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 #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/sysctl.h>
52 #include <sys/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/random.h>
55
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_private.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/route.h>
62 #include <net/route/nhop.h>
63 #include <net/route/route_ctl.h>
64 #include <net/radix.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68 #include <net/if_llatbl.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/nd6.h>
74 #include <netinet/icmp6.h>
75 #include <netinet6/scope6_var.h>
76
77 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
78 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
79 struct mbuf *, int);
80 static int nd6_prefix_onlink(struct nd_prefix *);
81 static int in6_get_tmp_ifid(struct in6_aliasreq *);
82
83 TAILQ_HEAD(nd6_drhead, nd_defrouter);
84 VNET_DEFINE_STATIC(struct nd6_drhead, nd6_defrouter);
85 #define V_nd6_defrouter VNET(nd6_defrouter)
86
87 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
88 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
89
90 VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp);
91 VNET_DEFINE(int, nd6_defifindex);
92 #define V_nd6_defifp VNET(nd6_defifp)
93
94 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
95
96 VNET_DEFINE(int, ip6_desync_factor);
97 VNET_DEFINE(uint32_t, ip6_temp_max_desync_factor) = TEMP_MAX_DESYNC_FACTOR_BASE;
98 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
99 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
100
101 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
102
103 #ifdef EXPERIMENTAL
104 VNET_DEFINE_STATIC(int, nd6_ignore_ipv6_only_ra) = 1;
105 #define V_nd6_ignore_ipv6_only_ra VNET(nd6_ignore_ipv6_only_ra)
106 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO,
107 nd6_ignore_ipv6_only_ra, CTLFLAG_VNET | CTLFLAG_RW,
108 &VNET_NAME(nd6_ignore_ipv6_only_ra), 0,
109 "Ignore the 'IPv6-Only flag' in RA messages in compliance with "
110 "draft-ietf-6man-ipv6only-flag");
111 #endif
112
113 /* RTPREF_MEDIUM has to be 0! */
114 #define RTPREF_HIGH 1
115 #define RTPREF_MEDIUM 0
116 #define RTPREF_LOW (-1)
117 #define RTPREF_RESERVED (-2)
118 #define RTPREF_INVALID (-3) /* internal */
119
120 static void
defrouter_ref(struct nd_defrouter * dr)121 defrouter_ref(struct nd_defrouter *dr)
122 {
123
124 refcount_acquire(&dr->refcnt);
125 }
126
127 void
defrouter_rele(struct nd_defrouter * dr)128 defrouter_rele(struct nd_defrouter *dr)
129 {
130
131 if (refcount_release(&dr->refcnt))
132 free(dr, M_IP6NDP);
133 }
134
135 /*
136 * Remove a router from the global list and optionally stash it in a
137 * caller-supplied queue.
138 */
139 static void
defrouter_unlink(struct nd_defrouter * dr,struct nd6_drhead * drq)140 defrouter_unlink(struct nd_defrouter *dr, struct nd6_drhead *drq)
141 {
142
143 ND6_WLOCK_ASSERT();
144
145 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
146 V_nd6_list_genid++;
147 if (drq != NULL)
148 TAILQ_INSERT_TAIL(drq, dr, dr_entry);
149 }
150
151 /*
152 * Receive Router Solicitation Message - just for routers.
153 * Router solicitation/advertisement is mostly managed by userland program
154 * (rtadvd) so here we have no function like nd6_ra_output().
155 *
156 * Based on RFC 2461
157 */
158 void
nd6_rs_input(struct mbuf * m,int off,int icmp6len)159 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
160 {
161 struct ifnet *ifp;
162 struct ip6_hdr *ip6;
163 struct nd_router_solicit *nd_rs;
164 struct in6_addr saddr6;
165 union nd_opts ndopts;
166 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
167 char *lladdr;
168 int lladdrlen;
169
170 ifp = m->m_pkthdr.rcvif;
171
172 /*
173 * Accept RS only when V_ip6_forwarding=1 and the interface has
174 * no ND6_IFF_ACCEPT_RTADV.
175 */
176 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
177 goto freeit;
178
179 /* RFC 6980: Nodes MUST silently ignore fragments */
180 if(m->m_flags & M_FRAGMENTED)
181 goto freeit;
182
183 /* Sanity checks */
184 ip6 = mtod(m, struct ip6_hdr *);
185 if (__predict_false(ip6->ip6_hlim != 255)) {
186 ICMP6STAT_INC(icp6s_invlhlim);
187 nd6log((LOG_ERR,
188 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
189 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
190 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
191 goto bad;
192 }
193
194 /*
195 * Don't update the neighbor cache, if src = ::.
196 * This indicates that the src has no IP address assigned yet.
197 */
198 saddr6 = ip6->ip6_src;
199 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
200 goto freeit;
201
202 if (m->m_len < off + icmp6len) {
203 m = m_pullup(m, off + icmp6len);
204 if (m == NULL) {
205 IP6STAT_INC(ip6s_exthdrtoolong);
206 return;
207 }
208 }
209 ip6 = mtod(m, struct ip6_hdr *);
210 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
211
212 icmp6len -= sizeof(*nd_rs);
213 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
214 if (nd6_options(&ndopts) < 0) {
215 nd6log((LOG_INFO,
216 "%s: invalid ND option, ignored\n", __func__));
217 /* nd6_options have incremented stats */
218 goto freeit;
219 }
220
221 lladdr = NULL;
222 lladdrlen = 0;
223 if (ndopts.nd_opts_src_lladdr) {
224 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
225 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
226 }
227
228 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
229 nd6log((LOG_INFO,
230 "%s: lladdrlen mismatch for %s (if %d, RS packet %d)\n",
231 __func__, ip6_sprintf(ip6bufs, &saddr6),
232 ifp->if_addrlen, lladdrlen - 2));
233 goto bad;
234 }
235
236 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
237
238 freeit:
239 m_freem(m);
240 return;
241
242 bad:
243 ICMP6STAT_INC(icp6s_badrs);
244 m_freem(m);
245 }
246
247 #ifdef EXPERIMENTAL
248 /*
249 * An initial update routine for draft-ietf-6man-ipv6only-flag.
250 * We need to iterate over all default routers for the given
251 * interface to see whether they are all advertising the "S"
252 * (IPv6-Only) flag. If they do set, otherwise unset, the
253 * interface flag we later use to filter on.
254 */
255 static void
defrtr_ipv6_only_ifp(struct ifnet * ifp)256 defrtr_ipv6_only_ifp(struct ifnet *ifp)
257 {
258 struct nd_defrouter *dr;
259 bool ipv6_only, ipv6_only_old;
260 #ifdef INET
261 struct epoch_tracker et;
262 struct ifaddr *ifa;
263 bool has_ipv4_addr;
264 #endif
265
266 if (V_nd6_ignore_ipv6_only_ra != 0)
267 return;
268
269 ipv6_only = true;
270 ND6_RLOCK();
271 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
272 if (dr->ifp == ifp &&
273 (dr->raflags & ND_RA_FLAG_IPV6_ONLY) == 0)
274 ipv6_only = false;
275 ND6_RUNLOCK();
276
277 IF_AFDATA_WLOCK(ifp);
278 ipv6_only_old = ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY;
279 IF_AFDATA_WUNLOCK(ifp);
280
281 /* If nothing changed, we have an early exit. */
282 if (ipv6_only == ipv6_only_old)
283 return;
284
285 #ifdef INET
286 /*
287 * Should we want to set the IPV6-ONLY flag, check if the
288 * interface has a non-0/0 and non-link-local IPv4 address
289 * configured on it. If it has we will assume working
290 * IPv4 operations and will clear the interface flag.
291 */
292 has_ipv4_addr = false;
293 if (ipv6_only) {
294 NET_EPOCH_ENTER(et);
295 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
296 if (ifa->ifa_addr->sa_family != AF_INET)
297 continue;
298 if (in_canforward(
299 satosin(ifa->ifa_addr)->sin_addr)) {
300 has_ipv4_addr = true;
301 break;
302 }
303 }
304 NET_EPOCH_EXIT(et);
305 }
306 if (ipv6_only && has_ipv4_addr) {
307 log(LOG_NOTICE, "%s rcvd RA w/ IPv6-Only flag set but has IPv4 "
308 "configured, ignoring IPv6-Only flag.\n", ifp->if_xname);
309 ipv6_only = false;
310 }
311 #endif
312
313 IF_AFDATA_WLOCK(ifp);
314 if (ipv6_only)
315 ND_IFINFO(ifp)->flags |= ND6_IFF_IPV6_ONLY;
316 else
317 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
318 IF_AFDATA_WUNLOCK(ifp);
319
320 #ifdef notyet
321 /* Send notification of flag change. */
322 #endif
323 }
324
325 static void
defrtr_ipv6_only_ipf_down(struct ifnet * ifp)326 defrtr_ipv6_only_ipf_down(struct ifnet *ifp)
327 {
328
329 IF_AFDATA_WLOCK(ifp);
330 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
331 IF_AFDATA_WUNLOCK(ifp);
332 }
333 #endif /* EXPERIMENTAL */
334
335 void
nd6_ifnet_link_event(void * arg __unused,struct ifnet * ifp,int linkstate)336 nd6_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate)
337 {
338
339 /*
340 * XXX-BZ we might want to trigger re-evaluation of our default router
341 * availability. E.g., on link down the default router might be
342 * unreachable but a different interface might still have connectivity.
343 */
344
345 #ifdef EXPERIMENTAL
346 if (linkstate == LINK_STATE_DOWN)
347 defrtr_ipv6_only_ipf_down(ifp);
348 #endif
349 }
350
351 /*
352 * Receive Router Advertisement Message.
