1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/11.2/sys/netinet6/nd6_rtr.c 331722 2018-03-29 02:50:57Z eadler $");
34
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/refcount.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/errno.h>
49 #include <sys/rmlock.h>
50 #include <sys/rwlock.h>
51 #include <sys/syslog.h>
52 #include <sys/queue.h>
53
54 #include <net/if.h>
55 #include <net/if_var.h>
56 #include <net/if_types.h>
57 #include <net/if_dl.h>
58 #include <net/route.h>
59 #include <net/route_var.h>
60 #include <net/radix.h>
61 #include <net/vnet.h>
62
63 #include <netinet/in.h>
64 #include <net/if_llatbl.h>
65 #include <netinet6/in6_var.h>
66 #include <netinet6/in6_ifattach.h>
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
69 #include <netinet6/nd6.h>
70 #include <netinet/icmp6.h>
71 #include <netinet6/scope6_var.h>
72
73 static int rtpref(struct nd_defrouter *);
74 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
75 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
76 struct mbuf *, int);
77 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
78 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
79 struct nd_defrouter *);
80 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
81 static void pfxrtr_del(struct nd_pfxrouter *);
82 static struct nd_pfxrouter *find_pfxlist_reachable_router(struct nd_prefix *);
83 static void defrouter_delreq(struct nd_defrouter *);
84 static void nd6_rtmsg(int, struct rtentry *);
85
86 static int in6_init_prefix_ltimes(struct nd_prefix *);
87 static void in6_init_address_ltimes(struct nd_prefix *,
88 struct in6_addrlifetime *);
89
90 static int rt6_deleteroute(const struct rtentry *, void *);
91
92 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
93 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
94
95 static VNET_DEFINE(struct ifnet *, nd6_defifp);
96 VNET_DEFINE(int, nd6_defifindex);
97 #define V_nd6_defifp VNET(nd6_defifp)
98
99 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
100
101 VNET_DEFINE(int, ip6_desync_factor);
102 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
103 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
104
105 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
106
107 /* RTPREF_MEDIUM has to be 0! */
108 #define RTPREF_HIGH 1
109 #define RTPREF_MEDIUM 0
110 #define RTPREF_LOW (-1)
111 #define RTPREF_RESERVED (-2)
112 #define RTPREF_INVALID (-3) /* internal */
113
114 /*
115 * Receive Router Solicitation Message - just for routers.
116 * Router solicitation/advertisement is mostly managed by userland program
117 * (rtadvd) so here we have no function like nd6_ra_output().
118 *
119 * Based on RFC 2461
120 */
121 void
122 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
123 {
124 struct ifnet *ifp = m->m_pkthdr.rcvif;
125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
126 struct nd_router_solicit *nd_rs;
127 struct in6_addr saddr6 = ip6->ip6_src;
128 char *lladdr = NULL;
129 int lladdrlen = 0;
130 union nd_opts ndopts;
131 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
132
133 /*
134 * Accept RS only when V_ip6_forwarding=1 and the interface has
135 * no ND6_IFF_ACCEPT_RTADV.
136 */
137 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
138 goto freeit;
139
140 /* RFC 6980: Nodes MUST silently ignore fragments */
141 if(m->m_flags & M_FRAGMENTED)
142 goto freeit;
143
144 /* Sanity checks */
145 if (ip6->ip6_hlim != 255) {
146 nd6log((LOG_ERR,
147 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
148 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
149 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
150 goto bad;
151 }
152
153 /*
154 * Don't update the neighbor cache, if src = ::.
155 * This indicates that the src has no IP address assigned yet.
156 */
157 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
158 goto freeit;
159
160 #ifndef PULLDOWN_TEST
161 IP6_EXTHDR_CHECK(m, off, icmp6len,);
162 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
163 #else
164 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
165 if (nd_rs == NULL) {
166 ICMP6STAT_INC(icp6s_tooshort);
167 return;
168 }
169 #endif
170
171 icmp6len -= sizeof(*nd_rs);
172 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
173 if (nd6_options(&ndopts) < 0) {
174 nd6log((LOG_INFO,
175 "nd6_rs_input: invalid ND option, ignored\n"));
176 /* nd6_options have incremented stats */
177 goto freeit;
178 }
179
180 if (ndopts.nd_opts_src_lladdr) {
181 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
182 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
183 }
184
185 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
186 nd6log((LOG_INFO,
187 "nd6_rs_input: lladdrlen mismatch for %s "
188 "(if %d, RS packet %d)\n",
189 ip6_sprintf(ip6bufs, &saddr6),
190 ifp->if_addrlen, lladdrlen - 2));
191 goto bad;
192 }
193
194 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
195
196 freeit:
197 m_freem(m);
198 return;
199
200 bad:
201 ICMP6STAT_INC(icp6s_badrs);
202 m_freem(m);
203 }
204
205 /*
206 * Receive Router Advertisement Message.
207 *
208 * Based on RFC 2461
209 * TODO: on-link bit on prefix information
210 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
211 */
212 void
213 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
214 {
215 struct ifnet *ifp = m->m_pkthdr.rcvif;
216 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
217 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
218 struct nd_router_advert *nd_ra;
219 struct in6_addr saddr6 = ip6->ip6_src;
220 int mcast = 0;
221 union nd_opts ndopts;
222 struct nd_defrouter *dr;
223 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
224
225 dr = NULL;
226
227 /*
228 * We only accept RAs only when the per-interface flag
229 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
230 */
231 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
232 goto freeit;
233
234 /* RFC 6980: Nodes MUST silently ignore fragments */
235 if(m->m_flags & M_FRAGMENTED)
236 goto freeit;
237
238 if (ip6->ip6_hlim != 255) {
239 nd6log((LOG_ERR,
240 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
241 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
242 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
243 goto bad;
244 }
245
246 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
247 nd6log((LOG_ERR,
248 "nd6_ra_input: src %s is not link-local\n",
249 ip6_sprintf(ip6bufs, &saddr6)));
250 goto bad;
251 }
252
253 #ifndef PULLDOWN_TEST
254 IP6_EXTHDR_CHECK(m, off, icmp6len,);
255 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
256 #else
257 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
258 if (nd_ra == NULL) {
259 ICMP6STAT_INC(icp6s_tooshort);
260 return;
261 }
262 #endif
263
264 icmp6len -= sizeof(*nd_ra);
265 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
266 if (nd6_options(&ndopts) < 0) {
267 nd6log((LOG_INFO,
268 "nd6_ra_input: invalid ND option, ignored\n"));
269 /* nd6_options have incremented stats */
270 goto freeit;
271 }
272
273 {
274 struct nd_defrouter dr0;
275 u_int32_t advreachable = nd_ra->nd_ra_reachable;
276
277 /* remember if this is a multicasted advertisement */
278 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
279 mcast = 1;
280
281 bzero(&dr0, sizeof(dr0));
282 dr0.rtaddr = saddr6;
283 dr0.raflags = nd_ra->nd_ra_flags_reserved;
284 /*
285 * Effectively-disable routes from RA messages when
286 * ND6_IFF_NO_RADR enabled on the receiving interface or
287 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
288 */
289 if (ndi->flags & ND6_IFF_NO_RADR)
290 dr0.rtlifetime = 0;
291 else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
292 dr0.rtlifetime = 0;
293 else
294 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
295 dr0.expire = time_uptime + dr0.rtlifetime;
296 dr0.ifp = ifp;
297 /* unspecified or not? (RFC 2461 6.3.4) */
298 if (advreachable) {
299 advreachable = ntohl(advreachable);
300 if (advreachable <= MAX_REACHABLE_TIME &&
301 ndi->basereachable != advreachable) {
302 ndi->basereachable = advreachable;
303 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
304 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
305 }
306 }
307 if (nd_ra->nd_ra_retransmit)
308 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
309 if (nd_ra->nd_ra_curhoplimit) {
310 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
311 ndi->chlim = nd_ra->nd_ra_curhoplimit;
312 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
313 log(LOG_ERR, "RA with a lower CurHopLimit sent from "
314 "%s on %s (current = %d, received = %d). "
315 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
316 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
317 }
318 }
319 dr = defrtrlist_update(&dr0);
320 }
321
322 /*
323 * prefix
324 */
325 if (ndopts.nd_opts_pi) {
326 struct nd_opt_hdr *pt;
327 struct nd_opt_prefix_info *pi = NULL;
328 struct nd_prefixctl pr;
329
330 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
331 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
332 pt = (struct nd_opt_hdr *)((caddr_t)pt +
333 (pt->nd_opt_len << 3))) {
334 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
335 continue;
336 pi = (struct nd_opt_prefix_info *)pt;
337
338 if (pi->nd_opt_pi_len != 4) {
339 nd6log((LOG_INFO,
340 "nd6_ra_input: invalid option "
341 "len %d for prefix information option, "
342 "ignored\n", pi->nd_opt_pi_len));
343 continue;
344 }
345
346 if (128 < pi->nd_opt_pi_prefix_len) {
347 nd6log((LOG_INFO,
348 "nd6_ra_input: invalid prefix "
349 "len %d for prefix information option, "
350 "ignored\n", pi->nd_opt_pi_prefix_len));
351 continue;
352 }
353
354 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
355 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
356 nd6log((LOG_INFO,
357 "nd6_ra_input: invalid prefix "
358 "%s, ignored\n",
359 ip6_sprintf(ip6bufs,
360 &pi->nd_opt_pi_prefix)));
361 continue;
362 }
363
364 bzero(&pr, sizeof(pr));
365 pr.ndpr_prefix.sin6_family = AF_INET6;
366 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
367 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
368 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
369
370 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
371 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
372 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
373 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
374 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
375 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
376 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
377 (void)prelist_update(&pr, dr, m, mcast);
378 }
379 }
380 if (dr != NULL) {
381 defrouter_rele(dr);
382 dr = NULL;
383 }
384
385 /*
386 * MTU
387 */
388 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
389 u_long mtu;
390 u_long maxmtu;
391
392 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
393
394 /* lower bound */
395 if (mtu < IPV6_MMTU) {
396 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
397 "mtu=%lu sent from %s, ignoring\n",
398 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
399 goto skip;
400 }
401
402 /* upper bound */
403 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
404 ? ndi->maxmtu : ifp->if_mtu;
405 if (mtu <= maxmtu) {
406 int change = (ndi->linkmtu != mtu);
407
408 ndi->linkmtu = mtu;
409 if (change) {
410 /* in6_maxmtu may change */
411 in6_setmaxmtu();
412 rt_updatemtu(ifp);
413 }
414 } else {
415 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
416 "mtu=%lu sent from %s; "
417 "exceeds maxmtu %lu, ignoring\n",
418 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
419 }
420 }
421
422 skip:
423
424 /*
425 * Source link layer address
426 */
427 {
428 char *lladdr = NULL;
429 int lladdrlen = 0;
430
431 if (ndopts.nd_opts_src_lladdr) {
432 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
433 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
434 }
435
436 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
437 nd6log((LOG_INFO,
438 "nd6_ra_input: lladdrlen mismatch for %s "
439 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
440 ifp->if_addrlen, lladdrlen - 2));
441 goto bad;
442 }
443
444 nd6_cache_lladdr(ifp, &saddr6, lladdr,
445 lladdrlen, ND_ROUTER_ADVERT, 0);
446
447 /*
448 * Installing a link-layer address might change the state of the
449 * router's neighbor cache, which might also affect our on-link
450 * detection of adveritsed prefixes.
