname) {
2423         case IPV6_PKTINFO:
2424                 if (pktopt && pktopt->ip6po_pktinfo)
2425                         optdata = (void *)pktopt->ip6po_pktinfo;
2426                 else {
2427                         /* XXX: we don't have to do this every time... */
2428                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2429                         optdata = (void *)&null_pktinfo;
2430                 }
2431                 optdatalen = sizeof(struct in6_pktinfo);
2432                 break;
2433         case IPV6_TCLASS:
2434                 if (pktopt && pktopt->ip6po_tclass >= 0)
2435                         optdata = (void *)&pktopt->ip6po_tclass;
2436                 else
2437                         optdata = (void *)&deftclass;
2438                 optdatalen = sizeof(int);
2439                 break;
2440         case IPV6_HOPOPTS:
2441                 if (pktopt && pktopt->ip6po_hbh) {
2442                         optdata = (void *)pktopt->ip6po_hbh;
2443                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2444                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2445                 }
2446                 break;
2447         case IPV6_RTHDR:
2448                 if (pktopt && pktopt->ip6po_rthdr) {
2449                         optdata = (void *)pktopt->ip6po_rthdr;
2450                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2451                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2452                 }
2453                 break;
2454         case IPV6_RTHDRDSTOPTS:
2455                 if (pktopt && pktopt->ip6po_dest1) {
2456                         optdata = (void *)pktopt->ip6po_dest1;
2457                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2458                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2459                 }
2460                 break;
2461         case IPV6_DSTOPTS:
2462                 if (pktopt && pktopt->ip6po_dest2) {
2463                         optdata = (void *)pktopt->ip6po_dest2;
2464                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2465                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2466                 }
2467                 break;
2468         case IPV6_NEXTHOP:
2469                 if (pktopt && pktopt->ip6po_nexthop) {
2470                         optdata = (void *)pktopt->ip6po_nexthop;
2471                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2472                 }
2473                 break;
2474         case IPV6_USE_MIN_MTU:
2475                 if (pktopt)
2476                         optdata = (void *)&pktopt->ip6po_minmtu;
2477                 else
2478                         optdata = (void *)&defminmtu;
2479                 optdatalen = sizeof(int);
2480                 break;
2481         case IPV6_DONTFRAG:
2482                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2483                         on = 1;
2484                 else
2485                         on = 0;
2486                 optdata = (void *)&on;
2487                 optdatalen = sizeof(on);
2488                 break;
2489         case IPV6_PREFER_TEMPADDR:
2490                 if (pktopt)
2491                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2492                 else
2493                         optdata = (void *)&defpreftemp;
2494                 optdatalen = sizeof(int);
2495                 break;
2496         default:                /* should not happen */
2497 #ifdef DIAGNOSTIC
2498                 panic("ip6_getpcbopt: unexpected option\n");
2499 #endif
2500                 return (ENOPROTOOPT);
2501         }
2502
2503         error = sooptcopyout(sopt, optdata, optdatalen);
2504
2505         return (error);
2506 }
2507
2508 void
2509 ip6_clearpktopts(pktopt, optname)
2510         struct ip6_pktopts *pktopt;
2511         int optname;
2512 {
2513         if (pktopt == NULL)
2514                 return;
2515
2516         if (optname == -1 || optname == IPV6_PKTINFO) {
2517                 if (pktopt->ip6po_pktinfo)
2518                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2519                 pktopt->ip6po_pktinfo = NULL;
2520         }
2521         if (optname == -1 || optname == IPV6_HOPLIMIT)
2522                 pktopt->ip6po_hlim = -1;
2523         if (optname == -1 || optname == IPV6_TCLASS)
2524                 pktopt->ip6po_tclass = -1;
2525         if (optname == -1 || optname == IPV6_NEXTHOP) {
2526                 if (pktopt->ip6po_nextroute.ro_rt) {
2527                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2528                         pktopt->ip6po_nextroute.ro_rt = NULL;
2529                 }
2530                 if (pktopt->ip6po_nexthop)
2531                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2532                 pktopt->ip6po_nexthop = NULL;
2533         }
2534         if (optname == -1 || optname == IPV6_HOPOPTS) {
2535                 if (pktopt->ip6po_hbh)
2536                         free(pktopt->ip6po_hbh, M_IP6OPT);
2537                 pktopt->ip6po_hbh = NULL;
2538         }
2539         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2540                 if (pktopt->ip6po_dest1)
2541                         free(pktopt->ip6po_dest1, M_IP6OPT);
2542                 pktopt->ip6po_dest1 = NULL;
2543         }
2544         if (optname == -1 || optname == IPV6_RTHDR) {