353 *
354 * Based on RFC 2461
355 * TODO: on-link bit on prefix information
356 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
357 */
358 void
nd6_ra_input(struct mbuf * m,int off,int icmp6len)359 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
360 {
361 struct ifnet *ifp;
362 struct nd_ifinfo *ndi;
363 struct ip6_hdr *ip6;
364 struct nd_router_advert *nd_ra;
365 struct in6_addr saddr6;
366 struct nd_defrouter *dr;
367 union nd_opts ndopts;
368 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
369 int mcast;
370
371 /*
372 * We only accept RAs only when the per-interface flag
373 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
374 */
375 ifp = m->m_pkthdr.rcvif;
376 ndi = ND_IFINFO(ifp);
377 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
378 goto freeit;
379
380 /* RFC 6980: Nodes MUST silently ignore fragments */
381 if(m->m_flags & M_FRAGMENTED)
382 goto freeit;
383
384 ip6 = mtod(m, struct ip6_hdr *);
385 if (__predict_false(ip6->ip6_hlim != 255)) {
386 ICMP6STAT_INC(icp6s_invlhlim);
387 nd6log((LOG_ERR,
388 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
389 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
390 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
391 goto bad;
392 }
393
394 saddr6 = ip6->ip6_src;
395 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
396 nd6log((LOG_ERR,
397 "%s: src %s is not link-local\n", __func__,
398 ip6_sprintf(ip6bufs, &saddr6)));
399 goto bad;
400 }
401
402 if (m->m_len < off + icmp6len) {
403 m = m_pullup(m, off + icmp6len);
404 if (m == NULL) {
405 IP6STAT_INC(ip6s_exthdrtoolong);
406 return;
407 }
408 }
409 ip6 = mtod(m, struct ip6_hdr *);
410 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
411
412 icmp6len -= sizeof(*nd_ra);
413 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
414 if (nd6_options(&ndopts) < 0) {
415 nd6log((LOG_INFO,
416 "%s: invalid ND option, ignored\n", __func__));
417 /* nd6_options have incremented stats */
418 goto freeit;
419 }
420
421 mcast = 0;
422 dr = NULL;
423 {
424 struct nd_defrouter dr0;
425 u_int32_t advreachable = nd_ra->nd_ra_reachable;
426
427 /* remember if this is a multicasted advertisement */
428 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
429 mcast = 1;
430
431 bzero(&dr0, sizeof(dr0));
432 dr0.rtaddr = saddr6;
433 dr0.raflags = nd_ra->nd_ra_flags_reserved;
434 /*
435 * Effectively-disable routes from RA messages when
436 * ND6_IFF_NO_RADR enabled on the receiving interface or
437 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
438 */
439 if (ndi->flags & ND6_IFF_NO_RADR)
440 dr0.rtlifetime = 0;
441 else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
442 dr0.rtlifetime = 0;
443 else
444 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
445 dr0.expire = time_uptime + dr0.rtlifetime;
446 dr0.ifp = ifp;
447 /* unspecified or not? (RFC 2461 6.3.4) */
448 if (advreachable) {
449 advreachable = ntohl(advreachable);
450 if (advreachable <= MAX_REACHABLE_TIME &&
451 ndi->basereachable != advreachable) {
452 ndi->basereachable = advreachable;
453 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
454 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
455 }
456 }
457 if (nd_ra->nd_ra_retransmit)
458 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
459 if (nd_ra->nd_ra_curhoplimit) {
460 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
461 ndi->chlim = nd_ra->nd_ra_curhoplimit;
462 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
463 log(LOG_ERR, "RA with a lower CurHopLimit sent from "
464 "%s on %s (current = %d, received = %d). "
465 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
466 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
467 }
468 }
469 dr = defrtrlist_update(&dr0);
470 #ifdef EXPERIMENTAL
471 defrtr_ipv6_only_ifp(ifp);
472 #endif
473 }
474
475 /*
476 * prefix
477 */
478 if (ndopts.nd_opts_pi) {
479 struct nd_opt_hdr *pt;
480 struct nd_opt_prefix_info *pi = NULL;
481 struct nd_prefixctl pr;
482
483 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
484 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
485 pt = (struct nd_opt_hdr *)((caddr_t)pt +
486 (pt->nd_opt_len << 3))) {
487 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
488 continue;
489 pi = (struct nd_opt_prefix_info *)pt;
490
491 if (pi->nd_opt_pi_len != 4) {
492 nd6log((LOG_INFO,
493 "%s: invalid option len %d for prefix "
494 "information option, ignored\n", __func__,
495 pi->nd_opt_pi_len));
496 continue;
497 }
498
499 if (128 < pi->nd_opt_pi_prefix_len) {
500 nd6log((LOG_INFO,
501 "%s: invalid prefix len %d for prefix "
502 "information option, ignored\n", __func__,
503 pi->nd_opt_pi_prefix_len));
504 continue;
505 }
506
507 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
508 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
509 nd6log((LOG_INFO,
510 "%s: invalid prefix %s, ignored\n",
511 __func__, ip6_sprintf(ip6bufs,
512 &pi->nd_opt_pi_prefix)));
513 continue;
514 }
515
516 bzero(&pr, sizeof(pr));
517 pr.ndpr_prefix.sin6_family = AF_INET6;
518 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
519 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
520 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
521
522 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
523 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
524 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
525 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
526 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
527 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
528 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
529 (void)prelist_update(&pr, dr, m, mcast);
530 }
531 }
532 if (dr != NULL) {
533 defrouter_rele(dr);
534 dr = NULL;
535 }
536
537 /*
538 * MTU
539 */
540 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
541 u_long mtu;
542 u_long maxmtu;
543
544 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
545
546 /* lower bound */
547 if (mtu < IPV6_MMTU) {
548 nd6log((LOG_INFO, "%s: bogus mtu option mtu=%lu sent "
549 "from %s, ignoring\n", __func__,
550 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
551 goto skip;
552 }
553
554 /* upper bound */
555 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
556 ? ndi->maxmtu : ifp->if_mtu;
557 if (mtu <= maxmtu) {
558 if (ndi->linkmtu != mtu) {
559 ndi->linkmtu = mtu;
560 rt_updatemtu(ifp);
561 }
562 } else {
563 nd6log((LOG_INFO, "%s: bogus mtu=%lu sent from %s; "
564 "exceeds maxmtu %lu, ignoring\n", __func__,
565 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
566 }
567 }
568
569 skip:
570
571 /*
572 * Source link layer address
573 */
574 {
575 char *lladdr = NULL;
576 int lladdrlen = 0;
577
578 if (ndopts.nd_opts_src_lladdr) {
579 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
580 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
581 }
582
583 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
584 nd6log((LOG_INFO,
585 "%s: lladdrlen mismatch for %s (if %d, RA packet %d)\n",
586 __func__, ip6_sprintf(ip6bufs, &saddr6),
587 ifp->if_addrlen, lladdrlen - 2));
588 goto bad;
589 }
590
591 nd6_cache_lladdr(ifp, &saddr6, lladdr,
592 lladdrlen, ND_ROUTER_ADVERT, 0);
593
594 /*
595 * Installing a link-layer address might change the state of the
596 * router's neighbor cache, which might also affect our on-link
597 * detection of adveritsed prefixes.
598 */
599 pfxlist_onlink_check();
600 }
601
602 freeit:
603 m_freem(m);
604 return;
605
606 bad:
607 ICMP6STAT_INC(icp6s_badra);
608 m_freem(m);
609 }
610
611 /* PFXRTR */
612 static struct nd_pfxrouter *
pfxrtr_lookup(struct nd_prefix * pr,struct nd_defrouter * dr)613 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
614 {
615 struct nd_pfxrouter *search;
616
617 ND6_LOCK_ASSERT();
618
619 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
620 if (search->router == dr)
621 break;
622 }
623 return (search);
624 }
625
626 static void
pfxrtr_add(struct nd_prefix * pr,struct nd_defrouter * dr)627 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
628 {
629 struct nd_pfxrouter *new;
630 bool update;
631
632 ND6_UNLOCK_ASSERT();
633
634 ND6_RLOCK();
635 if (pfxrtr_lookup(pr, dr) != NULL) {
636 ND6_RUNLOCK();
637 return;
638 }
639 ND6_RUNLOCK();
640
641 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
642 if (new == NULL)
643 return;
644 defrouter_ref(dr);
645 new->router = dr;
646
647 ND6_WLOCK();
648 if (pfxrtr_lookup(pr, dr) == NULL) {
649 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
650 update = true;
651 } else {
652 /* We lost a race to add the reference. */
653 defrouter_rele(dr);
654 free(new, M_IP6NDP);
655 update = false;
656 }
657 ND6_WUNLOCK();
658
659 if (update)
660 pfxlist_onlink_check();
661 }
662
663 static void
pfxrtr_del(struct nd_pfxrouter * pfr)664 pfxrtr_del(struct nd_pfxrouter *pfr)
665 {
666
667 ND6_WLOCK_ASSERT();
668
669 LIST_REMOVE(pfr, pfr_entry);
670 defrouter_rele(pfr->router);
671 free(pfr, M_IP6NDP);
672 }
673
674 /* Default router list processing sub routines. */
675 static void
defrouter_addreq(struct nd_defrouter * new)676 defrouter_addreq(struct nd_defrouter *new)
677 {
678 uint32_t fibnum = new->ifp->if_fib;
679 struct rib_cmd_info rc = {};
680 int error = 0;
681
682 NET_EPOCH_ASSERT();
683
684 struct sockaddr_in6 gw = {
685 .sin6_family = AF_INET6,
686 .sin6_len = sizeof(struct sockaddr_in6),
687 .sin6_addr = new->rtaddr,
688 };
689
690 error = rib_add_default_route(fibnum, AF_INET6, new->ifp,
691 (struct sockaddr *)&gw, &rc);
692
693 if (error == 0) {
694 struct nhop_object *nh = nhop_select_func(rc.rc_nh_new, 0);
695 rt_routemsg(RTM_ADD, rc.rc_rt, nh, fibnum);
696 new->installed = 1;
697 }
698 }
699
700 /*
701 * Remove the default route for a given router.
702 * This is just a subroutine function for defrouter_select_fib(), and
703 * should not be called from anywhere else.