451 */
452 pfxlist_onlink_check();
453 }
454
455 freeit:
456 m_freem(m);
457 return;
458
459 bad:
460 ICMP6STAT_INC(icp6s_badra);
461 m_freem(m);
462 }
463
464 /* tell the change to user processes watching the routing socket. */
465 static void
466 nd6_rtmsg(int cmd, struct rtentry *rt)
467 {
468 struct rt_addrinfo info;
469 struct ifnet *ifp;
470 struct ifaddr *ifa;
471
472 bzero((caddr_t)&info, sizeof(info));
473 info.rti_info[RTAX_DST] = rt_key(rt);
474 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
475 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
476 ifp = rt->rt_ifp;
477 if (ifp != NULL) {
478 IF_ADDR_RLOCK(ifp);
479 ifa = TAILQ_FIRST(&ifp->if_addrhead);
480 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
481 ifa_ref(ifa);
482 IF_ADDR_RUNLOCK(ifp);
483 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
484 } else
485 ifa = NULL;
486
487 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
488 if (ifa != NULL)
489 ifa_free(ifa);
490 }
491
492 /*
493 * default router list processing sub routines
494 */
495
496 static void
497 defrouter_addreq(struct nd_defrouter *new)
498 {
499 struct sockaddr_in6 def, mask, gate;
500 struct rtentry *newrt = NULL;
501 int error;
502
503 bzero(&def, sizeof(def));
504 bzero(&mask, sizeof(mask));
505 bzero(&gate, sizeof(gate));
506
507 def.sin6_len = mask.sin6_len = gate.sin6_len =
508 sizeof(struct sockaddr_in6);
509 def.sin6_family = gate.sin6_family = AF_INET6;
510 gate.sin6_addr = new->rtaddr;
511
512 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
513 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
514 RTF_GATEWAY, &newrt, new->ifp->if_fib);
515 if (newrt) {
516 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
517 RTFREE(newrt);
518 }
519 if (error == 0)
520 new->installed = 1;
521 }
522
523 struct nd_defrouter *
524 defrouter_lookup_locked(struct in6_addr *addr, struct ifnet *ifp)
525 {
526 struct nd_defrouter *dr;
527
528 ND6_LOCK_ASSERT();
529 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
530 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
531 defrouter_ref(dr);
532 return (dr);
533 }
534 return (NULL);
535 }
536
537 struct nd_defrouter *
538 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
539 {
540 struct nd_defrouter *dr;
541
542 ND6_RLOCK();
543 dr = defrouter_lookup_locked(addr, ifp);
544 ND6_RUNLOCK();
545 return (dr);
546 }
547
548 void
549 defrouter_ref(struct nd_defrouter *dr)
550 {
551
552 refcount_acquire(&dr->refcnt);
553 }
554
555 void
556 defrouter_rele(struct nd_defrouter *dr)
557 {
558
559 if (refcount_release(&dr->refcnt))
560 free(dr, M_IP6NDP);
561 }
562
563 /*
564 * Remove the default route for a given router.
565 * This is just a subroutine function for defrouter_select_fib(), and
566 * should not be called from anywhere else.
567 */
568 static void
569 defrouter_delreq(struct nd_defrouter *dr)
570 {
571 struct sockaddr_in6 def, mask, gate;
572 struct rtentry *oldrt = NULL;
573
574 bzero(&def, sizeof(def));
575 bzero(&mask, sizeof(mask));
576 bzero(&gate, sizeof(gate));
577
578 def.sin6_len = mask.sin6_len = gate.sin6_len =
579 sizeof(struct sockaddr_in6);
580 def.sin6_family = gate.sin6_family = AF_INET6;
581 gate.sin6_addr = dr->rtaddr;
582
583 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
584 (struct sockaddr *)&gate,
585 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, dr->ifp->if_fib);
586 if (oldrt) {
587 nd6_rtmsg(RTM_DELETE, oldrt);
588 RTFREE(oldrt);
589 }
590
591 dr->installed = 0;
592 }
593
594 /*
595 * Remove all default routes from default router list.
596 */
597 void
598 defrouter_reset(void)
599 {
600 struct nd_defrouter *dr, **dra;
601 int count, i;
602
603 count = i = 0;
604
605 /*
606 * We can't delete routes with the ND lock held, so make a copy of the
607 * current default router list and use that when deleting routes.
608 */
609 ND6_RLOCK();
610 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
611 count++;
612 ND6_RUNLOCK();
613
614 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
615
616 ND6_RLOCK();
617 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
618 if (i == count)
619 break;
620 defrouter_ref(dr);
621 dra[i++] = dr;
622 }
623 ND6_RUNLOCK();
624
625 for (i = 0; i < count && dra[i] != NULL; i++) {
626 defrouter_delreq(dra[i]);
627 defrouter_rele(dra[i]);
628 }
629 free(dra, M_TEMP);
630
631 /*
632 * XXX should we also nuke any default routers in the kernel, by
633 * going through them by rtalloc1()?
634 */
635 }
636
637 /*
638 * Look up a matching default router list entry and remove it. Returns true if a
639 * matching entry was found, false otherwise.
640 */
641 bool
642 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
643 {
644 struct nd_defrouter *dr;
645
646 ND6_WLOCK();
647 dr = defrouter_lookup_locked(addr, ifp);
648 if (dr == NULL) {
649 ND6_WUNLOCK();
650 return (false);
651 }
652
653 defrouter_unlink(dr, NULL);
654 ND6_WUNLOCK();
655 defrouter_del(dr);
656 defrouter_rele(dr);
657 return (true);
658 }
659
660 /*
661 * Remove a router from the global list and optionally stash it in a
662 * caller-supplied queue.
663 *
664 * The ND lock must be held.
665 */
666 void
667 defrouter_unlink(struct nd_defrouter *dr, struct nd_drhead *drq)
668 {
669
670 ND6_WLOCK_ASSERT();
671 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
672 V_nd6_list_genid++;
673 if (drq != NULL)
674 TAILQ_INSERT_TAIL(drq, dr, dr_entry);
675 }
676
677 void
678 defrouter_del(struct nd_defrouter *dr)
679 {
680 struct nd_defrouter *deldr = NULL;
681 struct nd_prefix *pr;
682 struct nd_pfxrouter *pfxrtr;
683
684 ND6_UNLOCK_ASSERT();
685
686 /*
687 * Flush all the routing table entries that use the router
688 * as a next hop.
