1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
32 */
33
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD: releng/12.0/sys/netinet6/nd6_rtr.c 336676 2018-07-24 16:35:52Z andrew $");
36
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/refcount.h>
45 #include <sys/socket.h>
46 #include <sys/sockio.h>
47 #include <sys/time.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/errno.h>
51 #include <sys/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_types.h>
59 #include <net/if_dl.h>
60 #include <net/route.h>
61 #include <net/route_var.h>
62 #include <net/radix.h>
63 #include <net/vnet.h>
64
65 #include <netinet/in.h>
66 #include <net/if_llatbl.h>
67 #include <netinet6/in6_var.h>
68 #include <netinet6/in6_ifattach.h>
69 #include <netinet/ip6.h>
70 #include <netinet6/ip6_var.h>
71 #include <netinet6/nd6.h>
72 #include <netinet/icmp6.h>
73 #include <netinet6/scope6_var.h>
74
75 static int rtpref(struct nd_defrouter *);
76 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
77 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
78 struct mbuf *, int);
79 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
80 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
81 struct nd_defrouter *);
82 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
83 static void pfxrtr_del(struct nd_pfxrouter *);
84 static struct nd_pfxrouter *find_pfxlist_reachable_router(struct nd_prefix *);
85 static void defrouter_delreq(struct nd_defrouter *);
86 static void nd6_rtmsg(int, struct rtentry *);
87
88 static int in6_init_prefix_ltimes(struct nd_prefix *);
89 static void in6_init_address_ltimes(struct nd_prefix *,
90 struct in6_addrlifetime *);
91
92 static int rt6_deleteroute(const struct rtentry *, void *);
93
94 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
95 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
96
97 VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp);
98 VNET_DEFINE(int, nd6_defifindex);
99 #define V_nd6_defifp VNET(nd6_defifp)
100
101 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
102
103 VNET_DEFINE(int, ip6_desync_factor);
104 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
105 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
106
107 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
108
109 /* RTPREF_MEDIUM has to be 0! */
110 #define RTPREF_HIGH 1
111 #define RTPREF_MEDIUM 0
112 #define RTPREF_LOW (-1)
113 #define RTPREF_RESERVED (-2)
114 #define RTPREF_INVALID (-3) /* internal */
115
116 /*
117 * Receive Router Solicitation Message - just for routers.
118 * Router solicitation/advertisement is mostly managed by userland program
119 * (rtadvd) so here we have no function like nd6_ra_output().
120 *
121 * Based on RFC 2461
122 */
123 void
124 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
125 {
126 struct ifnet *ifp = m->m_pkthdr.rcvif;
127 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
128 struct nd_router_solicit *nd_rs;
129 struct in6_addr saddr6 = ip6->ip6_src;
130 char *lladdr = NULL;
131 int lladdrlen = 0;
132 union nd_opts ndopts;
133 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
134
135 /*
136 * Accept RS only when V_ip6_forwarding=1 and the interface has
137 * no ND6_IFF_ACCEPT_RTADV.
138 */
139 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
140 goto freeit;
141
142 /* RFC 6980: Nodes MUST silently ignore fragments */
143 if(m->m_flags & M_FRAGMENTED)
144 goto freeit;
145
146 /* Sanity checks */
147 if (ip6->ip6_hlim != 255) {
148 nd6log((LOG_ERR,
149 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
150 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
151 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
152 goto bad;
153 }
154
155 /*
156 * Don't update the neighbor cache, if src = ::.
157 * This indicates that the src has no IP address assigned yet.
158 */
159 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
160 goto freeit;
161
162 #ifndef PULLDOWN_TEST
163 IP6_EXTHDR_CHECK(m, off, icmp6len,);
164 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
165 #else
166 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
167 if (nd_rs == NULL) {
168 ICMP6STAT_INC(icp6s_tooshort);
169 return;
170 }
171 #endif
172
173 icmp6len -= sizeof(*nd_rs);
174 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
175 if (nd6_options(&ndopts) < 0) {
176 nd6log((LOG_INFO,
177 "nd6_rs_input: invalid ND option, ignored\n"));
178 /* nd6_options have incremented stats */
179 goto freeit;
180 }
181
182 if (ndopts.nd_opts_src_lladdr) {
183 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
184 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
185 }
186
187 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
188 nd6log((LOG_INFO,
189 "nd6_rs_input: lladdrlen mismatch for %s "
190 "(if %d, RS packet %d)\n",
191 ip6_sprintf(ip6bufs, &saddr6),
192 ifp->if_addrlen, lladdrlen - 2));
193 goto bad;
194 }
195
196 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
197
198 freeit:
199 m_freem(m);
200 return;
201
202 bad:
203 ICMP6STAT_INC(icp6s_badrs);
204 m_freem(m);
205 }
206
207 /*
208 * Receive Router Advertisement Message.
209 *
210 * Based on RFC 2461
211 * TODO: on-link bit on prefix information
212 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
213 */
214 void
215 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
216 {
217 struct ifnet *ifp = m->m_pkthdr.rcvif;
218 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
219 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
220 struct nd_router_advert *nd_ra;
221 struct in6_addr saddr6 = ip6->ip6_src;
222 int mcast = 0;
223 union nd_opts ndopts;
224 struct nd_defrouter *dr;
225 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
226
227 dr = NULL;
228
229 /*
230 * We only accept RAs only when the per-interface flag
231 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
232 */
233 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
234 goto freeit;
235
236 /* RFC 6980: Nodes MUST silently ignore fragments */
237 if(m->m_flags & M_FRAGMENTED)
238 goto freeit;
239
240 if (ip6->ip6_hlim != 255) {
241 nd6log((LOG_ERR,
242 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
243 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
244 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
245 goto bad;
246 }
247
248 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
249 nd6log((LOG_ERR,
250 "nd6_ra_input: src %s is not link-local\n",
251 ip6_sprintf(ip6bufs, &saddr6)));
252 goto bad;
253 }
254
255 #ifndef PULLDOWN_TEST
256 IP6_EXTHDR_CHECK(m, off, icmp6len,);
257 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
258 #else
259 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
260 if (nd_ra == NULL) {
261 ICMP6STAT_INC(icp6s_tooshort);
262 return;
263 }
264 #endif
265
266 icmp6len -= sizeof(*nd_ra);
267 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
268 if (nd6_options(&ndopts) < 0) {
269 nd6log((LOG_INFO,
270 "nd6_ra_input: invalid ND option, ignored\n"));
271 /* nd6_options have incremented stats */
272 goto freeit;
273 }
274
275 {
276 struct nd_defrouter dr0;
277 u_int32_t advreachable = nd_ra->nd_ra_reachable;
278
279 /* remember if this is a multicasted advertisement */
280 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
281 mcast = 1;
282
283 bzero(&dr0, sizeof(dr0));
284 dr0.rtaddr = saddr6;
285 dr0.raflags = nd_ra->nd_ra_flags_reserved;
286 /*
287 * Effectively-disable routes from RA messages when
288 * ND6_IFF_NO_RADR enabled on the receiving interface or
289 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
290 */
291 if (ndi->flags & ND6_IFF_NO_RADR)
292 dr0.rtlifetime = 0;
293 else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
294 dr0.rtlifetime = 0;
295 else
296 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
297 dr0.expire = time_uptime + dr0.rtlifetime;
298 dr0.ifp = ifp;
299 /* unspecified or not? (RFC 2461 6.3.4) */
300 if (advreachable) {
301 advreachable = ntohl(advreachable);
302 if (advreachable <= MAX_REACHABLE_TIME &&
303 ndi->basereachable != advreachable) {
304 ndi->basereachable = advreachable;
305 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
306 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
307 }
308 }
309 if (nd_ra->nd_ra_retransmit)
310 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
311 if (nd_ra->nd_ra_curhoplimit) {
312 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
313 ndi->chlim = nd_ra->nd_ra_curhoplimit;
314 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
315 log(LOG_ERR, "RA with a lower CurHopLimit sent from "
316 "%s on %s (current = %d, received = %d). "
317 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
318 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
319 }
320 }
321 dr = defrtrlist_update(&dr0);
322 }
323
324 /*
325 * prefix
326 */
327 if (ndopts.nd_opts_pi) {
328 struct nd_opt_hdr *pt;
329 struct nd_opt_prefix_info *pi = NULL;
330 struct nd_prefixctl pr;
331
332 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
333 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
334 pt = (struct nd_opt_hdr *)((caddr_t)pt +
335 (pt->nd_opt_len << 3))) {
336 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
337 continue;
338 pi = (struct nd_opt_prefix_info *)pt;
339
340 if (pi->nd_opt_pi_len != 4) {
341 nd6log((LOG_INFO,
342 "nd6_ra_input: invalid option "
343 "len %d for prefix information option, "
344 "ignored\n", pi->nd_opt_pi_len));
345 continue;
346 }
347
348 if (128 < pi->nd_opt_pi_prefix_len) {
349 nd6log((LOG_INFO,
350 "nd6_ra_input: invalid prefix "
351 "len %d for prefix information option, "
352 "ignored\n", pi->nd_opt_pi_prefix_len));
353 continue;
354 }
355
356 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
357 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
358 nd6log((LOG_INFO,
359 "nd6_ra_input: invalid prefix "
360 "%s, ignored\n",
361 ip6_sprintf(ip6bufs,
362 &pi->nd_opt_pi_prefix)));
363 continue;
364 }
365
366 bzero(&pr, sizeof(pr));
367 pr.ndpr_prefix.sin6_family = AF_INET6;
368 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
369 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
370 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
371
372 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
373 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
374 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
375 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
376 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
377 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
378 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
379 (void)prelist_update(&pr, dr, m, mcast);
380 }
381 }
382 if (dr != NULL) {
383 defrouter_rele(dr);
384 dr = NULL;
385 }
386
387 /*
388 * MTU
389 */
390 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
391 u_long mtu;
392 u_long maxmtu;
393
394 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
395
396 /* lower bound */
397 if (mtu < IPV6_MMTU) {
398 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
399 "mtu=%lu sent from %s, ignoring\n",
400 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
401 goto skip;
402 }
403
404 /* upper bound */
405 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
406 ? ndi->maxmtu : ifp->if_mtu;
407 if (mtu <= maxmtu) {
408 int change = (ndi->linkmtu != mtu);
409
410 ndi->linkmtu = mtu;
411 if (change) {
412 /* in6_maxmtu may change */
413 in6_setmaxmtu();
414 rt_updatemtu(ifp);
415 }
416 } else {
417 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
418 "mtu=%lu sent from %s; "
419 "exceeds maxmtu %lu, ignoring\n",
420 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
421 }
422 }
423
424 skip:
425
426 /*
427 * Source link layer address
428 */
429 {
430 char *lladdr = NULL;
431 int lladdrlen = 0;
432
433 if (ndopts.nd_opts_src_lladdr) {
434 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
435 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
436 }
437
438 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
439 nd6log((LOG_INFO,
440 "nd6_ra_input: lladdrlen mismatch for %s "
441 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
442 ifp->if_addrlen, lladdrlen - 2));
443 goto bad;
444 }
445
446 nd6_cache_lladdr(ifp, &saddr6, lladdr,
447 lladdrlen, ND_ROUTER_ADVERT, 0);
448
449 /*
450 * Installing a link-layer address might change the state of the
451 * router's neighbor cache, which might also affect our on-link
452 * detection of adveritsed prefixes.
