Deleted Added
full compact
ip6_output.c (102291) ip6_output.c (105194)
1/* $FreeBSD: head/sys/netinet6/ip6_output.c 102291 2002-08-22 21:24:01Z archie $ */
1/* $FreeBSD: head/sys/netinet6/ip6_output.c 105194 2002-10-16 01:54:46Z sam $ */
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 * 3. All advertising materials mentioning features or use of this software
46 * must display the following acknowledgement:
47 * This product includes software developed by the University of
48 * California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
66 */
67
68#include "opt_ip6fw.h"
69#include "opt_inet.h"
70#include "opt_inet6.h"
71#include "opt_ipsec.h"
72#include "opt_pfil_hooks.h"
73
74#include <sys/param.h>
75#include <sys/malloc.h>
76#include <sys/mbuf.h>
77#include <sys/proc.h>
78#include <sys/errno.h>
79#include <sys/protosw.h>
80#include <sys/socket.h>
81#include <sys/socketvar.h>
82#include <sys/systm.h>
83#include <sys/kernel.h>
84
85#include <net/if.h>
86#include <net/route.h>
87#ifdef PFIL_HOOKS
88#include <net/pfil.h>
89#endif
90
91#include <netinet/in.h>
92#include <netinet/in_var.h>
93#include <netinet6/in6_var.h>
94#include <netinet/ip6.h>
95#include <netinet/icmp6.h>
96#include <netinet6/ip6_var.h>
97#include <netinet/in_pcb.h>
98#include <netinet6/nd6.h>
99
100#ifdef IPSEC
101#include <netinet6/ipsec.h>
102#ifdef INET6
103#include <netinet6/ipsec6.h>
104#endif
105#include <netkey/key.h>
106#endif /* IPSEC */
107
108#include <netinet6/ip6_fw.h>
109
110#include <net/net_osdep.h>
111
112#include <netinet6/ip6protosw.h>
113
114static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
115
116struct ip6_exthdrs {
117 struct mbuf *ip6e_ip6;
118 struct mbuf *ip6e_hbh;
119 struct mbuf *ip6e_dest1;
120 struct mbuf *ip6e_rthdr;
121 struct mbuf *ip6e_dest2;
122};
123
124static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
125 struct socket *, struct sockopt *sopt));
126static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
127static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
128static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
129static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
130 struct ip6_frag **));
131static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
132static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
133
134/*
135 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
136 * header (with pri, len, nxt, hlim, src, dst).
137 * This function may modify ver and hlim only.
138 * The mbuf chain containing the packet will be freed.
139 * The mbuf opt, if present, will not be freed.
140 *
141 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
142 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
143 * which is rt_rmx.rmx_mtu.
144 */
145int
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 * 3. All advertising materials mentioning features or use of this software
46 * must display the following acknowledgement:
47 * This product includes software developed by the University of
48 * California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
66 */
67
68#include "opt_ip6fw.h"
69#include "opt_inet.h"
70#include "opt_inet6.h"
71#include "opt_ipsec.h"
72#include "opt_pfil_hooks.h"
73
74#include <sys/param.h>
75#include <sys/malloc.h>
76#include <sys/mbuf.h>
77#include <sys/proc.h>
78#include <sys/errno.h>
79#include <sys/protosw.h>
80#include <sys/socket.h>
81#include <sys/socketvar.h>
82#include <sys/systm.h>
83#include <sys/kernel.h>
84
85#include <net/if.h>
86#include <net/route.h>
87#ifdef PFIL_HOOKS
88#include <net/pfil.h>
89#endif
90
91#include <netinet/in.h>
92#include <netinet/in_var.h>
93#include <netinet6/in6_var.h>
94#include <netinet/ip6.h>
95#include <netinet/icmp6.h>
96#include <netinet6/ip6_var.h>
97#include <netinet/in_pcb.h>
98#include <netinet6/nd6.h>
99
100#ifdef IPSEC
101#include <netinet6/ipsec.h>
102#ifdef INET6
103#include <netinet6/ipsec6.h>
104#endif
105#include <netkey/key.h>
106#endif /* IPSEC */
107
108#include <netinet6/ip6_fw.h>
109
110#include <net/net_osdep.h>
111
112#include <netinet6/ip6protosw.h>
113
114static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
115
116struct ip6_exthdrs {
117 struct mbuf *ip6e_ip6;
118 struct mbuf *ip6e_hbh;
119 struct mbuf *ip6e_dest1;
120 struct mbuf *ip6e_rthdr;
121 struct mbuf *ip6e_dest2;
122};
123
124static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
125 struct socket *, struct sockopt *sopt));
126static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
127static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
128static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
129static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
130 struct ip6_frag **));
131static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
132static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
133
134/*
135 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
136 * header (with pri, len, nxt, hlim, src, dst).
137 * This function may modify ver and hlim only.
138 * The mbuf chain containing the packet will be freed.
139 * The mbuf opt, if present, will not be freed.
140 *
141 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
142 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
143 * which is rt_rmx.rmx_mtu.
144 */
145int
146ip6_output(m0, opt, ro, flags, im6o, ifpp)
146ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
147 struct mbuf *m0;
148 struct ip6_pktopts *opt;
149 struct route_in6 *ro;
150 int flags;
151 struct ip6_moptions *im6o;
152 struct ifnet **ifpp; /* XXX: just for statistics */
147 struct mbuf *m0;
148 struct ip6_pktopts *opt;
149 struct route_in6 *ro;
150 int flags;
151 struct ip6_moptions *im6o;
152 struct ifnet **ifpp; /* XXX: just for statistics */
153 struct inpcb *inp;
153{
154 struct ip6_hdr *ip6, *mhip6;
155 struct ifnet *ifp, *origifp;
156 struct mbuf *m = m0;
157 int hlen, tlen, len, off;
158 struct route_in6 ip6route;
159 struct sockaddr_in6 *dst;
160 int error = 0;
161 struct in6_ifaddr *ia = NULL;
162 u_long mtu;
163 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
164 struct ip6_exthdrs exthdrs;
165 struct in6_addr finaldst;
166 struct route_in6 *ro_pmtu = NULL;
167 int hdrsplit = 0;
168 int needipsec = 0;
169#ifdef PFIL_HOOKS
170 struct packet_filter_hook *pfh;
171 struct mbuf *m1;
172 int rv;
173#endif /* PFIL_HOOKS */
174#ifdef IPSEC
175 int needipsectun = 0;
154{
155 struct ip6_hdr *ip6, *mhip6;
156 struct ifnet *ifp, *origifp;
157 struct mbuf *m = m0;
158 int hlen, tlen, len, off;
159 struct route_in6 ip6route;
160 struct sockaddr_in6 *dst;
161 int error = 0;
162 struct in6_ifaddr *ia = NULL;
163 u_long mtu;
164 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
165 struct ip6_exthdrs exthdrs;
166 struct in6_addr finaldst;
167 struct route_in6 *ro_pmtu = NULL;
168 int hdrsplit = 0;
169 int needipsec = 0;
170#ifdef PFIL_HOOKS
171 struct packet_filter_hook *pfh;
172 struct mbuf *m1;
173 int rv;
174#endif /* PFIL_HOOKS */
175#ifdef IPSEC
176 int needipsectun = 0;
176 struct socket *so;
177 struct secpolicy *sp = NULL;
177 struct secpolicy *sp = NULL;
178 struct socket *so = inp ? inp->inp_socket : NULL;
178
179
179 /* for AH processing. stupid to have "socket" variable in IP layer... */
180 so = ipsec_getsocket(m);
181 (void)ipsec_setsocket(m, NULL);
182 ip6 = mtod(m, struct ip6_hdr *);
183#endif /* IPSEC */
184
185#define MAKE_EXTHDR(hp, mp) \
186 do { \
187 if (hp) { \
188 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
189 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
190 ((eh)->ip6e_len + 1) << 3); \
191 if (error) \
192 goto freehdrs; \
193 } \
194 } while (0)
195
196 bzero(&exthdrs, sizeof(exthdrs));
197
198 if (opt) {
199 /* Hop-by-Hop options header */
200 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
201 /* Destination options header(1st part) */
202 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
203 /* Routing header */
204 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
205 /* Destination options header(2nd part) */
206 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
207 }
208
209#ifdef IPSEC
210 /* get a security policy for this packet */
211 if (so == NULL)
212 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
213 else
214 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
215
216 if (sp == NULL) {
217 ipsec6stat.out_inval++;
218 goto freehdrs;
219 }
220
221 error = 0;
222
223 /* check policy */
224 switch (sp->policy) {
225 case IPSEC_POLICY_DISCARD:
226 /*
227 * This packet is just discarded.
228 */
229 ipsec6stat.out_polvio++;
230 goto freehdrs;
231
232 case IPSEC_POLICY_BYPASS:
233 case IPSEC_POLICY_NONE:
234 /* no need to do IPsec. */
235 needipsec = 0;
236 break;
237
238 case IPSEC_POLICY_IPSEC:
239 if (sp->req == NULL) {
240 /* acquire a policy */
241 error = key_spdacquire(sp);
242 goto freehdrs;
243 }
244 needipsec = 1;
245 break;
246
247 case IPSEC_POLICY_ENTRUST:
248 default:
249 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
250 }
251#endif /* IPSEC */
252
253 /*
254 * Calculate the total length of the extension header chain.
255 * Keep the length of the unfragmentable part for fragmentation.
256 */
257 optlen = 0;
258 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
259 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
260 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
261 unfragpartlen = optlen + sizeof(struct ip6_hdr);
262 /* NOTE: we don't add AH/ESP length here. do that later. */
263 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
264
265 /*
266 * If we need IPsec, or there is at least one extension header,
267 * separate IP6 header from the payload.
268 */
269 if ((needipsec || optlen) && !hdrsplit) {
270 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
271 m = NULL;
272 goto freehdrs;
273 }
274 m = exthdrs.ip6e_ip6;
275 hdrsplit++;
276 }
277
278 /* adjust pointer */
279 ip6 = mtod(m, struct ip6_hdr *);
280
281 /* adjust mbuf packet header length */
282 m->m_pkthdr.len += optlen;
283 plen = m->m_pkthdr.len - sizeof(*ip6);
284
285 /* If this is a jumbo payload, insert a jumbo payload option. */
286 if (plen > IPV6_MAXPACKET) {
287 if (!hdrsplit) {
288 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
289 m = NULL;
290 goto freehdrs;
291 }
292 m = exthdrs.ip6e_ip6;
293 hdrsplit++;
294 }
295 /* adjust pointer */
296 ip6 = mtod(m, struct ip6_hdr *);
297 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
298 goto freehdrs;
299 ip6->ip6_plen = 0;
300 } else
301 ip6->ip6_plen = htons(plen);
302
303 /*
304 * Concatenate headers and fill in next header fields.
305 * Here we have, on "m"
306 * IPv6 payload
307 * and we insert headers accordingly. Finally, we should be getting:
308 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
309 *
310 * during the header composing process, "m" points to IPv6 header.
311 * "mprev" points to an extension header prior to esp.
312 */
313 {
314 u_char *nexthdrp = &ip6->ip6_nxt;
315 struct mbuf *mprev = m;
316
317 /*
318 * we treat dest2 specially. this makes IPsec processing
319 * much easier. the goal here is to make mprev point the
320 * mbuf prior to dest2.
321 *
322 * result: IPv6 dest2 payload
323 * m and mprev will point to IPv6 header.
324 */
325 if (exthdrs.ip6e_dest2) {
326 if (!hdrsplit)
327 panic("assumption failed: hdr not split");
328 exthdrs.ip6e_dest2->m_next = m->m_next;
329 m->m_next = exthdrs.ip6e_dest2;
330 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
331 ip6->ip6_nxt = IPPROTO_DSTOPTS;
332 }
333
334#define MAKE_CHAIN(m, mp, p, i)\
335 do {\
336 if (m) {\
337 if (!hdrsplit) \
338 panic("assumption failed: hdr not split"); \
339 *mtod((m), u_char *) = *(p);\
340 *(p) = (i);\
341 p = mtod((m), u_char *);\
342 (m)->m_next = (mp)->m_next;\
343 (mp)->m_next = (m);\
344 (mp) = (m);\
345 }\
346 } while (0)
347 /*
348 * result: IPv6 hbh dest1 rthdr dest2 payload
349 * m will point to IPv6 header. mprev will point to the
350 * extension header prior to dest2 (rthdr in the above case).
351 */
352 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
353 nexthdrp, IPPROTO_HOPOPTS);
354 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
355 nexthdrp, IPPROTO_DSTOPTS);
356 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
357 nexthdrp, IPPROTO_ROUTING);
358
359#ifdef IPSEC
360 if (!needipsec)
361 goto skip_ipsec2;
362
363 /*
364 * pointers after IPsec headers are not valid any more.
365 * other pointers need a great care too.
366 * (IPsec routines should not mangle mbufs prior to AH/ESP)
367 */
368 exthdrs.ip6e_dest2 = NULL;
369
370 {
371 struct ip6_rthdr *rh = NULL;
372 int segleft_org = 0;
373 struct ipsec_output_state state;
374
375 if (exthdrs.ip6e_rthdr) {
376 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
377 segleft_org = rh->ip6r_segleft;
378 rh->ip6r_segleft = 0;
379 }
380
381 bzero(&state, sizeof(state));
382 state.m = m;
383 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
384 &needipsectun);
385 m = state.m;
386 if (error) {
387 /* mbuf is already reclaimed in ipsec6_output_trans. */
388 m = NULL;
389 switch (error) {
390 case EHOSTUNREACH:
391 case ENETUNREACH:
392 case EMSGSIZE:
393 case ENOBUFS:
394 case ENOMEM:
395 break;
396 default:
397 printf("ip6_output (ipsec): error code %d\n", error);
398 /* fall through */
399 case ENOENT:
400 /* don't show these error codes to the user */
401 error = 0;
402 break;
403 }
404 goto bad;
405 }
406 if (exthdrs.ip6e_rthdr) {
407 /* ah6_output doesn't modify mbuf chain */
408 rh->ip6r_segleft = segleft_org;
409 }
410 }
411skip_ipsec2:;
412#endif
413 }
414
415 /*
416 * If there is a routing header, replace destination address field
417 * with the first hop of the routing header.
418 */
419 if (exthdrs.ip6e_rthdr) {
420 struct ip6_rthdr *rh =
421 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
422 struct ip6_rthdr *));
423 struct ip6_rthdr0 *rh0;
424
425 finaldst = ip6->ip6_dst;
426 switch (rh->ip6r_type) {
427 case IPV6_RTHDR_TYPE_0:
428 rh0 = (struct ip6_rthdr0 *)rh;
429 ip6->ip6_dst = rh0->ip6r0_addr[0];
430 bcopy((caddr_t)&rh0->ip6r0_addr[1],
431 (caddr_t)&rh0->ip6r0_addr[0],
432 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
433 );
434 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
435 break;
436 default: /* is it possible? */
437 error = EINVAL;
438 goto bad;
439 }
440 }
441
442 /* Source address validation */
443 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
444 (flags & IPV6_DADOUTPUT) == 0) {
445 error = EOPNOTSUPP;
446 ip6stat.ip6s_badscope++;
447 goto bad;
448 }
449 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
450 error = EOPNOTSUPP;
451 ip6stat.ip6s_badscope++;
452 goto bad;
453 }
454
455 ip6stat.ip6s_localout++;
456
457 /*
458 * Route packet.
459 */
460 if (ro == 0) {
461 ro = &ip6route;
462 bzero((caddr_t)ro, sizeof(*ro));
463 }
464 ro_pmtu = ro;
465 if (opt && opt->ip6po_rthdr)
466 ro = &opt->ip6po_route;
467 dst = (struct sockaddr_in6 *)&ro->ro_dst;
468 /*
469 * If there is a cached route,
470 * check that it is to the same destination
471 * and is still up. If not, free it and try again.
472 */
473 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
474 dst->sin6_family != AF_INET6 ||
475 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
476 RTFREE(ro->ro_rt);
477 ro->ro_rt = (struct rtentry *)0;
478 }
479 if (ro->ro_rt == 0) {
480 bzero(dst, sizeof(*dst));
481 dst->sin6_family = AF_INET6;
482 dst->sin6_len = sizeof(struct sockaddr_in6);
483 dst->sin6_addr = ip6->ip6_dst;
484#ifdef SCOPEDROUTING
485 /* XXX: sin6_scope_id should already be fixed at this point */
486 if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
487 dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
488#endif
489 }
490#ifdef IPSEC
491 if (needipsec && needipsectun) {
492 struct ipsec_output_state state;
493
494 /*
495 * All the extension headers will become inaccessible
496 * (since they can be encrypted).
497 * Don't panic, we need no more updates to extension headers
498 * on inner IPv6 packet (since they are now encapsulated).
499 *
500 * IPv6 [ESP|AH] IPv6 [extension headers] payload
501 */
502 bzero(&exthdrs, sizeof(exthdrs));
503 exthdrs.ip6e_ip6 = m;
504
505 bzero(&state, sizeof(state));
506 state.m = m;
507 state.ro = (struct route *)ro;
508 state.dst = (struct sockaddr *)dst;
509
510 error = ipsec6_output_tunnel(&state, sp, flags);
511
512 m = state.m;
513 ro = (struct route_in6 *)state.ro;
514 dst = (struct sockaddr_in6 *)state.dst;
515 if (error) {
516 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
517 m0 = m = NULL;
518 m = NULL;
519 switch (error) {
520 case EHOSTUNREACH:
521 case ENETUNREACH:
522 case EMSGSIZE:
523 case ENOBUFS:
524 case ENOMEM:
525 break;
526 default:
527 printf("ip6_output (ipsec): error code %d\n", error);
528 /* fall through */
529 case ENOENT:
530 /* don't show these error codes to the user */
531 error = 0;
532 break;
533 }
534 goto bad;
535 }
536
537 exthdrs.ip6e_ip6 = m;
538 }
539#endif /* IPSEC */
540
541 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
542 /* Unicast */
543
544#define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
545#define sin6tosa(sin6) ((struct sockaddr *)(sin6))
546 /* xxx
547 * interface selection comes here
548 * if an interface is specified from an upper layer,
549 * ifp must point it.
550 */
551 if (ro->ro_rt == 0) {
552 /*
553 * non-bsdi always clone routes, if parent is
554 * PRF_CLONING.
555 */
556 rtalloc((struct route *)ro);
557 }
558 if (ro->ro_rt == 0) {
559 ip6stat.ip6s_noroute++;
560 error = EHOSTUNREACH;
561 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
562 goto bad;
563 }
564 ia = ifatoia6(ro->ro_rt->rt_ifa);
565 ifp = ro->ro_rt->rt_ifp;
566 ro->ro_rt->rt_use++;
567 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
568 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
569 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
570
571 in6_ifstat_inc(ifp, ifs6_out_request);
572
573 /*
574 * Check if the outgoing interface conflicts with
575 * the interface specified by ifi6_ifindex (if specified).