689 */
690 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
691 rt6_flush(&dr->rtaddr, dr->ifp);
692
693 if (dr->installed) {
694 deldr = dr;
695 defrouter_delreq(dr);
696 }
697
698 /*
699 * Also delete all the pointers to the router in each prefix lists.
700 */
701 ND6_WLOCK();
702 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
703 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
704 pfxrtr_del(pfxrtr);
705 }
706 ND6_WUNLOCK();
707
708 pfxlist_onlink_check();
709
710 /*
711 * If the router is the primary one, choose a new one.
712 * Note that defrouter_select_fib() will remove the current
713 * gateway from the routing table.
714 */
715 if (deldr)
716 defrouter_select_fib(deldr->ifp->if_fib);
717
718 /*
719 * Release the list reference.
720 */
721 defrouter_rele(dr);
722 }
723
724 /*
725 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
726 * draft-ietf-ipngwg-router-selection:
727 * 1) Routers that are reachable or probably reachable should be preferred.
728 * If we have more than one (probably) reachable router, prefer ones
729 * with the highest router preference.
730 * 2) When no routers on the list are known to be reachable or
731 * probably reachable, routers SHOULD be selected in a round-robin
732 * fashion, regardless of router preference values.
733 * 3) If the Default Router List is empty, assume that all
734 * destinations are on-link.
735 *
736 * We assume nd_defrouter is sorted by router preference value.
737 * Since the code below covers both with and without router preference cases,
738 * we do not need to classify the cases by ifdef.
739 *
740 * At this moment, we do not try to install more than one default router,
741 * even when the multipath routing is available, because we're not sure about
742 * the benefits for stub hosts comparing to the risk of making the code
743 * complicated and the possibility of introducing bugs.
744 *
745 * We maintain a single list of routers for multiple FIBs, only considering one
746 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
747 * we do the whole thing multiple times.
748 */
749 void
750 defrouter_select_fib(int fibnum)
751 {
752 struct nd_defrouter *dr, *selected_dr, *installed_dr;
753 struct llentry *ln = NULL;
754
755 if (fibnum == RT_ALL_FIBS) {
756 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
757 defrouter_select_fib(fibnum);
758 }
759 }
760
761 ND6_RLOCK();
762 /*
763 * Let's handle easy case (3) first:
764 * If default router list is empty, there's nothing to be done.
765 */
766 if (TAILQ_EMPTY(&V_nd_defrouter)) {
767 ND6_RUNLOCK();
768 return;
769 }
770
771 /*
772 * Search for a (probably) reachable router from the list.
773 * We just pick up the first reachable one (if any), assuming that
774 * the ordering rule of the list described in defrtrlist_update().
775 */
776 selected_dr = installed_dr = NULL;
777 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
778 IF_AFDATA_RLOCK(dr->ifp);
779 if (selected_dr == NULL && dr->ifp->if_fib == fibnum &&
780 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
781 ND6_IS_LLINFO_PROBREACH(ln)) {
782 selected_dr = dr;
783 defrouter_ref(selected_dr);
784 }
785 IF_AFDATA_RUNLOCK(dr->ifp);
786 if (ln != NULL) {
787 LLE_RUNLOCK(ln);
788 ln = NULL;
789 }
790
791 if (dr->installed && dr->ifp->if_fib == fibnum) {
792 if (installed_dr == NULL) {
793 installed_dr = dr;
794 defrouter_ref(installed_dr);
795 } else {
796 /*
797 * this should not happen.
798 * warn for diagnosis.
799 */
800 log(LOG_ERR, "defrouter_select_fib: more than "
801 "one router is installed\n");
802 }
803 }
804 }
805 /*
806 * If none of the default routers was found to be reachable,
807 * round-robin the list regardless of preference.
808 * Otherwise, if we have an installed router, check if the selected
809 * (reachable) router should really be preferred to the installed one.
810 * We only prefer the new router when the old one is not reachable
811 * or when the new one has a really higher preference value.
812 */
813 if (selected_dr == NULL) {
814 if (installed_dr == NULL ||
815 TAILQ_NEXT(installed_dr, dr_entry) == NULL)
816 dr = TAILQ_FIRST(&V_nd_defrouter);
817 else
818 dr = TAILQ_NEXT(installed_dr, dr_entry);
819
820 /* Ensure we select a router for this FIB. */
821 TAILQ_FOREACH_FROM(dr, &V_nd_defrouter, dr_entry) {
822 if (dr->ifp->if_fib == fibnum) {
823 selected_dr = dr;
824 defrouter_ref(selected_dr);
825 break;
826 }
827 }
828 } else if (installed_dr != NULL) {
829 IF_AFDATA_RLOCK(installed_dr->ifp);
830 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0,
831 installed_dr->ifp)) &&
832 ND6_IS_LLINFO_PROBREACH(ln) &&
833 installed_dr->ifp->if_fib == fibnum &&
834 rtpref(selected_dr) <= rtpref(installed_dr)) {
835 defrouter_rele(selected_dr);
836 selected_dr = installed_dr;
837 }
838 IF_AFDATA_RUNLOCK(installed_dr->ifp);
839 if (ln != NULL)
840 LLE_RUNLOCK(ln);
841 }
842 ND6_RUNLOCK();
843
844 /*
845 * If we selected a router for this FIB and it's different
846 * than the installed one, remove the installed router and
847 * install the selected one in its place.
848 */
849 if (installed_dr != selected_dr) {
850 if (installed_dr != NULL) {
851 defrouter_delreq(installed_dr);
852 defrouter_rele(installed_dr);
853 }
854 if (selected_dr != NULL)
855 defrouter_addreq(selected_dr);
856 }
857 if (selected_dr != NULL)
858 defrouter_rele(selected_dr);
859 }
860
861 /*
862 * Maintain old KPI for default router selection.
863 * If unspecified, we can re-select routers for all FIBs.
864 */
865 void
866 defrouter_select(void)
867 {
868 defrouter_select_fib(RT_ALL_FIBS);
869 }
870
871 /*
872 * for default router selection
873 * regards router-preference field as a 2-bit signed integer
874 */
875 static int
876 rtpref(struct nd_defrouter *dr)
877 {
878 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
879 case ND_RA_FLAG_RTPREF_HIGH:
880 return (RTPREF_HIGH);
881 case ND_RA_FLAG_RTPREF_MEDIUM:
882 case ND_RA_FLAG_RTPREF_RSV:
883 return (RTPREF_MEDIUM);
884 case ND_RA_FLAG_RTPREF_LOW:
885 return (RTPREF_LOW);
886 default:
887 /*
888 * This case should never happen. If it did, it would mean a
889 * serious bug of kernel internal. We thus always bark here.
890 * Or, can we even panic?
891 */
892 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
893 return (RTPREF_INVALID);
894 }
895 /* NOTREACHED */
896 }
897
898 static struct nd_defrouter *
899 defrtrlist_update(struct nd_defrouter *new)
900 {
901 struct nd_defrouter *dr, *n;
902 uint64_t genid;
903 int oldpref;
904 bool writelocked;
905
906 if (new->rtlifetime == 0) {
907 defrouter_remove(&new->rtaddr, new->ifp);
908 return (NULL);
909 }
910
911 ND6_RLOCK();
912 writelocked = false;
913 restart:
914 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
915 if (dr != NULL) {
916 oldpref = rtpref(dr);
917
918 /* override */
919 dr->raflags = new->raflags; /* XXX flag check */
920 dr->rtlifetime = new->rtlifetime;
921 dr->expire = new->expire;
922
923 /*
924 * If the preference does not change, there's no need
925 * to sort the entries. Also make sure the selected
926 * router is still installed in the kernel.
927 */
928 if (dr->installed && rtpref(new) == oldpref) {
929 if (writelocked)
930 ND6_WUNLOCK();
931 else
932 ND6_RUNLOCK();
933 return (dr);
934 }
935 }
936
937 /*
938 * The router needs to be reinserted into the default router
939 * list, so upgrade to a write lock. If that fails and the list
940 * has potentially changed while the lock was dropped, we'll
941 * redo the lookup with the write lock held.
942 */
943 if (!writelocked) {
944 writelocked = true;
945 if (!ND6_TRY_UPGRADE()) {
946 genid = V_nd6_list_genid;
947 ND6_RUNLOCK();
948 ND6_WLOCK();
949 if (genid != V_nd6_list_genid)
950 goto restart;
951 }
952 }
953
954 if (dr != NULL) {
955 /*
956 * The preferred router may have changed, so relocate this
957 * router.
958 */
959 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
960 n = dr;
961 } else {
962 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
963 if (n == NULL) {
964 ND6_WUNLOCK();
965 return (NULL);
966 }
967 memcpy(n, new, sizeof(*n));
968 /* Initialize with an extra reference for the caller. */
969 refcount_init(&n->refcnt, 2);
970 }
971
972 /*
973 * Insert the new router in the Default Router List;
974 * The Default Router List should be in the descending order
975 * of router-preferece. Routers with the same preference are
976 * sorted in the arriving time order.