453 */
454 pfxlist_onlink_check();
455 }
456
457 freeit:
458 m_freem(m);
459 return;
460
461 bad:
462 ICMP6STAT_INC(icp6s_badra);
463 m_freem(m);
464 }
465
466 /* tell the change to user processes watching the routing socket. */
467 static void
468 nd6_rtmsg(int cmd, struct rtentry *rt)
469 {
470 struct rt_addrinfo info;
471 struct ifnet *ifp;
472 struct ifaddr *ifa;
473
474 bzero((caddr_t)&info, sizeof(info));
475 info.rti_info[RTAX_DST] = rt_key(rt);
476 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
477 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
478 ifp = rt->rt_ifp;
479 if (ifp != NULL) {
480 IF_ADDR_RLOCK(ifp);
481 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
482 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
483 ifa_ref(ifa);
484 IF_ADDR_RUNLOCK(ifp);
485 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
486 } else
487 ifa = NULL;
488
489 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
490 if (ifa != NULL)
491 ifa_free(ifa);
492 }
493
494 /*
495 * default router list processing sub routines
496 */
497
498 static void
499 defrouter_addreq(struct nd_defrouter *new)
500 {
501 struct sockaddr_in6 def, mask, gate;
502 struct rtentry *newrt = NULL;
503 int error;
504
505 bzero(&def, sizeof(def));
506 bzero(&mask, sizeof(mask));
507 bzero(&gate, sizeof(gate));
508
509 def.sin6_len = mask.sin6_len = gate.sin6_len =
510 sizeof(struct sockaddr_in6);
511 def.sin6_family = gate.sin6_family = AF_INET6;
512 gate.sin6_addr = new->rtaddr;
513
514 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
515 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
516 RTF_GATEWAY, &newrt, new->ifp->if_fib);
517 if (newrt) {
518 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
519 RTFREE(newrt);
520 }
521 if (error == 0)
522 new->installed = 1;
523 }
524
525 struct nd_defrouter *
526 defrouter_lookup_locked(struct in6_addr *addr, struct ifnet *ifp)
527 {
528 struct nd_defrouter *dr;
529
530 ND6_LOCK_ASSERT();
531 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
532 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
533 defrouter_ref(dr);
534 return (dr);
535 }
536 return (NULL);
537 }
538
539 struct nd_defrouter *
540 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
541 {
542 struct nd_defrouter *dr;
543
544 ND6_RLOCK();
545 dr = defrouter_lookup_locked(addr, ifp);
546 ND6_RUNLOCK();
547 return (dr);
548 }
549
550 void
551 defrouter_ref(struct nd_defrouter *dr)
552 {
553
554 refcount_acquire(&dr->refcnt);
555 }
556
557 void
558 defrouter_rele(struct nd_defrouter *dr)
559 {
560
561 if (refcount_release(&dr->refcnt))
562 free(dr, M_IP6NDP);
563 }
564
565 /*
566 * Remove the default route for a given router.
567 * This is just a subroutine function for defrouter_select_fib(), and
568 * should not be called from anywhere else.
569 */
570 static void
571 defrouter_delreq(struct nd_defrouter *dr)
572 {
573 struct sockaddr_in6 def, mask, gate;
574 struct rtentry *oldrt = NULL;
575
576 bzero(&def, sizeof(def));
577 bzero(&mask, sizeof(mask));
578 bzero(&gate, sizeof(gate));
579
580 def.sin6_len = mask.sin6_len = gate.sin6_len =
581 sizeof(struct sockaddr_in6);
582 def.sin6_family = gate.sin6_family = AF_INET6;
583 gate.sin6_addr = dr->rtaddr;
584
585 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
586 (struct sockaddr *)&gate,
587 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, dr->ifp->if_fib);
588 if (oldrt) {
589 nd6_rtmsg(RTM_DELETE, oldrt);
590 RTFREE(oldrt);
591 }
592
593 dr->installed = 0;
594 }
595
596 /*
597 * Remove all default routes from default router list.
598 */
599 void
600 defrouter_reset(void)
601 {
602 struct nd_defrouter *dr, **dra;
603 int count, i;
604
605 count = i = 0;
606
607 /*
608 * We can't delete routes with the ND lock held, so make a copy of the
609 * current default router list and use that when deleting routes.
610 */
611 ND6_RLOCK();
612 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
613 count++;
614 ND6_RUNLOCK();
615
616 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
617
618 ND6_RLOCK();
619 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
620 if (i == count)
621 break;
622 defrouter_ref(dr);
623 dra[i++] = dr;
624 }
625 ND6_RUNLOCK();
626
627 for (i = 0; i < count && dra[i] != NULL; i++) {
628 defrouter_delreq(dra[i]);
629 defrouter_rele(dra[i]);
630 }
631 free(dra, M_TEMP);
632
633 /*
634 * XXX should we also nuke any default routers in the kernel, by
635 * going through them by rtalloc1()?
636 */
637 }
638
639 /*
640 * Look up a matching default router list entry and remove it. Returns true if a
641 * matching entry was found, false otherwise.
642 */
643 bool
644 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
645 {
646 struct nd_defrouter *dr;
647
648 ND6_WLOCK();
649 dr = defrouter_lookup_locked(addr, ifp);
650 if (dr == NULL) {
651 ND6_WUNLOCK();
652 return (false);
653 }
654
655 defrouter_unlink(dr, NULL);
656 ND6_WUNLOCK();
657 defrouter_del(dr);
658 defrouter_rele(dr);
659 return (true);
660 }
661
662 /*
663 * Remove a router from the global list and optionally stash it in a
664 * caller-supplied queue.
665 *
666 * The ND lock must be held.
667 */
668 void
669 defrouter_unlink(struct nd_defrouter *dr, struct nd_drhead *drq)
670 {
671
672 ND6_WLOCK_ASSERT();
673 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
674 V_nd6_list_genid++;
675 if (drq != NULL)
676 TAILQ_INSERT_TAIL(drq, dr, dr_entry);
677 }
678
679 void
680 defrouter_del(struct nd_defrouter *dr)
681 {
682 struct nd_defrouter *deldr = NULL;
683 struct nd_prefix *pr;
684 struct nd_pfxrouter *pfxrtr;
685
686 ND6_UNLOCK_ASSERT();
687
688 /*
689 * Flush all the routing table entries that use the router
690 * as a next hop.