576 * Note that loopback interface is always okay.
577 * (this may happen when we are sending a packet to one of
578 * our own addresses.)
579 */
580 if (opt && opt->ip6po_pktinfo
581 && opt->ip6po_pktinfo->ipi6_ifindex) {
582 if (!(ifp->if_flags & IFF_LOOPBACK)
583 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
584 ip6stat.ip6s_noroute++;
585 in6_ifstat_inc(ifp, ifs6_out_discard);
586 error = EHOSTUNREACH;
587 goto bad;
588 }
589 }
590
591 if (opt && opt->ip6po_hlim != -1)
592 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
593 } else {
594 /* Multicast */
595 struct in6_multi *in6m;
596
597 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
598
599 /*
600 * See if the caller provided any multicast options
601 */
602 ifp = NULL;
603 if (im6o != NULL) {
604 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
605 if (im6o->im6o_multicast_ifp != NULL)
606 ifp = im6o->im6o_multicast_ifp;
607 } else
608 ip6->ip6_hlim = ip6_defmcasthlim;
609
610 /*
611 * See if the caller provided the outgoing interface
612 * as an ancillary data.
613 * Boundary check for ifindex is assumed to be already done.
614 */
615 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
616 ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
617
618 /*
619 * If the destination is a node-local scope multicast,
620 * the packet should be loop-backed only.
621 */
622 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
623 /*
624 * If the outgoing interface is already specified,
625 * it should be a loopback interface.
626 */
627 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
628 ip6stat.ip6s_badscope++;
629 error = ENETUNREACH; /* XXX: better error? */
630 /* XXX correct ifp? */
631 in6_ifstat_inc(ifp, ifs6_out_discard);
632 goto bad;
633 } else {
634 ifp = &loif[0];
635 }
636 }
637
638 if (opt && opt->ip6po_hlim != -1)
639 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
640
641 /*
642 * If caller did not provide an interface lookup a
643 * default in the routing table. This is either a
644 * default for the speicfied group (i.e. a host
645 * route), or a multicast default (a route for the
646 * ``net'' ff00::/8).
647 */
648 if (ifp == NULL) {
649 if (ro->ro_rt == 0) {
650 ro->ro_rt = rtalloc1((struct sockaddr *)
651 &ro->ro_dst, 0, 0UL);
652 }
653 if (ro->ro_rt == 0) {
654 ip6stat.ip6s_noroute++;
655 error = EHOSTUNREACH;
656 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
657 goto bad;
658 }
659 ia = ifatoia6(ro->ro_rt->rt_ifa);
660 ifp = ro->ro_rt->rt_ifp;
661 ro->ro_rt->rt_use++;
662 }
663
664 if ((flags & IPV6_FORWARDING) == 0)
665 in6_ifstat_inc(ifp, ifs6_out_request);
666 in6_ifstat_inc(ifp, ifs6_out_mcast);
667
668 /*
669 * Confirm that the outgoing interface supports multicast.
670 */
671 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
672 ip6stat.ip6s_noroute++;
673 in6_ifstat_inc(ifp, ifs6_out_discard);
674 error = ENETUNREACH;
675 goto bad;
676 }
677 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
678 if (in6m != NULL &&
679 (im6o == NULL || im6o->im6o_multicast_loop)) {
680 /*
681 * If we belong to the destination multicast group
682 * on the outgoing interface, and the caller did not
683 * forbid loopback, loop back a copy.
684 */
685 ip6_mloopback(ifp, m, dst);
686 } else {
687 /*
688 * If we are acting as a multicast router, perform
689 * multicast forwarding as if the packet had just
690 * arrived on the interface to which we are about
691 * to send. The multicast forwarding function
692 * recursively calls this function, using the
693 * IPV6_FORWARDING flag to prevent infinite recursion.
694 *
695 * Multicasts that are looped back by ip6_mloopback(),
696 * above, will be forwarded by the ip6_input() routine,
697 * if necessary.
698 */
699 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
700 if (ip6_mforward(ip6, ifp, m) != 0) {
701 m_freem(m);
702 goto done;
703 }
704 }
705 }
706 /*
707 * Multicasts with a hoplimit of zero may be looped back,
708 * above, but must not be transmitted on a network.
709 * Also, multicasts addressed to the loopback interface
710 * are not sent -- the above call to ip6_mloopback() will
711 * loop back a copy if this host actually belongs to the
712 * destination group on the loopback interface.
713 */
714 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
715 m_freem(m);
716 goto done;
717 }
718 }
719
720 /*
721 * Fill the outgoing inteface to tell the upper layer
722 * to increment per-interface statistics.
723 */
724 if (ifpp)
725 *ifpp = ifp;
726
727 /*
728 * Determine path MTU.
729 */
730 if (ro_pmtu != ro) {
731 /* The first hop and the final destination may differ. */
732 struct sockaddr_in6 *sin6_fin =
733 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
734 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
735 !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
736 &finaldst))) {
737 RTFREE(ro_pmtu->ro_rt);
738 ro_pmtu->ro_rt = (struct rtentry *)0;
739 }
740 if (ro_pmtu->ro_rt == 0) {
741 bzero(sin6_fin, sizeof(*sin6_fin));
742 sin6_fin->sin6_family = AF_INET6;
743 sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
744 sin6_fin->sin6_addr = finaldst;
745
746 rtalloc((struct route *)ro_pmtu);
747 }
748 }
749 if (ro_pmtu->ro_rt != NULL) {
750 u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
751
752 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
753 if (mtu > ifmtu || mtu == 0) {
754 /*
755 * The MTU on the route is larger than the MTU on
756 * the interface! This shouldn't happen, unless the
757 * MTU of the interface has been changed after the
758 * interface was brought up. Change the MTU in the
759 * route to match the interface MTU (as long as the
760 * field isn't locked).
761 *
762 * if MTU on the route is 0, we need to fix the MTU.
763 * this case happens with path MTU discovery timeouts.
764 */
765 mtu = ifmtu;
766 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
767 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
768 }
769 } else {
770 mtu = nd_ifinfo[ifp->if_index].linkmtu;
771 }
772
773 /*
774 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
775 */
776 if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
777 mtu = IPV6_MMTU;
778
779 /* Fake scoped addresses */
780 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
781 /*
782 * If source or destination address is a scoped address, and
783 * the packet is going to be sent to a loopback interface,
784 * we should keep the original interface.
785 */
786
787 /*
788 * XXX: this is a very experimental and temporary solution.
789 * We eventually have sockaddr_in6 and use the sin6_scope_id
790 * field of the structure here.
791 * We rely on the consistency between two scope zone ids
792 * of source and destination, which should already be assured.
793 * Larger scopes than link will be supported in the future.
794 */
795 origifp = NULL;
796 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
797 origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
798 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
799 origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
800 /*
801 * XXX: origifp can be NULL even in those two cases above.
802 * For example, if we remove the (only) link-local address
803 * from the loopback interface, and try to send a link-local
804 * address without link-id information. Then the source
805 * address is ::1, and the destination address is the
806 * link-local address with its s6_addr16[1] being zero.
807 * What is worse, if the packet goes to the loopback interface
808 * by a default rejected route, the null pointer would be
809 * passed to looutput, and the kernel would hang.
810 * The following last resort would prevent such disaster.
811 */
812 if (origifp == NULL)
813 origifp = ifp;
814 }
815 else
816 origifp = ifp;
817#ifndef SCOPEDROUTING
818 /*
819 * clear embedded scope identifiers if necessary.
820 * in6_clearscope will touch the addresses only when necessary.
821 */
822 in6_clearscope(&ip6->ip6_src);
823 in6_clearscope(&ip6->ip6_dst);
824#endif
825
826 /*
827 * Check with the firewall...
828 */
829 if (ip6_fw_enable && ip6_fw_chk_ptr) {
830 u_short port = 0;
831 m->m_pkthdr.rcvif = NULL; /* XXX */
832 /* If ipfw says divert, we have to just drop packet */
833 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
834 m_freem(m);
835 goto done;
836 }
837 if (!m) {
838 error = EACCES;
839 goto done;
840 }
841 }
842
843 /*
844 * If the outgoing packet contains a hop-by-hop options header,
845 * it must be examined and processed even by the source node.
846 * (RFC 2460, section 4.)
847 */
848 if (exthdrs.ip6e_hbh) {
849 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
850 u_int32_t dummy1; /* XXX unused */
851 u_int32_t dummy2; /* XXX unused */
852
853#ifdef DIAGNOSTIC
854 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
855 panic("ip6e_hbh is not continuous");
856#endif
857 /*
858 * XXX: if we have to send an ICMPv6 error to the sender,
859 * we need the M_LOOP flag since icmp6_error() expects
860 * the IPv6 and the hop-by-hop options header are
861 * continuous unless the flag is set.
862 */
863 m->m_flags |= M_LOOP;
864 m->m_pkthdr.rcvif = ifp;
865 if (ip6_process_hopopts(m,
866 (u_int8_t *)(hbh + 1),
867 ((hbh->ip6h_len + 1) << 3) -
868 sizeof(struct ip6_hbh),
869 &dummy1, &dummy2) < 0) {
870 /* m was already freed at this point */
871 error = EINVAL;/* better error? */
872 goto done;
873 }
874 m->m_flags &= ~M_LOOP; /* XXX */
875 m->m_pkthdr.rcvif = NULL;
876 }
877
878#ifdef PFIL_HOOKS
879 /*
880 * Run through list of hooks for output packets.
881 */
882 m1 = m;
883 pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
884 for (; pfh; pfh = pfh->pfil_link.tqe_next)
885 if (pfh->pfil_func) {
886 rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
887 if (rv) {
888 error = EHOSTUNREACH;
889 goto done;
890 }
891 m = m1;
892 if (m == NULL)
893 goto done;
894 ip6 = mtod(m, struct ip6_hdr *);
895 }
896#endif /* PFIL_HOOKS */
897 /*
898 * Send the packet to the outgoing interface.
899 * If necessary, do IPv6 fragmentation before sending.
900 */
901 tlen = m->m_pkthdr.len;
902 if (tlen <= mtu
903#ifdef notyet
904 /*
905 * On any link that cannot convey a 1280-octet packet in one piece,
906 * link-specific fragmentation and reassembly must be provided at
907 * a layer below IPv6. [RFC 2460, sec.5]
908 * Thus if the interface has ability of link-level fragmentation,
909 * we can just send the packet even if the packet size is
910 * larger than the link's MTU.
911 * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
912 */
913
914 || ifp->if_flags & IFF_FRAGMENTABLE
915#endif
916 )
917 {
918 /* Record statistics for this interface address. */
919 if (ia && !(flags & IPV6_FORWARDING)) {
920 ia->ia_ifa.if_opackets++;
921 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
922 }
923#ifdef IPSEC
924 /* clean ipsec history once it goes out of the node */
925 ipsec_delaux(m);
926#endif
927 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
928 goto done;
929 } else if (mtu < IPV6_MMTU) {
930 /*
931 * note that path MTU is never less than IPV6_MMTU
932 * (see icmp6_input).
933 */
934 error = EMSGSIZE;
935 in6_ifstat_inc(ifp, ifs6_out_fragfail);
936 goto bad;
937 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
938 error = EMSGSIZE;
939 in6_ifstat_inc(ifp, ifs6_out_fragfail);
940 goto bad;
941 } else {
942 struct mbuf **mnext, *m_frgpart;
943 struct ip6_frag *ip6f;
944 u_int32_t id = htonl(ip6_id++);
945 u_char nextproto;
946
947 /*
948 * Too large for the destination or interface;
949 * fragment if possible.
950 * Must be able to put at least 8 bytes per fragment.
951 */
952 hlen = unfragpartlen;
953 if (mtu > IPV6_MAXPACKET)
954 mtu = IPV6_MAXPACKET;
955
956 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
957 if (len < 8) {
958 error = EMSGSIZE;
959 in6_ifstat_inc(ifp, ifs6_out_fragfail);
960 goto bad;
961 }
962
963 mnext = &m->m_nextpkt;
964
965 /*
966 * Change the next header field of the last header in the
967 * unfragmentable part.
968 */
969 if (exthdrs.ip6e_rthdr) {
970 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
971 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
972 } else if (exthdrs.ip6e_dest1) {
973 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
974 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
975 } else if (exthdrs.ip6e_hbh) {
976 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
977 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
978 } else {
979 nextproto = ip6->ip6_nxt;
980 ip6->ip6_nxt = IPPROTO_FRAGMENT;
981 }
982
983 /*
984 * Loop through length of segment after first fragment,
985 * make new header and copy data of each part and link onto
986 * chain.
987 */
988 m0 = m;
989 for (off = hlen; off < tlen; off += len) {
990 MGETHDR(m, M_DONTWAIT, MT_HEADER);
991 if (!m) {
992 error = ENOBUFS;
993 ip6stat.ip6s_odropped++;
994 goto sendorfree;
995 }
996 m->m_pkthdr.rcvif = NULL;
997 m->m_flags = m0->m_flags & M_COPYFLAGS;
998 *mnext = m;
999 mnext = &m->m_nextpkt;
1000 m->m_data += max_linkhdr;
1001 mhip6 = mtod(m, struct ip6_hdr *);
1002 *mhip6 = *ip6;
1003 m->m_len = sizeof(*mhip6);
1004 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1005 if (error) {
1006 ip6stat.ip6s_odropped++;
1007 goto sendorfree;
1008 }
1009 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1010 if (off + len >= tlen)
1011 len = tlen - off;
1012 else
1013 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1014 mhip6->ip6_plen = htons((u_short)(len + hlen +
1015 sizeof(*ip6f) -
1016 sizeof(struct ip6_hdr)));
1017 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1018 error = ENOBUFS;
1019 ip6stat.ip6s_odropped++;
1020 goto sendorfree;
1021 }
1022 m_cat(m, m_frgpart);
1023 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1024 m->m_pkthdr.rcvif = (struct ifnet *)0;
1025 ip6f->ip6f_reserved = 0;
1026 ip6f->ip6f_ident = id;
1027 ip6f->ip6f_nxt = nextproto;
1028 ip6stat.ip6s_ofragments++;
1029 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1030 }
1031
1032 in6_ifstat_inc(ifp, ifs6_out_fragok);
1033 }
1034
1035 /*
1036 * Remove leading garbages.
1037 */
1038sendorfree:
1039 m = m0->m_nextpkt;
1040 m0->m_nextpkt = 0;
1041 m_freem(m0);
1042 for (m0 = m; m; m = m0) {
1043 m0 = m->m_nextpkt;
1044 m->m_nextpkt = 0;
1045 if (error == 0) {
1046 /* Record statistics for this interface address. */
1047 if (ia) {
1048 ia->ia_ifa.if_opackets++;
1049 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1050 }
1051#ifdef IPSEC
1052 /* clean ipsec history once it goes out of the node */
1053 ipsec_delaux(m);
1054#endif
1055 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1056 } else
1057 m_freem(m);
1058 }
1059
1060 if (error == 0)
1061 ip6stat.ip6s_fragmented++;
1062
1063done:
1064 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1065 RTFREE(ro->ro_rt);
1066 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1067 RTFREE(ro_pmtu->ro_rt);
1068 }
1069
1070#ifdef IPSEC
1071 if (sp != NULL)
1072 key_freesp(sp);
1073#endif /* IPSEC */
1074
1075 return(error);
1076
1077freehdrs:
1078 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1079 m_freem(exthdrs.ip6e_dest1);
1080 m_freem(exthdrs.ip6e_rthdr);
1081 m_freem(exthdrs.ip6e_dest2);
1082 /* fall through */
1083bad:
1084 m_freem(m);
1085 goto done;
1086}
1087
1088static int
1089ip6_copyexthdr(mp, hdr, hlen)
1090 struct mbuf **mp;
1091 caddr_t hdr;
1092 int hlen;
1093{
1094 struct mbuf *m;
1095
1096 if (hlen > MCLBYTES)
1097 return(ENOBUFS); /* XXX */
1098
1099 MGET(m, M_DONTWAIT, MT_DATA);
1100 if (!m)
1101 return(ENOBUFS);
1102
1103 if (hlen > MLEN) {
1104 MCLGET(m, M_DONTWAIT);
1105 if ((m->m_flags & M_EXT) == 0) {
1106 m_free(m);
1107 return(ENOBUFS);
1108 }
1109 }
1110 m->m_len = hlen;
1111 if (hdr)
1112 bcopy(hdr, mtod(m, caddr_t), hlen);
1113
1114 *mp = m;
1115 return(0);
1116}
1117
1118/*
1119 * Insert jumbo payload option.
1120 */
1121static int
1122ip6_insert_jumboopt(exthdrs, plen)
1123 struct ip6_exthdrs *exthdrs;
1124 u_int32_t plen;
1125{
1126 struct mbuf *mopt;
1127 u_char *optbuf;
1128 u_int32_t v;
1129
1130#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1131
1132 /*
1133 * If there is no hop-by-hop options header, allocate new one.
1134 * If there is one but it doesn't have enough space to store the
1135 * jumbo payload option, allocate a cluster to store the whole options.
1136 * Otherwise, use it to store the options.
1137 */
1138 if (exthdrs->ip6e_hbh == 0) {
1139 MGET(mopt, M_DONTWAIT, MT_DATA);
1140 if (mopt == 0)
1141 return(ENOBUFS);
1142 mopt->m_len = JUMBOOPTLEN;
1143 optbuf = mtod(mopt, u_char *);
1144 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1145 exthdrs->ip6e_hbh = mopt;
1146 } else {
1147 struct ip6_hbh *hbh;
1148
1149 mopt = exthdrs->ip6e_hbh;
1150 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1151 /*
1152 * XXX assumption:
1153 * - exthdrs->ip6e_hbh is not referenced from places
1154 * other than exthdrs.
1155 * - exthdrs->ip6e_hbh is not an mbuf chain.
1156 */
1157 int oldoptlen = mopt->m_len;
1158 struct mbuf *n;
1159
1160 /*
1161 * XXX: give up if the whole (new) hbh header does
1162 * not fit even in an mbuf cluster.
1163 */
1164 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1165 return(ENOBUFS);
1166
1167 /*
1168 * As a consequence, we must always prepare a cluster
1169 * at this point.