2545                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2546                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2547                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2548                 if (pktopt->ip6po_route.ro_rt) {
2549                         RTFREE(pktopt->ip6po_route.ro_rt);
2550                         pktopt->ip6po_route.ro_rt = NULL;
2551                 }
2552         }
2553         if (optname == -1 || optname == IPV6_DSTOPTS) {
2554                 if (pktopt->ip6po_dest2)
2555                         free(pktopt->ip6po_dest2, M_IP6OPT);
2556                 pktopt->ip6po_dest2 = NULL;
2557         }
2558 }
2559
2560 #define PKTOPT_EXTHDRCPY(type) \
2561 do {\
2562         if (src->type) {\
2563                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2564                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2565                 if (dst->type == NULL && canwait == M_NOWAIT)\
2566                         goto bad;\
2567                 bcopy(src->type, dst->type, hlen);\
2568         }\
2569 } while (/*CONSTCOND*/ 0)
2570
2571 static int
2572 copypktopts(dst, src, canwait)
2573         struct ip6_pktopts *dst, *src;
2574         int canwait;
2575 {
2576         if (dst == NULL || src == NULL)  {
2577                 printf("ip6_clearpktopts: invalid argument\n");
2578                 return (EINVAL);
2579         }
2580
2581         dst->ip6po_hlim = src->ip6po_hlim;
2582         dst->ip6po_tclass = src->ip6po_tclass;
2583         dst->ip6po_flags = src->ip6po_flags;
2584         if (src->ip6po_pktinfo) {
2585                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2586                     M_IP6OPT, canwait);
2587                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2588                         goto bad;
2589                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2590         }
2591         if (src->ip6po_nexthop) {
2592                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2593                     M_IP6OPT, canwait);
2594                 if (dst->ip6po_nexthop == NULL)
2595                         goto bad;
2596                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2597                     src->ip6po_nexthop->sa_len);
2598         }
2599         PKTOPT_EXTHDRCPY(ip6po_hbh);
2600         PKTOPT_EXTHDRCPY(ip6po_dest1);
2601         PKTOPT_EXTHDRCPY(ip6po_dest2);
2602         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2603         return (0);
2604
2605   bad:
2606         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2607         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2608         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2609         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2610         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2611         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2612         return (ENOBUFS);
2613 }
2614 #undef PKTOPT_EXTHDRCPY
2615
2616 struct ip6_pktopts *
2617 ip6_copypktopts(src, canwait)
2618         struct ip6_pktopts *src;
2619         int canwait;
2620 {
2621         int error;
2622         struct ip6_pktopts *dst;
2623
2624         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2625         if (dst == NULL && canwait == M_NOWAIT)
2626                 return (NULL);
2627         ip6_initpktopts(dst);
2628
2629         if ((error = copypktopts(dst, src, canwait)) != 0) {
2630                 free(dst, M_IP6OPT);
2631                 return (NULL);
2632         }
2633
2634         return (dst);
2635 }
2636
2637 void
2638 ip6_freepcbopts(pktopt)
2639         struct ip6_pktopts *pktopt;
2640 {
2641         if (pktopt == NULL)
2642                 return;
2643
2644         ip6_clearpktopts(pktopt, -1);
2645
2646         free(pktopt, M_IP6OPT);
2647 }
2648
2649 /*
2650  * Set the IP6 multicast options in response to user setsockopt().
2651  */
2652 static int
2653 ip6_setmoptions(optname, im6op, m)
2654         int optname;
2655         struct ip6_moptions **im6op;
2656         struct mbuf *m;
2657 {
2658         int error = 0;
2659         u_int loop, ifindex;
2660         struct ipv6_mreq *mreq;
2661         struct ifnet *ifp;
2662         struct ip6_moptions *im6o = *im6op;
2663         struct route_in6 ro;
2664         struct sockaddr_in6 *dst;
2665         struct in6_multi_mship *imm;
2666         struct thread *td = curthread;
2667
2668         if (im6o == NULL) {
2669                 /*
2670                  * No multicast option buffer attached to the pcb;
2671                  * allocate one and initialize to default values.
2672                  */
2673                 im6o = (struct ip6_moptions *)
2674                         malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2675
2676                 if (im6o == NULL)
2677                         return (ENOBUFS);
2678                 *im6op = im6o;
2679                 im6o->im6o_multicast_ifp = NULL;
2680                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2681                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2682                 LIST_INIT(&im6o->im6o_memberships);
2683         }
2684
2685         switch (optname) {
2686
2687         case IPV6_MULTICAST_IF:
2688                 /*
2689                  * Select the interface for outgoing multicast packets.