691 */
692 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
693 rt6_flush(&dr->rtaddr, dr->ifp);
694
695 if (dr->installed) {
696 deldr = dr;
697 defrouter_delreq(dr);
698 }
699
700 /*
701 * Also delete all the pointers to the router in each prefix lists.
702 */
703 ND6_WLOCK();
704 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
705 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
706 pfxrtr_del(pfxrtr);
707 }
708 ND6_WUNLOCK();
709
710 pfxlist_onlink_check();
711
712 /*
713 * If the router is the primary one, choose a new one.
714 * Note that defrouter_select_fib() will remove the current
715 * gateway from the routing table.
716 */
717 if (deldr)
718 defrouter_select_fib(deldr->ifp->if_fib);
719
720 /*
721 * Release the list reference.
722 */
723 defrouter_rele(dr);
724 }
725
726 /*
727 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
728 * draft-ietf-ipngwg-router-selection:
729 * 1) Routers that are reachable or probably reachable should be preferred.
730 * If we have more than one (probably) reachable router, prefer ones
731 * with the highest router preference.
732 * 2) When no routers on the list are known to be reachable or
733 * probably reachable, routers SHOULD be selected in a round-robin
734 * fashion, regardless of router preference values.
735 * 3) If the Default Router List is empty, assume that all
736 * destinations are on-link.
737 *
738 * We assume nd_defrouter is sorted by router preference value.
739 * Since the code below covers both with and without router preference cases,
740 * we do not need to classify the cases by ifdef.
741 *
742 * At this moment, we do not try to install more than one default router,
743 * even when the multipath routing is available, because we're not sure about
744 * the benefits for stub hosts comparing to the risk of making the code
745 * complicated and the possibility of introducing bugs.
746 *
747 * We maintain a single list of routers for multiple FIBs, only considering one
748 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
749 * we do the whole thing multiple times.
750 */
751 void
752 defrouter_select_fib(int fibnum)
753 {
754 struct nd_defrouter *dr, *selected_dr, *installed_dr;
755 struct llentry *ln = NULL;
756
757 if (fibnum == RT_ALL_FIBS) {
758 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
759 defrouter_select_fib(fibnum);
760 }
761 }
762
763 ND6_RLOCK();
764 /*
765 * Let's handle easy case (3) first:
766 * If default router list is empty, there's nothing to be done.
767 */
768 if (TAILQ_EMPTY(&V_nd_defrouter)) {
769 ND6_RUNLOCK();
770 return;
771 }
772
773 /*
774 * Search for a (probably) reachable router from the list.
775 * We just pick up the first reachable one (if any), assuming that
776 * the ordering rule of the list described in defrtrlist_update().
777 */
778 selected_dr = installed_dr = NULL;
779 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
780 IF_AFDATA_RLOCK(dr->ifp);
781 if (selected_dr == NULL && dr->ifp->if_fib == fibnum &&
782 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
783 ND6_IS_LLINFO_PROBREACH(ln)) {
784 selected_dr = dr;
785 defrouter_ref(selected_dr);
786 }
787 IF_AFDATA_RUNLOCK(dr->ifp);
788 if (ln != NULL) {
789 LLE_RUNLOCK(ln);
790 ln = NULL;
791 }
792
793 if (dr->installed && dr->ifp->if_fib == fibnum) {
794 if (installed_dr == NULL) {
795 installed_dr = dr;
796 defrouter_ref(installed_dr);
797 } else {
798 /*
799 * this should not happen.
800 * warn for diagnosis.
801 */
802 log(LOG_ERR, "defrouter_select_fib: more than "
803 "one router is installed\n");
804 }
805 }
806 }
807 /*
808 * If none of the default routers was found to be reachable,
809 * round-robin the list regardless of preference.
810 * Otherwise, if we have an installed router, check if the selected
811 * (reachable) router should really be preferred to the installed one.
812 * We only prefer the new router when the old one is not reachable
813 * or when the new one has a really higher preference value.
814 */
815 if (selected_dr == NULL) {
816 if (installed_dr == NULL ||
817 TAILQ_NEXT(installed_dr, dr_entry) == NULL)
818 dr = TAILQ_FIRST(&V_nd_defrouter);
819 else
820 dr = TAILQ_NEXT(installed_dr, dr_entry);
821
822 /* Ensure we select a router for this FIB. */
823 TAILQ_FOREACH_FROM(dr, &V_nd_defrouter, dr_entry) {
824 if (dr->ifp->if_fib == fibnum) {
825 selected_dr = dr;
826 defrouter_ref(selected_dr);
827 break;
828 }
829 }
830 } else if (installed_dr != NULL) {
831 IF_AFDATA_RLOCK(installed_dr->ifp);
832 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0,
833 installed_dr->ifp)) &&
834 ND6_IS_LLINFO_PROBREACH(ln) &&
835 installed_dr->ifp->if_fib == fibnum &&
836 rtpref(selected_dr) <= rtpref(installed_dr)) {
837 defrouter_rele(selected_dr);
838 selected_dr = installed_dr;
839 }
840 IF_AFDATA_RUNLOCK(installed_dr->ifp);
841 if (ln != NULL)
842 LLE_RUNLOCK(ln);
843 }
844 ND6_RUNLOCK();
845
846 /*
847 * If we selected a router for this FIB and it's different
848 * than the installed one, remove the installed router and
849 * install the selected one in its place.
850 */
851 if (installed_dr != selected_dr) {
852 if (installed_dr != NULL) {
853 defrouter_delreq(installed_dr);
854 defrouter_rele(installed_dr);
855 }
856 if (selected_dr != NULL)
857 defrouter_addreq(selected_dr);
858 }
859 if (selected_dr != NULL)
860 defrouter_rele(selected_dr);
861 }
862
863 /*
864 * Maintain old KPI for default router selection.
865 * If unspecified, we can re-select routers for all FIBs.
866 */
867 void
868 defrouter_select(void)
869 {
870 defrouter_select_fib(RT_ALL_FIBS);
871 }
872
873 /*
874 * for default router selection
875 * regards router-preference field as a 2-bit signed integer
876 */
877 static int
878 rtpref(struct nd_defrouter *dr)
879 {
880 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
881 case ND_RA_FLAG_RTPREF_HIGH:
882 return (RTPREF_HIGH);
883 case ND_RA_FLAG_RTPREF_MEDIUM:
884 case ND_RA_FLAG_RTPREF_RSV:
885 return (RTPREF_MEDIUM);
886 case ND_RA_FLAG_RTPREF_LOW:
887 return (RTPREF_LOW);
888 default:
889 /*
890 * This case should never happen. If it did, it would mean a
891 * serious bug of kernel internal. We thus always bark here.
892 * Or, can we even panic?
893 */
894 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
895 return (RTPREF_INVALID);
896 }
897 /* NOTREACHED */
898 }
899
900 static struct nd_defrouter *
901 defrtrlist_update(struct nd_defrouter *new)
902 {
903 struct nd_defrouter *dr, *n;
904 uint64_t genid;
905 int oldpref;
906 bool writelocked;
907
908 if (new->rtlifetime == 0) {
909 defrouter_remove(&new->rtaddr, new->ifp);
910 return (NULL);
911 }
912
913 ND6_RLOCK();
914 writelocked = false;
915 restart:
916 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
917 if (dr != NULL) {
918 oldpref = rtpref(dr);
919
920 /* override */
921 dr->raflags = new->raflags; /* XXX flag check */
922 dr->rtlifetime = new->rtlifetime;
923 dr->expire = new->expire;
924
925 /*
926 * If the preference does not change, there's no need
927 * to sort the entries. Also make sure the selected
928 * router is still installed in the kernel.
929 */
930 if (dr->installed && rtpref(new) == oldpref) {
931 if (writelocked)
932 ND6_WUNLOCK();
933 else
934 ND6_RUNLOCK();
935 return (dr);
936 }
937 }
938
939 /*
940 * The router needs to be reinserted into the default router
941 * list, so upgrade to a write lock. If that fails and the list
942 * has potentially changed while the lock was dropped, we'll
943 * redo the lookup with the write lock held.
944 */
945 if (!writelocked) {
946 writelocked = true;
947 if (!ND6_TRY_UPGRADE()) {
948 genid = V_nd6_list_genid;
949 ND6_RUNLOCK();
950 ND6_WLOCK();
951 if (genid != V_nd6_list_genid)
952 goto restart;
953 }
954 }
955
956 if (dr != NULL) {
957 /*
958 * The preferred router may have changed, so relocate this
959 * router.
960 */
961 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
962 n = dr;
963 } else {
964 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
965 if (n == NULL) {
966 ND6_WUNLOCK();
967 return (NULL);
968 }
969 memcpy(n, new, sizeof(*n));
970 /* Initialize with an extra reference for the caller. */
971 refcount_init(&n->refcnt, 2);
972 }
973
974 /*
975 * Insert the new router in the Default Router List;
976 * The Default Router List should be in the descending order
977 * of router-preferece. Routers with the same preference are
978 * sorted in the arriving time order.