704 */
705 static void
defrouter_delreq(struct nd_defrouter * dr)706 defrouter_delreq(struct nd_defrouter *dr)
707 {
708 uint32_t fibnum = dr->ifp->if_fib;
709 struct epoch_tracker et;
710 struct rib_cmd_info rc;
711 int error;
712
713 struct sockaddr_in6 dst = {
714 .sin6_family = AF_INET6,
715 .sin6_len = sizeof(struct sockaddr_in6),
716 };
717
718 struct sockaddr_in6 gw = {
719 .sin6_family = AF_INET6,
720 .sin6_len = sizeof(struct sockaddr_in6),
721 .sin6_addr = dr->rtaddr,
722 };
723
724 NET_EPOCH_ENTER(et);
725 error = rib_del_route_px(fibnum, (struct sockaddr *)&dst, 0,
726 rib_match_gw, (struct sockaddr *)&gw, 0, &rc);
727 if (error == 0) {
728 struct nhop_object *nh = nhop_select_func(rc.rc_nh_old, 0);
729 rt_routemsg(RTM_DELETE, rc.rc_rt, nh, fibnum);
730 }
731 NET_EPOCH_EXIT(et);
732
733 dr->installed = 0;
734 }
735
736 static void
defrouter_del(struct nd_defrouter * dr)737 defrouter_del(struct nd_defrouter *dr)
738 {
739 struct nd_defrouter *deldr = NULL;
740 struct nd_prefix *pr;
741 struct nd_pfxrouter *pfxrtr;
742
743 ND6_UNLOCK_ASSERT();
744
745 /*
746 * Flush all the routing table entries that use the router
747 * as a next hop.
748 */
749 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
750 rt6_flush(&dr->rtaddr, dr->ifp);
751
752 #ifdef EXPERIMENTAL
753 defrtr_ipv6_only_ifp(dr->ifp);
754 #endif
755
756 if (dr->installed) {
757 deldr = dr;
758 defrouter_delreq(dr);
759 }
760
761 /*
762 * Also delete all the pointers to the router in each prefix lists.
763 */
764 ND6_WLOCK();
765 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
766 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
767 pfxrtr_del(pfxrtr);
768 }
769 ND6_WUNLOCK();
770
771 pfxlist_onlink_check();
772
773 /*
774 * If the router is the primary one, choose a new one.
775 * Note that defrouter_select_fib() will remove the current
776 * gateway from the routing table.
777 */
778 if (deldr)
779 defrouter_select_fib(deldr->ifp->if_fib);
780
781 /*
782 * Release the list reference.
783 */
784 defrouter_rele(dr);
785 }
786
787 struct nd_defrouter *
defrouter_lookup_locked(const struct in6_addr * addr,struct ifnet * ifp)788 defrouter_lookup_locked(const struct in6_addr *addr, struct ifnet *ifp)
789 {
790 struct nd_defrouter *dr;
791
792 ND6_LOCK_ASSERT();
793 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
794 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
795 defrouter_ref(dr);
796 return (dr);
797 }
798 return (NULL);
799 }
800
801 struct nd_defrouter *
defrouter_lookup(const struct in6_addr * addr,struct ifnet * ifp)802 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
803 {
804 struct nd_defrouter *dr;
805
806 ND6_RLOCK();
807 dr = defrouter_lookup_locked(addr, ifp);
808 ND6_RUNLOCK();
809 return (dr);
810 }
811
812 /*
813 * Remove all default routes from default router list.
814 */
815 void
defrouter_reset(void)816 defrouter_reset(void)
817 {
818 struct nd_defrouter *dr, **dra;
819 int count, i;
820
821 count = i = 0;
822
823 /*
824 * We can't delete routes with the ND lock held, so make a copy of the
825 * current default router list and use that when deleting routes.
826 */
827 ND6_RLOCK();
828 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
829 count++;
830 ND6_RUNLOCK();
831
832 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
833
834 ND6_RLOCK();
835 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
836 if (i == count)
837 break;
838 defrouter_ref(dr);
839 dra[i++] = dr;
840 }
841 ND6_RUNLOCK();
842
843 for (i = 0; i < count && dra[i] != NULL; i++) {
844 defrouter_delreq(dra[i]);
845 defrouter_rele(dra[i]);
846 }
847 free(dra, M_TEMP);
848
849 /*
850 * XXX should we also nuke any default routers in the kernel, by
851 * going through them by rtalloc1()?
852 */
853 }
854
855 /*
856 * Look up a matching default router list entry and remove it. Returns true if a
857 * matching entry was found, false otherwise.
858 */
859 bool
defrouter_remove(struct in6_addr * addr,struct ifnet * ifp)860 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
861 {
862 struct nd_defrouter *dr;
863
864 ND6_WLOCK();
865 dr = defrouter_lookup_locked(addr, ifp);
866 if (dr == NULL) {
867 ND6_WUNLOCK();
868 return (false);
869 }
870
871 defrouter_unlink(dr, NULL);
872 ND6_WUNLOCK();
873 defrouter_del(dr);
874 defrouter_rele(dr);
875 return (true);
876 }
877
878 /*
879 * for default router selection
880 * regards router-preference field as a 2-bit signed integer
881 */
882 static int
rtpref(struct nd_defrouter * dr)883 rtpref(struct nd_defrouter *dr)
884 {
885 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
886 case ND_RA_FLAG_RTPREF_HIGH:
887 return (RTPREF_HIGH);
888 case ND_RA_FLAG_RTPREF_MEDIUM:
889 case ND_RA_FLAG_RTPREF_RSV:
890 return (RTPREF_MEDIUM);
891 case ND_RA_FLAG_RTPREF_LOW:
892 return (RTPREF_LOW);
893 default:
894 /*
895 * This case should never happen. If it did, it would mean a
896 * serious bug of kernel internal. We thus always bark here.
897 * Or, can we even panic?
898 */
899 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
900 return (RTPREF_INVALID);
901 }
902 /* NOTREACHED */
903 }
904
905 static bool
is_dr_reachable(const struct nd_defrouter * dr)906 is_dr_reachable(const struct nd_defrouter *dr) {
907 struct llentry *ln = NULL;
908
909 ln = nd6_lookup(&dr->rtaddr, LLE_SF(AF_INET6, 0), dr->ifp);
910 if (ln == NULL)
911 return (false);
912 bool reachable = ND6_IS_LLINFO_PROBREACH(ln);
913 LLE_RUNLOCK(ln);
914 return reachable;
915 }
916
917 /*
918 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
919 * draft-ietf-ipngwg-router-selection:
920 * 1) Routers that are reachable or probably reachable should be preferred.
921 * If we have more than one (probably) reachable router, prefer ones
922 * with the highest router preference.
923 * 2) When no routers on the list are known to be reachable or
924 * probably reachable, routers SHOULD be selected in a round-robin
925 * fashion, regardless of router preference values.
926 * 3) If the Default Router List is empty, assume that all
927 * destinations are on-link.
928 *
929 * We assume nd_defrouter is sorted by router preference value.
930 * Since the code below covers both with and without router preference cases,
931 * we do not need to classify the cases by ifdef.
932 *
933 * At this moment, we do not try to install more than one default router,
934 * even when the multipath routing is available, because we're not sure about
935 * the benefits for stub hosts comparing to the risk of making the code
936 * complicated and the possibility of introducing bugs.
937 *
938 * We maintain a single list of routers for multiple FIBs, only considering one
939 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
940 * we do the whole thing multiple times.
941 */
942 void
defrouter_select_fib(int fibnum)943 defrouter_select_fib(int fibnum)
944 {
945 struct epoch_tracker et;
946 struct nd_defrouter *dr, *selected_dr, *installed_dr;
947
948 if (fibnum == RT_ALL_FIBS) {
949 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
950 defrouter_select_fib(fibnum);
951 }
952 return;
953 }
954
955 ND6_RLOCK();
956 /*
957 * Let's handle easy case (3) first:
958 * If default router list is empty, there's nothing to be done.
959 */
960 if (TAILQ_EMPTY(&V_nd6_defrouter)) {
961 ND6_RUNLOCK();
962 return;
963 }
964
965 /*
966 * Search for a (probably) reachable router from the list.
967 * We just pick up the first reachable one (if any), assuming that
968 * the ordering rule of the list described in defrtrlist_update().
969 */
970 selected_dr = installed_dr = NULL;
971 NET_EPOCH_ENTER(et);
972 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
973 if (dr->ifp->if_fib != fibnum)
974 continue;
975
976 if (selected_dr == NULL && is_dr_reachable(dr)) {
977 selected_dr = dr;
978 defrouter_ref(selected_dr);
979 }
980
981 if (dr->installed) {
982 if (installed_dr == NULL) {
983 installed_dr = dr;
984 defrouter_ref(installed_dr);
985 } else {
986 /*
987 * this should not happen.
988 * warn for diagnosis.
989 */
990 log(LOG_ERR, "defrouter_select_fib: more than "
991 "one router is installed\n");
992 }
993 }
994 }
995
996 /*
997 * If none of the default routers was found to be reachable,
998 * round-robin the list regardless of preference.
999 * Otherwise, if we have an installed router, check if the selected
1000 * (reachable) router should really be preferred to the installed one.
1001 * We only prefer the new router when the old one is not reachable
1002 * or when the new one has a really higher preference value.
1003 */
1004 if (selected_dr == NULL) {
1005 if (installed_dr == NULL ||
1006 TAILQ_NEXT(installed_dr, dr_entry) == NULL)
1007 dr = TAILQ_FIRST(&V_nd6_defrouter);
1008 else
1009 dr = TAILQ_NEXT(installed_dr, dr_entry);
1010
1011 /* Ensure we select a router for this FIB. */
1012 TAILQ_FOREACH_FROM(dr, &V_nd6_defrouter, dr_entry) {
1013 if (dr->ifp->if_fib == fibnum) {
1014 selected_dr = dr;
1015 defrouter_ref(selected_dr);
1016 break;
1017 }
1018 }
1019 } else if (installed_dr != NULL) {
1020 if (is_dr_reachable(installed_dr) &&
1021 rtpref(selected_dr) <= rtpref(installed_dr)) {
1022 defrouter_rele(selected_dr);
1023 selected_dr = installed_dr;
1024 }
1025 }
1026 ND6_RUNLOCK();
1027
1028 /*
1029 * If we selected a router for this FIB and it's different
1030 * than the installed one, remove the installed router and
1031 * install the selected one in its place.