2690                  */
2691                 if (m == NULL || m->m_len != sizeof(u_int)) {
2692                         error = EINVAL;
2693                         break;
2694                 }
2695                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2696                 if (ifindex < 0 || if_index < ifindex) {
2697                         error = ENXIO;  /* XXX EINVAL? */
2698                         break;
2699                 }
2700                 ifp = ifnet_byindex(ifindex);
2701                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2702                         error = EADDRNOTAVAIL;
2703                         break;
2704                 }
2705                 im6o->im6o_multicast_ifp = ifp;
2706                 break;
2707
2708         case IPV6_MULTICAST_HOPS:
2709             {
2710                 /*
2711                  * Set the IP6 hoplimit for outgoing multicast packets.
2712                  */
2713                 int optval;
2714                 if (m == NULL || m->m_len != sizeof(int)) {
2715                         error = EINVAL;
2716                         break;
2717                 }
2718                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2719                 if (optval < -1 || optval >= 256)
2720                         error = EINVAL;
2721                 else if (optval == -1)
2722                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2723                 else
2724                         im6o->im6o_multicast_hlim = optval;
2725                 break;
2726             }
2727
2728         case IPV6_MULTICAST_LOOP:
2729                 /*
2730                  * Set the loopback flag for outgoing multicast packets.
2731                  * Must be zero or one.
2732                  */
2733                 if (m == NULL || m->m_len != sizeof(u_int)) {
2734                         error = EINVAL;
2735                         break;
2736                 }
2737                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2738                 if (loop > 1) {
2739                         error = EINVAL;
2740                         break;
2741                 }
2742                 im6o->im6o_multicast_loop = loop;
2743                 break;
2744
2745         case IPV6_JOIN_GROUP:
2746                 /*
2747                  * Add a multicast group membership.
2748                  * Group must be a valid IP6 multicast address.
2749                  */
2750                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2751                         error = EINVAL;
2752                         break;
2753                 }
2754                 mreq = mtod(m, struct ipv6_mreq *);
2755                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2756                         /*
2757                          * We use the unspecified address to specify to accept
2758                          * all multicast addresses. Only super user is allowed
2759                          * to do this.
2760                          */
2761                         if (suser(td)) {
2762                                 error = EACCES;
2763                                 break;
2764                         }
2765                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2766                         error = EINVAL;
2767                         break;
2768                 }
2769
2770                 /*
2771                  * If the interface is specified, validate it.
2772                  */
2773                 if (mreq->ipv6mr_interface < 0 ||
2774                     if_index < mreq->ipv6mr_interface) {
2775                         error = ENXIO;  /* XXX EINVAL? */
2776                         break;
2777                 }
2778                 /*
2779                  * If no interface was explicitly specified, choose an
2780                  * appropriate one according to the given multicast address.
2781                  */
2782                 if (mreq->ipv6mr_interface == 0) {
2783                         /*
2784                          * If the multicast address is in node-local scope,
2785                          * the interface should be a loopback interface.
2786                          * Otherwise, look up the routing table for the
2787                          * address, and choose the outgoing interface.
2788                          *   XXX: is it a good approach?
2789                          */
2790                         if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) {
2791                                 ifp = &loif[0];
2792                         } else {
2793                                 ro.ro_rt = NULL;
2794                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2795                                 bzero(dst, sizeof(*dst));
2796                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2797                                 dst->sin6_family = AF_INET6;
2798                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2799                                 rtalloc((struct route *)&ro);
2800                                 if (ro.ro_rt == NULL) {
2801                                         error = EADDRNOTAVAIL;
2802                                         break;
2803                                 }
2804                                 ifp = ro.ro_rt->rt_ifp;
2805                                 RTFREE(ro.ro_rt);
2806                         }
2807                 } else
2808                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2809
2810                 /*
2811                  * See if we found an interface, and confirm that it
2812                  * supports multicast
2813                  */
2814                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2815                         error = EADDRNOTAVAIL;
2816                         break;
2817                 }
2818                 /*
2819                  * Put interface index into the multicast address,
2820                  * if the address has link-local scope.
2821                  */
2822                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2823                         mreq->ipv6mr_multiaddr.s6_addr16[1] =
2824                             htons(ifp->if_index);
2825                 }
2826                 /*
2827                  * See if the membership already exists.