979 */
980
981 /* insert at the end of the group */
982 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
983 if (rtpref(n) > rtpref(dr))
984 break;
985 }
986 if (dr != NULL)
987 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
988 else
989 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
990 V_nd6_list_genid++;
991 ND6_WUNLOCK();
992
993 defrouter_select_fib(new->ifp->if_fib);
994
995 return (n);
996 }
997
998 static struct nd_pfxrouter *
999 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
1000 {
1001 struct nd_pfxrouter *search;
1002
1003 ND6_LOCK_ASSERT();
1004
1005 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
1006 if (search->router == dr)
1007 break;
1008 }
1009 return (search);
1010 }
1011
1012 static void
1013 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
1014 {
1015 struct nd_pfxrouter *new;
1016 bool update;
1017
1018 ND6_UNLOCK_ASSERT();
1019
1020 ND6_RLOCK();
1021 if (pfxrtr_lookup(pr, dr) != NULL) {
1022 ND6_RUNLOCK();
1023 return;
1024 }
1025 ND6_RUNLOCK();
1026
1027 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1028 if (new == NULL)
1029 return;
1030 defrouter_ref(dr);
1031 new->router = dr;
1032
1033 ND6_WLOCK();
1034 if (pfxrtr_lookup(pr, dr) == NULL) {
1035 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
1036 update = true;
1037 } else {
1038 /* We lost a race to add the reference. */
1039 defrouter_rele(dr);
1040 free(new, M_IP6NDP);
1041 update = false;
1042 }
1043 ND6_WUNLOCK();
1044
1045 if (update)
1046 pfxlist_onlink_check();
1047 }
1048
1049 static void
1050 pfxrtr_del(struct nd_pfxrouter *pfr)
1051 {
1052
1053 ND6_WLOCK_ASSERT();
1054
1055 LIST_REMOVE(pfr, pfr_entry);
1056 defrouter_rele(pfr->router);
1057 free(pfr, M_IP6NDP);
1058 }
1059
1060 static struct nd_prefix *
1061 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1062 {
1063 struct nd_prefix *search;
1064
1065 ND6_LOCK_ASSERT();
1066
1067 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1068 if (key->ndpr_ifp == search->ndpr_ifp &&
1069 key->ndpr_plen == search->ndpr_plen &&
1070 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1071 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1072 nd6_prefix_ref(search);
1073 break;
1074 }
1075 }
1076 return (search);
1077 }
1078
1079 struct nd_prefix *
1080 nd6_prefix_lookup(struct nd_prefixctl *key)
1081 {
1082 struct nd_prefix *search;
1083
1084 ND6_RLOCK();
1085 search = nd6_prefix_lookup_locked(key);
1086 ND6_RUNLOCK();
1087 return (search);
1088 }
1089
1090 void
1091 nd6_prefix_ref(struct nd_prefix *pr)
1092 {
1093
1094 refcount_acquire(&pr->ndpr_refcnt);
1095 }
1096
1097 void
1098 nd6_prefix_rele(struct nd_prefix *pr)
1099 {
1100
1101 if (refcount_release(&pr->ndpr_refcnt)) {
1102 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1103 ("prefix %p has advertising routers", pr));
1104 free(pr, M_IP6NDP);
1105 }
1106 }
1107
1108 int
1109 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1110 struct nd_prefix **newp)
1111 {
1112 struct nd_prefix *new;
1113 char ip6buf[INET6_ADDRSTRLEN];
1114 int error;
1115
1116 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1117 if (new == NULL)
1118 return (ENOMEM);
1119 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1120 new->ndpr_ifp = pr->ndpr_ifp;
1121 new->ndpr_prefix = pr->ndpr_prefix;
1122 new->ndpr_plen = pr->ndpr_plen;
1123 new->ndpr_vltime = pr->ndpr_vltime;
1124 new->ndpr_pltime = pr->ndpr_pltime;
1125 new->ndpr_flags = pr->ndpr_flags;
1126 if ((error = in6_init_prefix_ltimes(new)) != 0) {
1127 free(new, M_IP6NDP);
1128 return (error);
1129 }
1130 new->ndpr_lastupdate = time_uptime;
1131
1132 /* initialization */
1133 LIST_INIT(&new->ndpr_advrtrs);
1134 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1135 /* make prefix in the canonical form */
1136 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1137
1138 ND6_WLOCK();
1139 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1140 V_nd6_list_genid++;
1141 ND6_WUNLOCK();
1142
1143 /* ND_OPT_PI_FLAG_ONLINK processing */
1144 if (new->ndpr_raf_onlink) {
1145 ND6_ONLINK_LOCK();
1146 if ((error = nd6_prefix_onlink(new)) != 0) {
1147 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
1148 "the prefix %s/%d on-link on %s (errno=%d)\n",
1149 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1150 pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1151 /* proceed anyway. XXX: is it correct? */
1152 }
1153 ND6_ONLINK_UNLOCK();
1154 }
1155
1156 if (dr != NULL)
1157 pfxrtr_add(new, dr);
1158 if (newp != NULL)
1159 *newp = new;
1160 return (0);
1161 }
1162
1163 /*
1164 * Remove a prefix from the prefix list and optionally stash it in a
1165 * caller-provided list.
1166 *
1167 * The ND6 lock must be held.
1168 */
1169 void
1170 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1171 {
1172
1173 ND6_WLOCK_ASSERT();
1174
1175 LIST_REMOVE(pr, ndpr_entry);
1176 V_nd6_list_genid++;
1177 if (list != NULL)
1178 LIST_INSERT_HEAD(list, pr, ndpr_entry);
1179 }
1180
1181 /*
1182 * Free an unlinked prefix, first marking it off-link if necessary.
1183 */
1184 void
1185 nd6_prefix_del(struct nd_prefix *pr)
1186 {
1187 struct nd_pfxrouter *pfr, *next;
1188 int e;
1189 char ip6buf[INET6_ADDRSTRLEN];
1190
1191 KASSERT(pr->ndpr_addrcnt == 0,
1192 ("prefix %p has referencing addresses", pr));
1193 ND6_UNLOCK_ASSERT();
1194
1195 /*
1196 * Though these flags are now meaningless, we'd rather keep the value
1197 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1198 * when executing "ndp -p".
1199 */
1200 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1201 ND6_ONLINK_LOCK();
1202 if ((e = nd6_prefix_offlink(pr)) != 0) {
1203 nd6log((LOG_ERR,
1204 "nd6_prefix_del: failed to make %s/%d offlink "
1205 "on %s, errno=%d\n",
1206 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1207 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1208 /* what should we do? */
1209 }
1210 ND6_ONLINK_UNLOCK();
1211 }
1212
1213 /* Release references to routers that have advertised this prefix. */
1214 ND6_WLOCK();
1215 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1216 pfxrtr_del(pfr);
1217 ND6_WUNLOCK();
1218
1219 nd6_prefix_rele(pr);
1220
1221 pfxlist_onlink_check();
1222 }
1223
1224 static int
1225 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1226 struct mbuf *m, int mcast)
1227 {
1228 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1229 struct ifaddr *ifa;
1230 struct ifnet *ifp = new->ndpr_ifp;
1231 struct nd_prefix *pr;
1232 int error = 0;
1233 int auth;
1234 struct in6_addrlifetime lt6_tmp;
1235 char ip6buf[INET6_ADDRSTRLEN];
1236
1237 auth = 0;
1238 if (m) {
1239 /*
1240 * Authenticity for NA consists authentication for
1241 * both IP header and IP datagrams, doesn't it ?
1242 */
1243 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1244 auth = ((m->m_flags & M_AUTHIPHDR) &&
1245 (m->m_flags & M_AUTHIPDGM));
1246 #endif
1247 }
1248
1249 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1250 /*
1251 * nd6_prefix_lookup() ensures that pr and new have the same
1252 * prefix on a same interface.
1253 */
1254
1255 /*
1256 * Update prefix information. Note that the on-link (L) bit
1257 * and the autonomous (A) bit should NOT be changed from 1
1258 * to 0.