977 */
978
979 /* insert at the end of the group */
980 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
981 if (rtpref(n) > rtpref(dr))
982 break;
983 }
984 if (dr != NULL)
985 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
986 else
987 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
988 V_nd6_list_genid++;
989 ND6_WUNLOCK();
990
991 defrouter_select_fib(new->ifp->if_fib);
992
993 return (n);
994 }
995
996 static struct nd_pfxrouter *
997 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
998 {
999 struct nd_pfxrouter *search;
1000
1001 ND6_LOCK_ASSERT();
1002
1003 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
1004 if (search->router == dr)
1005 break;
1006 }
1007 return (search);
1008 }
1009
1010 static void
1011 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
1012 {
1013 struct nd_pfxrouter *new;
1014 bool update;
1015
1016 ND6_UNLOCK_ASSERT();
1017
1018 ND6_RLOCK();
1019 if (pfxrtr_lookup(pr, dr) != NULL) {
1020 ND6_RUNLOCK();
1021 return;
1022 }
1023 ND6_RUNLOCK();
1024
1025 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1026 if (new == NULL)
1027 return;
1028 defrouter_ref(dr);
1029 new->router = dr;
1030
1031 ND6_WLOCK();
1032 if (pfxrtr_lookup(pr, dr) == NULL) {
1033 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
1034 update = true;
1035 } else {
1036 /* We lost a race to add the reference. */
1037 defrouter_rele(dr);
1038 free(new, M_IP6NDP);
1039 update = false;
1040 }
1041 ND6_WUNLOCK();
1042
1043 if (update)
1044 pfxlist_onlink_check();
1045 }
1046
1047 static void
1048 pfxrtr_del(struct nd_pfxrouter *pfr)
1049 {
1050
1051 ND6_WLOCK_ASSERT();
1052
1053 LIST_REMOVE(pfr, pfr_entry);
1054 defrouter_rele(pfr->router);
1055 free(pfr, M_IP6NDP);
1056 }
1057
1058 static struct nd_prefix *
1059 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1060 {
1061 struct nd_prefix *search;
1062
1063 ND6_LOCK_ASSERT();
1064
1065 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1066 if (key->ndpr_ifp == search->ndpr_ifp &&
1067 key->ndpr_plen == search->ndpr_plen &&
1068 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1069 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1070 nd6_prefix_ref(search);
1071 break;
1072 }
1073 }
1074 return (search);
1075 }
1076
1077 struct nd_prefix *
1078 nd6_prefix_lookup(struct nd_prefixctl *key)
1079 {
1080 struct nd_prefix *search;
1081
1082 ND6_RLOCK();
1083 search = nd6_prefix_lookup_locked(key);
1084 ND6_RUNLOCK();
1085 return (search);
1086 }
1087
1088 void
1089 nd6_prefix_ref(struct nd_prefix *pr)
1090 {
1091
1092 refcount_acquire(&pr->ndpr_refcnt);
1093 }
1094
1095 void
1096 nd6_prefix_rele(struct nd_prefix *pr)
1097 {
1098
1099 if (refcount_release(&pr->ndpr_refcnt)) {
1100 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1101 ("prefix %p has advertising routers", pr));
1102 free(pr, M_IP6NDP);
1103 }
1104 }
1105
1106 int
1107 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1108 struct nd_prefix **newp)
1109 {
1110 struct nd_prefix *new;
1111 char ip6buf[INET6_ADDRSTRLEN];
1112 int error;
1113
1114 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1115 if (new == NULL)
1116 return (ENOMEM);
1117 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1118 new->ndpr_ifp = pr->ndpr_ifp;
1119 new->ndpr_prefix = pr->ndpr_prefix;
1120 new->ndpr_plen = pr->ndpr_plen;
1121 new->ndpr_vltime = pr->ndpr_vltime;
1122 new->ndpr_pltime = pr->ndpr_pltime;
1123 new->ndpr_flags = pr->ndpr_flags;
1124 if ((error = in6_init_prefix_ltimes(new)) != 0) {
1125 free(new, M_IP6NDP);
1126 return (error);
1127 }
1128 new->ndpr_lastupdate = time_uptime;
1129
1130 /* initialization */
1131 LIST_INIT(&new->ndpr_advrtrs);
1132 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1133 /* make prefix in the canonical form */
1134 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1135
1136 ND6_WLOCK();
1137 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1138 V_nd6_list_genid++;
1139 ND6_WUNLOCK();
1140
1141 /* ND_OPT_PI_FLAG_ONLINK processing */
1142 if (new->ndpr_raf_onlink) {
1143 ND6_ONLINK_LOCK();
1144 if ((error = nd6_prefix_onlink(new)) != 0) {
1145 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
1146 "the prefix %s/%d on-link on %s (errno=%d)\n",
1147 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1148 pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1149 /* proceed anyway. XXX: is it correct? */
1150 }
1151 ND6_ONLINK_UNLOCK();
1152 }
1153
1154 if (dr != NULL)
1155 pfxrtr_add(new, dr);
1156 if (newp != NULL)
1157 *newp = new;
1158 return (0);
1159 }
1160
1161 /*
1162 * Remove a prefix from the prefix list and optionally stash it in a
1163 * caller-provided list.
1164 *
1165 * The ND6 lock must be held.
1166 */
1167 void
1168 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1169 {
1170
1171 ND6_WLOCK_ASSERT();
1172
1173 LIST_REMOVE(pr, ndpr_entry);
1174 V_nd6_list_genid++;
1175 if (list != NULL)
1176 LIST_INSERT_HEAD(list, pr, ndpr_entry);
1177 }
1178
1179 /*
1180 * Free an unlinked prefix, first marking it off-link if necessary.
1181 */
1182 void
1183 nd6_prefix_del(struct nd_prefix *pr)
1184 {
1185 struct nd_pfxrouter *pfr, *next;
1186 int e;
1187 char ip6buf[INET6_ADDRSTRLEN];
1188
1189 KASSERT(pr->ndpr_addrcnt == 0,
1190 ("prefix %p has referencing addresses", pr));
1191 ND6_UNLOCK_ASSERT();
1192
1193 /*
1194 * Though these flags are now meaningless, we'd rather keep the value
1195 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1196 * when executing "ndp -p".
1197 */
1198 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1199 ND6_ONLINK_LOCK();
1200 if ((e = nd6_prefix_offlink(pr)) != 0) {
1201 nd6log((LOG_ERR,
1202 "nd6_prefix_del: failed to make %s/%d offlink "
1203 "on %s, errno=%d\n",
1204 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1205 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1206 /* what should we do? */
1207 }
1208 ND6_ONLINK_UNLOCK();
1209 }
1210
1211 /* Release references to routers that have advertised this prefix. */
1212 ND6_WLOCK();
1213 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1214 pfxrtr_del(pfr);
1215 ND6_WUNLOCK();
1216
1217 nd6_prefix_rele(pr);
1218
1219 pfxlist_onlink_check();
1220 }
1221
1222 static int
1223 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1224 struct mbuf *m, int mcast)
1225 {
1226 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1227 struct ifaddr *ifa;
1228 struct ifnet *ifp = new->ndpr_ifp;
1229 struct nd_prefix *pr;
1230 int error = 0;
1231 int auth;
1232 struct in6_addrlifetime lt6_tmp;
1233 char ip6buf[INET6_ADDRSTRLEN];
1234
1235 auth = 0;
1236 if (m) {
1237 /*
1238 * Authenticity for NA consists authentication for
1239 * both IP header and IP datagrams, doesn't it ?
1240 */
1241 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1242 auth = ((m->m_flags & M_AUTHIPHDR) &&
1243 (m->m_flags & M_AUTHIPDGM));
1244 #endif
1245 }
1246
1247 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1248 /*
1249 * nd6_prefix_lookup() ensures that pr and new have the same
1250 * prefix on a same interface.
1251 */
1252
1253 /*
1254 * Update prefix information. Note that the on-link (L) bit
1255 * and the autonomous (A) bit should NOT be changed from 1
1256 * to 0.