2828                  */
2829                 for (imm = im6o->im6o_memberships.lh_first;
2830                      imm != NULL; imm = imm->i6mm_chain.le_next)
2831                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2832                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2833                                                &mreq->ipv6mr_multiaddr))
2834                                 break;
2835                 if (imm != NULL) {
2836                         error = EADDRINUSE;
2837                         break;
2838                 }
2839                 /*
2840                  * Everything looks good; add a new record to the multicast
2841                  * address list for the given interface.
2842                  */
2843                 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2844                 if (imm == NULL) {
2845                         error = ENOBUFS;
2846                         break;
2847                 }
2848                 if ((imm->i6mm_maddr =
2849                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2850                         free(imm, M_IPMADDR);
2851                         break;
2852                 }
2853                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2854                 break;
2855
2856         case IPV6_LEAVE_GROUP:
2857                 /*
2858                  * Drop a multicast group membership.
2859                  * Group must be a valid IP6 multicast address.
2860                  */
2861                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2862                         error = EINVAL;
2863                         break;
2864                 }
2865                 mreq = mtod(m, struct ipv6_mreq *);
2866                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2867                         if (suser(td)) {
2868                                 error = EACCES;
2869                                 break;
2870                         }
2871                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2872                         error = EINVAL;
2873                         break;
2874                 }
2875                 /*
2876                  * If an interface address was specified, get a pointer
2877                  * to its ifnet structure.
2878                  */
2879                 if (mreq->ipv6mr_interface < 0
2880                  || if_index < mreq->ipv6mr_interface) {
2881                         error = ENXIO;  /* XXX EINVAL? */
2882                         break;
2883                 }
2884                 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2885                 /*
2886                  * Put interface index into the multicast address,
2887                  * if the address has link-local scope.
2888                  */
2889                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2890                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2891                                 = htons(mreq->ipv6mr_interface);
2892                 }
2893
2894                 /*
2895                  * Find the membership in the membership list.
2896                  */
2897                 for (imm = im6o->im6o_memberships.lh_first;
2898                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2899                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2900                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2901                             &mreq->ipv6mr_multiaddr))
2902                                 break;
2903                 }
2904                 if (imm == NULL) {
2905                         /* Unable to resolve interface */
2906                         error = EADDRNOTAVAIL;
2907                         break;
2908                 }
2909                 /*
2910                  * Give up the multicast address record to which the
2911                  * membership points.
2912                  */
2913                 LIST_REMOVE(imm, i6mm_chain);
2914                 in6_delmulti(imm->i6mm_maddr);
2915                 free(imm, M_IPMADDR);
2916                 break;
2917
2918         default:
2919                 error = EOPNOTSUPP;
2920                 break;
2921         }
2922
2923         /*
2924          * If all options have default values, no need to keep the mbuf.
2925          */
2926         if (im6o->im6o_multicast_ifp == NULL &&
2927             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2928             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2929             im6o->im6o_memberships.lh_first == NULL) {
2930                 free(*im6op, M_IPMOPTS);
2931                 *im6op = NULL;
2932         }
2933
2934         return (error);
2935 }
2936
2937 /*
2938  * Return the IP6 multicast options in response to user getsockopt().
2939  */
2940 static int
2941 ip6_getmoptions(optname, im6o, mp)
2942         int optname;
2943         struct ip6_moptions *im6o;
2944         struct mbuf **mp;
2945 {
2946         u_int *hlim, *loop, *ifindex;
2947
2948         *mp = m_get(M_TRYWAIT, MT_HEADER);              /* XXX */
2949
2950         switch (optname) {
2951
2952         case IPV6_MULTICAST_IF:
2953                 ifindex = mtod(*mp, u_int *);
2954                 (*mp)->m_len = sizeof(u_int);
2955                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2956                         *ifindex = 0;
2957                 else
2958                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2959                 return (0);
2960
2961         case IPV6_MULTICAST_HOPS:
2962                 hlim = mtod(*mp, u_int *);
2963                 (*mp)->m_len = sizeof(u_int);
2964                 if (im6o == NULL)
2965                         *hlim = ip6_defmcasthlim;
2966                 else
2967                         *hlim = im6o->im6o_multicast_hlim;
2968                 return (0);
2969
2970         case IPV6_MULTICAST_LOOP:
2971                 loop = mtod(*mp, u_int *);
2972                 (*mp)->m_len = sizeof(u_int);
2973                 if (im6o == NULL)
2974                         *loop = ip6_defmcasthlim;
2975                 else
2976                         *loop = im6o->im6o_multicast_loop;
2977                 return (0);
2978
2979         default:
2980                 return (EOPNOTSUPP);
2981         }
2982 }
2983
2984 /*
2985  * Discard the IP6 multicast options.