1170 */
1171 MGET(n, M_DONTWAIT, MT_DATA);
1172 if (n) {
1173 MCLGET(n, M_DONTWAIT);
1174 if ((n->m_flags & M_EXT) == 0) {
1175 m_freem(n);
1176 n = NULL;
1177 }
1178 }
1179 if (!n)
1180 return(ENOBUFS);
1181 n->m_len = oldoptlen + JUMBOOPTLEN;
1182 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1183 oldoptlen);
1184 optbuf = mtod(n, caddr_t) + oldoptlen;
1185 m_freem(mopt);
1186 mopt = exthdrs->ip6e_hbh = n;
1187 } else {
1188 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1189 mopt->m_len += JUMBOOPTLEN;
1190 }
1191 optbuf[0] = IP6OPT_PADN;
1192 optbuf[1] = 1;
1193
1194 /*
1195 * Adjust the header length according to the pad and
1196 * the jumbo payload option.
1197 */
1198 hbh = mtod(mopt, struct ip6_hbh *);
1199 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1200 }
1201
1202 /* fill in the option. */
1203 optbuf[2] = IP6OPT_JUMBO;
1204 optbuf[3] = 4;
1205 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1206 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1207
1208 /* finally, adjust the packet header length */
1209 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1210
1211 return(0);
1212#undef JUMBOOPTLEN
1213}
1214
1215/*
1216 * Insert fragment header and copy unfragmentable header portions.
1217 */
1218static int
1219ip6_insertfraghdr(m0, m, hlen, frghdrp)
1220 struct mbuf *m0, *m;
1221 int hlen;
1222 struct ip6_frag **frghdrp;
1223{
1224 struct mbuf *n, *mlast;
1225
1226 if (hlen > sizeof(struct ip6_hdr)) {
1227 n = m_copym(m0, sizeof(struct ip6_hdr),
1228 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1229 if (n == 0)
1230 return(ENOBUFS);
1231 m->m_next = n;
1232 } else
1233 n = m;
1234
1235 /* Search for the last mbuf of unfragmentable part. */
1236 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1237 ;
1238
1239 if ((mlast->m_flags & M_EXT) == 0 &&
1240 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1241 /* use the trailing space of the last mbuf for the fragment hdr */
1242 *frghdrp =
1243 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1244 mlast->m_len += sizeof(struct ip6_frag);
1245 m->m_pkthdr.len += sizeof(struct ip6_frag);
1246 } else {
1247 /* allocate a new mbuf for the fragment header */
1248 struct mbuf *mfrg;
1249
1250 MGET(mfrg, M_DONTWAIT, MT_DATA);
1251 if (mfrg == 0)
1252 return(ENOBUFS);
1253 mfrg->m_len = sizeof(struct ip6_frag);
1254 *frghdrp = mtod(mfrg, struct ip6_frag *);
1255 mlast->m_next = mfrg;
1256 }
1257
1258 return(0);
1259}
1260
1261/*
1262 * IP6 socket option processing.
1263 */
1264int
1265ip6_ctloutput(so, sopt)
1266 struct socket *so;
1267 struct sockopt *sopt;
1268{
1269 int privileged;
1270 struct inpcb *in6p = sotoinpcb(so);
1271 int error, optval;
1272 int level, op, optname;
1273 int optlen;
1274 struct thread *td;
1275
1276 if (sopt) {
1277 level = sopt->sopt_level;
1278 op = sopt->sopt_dir;
1279 optname = sopt->sopt_name;
1280 optlen = sopt->sopt_valsize;
1281 td = sopt->sopt_td;
1282 } else {
1283 panic("ip6_ctloutput: arg soopt is NULL");
1284 }
1285 error = optval = 0;
1286
1287 privileged = (td == 0 || suser(td)) ? 0 : 1;
1288
1289 if (level == IPPROTO_IPV6) {
1290 switch (op) {
1291
1292 case SOPT_SET:
1293 switch (optname) {
1294 case IPV6_PKTOPTIONS:
1295 {
1296 struct mbuf *m;
1297
1298 error = soopt_getm(sopt, &m); /* XXX */
1299 if (error != 0)
1300 break;
1301 error = soopt_mcopyin(sopt, m); /* XXX */
1302 if (error != 0)
1303 break;
1304 error = ip6_pcbopts(&in6p->in6p_outputopts,
1305 m, so, sopt);
1306 m_freem(m); /* XXX */
1307 break;
1308 }
1309
1310 /*
1311 * Use of some Hop-by-Hop options or some
1312 * Destination options, might require special
1313 * privilege. That is, normal applications
1314 * (without special privilege) might be forbidden
1315 * from setting certain options in outgoing packets,
1316 * and might never see certain options in received
1317 * packets. [RFC 2292 Section 6]
1318 * KAME specific note:
1319 * KAME prevents non-privileged users from sending or
1320 * receiving ANY hbh/dst options in order to avoid
1321 * overhead of parsing options in the kernel.
1322 */
1323 case IPV6_UNICAST_HOPS:
1324 case IPV6_CHECKSUM:
1325 case IPV6_FAITH:
1326
1327 case IPV6_V6ONLY:
1328 if (optlen != sizeof(int)) {
1329 error = EINVAL;
1330 break;
1331 }
1332 error = sooptcopyin(sopt, &optval,
1333 sizeof optval, sizeof optval);
1334 if (error)
1335 break;
1336 switch (optname) {
1337
1338 case IPV6_UNICAST_HOPS:
1339 if (optval < -1 || optval >= 256)
1340 error = EINVAL;
1341 else {
1342 /* -1 = kernel default */
1343 in6p->in6p_hops = optval;
1344
1345 if ((in6p->in6p_vflag &
1346 INP_IPV4) != 0)
1347 in6p->inp_ip_ttl = optval;
1348 }
1349 break;
1350#define OPTSET(bit) \
1351do { \
1352 if (optval) \
1353 in6p->in6p_flags |= (bit); \
1354 else \
1355 in6p->in6p_flags &= ~(bit); \
1356} while (0)
1357#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1358
1359 case IPV6_CHECKSUM:
1360 in6p->in6p_cksum = optval;
1361 break;
1362
1363 case IPV6_FAITH:
1364 OPTSET(IN6P_FAITH);
1365 break;
1366
1367 case IPV6_V6ONLY:
1368 /*
1369 * make setsockopt(IPV6_V6ONLY)
1370 * available only prior to bind(2).
1371 * see ipng mailing list, Jun 22 2001.
1372 */
1373 if (in6p->in6p_lport ||
1374 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1375 {
1376 error = EINVAL;
1377 break;
1378 }
1379 OPTSET(IN6P_IPV6_V6ONLY);
1380 if (optval)
1381 in6p->in6p_vflag &= ~INP_IPV4;
1382 else
1383 in6p->in6p_vflag |= INP_IPV4;
1384 break;
1385 }
1386 break;
1387
1388 case IPV6_PKTINFO:
1389 case IPV6_HOPLIMIT:
1390 case IPV6_HOPOPTS:
1391 case IPV6_DSTOPTS:
1392 case IPV6_RTHDR:
1393 /* RFC 2292 */
1394 if (optlen != sizeof(int)) {
1395 error = EINVAL;
1396 break;
1397 }
1398 error = sooptcopyin(sopt, &optval,
1399 sizeof optval, sizeof optval);
1400 if (error)
1401 break;
1402 switch (optname) {
1403 case IPV6_PKTINFO:
1404 OPTSET(IN6P_PKTINFO);
1405 break;
1406 case IPV6_HOPLIMIT:
1407 OPTSET(IN6P_HOPLIMIT);
1408 break;
1409 case IPV6_HOPOPTS:
1410 /*
1411 * Check super-user privilege.
1412 * See comments for IPV6_RECVHOPOPTS.
1413 */
1414 if (!privileged)
1415 return(EPERM);
1416 OPTSET(IN6P_HOPOPTS);
1417 break;
1418 case IPV6_DSTOPTS:
1419 if (!privileged)
1420 return(EPERM);
1421 OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1422 break;
1423 case IPV6_RTHDR:
1424 OPTSET(IN6P_RTHDR);
1425 break;
1426 }
1427 break;
1428#undef OPTSET
1429
1430 case IPV6_MULTICAST_IF:
1431 case IPV6_MULTICAST_HOPS:
1432 case IPV6_MULTICAST_LOOP:
1433 case IPV6_JOIN_GROUP:
1434 case IPV6_LEAVE_GROUP:
1435 {
1436 struct mbuf *m;
1437 if (sopt->sopt_valsize > MLEN) {
1438 error = EMSGSIZE;
1439 break;
1440 }
1441 /* XXX */
1442 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1443 if (m == 0) {
1444 error = ENOBUFS;
1445 break;
1446 }
1447 m->m_len = sopt->sopt_valsize;
1448 error = sooptcopyin(sopt, mtod(m, char *),
1449 m->m_len, m->m_len);
1450 error = ip6_setmoptions(sopt->sopt_name,
1451 &in6p->in6p_moptions,
1452 m);
1453 (void)m_free(m);
1454 }
1455 break;
1456
1457 case IPV6_PORTRANGE:
1458 error = sooptcopyin(sopt, &optval,
1459 sizeof optval, sizeof optval);
1460 if (error)
1461 break;
1462
1463 switch (optval) {
1464 case IPV6_PORTRANGE_DEFAULT:
1465 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1466 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1467 break;
1468
1469 case IPV6_PORTRANGE_HIGH:
1470 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1471 in6p->in6p_flags |= IN6P_HIGHPORT;
1472 break;
1473
1474 case IPV6_PORTRANGE_LOW:
1475 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1476 in6p->in6p_flags |= IN6P_LOWPORT;
1477 break;
1478
1479 default:
1480 error = EINVAL;
1481 break;
1482 }
1483 break;
1484
1485#ifdef IPSEC
1486 case IPV6_IPSEC_POLICY:
1487 {
1488 caddr_t req = NULL;
1489 size_t len = 0;
1490 struct mbuf *m;
1491
1492 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1493 break;
1494 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1495 break;
1496 if (m) {
1497 req = mtod(m, caddr_t);
1498 len = m->m_len;
1499 }
1500 error = ipsec6_set_policy(in6p, optname, req,
1501 len, privileged);
1502 m_freem(m);
1503 }
1504 break;
1505#endif /* KAME IPSEC */
1506
1507 case IPV6_FW_ADD:
1508 case IPV6_FW_DEL:
1509 case IPV6_FW_FLUSH:
1510 case IPV6_FW_ZERO:
1511 {
1512 struct mbuf *m;
1513 struct mbuf **mp = &m;
1514
1515 if (ip6_fw_ctl_ptr == NULL)
1516 return EINVAL;
1517 /* XXX */
1518 if ((error = soopt_getm(sopt, &m)) != 0)
1519 break;
1520 /* XXX */
1521 if ((error = soopt_mcopyin(sopt, m)) != 0)
1522 break;
1523 error = (*ip6_fw_ctl_ptr)(optname, mp);
1524 m = *mp;
1525 }
1526 break;
1527
1528 default:
1529 error = ENOPROTOOPT;
1530 break;
1531 }
1532 break;
1533
1534 case SOPT_GET:
1535 switch (optname) {
1536
1537 case IPV6_PKTOPTIONS:
1538 if (in6p->in6p_options) {
1539 struct mbuf *m;
1540 m = m_copym(in6p->in6p_options,
1541 0, M_COPYALL, M_TRYWAIT);
1542 error = soopt_mcopyout(sopt, m);
1543 if (error == 0)
1544 m_freem(m);
1545 } else
1546 sopt->sopt_valsize = 0;
1547 break;
1548
1549 case IPV6_UNICAST_HOPS:
1550 case IPV6_CHECKSUM:
1551
1552 case IPV6_FAITH:
1553 case IPV6_V6ONLY:
1554 case IPV6_PORTRANGE:
1555 switch (optname) {
1556
1557 case IPV6_UNICAST_HOPS:
1558 optval = in6p->in6p_hops;
1559 break;
1560
1561 case IPV6_CHECKSUM:
1562 optval = in6p->in6p_cksum;
1563 break;
1564
1565 case IPV6_FAITH:
1566 optval = OPTBIT(IN6P_FAITH);
1567 break;
1568
1569 case IPV6_V6ONLY:
1570 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1571 break;
1572
1573 case IPV6_PORTRANGE:
1574 {
1575 int flags;
1576 flags = in6p->in6p_flags;
1577 if (flags & IN6P_HIGHPORT)
1578 optval = IPV6_PORTRANGE_HIGH;
1579 else if (flags & IN6P_LOWPORT)
1580 optval = IPV6_PORTRANGE_LOW;
1581 else
1582 optval = 0;
1583 break;
1584 }
1585 }
1586 error = sooptcopyout(sopt, &optval,
1587 sizeof optval);
1588 break;
1589
1590 case IPV6_PKTINFO:
1591 case IPV6_HOPLIMIT:
1592 case IPV6_HOPOPTS:
1593 case IPV6_RTHDR:
1594 case IPV6_DSTOPTS:
1595 if (optname == IPV6_HOPOPTS ||
1596 optname == IPV6_DSTOPTS ||
1597 !privileged)
1598 return(EPERM);
1599 switch (optname) {
1600 case IPV6_PKTINFO:
1601 optval = OPTBIT(IN6P_PKTINFO);
1602 break;
1603 case IPV6_HOPLIMIT:
1604 optval = OPTBIT(IN6P_HOPLIMIT);
1605 break;
1606 case IPV6_HOPOPTS:
1607 if (!privileged)
1608 return(EPERM);
1609 optval = OPTBIT(IN6P_HOPOPTS);
1610 break;
1611 case IPV6_RTHDR:
1612 optval = OPTBIT(IN6P_RTHDR);
1613 break;
1614 case IPV6_DSTOPTS:
1615 if (!privileged)
1616 return(EPERM);
1617 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1618 break;
1619 }
1620 error = sooptcopyout(sopt, &optval,
1621 sizeof optval);
1622 break;
1623
1624 case IPV6_MULTICAST_IF:
1625 case IPV6_MULTICAST_HOPS:
1626 case IPV6_MULTICAST_LOOP:
1627 case IPV6_JOIN_GROUP:
1628 case IPV6_LEAVE_GROUP:
1629 {
1630 struct mbuf *m;
1631 error = ip6_getmoptions(sopt->sopt_name,
1632 in6p->in6p_moptions, &m);
1633 if (error == 0)
1634 error = sooptcopyout(sopt,
1635 mtod(m, char *), m->m_len);
1636 m_freem(m);
1637 }
1638 break;
1639
1640#ifdef IPSEC
1641 case IPV6_IPSEC_POLICY:
1642 {
1643 caddr_t req = NULL;
1644 size_t len = 0;
1645 struct mbuf *m = NULL;
1646 struct mbuf **mp = &m;
1647
1648 error = soopt_getm(sopt, &m); /* XXX */
1649 if (error != 0)
1650 break;
1651 error = soopt_mcopyin(sopt, m); /* XXX */
1652 if (error != 0)
1653 break;
1654 if (m) {
1655 req = mtod(m, caddr_t);
1656 len = m->m_len;
1657 }
1658 error = ipsec6_get_policy(in6p, req, len, mp);
1659 if (error == 0)
1660 error = soopt_mcopyout(sopt, m); /*XXX*/
1661 if (error == 0 && m)
1662 m_freem(m);
1663 break;
1664 }
1665#endif /* KAME IPSEC */
1666
1667 case IPV6_FW_GET:
1668 {
1669 struct mbuf *m;
1670 struct mbuf **mp = &m;
1671
1672 if (ip6_fw_ctl_ptr == NULL)
1673 {
1674 return EINVAL;
1675 }
1676 error = (*ip6_fw_ctl_ptr)(optname, mp);
1677 if (error == 0)
1678 error = soopt_mcopyout(sopt, m); /* XXX */
1679 if (error == 0 && m)
1680 m_freem(m);
1681 }
1682 break;
1683
1684 default:
1685 error = ENOPROTOOPT;
1686 break;
1687 }
1688 break;
1689 }
1690 } else {
1691 error = EINVAL;
1692 }
1693 return(error);
1694}
1695
1696/*
1697 * Set up IP6 options in pcb for insertion in output packets or
1698 * specifying behavior of outgoing packets.
1699 */
1700static int
1701ip6_pcbopts(pktopt, m, so, sopt)
1702 struct ip6_pktopts **pktopt;
1703 struct mbuf *m;
1704 struct socket *so;
1705 struct sockopt *sopt;
1706{
1707 struct ip6_pktopts *opt = *pktopt;
1708 int error = 0;
1709 struct thread *td = sopt->sopt_td;
1710 int priv = 0;
1711
1712 /* turn off any old options. */
1713 if (opt) {
1714#ifdef DIAGNOSTIC
1715 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1716 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1717 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1718 printf("ip6_pcbopts: all specified options are cleared.\n");
1719#endif
1720 ip6_clearpktopts(opt, 1, -1);
1721 } else
1722 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1723 *pktopt = NULL;
1724
1725 if (!m || m->m_len == 0) {
1726 /*
1727 * Only turning off any previous options.
1728 */
1729 if (opt)
1730 free(opt, M_IP6OPT);
1731 return(0);
1732 }
1733
1734 /* set options specified by user. */
1735 if (td && !suser(td))
1736 priv = 1;
1737 if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1738 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1739 return(error);
1740 }
1741 *pktopt = opt;
1742 return(0);
1743}
1744
1745/*
1746 * initialize ip6_pktopts. beware that there are non-zero default values in
1747 * the struct.