1032 */
1033 if (installed_dr != selected_dr) {
1034 if (installed_dr != NULL) {
1035 defrouter_delreq(installed_dr);
1036 defrouter_rele(installed_dr);
1037 }
1038 if (selected_dr != NULL)
1039 defrouter_addreq(selected_dr);
1040 }
1041 if (selected_dr != NULL)
1042 defrouter_rele(selected_dr);
1043 NET_EPOCH_EXIT(et);
1044 }
1045
1046 static struct nd_defrouter *
defrtrlist_update(struct nd_defrouter * new)1047 defrtrlist_update(struct nd_defrouter *new)
1048 {
1049 struct nd_defrouter *dr, *n;
1050 uint64_t genid;
1051 int oldpref;
1052 bool writelocked;
1053
1054 if (new->rtlifetime == 0) {
1055 defrouter_remove(&new->rtaddr, new->ifp);
1056 return (NULL);
1057 }
1058
1059 ND6_RLOCK();
1060 writelocked = false;
1061 restart:
1062 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
1063 if (dr != NULL) {
1064 oldpref = rtpref(dr);
1065
1066 /* override */
1067 dr->raflags = new->raflags; /* XXX flag check */
1068 dr->rtlifetime = new->rtlifetime;
1069 dr->expire = new->expire;
1070
1071 /*
1072 * If the preference does not change, there's no need
1073 * to sort the entries. Also make sure the selected
1074 * router is still installed in the kernel.
1075 */
1076 if (dr->installed && rtpref(new) == oldpref) {
1077 if (writelocked)
1078 ND6_WUNLOCK();
1079 else
1080 ND6_RUNLOCK();
1081 return (dr);
1082 }
1083 }
1084
1085 /*
1086 * The router needs to be reinserted into the default router
1087 * list, so upgrade to a write lock. If that fails and the list
1088 * has potentially changed while the lock was dropped, we'll
1089 * redo the lookup with the write lock held.
1090 */
1091 if (!writelocked) {
1092 writelocked = true;
1093 if (!ND6_TRY_UPGRADE()) {
1094 genid = V_nd6_list_genid;
1095 ND6_RUNLOCK();
1096 ND6_WLOCK();
1097 if (genid != V_nd6_list_genid)
1098 goto restart;
1099 }
1100 }
1101
1102 if (dr != NULL) {
1103 /*
1104 * The preferred router may have changed, so relocate this
1105 * router.
1106 */
1107 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
1108 n = dr;
1109 } else {
1110 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
1111 if (n == NULL) {
1112 ND6_WUNLOCK();
1113 return (NULL);
1114 }
1115 memcpy(n, new, sizeof(*n));
1116 /* Initialize with an extra reference for the caller. */
1117 refcount_init(&n->refcnt, 2);
1118 }
1119
1120 /*
1121 * Insert the new router in the Default Router List;
1122 * The Default Router List should be in the descending order
1123 * of router-preferece. Routers with the same preference are
1124 * sorted in the arriving time order.
1125 */
1126
1127 /* insert at the end of the group */
1128 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1129 if (rtpref(n) > rtpref(dr))
1130 break;
1131 }
1132 if (dr != NULL)
1133 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
1134 else
1135 TAILQ_INSERT_TAIL(&V_nd6_defrouter, n, dr_entry);
1136 V_nd6_list_genid++;
1137 ND6_WUNLOCK();
1138
1139 defrouter_select_fib(new->ifp->if_fib);
1140
1141 return (n);
1142 }
1143
1144 static int
in6_init_prefix_ltimes(struct nd_prefix * ndpr)1145 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1146 {
1147 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1148 ndpr->ndpr_preferred = 0;
1149 else
1150 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
1151 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1152 ndpr->ndpr_expire = 0;
1153 else
1154 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
1155
1156 return 0;
1157 }
1158
1159 static void
in6_init_address_ltimes(struct nd_prefix * new,struct in6_addrlifetime * lt6)1160 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1161 {
1162 /* init ia6t_expire */
1163 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1164 lt6->ia6t_expire = 0;
1165 else {
1166 lt6->ia6t_expire = time_uptime;
1167 lt6->ia6t_expire += lt6->ia6t_vltime;
1168 }
1169
1170 /* init ia6t_preferred */
1171 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1172 lt6->ia6t_preferred = 0;
1173 else {
1174 lt6->ia6t_preferred = time_uptime;
1175 lt6->ia6t_preferred += lt6->ia6t_pltime;
1176 }
1177 }
1178
1179 static struct in6_ifaddr *
in6_ifadd(struct nd_prefixctl * pr,int mcast)1180 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1181 {
1182 struct ifnet *ifp = pr->ndpr_ifp;
1183 struct ifaddr *ifa;
1184 struct in6_aliasreq ifra;
1185 struct in6_ifaddr *ia = NULL, *ib = NULL;
1186 int error, plen0;
1187 struct in6_addr *ifid_addr = NULL, mask;
1188 int prefixlen = pr->ndpr_plen;
1189 int updateflags;
1190 char ip6buf[INET6_ADDRSTRLEN];
1191
1192 in6_prefixlen2mask(&mask, prefixlen);
1193
1194 /*
1195 * find a link-local address (will be interface ID).
1196 * Is it really mandatory? Theoretically, a global or a site-local
1197 * address can be configured without a link-local address, if we
1198 * have a unique interface identifier...
1199 *
1200 * it is not mandatory to have a link-local address, we can generate
1201 * interface identifier on the fly. we do this because:
1202 * (1) it should be the easiest way to find interface identifier.
1203 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1204 * for multiple addresses on a single interface, and possible shortcut
1205 * of DAD. we omitted DAD for this reason in the past.
1206 * (3) a user can prevent autoconfiguration of global address
1207 * by removing link-local address by hand (this is partly because we
1208 * don't have other way to control the use of IPv6 on an interface.
1209 * this has been our design choice - cf. NRL's "ifconfig auto").
1210 * (4) it is easier to manage when an interface has addresses
1211 * with the same interface identifier, than to have multiple addresses
1212 * with different interface identifiers.
1213 */
1214 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1215 if (ifa) {
1216 ib = (struct in6_ifaddr *)ifa;
1217 ifid_addr = &ib->ia_addr.sin6_addr;
1218
1219 /* prefixlen + ifidlen must be equal to 128 */
1220 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1221 if (prefixlen != plen0) {
1222 ifa_free(ifa);
1223 ifid_addr = NULL;
1224 nd6log((LOG_DEBUG,
1225 "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n",
1226 __func__, if_name(ifp), prefixlen, 128 - plen0));
1227 }
1228 }
1229
1230 /* No suitable LL address, get the ifid directly */
1231 if (ifid_addr == NULL) {
1232 struct in6_addr taddr;
1233 ifa = ifa_alloc(sizeof(taddr), M_WAITOK);
1234 if (ifa) {
1235 ib = (struct in6_ifaddr *)ifa;
1236 ifid_addr = &ib->ia_addr.sin6_addr;
1237 if(in6_get_ifid(ifp, NULL, ifid_addr) != 0) {
1238 nd6log((LOG_DEBUG,
1239 "%s: failed to get ifid for %s\n",
1240 __func__, if_name(ifp)));
1241 ifa_free(ifa);
1242 ifid_addr = NULL;
1243 }
1244 }
1245 }
1246
1247 if (ifid_addr == NULL) {
1248 nd6log((LOG_INFO,
1249 "%s: could not determine ifid for %s\n",
1250 __func__, if_name(ifp)));
1251 return NULL;
1252 }
1253
1254 /* make ifaddr */
1255 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1256
1257 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1258 /* interface ID */
1259 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1260 (ifid_addr->s6_addr32[0] & ~mask.s6_addr32[0]);
1261 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1262 (ifid_addr->s6_addr32[1] & ~mask.s6_addr32[1]);
1263 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1264 (ifid_addr->s6_addr32[2] & ~mask.s6_addr32[2]);
1265 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1266 (ifid_addr->s6_addr32[3] & ~mask.s6_addr32[3]);
1267 ifa_free(ifa);
1268
1269 /* lifetimes. */
1270 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1271 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1272
1273 /* XXX: scope zone ID? */
1274
1275 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1276
1277 /*
1278 * Make sure that we do not have this address already. This should
1279 * usually not happen, but we can still see this case, e.g., if we
1280 * have manually configured the exact address to be configured.
1281 */
1282 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1283 &ifra.ifra_addr.sin6_addr);
1284 if (ifa != NULL) {
1285 ifa_free(ifa);
1286 /* this should be rare enough to make an explicit log */
1287 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1288 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1289 return (NULL);
1290 }
1291
1292 /*
1293 * Allocate ifaddr structure, link into chain, etc.
1294 * If we are going to create a new address upon receiving a multicasted
1295 * RA, we need to impose a random delay before starting DAD.
1296 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1297 */
1298 updateflags = 0;
1299 if (mcast)
1300 updateflags |= IN6_IFAUPDATE_DADDELAY;
1301 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1302 nd6log((LOG_ERR,
1303 "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__,
1304 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1305 if_name(ifp), error));
1306 return (NULL); /* ifaddr must not have been allocated. */
1307 }
1308
1309 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1310 /*
1311 * XXXRW: Assumption of non-NULLness here might not be true with
1312 * fine-grained locking -- should we validate it? Or just return
1313 * earlier ifa rather than looking it up again?