2986  */
2987 void
2988 ip6_freemoptions(im6o)
2989         struct ip6_moptions *im6o;
2990 {
2991         struct in6_multi_mship *imm;
2992
2993         if (im6o == NULL)
2994                 return;
2995
2996         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2997                 LIST_REMOVE(imm, i6mm_chain);
2998                 if (imm->i6mm_maddr)
2999                         in6_delmulti(imm->i6mm_maddr);
3000                 free(imm, M_IPMADDR);
3001         }
3002         free(im6o, M_IPMOPTS);
3003 }
3004
3005 /*
3006  * Set IPv6 outgoing packet options based on advanced API.
3007  */
3008 int
3009 ip6_setpktopts(control, opt, stickyopt, priv, uproto)
3010         struct mbuf *control;
3011         struct ip6_pktopts *opt, *stickyopt;
3012         int priv, uproto;
3013 {
3014         struct cmsghdr *cm = 0;
3015
3016         if (control == NULL || opt == NULL)
3017                 return (EINVAL);
3018
3019         ip6_initpktopts(opt);
3020         if (stickyopt) {
3021                 int error;
3022
3023                 /*
3024                  * If stickyopt is provided, make a local copy of the options
3025                  * for this particular packet, then override them by ancillary
3026                  * objects.
3027                  * XXX: copypktopts() does not copy the cached route to a next
3028                  * hop (if any).  This is not very good in terms of efficiency,
3029                  * but we can allow this since this option should be rarely
3030                  * used.
3031                  */
3032                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3033                         return (error);
3034         }
3035
3036         /*
3037          * XXX: Currently, we assume all the optional information is stored
3038          * in a single mbuf.
3039          */
3040         if (control->m_next)
3041                 return (EINVAL);
3042
3043         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3044             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3045                 int error;
3046
3047                 if (control->m_len < CMSG_LEN(0))
3048                         return (EINVAL);
3049
3050                 cm = mtod(control, struct cmsghdr *);
3051                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3052                         return (EINVAL);
3053                 if (cm->cmsg_level != IPPROTO_IPV6)
3054                         continue;
3055
3056                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
3057                     cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
3058                 if (error)
3059                         return (error);
3060         }
3061
3062         return (0);
3063 }
3064
3065 /*
3066  * Set a particular packet option, as a sticky option or an ancillary data
3067  * item.  "len" can be 0 only when it's a sticky option.
3068  * We have 4 cases of combination of "sticky" and "cmsg":
3069  * "sticky=0, cmsg=0": impossible
3070  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
3071  * "sticky=1, cmsg=0": RFC3542 socket option
3072  * "sticky=1, cmsg=1": RFC2292 socket option
3073  */
3074 static int
3075 ip6_setpktopt(optname, buf, len, opt, priv, sticky, cmsg, uproto)
3076         int optname, len, priv, sticky, cmsg, uproto;
3077         u_char *buf;
3078         struct ip6_pktopts *opt;
3079 {
3080         int minmtupolicy, preftemp;
3081
3082         if (!sticky && !cmsg) {
3083 #ifdef DIAGNOSTIC
3084                 printf("ip6_setpktopt: impossible case\n");
3085 #endif
3086                 return (EINVAL);
3087         }
3088
3089         /*
3090          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3091          * not be specified in the context of RFC3542.  Conversely,
3092          * RFC3542 types should not be specified in the context of RFC2292.
3093          */
3094         if (!cmsg) {
3095                 switch (optname) {
3096                 case IPV6_2292PKTINFO:
3097                 case IPV6_2292HOPLIMIT:
3098                 case IPV6_2292NEXTHOP:
3099                 case IPV6_2292HOPOPTS:
3100                 case IPV6_2292DSTOPTS:
3101                 case IPV6_2292RTHDR:
3102                 case IPV6_2292PKTOPTIONS:
3103                         return (ENOPROTOOPT);
3104                 }
3105         }
3106         if (sticky && cmsg) {
3107                 switch (optname) {
3108                 case IPV6_PKTINFO:
3109                 case IPV6_HOPLIMIT:
3110                 case IPV6_NEXTHOP:
3111                 case IPV6_HOPOPTS:
3112                 case IPV6_DSTOPTS:
3113                 case IPV6_RTHDRDSTOPTS:
3114                 case IPV6_RTHDR:
3115                 case IPV6_USE_MIN_MTU:
3116                 case IPV6_DONTFRAG:
3117                 case IPV6_TCLASS:
3118                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
3119                         return (ENOPROTOOPT);
3120                 }
3121         }
3122
3123         switch (optname) {
3124         case IPV6_2292PKTINFO:
3125         case IPV6_PKTINFO:
3126         {
3127                 struct ifnet *ifp = NULL;
3128                 struct in6_pktinfo *pktinfo;
3129
3130                 if (len != sizeof(struct in6_pktinfo))
3131                         return (EINVAL);
3132
3133                 pktinfo = (struct in6_pktinfo *)buf;
3134
3135                 /*
3136                  * An application can clear any sticky IPV6_PKTINFO option by
3137                  * doing a "regular" setsockopt with ipi6_addr being
3138                  * in6addr_any and ipi6_ifindex being zero.
