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, *ib;
1186 int error, plen0;
1187 struct in6_addr 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 else
1218 return NULL;
1219
1220 /* prefixlen + ifidlen must be equal to 128 */
1221 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1222 if (prefixlen != plen0) {
1223 ifa_free(ifa);
1224 nd6log((LOG_INFO,
1225 "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n",
1226 __func__, if_name(ifp), prefixlen, 128 - plen0));
1227 return NULL;
1228 }
1229
1230 /* make ifaddr */
1231 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1232
1233 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1234 /* interface ID */
1235 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1236 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1237 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1238 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1239 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1240 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1241 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1242 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1243 ifa_free(ifa);
1244
1245 /* lifetimes. */
1246 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1247 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1248
1249 /* XXX: scope zone ID? */
1250
1251 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1252
1253 /*
1254 * Make sure that we do not have this address already. This should
1255 * usually not happen, but we can still see this case, e.g., if we
1256 * have manually configured the exact address to be configured.
1257 */
1258 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1259 &ifra.ifra_addr.sin6_addr);
1260 if (ifa != NULL) {
1261 ifa_free(ifa);
1262 /* this should be rare enough to make an explicit log */
1263 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1264 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1265 return (NULL);
1266 }
1267
1268 /*
1269 * Allocate ifaddr structure, link into chain, etc.
1270 * If we are going to create a new address upon receiving a multicasted
1271 * RA, we need to impose a random delay before starting DAD.
1272 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1273 */
1274 updateflags = 0;
1275 if (mcast)
1276 updateflags |= IN6_IFAUPDATE_DADDELAY;
1277 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1278 nd6log((LOG_ERR,
1279 "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__,
1280 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1281 if_name(ifp), error));
1282 return (NULL); /* ifaddr must not have been allocated. */
1283 }
1284
1285 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1286 /*
1287 * XXXRW: Assumption of non-NULLness here might not be true with
1288 * fine-grained locking -- should we validate it? Or just return
1289 * earlier ifa rather than looking it up again?
1290 */
1291 return (ia); /* this is always non-NULL and referenced. */
1292 }
1293
1294 static struct nd_prefix *
nd6_prefix_lookup_locked(struct nd_prefixctl * key)1295 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1296 {
1297 struct nd_prefix *search;
1298
1299 ND6_LOCK_ASSERT();
1300
1301 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1302 if (key->ndpr_ifp == search->ndpr_ifp &&
1303 key->ndpr_plen == search->ndpr_plen &&
1304 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1305 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1306 nd6_prefix_ref(search);
1307 break;
1308 }
1309 }
1310 return (search);
1311 }
1312
1313 struct nd_prefix *
nd6_prefix_lookup(struct nd_prefixctl * key)1314 nd6_prefix_lookup(struct nd_prefixctl *key)
1315 {
1316 struct nd_prefix *search;
1317
1318 ND6_RLOCK();
1319 search = nd6_prefix_lookup_locked(key);
1320 ND6_RUNLOCK();
1321 return (search);
1322 }
1323
1324 void
nd6_prefix_ref(struct nd_prefix * pr)1325 nd6_prefix_ref(struct nd_prefix *pr)
1326 {
1327
1328 refcount_acquire(&pr->ndpr_refcnt);
1329 }
1330
1331 void
nd6_prefix_rele(struct nd_prefix * pr)1332 nd6_prefix_rele(struct nd_prefix *pr)
1333 {
1334
1335 if (refcount_release(&pr->ndpr_refcnt)) {
1336 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1337 ("prefix %p has advertising routers", pr));
1338 free(pr, M_IP6NDP);