1259 */
1260 if (new->ndpr_raf_onlink == 1)
1261 pr->ndpr_raf_onlink = 1;
1262 if (new->ndpr_raf_auto == 1)
1263 pr->ndpr_raf_auto = 1;
1264 if (new->ndpr_raf_onlink) {
1265 pr->ndpr_vltime = new->ndpr_vltime;
1266 pr->ndpr_pltime = new->ndpr_pltime;
1267 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1268 pr->ndpr_lastupdate = time_uptime;
1269 }
1270
1271 if (new->ndpr_raf_onlink &&
1272 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1273 ND6_ONLINK_LOCK();
1274 if ((error = nd6_prefix_onlink(pr)) != 0) {
1275 nd6log((LOG_ERR,
1276 "prelist_update: failed to make "
1277 "the prefix %s/%d on-link on %s "
1278 "(errno=%d)\n",
1279 ip6_sprintf(ip6buf,
1280 &pr->ndpr_prefix.sin6_addr),
1281 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1282 error));
1283 /* proceed anyway. XXX: is it correct? */
1284 }
1285 ND6_ONLINK_UNLOCK();
1286 }
1287
1288 if (dr != NULL)
1289 pfxrtr_add(pr, dr);
1290 } else {
1291 if (new->ndpr_vltime == 0)
1292 goto end;
1293 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1294 goto end;
1295
1296 error = nd6_prelist_add(new, dr, &pr);
1297 if (error != 0) {
1298 nd6log((LOG_NOTICE, "prelist_update: "
1299 "nd6_prelist_add failed for %s/%d on %s errno=%d\n",
1300 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1301 new->ndpr_plen, if_name(new->ndpr_ifp), error));
1302 goto end; /* we should just give up in this case. */
1303 }
1304
1305 /*
1306 * XXX: from the ND point of view, we can ignore a prefix
1307 * with the on-link bit being zero. However, we need a
1308 * prefix structure for references from autoconfigured
1309 * addresses. Thus, we explicitly make sure that the prefix
1310 * itself expires now.
1311 */
1312 if (pr->ndpr_raf_onlink == 0) {
1313 pr->ndpr_vltime = 0;
1314 pr->ndpr_pltime = 0;
1315 in6_init_prefix_ltimes(pr);
1316 }
1317 }
1318
1319 /*
1320 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1321 * Note that pr must be non NULL at this point.
1322 */
1323
1324 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1325 if (!new->ndpr_raf_auto)
1326 goto end;
1327
1328 /*
1329 * 5.5.3 (b). the link-local prefix should have been ignored in
1330 * nd6_ra_input.
1331 */
1332
1333 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1334 if (new->ndpr_pltime > new->ndpr_vltime) {
1335 error = EINVAL; /* XXX: won't be used */
1336 goto end;
1337 }
1338
1339 /*
1340 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1341 * an address configured by stateless autoconfiguration already in the
1342 * list of addresses associated with the interface, and the Valid
1343 * Lifetime is not 0, form an address. We first check if we have
1344 * a matching prefix.
1345 * Note: we apply a clarification in rfc2462bis-02 here. We only
1346 * consider autoconfigured addresses while RFC2462 simply said
1347 * "address".
1348 */
1349 IF_ADDR_RLOCK(ifp);
1350 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1351 struct in6_ifaddr *ifa6;
1352 u_int32_t remaininglifetime;
1353
1354 if (ifa->ifa_addr->sa_family != AF_INET6)
1355 continue;
1356
1357 ifa6 = (struct in6_ifaddr *)ifa;
1358
1359 /*
1360 * We only consider autoconfigured addresses as per rfc2462bis.
1361 */
1362 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1363 continue;
1364
1365 /*
1366 * Spec is not clear here, but I believe we should concentrate
1367 * on unicast (i.e. not anycast) addresses.
1368 * XXX: other ia6_flags? detached or duplicated?
1369 */
1370 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1371 continue;
1372
1373 /*
1374 * Ignore the address if it is not associated with a prefix
1375 * or is associated with a prefix that is different from this
1376 * one. (pr is never NULL here)
1377 */
1378 if (ifa6->ia6_ndpr != pr)
1379 continue;
1380
1381 if (ia6_match == NULL) /* remember the first one */
1382 ia6_match = ifa6;
1383
1384 /*
1385 * An already autoconfigured address matched. Now that we
1386 * are sure there is at least one matched address, we can
1387 * proceed to 5.5.3. (e): update the lifetimes according to the
1388 * "two hours" rule and the privacy extension.
1389 * We apply some clarifications in rfc2462bis:
1390 * - use remaininglifetime instead of storedlifetime as a
1391 * variable name
1392 * - remove the dead code in the "two-hour" rule
1393 */
1394 #define TWOHOUR (120*60)
1395 lt6_tmp = ifa6->ia6_lifetime;
1396
1397 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1398 remaininglifetime = ND6_INFINITE_LIFETIME;
1399 else if (time_uptime - ifa6->ia6_updatetime >
1400 lt6_tmp.ia6t_vltime) {
1401 /*
1402 * The case of "invalid" address. We should usually
1403 * not see this case.
1404 */
1405 remaininglifetime = 0;
1406 } else
1407 remaininglifetime = lt6_tmp.ia6t_vltime -
1408 (time_uptime - ifa6->ia6_updatetime);
1409
1410 /* when not updating, keep the current stored lifetime. */
1411 lt6_tmp.ia6t_vltime = remaininglifetime;
1412
1413 if (TWOHOUR < new->ndpr_vltime ||
1414 remaininglifetime < new->ndpr_vltime) {
1415 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1416 } else if (remaininglifetime <= TWOHOUR) {
1417 if (auth) {
1418 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1419 }
1420 } else {
1421 /*
1422 * new->ndpr_vltime <= TWOHOUR &&
1423 * TWOHOUR < remaininglifetime
1424 */
1425 lt6_tmp.ia6t_vltime = TWOHOUR;
1426 }
1427
1428 /* The 2 hour rule is not imposed for preferred lifetime. */
1429 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1430
1431 in6_init_address_ltimes(pr, <6_tmp);
1432
1433 /*
1434 * We need to treat lifetimes for temporary addresses
1435 * differently, according to
1436 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1437 * we only update the lifetimes when they are in the maximum
1438 * intervals.
1439 */
1440 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1441 u_int32_t maxvltime, maxpltime;
1442
1443 if (V_ip6_temp_valid_lifetime >
1444 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1445 V_ip6_desync_factor)) {
1446 maxvltime = V_ip6_temp_valid_lifetime -
1447 (time_uptime - ifa6->ia6_createtime) -
1448 V_ip6_desync_factor;
1449 } else
1450 maxvltime = 0;
1451 if (V_ip6_temp_preferred_lifetime >
1452 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1453 V_ip6_desync_factor)) {
1454 maxpltime = V_ip6_temp_preferred_lifetime -
1455 (time_uptime - ifa6->ia6_createtime) -
1456 V_ip6_desync_factor;
1457 } else
1458 maxpltime = 0;
1459
1460 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1461 lt6_tmp.ia6t_vltime > maxvltime) {
1462 lt6_tmp.ia6t_vltime = maxvltime;
1463 }
1464 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1465 lt6_tmp.ia6t_pltime > maxpltime) {
1466 lt6_tmp.ia6t_pltime = maxpltime;
1467 }
1468 }
1469 ifa6->ia6_lifetime = lt6_tmp;
1470 ifa6->ia6_updatetime = time_uptime;
1471 }
1472 IF_ADDR_RUNLOCK(ifp);
1473 if (ia6_match == NULL && new->ndpr_vltime) {
1474 int ifidlen;
1475
1476 /*
1477 * 5.5.3 (d) (continued)
1478 * No address matched and the valid lifetime is non-zero.
1479 * Create a new address.
1480 */
1481
1482 /*
1483 * Prefix Length check:
1484 * If the sum of the prefix length and interface identifier
1485 * length does not equal 128 bits, the Prefix Information
1486 * option MUST be ignored. The length of the interface
1487 * identifier is defined in a separate link-type specific
1488 * document.
1489 */
1490 ifidlen = in6_if2idlen(ifp);
1491 if (ifidlen < 0) {
1492 /* this should not happen, so we always log it. */
1493 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1494 if_name(ifp));
1495 goto end;
1496 }
1497 if (ifidlen + pr->ndpr_plen != 128) {
1498 nd6log((LOG_INFO,
1499 "prelist_update: invalid prefixlen "
1500 "%d for %s, ignored\n",
1501 pr->ndpr_plen, if_name(ifp)));
1502 goto end;
1503 }
1504
1505 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1506 /*
1507 * note that we should use pr (not new) for reference.
1508 */
1509 pr->ndpr_addrcnt++;
1510 ia6->ia6_ndpr = pr;
1511
1512 /*
1513 * RFC 3041 3.3 (2).
1514 * When a new public address is created as described
1515 * in RFC2462, also create a new temporary address.
1516 *
1517 * RFC 3041 3.5.
1518 * When an interface connects to a new link, a new
1519 * randomized interface identifier should be generated
1520 * immediately together with a new set of temporary
1521 * addresses. Thus, we specifiy 1 as the 2nd arg of
1522 * in6_tmpifadd().
1523 */
1524 if (V_ip6_use_tempaddr) {
1525 int e;
1526 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1527 nd6log((LOG_NOTICE, "prelist_update: "
1528 "failed to create a temporary "
1529 "address, errno=%d\n",
1530 e));
1531 }
1532 }
1533 ifa_free(&ia6->ia_ifa);
1534
1535 /*
1536 * A newly added address might affect the status
1537 * of other addresses, so we check and update it.