1314 */
1315 return (ia); /* this is always non-NULL and referenced. */
1316 }
1317
1318 static struct nd_prefix *
nd6_prefix_lookup_locked(struct nd_prefixctl * key)1319 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1320 {
1321 struct nd_prefix *search;
1322
1323 ND6_LOCK_ASSERT();
1324
1325 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1326 if (key->ndpr_ifp == search->ndpr_ifp &&
1327 key->ndpr_plen == search->ndpr_plen &&
1328 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1329 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1330 nd6_prefix_ref(search);
1331 break;
1332 }
1333 }
1334 return (search);
1335 }
1336
1337 struct nd_prefix *
nd6_prefix_lookup(struct nd_prefixctl * key)1338 nd6_prefix_lookup(struct nd_prefixctl *key)
1339 {
1340 struct nd_prefix *search;
1341
1342 ND6_RLOCK();
1343 search = nd6_prefix_lookup_locked(key);
1344 ND6_RUNLOCK();
1345 return (search);
1346 }
1347
1348 void
nd6_prefix_ref(struct nd_prefix * pr)1349 nd6_prefix_ref(struct nd_prefix *pr)
1350 {
1351
1352 refcount_acquire(&pr->ndpr_refcnt);
1353 }
1354
1355 void
nd6_prefix_rele(struct nd_prefix * pr)1356 nd6_prefix_rele(struct nd_prefix *pr)
1357 {
1358
1359 if (refcount_release(&pr->ndpr_refcnt)) {
1360 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1361 ("prefix %p has advertising routers", pr));
1362 free(pr, M_IP6NDP);
1363 }
1364 }
1365
1366 int
nd6_prelist_add(struct nd_prefixctl * pr,struct nd_defrouter * dr,struct nd_prefix ** newp)1367 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1368 struct nd_prefix **newp)
1369 {
1370 struct nd_prefix *new;
1371 char ip6buf[INET6_ADDRSTRLEN];
1372 int error;
1373
1374 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1375 if (new == NULL)
1376 return (ENOMEM);
1377 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1378 new->ndpr_ifp = pr->ndpr_ifp;
1379 new->ndpr_prefix = pr->ndpr_prefix;
1380 new->ndpr_plen = pr->ndpr_plen;
1381 new->ndpr_vltime = pr->ndpr_vltime;
1382 new->ndpr_pltime = pr->ndpr_pltime;
1383 new->ndpr_flags = pr->ndpr_flags;
1384 if ((error = in6_init_prefix_ltimes(new)) != 0) {
1385 free(new, M_IP6NDP);
1386 return (error);
1387 }
1388 new->ndpr_lastupdate = time_uptime;
1389
1390 /* initialization */
1391 LIST_INIT(&new->ndpr_advrtrs);
1392 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1393 /* make prefix in the canonical form */
1394 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1395
1396 ND6_WLOCK();
1397 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1398 V_nd6_list_genid++;
1399 ND6_WUNLOCK();
1400
1401 /* ND_OPT_PI_FLAG_ONLINK processing */
1402 if (new->ndpr_raf_onlink) {
1403 struct epoch_tracker et;
1404
1405 ND6_ONLINK_LOCK();
1406 NET_EPOCH_ENTER(et);
1407 if ((error = nd6_prefix_onlink(new)) != 0) {
1408 nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d "
1409 "on-link on %s (errno=%d)\n", __func__,
1410 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1411 pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1412 /* proceed anyway. XXX: is it correct? */
1413 }
1414 NET_EPOCH_EXIT(et);
1415 ND6_ONLINK_UNLOCK();
1416 }
1417
1418 if (dr != NULL)
1419 pfxrtr_add(new, dr);
1420 if (newp != NULL)
1421 *newp = new;
1422 return (0);
1423 }
1424
1425 /*
1426 * Remove a prefix from the prefix list and optionally stash it in a
1427 * caller-provided list.
1428 *
1429 * The ND6 lock must be held.
1430 */
1431 void
nd6_prefix_unlink(struct nd_prefix * pr,struct nd_prhead * list)1432 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1433 {
1434
1435 ND6_WLOCK_ASSERT();
1436
1437 LIST_REMOVE(pr, ndpr_entry);
1438 V_nd6_list_genid++;
1439 if (list != NULL)
1440 LIST_INSERT_HEAD(list, pr, ndpr_entry);
1441 }
1442
1443 /*
1444 * Free an unlinked prefix, first marking it off-link if necessary.
1445 */
1446 void
nd6_prefix_del(struct nd_prefix * pr)1447 nd6_prefix_del(struct nd_prefix *pr)
1448 {
1449 struct nd_pfxrouter *pfr, *next;
1450 int e;
1451 char ip6buf[INET6_ADDRSTRLEN];
1452
1453 KASSERT(pr->ndpr_addrcnt == 0,
1454 ("prefix %p has referencing addresses", pr));
1455 ND6_UNLOCK_ASSERT();
1456
1457 /*
1458 * Though these flags are now meaningless, we'd rather keep the value
1459 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1460 * when executing "ndp -p".
1461 */
1462 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1463 ND6_ONLINK_LOCK();
1464 if ((e = nd6_prefix_offlink(pr)) != 0) {
1465 nd6log((LOG_ERR,
1466 "%s: failed to make the prefix %s/%d offlink on %s "
1467 "(errno=%d)\n", __func__,
1468 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1469 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1470 /* what should we do? */
1471 }
1472 ND6_ONLINK_UNLOCK();
1473 }
1474
1475 /* Release references to routers that have advertised this prefix. */
1476 ND6_WLOCK();
1477 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1478 pfxrtr_del(pfr);
1479 ND6_WUNLOCK();
1480
1481 nd6_prefix_rele(pr);
1482
1483 pfxlist_onlink_check();
1484 }
1485
1486 static int
prelist_update(struct nd_prefixctl * new,struct nd_defrouter * dr,struct mbuf * m,int mcast)1487 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1488 struct mbuf *m, int mcast)
1489 {
1490 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1491 struct ifaddr *ifa;
1492 struct ifnet *ifp = new->ndpr_ifp;
1493 struct nd_prefix *pr;
1494 int error = 0;
1495 int auth;
1496 struct in6_addrlifetime lt6_tmp;
1497 char ip6buf[INET6_ADDRSTRLEN];
1498
1499 NET_EPOCH_ASSERT();
1500
1501 auth = 0;
1502 if (m) {
1503 /*
1504 * Authenticity for NA consists authentication for
1505 * both IP header and IP datagrams, doesn't it ?
1506 */
1507 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1508 auth = ((m->m_flags & M_AUTHIPHDR) &&
1509 (m->m_flags & M_AUTHIPDGM));
1510 #endif
1511 }
1512
1513 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1514 /*
1515 * nd6_prefix_lookup() ensures that pr and new have the same
1516 * prefix on a same interface.
1517 */
1518
1519 /*
1520 * Update prefix information. Note that the on-link (L) bit
1521 * and the autonomous (A) bit should NOT be changed from 1
1522 * to 0.
1523 */
1524 if (new->ndpr_raf_onlink == 1)
1525 pr->ndpr_raf_onlink = 1;
1526 if (new->ndpr_raf_auto == 1)
1527 pr->ndpr_raf_auto = 1;
1528 if (new->ndpr_raf_onlink) {
1529 pr->ndpr_vltime = new->ndpr_vltime;
1530 pr->ndpr_pltime = new->ndpr_pltime;
1531 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1532 pr->ndpr_lastupdate = time_uptime;
1533 }
1534
1535 if (new->ndpr_raf_onlink &&
1536 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1537 ND6_ONLINK_LOCK();
1538 if ((error = nd6_prefix_onlink(pr)) != 0) {
1539 nd6log((LOG_ERR,
1540 "%s: failed to make the prefix %s/%d "
1541 "on-link on %s (errno=%d)\n", __func__,
1542 ip6_sprintf(ip6buf,
1543 &pr->ndpr_prefix.sin6_addr),
1544 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1545 error));
1546 /* proceed anyway. XXX: is it correct? */
1547 }
1548 ND6_ONLINK_UNLOCK();
1549 }
1550
1551 if (dr != NULL)
1552 pfxrtr_add(pr, dr);
1553 } else {
1554 if (new->ndpr_vltime == 0)
1555 goto end;
1556 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1557 goto end;
1558
1559 error = nd6_prelist_add(new, dr, &pr);
1560 if (error != 0) {
1561 nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for "
1562 "the prefix %s/%d on %s (errno=%d)\n", __func__,
1563 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1564 new->ndpr_plen, if_name(new->ndpr_ifp), error));
1565 goto end; /* we should just give up in this case. */
1566 }
1567
1568 /*
1569 * XXX: from the ND point of view, we can ignore a prefix
1570 * with the on-link bit being zero. However, we need a
1571 * prefix structure for references from autoconfigured
1572 * addresses. Thus, we explicitly make sure that the prefix
1573 * itself expires now.
1574 */
1575 if (pr->ndpr_raf_onlink == 0) {
1576 pr->ndpr_vltime = 0;
1577 pr->ndpr_pltime = 0;
1578 in6_init_prefix_ltimes(pr);
1579 }
1580 }
1581
1582 /*
1583 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1584 * Note that pr must be non NULL at this point.
1585 */
1586
1587 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1588 if (!new->ndpr_raf_auto)
1589 goto end;
1590
1591 /*
1592 * 5.5.3 (b). the link-local prefix should have been ignored in
1593 * nd6_ra_input.
1594 */
1595
1596 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1597 if (new->ndpr_pltime > new->ndpr_vltime) {
1598 error = EINVAL; /* XXX: won't be used */
1599 goto end;
1600 }
1601
1602 /*
1603 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1604 * an address configured by stateless autoconfiguration already in the
1605 * list of addresses associated with the interface, and the Valid
1606 * Lifetime is not 0, form an address. We first check if we have
1607 * a matching prefix.
1608 * Note: we apply a clarification in rfc2462bis-02 here. We only
1609 * consider autoconfigured addresses while RFC2462 simply said
1610 * "address".
1611 */
1612 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1613 struct in6_ifaddr *ifa6;
1614 u_int32_t remaininglifetime;
1615
1616 if (ifa->ifa_addr->sa_family != AF_INET6)
1617 continue;
1618
1619 ifa6 = (struct in6_ifaddr *)ifa;
1620
1621 /*
1622 * We only consider autoconfigured addresses as per rfc2462bis.
1623 */
1624 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1625 continue;
1626
1627 /*
1628 * Spec is not clear here, but I believe we should concentrate
1629 * on unicast (i.e. not anycast) addresses.
1630 * XXX: other ia6_flags? detached or duplicated?
1631 */
1632 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1633 continue;
1634
1635 /*
1636 * Ignore the address if it is not associated with a prefix
1637 * or is associated with a prefix that is different from this
1638 * one. (pr is never NULL here)
1639 */
1640 if (ifa6->ia6_ndpr != pr)
1641 continue;
1642
1643 if (ia6_match == NULL) /* remember the first one */
1644 ia6_match = ifa6;
1645
1646 /*
1647 * An already autoconfigured address matched. Now that we
1648 * are sure there is at least one matched address, we can
1649 * proceed to 5.5.3. (e): update the lifetimes according to the
1650 * "two hours" rule and the privacy extension.
1651 * We apply some clarifications in rfc2462bis:
1652 * - use remaininglifetime instead of storedlifetime as a
1653 * variable name
1654 * - remove the dead code in the "two-hour" rule
1655 */
1656 #define TWOHOUR (120*60)
1657 lt6_tmp = ifa6->ia6_lifetime;
1658
1659 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1660 remaininglifetime = ND6_INFINITE_LIFETIME;
1661 else if (time_uptime - ifa6->ia6_updatetime >
1662 lt6_tmp.ia6t_vltime) {
1663 /*
1664 * The case of "invalid" address. We should usually
1665 * not see this case.
