proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[dragonfly.git] / sys / netinet6 / in6.c
CommitLineData
984263bc 1/* $FreeBSD: src/sys/netinet6/in6.c,v 1.7.2.9 2002/04/28 05:40:26 suz Exp $ */
dadab5e9 2/* $DragonFly: src/sys/netinet6/in6.c,v 1.4 2003/06/25 03:56:04 dillon Exp $ */
984263bc
MD
3/* $KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $ */
4
5/*
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34/*
35 * Copyright (c) 1982, 1986, 1991, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)in.c 8.2 (Berkeley) 11/15/93
67 */
68
69#include "opt_inet.h"
70#include "opt_inet6.h"
71
72#include <sys/param.h>
73#include <sys/errno.h>
74#include <sys/malloc.h>
75#include <sys/socket.h>
76#include <sys/socketvar.h>
77#include <sys/sockio.h>
78#include <sys/systm.h>
79#include <sys/proc.h>
80#include <sys/time.h>
81#include <sys/kernel.h>
82#include <sys/syslog.h>
83
84#include <net/if.h>
85#include <net/if_types.h>
86#include <net/route.h>
87#include <net/if_dl.h>
88
89#include <netinet/in.h>
90#include <netinet/in_var.h>
91#include <netinet/if_ether.h>
92#ifndef SCOPEDROUTING
93#include <netinet/in_systm.h>
94#include <netinet/ip.h>
95#include <netinet/in_pcb.h>
96#endif
97
98#include <netinet/ip6.h>
99#include <netinet6/ip6_var.h>
100#include <netinet6/nd6.h>
101#include <netinet6/mld6_var.h>
102#include <netinet6/ip6_mroute.h>
103#include <netinet6/in6_ifattach.h>
104#include <netinet6/scope6_var.h>
105#ifndef SCOPEDROUTING
106#include <netinet6/in6_pcb.h>
107#endif
108
109#include <net/net_osdep.h>
110
111MALLOC_DEFINE(M_IPMADDR, "in6_multi", "internet multicast address");
112
113/*
114 * Definitions of some costant IP6 addresses.
115 */
116const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
117const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
118const struct in6_addr in6addr_nodelocal_allnodes =
119 IN6ADDR_NODELOCAL_ALLNODES_INIT;
120const struct in6_addr in6addr_linklocal_allnodes =
121 IN6ADDR_LINKLOCAL_ALLNODES_INIT;
122const struct in6_addr in6addr_linklocal_allrouters =
123 IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
124
125const struct in6_addr in6mask0 = IN6MASK0;
126const struct in6_addr in6mask32 = IN6MASK32;
127const struct in6_addr in6mask64 = IN6MASK64;
128const struct in6_addr in6mask96 = IN6MASK96;
129const struct in6_addr in6mask128 = IN6MASK128;
130
131const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
132 0, 0, IN6ADDR_ANY_INIT, 0};
133
134static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
dadab5e9 135 struct ifnet *, struct thread *));
984263bc
MD
136static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
137 struct sockaddr_in6 *, int));
138static void in6_unlink_ifa __P((struct in6_ifaddr *, struct ifnet *));
139
140struct in6_multihead in6_multihead; /* XXX BSS initialization */
141
142int (*faithprefix_p)(struct in6_addr *);
143
144/*
145 * Subroutine for in6_ifaddloop() and in6_ifremloop().
146 * This routine does actual work.
147 */
148static void
149in6_ifloop_request(int cmd, struct ifaddr *ifa)
150{
151 struct sockaddr_in6 all1_sa;
152 struct rtentry *nrt = NULL;
153 int e;
154
155 bzero(&all1_sa, sizeof(all1_sa));
156 all1_sa.sin6_family = AF_INET6;
157 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
158 all1_sa.sin6_addr = in6mask128;
159
160 /*
161 * We specify the address itself as the gateway, and set the
162 * RTF_LLINFO flag, so that the corresponding host route would have
163 * the flag, and thus applications that assume traditional behavior
164 * would be happy. Note that we assume the caller of the function
165 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
166 * which changes the outgoing interface to the loopback interface.
167 */
168 e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
169 (struct sockaddr *)&all1_sa,
170 RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
171 if (e != 0) {
172 log(LOG_ERR, "in6_ifloop_request: "
173 "%s operation failed for %s (errno=%d)\n",
174 cmd == RTM_ADD ? "ADD" : "DELETE",
175 ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
176 e);
177 }
178
179 /*
180 * Make sure rt_ifa be equal to IFA, the second argument of the
181 * function.
182 * We need this because when we refer to rt_ifa->ia6_flags in
183 * ip6_input, we assume that the rt_ifa points to the address instead
184 * of the loopback address.
185 */
186 if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa) {
187 IFAFREE(nrt->rt_ifa);
188 IFAREF(ifa);
189 nrt->rt_ifa = ifa;
190 }
191
192 /*
193 * Report the addition/removal of the address to the routing socket.
194 * XXX: since we called rtinit for a p2p interface with a destination,
195 * we end up reporting twice in such a case. Should we rather
196 * omit the second report?
197 */
198 if (nrt) {
199 rt_newaddrmsg(cmd, ifa, e, nrt);
200 if (cmd == RTM_DELETE) {
201 if (nrt->rt_refcnt <= 0) {
202 /* XXX: we should free the entry ourselves. */
203 nrt->rt_refcnt++;
204 rtfree(nrt);
205 }
206 } else {
207 /* the cmd must be RTM_ADD here */
208 nrt->rt_refcnt--;
209 }
210 }
211}
212
213/*
214 * Add ownaddr as loopback rtentry. We previously add the route only if
215 * necessary (ex. on a p2p link). However, since we now manage addresses
216 * separately from prefixes, we should always add the route. We can't
217 * rely on the cloning mechanism from the corresponding interface route
218 * any more.
219 */
220static void
221in6_ifaddloop(struct ifaddr *ifa)
222{
223 struct rtentry *rt;
224
225 /* If there is no loopback entry, allocate one. */
226 rt = rtalloc1(ifa->ifa_addr, 0, 0);
227 if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
228 (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
229 in6_ifloop_request(RTM_ADD, ifa);
230 if (rt)
231 rt->rt_refcnt--;
232}
233
234/*
235 * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
236 * if it exists.
237 */
238static void
239in6_ifremloop(struct ifaddr *ifa)
240{
241 struct in6_ifaddr *ia;
242 struct rtentry *rt;
243 int ia_count = 0;
244
245 /*
246 * Some of BSD variants do not remove cloned routes
247 * from an interface direct route, when removing the direct route
248 * (see comments in net/net_osdep.h). Even for variants that do remove
249 * cloned routes, they could fail to remove the cloned routes when
250 * we handle multple addresses that share a common prefix.
251 * So, we should remove the route corresponding to the deleted address
252 * regardless of the result of in6_is_ifloop_auto().
253 */
254
255 /*
256 * Delete the entry only if exact one ifa exists. More than one ifa
257 * can exist if we assign a same single address to multiple
258 * (probably p2p) interfaces.
259 * XXX: we should avoid such a configuration in IPv6...
260 */
261 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
262 if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
263 ia_count++;
264 if (ia_count > 1)
265 break;
266 }
267 }
268
269 if (ia_count == 1) {
270 /*
271 * Before deleting, check if a corresponding loopbacked host
272 * route surely exists. With this check, we can avoid to
273 * delete an interface direct route whose destination is same
274 * as the address being removed. This can happen when remofing
275 * a subnet-router anycast address on an interface attahced
276 * to a shared medium.
277 */
278 rt = rtalloc1(ifa->ifa_addr, 0, 0);
279 if (rt != NULL && (rt->rt_flags & RTF_HOST) != 0 &&
280 (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
281 rt->rt_refcnt--;
282 in6_ifloop_request(RTM_DELETE, ifa);
283 }
284 }
285}
286
287int
288in6_ifindex2scopeid(idx)
289 int idx;
290{
291 struct ifnet *ifp;
292 struct ifaddr *ifa;
293 struct sockaddr_in6 *sin6;
294
295 if (idx < 0 || if_index < idx)
296 return -1;
297 ifp = ifindex2ifnet[idx];
298
299 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
300 {
301 if (ifa->ifa_addr->sa_family != AF_INET6)
302 continue;
303 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
304 if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))
305 return sin6->sin6_scope_id & 0xffff;
306 }
307
308 return -1;
309}
310
311int
312in6_mask2len(mask, lim0)
313 struct in6_addr *mask;
314 u_char *lim0;
315{
316 int x = 0, y;
317 u_char *lim = lim0, *p;
318
319 if (lim0 == NULL ||
320 lim0 - (u_char *)mask > sizeof(*mask)) /* ignore the scope_id part */
321 lim = (u_char *)mask + sizeof(*mask);
322 for (p = (u_char *)mask; p < lim; x++, p++) {
323 if (*p != 0xff)
324 break;
325 }
326 y = 0;
327 if (p < lim) {
328 for (y = 0; y < 8; y++) {
329 if ((*p & (0x80 >> y)) == 0)
330 break;
331 }
332 }
333
334 /*
335 * when the limit pointer is given, do a stricter check on the
336 * remaining bits.