1538 * XXX: what if address duplication happens?
1539 */
1540 pfxlist_onlink_check();
1541 } else {
1542 /* just set an error. do not bark here. */
1543 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1544 }
1545 }
1546
1547 end:
1548 if (pr != NULL)
1549 nd6_prefix_rele(pr);
1550 return (error);
1551 }
1552
1553 /*
1554 * A supplement function used in the on-link detection below;
1555 * detect if a given prefix has a (probably) reachable advertising router.
1556 * XXX: lengthy function name...
1557 */
1558 static struct nd_pfxrouter *
1559 find_pfxlist_reachable_router(struct nd_prefix *pr)
1560 {
1561 struct nd_pfxrouter *pfxrtr;
1562 struct llentry *ln;
1563 int canreach;
1564
1565 ND6_LOCK_ASSERT();
1566
1567 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1568 IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1569 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1570 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1571 if (ln == NULL)
1572 continue;
1573 canreach = ND6_IS_LLINFO_PROBREACH(ln);
1574 LLE_RUNLOCK(ln);
1575 if (canreach)
1576 break;
1577 }
1578 return (pfxrtr);
1579 }
1580
1581 /*
1582 * Check if each prefix in the prefix list has at least one available router
1583 * that advertised the prefix (a router is "available" if its neighbor cache
1584 * entry is reachable or probably reachable).
1585 * If the check fails, the prefix may be off-link, because, for example,
1586 * we have moved from the network but the lifetime of the prefix has not
1587 * expired yet. So we should not use the prefix if there is another prefix
1588 * that has an available router.
1589 * But, if there is no prefix that has an available router, we still regard
1590 * all the prefixes as on-link. This is because we can't tell if all the
1591 * routers are simply dead or if we really moved from the network and there
1592 * is no router around us.
1593 */
1594 void
1595 pfxlist_onlink_check(void)
1596 {
1597 struct nd_prefix *pr;
1598 struct in6_ifaddr *ifa;
1599 struct nd_defrouter *dr;
1600 struct nd_pfxrouter *pfxrtr = NULL;
1601 struct rm_priotracker in6_ifa_tracker;
1602 uint64_t genid;
1603 uint32_t flags;
1604
1605 ND6_ONLINK_LOCK();
1606 ND6_RLOCK();
1607
1608 /*
1609 * Check if there is a prefix that has a reachable advertising
1610 * router.
1611 */
1612 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1613 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1614 break;
1615 }
1616
1617 /*
1618 * If we have no such prefix, check whether we still have a router
1619 * that does not advertise any prefixes.
1620 */
1621 if (pr == NULL) {
1622 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1623 struct nd_prefix *pr0;
1624
1625 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1626 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1627 break;
1628 }
1629 if (pfxrtr != NULL)
1630 break;
1631 }
1632 }
1633 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1634 /*
1635 * There is at least one prefix that has a reachable router,
1636 * or at least a router which probably does not advertise
1637 * any prefixes. The latter would be the case when we move
1638 * to a new link where we have a router that does not provide
1639 * prefixes and we configure an address by hand.
1640 * Detach prefixes which have no reachable advertising
1641 * router, and attach other prefixes.
1642 */
1643 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1644 /* XXX: a link-local prefix should never be detached */
1645 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1646 pr->ndpr_raf_onlink == 0 ||
1647 pr->ndpr_raf_auto == 0)
1648 continue;
1649
1650 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1651 find_pfxlist_reachable_router(pr) == NULL)
1652 pr->ndpr_stateflags |= NDPRF_DETACHED;
1653 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1654 find_pfxlist_reachable_router(pr) != NULL)
1655 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1656 }
1657 } else {
1658 /* there is no prefix that has a reachable router */
1659 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1660 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1661 pr->ndpr_raf_onlink == 0 ||
1662 pr->ndpr_raf_auto == 0)
1663 continue;
1664 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1665 }
1666 }
1667
1668 /*
1669 * Remove each interface route associated with a (just) detached
1670 * prefix, and reinstall the interface route for a (just) attached
1671 * prefix. Note that all attempt of reinstallation does not
1672 * necessarily success, when a same prefix is shared among multiple
1673 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1674 * so we don't have to care about them.
1675 */
1676 restart:
1677 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1678 char ip6buf[INET6_ADDRSTRLEN];
1679 int e;
1680
1681 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1682 pr->ndpr_raf_onlink == 0 ||
1683 pr->ndpr_raf_auto == 0)
1684 continue;
1685
1686 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1687 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1688 genid = V_nd6_list_genid;
1689 ND6_RUNLOCK();
1690 if ((flags & NDPRF_ONLINK) != 0 &&
1691 (e = nd6_prefix_offlink(pr)) != 0) {
1692 nd6log((LOG_ERR,
1693 "pfxlist_onlink_check: failed to "
1694 "make %s/%d offlink, errno=%d\n",
1695 ip6_sprintf(ip6buf,
1696 &pr->ndpr_prefix.sin6_addr),
1697 pr->ndpr_plen, e));
1698 } else if ((flags & NDPRF_ONLINK) == 0 &&
1699 (e = nd6_prefix_onlink(pr)) != 0) {
1700 nd6log((LOG_ERR,
1701 "pfxlist_onlink_check: failed to "
1702 "make %s/%d onlink, errno=%d\n",
1703 ip6_sprintf(ip6buf,
1704 &pr->ndpr_prefix.sin6_addr),
1705 pr->ndpr_plen, e));
1706 }
1707 ND6_RLOCK();
1708 if (genid != V_nd6_list_genid)
1709 goto restart;
1710 }
1711 }
1712
1713 /*
1714 * Changes on the prefix status might affect address status as well.
1715 * Make sure that all addresses derived from an attached prefix are
1716 * attached, and that all addresses derived from a detached prefix are
1717 * detached. Note, however, that a manually configured address should
1718 * always be attached.
1719 * The precise detection logic is same as the one for prefixes.
1720 */
1721 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1722 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1723 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1724 continue;
1725
1726 if (ifa->ia6_ndpr == NULL) {
1727 /*
1728 * This can happen when we first configure the address
1729 * (i.e. the address exists, but the prefix does not).
1730 * XXX: complicated relationships...
1731 */
1732 continue;
1733 }
1734
1735 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1736 break;
1737 }
1738 if (ifa) {
1739 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1740 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1741 continue;
1742
1743 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1744 continue;
1745
1746 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1747 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1748 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1749 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1750 nd6_dad_start((struct ifaddr *)ifa, 0);
1751 }
1752 } else {
1753 ifa->ia6_flags |= IN6_IFF_DETACHED;
1754 }
1755 }
1756 } else {
1757 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1758 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1759 continue;
1760
1761 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1762 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1763 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1764 /* Do we need a delay in this case? */
1765 nd6_dad_start((struct ifaddr *)ifa, 0);
1766 }
1767 }
1768 }
1769 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1770 ND6_RUNLOCK();
1771 ND6_ONLINK_UNLOCK();
1772 }
1773
1774 static int
1775 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1776 {
1777 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1778 struct rib_head *rnh;
1779 struct rtentry *rt;
1780 struct sockaddr_in6 mask6;
1781 u_long rtflags;
1782 int error, a_failure, fibnum, maxfib;
1783
1784 /*
1785 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1786 * ifa->ifa_rtrequest = nd6_rtrequest;
1787 */
1788 bzero(&mask6, sizeof(mask6));
1789 mask6.sin6_len = sizeof(mask6);
1790 mask6.sin6_addr = pr->ndpr_mask;
1791 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1792
1793 if(V_rt_add_addr_allfibs) {
1794 fibnum = 0;
1795 maxfib = rt_numfibs;
1796 } else {
1797 fibnum = ifa->ifa_ifp->if_fib;
1798 maxfib = fibnum + 1;
1799 }
1800 a_failure = 0;
1801 for (; fibnum < maxfib; fibnum++) {
1802
1803 rt = NULL;
1804 error = in6_rtrequest(RTM_ADD,
1805 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1806 (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1807 if (error == 0) {
1808 KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1809 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1810 error, pr, ifa));
1811
1812 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1813 /* XXX what if rhn == NULL? */
1814 RIB_WLOCK(rnh);
1815 RT_LOCK(rt);
1816 if (rt_setgate(rt, rt_key(rt),
1817 (struct sockaddr *)&null_sdl) == 0) {
1818 struct sockaddr_dl *dl;
1819
1820 dl = (struct sockaddr_dl *)rt->rt_gateway;
1821 dl->sdl_type = rt->rt_ifp->if_type;
1822 dl->sdl_index = rt->rt_ifp->if_index;
1823 }
1824 RIB_WUNLOCK(rnh);
1825 nd6_rtmsg(RTM_ADD, rt);
1826 RT_UNLOCK(rt);
1827 pr->ndpr_stateflags |= NDPRF_ONLINK;
1828 } else {
1829 char ip6buf[INET6_ADDRSTRLEN];
1830 char ip6bufg[INET6_ADDRSTRLEN];
1831 char ip6bufm[INET6_ADDRSTRLEN];
1832 struct sockaddr_in6 *sin6;
1833
1834 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1835 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1836 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1837 "flags=%lx errno = %d\n",
1838 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1839 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1840 ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1841 ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1842 rtflags, error));
1843
1844 /* Save last error to return, see rtinit(). */
1845 a_failure = error;
1846 }
1847
1848 if (rt != NULL) {
1849 RT_LOCK(rt);
1850 RT_REMREF(rt);
1851 RT_UNLOCK(rt);
1852 }
1853 }
1854
1855 /* Return the last error we got. */
1856 return (a_failure);
1857 }
1858
1859 int
1860 nd6_prefix_onlink(struct nd_prefix *pr)
1861 {
1862 struct ifaddr *ifa;
1863 struct ifnet *ifp = pr->ndpr_ifp;
1864 struct nd_prefix *opr;
1865 char ip6buf[INET6_ADDRSTRLEN];
1866 int error;
1867
1868 ND6_ONLINK_LOCK_ASSERT();
1869 ND6_UNLOCK_ASSERT();
1870
1871 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1872 return (EEXIST);
1873
1874 /*
1875 * Add the interface route associated with the prefix. Before
1876 * installing the route, check if there's the same prefix on another
1877 * interface, and the prefix has already installed the interface route.