3139                  * [RFC 3542, Section 6]
3140                  */
3141                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3142                     pktinfo->ipi6_ifindex == 0 &&
3143                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3144                         ip6_clearpktopts(opt, optname);
3145                         break;
3146                 }
3147
3148                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3149                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3150                         return (EINVAL);
3151                 }
3152
3153                 /* validate the interface index if specified. */
3154                 if (pktinfo->ipi6_ifindex > if_index ||
3155                     pktinfo->ipi6_ifindex < 0) {
3156                          return (ENXIO);
3157                 }
3158                 if (pktinfo->ipi6_ifindex) {
3159                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3160                         if (ifp == NULL)
3161                                 return (ENXIO);
3162                 }
3163
3164                 /*
3165                  * We store the address anyway, and let in6_selectsrc()
3166                  * validate the specified address.  This is because ipi6_addr
3167                  * may not have enough information about its scope zone, and
3168                  * we may need additional information (such as outgoing
3169                  * interface or the scope zone of a destination address) to
3170                  * disambiguate the scope.
3171                  * XXX: the delay of the validation may confuse the
3172                  * application when it is used as a sticky option.
3173                  */
3174                 if (opt->ip6po_pktinfo == NULL) {
3175                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3176                             M_IP6OPT, M_NOWAIT);
3177                         if (opt->ip6po_pktinfo == NULL)
3178                                 return (ENOBUFS);
3179                 }
3180                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3181                 break;
3182         }
3183
3184         case IPV6_2292HOPLIMIT:
3185         case IPV6_HOPLIMIT:
3186         {
3187                 int *hlimp;
3188
3189                 /*
3190                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3191                  * to simplify the ordering among hoplimit options.
3192                  */
3193                 if (optname == IPV6_HOPLIMIT && sticky)
3194                         return (ENOPROTOOPT);
3195
3196                 if (len != sizeof(int))
3197                         return (EINVAL);
3198                 hlimp = (int *)buf;
3199                 if (*hlimp < -1 || *hlimp > 255)
3200                         return (EINVAL);
3201
3202                 opt->ip6po_hlim = *hlimp;
3203                 break;
3204         }
3205
3206         case IPV6_TCLASS:
3207         {
3208                 int tclass;
3209
3210                 if (len != sizeof(int))
3211                         return (EINVAL);
3212                 tclass = *(int *)buf;
3213                 if (tclass < -1 || tclass > 255)
3214                         return (EINVAL);
3215
3216                 opt->ip6po_tclass = tclass;
3217                 break;
3218         }
3219
3220         case IPV6_2292NEXTHOP:
3221         case IPV6_NEXTHOP:
3222                 if (!priv)
3223                         return (EPERM);
3224
3225                 if (len == 0) { /* just remove the option */
3226                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3227                         break;
3228                 }
3229
3230                 /* check if cmsg_len is large enough for sa_len */
3231                 if (len < sizeof(struct sockaddr) || len < *buf)
3232                         return (EINVAL);
3233
3234                 switch (((struct sockaddr *)buf)->sa_family) {
3235                 case AF_INET6:
3236                 {
3237                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3238 #if 0
3239                         int error;
3240 #endif
3241
3242                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3243                                 return (EINVAL);
3244
3245                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3246                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3247                                 return (EINVAL);
3248                         }
3249 #if 0
3250                         if ((error = scope6_check_id(sa6, ip6_use_defzone))
3251                             != 0) {
3252                                 return (error);
3253                         }
3254 #endif
3255                         sa6->sin6_scope_id = 0; /* XXX */
3256                         break;
3257                 }
3258                 case AF_LINK:   /* should eventually be supported */
3259                 default:
3260                         return (EAFNOSUPPORT);
3261                 }
3262
3263                 /* turn off the previous option, then set the new option. */
3264                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3265                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_WAITOK);
3266                 bcopy(buf, opt->ip6po_nexthop, *buf);
3267                 break;
3268
3269         case IPV6_2292HOPOPTS:
3270         case IPV6_HOPOPTS:
3271         {
3272                 struct ip6_hbh *hbh;
3273                 int hbhlen;
3274
3275                 /*
3276                  * XXX: We don't allow a non-privileged user to set ANY HbH
3277                  * options, since per-option restriction has too much
3278                  * overhead.