337 */
338 if (p < lim) {
339 if (y != 0 && (*p & (0x00ff >> y)) != 0)
340 return(-1);
341 for (p = p + 1; p < lim; p++)
342 if (*p != 0)
343 return(-1);
344 }
345
346 return x * 8 + y;
347}
348
349void
350in6_len2mask(mask, len)
351 struct in6_addr *mask;
352 int len;
353{
354 int i;
355
356 bzero(mask, sizeof(*mask));
357 for (i = 0; i < len / 8; i++)
358 mask->s6_addr8[i] = 0xff;
359 if (len % 8)
360 mask->s6_addr8[i] = (0xff00 >> (len % 8)) & 0xff;
361}
362
363#define ifa2ia6(ifa) ((struct in6_ifaddr *)(ifa))
364#define ia62ifa(ia6) (&((ia6)->ia_ifa))
365
366int
dadab5e9
MD
367in6_control(struct socket *so, u_long cmd, caddr_t data,
368 struct ifnet *ifp, struct thread *td)
984263bc
MD
369{
370 struct in6_ifreq *ifr = (struct in6_ifreq *)data;
371 struct in6_ifaddr *ia = NULL;
372 struct in6_aliasreq *ifra = (struct in6_aliasreq *)data;
373 int privileged;
374
375 privileged = 0;
dadab5e9 376 if (suser(td) == 0)
984263bc
MD
377 privileged++;
378
379 switch (cmd) {
380 case SIOCGETSGCNT_IN6:
381 case SIOCGETMIFCNT_IN6:
382 return (mrt6_ioctl(cmd, data));
383 }
384
385 if (ifp == NULL)
386 return(EOPNOTSUPP);
387
388 switch (cmd) {
389 case SIOCSNDFLUSH_IN6:
390 case SIOCSPFXFLUSH_IN6:
391 case SIOCSRTRFLUSH_IN6:
392 case SIOCSDEFIFACE_IN6:
393 case SIOCSIFINFO_FLAGS:
394 if (!privileged)
395 return(EPERM);
396 /* fall through */
397 case OSIOCGIFINFO_IN6:
398 case SIOCGIFINFO_IN6:
399 case SIOCGDRLST_IN6:
400 case SIOCGPRLST_IN6:
401 case SIOCGNBRINFO_IN6:
402 case SIOCGDEFIFACE_IN6:
403 return(nd6_ioctl(cmd, data, ifp));
404 }
405
406 switch (cmd) {
407 case SIOCSIFPREFIX_IN6:
408 case SIOCDIFPREFIX_IN6:
409 case SIOCAIFPREFIX_IN6:
410 case SIOCCIFPREFIX_IN6:
411 case SIOCSGIFPREFIX_IN6:
412 case SIOCGIFPREFIX_IN6:
413 log(LOG_NOTICE,
414 "prefix ioctls are now invalidated. "
415 "please use ifconfig.\n");
416 return(EOPNOTSUPP);
417 }
418
419 switch (cmd) {
420 case SIOCSSCOPE6:
421 if (!privileged)
422 return(EPERM);
423 return(scope6_set(ifp, ifr->ifr_ifru.ifru_scope_id));
424 break;
425 case SIOCGSCOPE6:
426 return(scope6_get(ifp, ifr->ifr_ifru.ifru_scope_id));
427 break;
428 case SIOCGSCOPE6DEF:
429 return(scope6_get_default(ifr->ifr_ifru.ifru_scope_id));
430 break;
431 }
432
433 switch (cmd) {
434 case SIOCALIFADDR:
435 case SIOCDLIFADDR:
436 if (!privileged)
437 return(EPERM);
438 /* fall through */
439 case SIOCGLIFADDR:
440 return in6_lifaddr_ioctl(so, cmd, data, ifp, p);
441 }
442
443 /*
444 * Find address for this interface, if it exists.
445 */
446 if (ifra->ifra_addr.sin6_family == AF_INET6) { /* XXX */
447 struct sockaddr_in6 *sa6 =
448 (struct sockaddr_in6 *)&ifra->ifra_addr;
449
450 if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
451 if (sa6->sin6_addr.s6_addr16[1] == 0) {
452 /* link ID is not embedded by the user */
453 sa6->sin6_addr.s6_addr16[1] =
454 htons(ifp->if_index);
455 } else if (sa6->sin6_addr.s6_addr16[1] !=
456 htons(ifp->if_index)) {
457 return(EINVAL); /* link ID contradicts */
458 }
459 if (sa6->sin6_scope_id) {
460 if (sa6->sin6_scope_id !=
461 (u_int32_t)ifp->if_index)
462 return(EINVAL);
463 sa6->sin6_scope_id = 0; /* XXX: good way? */
464 }
465 }
466 ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
467 }
468
469 switch (cmd) {
470 case SIOCSIFADDR_IN6:
471 case SIOCSIFDSTADDR_IN6:
472 case SIOCSIFNETMASK_IN6:
473 /*
474 * Since IPv6 allows a node to assign multiple addresses
475 * on a single interface, SIOCSIFxxx ioctls are not suitable
476 * and should be unused.
477 */
478 /* we decided to obsolete this command (20000704) */
479 return(EINVAL);
480
481 case SIOCDIFADDR_IN6:
482 /*
483 * for IPv4, we look for existing in_ifaddr here to allow
484 * "ifconfig if0 delete" to remove first IPv4 address on the
485 * interface. For IPv6, as the spec allow multiple interface
486 * address from the day one, we consider "remove the first one"
487 * semantics to be not preferable.
488 */
489 if (ia == NULL)
490 return(EADDRNOTAVAIL);
491 /* FALLTHROUGH */
492 case SIOCAIFADDR_IN6:
493 /*
494 * We always require users to specify a valid IPv6 address for
495 * the corresponding operation.
496 */
497 if (ifra->ifra_addr.sin6_family != AF_INET6 ||
498 ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
499 return(EAFNOSUPPORT);
500 if (!privileged)
501 return(EPERM);
502
503 break;
504
505 case SIOCGIFADDR_IN6:
506 /* This interface is basically deprecated. use SIOCGIFCONF. */
507 /* fall through */
508 case SIOCGIFAFLAG_IN6:
509 case SIOCGIFNETMASK_IN6:
510 case SIOCGIFDSTADDR_IN6:
511 case SIOCGIFALIFETIME_IN6:
512 /* must think again about its semantics */
513 if (ia == NULL)
514 return(EADDRNOTAVAIL);
515 break;
516 case SIOCSIFALIFETIME_IN6:
517 {
518 struct in6_addrlifetime *lt;
519
520 if (!privileged)
521 return(EPERM);
522 if (ia == NULL)
523 return(EADDRNOTAVAIL);
524 /* sanity for overflow - beware unsigned */
525 lt = &ifr->ifr_ifru.ifru_lifetime;
526 if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
527 && lt->ia6t_vltime + time_second < time_second) {
528 return EINVAL;
529 }
530 if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
531 && lt->ia6t_pltime + time_second < time_second) {
532 return EINVAL;
533 }
534 break;
535 }
536 }
537
538 switch (cmd) {
539
540 case SIOCGIFADDR_IN6:
541 ifr->ifr_addr = ia->ia_addr;
542 break;
543
544 case SIOCGIFDSTADDR_IN6:
545 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
546 return(EINVAL);
547 /*
548 * XXX: should we check if ifa_dstaddr is NULL and return
549 * an error?
550 */
551 ifr->ifr_dstaddr = ia->ia_dstaddr;
552 break;
553
554 case SIOCGIFNETMASK_IN6:
555 ifr->ifr_addr = ia->ia_prefixmask;
556 break;
557
558 case SIOCGIFAFLAG_IN6:
559 ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
560 break;
561
562 case SIOCGIFSTAT_IN6:
563 if (ifp == NULL)
564 return EINVAL;
565 if (in6_ifstat == NULL || ifp->if_index >= in6_ifstatmax
566 || in6_ifstat[ifp->if_index] == NULL) {
567 /* return EAFNOSUPPORT? */
568 bzero(&ifr->ifr_ifru.ifru_stat,
569 sizeof(ifr->ifr_ifru.ifru_stat));
570 } else
571 ifr->ifr_ifru.ifru_stat = *in6_ifstat[ifp->if_index];
572 break;
573
574 case SIOCGIFSTAT_ICMP6:
575 if (ifp == NULL)
576 return EINVAL;
577 if (icmp6_ifstat == NULL || ifp->if_index >= icmp6_ifstatmax ||
578 icmp6_ifstat[ifp->if_index] == NULL) {
579 /* return EAFNOSUPPORT? */
580 bzero(&ifr->ifr_ifru.ifru_stat,
581 sizeof(ifr->ifr_ifru.ifru_icmp6stat));
582 } else
583 ifr->ifr_ifru.ifru_icmp6stat =
584 *icmp6_ifstat[ifp->if_index];
585 break;
586
587 case SIOCGIFALIFETIME_IN6:
588 ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
589 break;
590
591 case SIOCSIFALIFETIME_IN6:
592 ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
593 /* for sanity */
594 if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
595 ia->ia6_lifetime.ia6t_expire =
596 time_second + ia->ia6_lifetime.ia6t_vltime;
597 } else
598 ia->ia6_lifetime.ia6t_expire = 0;
599 if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
600 ia->ia6_lifetime.ia6t_preferred =
601 time_second + ia->ia6_lifetime.ia6t_pltime;
602 } else
603 ia->ia6_lifetime.ia6t_preferred = 0;
604 break;
605
606 case SIOCAIFADDR_IN6:
607 {
608 int i, error = 0;
609 struct nd_prefix pr0, *pr;
610
611 /*
612 * first, make or update the interface address structure,
613 * and link it to the list.
614 */
615 if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
616 return(error);
617
618 /*
619 * then, make the prefix on-link on the interface.
620 * XXX: we'd rather create the prefix before the address, but
621 * we need at least one address to install the corresponding
622 * interface route, so we configure the address first.
623 */
624
625 /*
626 * convert mask to prefix length (prefixmask has already
627 * been validated in in6_update_ifa().
628 */
629 bzero(&pr0, sizeof(pr0));
630 pr0.ndpr_ifp = ifp;
631 pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
632 NULL);
633 if (pr0.ndpr_plen == 128)
634 break; /* we don't need to install a host route. */
635 pr0.ndpr_prefix = ifra->ifra_addr;
636 pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
637 /* apply the mask for safety. */
638 for (i = 0; i < 4; i++) {
639 pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
640 ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
641 }
642 /*
643 * XXX: since we don't have an API to set prefix (not address)
644 * lifetimes, we just use the same lifetimes as addresses.