1257 */
1258 if (new->ndpr_raf_onlink == 1)
1259 pr->ndpr_raf_onlink = 1;
1260 if (new->ndpr_raf_auto == 1)
1261 pr->ndpr_raf_auto = 1;
1262 if (new->ndpr_raf_onlink) {
1263 pr->ndpr_vltime = new->ndpr_vltime;
1264 pr->ndpr_pltime = new->ndpr_pltime;
1265 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1266 pr->ndpr_lastupdate = time_uptime;
1267 }
1268
1269 if (new->ndpr_raf_onlink &&
1270 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1271 ND6_ONLINK_LOCK();
1272 if ((error = nd6_prefix_onlink(pr)) != 0) {
1273 nd6log((LOG_ERR,
1274 "prelist_update: failed to make "
1275 "the prefix %s/%d on-link on %s "
1276 "(errno=%d)\n",
1277 ip6_sprintf(ip6buf,
1278 &pr->ndpr_prefix.sin6_addr),
1279 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1280 error));
1281 /* proceed anyway. XXX: is it correct? */
1282 }
1283 ND6_ONLINK_UNLOCK();
1284 }
1285
1286 if (dr != NULL)
1287 pfxrtr_add(pr, dr);
1288 } else {
1289 if (new->ndpr_vltime == 0)
1290 goto end;
1291 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1292 goto end;
1293
1294 error = nd6_prelist_add(new, dr, &pr);
1295 if (error != 0) {
1296 nd6log((LOG_NOTICE, "prelist_update: "
1297 "nd6_prelist_add failed for %s/%d on %s errno=%d\n",
1298 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1299 new->ndpr_plen, if_name(new->ndpr_ifp), error));
1300 goto end; /* we should just give up in this case. */
1301 }
1302
1303 /*
1304 * XXX: from the ND point of view, we can ignore a prefix
1305 * with the on-link bit being zero. However, we need a
1306 * prefix structure for references from autoconfigured
1307 * addresses. Thus, we explicitly make sure that the prefix
1308 * itself expires now.
1309 */
1310 if (pr->ndpr_raf_onlink == 0) {
1311 pr->ndpr_vltime = 0;
1312 pr->ndpr_pltime = 0;
1313 in6_init_prefix_ltimes(pr);
1314 }
1315 }
1316
1317 /*
1318 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1319 * Note that pr must be non NULL at this point.
1320 */
1321
1322 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1323 if (!new->ndpr_raf_auto)
1324 goto end;
1325
1326 /*
1327 * 5.5.3 (b). the link-local prefix should have been ignored in
1328 * nd6_ra_input.
1329 */
1330
1331 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1332 if (new->ndpr_pltime > new->ndpr_vltime) {
1333 error = EINVAL; /* XXX: won't be used */
1334 goto end;
1335 }
1336
1337 /*
1338 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1339 * an address configured by stateless autoconfiguration already in the
1340 * list of addresses associated with the interface, and the Valid
1341 * Lifetime is not 0, form an address. We first check if we have
1342 * a matching prefix.
1343 * Note: we apply a clarification in rfc2462bis-02 here. We only
1344 * consider autoconfigured addresses while RFC2462 simply said
1345 * "address".
1346 */
1347 IF_ADDR_RLOCK(ifp);
1348 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1349 struct in6_ifaddr *ifa6;
1350 u_int32_t remaininglifetime;
1351
1352 if (ifa->ifa_addr->sa_family != AF_INET6)
1353 continue;
1354
1355 ifa6 = (struct in6_ifaddr *)ifa;
1356
1357 /*
1358 * We only consider autoconfigured addresses as per rfc2462bis.
1359 */
1360 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1361 continue;
1362
1363 /*
1364 * Spec is not clear here, but I believe we should concentrate
1365 * on unicast (i.e. not anycast) addresses.
1366 * XXX: other ia6_flags? detached or duplicated?
1367 */
1368 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1369 continue;
1370
1371 /*
1372 * Ignore the address if it is not associated with a prefix
1373 * or is associated with a prefix that is different from this
1374 * one. (pr is never NULL here)
1375 */
1376 if (ifa6->ia6_ndpr != pr)
1377 continue;
1378
1379 if (ia6_match == NULL) /* remember the first one */
1380 ia6_match = ifa6;
1381
1382 /*
1383 * An already autoconfigured address matched. Now that we
1384 * are sure there is at least one matched address, we can
1385 * proceed to 5.5.3. (e): update the lifetimes according to the
1386 * "two hours" rule and the privacy extension.
1387 * We apply some clarifications in rfc2462bis:
1388 * - use remaininglifetime instead of storedlifetime as a
1389 * variable name
1390 * - remove the dead code in the "two-hour" rule
1391 */
1392 #define TWOHOUR (120*60)
1393 lt6_tmp = ifa6->ia6_lifetime;
1394
1395 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1396 remaininglifetime = ND6_INFINITE_LIFETIME;
1397 else if (time_uptime - ifa6->ia6_updatetime >
1398 lt6_tmp.ia6t_vltime) {
1399 /*
1400 * The case of "invalid" address. We should usually
1401 * not see this case.
1402 */
1403 remaininglifetime = 0;
1404 } else
1405 remaininglifetime = lt6_tmp.ia6t_vltime -
1406 (time_uptime - ifa6->ia6_updatetime);
1407
1408 /* when not updating, keep the current stored lifetime. */
1409 lt6_tmp.ia6t_vltime = remaininglifetime;
1410
1411 if (TWOHOUR < new->ndpr_vltime ||
1412 remaininglifetime < new->ndpr_vltime) {
1413 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1414 } else if (remaininglifetime <= TWOHOUR) {
1415 if (auth) {
1416 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1417 }
1418 } else {
1419 /*
1420 * new->ndpr_vltime <= TWOHOUR &&
1421 * TWOHOUR < remaininglifetime
1422 */
1423 lt6_tmp.ia6t_vltime = TWOHOUR;
1424 }
1425
1426 /* The 2 hour rule is not imposed for preferred lifetime. */
1427 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1428
1429 in6_init_address_ltimes(pr, <6_tmp);
1430
1431 /*
1432 * We need to treat lifetimes for temporary addresses
1433 * differently, according to
1434 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1435 * we only update the lifetimes when they are in the maximum
1436 * intervals.
1437 */
1438 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1439 u_int32_t maxvltime, maxpltime;
1440
1441 if (V_ip6_temp_valid_lifetime >
1442 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1443 V_ip6_desync_factor)) {
1444 maxvltime = V_ip6_temp_valid_lifetime -
1445 (time_uptime - ifa6->ia6_createtime) -
1446 V_ip6_desync_factor;
1447 } else
1448 maxvltime = 0;
1449 if (V_ip6_temp_preferred_lifetime >
1450 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1451 V_ip6_desync_factor)) {
1452 maxpltime = V_ip6_temp_preferred_lifetime -
1453 (time_uptime - ifa6->ia6_createtime) -
1454 V_ip6_desync_factor;
1455 } else
1456 maxpltime = 0;
1457
1458 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1459 lt6_tmp.ia6t_vltime > maxvltime) {
1460 lt6_tmp.ia6t_vltime = maxvltime;
1461 }
1462 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1463 lt6_tmp.ia6t_pltime > maxpltime) {
1464 lt6_tmp.ia6t_pltime = maxpltime;
1465 }
1466 }
1467 ifa6->ia6_lifetime = lt6_tmp;
1468 ifa6->ia6_updatetime = time_uptime;
1469 }
1470 IF_ADDR_RUNLOCK(ifp);
1471 if (ia6_match == NULL && new->ndpr_vltime) {
1472 int ifidlen;
1473
1474 /*
1475 * 5.5.3 (d) (continued)
1476 * No address matched and the valid lifetime is non-zero.
1477 * Create a new address.
1478 */
1479
1480 /*
1481 * Prefix Length check:
1482 * If the sum of the prefix length and interface identifier
1483 * length does not equal 128 bits, the Prefix Information
1484 * option MUST be ignored. The length of the interface
1485 * identifier is defined in a separate link-type specific
1486 * document.
1487 */
1488 ifidlen = in6_if2idlen(ifp);
1489 if (ifidlen < 0) {
1490 /* this should not happen, so we always log it. */
1491 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1492 if_name(ifp));
1493 goto end;
1494 }
1495 if (ifidlen + pr->ndpr_plen != 128) {
1496 nd6log((LOG_INFO,
1497 "prelist_update: invalid prefixlen "
1498 "%d for %s, ignored\n",
1499 pr->ndpr_plen, if_name(ifp)));
1500 goto end;
1501 }
1502
1503 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1504 /*
1505 * note that we should use pr (not new) for reference.
1506 */
1507 pr->ndpr_addrcnt++;
1508 ia6->ia6_ndpr = pr;
1509
1510 /*
1511 * RFC 3041 3.3 (2).
1512 * When a new public address is created as described
1513 * in RFC2462, also create a new temporary address.
1514 *
1515 * RFC 3041 3.5.
1516 * When an interface connects to a new link, a new
1517 * randomized interface identifier should be generated
1518 * immediately together with a new set of temporary
1519 * addresses. Thus, we specifiy 1 as the 2nd arg of
1520 * in6_tmpifadd().
1521 */
1522 if (V_ip6_use_tempaddr) {
1523 int e;
1524 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1525 nd6log((LOG_NOTICE, "prelist_update: "
1526 "failed to create a temporary "
1527 "address, errno=%d\n",
1528 e));
1529 }
1530 }
1531 ifa_free(&ia6->ia_ifa);
1532
1533 /*
1534 * A newly added address might affect the status
1535 * of other addresses, so we check and update it.