1339 }
1340 }
1341
1342 int
nd6_prelist_add(struct nd_prefixctl * pr,struct nd_defrouter * dr,struct nd_prefix ** newp)1343 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1344 struct nd_prefix **newp)
1345 {
1346 struct nd_prefix *new;
1347 char ip6buf[INET6_ADDRSTRLEN];
1348 int error;
1349
1350 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1351 if (new == NULL)
1352 return (ENOMEM);
1353 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1354 new->ndpr_ifp = pr->ndpr_ifp;
1355 new->ndpr_prefix = pr->ndpr_prefix;
1356 new->ndpr_plen = pr->ndpr_plen;
1357 new->ndpr_vltime = pr->ndpr_vltime;
1358 new->ndpr_pltime = pr->ndpr_pltime;
1359 new->ndpr_flags = pr->ndpr_flags;
1360 if ((error = in6_init_prefix_ltimes(new)) != 0) {
1361 free(new, M_IP6NDP);
1362 return (error);
1363 }
1364 new->ndpr_lastupdate = time_uptime;
1365
1366 /* initialization */
1367 LIST_INIT(&new->ndpr_advrtrs);
1368 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1369 /* make prefix in the canonical form */
1370 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1371
1372 ND6_WLOCK();
1373 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1374 V_nd6_list_genid++;
1375 ND6_WUNLOCK();
1376
1377 /* ND_OPT_PI_FLAG_ONLINK processing */
1378 if (new->ndpr_raf_onlink) {
1379 struct epoch_tracker et;
1380
1381 ND6_ONLINK_LOCK();
1382 NET_EPOCH_ENTER(et);
1383 if ((error = nd6_prefix_onlink(new)) != 0) {
1384 nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d "
1385 "on-link on %s (errno=%d)\n", __func__,
1386 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1387 pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1388 /* proceed anyway. XXX: is it correct? */
1389 }
1390 NET_EPOCH_EXIT(et);
1391 ND6_ONLINK_UNLOCK();
1392 }
1393
1394 if (dr != NULL)
1395 pfxrtr_add(new, dr);
1396 if (newp != NULL)
1397 *newp = new;
1398 return (0);
1399 }
1400
1401 /*
1402 * Remove a prefix from the prefix list and optionally stash it in a
1403 * caller-provided list.
1404 *
1405 * The ND6 lock must be held.
1406 */
1407 void
nd6_prefix_unlink(struct nd_prefix * pr,struct nd_prhead * list)1408 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1409 {
1410
1411 ND6_WLOCK_ASSERT();
1412
1413 LIST_REMOVE(pr, ndpr_entry);
1414 V_nd6_list_genid++;
1415 if (list != NULL)
1416 LIST_INSERT_HEAD(list, pr, ndpr_entry);
1417 }
1418
1419 /*
1420 * Free an unlinked prefix, first marking it off-link if necessary.
1421 */
1422 void
nd6_prefix_del(struct nd_prefix * pr)1423 nd6_prefix_del(struct nd_prefix *pr)
1424 {
1425 struct nd_pfxrouter *pfr, *next;
1426 int e;
1427 char ip6buf[INET6_ADDRSTRLEN];
1428
1429 KASSERT(pr->ndpr_addrcnt == 0,
1430 ("prefix %p has referencing addresses", pr));
1431 ND6_UNLOCK_ASSERT();
1432
1433 /*
1434 * Though these flags are now meaningless, we'd rather keep the value
1435 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1436 * when executing "ndp -p".
1437 */
1438 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1439 ND6_ONLINK_LOCK();
1440 if ((e = nd6_prefix_offlink(pr)) != 0) {
1441 nd6log((LOG_ERR,
1442 "%s: failed to make the prefix %s/%d offlink on %s "
1443 "(errno=%d)\n", __func__,
1444 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1445 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1446 /* what should we do? */
1447 }
1448 ND6_ONLINK_UNLOCK();
1449 }
1450
1451 /* Release references to routers that have advertised this prefix. */
1452 ND6_WLOCK();
1453 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1454 pfxrtr_del(pfr);
1455 ND6_WUNLOCK();
1456
1457 nd6_prefix_rele(pr);
1458
1459 pfxlist_onlink_check();
1460 }
1461
1462 static int
prelist_update(struct nd_prefixctl * new,struct nd_defrouter * dr,struct mbuf * m,int mcast)1463 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1464 struct mbuf *m, int mcast)
1465 {
1466 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1467 struct ifaddr *ifa;
1468 struct ifnet *ifp = new->ndpr_ifp;
1469 struct nd_prefix *pr;
1470 int error = 0;
1471 int auth;
1472 struct in6_addrlifetime lt6_tmp;
1473 char ip6buf[INET6_ADDRSTRLEN];
1474
1475 NET_EPOCH_ASSERT();
1476
1477 auth = 0;
1478 if (m) {
1479 /*
1480 * Authenticity for NA consists authentication for
1481 * both IP header and IP datagrams, doesn't it ?