645 * The (temporarily) installed lifetimes can be overridden by
646 * later advertised RAs (when accept_rtadv is non 0), which is
647 * an intended behavior.
648 */
649 pr0.ndpr_raf_onlink = 1; /* should be configurable? */
650 pr0.ndpr_raf_auto =
651 ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
652 pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
653 pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
654
655 /* add the prefix if there's one. */
656 if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
657 /*
658 * nd6_prelist_add will install the corresponding
659 * interface route.
660 */
661 if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
662 return(error);
663 if (pr == NULL) {
664 log(LOG_ERR, "nd6_prelist_add succedded but "
665 "no prefix\n");
666 return(EINVAL); /* XXX panic here? */
667 }
668 }
669 if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
670 == NULL) {
671 /* XXX: this should not happen! */
672 log(LOG_ERR, "in6_control: addition succeeded, but"
673 " no ifaddr\n");
674 } else {
675 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0 &&
676 ia->ia6_ndpr == NULL) { /* new autoconfed addr */
677 ia->ia6_ndpr = pr;
678 pr->ndpr_refcnt++;
679
680 /*
681 * If this is the first autoconf address from
682 * the prefix, create a temporary address
683 * as well (when specified).
684 */
685 if (ip6_use_tempaddr &&
686 pr->ndpr_refcnt == 1) {
687 int e;
688 if ((e = in6_tmpifadd(ia, 1)) != 0) {
689 log(LOG_NOTICE, "in6_control: "
690 "failed to create a "
691 "temporary address, "
692 "errno=%d\n",
693 e);
694 }
695 }
696 }
697
698 /*
699 * this might affect the status of autoconfigured
700 * addresses, that is, this address might make
701 * other addresses detached.
702 */
703 pfxlist_onlink_check();
704 }
705 break;
706 }
707
708 case SIOCDIFADDR_IN6:
709 {
710 int i = 0;
711 struct nd_prefix pr0, *pr;
712
713 /*
714 * If the address being deleted is the only one that owns
715 * the corresponding prefix, expire the prefix as well.
716 * XXX: theoretically, we don't have to warry about such
717 * relationship, since we separate the address management
718 * and the prefix management. We do this, however, to provide
719 * as much backward compatibility as possible in terms of
720 * the ioctl operation.
721 */
722 bzero(&pr0, sizeof(pr0));
723 pr0.ndpr_ifp = ifp;
724 pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
725 NULL);
726 if (pr0.ndpr_plen == 128)
727 goto purgeaddr;
728 pr0.ndpr_prefix = ia->ia_addr;
729 pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
730 for (i = 0; i < 4; i++) {
731 pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
732 ia->ia_prefixmask.sin6_addr.s6_addr32[i];
733 }
734 /*
735 * The logic of the following condition is a bit complicated.
736 * We expire the prefix when
737 * 1. the address obeys autoconfiguration and it is the
738 * only owner of the associated prefix, or
739 * 2. the address does not obey autoconf and there is no
740 * other owner of the prefix.
741 */
742 if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
743 (((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0 &&
744 pr->ndpr_refcnt == 1) ||
745 ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0 &&
746 pr->ndpr_refcnt == 0))) {
747 pr->ndpr_expire = 1; /* XXX: just for expiration */
748 }
749
750 purgeaddr:
751 in6_purgeaddr(&ia->ia_ifa);
752 break;
753 }
754
755 default:
756 if (ifp == NULL || ifp->if_ioctl == 0)
757 return(EOPNOTSUPP);
758 return((*ifp->if_ioctl)(ifp, cmd, data));
759 }
760
761 return(0);
762}
763
764/*
765 * Update parameters of an IPv6 interface address.
766 * If necessary, a new entry is created and linked into address chains.
767 * This function is separated from in6_control().
768 * XXX: should this be performed under splnet()?
769 */
770int
771in6_update_ifa(ifp, ifra, ia)
772 struct ifnet *ifp;
773 struct in6_aliasreq *ifra;
774 struct in6_ifaddr *ia;
775{
776 int error = 0, hostIsNew = 0, plen = -1;
777 struct in6_ifaddr *oia;
778 struct sockaddr_in6 dst6;
779 struct in6_addrlifetime *lt;
780
781 /* Validate parameters */
782 if (ifp == NULL || ifra == NULL) /* this maybe redundant */
783 return(EINVAL);
784
785 /*
786 * The destination address for a p2p link must have a family
787 * of AF_UNSPEC or AF_INET6.
788 */
789 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
790 ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
791 ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
792 return(EAFNOSUPPORT);
793 /*
794 * validate ifra_prefixmask. don't check sin6_family, netmask
795 * does not carry fields other than sin6_len.
796 */
797 if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
798 return(EINVAL);
799 /*
800 * Because the IPv6 address architecture is classless, we require
801 * users to specify a (non 0) prefix length (mask) for a new address.
802 * We also require the prefix (when specified) mask is valid, and thus
803 * reject a non-consecutive mask.
804 */
805 if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
806 return(EINVAL);
807 if (ifra->ifra_prefixmask.sin6_len != 0) {
808 plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
809 (u_char *)&ifra->ifra_prefixmask +
810 ifra->ifra_prefixmask.sin6_len);
811 if (plen <= 0)
812 return(EINVAL);
813 }
814 else {
815 /*
816 * In this case, ia must not be NULL. We just use its prefix
817 * length.
818 */
819 plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
820 }
821 /*
822 * If the destination address on a p2p interface is specified,
823 * and the address is a scoped one, validate/set the scope
824 * zone identifier.
825 */
826 dst6 = ifra->ifra_dstaddr;
827 if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) &&
828 (dst6.sin6_family == AF_INET6)) {
829 int scopeid;
830
831#ifndef SCOPEDROUTING
832 if ((error = in6_recoverscope(&dst6,
833 &ifra->ifra_dstaddr.sin6_addr,
834 ifp)) != 0)
835 return(error);
836#endif
837 scopeid = in6_addr2scopeid(ifp, &dst6.sin6_addr);
838 if (dst6.sin6_scope_id == 0) /* user omit to specify the ID. */
839 dst6.sin6_scope_id = scopeid;
840 else if (dst6.sin6_scope_id != scopeid)
841 return(EINVAL); /* scope ID mismatch. */
842#ifndef SCOPEDROUTING
843 if ((error = in6_embedscope(&dst6.sin6_addr, &dst6, NULL, NULL))
844 != 0)
845 return(error);
846 dst6.sin6_scope_id = 0; /* XXX */
847#endif
848 }
849 /*
850 * The destination address can be specified only for a p2p or a
851 * loopback interface. If specified, the corresponding prefix length
852 * must be 128.
853 */
854 if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
855 if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
856 /* XXX: noisy message */
857 log(LOG_INFO, "in6_update_ifa: a destination can be "
858 "specified for a p2p or a loopback IF only\n");
859 return(EINVAL);
860 }
861 if (plen != 128) {
862 /*
863 * The following message seems noisy, but we dare to
864 * add it for diagnosis.
865 */
866 log(LOG_INFO, "in6_update_ifa: prefixlen must be 128 "
867 "when dstaddr is specified\n");
868 return(EINVAL);
869 }
870 }
871 /* lifetime consistency check */
872 lt = &ifra->ifra_lifetime;
873 if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
874 && lt->ia6t_vltime + time_second < time_second) {
875 return EINVAL;
876 }
877 if (lt->ia6t_vltime == 0) {
878 /*
879 * the following log might be noisy, but this is a typical
880 * configuration mistake or a tool's bug.
881 */
882 log(LOG_INFO,
883 "in6_update_ifa: valid lifetime is 0 for %s\n",
884 ip6_sprintf(&ifra->ifra_addr.sin6_addr));
885 }
886 if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
887 && lt->ia6t_pltime + time_second < time_second) {
888 return EINVAL;
889 }
890
891 /*
892 * If this is a new address, allocate a new ifaddr and link it
893 * into chains.
894 */
895 if (ia == NULL) {
896 hostIsNew = 1;
897 /*
898 * When in6_update_ifa() is called in a process of a received
899 * RA, it is called under splnet(). So, we should call malloc
900 * with M_NOWAIT.
901 */
902 ia = (struct in6_ifaddr *)
903 malloc(sizeof(*ia), M_IFADDR, M_NOWAIT);
904 if (ia == NULL)
905 return (ENOBUFS);
906 bzero((caddr_t)ia, sizeof(*ia));
907 /* Initialize the address and masks */
908 ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
909 ia->ia_addr.sin6_family = AF_INET6;
910 ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
911 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
912 /*
913 * XXX: some functions expect that ifa_dstaddr is not
914 * NULL for p2p interfaces.
915 */
916 ia->ia_ifa.ifa_dstaddr
917 = (struct sockaddr *)&ia->ia_dstaddr;
918 } else {
919 ia->ia_ifa.ifa_dstaddr = NULL;
920 }
921 ia->ia_ifa.ifa_netmask
922 = (struct sockaddr *)&ia->ia_prefixmask;
923
924 ia->ia_ifp = ifp;
925 if ((oia = in6_ifaddr) != NULL) {
926 for ( ; oia->ia_next; oia = oia->ia_next)
927 continue;
928 oia->ia_next = ia;
929 } else
930 in6_ifaddr = ia;
931
932 TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa,
933 ifa_list);
934 }
935
936 /* set prefix mask */
937 if (ifra->ifra_prefixmask.sin6_len) {
938 /*
939 * We prohibit changing the prefix length of an existing
940 * address, because
941 * + such an operation should be rare in IPv6, and
942 * + the operation would confuse prefix management.