1748 */
1749void
1750init_ip6pktopts(opt)
1751 struct ip6_pktopts *opt;
1752{
1753
1754 bzero(opt, sizeof(*opt));
1755 opt->ip6po_hlim = -1; /* -1 means default hop limit */
1756}
1757
1758void
1759ip6_clearpktopts(pktopt, needfree, optname)
1760 struct ip6_pktopts *pktopt;
1761 int needfree, optname;
1762{
1763 if (pktopt == NULL)
1764 return;
1765
1766 if (optname == -1) {
1767 if (needfree && pktopt->ip6po_pktinfo)
1768 free(pktopt->ip6po_pktinfo, M_IP6OPT);
1769 pktopt->ip6po_pktinfo = NULL;
1770 }
1771 if (optname == -1)
1772 pktopt->ip6po_hlim = -1;
1773 if (optname == -1) {
1774 if (needfree && pktopt->ip6po_nexthop)
1775 free(pktopt->ip6po_nexthop, M_IP6OPT);
1776 pktopt->ip6po_nexthop = NULL;
1777 }
1778 if (optname == -1) {
1779 if (needfree && pktopt->ip6po_hbh)
1780 free(pktopt->ip6po_hbh, M_IP6OPT);
1781 pktopt->ip6po_hbh = NULL;
1782 }
1783 if (optname == -1) {
1784 if (needfree && pktopt->ip6po_dest1)
1785 free(pktopt->ip6po_dest1, M_IP6OPT);
1786 pktopt->ip6po_dest1 = NULL;
1787 }
1788 if (optname == -1) {
1789 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1790 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1791 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1792 if (pktopt->ip6po_route.ro_rt) {
1793 RTFREE(pktopt->ip6po_route.ro_rt);
1794 pktopt->ip6po_route.ro_rt = NULL;
1795 }
1796 }
1797 if (optname == -1) {
1798 if (needfree && pktopt->ip6po_dest2)
1799 free(pktopt->ip6po_dest2, M_IP6OPT);
1800 pktopt->ip6po_dest2 = NULL;
1801 }
1802}
1803
1804#define PKTOPT_EXTHDRCPY(type) \
1805do {\
1806 if (src->type) {\
1807 int hlen =\
1808 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1809 dst->type = malloc(hlen, M_IP6OPT, canwait);\
1810 if (dst->type == NULL && canwait == M_NOWAIT)\
1811 goto bad;\
1812 bcopy(src->type, dst->type, hlen);\
1813 }\
1814} while (0)
1815
1816struct ip6_pktopts *
1817ip6_copypktopts(src, canwait)
1818 struct ip6_pktopts *src;
1819 int canwait;
1820{
1821 struct ip6_pktopts *dst;
1822
1823 if (src == NULL) {
1824 printf("ip6_clearpktopts: invalid argument\n");
1825 return(NULL);
1826 }
1827
1828 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1829 if (dst == NULL && canwait == M_NOWAIT)
1830 goto bad;
1831 bzero(dst, sizeof(*dst));
1832
1833 dst->ip6po_hlim = src->ip6po_hlim;
1834 if (src->ip6po_pktinfo) {
1835 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1836 M_IP6OPT, canwait);
1837 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1838 goto bad;
1839 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1840 }
1841 if (src->ip6po_nexthop) {
1842 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1843 M_IP6OPT, canwait);
1844 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1845 goto bad;
1846 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1847 src->ip6po_nexthop->sa_len);
1848 }
1849 PKTOPT_EXTHDRCPY(ip6po_hbh);
1850 PKTOPT_EXTHDRCPY(ip6po_dest1);
1851 PKTOPT_EXTHDRCPY(ip6po_dest2);
1852 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1853 return(dst);
1854
1855 bad:
1856 printf("ip6_copypktopts: copy failed");
1857 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1858 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1859 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1860 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1861 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1862 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1863 return(NULL);
1864}
1865#undef PKTOPT_EXTHDRCPY
1866
1867void
1868ip6_freepcbopts(pktopt)
1869 struct ip6_pktopts *pktopt;
1870{
1871 if (pktopt == NULL)
1872 return;
1873
1874 ip6_clearpktopts(pktopt, 1, -1);
1875
1876 free(pktopt, M_IP6OPT);
1877}
1878
1879/*
1880 * Set the IP6 multicast options in response to user setsockopt().
1881 */
1882static int
1883ip6_setmoptions(optname, im6op, m)
1884 int optname;
1885 struct ip6_moptions **im6op;
1886 struct mbuf *m;
1887{
1888 int error = 0;
1889 u_int loop, ifindex;
1890 struct ipv6_mreq *mreq;
1891 struct ifnet *ifp;
1892 struct ip6_moptions *im6o = *im6op;
1893 struct route_in6 ro;
1894 struct sockaddr_in6 *dst;
1895 struct in6_multi_mship *imm;
1896 struct thread *td = curthread; /* XXX */
1897
1898 if (im6o == NULL) {
1899 /*
1900 * No multicast option buffer attached to the pcb;
1901 * allocate one and initialize to default values.
1902 */
1903 im6o = (struct ip6_moptions *)
1904 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1905
1906 if (im6o == NULL)
1907 return(ENOBUFS);
1908 *im6op = im6o;
1909 im6o->im6o_multicast_ifp = NULL;
1910 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1911 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1912 LIST_INIT(&im6o->im6o_memberships);
1913 }
1914
1915 switch (optname) {
1916
1917 case IPV6_MULTICAST_IF:
1918 /*
1919 * Select the interface for outgoing multicast packets.
1920 */
1921 if (m == NULL || m->m_len != sizeof(u_int)) {
1922 error = EINVAL;
1923 break;
1924 }
1925 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1926 if (ifindex < 0 || if_index < ifindex) {
1927 error = ENXIO; /* XXX EINVAL? */
1928 break;
1929 }
1930 ifp = ifnet_byindex(ifindex);
1931 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1932 error = EADDRNOTAVAIL;
1933 break;
1934 }
1935 im6o->im6o_multicast_ifp = ifp;
1936 break;
1937
1938 case IPV6_MULTICAST_HOPS:
1939 {
1940 /*
1941 * Set the IP6 hoplimit for outgoing multicast packets.
1942 */
1943 int optval;
1944 if (m == NULL || m->m_len != sizeof(int)) {
1945 error = EINVAL;
1946 break;
1947 }
1948 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1949 if (optval < -1 || optval >= 256)
1950 error = EINVAL;
1951 else if (optval == -1)
1952 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1953 else
1954 im6o->im6o_multicast_hlim = optval;
1955 break;
1956 }
1957
1958 case IPV6_MULTICAST_LOOP:
1959 /*
1960 * Set the loopback flag for outgoing multicast packets.
1961 * Must be zero or one.
1962 */
1963 if (m == NULL || m->m_len != sizeof(u_int)) {
1964 error = EINVAL;
1965 break;
1966 }
1967 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1968 if (loop > 1) {
1969 error = EINVAL;
1970 break;
1971 }
1972 im6o->im6o_multicast_loop = loop;
1973 break;
1974
1975 case IPV6_JOIN_GROUP:
1976 /*
1977 * Add a multicast group membership.
1978 * Group must be a valid IP6 multicast address.
1979 */
1980 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1981 error = EINVAL;
1982 break;
1983 }
1984 mreq = mtod(m, struct ipv6_mreq *);
1985 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1986 /*
1987 * We use the unspecified address to specify to accept
1988 * all multicast addresses. Only super user is allowed
1989 * to do this.
1990 */
1991 if (suser(td))
1992 {
1993 error = EACCES;
1994 break;
1995 }
1996 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1997 error = EINVAL;
1998 break;
1999 }
2000
2001 /*
2002 * If the interface is specified, validate it.
2003 */
2004 if (mreq->ipv6mr_interface < 0
2005 || if_index < mreq->ipv6mr_interface) {
2006 error = ENXIO; /* XXX EINVAL? */
2007 break;
2008 }
2009 /*
2010 * If no interface was explicitly specified, choose an
2011 * appropriate one according to the given multicast address.
2012 */
2013 if (mreq->ipv6mr_interface == 0) {
2014 /*
2015 * If the multicast address is in node-local scope,
2016 * the interface should be a loopback interface.
2017 * Otherwise, look up the routing table for the
2018 * address, and choose the outgoing interface.
2019 * XXX: is it a good approach?
2020 */
2021 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2022 ifp = &loif[0];
2023 } else {
2024 ro.ro_rt = NULL;
2025 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2026 bzero(dst, sizeof(*dst));
2027 dst->sin6_len = sizeof(struct sockaddr_in6);
2028 dst->sin6_family = AF_INET6;
2029 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2030 rtalloc((struct route *)&ro);
2031 if (ro.ro_rt == NULL) {
2032 error = EADDRNOTAVAIL;
2033 break;
2034 }
2035 ifp = ro.ro_rt->rt_ifp;
2036 rtfree(ro.ro_rt);
2037 }
2038 } else
2039 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2040
2041 /*
2042 * See if we found an interface, and confirm that it
2043 * supports multicast
2044 */
2045 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2046 error = EADDRNOTAVAIL;
2047 break;
2048 }
2049 /*
2050 * Put interface index into the multicast address,
2051 * if the address has link-local scope.
2052 */
2053 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2054 mreq->ipv6mr_multiaddr.s6_addr16[1]
2055 = htons(mreq->ipv6mr_interface);
2056 }
2057 /*
2058 * See if the membership already exists.
2059 */
2060 for (imm = im6o->im6o_memberships.lh_first;
2061 imm != NULL; imm = imm->i6mm_chain.le_next)
2062 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2063 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2064 &mreq->ipv6mr_multiaddr))
2065 break;
2066 if (imm != NULL) {
2067 error = EADDRINUSE;
2068 break;
2069 }
2070 /*
2071 * Everything looks good; add a new record to the multicast
2072 * address list for the given interface.
2073 */
2074 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2075 if (imm == NULL) {
2076 error = ENOBUFS;
2077 break;
2078 }
2079 if ((imm->i6mm_maddr =
2080 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2081 free(imm, M_IPMADDR);
2082 break;
2083 }
2084 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2085 break;
2086
2087 case IPV6_LEAVE_GROUP:
2088 /*
2089 * Drop a multicast group membership.
2090 * Group must be a valid IP6 multicast address.
2091 */
2092 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2093 error = EINVAL;
2094 break;
2095 }
2096 mreq = mtod(m, struct ipv6_mreq *);
2097 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2098 if (suser(td)) {
2099 error = EACCES;
2100 break;
2101 }
2102 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2103 error = EINVAL;
2104 break;
2105 }
2106 /*
2107 * If an interface address was specified, get a pointer
2108 * to its ifnet structure.
2109 */
2110 if (mreq->ipv6mr_interface < 0
2111 || if_index < mreq->ipv6mr_interface) {
2112 error = ENXIO; /* XXX EINVAL? */
2113 break;
2114 }
2115 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2116 /*
2117 * Put interface index into the multicast address,
2118 * if the address has link-local scope.
2119 */
2120 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2121 mreq->ipv6mr_multiaddr.s6_addr16[1]
2122 = htons(mreq->ipv6mr_interface);
2123 }
2124 /*
2125 * Find the membership in the membership list.
2126 */
2127 for (imm = im6o->im6o_memberships.lh_first;
2128 imm != NULL; imm = imm->i6mm_chain.le_next) {
2129 if ((ifp == NULL ||
2130 imm->i6mm_maddr->in6m_ifp == ifp) &&
2131 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2132 &mreq->ipv6mr_multiaddr))
2133 break;
2134 }
2135 if (imm == NULL) {
2136 /* Unable to resolve interface */
2137 error = EADDRNOTAVAIL;
2138 break;
2139 }
2140 /*
2141 * Give up the multicast address record to which the
2142 * membership points.
2143 */
2144 LIST_REMOVE(imm, i6mm_chain);
2145 in6_delmulti(imm->i6mm_maddr);
2146 free(imm, M_IPMADDR);
2147 break;
2148
2149 default:
2150 error = EOPNOTSUPP;
2151 break;
2152 }
2153
2154 /*
2155 * If all options have default values, no need to keep the mbuf.
2156 */
2157 if (im6o->im6o_multicast_ifp == NULL &&
2158 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2159 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2160 im6o->im6o_memberships.lh_first == NULL) {
2161 free(*im6op, M_IPMOPTS);
2162 *im6op = NULL;
2163 }
2164
2165 return(error);
2166}
2167
2168/*
2169 * Return the IP6 multicast options in response to user getsockopt().
2170 */
2171static int
2172ip6_getmoptions(optname, im6o, mp)
2173 int optname;
2174 struct ip6_moptions *im6o;
2175 struct mbuf **mp;
2176{
2177 u_int *hlim, *loop, *ifindex;
2178
2179 *mp = m_get(M_TRYWAIT, MT_HEADER); /* XXX */
2180
2181 switch (optname) {
2182
2183 case IPV6_MULTICAST_IF:
2184 ifindex = mtod(*mp, u_int *);
2185 (*mp)->m_len = sizeof(u_int);
2186 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2187 *ifindex = 0;
2188 else
2189 *ifindex = im6o->im6o_multicast_ifp->if_index;
2190 return(0);
2191
2192 case IPV6_MULTICAST_HOPS:
2193 hlim = mtod(*mp, u_int *);
2194 (*mp)->m_len = sizeof(u_int);
2195 if (im6o == NULL)
2196 *hlim = ip6_defmcasthlim;
2197 else
2198 *hlim = im6o->im6o_multicast_hlim;
2199 return(0);
2200
2201 case IPV6_MULTICAST_LOOP:
2202 loop = mtod(*mp, u_int *);
2203 (*mp)->m_len = sizeof(u_int);
2204 if (im6o == NULL)
2205 *loop = ip6_defmcasthlim;
2206 else
2207 *loop = im6o->im6o_multicast_loop;
2208 return(0);
2209
2210 default:
2211 return(EOPNOTSUPP);
2212 }
2213}
2214
2215/*
2216 * Discard the IP6 multicast options.
2217 */
2218void
2219ip6_freemoptions(im6o)
2220 struct ip6_moptions *im6o;
2221{
2222 struct in6_multi_mship *imm;
2223
2224 if (im6o == NULL)
2225 return;
2226
2227 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2228 LIST_REMOVE(imm, i6mm_chain);
2229 if (imm->i6mm_maddr)
2230 in6_delmulti(imm->i6mm_maddr);
2231 free(imm, M_IPMADDR);
2232 }
2233 free(im6o, M_IPMOPTS);
2234}
2235
2236/*
2237 * Set IPv6 outgoing packet options based on advanced API.
2238 */
2239int
2240ip6_setpktoptions(control, opt, priv, needcopy)
2241 struct mbuf *control;
2242 struct ip6_pktopts *opt;
2243 int priv, needcopy;
2244{
2245 struct cmsghdr *cm = 0;
2246
2247 if (control == 0 || opt == 0)
2248 return(EINVAL);
2249
2250 init_ip6pktopts(opt);
2251
2252 /*
2253 * XXX: Currently, we assume all the optional information is stored
2254 * in a single mbuf.
2255 */
2256 if (control->m_next)
2257 return(EINVAL);
2258
2259 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2260 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2261 cm = mtod(control, struct cmsghdr *);
2262 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2263 return(EINVAL);
2264 if (cm->cmsg_level != IPPROTO_IPV6)
2265 continue;
2266
2267 /*
2268 * XXX should check if RFC2292 API is mixed with 2292bis API
2269 */
2270 switch (cm->cmsg_type) {
2271 case IPV6_PKTINFO:
2272 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2273 return(EINVAL);
2274 if (needcopy) {
2275 /* XXX: Is it really WAITOK? */
2276 opt->ip6po_pktinfo =
2277 malloc(sizeof(struct in6_pktinfo),
2278 M_IP6OPT, M_WAITOK);
2279 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2280 sizeof(struct in6_pktinfo));
2281 } else
2282 opt->ip6po_pktinfo =
2283 (struct in6_pktinfo *)CMSG_DATA(cm);
2284 if (opt->ip6po_pktinfo->ipi6_ifindex &&
2285 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2286 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2287 htons(opt->ip6po_pktinfo->ipi6_ifindex);
2288
2289 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2290 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2291 return(ENXIO);
2292 }
2293
2294 /*
2295 * Check if the requested source address is indeed a
2296 * unicast address assigned to the node, and can be
2297 * used as the packet's source address.
2298 */
2299 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2300 struct in6_ifaddr *ia6;
2301 struct sockaddr_in6 sin6;
2302
2303 bzero(&sin6, sizeof(sin6));
2304 sin6.sin6_len = sizeof(sin6);
2305 sin6.sin6_family = AF_INET6;
2306 sin6.sin6_addr =
2307 opt->ip6po_pktinfo->ipi6_addr;
2308 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2309 if (ia6 == NULL ||
2310 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2311 IN6_IFF_NOTREADY)) != 0)
2312 return(EADDRNOTAVAIL);
2313 }
2314 break;
2315
2316 case IPV6_HOPLIMIT:
2317 if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2318 return(EINVAL);
2319
2320 opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2321 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2322 return(EINVAL);
2323 break;
2324
2325 case IPV6_NEXTHOP:
2326 if (!priv)
2327 return(EPERM);
2328
2329 if (cm->cmsg_len < sizeof(u_char) ||
2330 /* check if cmsg_len is large enough for sa_len */
2331 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2332 return(EINVAL);
2333
2334 if (needcopy) {
2335 opt->ip6po_nexthop =
2336 malloc(*CMSG_DATA(cm),
2337 M_IP6OPT, M_WAITOK);
2338 bcopy(CMSG_DATA(cm),
2339 opt->ip6po_nexthop,
2340 *CMSG_DATA(cm));
2341 } else
2342 opt->ip6po_nexthop =
2343 (struct sockaddr *)CMSG_DATA(cm);
2344 break;
2345
2346 case IPV6_HOPOPTS:
2347 {
2348 struct ip6_hbh *hbh;
2349 int hbhlen;
2350
2351 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2352 return(EINVAL);
2353 hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2354 hbhlen = (hbh->ip6h_len + 1) << 3;
2355 if (cm->cmsg_len != CMSG_LEN(hbhlen))
2356 return(EINVAL);
2357
2358 if (needcopy) {
2359 opt->ip6po_hbh =
2360 malloc(hbhlen, M_IP6OPT, M_WAITOK);
2361 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2362 } else
2363 opt->ip6po_hbh = hbh;
2364 break;
2365 }
2366
2367 case IPV6_DSTOPTS:
2368 {
2369 struct ip6_dest *dest, **newdest;
2370 int destlen;
2371
2372 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2373 return(EINVAL);
2374 dest = (struct ip6_dest *)CMSG_DATA(cm);
2375 destlen = (dest->ip6d_len + 1) << 3;
2376 if (cm->cmsg_len != CMSG_LEN(destlen))
2377 return(EINVAL);
2378
2379 /*
2380 * The old advacned API is ambiguous on this
2381 * point. Our approach is to determine the
2382 * position based according to the existence
2383 * of a routing header. Note, however, that
2384 * this depends on the order of the extension
2385 * headers in the ancillary data; the 1st part
2386 * of the destination options header must
2387 * appear before the routing header in the
2388 * ancillary data, too.
2389 * RFC2292bis solved the ambiguity by
2390 * introducing separate cmsg types.
2391 */
2392 if (opt->ip6po_rthdr == NULL)
2393 newdest = &opt->ip6po_dest1;
2394 else
2395 newdest = &opt->ip6po_dest2;
2396
2397 if (needcopy) {
2398 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2399 bcopy(dest, *newdest, destlen);
2400 } else
2401 *newdest = dest;
2402
2403 break;
2404 }
2405
2406 case IPV6_RTHDR:
2407 {
2408 struct ip6_rthdr *rth;
2409 int rthlen;
2410
2411 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2412 return(EINVAL);
2413 rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2414 rthlen = (rth->ip6r_len + 1) << 3;
2415 if (cm->cmsg_len != CMSG_LEN(rthlen))
2416 return(EINVAL);
2417
2418 switch (rth->ip6r_type) {
2419 case IPV6_RTHDR_TYPE_0:
2420 /* must contain one addr */
2421 if (rth->ip6r_len == 0)
2422 return(EINVAL);
2423 /* length must be even */
2424 if (rth->ip6r_len % 2)
2425 return(EINVAL);
2426 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2427 return(EINVAL);
2428 break;
2429 default:
2430 return(EINVAL); /* not supported */
2431 }
2432
2433 if (needcopy) {
2434 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2435 M_WAITOK);
2436 bcopy(rth, opt->ip6po_rthdr, rthlen);
2437 } else
2438 opt->ip6po_rthdr = rth;
2439
2440 break;
2441 }
2442
2443 default:
2444 return(ENOPROTOOPT);
2445 }
2446 }
2447
2448 return(0);
2449}
2450
2451/*
2452 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2453 * packet to the input queue of a specified interface. Note that this
2454 * calls the output routine of the loopback "driver", but with an interface
2455 * pointer that might NOT be &loif -- easier than replicating that code here.