1536 * XXX: what if address duplication happens?
1537 */
1538 pfxlist_onlink_check();
1539 } else {
1540 /* just set an error. do not bark here. */
1541 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1542 }
1543 }
1544
1545 end:
1546 if (pr != NULL)
1547 nd6_prefix_rele(pr);
1548 return (error);
1549 }
1550
1551 /*
1552 * A supplement function used in the on-link detection below;
1553 * detect if a given prefix has a (probably) reachable advertising router.
1554 * XXX: lengthy function name...
1555 */
1556 static struct nd_pfxrouter *
1557 find_pfxlist_reachable_router(struct nd_prefix *pr)
1558 {
1559 struct nd_pfxrouter *pfxrtr;
1560 struct llentry *ln;
1561 int canreach;
1562
1563 ND6_LOCK_ASSERT();
1564
1565 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1566 IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1567 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1568 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1569 if (ln == NULL)
1570 continue;
1571 canreach = ND6_IS_LLINFO_PROBREACH(ln);
1572 LLE_RUNLOCK(ln);
1573 if (canreach)
1574 break;
1575 }
1576 return (pfxrtr);
1577 }
1578
1579 /*
1580 * Check if each prefix in the prefix list has at least one available router
1581 * that advertised the prefix (a router is "available" if its neighbor cache
1582 * entry is reachable or probably reachable).
1583 * If the check fails, the prefix may be off-link, because, for example,
1584 * we have moved from the network but the lifetime of the prefix has not
1585 * expired yet. So we should not use the prefix if there is another prefix
1586 * that has an available router.
1587 * But, if there is no prefix that has an available router, we still regard
1588 * all the prefixes as on-link. This is because we can't tell if all the
1589 * routers are simply dead or if we really moved from the network and there
1590 * is no router around us.
1591 */
1592 void
1593 pfxlist_onlink_check(void)
1594 {
1595 struct nd_prefix *pr;
1596 struct in6_ifaddr *ifa;
1597 struct nd_defrouter *dr;
1598 struct nd_pfxrouter *pfxrtr = NULL;
1599 struct rm_priotracker in6_ifa_tracker;
1600 uint64_t genid;
1601 uint32_t flags;
1602
1603 ND6_ONLINK_LOCK();
1604 ND6_RLOCK();
1605
1606 /*
1607 * Check if there is a prefix that has a reachable advertising
1608 * router.
1609 */
1610 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1611 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1612 break;
1613 }
1614
1615 /*
1616 * If we have no such prefix, check whether we still have a router
1617 * that does not advertise any prefixes.
1618 */
1619 if (pr == NULL) {
1620 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1621 struct nd_prefix *pr0;
1622
1623 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1624 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1625 break;
1626 }
1627 if (pfxrtr != NULL)
1628 break;
1629 }
1630 }
1631 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1632 /*
1633 * There is at least one prefix that has a reachable router,
1634 * or at least a router which probably does not advertise
1635 * any prefixes. The latter would be the case when we move
1636 * to a new link where we have a router that does not provide
1637 * prefixes and we configure an address by hand.
1638 * Detach prefixes which have no reachable advertising
1639 * router, and attach other prefixes.
1640 */
1641 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1642 /* XXX: a link-local prefix should never be detached */
1643 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1644 pr->ndpr_raf_onlink == 0 ||
1645 pr->ndpr_raf_auto == 0)
1646 continue;
1647
1648 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1649 find_pfxlist_reachable_router(pr) == NULL)
1650 pr->ndpr_stateflags |= NDPRF_DETACHED;
1651 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1652 find_pfxlist_reachable_router(pr) != NULL)
1653 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1654 }
1655 } else {
1656 /* there is no prefix that has a reachable router */
1657 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1658 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1659 pr->ndpr_raf_onlink == 0 ||
1660 pr->ndpr_raf_auto == 0)
1661 continue;
1662 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1663 }
1664 }
1665
1666 /*
1667 * Remove each interface route associated with a (just) detached
1668 * prefix, and reinstall the interface route for a (just) attached
1669 * prefix. Note that all attempt of reinstallation does not
1670 * necessarily success, when a same prefix is shared among multiple
1671 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1672 * so we don't have to care about them.
1673 */
1674 restart:
1675 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1676 char ip6buf[INET6_ADDRSTRLEN];
1677 int e;
1678
1679 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1680 pr->ndpr_raf_onlink == 0 ||
1681 pr->ndpr_raf_auto == 0)
1682 continue;
1683
1684 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1685 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1686 genid = V_nd6_list_genid;
1687 ND6_RUNLOCK();
1688 if ((flags & NDPRF_ONLINK) != 0 &&
1689 (e = nd6_prefix_offlink(pr)) != 0) {
1690 nd6log((LOG_ERR,
1691 "pfxlist_onlink_check: failed to "
1692 "make %s/%d offlink, errno=%d\n",
1693 ip6_sprintf(ip6buf,
1694 &pr->ndpr_prefix.sin6_addr),
1695 pr->ndpr_plen, e));
1696 } else if ((flags & NDPRF_ONLINK) == 0 &&
1697 (e = nd6_prefix_onlink(pr)) != 0) {
1698 nd6log((LOG_ERR,
1699 "pfxlist_onlink_check: failed to "
1700 "make %s/%d onlink, errno=%d\n",
1701 ip6_sprintf(ip6buf,
1702 &pr->ndpr_prefix.sin6_addr),
1703 pr->ndpr_plen, e));
1704 }
1705 ND6_RLOCK();
1706 if (genid != V_nd6_list_genid)
1707 goto restart;
1708 }
1709 }
1710
1711 /*
1712 * Changes on the prefix status might affect address status as well.
1713 * Make sure that all addresses derived from an attached prefix are
1714 * attached, and that all addresses derived from a detached prefix are
1715 * detached. Note, however, that a manually configured address should
1716 * always be attached.
1717 * The precise detection logic is same as the one for prefixes.
1718 */
1719 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1720 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1721 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1722 continue;
1723
1724 if (ifa->ia6_ndpr == NULL) {
1725 /*
1726 * This can happen when we first configure the address
1727 * (i.e. the address exists, but the prefix does not).
1728 * XXX: complicated relationships...
1729 */
1730 continue;
1731 }
1732
1733 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1734 break;
1735 }
1736 if (ifa) {
1737 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1738 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1739 continue;
1740
1741 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1742 continue;
1743
1744 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1745 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1746 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1747 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1748 nd6_dad_start((struct ifaddr *)ifa, 0);
1749 }
1750 } else {
1751 ifa->ia6_flags |= IN6_IFF_DETACHED;
1752 }
1753 }
1754 } else {
1755 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1756 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1757 continue;
1758
1759 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1760 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1761 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1762 /* Do we need a delay in this case? */
1763 nd6_dad_start((struct ifaddr *)ifa, 0);
1764 }
1765 }
1766 }
1767 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1768 ND6_RUNLOCK();
1769 ND6_ONLINK_UNLOCK();
1770 }
1771
1772 static int
1773 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1774 {
1775 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1776 struct rib_head *rnh;
1777 struct rtentry *rt;
1778 struct sockaddr_in6 mask6;
1779 u_long rtflags;
1780 int error, a_failure, fibnum, maxfib;
1781
1782 /*
1783 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1784 * ifa->ifa_rtrequest = nd6_rtrequest;
1785 */
1786 bzero(&mask6, sizeof(mask6));
1787 mask6.sin6_len = sizeof(mask6);
1788 mask6.sin6_addr = pr->ndpr_mask;
1789 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1790
1791 if(V_rt_add_addr_allfibs) {
1792 fibnum = 0;
1793 maxfib = rt_numfibs;
1794 } else {
1795 fibnum = ifa->ifa_ifp->if_fib;
1796 maxfib = fibnum + 1;
1797 }
1798 a_failure = 0;
1799 for (; fibnum < maxfib; fibnum++) {
1800
1801 rt = NULL;
1802 error = in6_rtrequest(RTM_ADD,
1803 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1804 (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1805 if (error == 0) {
1806 KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1807 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1808 error, pr, ifa));
1809
1810 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1811 /* XXX what if rhn == NULL? */
1812 RIB_WLOCK(rnh);
1813 RT_LOCK(rt);
1814 if (rt_setgate(rt, rt_key(rt),
1815 (struct sockaddr *)&null_sdl) == 0) {
1816 struct sockaddr_dl *dl;
1817
1818 dl = (struct sockaddr_dl *)rt->rt_gateway;
1819 dl->sdl_type = rt->rt_ifp->if_type;
1820 dl->sdl_index = rt->rt_ifp->if_index;
1821 }
1822 RIB_WUNLOCK(rnh);
1823 nd6_rtmsg(RTM_ADD, rt);
1824 RT_UNLOCK(rt);
1825 pr->ndpr_stateflags |= NDPRF_ONLINK;
1826 } else {
1827 char ip6buf[INET6_ADDRSTRLEN];
1828 char ip6bufg[INET6_ADDRSTRLEN];
1829 char ip6bufm[INET6_ADDRSTRLEN];
1830 struct sockaddr_in6 *sin6;
1831
1832 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1833 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1834 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1835 "flags=%lx errno = %d\n",
1836 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1837 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1838 ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1839 ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1840 rtflags, error));
1841
1842 /* Save last error to return, see rtinit(). */
1843 a_failure = error;
1844 }
1845
1846 if (rt != NULL) {
1847 RT_LOCK(rt);
1848 RT_REMREF(rt);
1849 RT_UNLOCK(rt);
1850 }
1851 }
1852
1853 /* Return the last error we got. */
1854 return (a_failure);
1855 }
1856
1857 int
1858 nd6_prefix_onlink(struct nd_prefix *pr)
1859 {
1860 struct ifaddr *ifa;
1861 struct ifnet *ifp = pr->ndpr_ifp;
1862 struct nd_prefix *opr;
1863 char ip6buf[INET6_ADDRSTRLEN];
1864 int error;
1865
1866 ND6_ONLINK_LOCK_ASSERT();
1867 ND6_UNLOCK_ASSERT();
1868
1869 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1870 return (EEXIST);
1871
1872 /*
1873 * Add the interface route associated with the prefix. Before
1874 * installing the route, check if there's the same prefix on another
1875 * interface, and the prefix has already installed the interface route.