1482 */
1483 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1484 auth = ((m->m_flags & M_AUTHIPHDR) &&
1485 (m->m_flags & M_AUTHIPDGM));
1486 #endif
1487 }
1488
1489 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1490 /*
1491 * nd6_prefix_lookup() ensures that pr and new have the same
1492 * prefix on a same interface.
1493 */
1494
1495 /*
1496 * Update prefix information. Note that the on-link (L) bit
1497 * and the autonomous (A) bit should NOT be changed from 1
1498 * to 0.
1499 */
1500 if (new->ndpr_raf_onlink == 1)
1501 pr->ndpr_raf_onlink = 1;
1502 if (new->ndpr_raf_auto == 1)
1503 pr->ndpr_raf_auto = 1;
1504 if (new->ndpr_raf_onlink) {
1505 pr->ndpr_vltime = new->ndpr_vltime;
1506 pr->ndpr_pltime = new->ndpr_pltime;
1507 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1508 pr->ndpr_lastupdate = time_uptime;
1509 }
1510
1511 if (new->ndpr_raf_onlink &&
1512 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1513 ND6_ONLINK_LOCK();
1514 if ((error = nd6_prefix_onlink(pr)) != 0) {
1515 nd6log((LOG_ERR,
1516 "%s: failed to make the prefix %s/%d "
1517 "on-link on %s (errno=%d)\n", __func__,
1518 ip6_sprintf(ip6buf,
1519 &pr->ndpr_prefix.sin6_addr),
1520 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1521 error));
1522 /* proceed anyway. XXX: is it correct? */
1523 }
1524 ND6_ONLINK_UNLOCK();
1525 }
1526
1527 if (dr != NULL)
1528 pfxrtr_add(pr, dr);
1529 } else {
1530 if (new->ndpr_vltime == 0)
1531 goto end;
1532 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1533 goto end;
1534
1535 error = nd6_prelist_add(new, dr, &pr);
1536 if (error != 0) {
1537 nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for "
1538 "the prefix %s/%d on %s (errno=%d)\n", __func__,
1539 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1540 new->ndpr_plen, if_name(new->ndpr_ifp), error));
1541 goto end; /* we should just give up in this case. */
1542 }
1543
1544 /*
1545 * XXX: from the ND point of view, we can ignore a prefix
1546 * with the on-link bit being zero. However, we need a
1547 * prefix structure for references from autoconfigured
1548 * addresses. Thus, we explicitly make sure that the prefix
1549 * itself expires now.
1550 */
1551 if (pr->ndpr_raf_onlink == 0) {
1552 pr->ndpr_vltime = 0;
1553 pr->ndpr_pltime = 0;
1554 in6_init_prefix_ltimes(pr);
1555 }
1556 }
1557
1558 /*
1559 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1560 * Note that pr must be non NULL at this point.
1561 */
1562
1563 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1564 if (!new->ndpr_raf_auto)
1565 goto end;
1566
1567 /*
1568 * 5.5.3 (b). the link-local prefix should have been ignored in
1569 * nd6_ra_input.
1570 */
1571
1572 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1573 if (new->ndpr_pltime > new->ndpr_vltime) {
1574 error = EINVAL; /* XXX: won't be used */
1575 goto end;
1576 }
1577
1578 /*
1579 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1580 * an address configured by stateless autoconfiguration already in the
1581 * list of addresses associated with the interface, and the Valid
1582 * Lifetime is not 0, form an address. We first check if we have
1583 * a matching prefix.
1584 * Note: we apply a clarification in rfc2462bis-02 here. We only
1585 * consider autoconfigured addresses while RFC2462 simply said
1586 * "address".
1587 */
1588 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1589 struct in6_ifaddr *ifa6;
1590 u_int32_t remaininglifetime;
1591
1592 if (ifa->ifa_addr->sa_family != AF_INET6)
1593 continue;
1594
1595 ifa6 = (struct in6_ifaddr *)ifa;
1596
1597 /*
1598 * We only consider autoconfigured addresses as per rfc2462bis.
1599 */
1600 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1601 continue;
1602
1603 /*
1604 * Spec is not clear here, but I believe we should concentrate
1605 * on unicast (i.e. not anycast) addresses.