2456 */
2457void
2458ip6_mloopback(ifp, m, dst)
2459 struct ifnet *ifp;
2460 struct mbuf *m;
2461 struct sockaddr_in6 *dst;
2462{
2463 struct mbuf *copym;
2464 struct ip6_hdr *ip6;
2465
2466 copym = m_copy(m, 0, M_COPYALL);
2467 if (copym == NULL)
2468 return;
2469
2470 /*
2471 * Make sure to deep-copy IPv6 header portion in case the data
2472 * is in an mbuf cluster, so that we can safely override the IPv6
2473 * header portion later.
2474 */
2475 if ((copym->m_flags & M_EXT) != 0 ||
2476 copym->m_len < sizeof(struct ip6_hdr)) {
2477 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2478 if (copym == NULL)
2479 return;
2480 }
2481
2482#ifdef DIAGNOSTIC
2483 if (copym->m_len < sizeof(*ip6)) {
2484 m_freem(copym);
2485 return;
2486 }
2487#endif
2488
2489 ip6 = mtod(copym, struct ip6_hdr *);
2490#ifndef SCOPEDROUTING
2491 /*
2492 * clear embedded scope identifiers if necessary.
2493 * in6_clearscope will touch the addresses only when necessary.
2494 */
2495 in6_clearscope(&ip6->ip6_src);
2496 in6_clearscope(&ip6->ip6_dst);
2497#endif
2498
2499 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
2500}
2501
2502/*
2503 * Chop IPv6 header off from the payload.
2504 */
2505static int
2506ip6_splithdr(m, exthdrs)
2507 struct mbuf *m;
2508 struct ip6_exthdrs *exthdrs;
2509{
2510 struct mbuf *mh;
2511 struct ip6_hdr *ip6;
2512
2513 ip6 = mtod(m, struct ip6_hdr *);
2514 if (m->m_len > sizeof(*ip6)) {
2515 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2516 if (mh == 0) {
2517 m_freem(m);
2518 return ENOBUFS;
2519 }
2520 M_COPY_PKTHDR(mh, m);
2521 MH_ALIGN(mh, sizeof(*ip6));
2522 m->m_flags &= ~M_PKTHDR;
2523 m->m_len -= sizeof(*ip6);
2524 m->m_data += sizeof(*ip6);
2525 mh->m_next = m;
2526 m = mh;
2527 m->m_len = sizeof(*ip6);
2528 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2529 }
2530 exthdrs->ip6e_ip6 = m;
2531 return 0;
2532}
2533
2534/*
2535 * Compute IPv6 extension header length.
2536 */
2537int
2538ip6_optlen(in6p)
2539 struct in6pcb *in6p;
2540{
2541 int len;
2542
2543 if (!in6p->in6p_outputopts)
2544 return 0;
2545
2546 len = 0;
2547#define elen(x) \
2548 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2549
2550 len += elen(in6p->in6p_outputopts->ip6po_hbh);
2551 if (in6p->in6p_outputopts->ip6po_rthdr)
2552 /* dest1 is valid with rthdr only */
2553 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2554 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2555 len += elen(in6p->in6p_outputopts->ip6po_dest2);
2556 return len;
2557#undef elen
2558}
180 ip6 = mtod(m, struct ip6_hdr *);
181#endif /* IPSEC */
182
183#define MAKE_EXTHDR(hp, mp) \
184 do { \
185 if (hp) { \
186 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
187 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
188 ((eh)->ip6e_len + 1) << 3); \
189 if (error) \
190 goto freehdrs; \
191 } \
192 } while (0)
193
194 bzero(&exthdrs, sizeof(exthdrs));
195
196 if (opt) {
197 /* Hop-by-Hop options header */
198 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
199 /* Destination options header(1st part) */
200 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
201 /* Routing header */
202 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
203 /* Destination options header(2nd part) */
204 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
205 }
206
207#ifdef IPSEC
208 /* get a security policy for this packet */
209 if (so == NULL)
210 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
211 else
212 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
213
214 if (sp == NULL) {
215 ipsec6stat.out_inval++;
216 goto freehdrs;
217 }
218
219 error = 0;
220
221 /* check policy */
222 switch (sp->policy) {
223 case IPSEC_POLICY_DISCARD:
224 /*
225 * This packet is just discarded.
226 */
227 ipsec6stat.out_polvio++;
228 goto freehdrs;
229
230 case IPSEC_POLICY_BYPASS:
231 case IPSEC_POLICY_NONE:
232 /* no need to do IPsec. */
233 needipsec = 0;
234 break;
235
236 case IPSEC_POLICY_IPSEC:
237 if (sp->req == NULL) {
238 /* acquire a policy */
239 error = key_spdacquire(sp);
240 goto freehdrs;
241 }
242 needipsec = 1;
243 break;
244
245 case IPSEC_POLICY_ENTRUST:
246 default:
247 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
248 }
249#endif /* IPSEC */
250
251 /*
252 * Calculate the total length of the extension header chain.
253 * Keep the length of the unfragmentable part for fragmentation.
254 */
255 optlen = 0;
256 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
257 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
258 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
259 unfragpartlen = optlen + sizeof(struct ip6_hdr);
260 /* NOTE: we don't add AH/ESP length here. do that later. */
261 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
262
263 /*
264 * If we need IPsec, or there is at least one extension header,
265 * separate IP6 header from the payload.
266 */
267 if ((needipsec || optlen) && !hdrsplit) {
268 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
269 m = NULL;
270 goto freehdrs;
271 }
272 m = exthdrs.ip6e_ip6;
273 hdrsplit++;
274 }
275
276 /* adjust pointer */
277 ip6 = mtod(m, struct ip6_hdr *);
278
279 /* adjust mbuf packet header length */
280 m->m_pkthdr.len += optlen;
281 plen = m->m_pkthdr.len - sizeof(*ip6);
282
283 /* If this is a jumbo payload, insert a jumbo payload option. */
284 if (plen > IPV6_MAXPACKET) {
285 if (!hdrsplit) {
286 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
287 m = NULL;
288 goto freehdrs;
289 }
290 m = exthdrs.ip6e_ip6;
291 hdrsplit++;
292 }
293 /* adjust pointer */
294 ip6 = mtod(m, struct ip6_hdr *);
295 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
296 goto freehdrs;
297 ip6->ip6_plen = 0;
298 } else
299 ip6->ip6_plen = htons(plen);
300
301 /*
302 * Concatenate headers and fill in next header fields.
303 * Here we have, on "m"
304 * IPv6 payload
305 * and we insert headers accordingly. Finally, we should be getting:
306 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
307 *
308 * during the header composing process, "m" points to IPv6 header.
309 * "mprev" points to an extension header prior to esp.
310 */
311 {
312 u_char *nexthdrp = &ip6->ip6_nxt;
313 struct mbuf *mprev = m;
314
315 /*
316 * we treat dest2 specially. this makes IPsec processing
317 * much easier. the goal here is to make mprev point the
318 * mbuf prior to dest2.
319 *
320 * result: IPv6 dest2 payload
321 * m and mprev will point to IPv6 header.
322 */
323 if (exthdrs.ip6e_dest2) {
324 if (!hdrsplit)
325 panic("assumption failed: hdr not split");
326 exthdrs.ip6e_dest2->m_next = m->m_next;
327 m->m_next = exthdrs.ip6e_dest2;
328 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
329 ip6->ip6_nxt = IPPROTO_DSTOPTS;
330 }
331
332#define MAKE_CHAIN(m, mp, p, i)\
333 do {\
334 if (m) {\
335 if (!hdrsplit) \
336 panic("assumption failed: hdr not split"); \
337 *mtod((m), u_char *) = *(p);\
338 *(p) = (i);\
339 p = mtod((m), u_char *);\
340 (m)->m_next = (mp)->m_next;\
341 (mp)->m_next = (m);\
342 (mp) = (m);\
343 }\
344 } while (0)
345 /*
346 * result: IPv6 hbh dest1 rthdr dest2 payload
347 * m will point to IPv6 header. mprev will point to the
348 * extension header prior to dest2 (rthdr in the above case).
349 */
350 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
351 nexthdrp, IPPROTO_HOPOPTS);
352 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
353 nexthdrp, IPPROTO_DSTOPTS);
354 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
355 nexthdrp, IPPROTO_ROUTING);
356
357#ifdef IPSEC
358 if (!needipsec)
359 goto skip_ipsec2;
360
361 /*
362 * pointers after IPsec headers are not valid any more.
363 * other pointers need a great care too.
364 * (IPsec routines should not mangle mbufs prior to AH/ESP)
365 */
366 exthdrs.ip6e_dest2 = NULL;
367
368 {
369 struct ip6_rthdr *rh = NULL;
370 int segleft_org = 0;
371 struct ipsec_output_state state;
372
373 if (exthdrs.ip6e_rthdr) {
374 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
375 segleft_org = rh->ip6r_segleft;
376 rh->ip6r_segleft = 0;
377 }
378
379 bzero(&state, sizeof(state));
380 state.m = m;
381 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
382 &needipsectun);
383 m = state.m;
384 if (error) {
385 /* mbuf is already reclaimed in ipsec6_output_trans. */
386 m = NULL;
387 switch (error) {
388 case EHOSTUNREACH:
389 case ENETUNREACH:
390 case EMSGSIZE:
391 case ENOBUFS:
392 case ENOMEM:
393 break;
394 default:
395 printf("ip6_output (ipsec): error code %d\n", error);
396 /* fall through */
397 case ENOENT:
398 /* don't show these error codes to the user */
399 error = 0;
400 break;
401 }
402 goto bad;
403 }
404 if (exthdrs.ip6e_rthdr) {
405 /* ah6_output doesn't modify mbuf chain */
406 rh->ip6r_segleft = segleft_org;
407 }
408 }
409skip_ipsec2:;
410#endif
411 }
412
413 /*
414 * If there is a routing header, replace destination address field
415 * with the first hop of the routing header.
416 */
417 if (exthdrs.ip6e_rthdr) {
418 struct ip6_rthdr *rh =
419 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
420 struct ip6_rthdr *));
421 struct ip6_rthdr0 *rh0;
422
423 finaldst = ip6->ip6_dst;
424 switch (rh->ip6r_type) {
425 case IPV6_RTHDR_TYPE_0:
426 rh0 = (struct ip6_rthdr0 *)rh;
427 ip6->ip6_dst = rh0->ip6r0_addr[0];
428 bcopy((caddr_t)&rh0->ip6r0_addr[1],
429 (caddr_t)&rh0->ip6r0_addr[0],
430 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
431 );
432 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
433 break;
434 default: /* is it possible? */
435 error = EINVAL;
436 goto bad;
437 }
438 }
439
440 /* Source address validation */
441 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
442 (flags & IPV6_DADOUTPUT) == 0) {
443 error = EOPNOTSUPP;
444 ip6stat.ip6s_badscope++;
445 goto bad;
446 }
447 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
448 error = EOPNOTSUPP;
449 ip6stat.ip6s_badscope++;
450 goto bad;
451 }
452
453 ip6stat.ip6s_localout++;
454
455 /*
456 * Route packet.
457 */
458 if (ro == 0) {
459 ro = &ip6route;
460 bzero((caddr_t)ro, sizeof(*ro));
461 }
462 ro_pmtu = ro;
463 if (opt && opt->ip6po_rthdr)
464 ro = &opt->ip6po_route;
465 dst = (struct sockaddr_in6 *)&ro->ro_dst;
466 /*
467 * If there is a cached route,
468 * check that it is to the same destination
469 * and is still up. If not, free it and try again.
470 */
471 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
472 dst->sin6_family != AF_INET6 ||
473 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
474 RTFREE(ro->ro_rt);
475 ro->ro_rt = (struct rtentry *)0;
476 }
477 if (ro->ro_rt == 0) {
478 bzero(dst, sizeof(*dst));
479 dst->sin6_family = AF_INET6;
480 dst->sin6_len = sizeof(struct sockaddr_in6);
481 dst->sin6_addr = ip6->ip6_dst;
482#ifdef SCOPEDROUTING
483 /* XXX: sin6_scope_id should already be fixed at this point */
484 if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
485 dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
486#endif
487 }
488#ifdef IPSEC
489 if (needipsec && needipsectun) {
490 struct ipsec_output_state state;
491
492 /*
493 * All the extension headers will become inaccessible
494 * (since they can be encrypted).
495 * Don't panic, we need no more updates to extension headers
496 * on inner IPv6 packet (since they are now encapsulated).
497 *
498 * IPv6 [ESP|AH] IPv6 [extension headers] payload
499 */
500 bzero(&exthdrs, sizeof(exthdrs));
501 exthdrs.ip6e_ip6 = m;
502
503 bzero(&state, sizeof(state));
504 state.m = m;
505 state.ro = (struct route *)ro;
506 state.dst = (struct sockaddr *)dst;
507
508 error = ipsec6_output_tunnel(&state, sp, flags);
509
510 m = state.m;
511 ro = (struct route_in6 *)state.ro;
512 dst = (struct sockaddr_in6 *)state.dst;
513 if (error) {
514 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
515 m0 = m = NULL;
516 m = NULL;
517 switch (error) {
518 case EHOSTUNREACH:
519 case ENETUNREACH:
520 case EMSGSIZE:
521 case ENOBUFS:
522 case ENOMEM:
523 break;
524 default:
525 printf("ip6_output (ipsec): error code %d\n", error);
526 /* fall through */
527 case ENOENT:
528 /* don't show these error codes to the user */
529 error = 0;
530 break;
531 }
532 goto bad;
533 }
534
535 exthdrs.ip6e_ip6 = m;
536 }
537#endif /* IPSEC */
538
539 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
540 /* Unicast */
541
542#define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
543#define sin6tosa(sin6) ((struct sockaddr *)(sin6))
544 /* xxx
545 * interface selection comes here
546 * if an interface is specified from an upper layer,
547 * ifp must point it.
548 */
549 if (ro->ro_rt == 0) {
550 /*
551 * non-bsdi always clone routes, if parent is
552 * PRF_CLONING.
553 */
554 rtalloc((struct route *)ro);
555 }
556 if (ro->ro_rt == 0) {
557 ip6stat.ip6s_noroute++;
558 error = EHOSTUNREACH;
559 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
560 goto bad;
561 }
562 ia = ifatoia6(ro->ro_rt->rt_ifa);
563 ifp = ro->ro_rt->rt_ifp;
564 ro->ro_rt->rt_use++;
565 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
566 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
567 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
568
569 in6_ifstat_inc(ifp, ifs6_out_request);
570
571 /*
572 * Check if the outgoing interface conflicts with
573 * the interface specified by ifi6_ifindex (if specified).
574 * Note that loopback interface is always okay.
575 * (this may happen when we are sending a packet to one of
576 * our own addresses.)
577 */
578 if (opt && opt->ip6po_pktinfo
579 && opt->ip6po_pktinfo->ipi6_ifindex) {
580 if (!(ifp->if_flags & IFF_LOOPBACK)
581 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
582 ip6stat.ip6s_noroute++;
583 in6_ifstat_inc(ifp, ifs6_out_discard);
584 error = EHOSTUNREACH;
585 goto bad;
586 }
587 }
588
589 if (opt && opt->ip6po_hlim != -1)
590 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
591 } else {
592 /* Multicast */
593 struct in6_multi *in6m;
594
595 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
596
597 /*
598 * See if the caller provided any multicast options
599 */
600 ifp = NULL;
601 if (im6o != NULL) {
602 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
603 if (im6o->im6o_multicast_ifp != NULL)
604 ifp = im6o->im6o_multicast_ifp;
605 } else
606 ip6->ip6_hlim = ip6_defmcasthlim;
607
608 /*
609 * See if the caller provided the outgoing interface
610 * as an ancillary data.
611 * Boundary check for ifindex is assumed to be already done.
612 */
613 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
614 ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
615
616 /*
617 * If the destination is a node-local scope multicast,
618 * the packet should be loop-backed only.
619 */
620 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
621 /*
622 * If the outgoing interface is already specified,
623 * it should be a loopback interface.
624 */
625 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
626 ip6stat.ip6s_badscope++;
627 error = ENETUNREACH; /* XXX: better error? */
628 /* XXX correct ifp? */
629 in6_ifstat_inc(ifp, ifs6_out_discard);
630 goto bad;
631 } else {
632 ifp = &loif[0];
633 }
634 }
635
636 if (opt && opt->ip6po_hlim != -1)
637 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
638
639 /*
640 * If caller did not provide an interface lookup a
641 * default in the routing table. This is either a
642 * default for the speicfied group (i.e. a host
643 * route), or a multicast default (a route for the
644 * ``net'' ff00::/8).
645 */
646 if (ifp == NULL) {
647 if (ro->ro_rt == 0) {
648 ro->ro_rt = rtalloc1((struct sockaddr *)
649 &ro->ro_dst, 0, 0UL);
650 }
651 if (ro->ro_rt == 0) {
652 ip6stat.ip6s_noroute++;
653 error = EHOSTUNREACH;
654 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
655 goto bad;
656 }
657 ia = ifatoia6(ro->ro_rt->rt_ifa);
658 ifp = ro->ro_rt->rt_ifp;
659 ro->ro_rt->rt_use++;
660 }
661
662 if ((flags & IPV6_FORWARDING) == 0)
663 in6_ifstat_inc(ifp, ifs6_out_request);
664 in6_ifstat_inc(ifp, ifs6_out_mcast);
665
666 /*
667 * Confirm that the outgoing interface supports multicast.
668 */
669 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
670 ip6stat.ip6s_noroute++;
671 in6_ifstat_inc(ifp, ifs6_out_discard);
672 error = ENETUNREACH;
673 goto bad;
674 }
675 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
676 if (in6m != NULL &&
677 (im6o == NULL || im6o->im6o_multicast_loop)) {
678 /*
679 * If we belong to the destination multicast group
680 * on the outgoing interface, and the caller did not
681 * forbid loopback, loop back a copy.
682 */
683 ip6_mloopback(ifp, m, dst);
684 } else {
685 /*
686 * If we are acting as a multicast router, perform
687 * multicast forwarding as if the packet had just
688 * arrived on the interface to which we are about
689 * to send. The multicast forwarding function
690 * recursively calls this function, using the
691 * IPV6_FORWARDING flag to prevent infinite recursion.