943 */
944 if (ia->ia_prefixmask.sin6_len &&
945 in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
946 log(LOG_INFO, "in6_update_ifa: the prefix length of an"
947 " existing (%s) address should not be changed\n",
948 ip6_sprintf(&ia->ia_addr.sin6_addr));
949 error = EINVAL;
950 goto unlink;
951 }
952 ia->ia_prefixmask = ifra->ifra_prefixmask;
953 }
954
955 /*
956 * If a new destination address is specified, scrub the old one and
957 * install the new destination. Note that the interface must be
958 * p2p or loopback (see the check above.)
959 */
960 if (dst6.sin6_family == AF_INET6 &&
961 !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
962 &ia->ia_dstaddr.sin6_addr)) {
963 int e;
964
965 if ((ia->ia_flags & IFA_ROUTE) != 0 &&
966 (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
967 != 0) {
968 log(LOG_ERR, "in6_update_ifa: failed to remove "
969 "a route to the old destination: %s\n",
970 ip6_sprintf(&ia->ia_addr.sin6_addr));
971 /* proceed anyway... */
972 }
973 else
974 ia->ia_flags &= ~IFA_ROUTE;
975 ia->ia_dstaddr = dst6;
976 }
977
978 /* reset the interface and routing table appropriately. */
979 if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
980 goto unlink;
981
982 /*
983 * Beyond this point, we should call in6_purgeaddr upon an error,
984 * not just go to unlink.
985 */
986
987#if 0 /* disable this mechanism for now */
988 /* update prefix list */
989 if (hostIsNew &&
990 (ifra->ifra_flags & IN6_IFF_NOPFX) == 0) { /* XXX */
991 int iilen;
992
993 iilen = (sizeof(ia->ia_prefixmask.sin6_addr) << 3) - plen;
994 if ((error = in6_prefix_add_ifid(iilen, ia)) != 0) {
995 in6_purgeaddr((struct ifaddr *)ia);
996 return(error);
997 }
998 }
999#endif
1000
1001 if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1002 struct sockaddr_in6 mltaddr, mltmask;
1003 struct in6_multi *in6m;
1004
1005 if (hostIsNew) {
1006 /*
1007 * join solicited multicast addr for new host id
1008 */
1009 struct in6_addr llsol;
1010 bzero(&llsol, sizeof(struct in6_addr));
1011 llsol.s6_addr16[0] = htons(0xff02);
1012 llsol.s6_addr16[1] = htons(ifp->if_index);
1013 llsol.s6_addr32[1] = 0;
1014 llsol.s6_addr32[2] = htonl(1);
1015 llsol.s6_addr32[3] =
1016 ifra->ifra_addr.sin6_addr.s6_addr32[3];
1017 llsol.s6_addr8[12] = 0xff;
1018 (void)in6_addmulti(&llsol, ifp, &error);
1019 if (error != 0) {
1020 log(LOG_WARNING,
1021 "in6_update_ifa: addmulti failed for "
1022 "%s on %s (errno=%d)\n",
1023 ip6_sprintf(&llsol), if_name(ifp),
1024 error);
1025 in6_purgeaddr((struct ifaddr *)ia);
1026 return(error);
1027 }
1028 }
1029
1030 bzero(&mltmask, sizeof(mltmask));
1031 mltmask.sin6_len = sizeof(struct sockaddr_in6);
1032 mltmask.sin6_family = AF_INET6;
1033 mltmask.sin6_addr = in6mask32;
1034
1035 /*
1036 * join link-local all-nodes address
1037 */
1038 bzero(&mltaddr, sizeof(mltaddr));
1039 mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1040 mltaddr.sin6_family = AF_INET6;
1041 mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1042 mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1043
1044 IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1045 if (in6m == NULL) {
1046 rtrequest(RTM_ADD,
1047 (struct sockaddr *)&mltaddr,
1048 (struct sockaddr *)&ia->ia_addr,
1049 (struct sockaddr *)&mltmask,
1050 RTF_UP|RTF_CLONING, /* xxx */
1051 (struct rtentry **)0);
1052 (void)in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1053 if (error != 0) {
1054 log(LOG_WARNING,
1055 "in6_update_ifa: addmulti failed for "
1056 "%s on %s (errno=%d)\n",
1057 ip6_sprintf(&mltaddr.sin6_addr),
1058 if_name(ifp), error);
1059 }
1060 }
1061
1062 /*
1063 * join node information group address
1064 */
1065#define hostnamelen strlen(hostname)
1066 if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1067 == 0) {
1068 IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1069 if (in6m == NULL && ia != NULL) {
1070 (void)in6_addmulti(&mltaddr.sin6_addr,
1071 ifp, &error);
1072 if (error != 0) {
1073 log(LOG_WARNING, "in6_update_ifa: "
1074 "addmulti failed for "
1075 "%s on %s (errno=%d)\n",
1076 ip6_sprintf(&mltaddr.sin6_addr),
1077 if_name(ifp), error);
1078 }
1079 }
1080 }
1081#undef hostnamelen
1082
1083 /*
1084 * join node-local all-nodes address, on loopback.
1085 * XXX: since "node-local" is obsoleted by interface-local,
1086 * we have to join the group on every interface with
1087 * some interface-boundary restriction.
1088 */
1089 if (ifp->if_flags & IFF_LOOPBACK) {
1090 struct in6_ifaddr *ia_loop;
1091
1092 struct in6_addr loop6 = in6addr_loopback;
1093 ia_loop = in6ifa_ifpwithaddr(ifp, &loop6);
1094
1095 mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1096
1097 IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1098 if (in6m == NULL && ia_loop != NULL) {
1099 rtrequest(RTM_ADD,
1100 (struct sockaddr *)&mltaddr,
1101 (struct sockaddr *)&ia_loop->ia_addr,
1102 (struct sockaddr *)&mltmask,
1103 RTF_UP,
1104 (struct rtentry **)0);
1105 (void)in6_addmulti(&mltaddr.sin6_addr, ifp,
1106 &error);
1107 if (error != 0) {
1108 log(LOG_WARNING, "in6_update_ifa: "
1109 "addmulti failed for %s on %s "
1110 "(errno=%d)\n",
1111 ip6_sprintf(&mltaddr.sin6_addr),
1112 if_name(ifp), error);
1113 }
1114 }
1115 }
1116 }
1117
1118 ia->ia6_flags = ifra->ifra_flags;
1119 ia->ia6_flags &= ~IN6_IFF_DUPLICATED; /*safety*/
1120 ia->ia6_flags &= ~IN6_IFF_NODAD; /* Mobile IPv6 */
1121
1122 ia->ia6_lifetime = ifra->ifra_lifetime;
1123 /* for sanity */
1124 if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1125 ia->ia6_lifetime.ia6t_expire =
1126 time_second + ia->ia6_lifetime.ia6t_vltime;
1127 } else
1128 ia->ia6_lifetime.ia6t_expire = 0;
1129 if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1130 ia->ia6_lifetime.ia6t_preferred =
1131 time_second + ia->ia6_lifetime.ia6t_pltime;
1132 } else
1133 ia->ia6_lifetime.ia6t_preferred = 0;
1134
1135 /*
1136 * make sure to initialize ND6 information. this is to workaround
1137 * issues with interfaces with IPv6 addresses, which have never brought
1138 * up. We are assuming that it is safe to nd6_ifattach multiple times.
1139 */
1140 nd6_ifattach(ifp);
1141
1142 /*
1143 * Perform DAD, if needed.
1144 * XXX It may be of use, if we can administratively
1145 * disable DAD.
1146 */
1147 if (in6if_do_dad(ifp) && (ifra->ifra_flags & IN6_IFF_NODAD) == 0) {
1148 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1149 nd6_dad_start((struct ifaddr *)ia, NULL);
1150 }
1151
1152 return(error);
1153
1154 unlink:
1155 /*
1156 * XXX: if a change of an existing address failed, keep the entry
1157 * anyway.
1158 */
1159 if (hostIsNew)
1160 in6_unlink_ifa(ia, ifp);
1161 return(error);
1162}
1163
1164void
1165in6_purgeaddr(ifa)
1166 struct ifaddr *ifa;
1167{
1168 struct ifnet *ifp = ifa->ifa_ifp;
1169 struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1170
1171 /* stop DAD processing */
1172 nd6_dad_stop(ifa);
1173
1174 /*
1175 * delete route to the destination of the address being purged.
1176 * The interface must be p2p or loopback in this case.