1606 * XXX: other ia6_flags? detached or duplicated?
1607 */
1608 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1609 continue;
1610
1611 /*
1612 * Ignore the address if it is not associated with a prefix
1613 * or is associated with a prefix that is different from this
1614 * one. (pr is never NULL here)
1615 */
1616 if (ifa6->ia6_ndpr != pr)
1617 continue;
1618
1619 if (ia6_match == NULL) /* remember the first one */
1620 ia6_match = ifa6;
1621
1622 /*
1623 * An already autoconfigured address matched. Now that we
1624 * are sure there is at least one matched address, we can
1625 * proceed to 5.5.3. (e): update the lifetimes according to the
1626 * "two hours" rule and the privacy extension.
1627 * We apply some clarifications in rfc2462bis:
1628 * - use remaininglifetime instead of storedlifetime as a
1629 * variable name
1630 * - remove the dead code in the "two-hour" rule
1631 */
1632 #define TWOHOUR (120*60)
1633 lt6_tmp = ifa6->ia6_lifetime;
1634
1635 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1636 remaininglifetime = ND6_INFINITE_LIFETIME;
1637 else if (time_uptime - ifa6->ia6_updatetime >
1638 lt6_tmp.ia6t_vltime) {
1639 /*
1640 * The case of "invalid" address. We should usually
1641 * not see this case.
1642 */
1643 remaininglifetime = 0;
1644 } else
1645 remaininglifetime = lt6_tmp.ia6t_vltime -
1646 (time_uptime - ifa6->ia6_updatetime);
1647
1648 /* when not updating, keep the current stored lifetime. */
1649 lt6_tmp.ia6t_vltime = remaininglifetime;
1650
1651 if (TWOHOUR < new->ndpr_vltime ||
1652 remaininglifetime < new->ndpr_vltime) {
1653 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1654 } else if (remaininglifetime <= TWOHOUR) {
1655 if (auth) {
1656 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1657 }
1658 } else {
1659 /*
1660 * new->ndpr_vltime <= TWOHOUR &&
1661 * TWOHOUR < remaininglifetime
1662 */
1663 lt6_tmp.ia6t_vltime = TWOHOUR;
1664 }
1665
1666 /* The 2 hour rule is not imposed for preferred lifetime. */
1667 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1668
1669 in6_init_address_ltimes(pr, <6_tmp);
1670
1671 /*
1672 * We need to treat lifetimes for temporary addresses
1673 * differently, according to
1674 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1675 * we only update the lifetimes when they are in the maximum
1676 * intervals.
1677 */
1678 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1679 u_int32_t maxvltime, maxpltime;
1680
1681 if (V_ip6_temp_valid_lifetime >
1682 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1683 V_ip6_desync_factor)) {
1684 maxvltime = V_ip6_temp_valid_lifetime -
1685 (time_uptime - ifa6->ia6_createtime) -
1686 V_ip6_desync_factor;
1687 } else
1688 maxvltime = 0;
1689 if (V_ip6_temp_preferred_lifetime >
1690 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1691 V_ip6_desync_factor)) {
1692 maxpltime = V_ip6_temp_preferred_lifetime -
1693 (time_uptime - ifa6->ia6_createtime) -
1694 V_ip6_desync_factor;
1695 } else
1696 maxpltime = 0;
1697
1698 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1699 lt6_tmp.ia6t_vltime > maxvltime) {
1700 lt6_tmp.ia6t_vltime = maxvltime;
1701 }
1702 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1703 lt6_tmp.ia6t_pltime > maxpltime) {
1704 lt6_tmp.ia6t_pltime = maxpltime;
1705 }
1706 }
1707 ifa6->ia6_lifetime = lt6_tmp;
1708 ifa6->ia6_updatetime = time_uptime;
1709 }
1710 if (ia6_match == NULL && new->ndpr_vltime) {
1711 int ifidlen;
1712
1713 /*
1714 * 5.5.3 (d) (continued)
1715 * No address matched and the valid lifetime is non-zero.
1716 * Create a new address.
1717 */
1718
1719 /*
1720 * Prefix Length check:
1721 * If the sum of the prefix length and interface identifier
1722 * length does not equal 128 bits, the Prefix Information
1723 * option MUST be ignored. The length of the interface
1724 * identifier is defined in a separate link-type specific
1725 * document.