1666 */
1667 remaininglifetime = 0;
1668 } else
1669 remaininglifetime = lt6_tmp.ia6t_vltime -
1670 (time_uptime - ifa6->ia6_updatetime);
1671
1672 /* when not updating, keep the current stored lifetime. */
1673 lt6_tmp.ia6t_vltime = remaininglifetime;
1674
1675 if (TWOHOUR < new->ndpr_vltime ||
1676 remaininglifetime < new->ndpr_vltime) {
1677 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1678 } else if (remaininglifetime <= TWOHOUR) {
1679 if (auth) {
1680 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1681 }
1682 } else {
1683 /*
1684 * new->ndpr_vltime <= TWOHOUR &&
1685 * TWOHOUR < remaininglifetime
1686 */
1687 lt6_tmp.ia6t_vltime = TWOHOUR;
1688 }
1689
1690 /* The 2 hour rule is not imposed for preferred lifetime. */
1691 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1692
1693 in6_init_address_ltimes(pr, <6_tmp);
1694
1695 /*
1696 * We need to treat lifetimes for temporary addresses
1697 * differently, according to
1698 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1699 * we only update the lifetimes when they are in the maximum
1700 * intervals.
1701 */
1702 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1703 u_int32_t maxvltime, maxpltime;
1704
1705 if (V_ip6_temp_valid_lifetime >
1706 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1707 V_ip6_desync_factor)) {
1708 maxvltime = V_ip6_temp_valid_lifetime -
1709 (time_uptime - ifa6->ia6_createtime) -
1710 V_ip6_desync_factor;
1711 } else
1712 maxvltime = 0;
1713 if (V_ip6_temp_preferred_lifetime >
1714 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1715 V_ip6_desync_factor)) {
1716 maxpltime = V_ip6_temp_preferred_lifetime -
1717 (time_uptime - ifa6->ia6_createtime) -
1718 V_ip6_desync_factor;
1719 } else
1720 maxpltime = 0;
1721
1722 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1723 lt6_tmp.ia6t_vltime > maxvltime) {
1724 lt6_tmp.ia6t_vltime = maxvltime;
1725 }
1726 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1727 lt6_tmp.ia6t_pltime > maxpltime) {
1728 lt6_tmp.ia6t_pltime = maxpltime;
1729 }
1730 }
1731 ifa6->ia6_lifetime = lt6_tmp;
1732 ifa6->ia6_updatetime = time_uptime;
1733 }
1734 if (ia6_match == NULL && new->ndpr_vltime) {
1735 int ifidlen;
1736
1737 /*
1738 * 5.5.3 (d) (continued)
1739 * No address matched and the valid lifetime is non-zero.
1740 * Create a new address.
1741 */
1742
1743 /*
1744 * Prefix Length check:
1745 * If the sum of the prefix length and interface identifier
1746 * length does not equal 128 bits, the Prefix Information
1747 * option MUST be ignored. The length of the interface
1748 * identifier is defined in a separate link-type specific
1749 * document.
1750 */
1751 ifidlen = in6_if2idlen(ifp);
1752 if (ifidlen < 0) {
1753 /* this should not happen, so we always log it. */
1754 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1755 if_name(ifp));
1756 goto end;
1757 }
1758 if (ifidlen + pr->ndpr_plen != 128) {
1759 nd6log((LOG_INFO,
1760 "%s: invalid prefixlen %d for %s, ignored\n",
1761 __func__, pr->ndpr_plen, if_name(ifp)));
1762 goto end;
1763 }
1764
1765 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1766 /*
1767 * note that we should use pr (not new) for reference.
1768 */
1769 pr->ndpr_addrcnt++;
1770 ia6->ia6_ndpr = pr;
1771
1772 /*
1773 * RFC 3041 3.3 (2).
1774 * When a new public address is created as described
1775 * in RFC2462, also create a new temporary address.
1776 *
1777 * RFC 3041 3.5.
1778 * When an interface connects to a new link, a new
1779 * randomized interface identifier should be generated
1780 * immediately together with a new set of temporary
1781 * addresses. Thus, we specifiy 1 as the 2nd arg of
1782 * in6_tmpifadd().
1783 */
1784 if (V_ip6_use_tempaddr) {
1785 int e;
1786 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1787 nd6log((LOG_NOTICE, "%s: failed to "
1788 "create a temporary address "
1789 "(errno=%d)\n", __func__, e));
1790 }
1791 }
1792 ifa_free(&ia6->ia_ifa);
1793
1794 /*
1795 * A newly added address might affect the status
1796 * of other addresses, so we check and update it.
1797 * XXX: what if address duplication happens?
1798 */
1799 pfxlist_onlink_check();
1800 } else {
1801 /* just set an error. do not bark here. */
1802 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1803 }
1804 }
1805
1806 end:
1807 if (pr != NULL)
1808 nd6_prefix_rele(pr);
1809 return (error);
1810 }
1811
1812 /*
1813 * A supplement function used in the on-link detection below;
1814 * detect if a given prefix has a (probably) reachable advertising router.
1815 * XXX: lengthy function name...
1816 */
1817 static struct nd_pfxrouter *
find_pfxlist_reachable_router(struct nd_prefix * pr)1818 find_pfxlist_reachable_router(struct nd_prefix *pr)
1819 {
1820 struct epoch_tracker et;
1821 struct nd_pfxrouter *pfxrtr;
1822
1823 ND6_LOCK_ASSERT();
1824
1825 NET_EPOCH_ENTER(et);
1826 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1827 if (is_dr_reachable(pfxrtr->router))
1828 break;
1829 }
1830 NET_EPOCH_EXIT(et);
1831 return (pfxrtr);
1832 }
1833
1834 /*
1835 * Check if each prefix in the prefix list has at least one available router
1836 * that advertised the prefix (a router is "available" if its neighbor cache
1837 * entry is reachable or probably reachable).
1838 * If the check fails, the prefix may be off-link, because, for example,
1839 * we have moved from the network but the lifetime of the prefix has not
1840 * expired yet. So we should not use the prefix if there is another prefix
1841 * that has an available router.
1842 * But, if there is no prefix that has an available router, we still regard
1843 * all the prefixes as on-link. This is because we can't tell if all the
1844 * routers are simply dead or if we really moved from the network and there
1845 * is no router around us.
1846 */
1847 void
pfxlist_onlink_check(void)1848 pfxlist_onlink_check(void)
1849 {
1850 struct nd_prefix *pr;
1851 struct in6_ifaddr *ifa;
1852 struct nd_defrouter *dr;
1853 struct nd_pfxrouter *pfxrtr = NULL;
1854 struct rm_priotracker in6_ifa_tracker;
1855 uint64_t genid;
1856 uint32_t flags;
1857
1858 ND6_ONLINK_LOCK();
1859 ND6_RLOCK();
1860
1861 /*
1862 * Check if there is a prefix that has a reachable advertising
1863 * router.
1864 */
1865 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1866 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1867 break;
1868 }
1869
1870 /*
1871 * If we have no such prefix, check whether we still have a router
1872 * that does not advertise any prefixes.
1873 */
1874 if (pr == NULL) {
1875 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1876 struct nd_prefix *pr0;
1877
1878 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1879 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1880 break;
1881 }
1882 if (pfxrtr != NULL)
1883 break;
1884 }
1885 }
1886 if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) {
1887 /*
1888 * There is at least one prefix that has a reachable router,
1889 * or at least a router which probably does not advertise
1890 * any prefixes. The latter would be the case when we move
1891 * to a new link where we have a router that does not provide
1892 * prefixes and we configure an address by hand.
1893 * Detach prefixes which have no reachable advertising
1894 * router, and attach other prefixes.
1895 */
1896 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1897 /* XXX: a link-local prefix should never be detached */
1898 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1899 pr->ndpr_raf_onlink == 0 ||
1900 pr->ndpr_raf_auto == 0)
1901 continue;
1902
1903 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1904 find_pfxlist_reachable_router(pr) == NULL)
1905 pr->ndpr_stateflags |= NDPRF_DETACHED;
1906 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1907 find_pfxlist_reachable_router(pr) != NULL)
1908 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1909 }
1910 } else {
1911 /* there is no prefix that has a reachable router */
1912 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1913 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1914 pr->ndpr_raf_onlink == 0 ||
1915 pr->ndpr_raf_auto == 0)
1916 continue;
1917 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1918 }
1919 }
1920
1921 /*
1922 * Remove each interface route associated with a (just) detached
1923 * prefix, and reinstall the interface route for a (just) attached
1924 * prefix. Note that all attempt of reinstallation does not
1925 * necessarily success, when a same prefix is shared among multiple
1926 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1927 * so we don't have to care about them.
1928 */
1929 restart:
1930 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1931 char ip6buf[INET6_ADDRSTRLEN];
1932 int e;
1933
1934 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1935 pr->ndpr_raf_onlink == 0 ||
1936 pr->ndpr_raf_auto == 0)
1937 continue;
1938
1939 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1940 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1941 genid = V_nd6_list_genid;
1942 ND6_RUNLOCK();
1943 if ((flags & NDPRF_ONLINK) != 0 &&
1944 (e = nd6_prefix_offlink(pr)) != 0) {
1945 nd6log((LOG_ERR,
1946 "%s: failed to make %s/%d offlink "
1947 "(errno=%d)\n", __func__,
1948 ip6_sprintf(ip6buf,
1949 &pr->ndpr_prefix.sin6_addr),
1950 pr->ndpr_plen, e));
1951 } else if ((flags & NDPRF_ONLINK) == 0 &&
1952 (e = nd6_prefix_onlink(pr)) != 0) {
1953 nd6log((LOG_ERR,
1954 "%s: failed to make %s/%d onlink "
1955 "(errno=%d)\n", __func__,
1956 ip6_sprintf(ip6buf,
1957 &pr->ndpr_prefix.sin6_addr),
1958 pr->ndpr_plen, e));
1959 }
1960 ND6_RLOCK();
1961 if (genid != V_nd6_list_genid)
1962 goto restart;
1963 }
1964 }
1965
1966 /*
1967 * Changes on the prefix status might affect address status as well.