1876 * Although such a configuration is expected to be rare, we explicitly
1877 * allow it.
1878 */
1879 ND6_RLOCK();
1880 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1881 if (opr == pr)
1882 continue;
1883
1884 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1885 continue;
1886
1887 if (!V_rt_add_addr_allfibs &&
1888 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
1889 continue;
1890
1891 if (opr->ndpr_plen == pr->ndpr_plen &&
1892 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1893 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1894 ND6_RUNLOCK();
1895 return (0);
1896 }
1897 }
1898 ND6_RUNLOCK();
1899
1900 /*
1901 * We prefer link-local addresses as the associated interface address.
1902 */
1903 /* search for a link-local addr */
1904 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1905 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1906 if (ifa == NULL) {
1907 /* XXX: freebsd does not have ifa_ifwithaf */
1908 IF_ADDR_RLOCK(ifp);
1909 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1910 if (ifa->ifa_addr->sa_family == AF_INET6) {
1911 ifa_ref(ifa);
1912 break;
1913 }
1914 }
1915 IF_ADDR_RUNLOCK(ifp);
1916 /* should we care about ia6_flags? */
1917 }
1918 if (ifa == NULL) {
1919 /*
1920 * This can still happen, when, for example, we receive an RA
1921 * containing a prefix with the L bit set and the A bit clear,
1922 * after removing all IPv6 addresses on the receiving
1923 * interface. This should, of course, be rare though.
1924 */
1925 nd6log((LOG_NOTICE,
1926 "nd6_prefix_onlink: failed to find any ifaddr"
1927 " to add route for a prefix(%s/%d) on %s\n",
1928 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1929 pr->ndpr_plen, if_name(ifp)));
1930 return (0);
1931 }
1932
1933 error = nd6_prefix_onlink_rtrequest(pr, ifa);
1934
1935 if (ifa != NULL)
1936 ifa_free(ifa);
1937
1938 return (error);
1939 }
1940
1941 int
1942 nd6_prefix_offlink(struct nd_prefix *pr)
1943 {
1944 int error = 0;
1945 struct ifnet *ifp = pr->ndpr_ifp;
1946 struct nd_prefix *opr;
1947 struct sockaddr_in6 sa6, mask6;
1948 struct rtentry *rt;
1949 char ip6buf[INET6_ADDRSTRLEN];
1950 uint64_t genid;
1951 int fibnum, maxfib, a_failure;
1952
1953 ND6_ONLINK_LOCK_ASSERT();
1954 ND6_UNLOCK_ASSERT();
1955
1956 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1957 return (EEXIST);
1958
1959 bzero(&sa6, sizeof(sa6));
1960 sa6.sin6_family = AF_INET6;
1961 sa6.sin6_len = sizeof(sa6);
1962 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1963 sizeof(struct in6_addr));
1964 bzero(&mask6, sizeof(mask6));
1965 mask6.sin6_family = AF_INET6;
1966 mask6.sin6_len = sizeof(sa6);
1967 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1968
1969 if (V_rt_add_addr_allfibs) {
1970 fibnum = 0;
1971 maxfib = rt_numfibs;
1972 } else {
1973 fibnum = ifp->if_fib;
1974 maxfib = fibnum + 1;
1975 }
1976
1977 a_failure = 0;
1978 for (; fibnum < maxfib; fibnum++) {
1979 rt = NULL;
1980 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1981 (struct sockaddr *)&mask6, 0, &rt, fibnum);
1982 if (error == 0) {
1983 /* report the route deletion to the routing socket. */
1984 if (rt != NULL)
1985 nd6_rtmsg(RTM_DELETE, rt);
1986 } else {
1987 /* Save last error to return, see rtinit(). */
1988 a_failure = error;
1989 }
1990 if (rt != NULL) {
1991 RTFREE(rt);
1992 }
1993 }
1994 error = a_failure;
1995 a_failure = 1;
1996 if (error == 0) {
1997 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1998
1999 /*
2000 * There might be the same prefix on another interface,
2001 * the prefix which could not be on-link just because we have
2002 * the interface route (see comments in nd6_prefix_onlink).
2003 * If there's one, try to make the prefix on-link on the
2004 * interface.
2005 */
2006 ND6_RLOCK();
2007 restart:
2008 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2009 /*
2010 * KAME specific: detached prefixes should not be
2011 * on-link.
2012 */
2013 if (opr == pr || (opr->ndpr_stateflags &
2014 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2015 continue;
2016
2017 if (opr->ndpr_plen == pr->ndpr_plen &&
2018 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2019 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2020 int e;
2021
2022 genid = V_nd6_list_genid;
2023 ND6_RUNLOCK();
2024 if ((e = nd6_prefix_onlink(opr)) != 0) {
2025 nd6log((LOG_ERR,
2026 "nd6_prefix_offlink: failed to "
2027 "recover a prefix %s/%d from %s "
2028 "to %s (errno = %d)\n",
2029 ip6_sprintf(ip6buf,
2030 &opr->ndpr_prefix.sin6_addr),
2031 opr->ndpr_plen, if_name(ifp),
2032 if_name(opr->ndpr_ifp), e));
2033 } else
2034 a_failure = 0;
2035 ND6_RLOCK();
2036 if (genid != V_nd6_list_genid)
2037 goto restart;
2038 }
2039 }
2040 ND6_RUNLOCK();
2041 } else {
2042 /* XXX: can we still set the NDPRF_ONLINK flag? */
2043 nd6log((LOG_ERR,
2044 "nd6_prefix_offlink: failed to delete route: "
2045 "%s/%d on %s (errno = %d)\n",
2046 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
2047 if_name(ifp), error));
2048 }
2049
2050 if (a_failure)
2051 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6,
2052 (struct sockaddr *)&mask6, LLE_STATIC);
2053
2054 return (error);
2055 }
2056
2057 static struct in6_ifaddr *
2058 in6_ifadd(struct nd_prefixctl *pr, int mcast)
2059 {
2060 struct ifnet *ifp = pr->ndpr_ifp;
2061 struct ifaddr *ifa;
2062 struct in6_aliasreq ifra;
2063 struct in6_ifaddr *ia, *ib;
2064 int error, plen0;
2065 struct in6_addr mask;
2066 int prefixlen = pr->ndpr_plen;
2067 int updateflags;
2068 char ip6buf[INET6_ADDRSTRLEN];
2069
2070 in6_prefixlen2mask(&mask, prefixlen);
2071
2072 /*
2073 * find a link-local address (will be interface ID).
2074 * Is it really mandatory? Theoretically, a global or a site-local
2075 * address can be configured without a link-local address, if we
2076 * have a unique interface identifier...
2077 *
2078 * it is not mandatory to have a link-local address, we can generate
2079 * interface identifier on the fly. we do this because:
2080 * (1) it should be the easiest way to find interface identifier.
2081 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
2082 * for multiple addresses on a single interface, and possible shortcut
2083 * of DAD. we omitted DAD for this reason in the past.
2084 * (3) a user can prevent autoconfiguration of global address
2085 * by removing link-local address by hand (this is partly because we
2086 * don't have other way to control the use of IPv6 on an interface.
2087 * this has been our design choice - cf. NRL's "ifconfig auto").
2088 * (4) it is easier to manage when an interface has addresses
2089 * with the same interface identifier, than to have multiple addresses
2090 * with different interface identifiers.