1878 * Although such a configuration is expected to be rare, we explicitly
1879 * allow it.
1880 */
1881 ND6_RLOCK();
1882 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1883 if (opr == pr)
1884 continue;
1885
1886 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1887 continue;
1888
1889 if (!V_rt_add_addr_allfibs &&
1890 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
1891 continue;
1892
1893 if (opr->ndpr_plen == pr->ndpr_plen &&
1894 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1895 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1896 ND6_RUNLOCK();
1897 return (0);
1898 }
1899 }
1900 ND6_RUNLOCK();
1901
1902 /*
1903 * We prefer link-local addresses as the associated interface address.
1904 */
1905 /* search for a link-local addr */
1906 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1907 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1908 if (ifa == NULL) {
1909 /* XXX: freebsd does not have ifa_ifwithaf */
1910 IF_ADDR_RLOCK(ifp);
1911 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1912 if (ifa->ifa_addr->sa_family == AF_INET6) {
1913 ifa_ref(ifa);
1914 break;
1915 }
1916 }
1917 IF_ADDR_RUNLOCK(ifp);
1918 /* should we care about ia6_flags? */
1919 }
1920 if (ifa == NULL) {
1921 /*
1922 * This can still happen, when, for example, we receive an RA
1923 * containing a prefix with the L bit set and the A bit clear,
1924 * after removing all IPv6 addresses on the receiving
1925 * interface. This should, of course, be rare though.
1926 */
1927 nd6log((LOG_NOTICE,
1928 "nd6_prefix_onlink: failed to find any ifaddr"
1929 " to add route for a prefix(%s/%d) on %s\n",
1930 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1931 pr->ndpr_plen, if_name(ifp)));
1932 return (0);
1933 }
1934
1935 error = nd6_prefix_onlink_rtrequest(pr, ifa);
1936
1937 if (ifa != NULL)
1938 ifa_free(ifa);
1939
1940 return (error);
1941 }
1942
1943 int
1944 nd6_prefix_offlink(struct nd_prefix *pr)
1945 {
1946 int error = 0;
1947 struct ifnet *ifp = pr->ndpr_ifp;
1948 struct nd_prefix *opr;
1949 struct sockaddr_in6 sa6, mask6;
1950 struct rtentry *rt;
1951 char ip6buf[INET6_ADDRSTRLEN];
1952 uint64_t genid;
1953 int fibnum, maxfib, a_failure;
1954
1955 ND6_ONLINK_LOCK_ASSERT();
1956 ND6_UNLOCK_ASSERT();
1957
1958 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1959 return (EEXIST);
1960
1961 bzero(&sa6, sizeof(sa6));
1962 sa6.sin6_family = AF_INET6;
1963 sa6.sin6_len = sizeof(sa6);
1964 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1965 sizeof(struct in6_addr));
1966 bzero(&mask6, sizeof(mask6));
1967 mask6.sin6_family = AF_INET6;
1968 mask6.sin6_len = sizeof(sa6);
1969 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1970
1971 if (V_rt_add_addr_allfibs) {
1972 fibnum = 0;
1973 maxfib = rt_numfibs;
1974 } else {
1975 fibnum = ifp->if_fib;
1976 maxfib = fibnum + 1;
1977 }
1978
1979 a_failure = 0;
1980 for (; fibnum < maxfib; fibnum++) {
1981 rt = NULL;
1982 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1983 (struct sockaddr *)&mask6, 0, &rt, fibnum);
1984 if (error == 0) {
1985 /* report the route deletion to the routing socket. */
1986 if (rt != NULL)
1987 nd6_rtmsg(RTM_DELETE, rt);
1988 } else {
1989 /* Save last error to return, see rtinit(). */
1990 a_failure = error;
1991 }
1992 if (rt != NULL) {
1993 RTFREE(rt);
1994 }
1995 }
1996 error = a_failure;
1997 a_failure = 1;
1998 if (error == 0) {
1999 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2000
2001 /*
2002 * There might be the same prefix on another interface,
2003 * the prefix which could not be on-link just because we have
2004 * the interface route (see comments in nd6_prefix_onlink).
2005 * If there's one, try to make the prefix on-link on the
2006 * interface.
2007 */
2008 ND6_RLOCK();
2009 restart:
2010 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2011 /*
2012 * KAME specific: detached prefixes should not be
2013 * on-link.
2014 */
2015 if (opr == pr || (opr->ndpr_stateflags &
2016 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2017 continue;
2018
2019 if (opr->ndpr_plen == pr->ndpr_plen &&
2020 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2021 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2022 int e;
2023
2024 genid = V_nd6_list_genid;
2025 ND6_RUNLOCK();
2026 if ((e = nd6_prefix_onlink(opr)) != 0) {
2027 nd6log((LOG_ERR,
2028 "nd6_prefix_offlink: failed to "
2029 "recover a prefix %s/%d from %s "
2030 "to %s (errno = %d)\n",
2031 ip6_sprintf(ip6buf,
2032 &opr->ndpr_prefix.sin6_addr),
2033 opr->ndpr_plen, if_name(ifp),
2034 if_name(opr->ndpr_ifp), e));
2035 } else
2036 a_failure = 0;
2037 ND6_RLOCK();
2038 if (genid != V_nd6_list_genid)
2039 goto restart;
2040 }
2041 }
2042 ND6_RUNLOCK();
2043 } else {
2044 /* XXX: can we still set the NDPRF_ONLINK flag? */
2045 nd6log((LOG_ERR,
2046 "nd6_prefix_offlink: failed to delete route: "
2047 "%s/%d on %s (errno = %d)\n",
2048 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
2049 if_name(ifp), error));
2050 }
2051
2052 if (a_failure)
2053 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6,
2054 (struct sockaddr *)&mask6, LLE_STATIC);
2055
2056 return (error);
2057 }
2058
2059 static struct in6_ifaddr *
2060 in6_ifadd(struct nd_prefixctl *pr, int mcast)
2061 {
2062 struct ifnet *ifp = pr->ndpr_ifp;
2063 struct ifaddr *ifa;
2064 struct in6_aliasreq ifra;
2065 struct in6_ifaddr *ia, *ib;
2066 int error, plen0;
2067 struct in6_addr mask;
2068 int prefixlen = pr->ndpr_plen;
2069 int updateflags;
2070 char ip6buf[INET6_ADDRSTRLEN];
2071
2072 in6_prefixlen2mask(&mask, prefixlen);
2073
2074 /*
2075 * find a link-local address (will be interface ID).
2076 * Is it really mandatory? Theoretically, a global or a site-local
2077 * address can be configured without a link-local address, if we
2078 * have a unique interface identifier...
2079 *
2080 * it is not mandatory to have a link-local address, we can generate
2081 * interface identifier on the fly. we do this because:
2082 * (1) it should be the easiest way to find interface identifier.
2083 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
2084 * for multiple addresses on a single interface, and possible shortcut
2085 * of DAD. we omitted DAD for this reason in the past.
2086 * (3) a user can prevent autoconfiguration of global address
2087 * by removing link-local address by hand (this is partly because we
2088 * don't have other way to control the use of IPv6 on an interface.
2089 * this has been our design choice - cf. NRL's "ifconfig auto").
2090 * (4) it is easier to manage when an interface has addresses
2091 * with the same interface identifier, than to have multiple addresses
2092 * with different interface identifiers.