1726 */
1727 ifidlen = in6_if2idlen(ifp);
1728 if (ifidlen < 0) {
1729 /* this should not happen, so we always log it. */
1730 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1731 if_name(ifp));
1732 goto end;
1733 }
1734 if (ifidlen + pr->ndpr_plen != 128) {
1735 nd6log((LOG_INFO,
1736 "%s: invalid prefixlen %d for %s, ignored\n",
1737 __func__, pr->ndpr_plen, if_name(ifp)));
1738 goto end;
1739 }
1740
1741 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1742 /*
1743 * note that we should use pr (not new) for reference.
1744 */
1745 pr->ndpr_addrcnt++;
1746 ia6->ia6_ndpr = pr;
1747
1748 /*
1749 * RFC 3041 3.3 (2).
1750 * When a new public address is created as described
1751 * in RFC2462, also create a new temporary address.
1752 *
1753 * RFC 3041 3.5.
1754 * When an interface connects to a new link, a new
1755 * randomized interface identifier should be generated
1756 * immediately together with a new set of temporary
1757 * addresses. Thus, we specifiy 1 as the 2nd arg of
1758 * in6_tmpifadd().
1759 */
1760 if (V_ip6_use_tempaddr) {
1761 int e;
1762 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1763 nd6log((LOG_NOTICE, "%s: failed to "
1764 "create a temporary address "
1765 "(errno=%d)\n", __func__, e));
1766 }
1767 }
1768 ifa_free(&ia6->ia_ifa);
1769
1770 /*
1771 * A newly added address might affect the status
1772 * of other addresses, so we check and update it.
1773 * XXX: what if address duplication happens?
1774 */
1775 pfxlist_onlink_check();
1776 } else {
1777 /* just set an error. do not bark here. */
1778 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1779 }
1780 }
1781
1782 end:
1783 if (pr != NULL)
1784 nd6_prefix_rele(pr);
1785 return (error);
1786 }
1787
1788 /*
1789 * A supplement function used in the on-link detection below;
1790 * detect if a given prefix has a (probably) reachable advertising router.
1791 * XXX: lengthy function name...
1792 */
1793 static struct nd_pfxrouter *
find_pfxlist_reachable_router(struct nd_prefix * pr)1794 find_pfxlist_reachable_router(struct nd_prefix *pr)
1795 {
1796 struct epoch_tracker et;
1797 struct nd_pfxrouter *pfxrtr;
1798
1799 ND6_LOCK_ASSERT();
1800
1801 NET_EPOCH_ENTER(et);
1802 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1803 if (is_dr_reachable(pfxrtr->router))
1804 break;
1805 }
1806 NET_EPOCH_EXIT(et);
1807 return (pfxrtr);
1808 }
1809
1810 /*
1811 * Check if each prefix in the prefix list has at least one available router
1812 * that advertised the prefix (a router is "available" if its neighbor cache
1813 * entry is reachable or probably reachable).
1814 * If the check fails, the prefix may be off-link, because, for example,
1815 * we have moved from the network but the lifetime of the prefix has not
1816 * expired yet. So we should not use the prefix if there is another prefix
1817 * that has an available router.
1818 * But, if there is no prefix that has an available router, we still regard
1819 * all the prefixes as on-link. This is because we can't tell if all the
1820 * routers are simply dead or if we really moved from the network and there
1821 * is no router around us.
1822 */
1823 void
pfxlist_onlink_check(void)1824 pfxlist_onlink_check(void)
1825 {
1826 struct nd_prefix *pr;
1827 struct in6_ifaddr *ifa;
1828 struct nd_defrouter *dr;
1829 struct nd_pfxrouter *pfxrtr = NULL;
1830 struct rm_priotracker in6_ifa_tracker;
1831 uint64_t genid;
1832 uint32_t flags;
1833
1834 ND6_ONLINK_LOCK();
1835 ND6_RLOCK();
1836
1837 /*
1838 * Check if there is a prefix that has a reachable advertising
1839 * router.
1840 */
1841 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1842 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1843 break;
1844 }
1845
1846 /*
1847 * If we have no such prefix, check whether we still have a router
1848 * that does not advertise any prefixes.