692 *
693 * Multicasts that are looped back by ip6_mloopback(),
694 * above, will be forwarded by the ip6_input() routine,
695 * if necessary.
696 */
697 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
698 if (ip6_mforward(ip6, ifp, m) != 0) {
699 m_freem(m);
700 goto done;
701 }
702 }
703 }
704 /*
705 * Multicasts with a hoplimit of zero may be looped back,
706 * above, but must not be transmitted on a network.
707 * Also, multicasts addressed to the loopback interface
708 * are not sent -- the above call to ip6_mloopback() will
709 * loop back a copy if this host actually belongs to the
710 * destination group on the loopback interface.
711 */
712 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
713 m_freem(m);
714 goto done;
715 }
716 }
717
718 /*
719 * Fill the outgoing inteface to tell the upper layer
720 * to increment per-interface statistics.
721 */
722 if (ifpp)
723 *ifpp = ifp;
724
725 /*
726 * Determine path MTU.
727 */
728 if (ro_pmtu != ro) {
729 /* The first hop and the final destination may differ. */
730 struct sockaddr_in6 *sin6_fin =
731 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
732 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
733 !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
734 &finaldst))) {
735 RTFREE(ro_pmtu->ro_rt);
736 ro_pmtu->ro_rt = (struct rtentry *)0;
737 }
738 if (ro_pmtu->ro_rt == 0) {
739 bzero(sin6_fin, sizeof(*sin6_fin));
740 sin6_fin->sin6_family = AF_INET6;
741 sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
742 sin6_fin->sin6_addr = finaldst;
743
744 rtalloc((struct route *)ro_pmtu);
745 }
746 }
747 if (ro_pmtu->ro_rt != NULL) {
748 u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
749
750 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
751 if (mtu > ifmtu || mtu == 0) {
752 /*
753 * The MTU on the route is larger than the MTU on
754 * the interface! This shouldn't happen, unless the
755 * MTU of the interface has been changed after the
756 * interface was brought up. Change the MTU in the
757 * route to match the interface MTU (as long as the
758 * field isn't locked).
759 *
760 * if MTU on the route is 0, we need to fix the MTU.
761 * this case happens with path MTU discovery timeouts.
762 */
763 mtu = ifmtu;
764 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
765 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
766 }
767 } else {
768 mtu = nd_ifinfo[ifp->if_index].linkmtu;
769 }
770
771 /*
772 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
773 */
774 if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
775 mtu = IPV6_MMTU;
776
777 /* Fake scoped addresses */
778 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
779 /*
780 * If source or destination address is a scoped address, and
781 * the packet is going to be sent to a loopback interface,
782 * we should keep the original interface.
783 */
784
785 /*
786 * XXX: this is a very experimental and temporary solution.
787 * We eventually have sockaddr_in6 and use the sin6_scope_id
788 * field of the structure here.
789 * We rely on the consistency between two scope zone ids
790 * of source and destination, which should already be assured.
791 * Larger scopes than link will be supported in the future.
792 */
793 origifp = NULL;
794 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
795 origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
796 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
797 origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
798 /*
799 * XXX: origifp can be NULL even in those two cases above.
800 * For example, if we remove the (only) link-local address
801 * from the loopback interface, and try to send a link-local
802 * address without link-id information. Then the source
803 * address is ::1, and the destination address is the
804 * link-local address with its s6_addr16[1] being zero.
805 * What is worse, if the packet goes to the loopback interface
806 * by a default rejected route, the null pointer would be
807 * passed to looutput, and the kernel would hang.
808 * The following last resort would prevent such disaster.
809 */
810 if (origifp == NULL)
811 origifp = ifp;
812 }
813 else
814 origifp = ifp;
815#ifndef SCOPEDROUTING
816 /*
817 * clear embedded scope identifiers if necessary.
818 * in6_clearscope will touch the addresses only when necessary.
819 */
820 in6_clearscope(&ip6->ip6_src);
821 in6_clearscope(&ip6->ip6_dst);
822#endif
823
824 /*
825 * Check with the firewall...
826 */
827 if (ip6_fw_enable && ip6_fw_chk_ptr) {
828 u_short port = 0;
829 m->m_pkthdr.rcvif = NULL; /* XXX */
830 /* If ipfw says divert, we have to just drop packet */
831 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
832 m_freem(m);
833 goto done;
834 }
835 if (!m) {
836 error = EACCES;
837 goto done;
838 }
839 }
840
841 /*
842 * If the outgoing packet contains a hop-by-hop options header,
843 * it must be examined and processed even by the source node.
844 * (RFC 2460, section 4.)
845 */
846 if (exthdrs.ip6e_hbh) {
847 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
848 u_int32_t dummy1; /* XXX unused */
849 u_int32_t dummy2; /* XXX unused */
850
851#ifdef DIAGNOSTIC
852 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
853 panic("ip6e_hbh is not continuous");
854#endif
855 /*
856 * XXX: if we have to send an ICMPv6 error to the sender,
857 * we need the M_LOOP flag since icmp6_error() expects
858 * the IPv6 and the hop-by-hop options header are
859 * continuous unless the flag is set.
860 */
861 m->m_flags |= M_LOOP;
862 m->m_pkthdr.rcvif = ifp;
863 if (ip6_process_hopopts(m,
864 (u_int8_t *)(hbh + 1),
865 ((hbh->ip6h_len + 1) << 3) -
866 sizeof(struct ip6_hbh),
867 &dummy1, &dummy2) < 0) {
868 /* m was already freed at this point */
869 error = EINVAL;/* better error? */
870 goto done;
871 }
872 m->m_flags &= ~M_LOOP; /* XXX */
873 m->m_pkthdr.rcvif = NULL;
874 }
875
876#ifdef PFIL_HOOKS
877 /*
878 * Run through list of hooks for output packets.
879 */
880 m1 = m;
881 pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
882 for (; pfh; pfh = pfh->pfil_link.tqe_next)
883 if (pfh->pfil_func) {
884 rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
885 if (rv) {
886 error = EHOSTUNREACH;
887 goto done;
888 }
889 m = m1;
890 if (m == NULL)
891 goto done;
892 ip6 = mtod(m, struct ip6_hdr *);
893 }
894#endif /* PFIL_HOOKS */
895 /*
896 * Send the packet to the outgoing interface.
897 * If necessary, do IPv6 fragmentation before sending.
898 */
899 tlen = m->m_pkthdr.len;
900 if (tlen <= mtu
901#ifdef notyet
902 /*
903 * On any link that cannot convey a 1280-octet packet in one piece,
904 * link-specific fragmentation and reassembly must be provided at
905 * a layer below IPv6. [RFC 2460, sec.5]
906 * Thus if the interface has ability of link-level fragmentation,
907 * we can just send the packet even if the packet size is
908 * larger than the link's MTU.
909 * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
910 */
911
912 || ifp->if_flags & IFF_FRAGMENTABLE
913#endif
914 )
915 {
916 /* Record statistics for this interface address. */
917 if (ia && !(flags & IPV6_FORWARDING)) {
918 ia->ia_ifa.if_opackets++;
919 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
920 }
921#ifdef IPSEC
922 /* clean ipsec history once it goes out of the node */
923 ipsec_delaux(m);
924#endif
925 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
926 goto done;
927 } else if (mtu < IPV6_MMTU) {
928 /*
929 * note that path MTU is never less than IPV6_MMTU
930 * (see icmp6_input).
931 */
932 error = EMSGSIZE;
933 in6_ifstat_inc(ifp, ifs6_out_fragfail);
934 goto bad;
935 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
936 error = EMSGSIZE;
937 in6_ifstat_inc(ifp, ifs6_out_fragfail);
938 goto bad;
939 } else {
940 struct mbuf **mnext, *m_frgpart;
941 struct ip6_frag *ip6f;
942 u_int32_t id = htonl(ip6_id++);
943 u_char nextproto;
944
945 /*
946 * Too large for the destination or interface;
947 * fragment if possible.
948 * Must be able to put at least 8 bytes per fragment.
949 */
950 hlen = unfragpartlen;
951 if (mtu > IPV6_MAXPACKET)
952 mtu = IPV6_MAXPACKET;
953
954 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
955 if (len < 8) {
956 error = EMSGSIZE;
957 in6_ifstat_inc(ifp, ifs6_out_fragfail);
958 goto bad;
959 }
960
961 mnext = &m->m_nextpkt;
962
963 /*
964 * Change the next header field of the last header in the
965 * unfragmentable part.
966 */
967 if (exthdrs.ip6e_rthdr) {
968 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
969 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
970 } else if (exthdrs.ip6e_dest1) {
971 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
972 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
973 } else if (exthdrs.ip6e_hbh) {
974 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
975 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
976 } else {
977 nextproto = ip6->ip6_nxt;
978 ip6->ip6_nxt = IPPROTO_FRAGMENT;
979 }
980
981 /*
982 * Loop through length of segment after first fragment,
983 * make new header and copy data of each part and link onto
984 * chain.
985 */
986 m0 = m;
987 for (off = hlen; off < tlen; off += len) {
988 MGETHDR(m, M_DONTWAIT, MT_HEADER);
989 if (!m) {
990 error = ENOBUFS;
991 ip6stat.ip6s_odropped++;
992 goto sendorfree;
993 }
994 m->m_pkthdr.rcvif = NULL;
995 m->m_flags = m0->m_flags & M_COPYFLAGS;
996 *mnext = m;
997 mnext = &m->m_nextpkt;
998 m->m_data += max_linkhdr;
999 mhip6 = mtod(m, struct ip6_hdr *);
1000 *mhip6 = *ip6;
1001 m->m_len = sizeof(*mhip6);
1002 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1003 if (error) {
1004 ip6stat.ip6s_odropped++;
1005 goto sendorfree;
1006 }
1007 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1008 if (off + len >= tlen)
1009 len = tlen - off;
1010 else
1011 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1012 mhip6->ip6_plen = htons((u_short)(len + hlen +
1013 sizeof(*ip6f) -
1014 sizeof(struct ip6_hdr)));
1015 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1016 error = ENOBUFS;
1017 ip6stat.ip6s_odropped++;
1018 goto sendorfree;
1019 }
1020 m_cat(m, m_frgpart);
1021 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1022 m->m_pkthdr.rcvif = (struct ifnet *)0;
1023 ip6f->ip6f_reserved = 0;
1024 ip6f->ip6f_ident = id;
1025 ip6f->ip6f_nxt = nextproto;
1026 ip6stat.ip6s_ofragments++;
1027 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1028 }
1029
1030 in6_ifstat_inc(ifp, ifs6_out_fragok);
1031 }
1032
1033 /*
1034 * Remove leading garbages.
1035 */
1036sendorfree:
1037 m = m0->m_nextpkt;
1038 m0->m_nextpkt = 0;
1039 m_freem(m0);
1040 for (m0 = m; m; m = m0) {
1041 m0 = m->m_nextpkt;
1042 m->m_nextpkt = 0;
1043 if (error == 0) {
1044 /* Record statistics for this interface address. */
1045 if (ia) {
1046 ia->ia_ifa.if_opackets++;
1047 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1048 }
1049#ifdef IPSEC
1050 /* clean ipsec history once it goes out of the node */
1051 ipsec_delaux(m);
1052#endif
1053 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1054 } else
1055 m_freem(m);
1056 }
1057
1058 if (error == 0)
1059 ip6stat.ip6s_fragmented++;
1060
1061done:
1062 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1063 RTFREE(ro->ro_rt);
1064 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1065 RTFREE(ro_pmtu->ro_rt);
1066 }
1067
1068#ifdef IPSEC
1069 if (sp != NULL)
1070 key_freesp(sp);
1071#endif /* IPSEC */
1072
1073 return(error);
1074
1075freehdrs:
1076 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1077 m_freem(exthdrs.ip6e_dest1);
1078 m_freem(exthdrs.ip6e_rthdr);
1079 m_freem(exthdrs.ip6e_dest2);
1080 /* fall through */
1081bad:
1082 m_freem(m);
1083 goto done;
1084}
1085
1086static int
1087ip6_copyexthdr(mp, hdr, hlen)
1088 struct mbuf **mp;
1089 caddr_t hdr;
1090 int hlen;
1091{
1092 struct mbuf *m;
1093
1094 if (hlen > MCLBYTES)
1095 return(ENOBUFS); /* XXX */
1096
1097 MGET(m, M_DONTWAIT, MT_DATA);
1098 if (!m)
1099 return(ENOBUFS);
1100
1101 if (hlen > MLEN) {
1102 MCLGET(m, M_DONTWAIT);
1103 if ((m->m_flags & M_EXT) == 0) {
1104 m_free(m);
1105 return(ENOBUFS);
1106 }
1107 }
1108 m->m_len = hlen;
1109 if (hdr)
1110 bcopy(hdr, mtod(m, caddr_t), hlen);
1111
1112 *mp = m;
1113 return(0);
1114}
1115
1116/*
1117 * Insert jumbo payload option.
1118 */
1119static int
1120ip6_insert_jumboopt(exthdrs, plen)
1121 struct ip6_exthdrs *exthdrs;
1122 u_int32_t plen;
1123{
1124 struct mbuf *mopt;
1125 u_char *optbuf;
1126 u_int32_t v;
1127
1128#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1129
1130 /*
1131 * If there is no hop-by-hop options header, allocate new one.
1132 * If there is one but it doesn't have enough space to store the
1133 * jumbo payload option, allocate a cluster to store the whole options.
1134 * Otherwise, use it to store the options.
1135 */
1136 if (exthdrs->ip6e_hbh == 0) {
1137 MGET(mopt, M_DONTWAIT, MT_DATA);
1138 if (mopt == 0)
1139 return(ENOBUFS);
1140 mopt->m_len = JUMBOOPTLEN;
1141 optbuf = mtod(mopt, u_char *);
1142 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1143 exthdrs->ip6e_hbh = mopt;
1144 } else {
1145 struct ip6_hbh *hbh;
1146
1147 mopt = exthdrs->ip6e_hbh;
1148 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1149 /*
1150 * XXX assumption:
1151 * - exthdrs->ip6e_hbh is not referenced from places
1152 * other than exthdrs.
1153 * - exthdrs->ip6e_hbh is not an mbuf chain.
1154 */
1155 int oldoptlen = mopt->m_len;
1156 struct mbuf *n;
1157
1158 /*
1159 * XXX: give up if the whole (new) hbh header does
1160 * not fit even in an mbuf cluster.
1161 */
1162 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1163 return(ENOBUFS);
1164
1165 /*
1166 * As a consequence, we must always prepare a cluster
1167 * at this point.
1168 */
1169 MGET(n, M_DONTWAIT, MT_DATA);
1170 if (n) {
1171 MCLGET(n, M_DONTWAIT);
1172 if ((n->m_flags & M_EXT) == 0) {
1173 m_freem(n);
1174 n = NULL;
1175 }
1176 }
1177 if (!n)
1178 return(ENOBUFS);
1179 n->m_len = oldoptlen + JUMBOOPTLEN;
1180 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1181 oldoptlen);
1182 optbuf = mtod(n, caddr_t) + oldoptlen;
1183 m_freem(mopt);
1184 mopt = exthdrs->ip6e_hbh = n;
1185 } else {
1186 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1187 mopt->m_len += JUMBOOPTLEN;
1188 }
1189 optbuf[0] = IP6OPT_PADN;
1190 optbuf[1] = 1;
1191
1192 /*
1193 * Adjust the header length according to the pad and
1194 * the jumbo payload option.
1195 */
1196 hbh = mtod(mopt, struct ip6_hbh *);
1197 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1198 }
1199
1200 /* fill in the option. */
1201 optbuf[2] = IP6OPT_JUMBO;
1202 optbuf[3] = 4;
1203 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1204 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1205
1206 /* finally, adjust the packet header length */
1207 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1208
1209 return(0);
1210#undef JUMBOOPTLEN
1211}
1212
1213/*
1214 * Insert fragment header and copy unfragmentable header portions.
1215 */
1216static int
1217ip6_insertfraghdr(m0, m, hlen, frghdrp)
1218 struct mbuf *m0, *m;
1219 int hlen;
1220 struct ip6_frag **frghdrp;
1221{
1222 struct mbuf *n, *mlast;
1223
1224 if (hlen > sizeof(struct ip6_hdr)) {
1225 n = m_copym(m0, sizeof(struct ip6_hdr),
1226 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1227 if (n == 0)
1228 return(ENOBUFS);
1229 m->m_next = n;
1230 } else
1231 n = m;
1232
1233 /* Search for the last mbuf of unfragmentable part. */
1234 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1235 ;
1236
1237 if ((mlast->m_flags & M_EXT) == 0 &&
1238 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1239 /* use the trailing space of the last mbuf for the fragment hdr */
1240 *frghdrp =
1241 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1242 mlast->m_len += sizeof(struct ip6_frag);
1243 m->m_pkthdr.len += sizeof(struct ip6_frag);
1244 } else {
1245 /* allocate a new mbuf for the fragment header */
1246 struct mbuf *mfrg;
1247
1248 MGET(mfrg, M_DONTWAIT, MT_DATA);
1249 if (mfrg == 0)
1250 return(ENOBUFS);
1251 mfrg->m_len = sizeof(struct ip6_frag);
1252 *frghdrp = mtod(mfrg, struct ip6_frag *);
1253 mlast->m_next = mfrg;
1254 }
1255
1256 return(0);
1257}
1258
1259/*
1260 * IP6 socket option processing.
1261 */
1262int
1263ip6_ctloutput(so, sopt)
1264 struct socket *so;
1265 struct sockopt *sopt;
1266{
1267 int privileged;
1268 struct inpcb *in6p = sotoinpcb(so);
1269 int error, optval;
1270 int level, op, optname;
1271 int optlen;
1272 struct thread *td;
1273
1274 if (sopt) {
1275 level = sopt->sopt_level;
1276 op = sopt->sopt_dir;
1277 optname = sopt->sopt_name;
1278 optlen = sopt->sopt_valsize;
1279 td = sopt->sopt_td;
1280 } else {
1281 panic("ip6_ctloutput: arg soopt is NULL");
1282 }
1283 error = optval = 0;
1284
1285 privileged = (td == 0 || suser(td)) ? 0 : 1;
1286
1287 if (level == IPPROTO_IPV6) {
1288 switch (op) {
1289
1290 case SOPT_SET:
1291 switch (optname) {
1292 case IPV6_PKTOPTIONS:
1293 {
1294 struct mbuf *m;
1295
1296 error = soopt_getm(sopt, &m); /* XXX */
1297 if (error != 0)
1298 break;
1299 error = soopt_mcopyin(sopt, m); /* XXX */
1300 if (error != 0)
1301 break;
1302 error = ip6_pcbopts(&in6p->in6p_outputopts,
1303 m, so, sopt);
1304 m_freem(m); /* XXX */
1305 break;
1306 }
1307
1308 /*
1309 * Use of some Hop-by-Hop options or some
1310 * Destination options, might require special
1311 * privilege. That is, normal applications
1312 * (without special privilege) might be forbidden
1313 * from setting certain options in outgoing packets,
1314 * and might never see certain options in received
1315 * packets. [RFC 2292 Section 6]
1316 * KAME specific note:
1317 * KAME prevents non-privileged users from sending or
1318 * receiving ANY hbh/dst options in order to avoid
1319 * overhead of parsing options in the kernel.