1177 */
1178 if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1179 int e;
1180
1181 if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1182 != 0) {
1183 log(LOG_ERR, "in6_purgeaddr: failed to remove "
1184 "a route to the p2p destination: %s on %s, "
1185 "errno=%d\n",
1186 ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1187 e);
1188 /* proceed anyway... */
1189 }
1190 else
1191 ia->ia_flags &= ~IFA_ROUTE;
1192 }
1193
1194 /* Remove ownaddr's loopback rtentry, if it exists. */
1195 in6_ifremloop(&(ia->ia_ifa));
1196
1197 if (ifp->if_flags & IFF_MULTICAST) {
1198 /*
1199 * delete solicited multicast addr for deleting host id
1200 */
1201 struct in6_multi *in6m;
1202 struct in6_addr llsol;
1203 bzero(&llsol, sizeof(struct in6_addr));
1204 llsol.s6_addr16[0] = htons(0xff02);
1205 llsol.s6_addr16[1] = htons(ifp->if_index);
1206 llsol.s6_addr32[1] = 0;
1207 llsol.s6_addr32[2] = htonl(1);
1208 llsol.s6_addr32[3] =
1209 ia->ia_addr.sin6_addr.s6_addr32[3];
1210 llsol.s6_addr8[12] = 0xff;
1211
1212 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1213 if (in6m)
1214 in6_delmulti(in6m);
1215 }
1216
1217 in6_unlink_ifa(ia, ifp);
1218}
1219
1220static void
1221in6_unlink_ifa(ia, ifp)
1222 struct in6_ifaddr *ia;
1223 struct ifnet *ifp;
1224{
1225 int plen, iilen;
1226 struct in6_ifaddr *oia;
1227 int s = splnet();
1228
1229 TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1230
1231 oia = ia;
1232 if (oia == (ia = in6_ifaddr))
1233 in6_ifaddr = ia->ia_next;
1234 else {
1235 while (ia->ia_next && (ia->ia_next != oia))
1236 ia = ia->ia_next;
1237 if (ia->ia_next)
1238 ia->ia_next = oia->ia_next;
1239 else {
1240 /* search failed */
1241 printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1242 }
1243 }
1244
1245 if (oia->ia6_ifpr) { /* check for safety */
1246 plen = in6_mask2len(&oia->ia_prefixmask.sin6_addr, NULL);
1247 iilen = (sizeof(oia->ia_prefixmask.sin6_addr) << 3) - plen;
1248 in6_prefix_remove_ifid(iilen, oia);
1249 }
1250
1251 /*
1252 * When an autoconfigured address is being removed, release the
1253 * reference to the base prefix. Also, since the release might
1254 * affect the status of other (detached) addresses, call
1255 * pfxlist_onlink_check().
1256 */
1257 if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1258 if (oia->ia6_ndpr == NULL) {
1259 log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1260 "%p has no prefix\n", oia);
1261 } else {
1262 oia->ia6_ndpr->ndpr_refcnt--;
1263 oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1264 oia->ia6_ndpr = NULL;
1265 }
1266
1267 pfxlist_onlink_check();
1268 }
1269
1270 /*
1271 * release another refcnt for the link from in6_ifaddr.
1272 * Note that we should decrement the refcnt at least once for all *BSD.
1273 */
1274 IFAFREE(&oia->ia_ifa);
1275
1276 splx(s);
1277}
1278
1279void
1280in6_purgeif(ifp)
1281 struct ifnet *ifp;
1282{
1283 struct ifaddr *ifa, *nifa;
1284
1285 for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa)
1286 {
1287 nifa = TAILQ_NEXT(ifa, ifa_list);
1288 if (ifa->ifa_addr->sa_family != AF_INET6)
1289 continue;
1290 in6_purgeaddr(ifa);
1291 }
1292
1293 in6_ifdetach(ifp);
1294}
1295
1296/*
1297 * SIOC[GAD]LIFADDR.
1298 * SIOCGLIFADDR: get first address. (?)
1299 * SIOCGLIFADDR with IFLR_PREFIX:
1300 * get first address that matches the specified prefix.
1301 * SIOCALIFADDR: add the specified address.
1302 * SIOCALIFADDR with IFLR_PREFIX:
1303 * add the specified prefix, filling hostid part from
1304 * the first link-local address. prefixlen must be <= 64.
1305 * SIOCDLIFADDR: delete the specified address.
1306 * SIOCDLIFADDR with IFLR_PREFIX:
1307 * delete the first address that matches the specified prefix.
1308 * return values:
1309 * EINVAL on invalid parameters
1310 * EADDRNOTAVAIL on prefix match failed/specified address not found
1311 * other values may be returned from in6_ioctl()
1312 *
1313 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1314 * this is to accomodate address naming scheme other than RFC2374,
1315 * in the future.
1316 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1317 * address encoding scheme. (see figure on page 8)
1318 */
1319static int
dadab5e9
MD
1320in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1321 struct ifnet *ifp, struct thread *td)
984263bc
MD
1322{
1323 struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1324 struct ifaddr *ifa;
1325 struct sockaddr *sa;
1326
1327 /* sanity checks */
1328 if (!data || !ifp) {
1329 panic("invalid argument to in6_lifaddr_ioctl");
1330 /*NOTRECHED*/
1331 }
1332
1333 switch (cmd) {
1334 case SIOCGLIFADDR:
1335 /* address must be specified on GET with IFLR_PREFIX */
1336 if ((iflr->flags & IFLR_PREFIX) == 0)
1337 break;
1338 /* FALLTHROUGH */
1339 case SIOCALIFADDR:
1340 case SIOCDLIFADDR:
1341 /* address must be specified on ADD and DELETE */
1342 sa = (struct sockaddr *)&iflr->addr;
1343 if (sa->sa_family != AF_INET6)
1344 return EINVAL;
1345 if (sa->sa_len != sizeof(struct sockaddr_in6))
1346 return EINVAL;
1347 /* XXX need improvement */
1348 sa = (struct sockaddr *)&iflr->dstaddr;
1349 if (sa->sa_family && sa->sa_family != AF_INET6)
1350 return EINVAL;
1351 if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1352 return EINVAL;
1353 break;
1354 default: /* shouldn't happen */
1355#if 0
1356 panic("invalid cmd to in6_lifaddr_ioctl");
1357 /* NOTREACHED */
1358#else
1359 return EOPNOTSUPP;
1360#endif
1361 }
1362 if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1363 return EINVAL;
1364
1365 switch (cmd) {
1366 case SIOCALIFADDR:
1367 {
1368 struct in6_aliasreq ifra;
1369 struct in6_addr *hostid = NULL;
1370 int prefixlen;
1371
1372 if ((iflr->flags & IFLR_PREFIX) != 0) {
1373 struct sockaddr_in6 *sin6;
1374
1375 /*
1376 * hostid is to fill in the hostid part of the
1377 * address. hostid points to the first link-local
1378 * address attached to the interface.
1379 */
1380 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1381 if (!ifa)
1382 return EADDRNOTAVAIL;
1383 hostid = IFA_IN6(ifa);
1384
1385 /* prefixlen must be <= 64. */
1386 if (64 < iflr->prefixlen)
1387 return EINVAL;
1388 prefixlen = iflr->prefixlen;
1389
1390 /* hostid part must be zero. */
1391 sin6 = (struct sockaddr_in6 *)&iflr->addr;
1392 if (sin6->sin6_addr.s6_addr32[2] != 0
1393 || sin6->sin6_addr.s6_addr32[3] != 0) {
1394 return EINVAL;
1395 }
1396 } else
1397 prefixlen = iflr->prefixlen;
1398
1399 /* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1400 bzero(&ifra, sizeof(ifra));
1401 bcopy(iflr->iflr_name, ifra.ifra_name,
1402 sizeof(ifra.ifra_name));
1403
1404 bcopy(&iflr->addr, &ifra.ifra_addr,
1405 ((struct sockaddr *)&iflr->addr)->sa_len);
1406 if (hostid) {
1407 /* fill in hostid part */
1408 ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1409 hostid->s6_addr32[2];
1410 ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1411 hostid->s6_addr32[3];
1412 }
1413
1414 if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /*XXX*/
1415 bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1416 ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1417 if (hostid) {
1418 ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1419 hostid->s6_addr32[2];
1420 ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1421 hostid->s6_addr32[3];
1422 }
1423 }
1424
1425 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1426 in6_len2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1427
1428 ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1429 return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, p);
1430 }
1431 case SIOCGLIFADDR:
1432 case SIOCDLIFADDR:
1433 {
1434 struct in6_ifaddr *ia;
1435 struct in6_addr mask, candidate, match;
1436 struct sockaddr_in6 *sin6;
1437 int cmp;
1438
1439 bzero(&mask, sizeof(mask));
1440 if (iflr->flags & IFLR_PREFIX) {
1441 /* lookup a prefix rather than address. */
1442 in6_len2mask(&mask, iflr->prefixlen);
1443
1444 sin6 = (struct sockaddr_in6 *)&iflr->addr;
1445 bcopy(&sin6->sin6_addr, &match, sizeof(match));
1446 match.s6_addr32[0] &= mask.s6_addr32[0];
1447 match.s6_addr32[1] &= mask.s6_addr32[1];
1448 match.s6_addr32[2] &= mask.s6_addr32[2];
1449 match.s6_addr32[3] &= mask.s6_addr32[3];
1450
1451 /* if you set extra bits, that's wrong */
1452 if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1453 return EINVAL;
1454
1455 cmp = 1;
1456 } else {
1457 if (cmd == SIOCGLIFADDR) {
1458 /* on getting an address, take the 1st match */
1459 cmp = 0; /* XXX */
1460 } else {
1461 /* on deleting an address, do exact match */
1462 in6_len2mask(&mask, 128);
1463 sin6 = (struct sockaddr_in6 *)&iflr->addr;
1464 bcopy(&sin6->sin6_addr, &match, sizeof(match));
1465
1466 cmp = 1;
1467 }
1468 }
1469
1470 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1471 {
1472 if (ifa->ifa_addr->sa_family != AF_INET6)
1473 continue;
1474 if (!cmp)
1475 break;
1476
1477 bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1478#ifndef SCOPEDROUTING
1479 /*
1480 * XXX: this is adhoc, but is necessary to allow
1481 * a user to specify fe80::/64 (not /10) for a
1482 * link-local address.