1849 */
1850 if (pr == NULL) {
1851 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1852 struct nd_prefix *pr0;
1853
1854 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1855 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1856 break;
1857 }
1858 if (pfxrtr != NULL)
1859 break;
1860 }
1861 }
1862 if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) {
1863 /*
1864 * There is at least one prefix that has a reachable router,
1865 * or at least a router which probably does not advertise
1866 * any prefixes. The latter would be the case when we move
1867 * to a new link where we have a router that does not provide
1868 * prefixes and we configure an address by hand.
1869 * Detach prefixes which have no reachable advertising
1870 * router, and attach other prefixes.
1871 */
1872 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1873 /* XXX: a link-local prefix should never be detached */
1874 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1875 pr->ndpr_raf_onlink == 0 ||
1876 pr->ndpr_raf_auto == 0)
1877 continue;
1878
1879 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1880 find_pfxlist_reachable_router(pr) == NULL)
1881 pr->ndpr_stateflags |= NDPRF_DETACHED;
1882 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1883 find_pfxlist_reachable_router(pr) != NULL)
1884 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1885 }
1886 } else {
1887 /* there is no prefix that has a reachable router */
1888 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1889 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1890 pr->ndpr_raf_onlink == 0 ||
1891 pr->ndpr_raf_auto == 0)
1892 continue;
1893 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1894 }
1895 }
1896
1897 /*
1898 * Remove each interface route associated with a (just) detached
1899 * prefix, and reinstall the interface route for a (just) attached
1900 * prefix. Note that all attempt of reinstallation does not
1901 * necessarily success, when a same prefix is shared among multiple
1902 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1903 * so we don't have to care about them.
1904 */
1905 restart:
1906 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1907 char ip6buf[INET6_ADDRSTRLEN];
1908 int e;
1909
1910 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1911 pr->ndpr_raf_onlink == 0 ||
1912 pr->ndpr_raf_auto == 0)
1913 continue;
1914
1915 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1916 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1917 genid = V_nd6_list_genid;
1918 ND6_RUNLOCK();
1919 if ((flags & NDPRF_ONLINK) != 0 &&
1920 (e = nd6_prefix_offlink(pr)) != 0) {
1921 nd6log((LOG_ERR,
1922 "%s: failed to make %s/%d offlink "
1923 "(errno=%d)\n", __func__,
1924 ip6_sprintf(ip6buf,
1925 &pr->ndpr_prefix.sin6_addr),
1926 pr->ndpr_plen, e));
1927 } else if ((flags & NDPRF_ONLINK) == 0 &&
1928 (e = nd6_prefix_onlink(pr)) != 0) {
1929 nd6log((LOG_ERR,
1930 "%s: failed to make %s/%d onlink "
1931 "(errno=%d)\n", __func__,
1932 ip6_sprintf(ip6buf,
1933 &pr->ndpr_prefix.sin6_addr),
1934 pr->ndpr_plen, e));
1935 }
1936 ND6_RLOCK();
1937 if (genid != V_nd6_list_genid)
1938 goto restart;
1939 }
1940 }
1941
1942 /*
1943 * Changes on the prefix status might affect address status as well.
1944 * Make sure that all addresses derived from an attached prefix are
1945 * attached, and that all addresses derived from a detached prefix are
1946 * detached. Note, however, that a manually configured address should
1947 * always be attached.
1948 * The precise detection logic is same as the one for prefixes.
1949 */
1950 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1951 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1952 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1953 continue;
1954
1955 if (ifa->ia6_ndpr == NULL) {
1956 /*
1957 * This can happen when we first configure the address
1958 * (i.e. the address exists, but the prefix does not).
1959 * XXX: complicated relationships...
1960 */
1961 continue;
1962 }
1963
1964 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1965 break;
1966 }
1967 if (ifa) {
1968 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1969 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1970 continue;
1971
1972 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1973 continue;
1974
1975 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1976 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1977 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1978 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1979 nd6_dad_start((struct ifaddr *)ifa, 0);
1980 }
1981 } else {
1982 ifa->ia6_flags |= IN6_IFF_DETACHED;
1983 }
1984 }
1985 } else {
1986 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1987 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1988 continue;
1989
1990 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1991 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1992 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1993 /* Do we need a delay in this case? */
1994 nd6_dad_start((struct ifaddr *)ifa, 0);
1995 }
1996 }
1997 }
1998 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1999 ND6_RUNLOCK();
2000 ND6_ONLINK_UNLOCK();
2001 }
2002
2003 /*
2004 * Add or remove interface route specified by @dst, @netmask and @ifp.