1320 */
1321 case IPV6_UNICAST_HOPS:
1322 case IPV6_CHECKSUM:
1323 case IPV6_FAITH:
1324
1325 case IPV6_V6ONLY:
1326 if (optlen != sizeof(int)) {
1327 error = EINVAL;
1328 break;
1329 }
1330 error = sooptcopyin(sopt, &optval,
1331 sizeof optval, sizeof optval);
1332 if (error)
1333 break;
1334 switch (optname) {
1335
1336 case IPV6_UNICAST_HOPS:
1337 if (optval < -1 || optval >= 256)
1338 error = EINVAL;
1339 else {
1340 /* -1 = kernel default */
1341 in6p->in6p_hops = optval;
1342
1343 if ((in6p->in6p_vflag &
1344 INP_IPV4) != 0)
1345 in6p->inp_ip_ttl = optval;
1346 }
1347 break;
1348#define OPTSET(bit) \
1349do { \
1350 if (optval) \
1351 in6p->in6p_flags |= (bit); \
1352 else \
1353 in6p->in6p_flags &= ~(bit); \
1354} while (0)
1355#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1356
1357 case IPV6_CHECKSUM:
1358 in6p->in6p_cksum = optval;
1359 break;
1360
1361 case IPV6_FAITH:
1362 OPTSET(IN6P_FAITH);
1363 break;
1364
1365 case IPV6_V6ONLY:
1366 /*
1367 * make setsockopt(IPV6_V6ONLY)
1368 * available only prior to bind(2).
1369 * see ipng mailing list, Jun 22 2001.
1370 */
1371 if (in6p->in6p_lport ||
1372 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1373 {
1374 error = EINVAL;
1375 break;
1376 }
1377 OPTSET(IN6P_IPV6_V6ONLY);
1378 if (optval)
1379 in6p->in6p_vflag &= ~INP_IPV4;
1380 else
1381 in6p->in6p_vflag |= INP_IPV4;
1382 break;
1383 }
1384 break;
1385
1386 case IPV6_PKTINFO:
1387 case IPV6_HOPLIMIT:
1388 case IPV6_HOPOPTS:
1389 case IPV6_DSTOPTS:
1390 case IPV6_RTHDR:
1391 /* RFC 2292 */
1392 if (optlen != sizeof(int)) {
1393 error = EINVAL;
1394 break;
1395 }
1396 error = sooptcopyin(sopt, &optval,
1397 sizeof optval, sizeof optval);
1398 if (error)
1399 break;
1400 switch (optname) {
1401 case IPV6_PKTINFO:
1402 OPTSET(IN6P_PKTINFO);
1403 break;
1404 case IPV6_HOPLIMIT:
1405 OPTSET(IN6P_HOPLIMIT);
1406 break;
1407 case IPV6_HOPOPTS:
1408 /*
1409 * Check super-user privilege.
1410 * See comments for IPV6_RECVHOPOPTS.
1411 */
1412 if (!privileged)
1413 return(EPERM);
1414 OPTSET(IN6P_HOPOPTS);
1415 break;
1416 case IPV6_DSTOPTS:
1417 if (!privileged)
1418 return(EPERM);
1419 OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1420 break;
1421 case IPV6_RTHDR:
1422 OPTSET(IN6P_RTHDR);
1423 break;
1424 }
1425 break;
1426#undef OPTSET
1427
1428 case IPV6_MULTICAST_IF:
1429 case IPV6_MULTICAST_HOPS:
1430 case IPV6_MULTICAST_LOOP:
1431 case IPV6_JOIN_GROUP:
1432 case IPV6_LEAVE_GROUP:
1433 {
1434 struct mbuf *m;
1435 if (sopt->sopt_valsize > MLEN) {
1436 error = EMSGSIZE;
1437 break;
1438 }
1439 /* XXX */
1440 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1441 if (m == 0) {
1442 error = ENOBUFS;
1443 break;
1444 }
1445 m->m_len = sopt->sopt_valsize;
1446 error = sooptcopyin(sopt, mtod(m, char *),
1447 m->m_len, m->m_len);
1448 error = ip6_setmoptions(sopt->sopt_name,
1449 &in6p->in6p_moptions,
1450 m);
1451 (void)m_free(m);
1452 }
1453 break;
1454
1455 case IPV6_PORTRANGE:
1456 error = sooptcopyin(sopt, &optval,
1457 sizeof optval, sizeof optval);
1458 if (error)
1459 break;
1460
1461 switch (optval) {
1462 case IPV6_PORTRANGE_DEFAULT:
1463 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1464 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1465 break;
1466
1467 case IPV6_PORTRANGE_HIGH:
1468 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1469 in6p->in6p_flags |= IN6P_HIGHPORT;
1470 break;
1471
1472 case IPV6_PORTRANGE_LOW:
1473 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1474 in6p->in6p_flags |= IN6P_LOWPORT;
1475 break;
1476
1477 default:
1478 error = EINVAL;
1479 break;
1480 }
1481 break;
1482
1483#ifdef IPSEC
1484 case IPV6_IPSEC_POLICY:
1485 {
1486 caddr_t req = NULL;
1487 size_t len = 0;
1488 struct mbuf *m;
1489
1490 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1491 break;
1492 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1493 break;
1494 if (m) {
1495 req = mtod(m, caddr_t);
1496 len = m->m_len;
1497 }
1498 error = ipsec6_set_policy(in6p, optname, req,
1499 len, privileged);
1500 m_freem(m);
1501 }
1502 break;
1503#endif /* KAME IPSEC */
1504
1505 case IPV6_FW_ADD:
1506 case IPV6_FW_DEL:
1507 case IPV6_FW_FLUSH:
1508 case IPV6_FW_ZERO:
1509 {
1510 struct mbuf *m;
1511 struct mbuf **mp = &m;
1512
1513 if (ip6_fw_ctl_ptr == NULL)
1514 return EINVAL;
1515 /* XXX */
1516 if ((error = soopt_getm(sopt, &m)) != 0)
1517 break;
1518 /* XXX */
1519 if ((error = soopt_mcopyin(sopt, m)) != 0)
1520 break;
1521 error = (*ip6_fw_ctl_ptr)(optname, mp);
1522 m = *mp;
1523 }
1524 break;
1525
1526 default:
1527 error = ENOPROTOOPT;
1528 break;
1529 }
1530 break;
1531
1532 case SOPT_GET:
1533 switch (optname) {
1534
1535 case IPV6_PKTOPTIONS:
1536 if (in6p->in6p_options) {
1537 struct mbuf *m;
1538 m = m_copym(in6p->in6p_options,
1539 0, M_COPYALL, M_TRYWAIT);
1540 error = soopt_mcopyout(sopt, m);
1541 if (error == 0)
1542 m_freem(m);
1543 } else
1544 sopt->sopt_valsize = 0;
1545 break;
1546
1547 case IPV6_UNICAST_HOPS:
1548 case IPV6_CHECKSUM:
1549
1550 case IPV6_FAITH:
1551 case IPV6_V6ONLY:
1552 case IPV6_PORTRANGE:
1553 switch (optname) {
1554
1555 case IPV6_UNICAST_HOPS:
1556 optval = in6p->in6p_hops;
1557 break;
1558
1559 case IPV6_CHECKSUM:
1560 optval = in6p->in6p_cksum;
1561 break;
1562
1563 case IPV6_FAITH:
1564 optval = OPTBIT(IN6P_FAITH);
1565 break;
1566
1567 case IPV6_V6ONLY:
1568 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1569 break;
1570
1571 case IPV6_PORTRANGE:
1572 {
1573 int flags;
1574 flags = in6p->in6p_flags;
1575 if (flags & IN6P_HIGHPORT)
1576 optval = IPV6_PORTRANGE_HIGH;
1577 else if (flags & IN6P_LOWPORT)
1578 optval = IPV6_PORTRANGE_LOW;
1579 else
1580 optval = 0;
1581 break;
1582 }
1583 }
1584 error = sooptcopyout(sopt, &optval,
1585 sizeof optval);
1586 break;
1587
1588 case IPV6_PKTINFO:
1589 case IPV6_HOPLIMIT:
1590 case IPV6_HOPOPTS:
1591 case IPV6_RTHDR:
1592 case IPV6_DSTOPTS:
1593 if (optname == IPV6_HOPOPTS ||
1594 optname == IPV6_DSTOPTS ||
1595 !privileged)
1596 return(EPERM);
1597 switch (optname) {
1598 case IPV6_PKTINFO:
1599 optval = OPTBIT(IN6P_PKTINFO);
1600 break;
1601 case IPV6_HOPLIMIT:
1602 optval = OPTBIT(IN6P_HOPLIMIT);
1603 break;
1604 case IPV6_HOPOPTS:
1605 if (!privileged)
1606 return(EPERM);
1607 optval = OPTBIT(IN6P_HOPOPTS);
1608 break;
1609 case IPV6_RTHDR:
1610 optval = OPTBIT(IN6P_RTHDR);
1611 break;
1612 case IPV6_DSTOPTS:
1613 if (!privileged)
1614 return(EPERM);
1615 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1616 break;
1617 }
1618 error = sooptcopyout(sopt, &optval,
1619 sizeof optval);
1620 break;
1621
1622 case IPV6_MULTICAST_IF:
1623 case IPV6_MULTICAST_HOPS:
1624 case IPV6_MULTICAST_LOOP:
1625 case IPV6_JOIN_GROUP:
1626 case IPV6_LEAVE_GROUP:
1627 {
1628 struct mbuf *m;
1629 error = ip6_getmoptions(sopt->sopt_name,
1630 in6p->in6p_moptions, &m);
1631 if (error == 0)
1632 error = sooptcopyout(sopt,
1633 mtod(m, char *), m->m_len);
1634 m_freem(m);
1635 }
1636 break;
1637
1638#ifdef IPSEC
1639 case IPV6_IPSEC_POLICY:
1640 {
1641 caddr_t req = NULL;
1642 size_t len = 0;
1643 struct mbuf *m = NULL;
1644 struct mbuf **mp = &m;
1645
1646 error = soopt_getm(sopt, &m); /* XXX */
1647 if (error != 0)
1648 break;
1649 error = soopt_mcopyin(sopt, m); /* XXX */
1650 if (error != 0)
1651 break;
1652 if (m) {
1653 req = mtod(m, caddr_t);
1654 len = m->m_len;
1655 }
1656 error = ipsec6_get_policy(in6p, req, len, mp);
1657 if (error == 0)
1658 error = soopt_mcopyout(sopt, m); /*XXX*/
1659 if (error == 0 && m)
1660 m_freem(m);
1661 break;
1662 }
1663#endif /* KAME IPSEC */
1664
1665 case IPV6_FW_GET:
1666 {
1667 struct mbuf *m;
1668 struct mbuf **mp = &m;
1669
1670 if (ip6_fw_ctl_ptr == NULL)
1671 {
1672 return EINVAL;
1673 }
1674 error = (*ip6_fw_ctl_ptr)(optname, mp);
1675 if (error == 0)
1676 error = soopt_mcopyout(sopt, m); /* XXX */
1677 if (error == 0 && m)
1678 m_freem(m);
1679 }
1680 break;
1681
1682 default:
1683 error = ENOPROTOOPT;
1684 break;
1685 }
1686 break;
1687 }
1688 } else {
1689 error = EINVAL;
1690 }
1691 return(error);
1692}
1693
1694/*
1695 * Set up IP6 options in pcb for insertion in output packets or
1696 * specifying behavior of outgoing packets.
1697 */
1698static int
1699ip6_pcbopts(pktopt, m, so, sopt)
1700 struct ip6_pktopts **pktopt;
1701 struct mbuf *m;
1702 struct socket *so;
1703 struct sockopt *sopt;
1704{
1705 struct ip6_pktopts *opt = *pktopt;
1706 int error = 0;
1707 struct thread *td = sopt->sopt_td;
1708 int priv = 0;
1709
1710 /* turn off any old options. */
1711 if (opt) {
1712#ifdef DIAGNOSTIC
1713 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1714 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1715 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1716 printf("ip6_pcbopts: all specified options are cleared.\n");
1717#endif
1718 ip6_clearpktopts(opt, 1, -1);
1719 } else
1720 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1721 *pktopt = NULL;
1722
1723 if (!m || m->m_len == 0) {
1724 /*
1725 * Only turning off any previous options.
1726 */
1727 if (opt)
1728 free(opt, M_IP6OPT);
1729 return(0);
1730 }
1731
1732 /* set options specified by user. */
1733 if (td && !suser(td))
1734 priv = 1;
1735 if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1736 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1737 return(error);
1738 }
1739 *pktopt = opt;
1740 return(0);
1741}
1742
1743/*
1744 * initialize ip6_pktopts. beware that there are non-zero default values in
1745 * the struct.
1746 */
1747void
1748init_ip6pktopts(opt)
1749 struct ip6_pktopts *opt;
1750{
1751
1752 bzero(opt, sizeof(*opt));
1753 opt->ip6po_hlim = -1; /* -1 means default hop limit */
1754}
1755
1756void
1757ip6_clearpktopts(pktopt, needfree, optname)
1758 struct ip6_pktopts *pktopt;
1759 int needfree, optname;
1760{
1761 if (pktopt == NULL)
1762 return;
1763
1764 if (optname == -1) {
1765 if (needfree && pktopt->ip6po_pktinfo)
1766 free(pktopt->ip6po_pktinfo, M_IP6OPT);
1767 pktopt->ip6po_pktinfo = NULL;
1768 }
1769 if (optname == -1)
1770 pktopt->ip6po_hlim = -1;
1771 if (optname == -1) {
1772 if (needfree && pktopt->ip6po_nexthop)
1773 free(pktopt->ip6po_nexthop, M_IP6OPT);
1774 pktopt->ip6po_nexthop = NULL;
1775 }
1776 if (optname == -1) {
1777 if (needfree && pktopt->ip6po_hbh)
1778 free(pktopt->ip6po_hbh, M_IP6OPT);
1779 pktopt->ip6po_hbh = NULL;
1780 }
1781 if (optname == -1) {
1782 if (needfree && pktopt->ip6po_dest1)
1783 free(pktopt->ip6po_dest1, M_IP6OPT);
1784 pktopt->ip6po_dest1 = NULL;
1785 }
1786 if (optname == -1) {
1787 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1788 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1789 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1790 if (pktopt->ip6po_route.ro_rt) {
1791 RTFREE(pktopt->ip6po_route.ro_rt);
1792 pktopt->ip6po_route.ro_rt = NULL;
1793 }
1794 }
1795 if (optname == -1) {
1796 if (needfree && pktopt->ip6po_dest2)
1797 free(pktopt->ip6po_dest2, M_IP6OPT);
1798 pktopt->ip6po_dest2 = NULL;
1799 }
1800}
1801
1802#define PKTOPT_EXTHDRCPY(type) \
1803do {\
1804 if (src->type) {\
1805 int hlen =\
1806 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1807 dst->type = malloc(hlen, M_IP6OPT, canwait);\
1808 if (dst->type == NULL && canwait == M_NOWAIT)\
1809 goto bad;\
1810 bcopy(src->type, dst->type, hlen);\
1811 }\
1812} while (0)
1813
1814struct ip6_pktopts *
1815ip6_copypktopts(src, canwait)
1816 struct ip6_pktopts *src;
1817 int canwait;
1818{
1819 struct ip6_pktopts *dst;
1820
1821 if (src == NULL) {
1822 printf("ip6_clearpktopts: invalid argument\n");
1823 return(NULL);
1824 }
1825
1826 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1827 if (dst == NULL && canwait == M_NOWAIT)
1828 goto bad;
1829 bzero(dst, sizeof(*dst));
1830
1831 dst->ip6po_hlim = src->ip6po_hlim;
1832 if (src->ip6po_pktinfo) {
1833 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1834 M_IP6OPT, canwait);
1835 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1836 goto bad;
1837 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1838 }
1839 if (src->ip6po_nexthop) {
1840 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1841 M_IP6OPT, canwait);
1842 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1843 goto bad;
1844 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1845 src->ip6po_nexthop->sa_len);
1846 }
1847 PKTOPT_EXTHDRCPY(ip6po_hbh);
1848 PKTOPT_EXTHDRCPY(ip6po_dest1);
1849 PKTOPT_EXTHDRCPY(ip6po_dest2);
1850 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1851 return(dst);
1852
1853 bad:
1854 printf("ip6_copypktopts: copy failed");
1855 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1856 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1857 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1858 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1859 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1860 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1861 return(NULL);
1862}
1863#undef PKTOPT_EXTHDRCPY
1864
1865void
1866ip6_freepcbopts(pktopt)
1867 struct ip6_pktopts *pktopt;
1868{
1869 if (pktopt == NULL)
1870 return;
1871
1872 ip6_clearpktopts(pktopt, 1, -1);
1873
1874 free(pktopt, M_IP6OPT);
1875}
1876
1877/*
1878 * Set the IP6 multicast options in response to user setsockopt().
1879 */
1880static int
1881ip6_setmoptions(optname, im6op, m)
1882 int optname;
1883 struct ip6_moptions **im6op;
1884 struct mbuf *m;
1885{
1886 int error = 0;
1887 u_int loop, ifindex;
1888 struct ipv6_mreq *mreq;
1889 struct ifnet *ifp;
1890 struct ip6_moptions *im6o = *im6op;
1891 struct route_in6 ro;
1892 struct sockaddr_in6 *dst;
1893 struct in6_multi_mship *imm;
1894 struct thread *td = curthread; /* XXX */
1895
1896 if (im6o == NULL) {
1897 /*
1898 * No multicast option buffer attached to the pcb;
1899 * allocate one and initialize to default values.
1900 */
1901 im6o = (struct ip6_moptions *)
1902 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1903
1904 if (im6o == NULL)
1905 return(ENOBUFS);
1906 *im6op = im6o;
1907 im6o->im6o_multicast_ifp = NULL;
1908 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1909 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1910 LIST_INIT(&im6o->im6o_memberships);
1911 }
1912
1913 switch (optname) {
1914
1915 case IPV6_MULTICAST_IF:
1916 /*
1917 * Select the interface for outgoing multicast packets.