2091 */
2092 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
2093 if (ifa)
2094 ib = (struct in6_ifaddr *)ifa;
2095 else
2096 return NULL;
2097
2098 /* prefixlen + ifidlen must be equal to 128 */
2099 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
2100 if (prefixlen != plen0) {
2101 ifa_free(ifa);
2102 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
2103 "(prefix=%d ifid=%d)\n",
2104 if_name(ifp), prefixlen, 128 - plen0));
2105 return NULL;
2106 }
2107
2108 /* make ifaddr */
2109 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
2110
2111 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
2112 /* interface ID */
2113 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
2114 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
2115 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
2116 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
2117 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2118 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
2119 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2120 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
2121 ifa_free(ifa);
2122
2123 /* lifetimes. */
2124 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
2125 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
2126
2127 /* XXX: scope zone ID? */
2128
2129 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
2130
2131 /*
2132 * Make sure that we do not have this address already. This should
2133 * usually not happen, but we can still see this case, e.g., if we
2134 * have manually configured the exact address to be configured.
2135 */
2136 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
2137 &ifra.ifra_addr.sin6_addr);
2138 if (ifa != NULL) {
2139 ifa_free(ifa);
2140 /* this should be rare enough to make an explicit log */
2141 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
2142 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
2143 return (NULL);
2144 }
2145
2146 /*
2147 * Allocate ifaddr structure, link into chain, etc.
2148 * If we are going to create a new address upon receiving a multicasted
2149 * RA, we need to impose a random delay before starting DAD.
2150 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
2151 */
2152 updateflags = 0;
2153 if (mcast)
2154 updateflags |= IN6_IFAUPDATE_DADDELAY;
2155 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
2156 nd6log((LOG_ERR,
2157 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
2158 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
2159 if_name(ifp), error));
2160 return (NULL); /* ifaddr must not have been allocated. */
2161 }
2162
2163 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2164 /*
2165 * XXXRW: Assumption of non-NULLness here might not be true with
2166 * fine-grained locking -- should we validate it? Or just return
2167 * earlier ifa rather than looking it up again?
2168 */
2169 return (ia); /* this is always non-NULL and referenced. */
2170 }
2171
2172 /*
2173 * ia0 - corresponding public address
2174 */
2175 int
2176 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2177 {
2178 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2179 struct in6_ifaddr *newia;
2180 struct in6_aliasreq ifra;
2181 int error;
2182 int trylimit = 3; /* XXX: adhoc value */
2183 int updateflags;
2184 u_int32_t randid[2];
2185 time_t vltime0, pltime0;
2186
2187 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2188 &ia0->ia_prefixmask.sin6_addr);
2189
2190 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */
2191 /* clear the old IFID */
2192 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2193 &ifra.ifra_prefixmask.sin6_addr);
2194
2195 again:
2196 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2197 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2198 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
2199 "random IFID\n"));
2200 return (EINVAL);
2201 }
2202 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2203 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2204 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2205 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2206
2207 /*
2208 * in6_get_tmpifid() quite likely provided a unique interface ID.
2209 * However, we may still have a chance to see collision, because
2210 * there may be a time lag between generation of the ID and generation
2211 * of the address. So, we'll do one more sanity check.
2212 */
2213
2214 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2215 if (trylimit-- > 0) {
2216 forcegen = 1;
2217 goto again;
2218 }
2219
2220 /* Give up. Something strange should have happened. */
2221 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
2222 "find a unique random IFID\n"));
2223 return (EEXIST);
2224 }
2225
2226 /*
2227 * The Valid Lifetime is the lower of the Valid Lifetime of the
2228 * public address or TEMP_VALID_LIFETIME.
2229 * The Preferred Lifetime is the lower of the Preferred Lifetime
2230 * of the public address or TEMP_PREFERRED_LIFETIME -
2231 * DESYNC_FACTOR.
2232 */
2233 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2234 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2235 (ia0->ia6_lifetime.ia6t_vltime -
2236 (time_uptime - ia0->ia6_updatetime));
2237 if (vltime0 > V_ip6_temp_valid_lifetime)
2238 vltime0 = V_ip6_temp_valid_lifetime;
2239 } else
2240 vltime0 = V_ip6_temp_valid_lifetime;
2241 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2242 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2243 (ia0->ia6_lifetime.ia6t_pltime -
2244 (time_uptime - ia0->ia6_updatetime));
2245 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2246 pltime0 = V_ip6_temp_preferred_lifetime -
2247 V_ip6_desync_factor;
2248 }
2249 } else
2250 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2251 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2252 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2253
2254 /*
2255 * A temporary address is created only if this calculated Preferred
2256 * Lifetime is greater than REGEN_ADVANCE time units.
2257 */
2258 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2259 return (0);
2260
2261 /* XXX: scope zone ID? */
2262
2263 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2264
2265 /* allocate ifaddr structure, link into chain, etc. */
2266 updateflags = 0;
2267 if (delay)
2268 updateflags |= IN6_IFAUPDATE_DADDELAY;
2269 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2270 return (error);
2271
2272 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2273 if (newia == NULL) { /* XXX: can it happen? */
2274 nd6log((LOG_ERR,
2275 "in6_tmpifadd: ifa update succeeded, but we got "
2276 "no ifaddr\n"));
2277 return (EINVAL); /* XXX */
2278 }
2279 newia->ia6_ndpr = ia0->ia6_ndpr;
2280 newia->ia6_ndpr->ndpr_addrcnt++;
2281 ifa_free(&newia->ia_ifa);
2282
2283 /*
2284 * A newly added address might affect the status of other addresses.
2285 * XXX: when the temporary address is generated with a new public
2286 * address, the onlink check is redundant. However, it would be safe
2287 * to do the check explicitly everywhere a new address is generated,
2288 * and, in fact, we surely need the check when we create a new
2289 * temporary address due to deprecation of an old temporary address.
2290 */
2291 pfxlist_onlink_check();
2292
2293 return (0);
2294 }
2295
2296 static int
2297 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2298 {
2299 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2300 ndpr->ndpr_preferred = 0;
2301 else
2302 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2303 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2304 ndpr->ndpr_expire = 0;
2305 else
2306 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2307
2308 return 0;
2309 }
2310
2311 static void
2312 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2313 {
2314 /* init ia6t_expire */
2315 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2316 lt6->ia6t_expire = 0;
2317 else {
2318 lt6->ia6t_expire = time_uptime;
2319 lt6->ia6t_expire += lt6->ia6t_vltime;
2320 }
2321
2322 /* init ia6t_preferred */
2323 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2324 lt6->ia6t_preferred = 0;
2325 else {
2326 lt6->ia6t_preferred = time_uptime;
2327 lt6->ia6t_preferred += lt6->ia6t_pltime;
2328 }
2329 }
2330
2331 /*
2332 * Delete all the routing table entries that use the specified gateway.
2333 * XXX: this function causes search through all entries of routing table, so
2334 * it shouldn't be called when acting as a router.
2335 */
2336 void
2337 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2338 {
2339
2340 /* We'll care only link-local addresses */
2341 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2342 return;
2343
2344 /* XXX Do we really need to walk any but the default FIB? */
2345 rt_foreach_fib_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2346 }
2347
2348 static int
2349 rt6_deleteroute(const struct rtentry *rt, void *arg)
2350 {
2351 #define SIN6(s) ((struct sockaddr_in6 *)s)
2352 struct in6_addr *gate = (struct in6_addr *)arg;
2353
2354 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2355 return (0);
2356
2357 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2358 return (0);
2359 }
2360
2361 /*
2362 * Do not delete a static route.
2363 * XXX: this seems to be a bit ad-hoc. Should we consider the
2364 * 'cloned' bit instead?
2365 */
2366 if ((rt->rt_flags & RTF_STATIC) != 0)
2367 return (0);
2368
2369 /*
2370 * We delete only host route. This means, in particular, we don't
2371 * delete default route.
2372 */
2373 if ((rt->rt_flags & RTF_HOST) == 0)
2374 return (0);
2375
2376 return (1);
2377 #undef SIN6
2378 }
2379
2380 int
2381 nd6_setdefaultiface(int ifindex)
2382 {
2383 int error = 0;
2384
2385 if (ifindex < 0 || V_if_index < ifindex)
2386 return (EINVAL);
2387 if (ifindex != 0 && !ifnet_byindex(ifindex))
2388 return (EINVAL);
2389
2390 if (V_nd6_defifindex != ifindex) {
2391 V_nd6_defifindex = ifindex;
2392 if (V_nd6_defifindex > 0)
2393 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2394 else
2395 V_nd6_defifp = NULL;
2396
2397 /*
2398 * Our current implementation assumes one-to-one maping between
2399 * interfaces and links, so it would be natural to use the
2400 * default interface as the default link.
2401 */
2402 scope6_setdefault(V_nd6_defifp);
2403 }
2404
2405 return (error);
2406 }
Cache object: c190c12490f45d525daae2ab79a8ca97
|