1483 */
1484 if (IN6_IS_ADDR_LINKLOCAL(&candidate))
1485 candidate.s6_addr16[1] = 0;
1486#endif
1487 candidate.s6_addr32[0] &= mask.s6_addr32[0];
1488 candidate.s6_addr32[1] &= mask.s6_addr32[1];
1489 candidate.s6_addr32[2] &= mask.s6_addr32[2];
1490 candidate.s6_addr32[3] &= mask.s6_addr32[3];
1491 if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1492 break;
1493 }
1494 if (!ifa)
1495 return EADDRNOTAVAIL;
1496 ia = ifa2ia6(ifa);
1497
1498 if (cmd == SIOCGLIFADDR) {
1499#ifndef SCOPEDROUTING
1500 struct sockaddr_in6 *s6;
1501#endif
1502
1503 /* fill in the if_laddrreq structure */
1504 bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1505#ifndef SCOPEDROUTING /* XXX see above */
1506 s6 = (struct sockaddr_in6 *)&iflr->addr;
1507 if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1508 s6->sin6_addr.s6_addr16[1] = 0;
1509 s6->sin6_scope_id =
1510 in6_addr2scopeid(ifp, &s6->sin6_addr);
1511 }
1512#endif
1513 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1514 bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1515 ia->ia_dstaddr.sin6_len);
1516#ifndef SCOPEDROUTING /* XXX see above */
1517 s6 = (struct sockaddr_in6 *)&iflr->dstaddr;
1518 if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1519 s6->sin6_addr.s6_addr16[1] = 0;
1520 s6->sin6_scope_id =
1521 in6_addr2scopeid(ifp,
1522 &s6->sin6_addr);
1523 }
1524#endif
1525 } else
1526 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1527
1528 iflr->prefixlen =
1529 in6_mask2len(&ia->ia_prefixmask.sin6_addr,
1530 NULL);
1531
1532 iflr->flags = ia->ia6_flags; /* XXX */
1533
1534 return 0;
1535 } else {
1536 struct in6_aliasreq ifra;
1537
1538 /* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1539 bzero(&ifra, sizeof(ifra));
1540 bcopy(iflr->iflr_name, ifra.ifra_name,
1541 sizeof(ifra.ifra_name));
1542
1543 bcopy(&ia->ia_addr, &ifra.ifra_addr,
1544 ia->ia_addr.sin6_len);
1545 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1546 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1547 ia->ia_dstaddr.sin6_len);
1548 } else {
1549 bzero(&ifra.ifra_dstaddr,
1550 sizeof(ifra.ifra_dstaddr));
1551 }
1552 bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1553 ia->ia_prefixmask.sin6_len);
1554
1555 ifra.ifra_flags = ia->ia6_flags;
1556 return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1557 ifp, p);
1558 }
1559 }
1560 }
1561
1562 return EOPNOTSUPP; /* just for safety */
1563}
1564
1565/*
1566 * Initialize an interface's intetnet6 address
1567 * and routing table entry.
1568 */
1569static int
1570in6_ifinit(ifp, ia, sin6, newhost)
1571 struct ifnet *ifp;
1572 struct in6_ifaddr *ia;
1573 struct sockaddr_in6 *sin6;
1574 int newhost;
1575{
1576 int error = 0, plen, ifacount = 0;
1577 int s = splimp();
1578 struct ifaddr *ifa;
1579
1580 /*
1581 * Give the interface a chance to initialize
1582 * if this is its first address,
1583 * and to validate the address if necessary.
1584 */
1585 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1586 {
1587 if (ifa->ifa_addr == NULL)
1588 continue; /* just for safety */
1589 if (ifa->ifa_addr->sa_family != AF_INET6)
1590 continue;
1591 ifacount++;
1592 }
1593
1594 ia->ia_addr = *sin6;
1595
1596 if (ifacount <= 1 && ifp->if_ioctl &&
1597 (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia))) {
1598 splx(s);
1599 return(error);
1600 }
1601 splx(s);
1602
1603 ia->ia_ifa.ifa_metric = ifp->if_metric;
1604
1605 /* we could do in(6)_socktrim here, but just omit it at this moment. */
1606
1607 /*
1608 * Special case:
1609 * If the destination address is specified for a point-to-point
1610 * interface, install a route to the destination as an interface
1611 * direct route.
1612 */
1613 plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1614 if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1615 if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1616 RTF_UP | RTF_HOST)) != 0)
1617 return(error);
1618 ia->ia_flags |= IFA_ROUTE;
1619 }
1620 if (plen < 128) {
1621 /*
1622 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1623 */
1624 ia->ia_ifa.ifa_flags |= RTF_CLONING;
1625 }
1626
1627 /* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1628 if (newhost) {
1629 /* set the rtrequest function to create llinfo */
1630 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1631 in6_ifaddloop(&(ia->ia_ifa));
1632 }
1633
1634 return(error);
1635}
1636
1637/*
1638 * Add an address to the list of IP6 multicast addresses for a
1639 * given interface.
1640 */
1641struct in6_multi *
1642in6_addmulti(maddr6, ifp, errorp)
1643 struct in6_addr *maddr6;
1644 struct ifnet *ifp;
1645 int *errorp;
1646{
1647 struct in6_multi *in6m;
1648 struct sockaddr_in6 sin6;
1649 struct ifmultiaddr *ifma;
1650 int s = splnet();
1651
1652 *errorp = 0;
1653
1654 /*
1655 * Call generic routine to add membership or increment
1656 * refcount. It wants addresses in the form of a sockaddr,
1657 * so we build one here (being careful to zero the unused bytes).
1658 */
1659 bzero(&sin6, sizeof sin6);
1660 sin6.sin6_family = AF_INET6;
1661 sin6.sin6_len = sizeof sin6;
1662 sin6.sin6_addr = *maddr6;
1663 *errorp = if_addmulti(ifp, (struct sockaddr *)&sin6, &ifma);
1664 if (*errorp) {
1665 splx(s);
1666 return 0;
1667 }
1668
1669 /*
1670 * If ifma->ifma_protospec is null, then if_addmulti() created
1671 * a new record. Otherwise, we are done.
1672 */
1673 if (ifma->ifma_protospec != 0)
1674 return ifma->ifma_protospec;
1675
1676 /* XXX - if_addmulti uses M_WAITOK. Can this really be called
1677 at interrupt time? If so, need to fix if_addmulti. XXX */
1678 in6m = (struct in6_multi *)malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT);
1679 if (in6m == NULL) {
1680 splx(s);
1681 return (NULL);
1682 }
1683
1684 bzero(in6m, sizeof *in6m);
1685 in6m->in6m_addr = *maddr6;
1686 in6m->in6m_ifp = ifp;
1687 in6m->in6m_ifma = ifma;
1688 ifma->ifma_protospec = in6m;
1689 LIST_INSERT_HEAD(&in6_multihead, in6m, in6m_entry);
1690
1691 /*
1692 * Let MLD6 know that we have joined a new IP6 multicast
1693 * group.
1694 */
1695 mld6_start_listening(in6m);
1696 splx(s);
1697 return(in6m);
1698}
1699
1700/*
1701 * Delete a multicast address record.
1702 */
1703void
1704in6_delmulti(in6m)
1705 struct in6_multi *in6m;
1706{
1707 struct ifmultiaddr *ifma = in6m->in6m_ifma;
1708 int s = splnet();
1709
1710 if (ifma->ifma_refcount == 1) {
1711 /*
1712 * No remaining claims to this record; let MLD6 know
1713 * that we are leaving the multicast group.
1714 */
1715 mld6_stop_listening(in6m);
1716 ifma->ifma_protospec = 0;
1717 LIST_REMOVE(in6m, in6m_entry);
1718 free(in6m, M_IPMADDR);
1719 }
1720 /* XXX - should be separate API for when we have an ifma? */
1721 if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1722 splx(s);
1723}
1724
1725/*
1726 * Find an IPv6 interface link-local address specific to an interface.
1727 */
1728struct in6_ifaddr *
1729in6ifa_ifpforlinklocal(ifp, ignoreflags)
1730 struct ifnet *ifp;
1731 int ignoreflags;
1732{
1733 struct ifaddr *ifa;
1734
1735 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1736 {
1737 if (ifa->ifa_addr == NULL)
1738 continue; /* just for safety */
1739 if (ifa->ifa_addr->sa_family != AF_INET6)
1740 continue;
1741 if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1742 if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1743 ignoreflags) != 0)
1744 continue;
1745 break;
1746 }
1747 }
1748
1749 return((struct in6_ifaddr *)ifa);
1750}
1751
1752
1753/*
1754 * find the internet address corresponding to a given interface and address.
1755 */
1756struct in6_ifaddr *
1757in6ifa_ifpwithaddr(ifp, addr)
1758 struct ifnet *ifp;
1759 struct in6_addr *addr;
1760{
1761 struct ifaddr *ifa;
1762
1763 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1764 {
1765 if (ifa->ifa_addr == NULL)
1766 continue; /* just for safety */
1767 if (ifa->ifa_addr->sa_family != AF_INET6)
1768 continue;
1769 if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1770 break;
1771 }
1772
1773 return((struct in6_ifaddr *)ifa);
1774}
1775
1776/*
1777 * Convert IP6 address to printable (loggable) representation.