1918 */
1919 if (m == NULL || m->m_len != sizeof(u_int)) {
1920 error = EINVAL;
1921 break;
1922 }
1923 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1924 if (ifindex < 0 || if_index < ifindex) {
1925 error = ENXIO; /* XXX EINVAL? */
1926 break;
1927 }
1928 ifp = ifnet_byindex(ifindex);
1929 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1930 error = EADDRNOTAVAIL;
1931 break;
1932 }
1933 im6o->im6o_multicast_ifp = ifp;
1934 break;
1935
1936 case IPV6_MULTICAST_HOPS:
1937 {
1938 /*
1939 * Set the IP6 hoplimit for outgoing multicast packets.
1940 */
1941 int optval;
1942 if (m == NULL || m->m_len != sizeof(int)) {
1943 error = EINVAL;
1944 break;
1945 }
1946 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1947 if (optval < -1 || optval >= 256)
1948 error = EINVAL;
1949 else if (optval == -1)
1950 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1951 else
1952 im6o->im6o_multicast_hlim = optval;
1953 break;
1954 }
1955
1956 case IPV6_MULTICAST_LOOP:
1957 /*
1958 * Set the loopback flag for outgoing multicast packets.
1959 * Must be zero or one.
1960 */
1961 if (m == NULL || m->m_len != sizeof(u_int)) {
1962 error = EINVAL;
1963 break;
1964 }
1965 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1966 if (loop > 1) {
1967 error = EINVAL;
1968 break;
1969 }
1970 im6o->im6o_multicast_loop = loop;
1971 break;
1972
1973 case IPV6_JOIN_GROUP:
1974 /*
1975 * Add a multicast group membership.
1976 * Group must be a valid IP6 multicast address.
1977 */
1978 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1979 error = EINVAL;
1980 break;
1981 }
1982 mreq = mtod(m, struct ipv6_mreq *);
1983 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1984 /*
1985 * We use the unspecified address to specify to accept
1986 * all multicast addresses. Only super user is allowed
1987 * to do this.
1988 */
1989 if (suser(td))
1990 {
1991 error = EACCES;
1992 break;
1993 }
1994 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1995 error = EINVAL;
1996 break;
1997 }
1998
1999 /*
2000 * If the interface is specified, validate it.
2001 */
2002 if (mreq->ipv6mr_interface < 0
2003 || if_index < mreq->ipv6mr_interface) {
2004 error = ENXIO; /* XXX EINVAL? */
2005 break;
2006 }
2007 /*
2008 * If no interface was explicitly specified, choose an
2009 * appropriate one according to the given multicast address.
2010 */
2011 if (mreq->ipv6mr_interface == 0) {
2012 /*
2013 * If the multicast address is in node-local scope,
2014 * the interface should be a loopback interface.
2015 * Otherwise, look up the routing table for the
2016 * address, and choose the outgoing interface.
2017 * XXX: is it a good approach?
2018 */
2019 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2020 ifp = &loif[0];
2021 } else {
2022 ro.ro_rt = NULL;
2023 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2024 bzero(dst, sizeof(*dst));
2025 dst->sin6_len = sizeof(struct sockaddr_in6);
2026 dst->sin6_family = AF_INET6;
2027 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2028 rtalloc((struct route *)&ro);
2029 if (ro.ro_rt == NULL) {
2030 error = EADDRNOTAVAIL;
2031 break;
2032 }
2033 ifp = ro.ro_rt->rt_ifp;
2034 rtfree(ro.ro_rt);
2035 }
2036 } else
2037 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2038
2039 /*
2040 * See if we found an interface, and confirm that it
2041 * supports multicast
2042 */
2043 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2044 error = EADDRNOTAVAIL;
2045 break;
2046 }
2047 /*
2048 * Put interface index into the multicast address,
2049 * if the address has link-local scope.
2050 */
2051 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2052 mreq->ipv6mr_multiaddr.s6_addr16[1]
2053 = htons(mreq->ipv6mr_interface);
2054 }
2055 /*
2056 * See if the membership already exists.
2057 */
2058 for (imm = im6o->im6o_memberships.lh_first;
2059 imm != NULL; imm = imm->i6mm_chain.le_next)
2060 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2061 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2062 &mreq->ipv6mr_multiaddr))
2063 break;
2064 if (imm != NULL) {
2065 error = EADDRINUSE;
2066 break;
2067 }
2068 /*
2069 * Everything looks good; add a new record to the multicast
2070 * address list for the given interface.
2071 */
2072 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2073 if (imm == NULL) {
2074 error = ENOBUFS;
2075 break;
2076 }
2077 if ((imm->i6mm_maddr =
2078 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2079 free(imm, M_IPMADDR);
2080 break;
2081 }
2082 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2083 break;
2084
2085 case IPV6_LEAVE_GROUP:
2086 /*
2087 * Drop a multicast group membership.
2088 * Group must be a valid IP6 multicast address.
2089 */
2090 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2091 error = EINVAL;
2092 break;
2093 }
2094 mreq = mtod(m, struct ipv6_mreq *);
2095 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2096 if (suser(td)) {
2097 error = EACCES;
2098 break;
2099 }
2100 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2101 error = EINVAL;
2102 break;
2103 }
2104 /*
2105 * If an interface address was specified, get a pointer
2106 * to its ifnet structure.
2107 */
2108 if (mreq->ipv6mr_interface < 0
2109 || if_index < mreq->ipv6mr_interface) {
2110 error = ENXIO; /* XXX EINVAL? */
2111 break;
2112 }
2113 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2114 /*
2115 * Put interface index into the multicast address,
2116 * if the address has link-local scope.
2117 */
2118 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2119 mreq->ipv6mr_multiaddr.s6_addr16[1]
2120 = htons(mreq->ipv6mr_interface);
2121 }
2122 /*
2123 * Find the membership in the membership list.
2124 */
2125 for (imm = im6o->im6o_memberships.lh_first;
2126 imm != NULL; imm = imm->i6mm_chain.le_next) {
2127 if ((ifp == NULL ||
2128 imm->i6mm_maddr->in6m_ifp == ifp) &&
2129 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2130 &mreq->ipv6mr_multiaddr))
2131 break;
2132 }
2133 if (imm == NULL) {
2134 /* Unable to resolve interface */
2135 error = EADDRNOTAVAIL;
2136 break;
2137 }
2138 /*
2139 * Give up the multicast address record to which the
2140 * membership points.
2141 */
2142 LIST_REMOVE(imm, i6mm_chain);
2143 in6_delmulti(imm->i6mm_maddr);
2144 free(imm, M_IPMADDR);
2145 break;
2146
2147 default:
2148 error = EOPNOTSUPP;
2149 break;
2150 }
2151
2152 /*
2153 * If all options have default values, no need to keep the mbuf.
2154 */
2155 if (im6o->im6o_multicast_ifp == NULL &&
2156 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2157 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2158 im6o->im6o_memberships.lh_first == NULL) {
2159 free(*im6op, M_IPMOPTS);
2160 *im6op = NULL;
2161 }
2162
2163 return(error);
2164}
2165
2166/*
2167 * Return the IP6 multicast options in response to user getsockopt().
2168 */
2169static int
2170ip6_getmoptions(optname, im6o, mp)
2171 int optname;
2172 struct ip6_moptions *im6o;
2173 struct mbuf **mp;
2174{
2175 u_int *hlim, *loop, *ifindex;
2176
2177 *mp = m_get(M_TRYWAIT, MT_HEADER); /* XXX */
2178
2179 switch (optname) {
2180
2181 case IPV6_MULTICAST_IF:
2182 ifindex = mtod(*mp, u_int *);
2183 (*mp)->m_len = sizeof(u_int);
2184 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2185 *ifindex = 0;
2186 else
2187 *ifindex = im6o->im6o_multicast_ifp->if_index;
2188 return(0);
2189
2190 case IPV6_MULTICAST_HOPS:
2191 hlim = mtod(*mp, u_int *);
2192 (*mp)->m_len = sizeof(u_int);
2193 if (im6o == NULL)
2194 *hlim = ip6_defmcasthlim;
2195 else
2196 *hlim = im6o->im6o_multicast_hlim;
2197 return(0);
2198
2199 case IPV6_MULTICAST_LOOP:
2200 loop = mtod(*mp, u_int *);
2201 (*mp)->m_len = sizeof(u_int);
2202 if (im6o == NULL)
2203 *loop = ip6_defmcasthlim;
2204 else
2205 *loop = im6o->im6o_multicast_loop;
2206 return(0);
2207
2208 default:
2209 return(EOPNOTSUPP);
2210 }
2211}
2212
2213/*
2214 * Discard the IP6 multicast options.
2215 */
2216void
2217ip6_freemoptions(im6o)
2218 struct ip6_moptions *im6o;
2219{
2220 struct in6_multi_mship *imm;
2221
2222 if (im6o == NULL)
2223 return;
2224
2225 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2226 LIST_REMOVE(imm, i6mm_chain);
2227 if (imm->i6mm_maddr)
2228 in6_delmulti(imm->i6mm_maddr);
2229 free(imm, M_IPMADDR);
2230 }
2231 free(im6o, M_IPMOPTS);
2232}
2233
2234/*
2235 * Set IPv6 outgoing packet options based on advanced API.
2236 */
2237int
2238ip6_setpktoptions(control, opt, priv, needcopy)
2239 struct mbuf *control;
2240 struct ip6_pktopts *opt;
2241 int priv, needcopy;
2242{
2243 struct cmsghdr *cm = 0;
2244
2245 if (control == 0 || opt == 0)
2246 return(EINVAL);
2247
2248 init_ip6pktopts(opt);
2249
2250 /*
2251 * XXX: Currently, we assume all the optional information is stored
2252 * in a single mbuf.
2253 */
2254 if (control->m_next)
2255 return(EINVAL);
2256
2257 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2258 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2259 cm = mtod(control, struct cmsghdr *);
2260 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2261 return(EINVAL);
2262 if (cm->cmsg_level != IPPROTO_IPV6)
2263 continue;
2264
2265 /*
2266 * XXX should check if RFC2292 API is mixed with 2292bis API
2267 */
2268 switch (cm->cmsg_type) {
2269 case IPV6_PKTINFO:
2270 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2271 return(EINVAL);
2272 if (needcopy) {
2273 /* XXX: Is it really WAITOK? */
2274 opt->ip6po_pktinfo =
2275 malloc(sizeof(struct in6_pktinfo),
2276 M_IP6OPT, M_WAITOK);
2277 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2278 sizeof(struct in6_pktinfo));
2279 } else
2280 opt->ip6po_pktinfo =
2281 (struct in6_pktinfo *)CMSG_DATA(cm);
2282 if (opt->ip6po_pktinfo->ipi6_ifindex &&
2283 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2284 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2285 htons(opt->ip6po_pktinfo->ipi6_ifindex);
2286
2287 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2288 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2289 return(ENXIO);
2290 }
2291
2292 /*
2293 * Check if the requested source address is indeed a
2294 * unicast address assigned to the node, and can be
2295 * used as the packet's source address.
2296 */
2297 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2298 struct in6_ifaddr *ia6;
2299 struct sockaddr_in6 sin6;
2300
2301 bzero(&sin6, sizeof(sin6));
2302 sin6.sin6_len = sizeof(sin6);
2303 sin6.sin6_family = AF_INET6;
2304 sin6.sin6_addr =
2305 opt->ip6po_pktinfo->ipi6_addr;
2306 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2307 if (ia6 == NULL ||
2308 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2309 IN6_IFF_NOTREADY)) != 0)
2310 return(EADDRNOTAVAIL);
2311 }
2312 break;
2313
2314 case IPV6_HOPLIMIT:
2315 if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2316 return(EINVAL);
2317
2318 opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2319 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2320 return(EINVAL);
2321 break;
2322
2323 case IPV6_NEXTHOP:
2324 if (!priv)
2325 return(EPERM);
2326
2327 if (cm->cmsg_len < sizeof(u_char) ||
2328 /* check if cmsg_len is large enough for sa_len */
2329 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2330 return(EINVAL);
2331
2332 if (needcopy) {
2333 opt->ip6po_nexthop =
2334 malloc(*CMSG_DATA(cm),
2335 M_IP6OPT, M_WAITOK);
2336 bcopy(CMSG_DATA(cm),
2337 opt->ip6po_nexthop,
2338 *CMSG_DATA(cm));
2339 } else
2340 opt->ip6po_nexthop =
2341 (struct sockaddr *)CMSG_DATA(cm);
2342 break;
2343
2344 case IPV6_HOPOPTS:
2345 {
2346 struct ip6_hbh *hbh;
2347 int hbhlen;
2348
2349 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2350 return(EINVAL);
2351 hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2352 hbhlen = (hbh->ip6h_len + 1) << 3;
2353 if (cm->cmsg_len != CMSG_LEN(hbhlen))
2354 return(EINVAL);
2355
2356 if (needcopy) {
2357 opt->ip6po_hbh =
2358 malloc(hbhlen, M_IP6OPT, M_WAITOK);
2359 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2360 } else
2361 opt->ip6po_hbh = hbh;
2362 break;
2363 }
2364
2365 case IPV6_DSTOPTS:
2366 {
2367 struct ip6_dest *dest, **newdest;
2368 int destlen;
2369
2370 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2371 return(EINVAL);
2372 dest = (struct ip6_dest *)CMSG_DATA(cm);
2373 destlen = (dest->ip6d_len + 1) << 3;
2374 if (cm->cmsg_len != CMSG_LEN(destlen))
2375 return(EINVAL);
2376
2377 /*
2378 * The old advacned API is ambiguous on this
2379 * point. Our approach is to determine the
2380 * position based according to the existence
2381 * of a routing header. Note, however, that
2382 * this depends on the order of the extension
2383 * headers in the ancillary data; the 1st part
2384 * of the destination options header must
2385 * appear before the routing header in the
2386 * ancillary data, too.
2387 * RFC2292bis solved the ambiguity by
2388 * introducing separate cmsg types.
2389 */
2390 if (opt->ip6po_rthdr == NULL)
2391 newdest = &opt->ip6po_dest1;
2392 else
2393 newdest = &opt->ip6po_dest2;
2394
2395 if (needcopy) {
2396 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2397 bcopy(dest, *newdest, destlen);
2398 } else
2399 *newdest = dest;
2400
2401 break;
2402 }
2403
2404 case IPV6_RTHDR:
2405 {
2406 struct ip6_rthdr *rth;
2407 int rthlen;
2408
2409 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2410 return(EINVAL);
2411 rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2412 rthlen = (rth->ip6r_len + 1) << 3;
2413 if (cm->cmsg_len != CMSG_LEN(rthlen))
2414 return(EINVAL);
2415
2416 switch (rth->ip6r_type) {
2417 case IPV6_RTHDR_TYPE_0:
2418 /* must contain one addr */
2419 if (rth->ip6r_len == 0)
2420 return(EINVAL);
2421 /* length must be even */
2422 if (rth->ip6r_len % 2)
2423 return(EINVAL);
2424 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2425 return(EINVAL);
2426 break;
2427 default:
2428 return(EINVAL); /* not supported */
2429 }
2430
2431 if (needcopy) {
2432 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2433 M_WAITOK);
2434 bcopy(rth, opt->ip6po_rthdr, rthlen);
2435 } else
2436 opt->ip6po_rthdr = rth;
2437
2438 break;
2439 }
2440
2441 default:
2442 return(ENOPROTOOPT);
2443 }
2444 }
2445
2446 return(0);
2447}
2448
2449/*
2450 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2451 * packet to the input queue of a specified interface. Note that this
2452 * calls the output routine of the loopback "driver", but with an interface
2453 * pointer that might NOT be &loif -- easier than replicating that code here.
2454 */
2455void
2456ip6_mloopback(ifp, m, dst)
2457 struct ifnet *ifp;
2458 struct mbuf *m;
2459 struct sockaddr_in6 *dst;
2460{
2461 struct mbuf *copym;
2462 struct ip6_hdr *ip6;
2463
2464 copym = m_copy(m, 0, M_COPYALL);
2465 if (copym == NULL)
2466 return;
2467
2468 /*
2469 * Make sure to deep-copy IPv6 header portion in case the data
2470 * is in an mbuf cluster, so that we can safely override the IPv6
2471 * header portion later.
2472 */
2473 if ((copym->m_flags & M_EXT) != 0 ||
2474 copym->m_len < sizeof(struct ip6_hdr)) {
2475 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2476 if (copym == NULL)
2477 return;
2478 }
2479
2480#ifdef DIAGNOSTIC
2481 if (copym->m_len < sizeof(*ip6)) {
2482 m_freem(copym);
2483 return;
2484 }
2485#endif
2486
2487 ip6 = mtod(copym, struct ip6_hdr *);
2488#ifndef SCOPEDROUTING
2489 /*
2490 * clear embedded scope identifiers if necessary.
2491 * in6_clearscope will touch the addresses only when necessary.
2492 */
2493 in6_clearscope(&ip6->ip6_src);
2494 in6_clearscope(&ip6->ip6_dst);
2495#endif
2496
2497 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
2498}
2499
2500/*
2501 * Chop IPv6 header off from the payload.
2502 */
2503static int
2504ip6_splithdr(m, exthdrs)
2505 struct mbuf *m;
2506 struct ip6_exthdrs *exthdrs;
2507{
2508 struct mbuf *mh;
2509 struct ip6_hdr *ip6;
2510
2511 ip6 = mtod(m, struct ip6_hdr *);
2512 if (m->m_len > sizeof(*ip6)) {
2513 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2514 if (mh == 0) {
2515 m_freem(m);
2516 return ENOBUFS;
2517 }
2518 M_COPY_PKTHDR(mh, m);
2519 MH_ALIGN(mh, sizeof(*ip6));
2520 m->m_flags &= ~M_PKTHDR;
2521 m->m_len -= sizeof(*ip6);
2522 m->m_data += sizeof(*ip6);
2523 mh->m_next = m;
2524 m = mh;
2525 m->m_len = sizeof(*ip6);
2526 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2527 }
2528 exthdrs->ip6e_ip6 = m;
2529 return 0;
2530}
2531
2532/*
2533 * Compute IPv6 extension header length.
2534 */
2535int
2536ip6_optlen(in6p)
2537 struct in6pcb *in6p;
2538{
2539 int len;
2540
2541 if (!in6p->in6p_outputopts)
2542 return 0;
2543
2544 len = 0;
2545#define elen(x) \
2546 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2547
2548 len += elen(in6p->in6p_outputopts->ip6po_hbh);
2549 if (in6p->in6p_outputopts->ip6po_rthdr)
2550 /* dest1 is valid with rthdr only */
2551 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2552 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2553 len += elen(in6p->in6p_outputopts->ip6po_dest2);
2554 return len;
2555#undef elen
2556}