2093 */
2094 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
2095 if (ifa)
2096 ib = (struct in6_ifaddr *)ifa;
2097 else
2098 return NULL;
2099
2100 /* prefixlen + ifidlen must be equal to 128 */
2101 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
2102 if (prefixlen != plen0) {
2103 ifa_free(ifa);
2104 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
2105 "(prefix=%d ifid=%d)\n",
2106 if_name(ifp), prefixlen, 128 - plen0));
2107 return NULL;
2108 }
2109
2110 /* make ifaddr */
2111 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
2112
2113 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
2114 /* interface ID */
2115 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
2116 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
2117 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
2118 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
2119 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2120 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
2121 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2122 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
2123 ifa_free(ifa);
2124
2125 /* lifetimes. */
2126 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
2127 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
2128
2129 /* XXX: scope zone ID? */
2130
2131 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
2132
2133 /*
2134 * Make sure that we do not have this address already. This should
2135 * usually not happen, but we can still see this case, e.g., if we
2136 * have manually configured the exact address to be configured.
2137 */
2138 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
2139 &ifra.ifra_addr.sin6_addr);
2140 if (ifa != NULL) {
2141 ifa_free(ifa);
2142 /* this should be rare enough to make an explicit log */
2143 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
2144 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
2145 return (NULL);
2146 }
2147
2148 /*
2149 * Allocate ifaddr structure, link into chain, etc.
2150 * If we are going to create a new address upon receiving a multicasted
2151 * RA, we need to impose a random delay before starting DAD.
2152 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
2153 */
2154 updateflags = 0;
2155 if (mcast)
2156 updateflags |= IN6_IFAUPDATE_DADDELAY;
2157 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
2158 nd6log((LOG_ERR,
2159 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
2160 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
2161 if_name(ifp), error));
2162 return (NULL); /* ifaddr must not have been allocated. */
2163 }
2164
2165 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2166 /*
2167 * XXXRW: Assumption of non-NULLness here might not be true with
2168 * fine-grained locking -- should we validate it? Or just return
2169 * earlier ifa rather than looking it up again?
2170 */
2171 return (ia); /* this is always non-NULL and referenced. */
2172 }
2173
2174 /*
2175 * ia0 - corresponding public address
2176 */
2177 int
2178 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2179 {
2180 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2181 struct in6_ifaddr *newia;
2182 struct in6_aliasreq ifra;
2183 int error;
2184 int trylimit = 3; /* XXX: adhoc value */
2185 int updateflags;
2186 u_int32_t randid[2];
2187 time_t vltime0, pltime0;
2188
2189 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2190 &ia0->ia_prefixmask.sin6_addr);
2191
2192 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */
2193 /* clear the old IFID */
2194 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2195 &ifra.ifra_prefixmask.sin6_addr);
2196
2197 again:
2198 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2199 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2200 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
2201 "random IFID\n"));
2202 return (EINVAL);
2203 }
2204 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2205 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2206 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2207 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2208
2209 /*
2210 * in6_get_tmpifid() quite likely provided a unique interface ID.
2211 * However, we may still have a chance to see collision, because
2212 * there may be a time lag between generation of the ID and generation
2213 * of the address. So, we'll do one more sanity check.
2214 */
2215
2216 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2217 if (trylimit-- > 0) {
2218 forcegen = 1;
2219 goto again;
2220 }
2221
2222 /* Give up. Something strange should have happened. */
2223 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
2224 "find a unique random IFID\n"));
2225 return (EEXIST);
2226 }
2227
2228 /*
2229 * The Valid Lifetime is the lower of the Valid Lifetime of the
2230 * public address or TEMP_VALID_LIFETIME.
2231 * The Preferred Lifetime is the lower of the Preferred Lifetime
2232 * of the public address or TEMP_PREFERRED_LIFETIME -
2233 * DESYNC_FACTOR.
2234 */
2235 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2236 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2237 (ia0->ia6_lifetime.ia6t_vltime -
2238 (time_uptime - ia0->ia6_updatetime));
2239 if (vltime0 > V_ip6_temp_valid_lifetime)
2240 vltime0 = V_ip6_temp_valid_lifetime;
2241 } else
2242 vltime0 = V_ip6_temp_valid_lifetime;
2243 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2244 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2245 (ia0->ia6_lifetime.ia6t_pltime -
2246 (time_uptime - ia0->ia6_updatetime));
2247 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2248 pltime0 = V_ip6_temp_preferred_lifetime -
2249 V_ip6_desync_factor;
2250 }
2251 } else
2252 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2253 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2254 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2255
2256 /*
2257 * A temporary address is created only if this calculated Preferred
2258 * Lifetime is greater than REGEN_ADVANCE time units.
2259 */
2260 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2261 return (0);
2262
2263 /* XXX: scope zone ID? */
2264
2265 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2266
2267 /* allocate ifaddr structure, link into chain, etc. */
2268 updateflags = 0;
2269 if (delay)
2270 updateflags |= IN6_IFAUPDATE_DADDELAY;
2271 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2272 return (error);
2273
2274 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2275 if (newia == NULL) { /* XXX: can it happen? */
2276 nd6log((LOG_ERR,
2277 "in6_tmpifadd: ifa update succeeded, but we got "
2278 "no ifaddr\n"));
2279 return (EINVAL); /* XXX */
2280 }
2281 newia->ia6_ndpr = ia0->ia6_ndpr;
2282 newia->ia6_ndpr->ndpr_addrcnt++;
2283 ifa_free(&newia->ia_ifa);
2284
2285 /*
2286 * A newly added address might affect the status of other addresses.
2287 * XXX: when the temporary address is generated with a new public
2288 * address, the onlink check is redundant. However, it would be safe
2289 * to do the check explicitly everywhere a new address is generated,
2290 * and, in fact, we surely need the check when we create a new
2291 * temporary address due to deprecation of an old temporary address.
2292 */
2293 pfxlist_onlink_check();
2294
2295 return (0);
2296 }
2297
2298 static int
2299 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2300 {
2301 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2302 ndpr->ndpr_preferred = 0;
2303 else
2304 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2305 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2306 ndpr->ndpr_expire = 0;
2307 else
2308 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2309
2310 return 0;
2311 }
2312
2313 static void
2314 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2315 {
2316 /* init ia6t_expire */
2317 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2318 lt6->ia6t_expire = 0;
2319 else {
2320 lt6->ia6t_expire = time_uptime;
2321 lt6->ia6t_expire += lt6->ia6t_vltime;
2322 }
2323
2324 /* init ia6t_preferred */
2325 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2326 lt6->ia6t_preferred = 0;
2327 else {
2328 lt6->ia6t_preferred = time_uptime;
2329 lt6->ia6t_preferred += lt6->ia6t_pltime;
2330 }
2331 }
2332
2333 /*
2334 * Delete all the routing table entries that use the specified gateway.
2335 * XXX: this function causes search through all entries of routing table, so
2336 * it shouldn't be called when acting as a router.
2337 */
2338 void
2339 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2340 {
2341
2342 /* We'll care only link-local addresses */
2343 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2344 return;
2345
2346 /* XXX Do we really need to walk any but the default FIB? */
2347 rt_foreach_fib_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2348 }
2349
2350 static int
2351 rt6_deleteroute(const struct rtentry *rt, void *arg)
2352 {
2353 #define SIN6(s) ((struct sockaddr_in6 *)s)
2354 struct in6_addr *gate = (struct in6_addr *)arg;
2355
2356 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2357 return (0);
2358
2359 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2360 return (0);
2361 }
2362
2363 /*
2364 * Do not delete a static route.
2365 * XXX: this seems to be a bit ad-hoc. Should we consider the
2366 * 'cloned' bit instead?
2367 */
2368 if ((rt->rt_flags & RTF_STATIC) != 0)
2369 return (0);
2370
2371 /*
2372 * We delete only host route. This means, in particular, we don't
2373 * delete default route.
2374 */
2375 if ((rt->rt_flags & RTF_HOST) == 0)
2376 return (0);
2377
2378 return (1);
2379 #undef SIN6
2380 }
2381
2382 int
2383 nd6_setdefaultiface(int ifindex)
2384 {
2385 int error = 0;
2386
2387 if (ifindex < 0 || V_if_index < ifindex)
2388 return (EINVAL);
2389 if (ifindex != 0 && !ifnet_byindex(ifindex))
2390 return (EINVAL);
2391
2392 if (V_nd6_defifindex != ifindex) {
2393 V_nd6_defifindex = ifindex;
2394 if (V_nd6_defifindex > 0)
2395 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2396 else
2397 V_nd6_defifp = NULL;
2398
2399 /*
2400 * Our current implementation assumes one-to-one maping between
2401 * interfaces and links, so it would be natural to use the
2402 * default interface as the default link.
2403 */
2404 scope6_setdefault(V_nd6_defifp);
2405 }
2406
2407 return (error);
2408 }
Cache object: dac4a17289ef9ad519e062413815b1a0
|