1778 */
1779static char digits[] = "0123456789abcdef";
1780static int ip6round = 0;
1781char *
1782ip6_sprintf(addr)
1783 const struct in6_addr *addr;
1784{
1785 static char ip6buf[8][48];
1786 int i;
1787 char *cp;
1788 u_short *a = (u_short *)addr;
1789 u_char *d;
1790 int dcolon = 0;
1791
1792 ip6round = (ip6round + 1) & 7;
1793 cp = ip6buf[ip6round];
1794
1795 for (i = 0; i < 8; i++) {
1796 if (dcolon == 1) {
1797 if (*a == 0) {
1798 if (i == 7)
1799 *cp++ = ':';
1800 a++;
1801 continue;
1802 } else
1803 dcolon = 2;
1804 }
1805 if (*a == 0) {
1806 if (dcolon == 0 && *(a + 1) == 0) {
1807 if (i == 0)
1808 *cp++ = ':';
1809 *cp++ = ':';
1810 dcolon = 1;
1811 } else {
1812 *cp++ = '0';
1813 *cp++ = ':';
1814 }
1815 a++;
1816 continue;
1817 }
1818 d = (u_char *)a;
1819 *cp++ = digits[*d >> 4];
1820 *cp++ = digits[*d++ & 0xf];
1821 *cp++ = digits[*d >> 4];
1822 *cp++ = digits[*d & 0xf];
1823 *cp++ = ':';
1824 a++;
1825 }
1826 *--cp = 0;
1827 return(ip6buf[ip6round]);
1828}
1829
1830int
1831in6_localaddr(in6)
1832 struct in6_addr *in6;
1833{
1834 struct in6_ifaddr *ia;
1835
1836 if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1837 return 1;
1838
1839 for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1840 if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1841 &ia->ia_prefixmask.sin6_addr))
1842 return 1;
1843
1844 return (0);
1845}
1846
1847int
1848in6_is_addr_deprecated(sa6)
1849 struct sockaddr_in6 *sa6;
1850{
1851 struct in6_ifaddr *ia;
1852
1853 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1854 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1855 &sa6->sin6_addr) &&
1856#ifdef SCOPEDROUTING
1857 ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1858#endif
1859 (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1860 return(1); /* true */
1861
1862 /* XXX: do we still have to go thru the rest of the list? */
1863 }
1864
1865 return(0); /* false */
1866}
1867
1868/*
1869 * return length of part which dst and src are equal
1870 * hard coding...
1871 */
1872int
1873in6_matchlen(src, dst)
1874struct in6_addr *src, *dst;
1875{
1876 int match = 0;
1877 u_char *s = (u_char *)src, *d = (u_char *)dst;
1878 u_char *lim = s + 16, r;
1879
1880 while (s < lim)
1881 if ((r = (*d++ ^ *s++)) != 0) {
1882 while (r < 128) {
1883 match++;
1884 r <<= 1;
1885 }
1886 break;
1887 } else
1888 match += 8;
1889 return match;
1890}
1891
1892/* XXX: to be scope conscious */
1893int
1894in6_are_prefix_equal(p1, p2, len)
1895 struct in6_addr *p1, *p2;
1896 int len;
1897{
1898 int bytelen, bitlen;
1899
1900 /* sanity check */
1901 if (0 > len || len > 128) {
1902 log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1903 len);
1904 return(0);
1905 }
1906
1907 bytelen = len / 8;
1908 bitlen = len % 8;
1909
1910 if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1911 return(0);
1912 if (p1->s6_addr[bytelen] >> (8 - bitlen) !=
1913 p2->s6_addr[bytelen] >> (8 - bitlen))
1914 return(0);
1915
1916 return(1);
1917}
1918
1919void
1920in6_prefixlen2mask(maskp, len)
1921 struct in6_addr *maskp;
1922 int len;
1923{
1924 u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1925 int bytelen, bitlen, i;
1926
1927 /* sanity check */
1928 if (0 > len || len > 128) {
1929 log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1930 len);
1931 return;
1932 }
1933
1934 bzero(maskp, sizeof(*maskp));
1935 bytelen = len / 8;
1936 bitlen = len % 8;
1937 for (i = 0; i < bytelen; i++)
1938 maskp->s6_addr[i] = 0xff;
1939 if (bitlen)
1940 maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1941}
1942
1943/*
1944 * return the best address out of the same scope
1945 */
1946struct in6_ifaddr *
1947in6_ifawithscope(oifp, dst)
1948 struct ifnet *oifp;
1949 struct in6_addr *dst;
1950{
1951 int dst_scope = in6_addrscope(dst), src_scope, best_scope = 0;
1952 int blen = -1;
1953 struct ifaddr *ifa;
1954 struct ifnet *ifp;
1955 struct in6_ifaddr *ifa_best = NULL;
1956
1957 if (oifp == NULL) {
1958#if 0
1959 printf("in6_ifawithscope: output interface is not specified\n");
1960#endif
1961 return(NULL);
1962 }
1963
1964 /*
1965 * We search for all addresses on all interfaces from the beginning.
1966 * Comparing an interface with the outgoing interface will be done
1967 * only at the final stage of tiebreaking.
1968 */
1969 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1970 {
1971 /*
1972 * We can never take an address that breaks the scope zone
1973 * of the destination.
1974 */
1975 if (in6_addr2scopeid(ifp, dst) != in6_addr2scopeid(oifp, dst))
1976 continue;
1977
1978 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1979 {
1980 int tlen = -1, dscopecmp, bscopecmp, matchcmp;
1981
1982 if (ifa->ifa_addr->sa_family != AF_INET6)
1983 continue;
1984
1985 src_scope = in6_addrscope(IFA_IN6(ifa));
1986
1987 /*
1988 * Don't use an address before completing DAD
1989 * nor a duplicated address.
1990 */
1991 if (((struct in6_ifaddr *)ifa)->ia6_flags &
1992 IN6_IFF_NOTREADY)
1993 continue;
1994
1995 /* XXX: is there any case to allow anycasts? */
1996 if (((struct in6_ifaddr *)ifa)->ia6_flags &
1997 IN6_IFF_ANYCAST)
1998 continue;
1999
2000 if (((struct in6_ifaddr *)ifa)->ia6_flags &
2001 IN6_IFF_DETACHED)
2002 continue;
2003
2004 /*
2005 * If this is the first address we find,
2006 * keep it anyway.
2007 */
2008 if (ifa_best == NULL)
2009 goto replace;
2010
2011 /*
2012 * ifa_best is never NULL beyond this line except
2013 * within the block labeled "replace".
2014 */
2015
2016 /*
2017 * If ifa_best has a smaller scope than dst and
2018 * the current address has a larger one than
2019 * (or equal to) dst, always replace ifa_best.
2020 * Also, if the current address has a smaller scope
2021 * than dst, ignore it unless ifa_best also has a
2022 * smaller scope.
2023 * Consequently, after the two if-clause below,
2024 * the followings must be satisfied:
2025 * (scope(src) < scope(dst) &&
2026 * scope(best) < scope(dst))
2027 * OR
2028 * (scope(best) >= scope(dst) &&
2029 * scope(src) >= scope(dst))
2030 */
2031 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2032 IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2033 goto replace; /* (A) */
2034 if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2035 IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2036 continue; /* (B) */
2037
2038 /*
2039 * A deprecated address SHOULD NOT be used in new
2040 * communications if an alternate (non-deprecated)
2041 * address is available and has sufficient scope.
2042 * RFC 2462, Section 5.5.4.
2043 */
2044 if (((struct in6_ifaddr *)ifa)->ia6_flags &
2045 IN6_IFF_DEPRECATED) {
2046 /*
2047 * Ignore any deprecated addresses if
2048 * specified by configuration.
2049 */
2050 if (!ip6_use_deprecated)
2051 continue;
2052
2053 /*
2054 * If we have already found a non-deprecated
2055 * candidate, just ignore deprecated addresses.
2056 */
2057 if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2058 == 0)
2059 continue;
2060 }
2061
2062 /*
2063 * A non-deprecated address is always preferred
2064 * to a deprecated one regardless of scopes and
2065 * address matching (Note invariants ensured by the
2066 * conditions (A) and (B) above.)
2067 */
2068 if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2069 (((struct in6_ifaddr *)ifa)->ia6_flags &
2070 IN6_IFF_DEPRECATED) == 0)
2071 goto replace;
2072
2073 /*
2074 * When we use temporary addresses described in
2075 * RFC 3041, we prefer temporary addresses to
2076 * public autoconf addresses. Again, note the
2077 * invariants from (A) and (B). Also note that we
2078 * don't have any preference between static addresses
2079 * and autoconf addresses (despite of whether or not
2080 * the latter is temporary or public.)
2081 */
2082 if (ip6_use_tempaddr) {
2083 struct in6_ifaddr *ifat;
2084
2085 ifat = (struct in6_ifaddr *)ifa;
2086 if ((ifa_best->ia6_flags &
2087 (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2088 == IN6_IFF_AUTOCONF &&
2089 (ifat->ia6_flags &
2090 (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2091 == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) {
2092 goto replace;
2093 }
2094 if ((ifa_best->ia6_flags &
2095 (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2096 == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY) &&
2097 (ifat->ia6_flags &
2098 (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2099 == IN6_IFF_AUTOCONF) {
2100 continue;
2101 }
2102 }
2103
2104 /*
2105 * At this point, we have two cases:
2106 * 1. we are looking at a non-deprecated address,
2107 * and ifa_best is also non-deprecated.
2108 * 2. we are looking at a deprecated address,
2109 * and ifa_best is also deprecated.
2110 * Also, we do not have to consider a case where
2111 * the scope of if_best is larger(smaller) than dst and
2112 * the scope of the current address is smaller(larger)
2113 * than dst. Such a case has already been covered.
2114 * Tiebreaking is done according to the following
2115 * items:
2116 * - the scope comparison between the address and
2117 * dst (dscopecmp)
2118 * - the scope comparison between the address and
2119 * ifa_best (bscopecmp)
2120 * - if the address match dst longer than ifa_best
2121 * (matchcmp)
2122 * - if the address is on the outgoing I/F (outI/F)
2123 *
2124 * Roughly speaking, the selection policy is
2125 * - the most important item is scope. The same scope
2126 * is best. Then search for a larger scope.
2127 * Smaller scopes are the last resort.
2128 * - A deprecated address is chosen only when we have
2129 * no address that has an enough scope, but is
2130 * prefered to any addresses of smaller scopes
2131 * (this must be already done above.)
2132 * - addresses on the outgoing I/F are preferred to
2133 * ones on other interfaces if none of above
2134 * tiebreaks. In the table below, the column "bI"
2135 * means if the best_ifa is on the outgoing
2136 * interface, and the column "sI" means if the ifa
2137 * is on the outgoing interface.
2138 * - If there is no other reasons to choose one,
2139 * longest address match against dst is considered.