3279                  */
3280                 if (!priv)
3281                         return (EPERM);
3282
3283                 if (len == 0) {
3284                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3285                         break;  /* just remove the option */
3286                 }
3287
3288                 /* message length validation */
3289                 if (len < sizeof(struct ip6_hbh))
3290                         return (EINVAL);
3291                 hbh = (struct ip6_hbh *)buf;
3292                 hbhlen = (hbh->ip6h_len + 1) << 3;
3293                 if (len != hbhlen)
3294                         return (EINVAL);
3295
3296                 /* turn off the previous option, then set the new option. */
3297                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3298                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_WAITOK);
3299                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3300
3301                 break;
3302         }
3303
3304         case IPV6_2292DSTOPTS:
3305         case IPV6_DSTOPTS:
3306         case IPV6_RTHDRDSTOPTS:
3307         {
3308                 struct ip6_dest *dest, **newdest = NULL;
3309                 int destlen;
3310
3311                 if (!priv)      /* XXX: see the comment for IPV6_HOPOPTS */
3312                         return (EPERM);
3313
3314                 if (len == 0) {
3315                         ip6_clearpktopts(opt, optname);
3316                         break;  /* just remove the option */
3317                 }
3318
3319                 /* message length validation */
3320                 if (len < sizeof(struct ip6_dest))
3321                         return (EINVAL);
3322                 dest = (struct ip6_dest *)buf;
3323                 destlen = (dest->ip6d_len + 1) << 3;
3324                 if (len != destlen)
3325                         return (EINVAL);
3326
3327                 /*
3328                  * Determine the position that the destination options header
3329                  * should be inserted; before or after the routing header.
3330                  */
3331                 switch (optname) {
3332                 case IPV6_2292DSTOPTS:
3333                         /*
3334                          * The old advacned API is ambiguous on this point.
3335                          * Our approach is to determine the position based
3336                          * according to the existence of a routing header.
3337                          * Note, however, that this depends on the order of the
3338                          * extension headers in the ancillary data; the 1st
3339                          * part of the destination options header must appear
3340                          * before the routing header in the ancillary data,
3341                          * too.
3342                          * RFC3542 solved the ambiguity by introducing
3343                          * separate ancillary data or option types.
3344                          */
3345                         if (opt->ip6po_rthdr == NULL)
3346                                 newdest = &opt->ip6po_dest1;
3347                         else
3348                                 newdest = &opt->ip6po_dest2;
3349                         break;
3350                 case IPV6_RTHDRDSTOPTS:
3351                         newdest = &opt->ip6po_dest1;
3352                         break;
3353                 case IPV6_DSTOPTS:
3354                         newdest = &opt->ip6po_dest2;
3355                         break;
3356                 }
3357
3358                 /* turn off the previous option, then set the new option. */
3359                 ip6_clearpktopts(opt, optname);
3360                 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
3361                 bcopy(dest, *newdest, destlen);
3362
3363                 break;
3364         }
3365
3366         case IPV6_2292RTHDR:
3367         case IPV6_RTHDR:
3368         {
3369                 struct ip6_rthdr *rth;
3370                 int rthlen;
3371
3372                 if (len == 0) {
3373                         ip6_clearpktopts(opt, IPV6_RTHDR);
3374                         break;  /* just remove the option */
3375                 }
3376
3377                 /* message length validation */
3378                 if (len < sizeof(struct ip6_rthdr))
3379                         return (EINVAL);
3380                 rth = (struct ip6_rthdr *)buf;
3381                 rthlen = (rth->ip6r_len + 1) << 3;
3382                 if (len != rthlen)
3383                         return (EINVAL);
3384
3385                 switch (rth->ip6r_type) {
3386                 case IPV6_RTHDR_TYPE_0:
3387                         if (rth->ip6r_len == 0) /* must contain one addr */
3388                                 return (EINVAL);
3389                         if (rth->ip6r_len % 2) /* length must be even */
3390                                 return (EINVAL);
3391                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3392                                 return (EINVAL);
3393                         break;
3394                 default:
3395                         return (EINVAL);        /* not supported */
3396                 }
3397
3398                 /* turn off the previous option */
3399                 ip6_clearpktopts(opt, IPV6_RTHDR);
3400                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_WAITOK);
3401                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3402
3403                 break;
3404         }
3405
3406         case IPV6_USE_MIN_MTU:
3407                 if (len != sizeof(int))
3408                         return (EINVAL);
3409                 minmtupolicy = *(int *)buf;
3410                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3411                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3412                     minmtupolicy != IP6PO_MINMTU_ALL) {
3413                         return (EINVAL);
3414                 }
3415                 opt->ip6po_minmtu = minmtupolicy;
3416                 break;
3417
3418         case IPV6_DONTFRAG:
3419                 if (len != sizeof(int))
3420                         return (EINVAL);
3421
3422                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3423                         /*
3424                          * we ignore this option for TCP sockets.