1968 * Make sure that all addresses derived from an attached prefix are
1969 * attached, and that all addresses derived from a detached prefix are
1970 * detached. Note, however, that a manually configured address should
1971 * always be attached.
1972 * The precise detection logic is same as the one for prefixes.
1973 */
1974 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1975 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1976 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1977 continue;
1978
1979 if (ifa->ia6_ndpr == NULL) {
1980 /*
1981 * This can happen when we first configure the address
1982 * (i.e. the address exists, but the prefix does not).
1983 * XXX: complicated relationships...
1984 */
1985 continue;
1986 }
1987
1988 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1989 break;
1990 }
1991 if (ifa) {
1992 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1993 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1994 continue;
1995
1996 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1997 continue;
1998
1999 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
2000 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
2001 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
2002 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
2003 nd6_dad_start((struct ifaddr *)ifa, 0);
2004 }
2005 } else {
2006 ifa->ia6_flags |= IN6_IFF_DETACHED;
2007 }
2008 }
2009 } else {
2010 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
2011 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
2012 continue;
2013
2014 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
2015 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
2016 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
2017 /* Do we need a delay in this case? */
2018 nd6_dad_start((struct ifaddr *)ifa, 0);
2019 }
2020 }
2021 }
2022 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
2023 ND6_RUNLOCK();
2024 ND6_ONLINK_UNLOCK();
2025 }
2026
2027 /*
2028 * Add or remove interface route specified by @dst, @netmask and @ifp.
2029 * ifa can be NULL.
2030 * Returns 0 on success
2031 */
2032 static int
nd6_prefix_rtrequest(uint32_t fibnum,int cmd,struct sockaddr_in6 * dst,struct sockaddr_in6 * netmask,struct ifnet * ifp,struct ifaddr * ifa)2033 nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst,
2034 struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa)
2035 {
2036 struct epoch_tracker et;
2037 int error;
2038
2039 /* Prepare gateway */
2040 struct sockaddr_dl_short sdl = {
2041 .sdl_family = AF_LINK,
2042 .sdl_len = sizeof(struct sockaddr_dl_short),
2043 .sdl_type = ifp->if_type,
2044 .sdl_index = ifp->if_index,
2045 };
2046
2047 struct rt_addrinfo info = {
2048 .rti_ifa = ifa,
2049 .rti_ifp = ifp,
2050 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
2051 .rti_info = {
2052 [RTAX_DST] = (struct sockaddr *)dst,
2053 [RTAX_NETMASK] = (struct sockaddr *)netmask,
2054 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
2055 },
2056 };
2057 /* Don't set additional per-gw filters on removal */
2058
2059 NET_EPOCH_ENTER(et);
2060 error = rib_handle_ifaddr_info(fibnum, cmd, &info);
2061 NET_EPOCH_EXIT(et);
2062 return (error);
2063 }
2064
2065 static int
nd6_prefix_onlink_rtrequest(struct nd_prefix * pr,struct ifaddr * ifa)2066 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
2067 {
2068 int error;
2069
2070 struct sockaddr_in6 mask6 = {
2071 .sin6_family = AF_INET6,
2072 .sin6_len = sizeof(struct sockaddr_in6),
2073 .sin6_addr = pr->ndpr_mask,
2074 };
2075 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2076
2077 error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD,
2078 &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa);
2079 if (error == 0)
2080 pr->ndpr_stateflags |= NDPRF_ONLINK;
2081
2082 return (error);
2083 }
2084
2085 static int
nd6_prefix_onlink(struct nd_prefix * pr)2086 nd6_prefix_onlink(struct nd_prefix *pr)
2087 {
2088 struct epoch_tracker et;
2089 struct ifaddr *ifa;
2090 struct ifnet *ifp = pr->ndpr_ifp;
2091 struct nd_prefix *opr;
2092 char ip6buf[INET6_ADDRSTRLEN];
2093 int error;
2094
2095 ND6_ONLINK_LOCK_ASSERT();
2096 ND6_UNLOCK_ASSERT();
2097
2098 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
2099 return (EEXIST);
2100
2101 /*
2102 * Add the interface route associated with the prefix. Before
2103 * installing the route, check if there's the same prefix on another
2104 * interface, and the prefix has already installed the interface route.
2105 * Although such a configuration is expected to be rare, we explicitly
2106 * allow it.
2107 */
2108 ND6_RLOCK();
2109 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2110 if (opr == pr)
2111 continue;
2112
2113 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2114 continue;
2115
2116 if (!V_rt_add_addr_allfibs &&
2117 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
2118 continue;
2119
2120 if (opr->ndpr_plen == pr->ndpr_plen &&
2121 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2122 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2123 ND6_RUNLOCK();
2124 return (0);
2125 }
2126 }
2127 ND6_RUNLOCK();
2128
2129 /*
2130 * We prefer link-local addresses as the associated interface address.
2131 */
2132 /* search for a link-local addr */
2133 NET_EPOCH_ENTER(et);
2134 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
2135 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
2136 if (ifa == NULL) {
2137 /* XXX: freebsd does not have ifa_ifwithaf */
2138 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2139 if (ifa->ifa_addr->sa_family == AF_INET6) {
2140 ifa_ref(ifa);
2141 break;
2142 }
2143 }
2144 /* should we care about ia6_flags? */
2145 }
2146 if (ifa == NULL) {
2147 /*
2148 * This can still happen, when, for example, we receive an RA
2149 * containing a prefix with the L bit set and the A bit clear,
2150 * after removing all IPv6 addresses on the receiving
2151 * interface. This should, of course, be rare though.
2152 */
2153 nd6log((LOG_NOTICE,
2154 "%s: failed to find any ifaddr to add route for a "
2155 "prefix(%s/%d) on %s\n", __func__,
2156 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2157 pr->ndpr_plen, if_name(ifp)));
2158 error = 0;
2159 } else {
2160 error = nd6_prefix_onlink_rtrequest(pr, ifa);
2161 ifa_free(ifa);
2162 }
2163 NET_EPOCH_EXIT(et);
2164
2165 return (error);
2166 }
2167
2168 int
nd6_prefix_offlink(struct nd_prefix * pr)2169 nd6_prefix_offlink(struct nd_prefix *pr)
2170 {
2171 int error = 0;
2172 struct ifnet *ifp = pr->ndpr_ifp;
2173 struct nd_prefix *opr;
2174 char ip6buf[INET6_ADDRSTRLEN];
2175 uint64_t genid;
2176 int a_failure;
2177
2178 ND6_ONLINK_LOCK_ASSERT();
2179 ND6_UNLOCK_ASSERT();
2180
2181 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2182 return (EEXIST);
2183
2184 struct sockaddr_in6 mask6 = {
2185 .sin6_family = AF_INET6,
2186 .sin6_len = sizeof(struct sockaddr_in6),
2187 .sin6_addr = pr->ndpr_mask,
2188 };
2189 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2190
2191 error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE,
2192 &pr->ndpr_prefix, pmask6, ifp, NULL);
2193
2194 a_failure = 1;
2195 if (error == 0) {
2196 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2197
2198 /*
2199 * There might be the same prefix on another interface,
2200 * the prefix which could not be on-link just because we have
2201 * the interface route (see comments in nd6_prefix_onlink).
2202 * If there's one, try to make the prefix on-link on the
2203 * interface.
2204 */
2205 ND6_RLOCK();
2206 restart:
2207 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2208 /*
2209 * KAME specific: detached prefixes should not be
2210 * on-link.
2211 */
2212 if (opr == pr || (opr->ndpr_stateflags &
2213 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2214 continue;
2215
2216 if (opr->ndpr_plen == pr->ndpr_plen &&
2217 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2218 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2219 int e;
2220
2221 genid = V_nd6_list_genid;
2222 ND6_RUNLOCK();
2223 if ((e = nd6_prefix_onlink(opr)) != 0) {
2224 nd6log((LOG_ERR,
2225 "%s: failed to recover a prefix "
2226 "%s/%d from %s to %s (errno=%d)\n",
2227 __func__, ip6_sprintf(ip6buf,
2228 &opr->ndpr_prefix.sin6_addr),
2229 opr->ndpr_plen, if_name(ifp),
2230 if_name(opr->ndpr_ifp), e));
2231 } else
2232 a_failure = 0;
2233 ND6_RLOCK();
2234 if (genid != V_nd6_list_genid)
2235 goto restart;
2236 }
2237 }
2238 ND6_RUNLOCK();
2239 } else {
2240 /* XXX: can we still set the NDPRF_ONLINK flag? */
2241 nd6log((LOG_ERR,
2242 "%s: failed to delete route: %s/%d on %s (errno=%d)\n",
2243 __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2244 pr->ndpr_plen, if_name(ifp), error));
2245 }
2246
2247 if (a_failure)
2248 lltable_prefix_free(AF_INET6,
2249 (struct sockaddr *)&pr->ndpr_prefix,
2250 (struct sockaddr *)&mask6, LLE_STATIC);
2251
2252 return (error);
2253 }
2254
2255 /*
2256 * Get a randomized interface identifier for a temporary address
2257 * Based on RFC 8981, Section 3.3.1.