2140 *
2141 * The precise decision table is as follows:
2142 * dscopecmp bscopecmp match bI oI | replace?
2143 * N/A equal N/A Y N | No (1)
2144 * N/A equal N/A N Y | Yes (2)
2145 * N/A equal larger N/A | Yes (3)
2146 * N/A equal !larger N/A | No (4)
2147 * larger larger N/A N/A | No (5)
2148 * larger smaller N/A N/A | Yes (6)
2149 * smaller larger N/A N/A | Yes (7)
2150 * smaller smaller N/A N/A | No (8)
2151 * equal smaller N/A N/A | Yes (9)
2152 * equal larger (already done at A above)
2153 */
2154 dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2155 bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2156
2157 if (bscopecmp == 0) {
2158 struct ifnet *bifp = ifa_best->ia_ifp;
2159
2160 if (bifp == oifp && ifp != oifp) /* (1) */
2161 continue;
2162 if (bifp != oifp && ifp == oifp) /* (2) */
2163 goto replace;
2164
2165 /*
2166 * Both bifp and ifp are on the outgoing
2167 * interface, or both two are on a different
2168 * interface from the outgoing I/F.
2169 * now we need address matching against dst
2170 * for tiebreaking.
2171 */
2172 tlen = in6_matchlen(IFA_IN6(ifa), dst);
2173 matchcmp = tlen - blen;
2174 if (matchcmp > 0) /* (3) */
2175 goto replace;
2176 continue; /* (4) */
2177 }
2178 if (dscopecmp > 0) {
2179 if (bscopecmp > 0) /* (5) */
2180 continue;
2181 goto replace; /* (6) */
2182 }
2183 if (dscopecmp < 0) {
2184 if (bscopecmp > 0) /* (7) */
2185 goto replace;
2186 continue; /* (8) */
2187 }
2188
2189 /* now dscopecmp must be 0 */
2190 if (bscopecmp < 0)
2191 goto replace; /* (9) */
2192
2193 replace:
2194 ifa_best = (struct in6_ifaddr *)ifa;
2195 blen = tlen >= 0 ? tlen :
2196 in6_matchlen(IFA_IN6(ifa), dst);
2197 best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2198 }
2199 }
2200
2201 /* count statistics for future improvements */
2202 if (ifa_best == NULL)
2203 ip6stat.ip6s_sources_none++;
2204 else {
2205 if (oifp == ifa_best->ia_ifp)
2206 ip6stat.ip6s_sources_sameif[best_scope]++;
2207 else
2208 ip6stat.ip6s_sources_otherif[best_scope]++;
2209
2210 if (best_scope == dst_scope)
2211 ip6stat.ip6s_sources_samescope[best_scope]++;
2212 else
2213 ip6stat.ip6s_sources_otherscope[best_scope]++;
2214
2215 if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) != 0)
2216 ip6stat.ip6s_sources_deprecated[best_scope]++;
2217 }
2218
2219 return(ifa_best);
2220}
2221
2222/*
2223 * return the best address out of the same scope. if no address was
2224 * found, return the first valid address from designated IF.
2225 */
2226struct in6_ifaddr *
2227in6_ifawithifp(ifp, dst)
2228 struct ifnet *ifp;
2229 struct in6_addr *dst;
2230{
2231 int dst_scope = in6_addrscope(dst), blen = -1, tlen;
2232 struct ifaddr *ifa;
2233 struct in6_ifaddr *besta = 0;
2234 struct in6_ifaddr *dep[2]; /* last-resort: deprecated */
2235
2236 dep[0] = dep[1] = NULL;
2237
2238 /*
2239 * We first look for addresses in the same scope.
2240 * If there is one, return it.
2241 * If two or more, return one which matches the dst longest.
2242 * If none, return one of global addresses assigned other ifs.
2243 */
2244 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
2245 {
2246 if (ifa->ifa_addr->sa_family != AF_INET6)
2247 continue;
2248 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2249 continue; /* XXX: is there any case to allow anycast? */
2250 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2251 continue; /* don't use this interface */
2252 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2253 continue;
2254 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2255 if (ip6_use_deprecated)
2256 dep[0] = (struct in6_ifaddr *)ifa;
2257 continue;
2258 }
2259
2260 if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2261 /*
2262 * call in6_matchlen() as few as possible
2263 */
2264 if (besta) {
2265 if (blen == -1)
2266 blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2267 tlen = in6_matchlen(IFA_IN6(ifa), dst);
2268 if (tlen > blen) {
2269 blen = tlen;
2270 besta = (struct in6_ifaddr *)ifa;
2271 }
2272 } else
2273 besta = (struct in6_ifaddr *)ifa;
2274 }
2275 }
2276 if (besta)
2277 return(besta);
2278
2279 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
2280 {
2281 if (ifa->ifa_addr->sa_family != AF_INET6)
2282 continue;
2283 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2284 continue; /* XXX: is there any case to allow anycast? */
2285 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2286 continue; /* don't use this interface */
2287 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2288 continue;
2289 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2290 if (ip6_use_deprecated)
2291 dep[1] = (struct in6_ifaddr *)ifa;
2292 continue;
2293 }
2294
2295 return (struct in6_ifaddr *)ifa;
2296 }
2297
2298 /* use the last-resort values, that are, deprecated addresses */
2299 if (dep[0])
2300 return dep[0];
2301 if (dep[1])
2302 return dep[1];
2303
2304 return NULL;
2305}
2306
2307/*
2308 * perform DAD when interface becomes IFF_UP.
2309 */
2310void
2311in6_if_up(ifp)
2312 struct ifnet *ifp;
2313{
2314 struct ifaddr *ifa;
2315 struct in6_ifaddr *ia;
2316 int dad_delay; /* delay ticks before DAD output */
2317
2318 /*
2319 * special cases, like 6to4, are handled in in6_ifattach
2320 */
2321 in6_ifattach(ifp, NULL);
2322
2323 dad_delay = 0;
2324 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
2325 {
2326 if (ifa->ifa_addr->sa_family != AF_INET6)
2327 continue;
2328 ia = (struct in6_ifaddr *)ifa;
2329 if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2330 nd6_dad_start(ifa, &dad_delay);
2331 }
2332}
2333
2334int
2335in6if_do_dad(ifp)
2336 struct ifnet *ifp;
2337{
2338 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2339 return(0);
2340
2341 switch (ifp->if_type) {
2342#ifdef IFT_DUMMY
2343 case IFT_DUMMY:
2344#endif
2345 case IFT_FAITH:
2346 /*
2347 * These interfaces do not have the IFF_LOOPBACK flag,
2348 * but loop packets back. We do not have to do DAD on such
2349 * interfaces. We should even omit it, because loop-backed
2350 * NS would confuse the DAD procedure.
2351 */
2352 return(0);
2353 default:
2354 /*
2355 * Our DAD routine requires the interface up and running.
2356 * However, some interfaces can be up before the RUNNING
2357 * status. Additionaly, users may try to assign addresses
2358 * before the interface becomes up (or running).
2359 * We simply skip DAD in such a case as a work around.
2360 * XXX: we should rather mark "tentative" on such addresses,
2361 * and do DAD after the interface becomes ready.
2362 */
2363 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2364 (IFF_UP|IFF_RUNNING))
2365 return(0);
2366
2367 return(1);
2368 }
2369}
2370
2371/*
2372 * Calculate max IPv6 MTU through all the interfaces and store it
2373 * to in6_maxmtu.
2374 */
2375void
2376in6_setmaxmtu()
2377{
2378 unsigned long maxmtu = 0;
2379 struct ifnet *ifp;
2380
2381 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2382 {
2383 if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2384 nd_ifinfo[ifp->if_index].linkmtu > maxmtu)
2385 maxmtu = nd_ifinfo[ifp->if_index].linkmtu;
2386 }
2387 if (maxmtu) /* update only when maxmtu is positive */
2388 in6_maxmtu = maxmtu;
2389}
2390
2391/*
2392 * Convert sockaddr_in6 to sockaddr_in. Original sockaddr_in6 must be
2393 * v4 mapped addr or v4 compat addr
2394 */
2395void
2396in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2397{
2398 bzero(sin, sizeof(*sin));
2399 sin->sin_len = sizeof(struct sockaddr_in);
2400 sin->sin_family = AF_INET;
2401 sin->sin_port = sin6->sin6_port;
2402 sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2403}
2404
2405/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2406void
2407in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2408{
2409 bzero(sin6, sizeof(*sin6));
2410 sin6->sin6_len = sizeof(struct sockaddr_in6);
2411 sin6->sin6_family = AF_INET6;
2412 sin6->sin6_port = sin->sin_port;
2413 sin6->sin6_addr.s6_addr32[0] = 0;
2414 sin6->sin6_addr.s6_addr32[1] = 0;
2415 sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2416 sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2417}
2418
2419/* Convert sockaddr_in6 into sockaddr_in. */
2420void
2421in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2422{
2423 struct sockaddr_in *sin_p;
2424 struct sockaddr_in6 sin6;
2425
2426 /*
2427 * Save original sockaddr_in6 addr and convert it
2428 * to sockaddr_in.
2429 */
2430 sin6 = *(struct sockaddr_in6 *)nam;
2431 sin_p = (struct sockaddr_in *)nam;
2432 in6_sin6_2_sin(sin_p, &sin6);
2433}
2434
2435/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2436void
2437in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2438{
2439 struct sockaddr_in *sin_p;
2440 struct sockaddr_in6 *sin6_p;
2441
2442 MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2443 M_WAITOK);
2444 sin_p = (struct sockaddr_in *)*nam;
2445 in6_sin_2_v4mapsin6(sin_p, sin6_p);
2446 FREE(*nam, M_SONAME);
2447 *nam = (struct sockaddr *)sin6_p;
2448}