3425                          * (RFC3542 leaves this case unspecified.)
3426                          */
3427                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3428                 } else
3429                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3430                 break;
3431
3432         case IPV6_PREFER_TEMPADDR:
3433                 if (len != sizeof(int))
3434                         return (EINVAL);
3435                 preftemp = *(int *)buf;
3436                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3437                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3438                     preftemp != IP6PO_TEMPADDR_PREFER) {
3439                         return (EINVAL);
3440                 }
3441                 opt->ip6po_prefer_tempaddr = preftemp;
3442                 break;
3443
3444         default:
3445                 return (ENOPROTOOPT);
3446         } /* end of switch */
3447
3448         return (0);
3449 }
3450
3451 /*
3452  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3453  * packet to the input queue of a specified interface.  Note that this
3454  * calls the output routine of the loopback "driver", but with an interface
3455  * pointer that might NOT be &loif -- easier than replicating that code here.
3456  */
3457 void
3458 ip6_mloopback(ifp, m, dst)
3459         struct ifnet *ifp;
3460         struct mbuf *m;
3461         struct sockaddr_in6 *dst;
3462 {
3463         struct mbuf *copym;
3464         struct ip6_hdr *ip6;
3465
3466         copym = m_copy(m, 0, M_COPYALL);
3467         if (copym == NULL)
3468                 return;
3469
3470         /*
3471          * Make sure to deep-copy IPv6 header portion in case the data
3472          * is in an mbuf cluster, so that we can safely override the IPv6
3473          * header portion later.
3474          */
3475         if ((copym->m_flags & M_EXT) != 0 ||
3476             copym->m_len < sizeof(struct ip6_hdr)) {
3477                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3478                 if (copym == NULL)
3479                         return;
3480         }
3481
3482 #ifdef DIAGNOSTIC
3483         if (copym->m_len < sizeof(*ip6)) {
3484                 m_freem(copym);
3485                 return;
3486         }
3487 #endif
3488
3489         ip6 = mtod(copym, struct ip6_hdr *);
3490         /*
3491          * clear embedded scope identifiers if necessary.
3492          * in6_clearscope will touch the addresses only when necessary.
3493          */
3494         in6_clearscope(&ip6->ip6_src);
3495         in6_clearscope(&ip6->ip6_dst);
3496
3497         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3498 }
3499
3500 /*
3501  * Chop IPv6 header off from the payload.
3502  */
3503 static int
3504 ip6_splithdr(m, exthdrs)
3505         struct mbuf *m;
3506         struct ip6_exthdrs *exthdrs;
3507 {
3508         struct mbuf *mh;
3509         struct ip6_hdr *ip6;
3510
3511         ip6 = mtod(m, struct ip6_hdr *);
3512         if (m->m_len > sizeof(*ip6)) {
3513                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3514                 if (mh == 0) {
3515                         m_freem(m);
3516                         return ENOBUFS;
3517                 }
3518                 M_MOVE_PKTHDR(mh, m);
3519                 MH_ALIGN(mh, sizeof(*ip6));
3520                 m->m_len -= sizeof(*ip6);
3521                 m->m_data += sizeof(*ip6);
3522                 mh->m_next = m;
3523                 m = mh;
3524                 m->m_len = sizeof(*ip6);
3525                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3526         }
3527         exthdrs->ip6e_ip6 = m;
3528         return 0;
3529 }
3530
3531 /*
3532  * Compute IPv6 extension header length.
3533  */
3534 int
3535 ip6_optlen(in6p)
3536         struct in6pcb *in6p;
3537 {
3538         int len;
3539
3540         if (!in6p->in6p_outputopts)
3541                 return 0;
3542
3543         len = 0;
3544 #define elen(x) \
3545     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3546
3547         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3548         if (in6p->in6p_outputopts->ip6po_rthdr)
3549                 /* dest1 is valid with rthdr only */
3550                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3551         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3552         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3553         return len;
3554 #undef elen
3555 }