2258 */
2259 static int
in6_get_tmp_ifid(struct in6_aliasreq * ifra)2260 in6_get_tmp_ifid(struct in6_aliasreq *ifra)
2261 {
2262 struct in6_addr *addr;
2263
2264 if(!is_random_seeded()){
2265 return 1;
2266 }
2267
2268 addr = &(ifra->ifra_addr.sin6_addr);
2269 regen:
2270 ifra->ifra_addr.sin6_addr.s6_addr32[2] |=
2271 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[2]));
2272 ifra->ifra_addr.sin6_addr.s6_addr32[3] |=
2273 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[3]));
2274
2275 /*
2276 * Check if generated address is not inappropriate:
2277 *
2278 * - Reserved IPv6 Interface aIdentifers
2279 * (https://www.iana.org/assignments/ipv6-interface-ids/)
2280 */
2281
2282 /* Subnet-router anycast: 0000:0000:0000:0000 */
2283 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2284 goto regen;
2285
2286 /*
2287 * IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2288 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2289 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2290 */
2291 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2292 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2293 goto regen;
2294
2295 /* Reserved subnet anycast addresses */
2296 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2297 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2298 goto regen;
2299
2300 return 0;
2301 }
2302
2303 /*
2304 * ia0 - corresponding public address
2305 */
2306 int
in6_tmpifadd(const struct in6_ifaddr * ia0,int forcegen,int delay)2307 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2308 {
2309 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2310 struct in6_ifaddr *newia;
2311 struct in6_aliasreq ifra;
2312 int error;
2313 int trylimit = 3; /* XXX: adhoc value */
2314 int updateflags;
2315 time_t vltime0, pltime0;
2316
2317 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2318 &ia0->ia_prefixmask.sin6_addr);
2319
2320 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */
2321 /* clear the old IFID */
2322 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2323 &ifra.ifra_prefixmask.sin6_addr);
2324
2325 again:
2326 if (in6_get_tmp_ifid(&ifra) != 0) {
2327 nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n",
2328 __func__));
2329 return (EINVAL);
2330 }
2331
2332 /*
2333 * in6_get_tmpifid() quite likely provided a unique interface ID.
2334 * However, we may still have a chance to see collision, because
2335 * there may be a time lag between generation of the ID and generation
2336 * of the address. So, we'll do one more sanity check.
2337 */
2338
2339 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2340 if (trylimit-- > 0) {
2341 forcegen = 1;
2342 goto again;
2343 }
2344
2345 /* Give up. Something strange should have happened. */
2346 nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n",
2347 __func__));
2348 return (EEXIST);
2349 }
2350
2351 /*
2352 * The Valid Lifetime is the lower of the Valid Lifetime of the
2353 * public address or TEMP_VALID_LIFETIME.
2354 * The Preferred Lifetime is the lower of the Preferred Lifetime
2355 * of the public address or TEMP_PREFERRED_LIFETIME -
2356 * DESYNC_FACTOR.
2357 */
2358 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2359 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2360 (ia0->ia6_lifetime.ia6t_vltime -
2361 (time_uptime - ia0->ia6_updatetime));
2362 if (vltime0 > V_ip6_temp_valid_lifetime)
2363 vltime0 = V_ip6_temp_valid_lifetime;
2364 } else
2365 vltime0 = V_ip6_temp_valid_lifetime;
2366 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2367 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2368 (ia0->ia6_lifetime.ia6t_pltime -
2369 (time_uptime - ia0->ia6_updatetime));
2370 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2371 pltime0 = V_ip6_temp_preferred_lifetime -
2372 V_ip6_desync_factor;
2373 }
2374 } else
2375 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2376 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2377 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2378
2379 /*
2380 * A temporary address is created only if this calculated Preferred
2381 * Lifetime is greater than REGEN_ADVANCE time units.
2382 */
2383 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2384 return (0);
2385
2386 /* XXX: scope zone ID? */
2387
2388 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2389
2390 /* allocate ifaddr structure, link into chain, etc. */
2391 updateflags = 0;
2392 if (delay)
2393 updateflags |= IN6_IFAUPDATE_DADDELAY;
2394 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2395 return (error);
2396
2397 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2398 if (newia == NULL) { /* XXX: can it happen? */
2399 nd6log((LOG_ERR,
2400 "%s: ifa update succeeded, but we got no ifaddr\n",
2401 __func__));
2402 return (EINVAL); /* XXX */
2403 }
2404 newia->ia6_ndpr = ia0->ia6_ndpr;
2405 newia->ia6_ndpr->ndpr_addrcnt++;
2406 ifa_free(&newia->ia_ifa);
2407
2408 /*
2409 * A newly added address might affect the status of other addresses.
2410 * XXX: when the temporary address is generated with a new public
2411 * address, the onlink check is redundant. However, it would be safe
2412 * to do the check explicitly everywhere a new address is generated,
2413 * and, in fact, we surely need the check when we create a new
2414 * temporary address due to deprecation of an old temporary address.
2415 */
2416 pfxlist_onlink_check();
2417
2418 return (0);
2419 }
2420
2421 static int
rt6_deleteroute(const struct rtentry * rt,const struct nhop_object * nh,void * arg)2422 rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh,
2423 void *arg)
2424 {
2425 struct in6_addr *gate = (struct in6_addr *)arg;
2426 int nh_rt_flags;
2427
2428 if (nh->gw_sa.sa_family != AF_INET6)
2429 return (0);
2430
2431 if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) {
2432 return (0);
2433 }
2434
2435 /*
2436 * Do not delete a static route.
2437 * XXX: this seems to be a bit ad-hoc. Should we consider the
2438 * 'cloned' bit instead?
2439 */
2440 nh_rt_flags = nhop_get_rtflags(nh);
2441 if ((nh_rt_flags & RTF_STATIC) != 0)
2442 return (0);
2443
2444 /*
2445 * We delete only host route. This means, in particular, we don't
2446 * delete default route.
2447 */
2448 if ((nh_rt_flags & RTF_HOST) == 0)
2449 return (0);
2450
2451 return (1);
2452 #undef SIN6
2453 }
2454
2455 /*
2456 * Delete all the routing table entries that use the specified gateway.
2457 * XXX: this function causes search through all entries of routing table, so
2458 * it shouldn't be called when acting as a router.
2459 */
2460 void
rt6_flush(struct in6_addr * gateway,struct ifnet * ifp)2461 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2462 {
2463
2464 /* We'll care only link-local addresses */
2465 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2466 return;
2467
2468 /* XXX Do we really need to walk any but the default FIB? */
2469 rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2470 }
2471
2472 int
nd6_setdefaultiface(int ifindex)2473 nd6_setdefaultiface(int ifindex)
2474 {
2475
2476 if (V_nd6_defifindex != ifindex) {
2477 V_nd6_defifindex = ifindex;
2478 if (V_nd6_defifindex != 0) {
2479 struct epoch_tracker et;
2480
2481 /*
2482 * XXXGL: this function should use ifnet_byindex_ref!
2483 */
2484 NET_EPOCH_ENTER(et);
2485 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2486 NET_EPOCH_EXIT(et);
2487 if (V_nd6_defifp == NULL)
2488 return (EINVAL);
2489 } else
2490 V_nd6_defifp = NULL;
2491
2492 /*
2493 * Our current implementation assumes one-to-one mapping between
2494 * interfaces and links, so it would be natural to use the
2495 * default interface as the default link.
2496 */
2497 scope6_setdefault(V_nd6_defifp);
2498 }
2499
2500 return (0);
2501 }
2502
2503 bool
nd6_defrouter_list_empty(void)2504 nd6_defrouter_list_empty(void)
2505 {
2506
2507 return (TAILQ_EMPTY(&V_nd6_defrouter));
2508 }
2509
2510 void
nd6_defrouter_timer(void)2511 nd6_defrouter_timer(void)
2512 {
2513 struct nd_defrouter *dr, *ndr;
2514 struct nd6_drhead drq;
2515
2516 TAILQ_INIT(&drq);
2517
2518 ND6_WLOCK();
2519 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr)
2520 if (dr->expire && dr->expire < time_uptime)
2521 defrouter_unlink(dr, &drq);
2522 ND6_WUNLOCK();
2523
2524 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2525 TAILQ_REMOVE(&drq, dr, dr_entry);
2526 defrouter_del(dr);
2527 }
2528 }
2529
2530 /*
2531 * Nuke default router list entries toward ifp.
2532 * We defer removal of default router list entries that is installed in the
2533 * routing table, in order to keep additional side effects as small as possible.
2534 */
2535 void
nd6_defrouter_purge(struct ifnet * ifp)2536 nd6_defrouter_purge(struct ifnet *ifp)
2537 {
2538 struct nd_defrouter *dr, *ndr;
2539 struct nd6_drhead drq;
2540
2541 TAILQ_INIT(&drq);
2542
2543 ND6_WLOCK();
2544 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2545 if (dr->installed)
2546 continue;
2547 if (dr->ifp == ifp)
2548 defrouter_unlink(dr, &drq);
2549 }
2550 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2551 if (!dr->installed)
2552 continue;
2553 if (dr->ifp == ifp)
2554 defrouter_unlink(dr, &drq);
2555 }
2556 ND6_WUNLOCK();
2557
2558 /* Delete the unlinked router objects. */
2559 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2560 TAILQ_REMOVE(&drq, dr, dr_entry);
2561 defrouter_del(dr);
2562 }
2563 }
2564
2565 void
nd6_defrouter_flush_all(void)2566 nd6_defrouter_flush_all(void)
2567 {
2568 struct nd_defrouter *dr;
2569 struct nd6_drhead drq;
2570
2571 TAILQ_INIT(&drq);
2572
2573 ND6_WLOCK();
2574 while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL)
2575 defrouter_unlink(dr, &drq);
2576 ND6_WUNLOCK();
2577
2578 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2579 TAILQ_REMOVE(&drq, dr, dr_entry);
2580 defrouter_del(dr);
2581 }
2582 }
2583
2584 void
nd6_defrouter_init(void)2585 nd6_defrouter_init(void)
2586 {
2587
2588 TAILQ_INIT(&V_nd6_defrouter);
2589 }
2590
2591 static int
nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)2592 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2593 {
2594 struct in6_defrouter d;
2595 struct nd_defrouter *dr;
2596 int error;
2597
2598 if (req->newptr != NULL)
2599 return (EPERM);
2600
2601 error = sysctl_wire_old_buffer(req, 0);
2602 if (error != 0)
2603 return (error);
2604
2605 bzero(&d, sizeof(d));
2606 d.rtaddr.sin6_family = AF_INET6;
2607 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2608
2609 ND6_RLOCK();
2610 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
2611 d.rtaddr.sin6_addr = dr->rtaddr;
2612 error = sa6_recoverscope(&d.rtaddr);
2613 if (error != 0)
2614 break;
2615 d.flags = dr->raflags;
2616 d.rtlifetime = dr->rtlifetime;
2617 d.expire = dr->expire + (time_second - time_uptime);
2618 d.if_index = dr->ifp->if_index;
2619 error = SYSCTL_OUT(req, &d, sizeof(d));
2620 if (error != 0)
2621 break;
2622 }
2623 ND6_RUNLOCK();
2624 return (error);
2625 }
2626 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2627 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2628 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2629 "NDP default router list");
2630