2005 * ifa can be NULL.
2006 * Returns 0 on success
2007 */
2008 static int
nd6_prefix_rtrequest(uint32_t fibnum,int cmd,struct sockaddr_in6 * dst,struct sockaddr_in6 * netmask,struct ifnet * ifp,struct ifaddr * ifa)2009 nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst,
2010 struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa)
2011 {
2012 struct epoch_tracker et;
2013 int error;
2014
2015 /* Prepare gateway */
2016 struct sockaddr_dl_short sdl = {
2017 .sdl_family = AF_LINK,
2018 .sdl_len = sizeof(struct sockaddr_dl_short),
2019 .sdl_type = ifp->if_type,
2020 .sdl_index = ifp->if_index,
2021 };
2022
2023 struct rt_addrinfo info = {
2024 .rti_ifa = ifa,
2025 .rti_ifp = ifp,
2026 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
2027 .rti_info = {
2028 [RTAX_DST] = (struct sockaddr *)dst,
2029 [RTAX_NETMASK] = (struct sockaddr *)netmask,
2030 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
2031 },
2032 };
2033 /* Don't set additional per-gw filters on removal */
2034
2035 NET_EPOCH_ENTER(et);
2036 error = rib_handle_ifaddr_info(fibnum, cmd, &info);
2037 NET_EPOCH_EXIT(et);
2038 return (error);
2039 }
2040
2041 static int
nd6_prefix_onlink_rtrequest(struct nd_prefix * pr,struct ifaddr * ifa)2042 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
2043 {
2044 int error;
2045
2046 struct sockaddr_in6 mask6 = {
2047 .sin6_family = AF_INET6,
2048 .sin6_len = sizeof(struct sockaddr_in6),
2049 .sin6_addr = pr->ndpr_mask,
2050 };
2051 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2052
2053 error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD,
2054 &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa);
2055 if (error == 0)
2056 pr->ndpr_stateflags |= NDPRF_ONLINK;
2057
2058 return (error);
2059 }
2060
2061 static int
nd6_prefix_onlink(struct nd_prefix * pr)2062 nd6_prefix_onlink(struct nd_prefix *pr)
2063 {
2064 struct epoch_tracker et;
2065 struct ifaddr *ifa;
2066 struct ifnet *ifp = pr->ndpr_ifp;
2067 struct nd_prefix *opr;
2068 char ip6buf[INET6_ADDRSTRLEN];
2069 int error;
2070
2071 ND6_ONLINK_LOCK_ASSERT();
2072 ND6_UNLOCK_ASSERT();
2073
2074 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
2075 return (EEXIST);
2076
2077 /*
2078 * Add the interface route associated with the prefix. Before
2079 * installing the route, check if there's the same prefix on another
2080 * interface, and the prefix has already installed the interface route.
2081 * Although such a configuration is expected to be rare, we explicitly
2082 * allow it.
2083 */
2084 ND6_RLOCK();
2085 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2086 if (opr == pr)
2087 continue;
2088
2089 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2090 continue;
2091
2092 if (!V_rt_add_addr_allfibs &&
2093 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
2094 continue;
2095
2096 if (opr->ndpr_plen == pr->ndpr_plen &&
2097 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2098 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2099 ND6_RUNLOCK();
2100 return (0);
2101 }
2102 }
2103 ND6_RUNLOCK();
2104
2105 /*
2106 * We prefer link-local addresses as the associated interface address.
2107 */
2108 /* search for a link-local addr */
2109 NET_EPOCH_ENTER(et);
2110 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
2111 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
2112 if (ifa == NULL) {
2113 /* XXX: freebsd does not have ifa_ifwithaf */
2114 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2115 if (ifa->ifa_addr->sa_family == AF_INET6) {
2116 ifa_ref(ifa);
2117 break;
2118 }
2119 }
2120 /* should we care about ia6_flags? */
2121 }
2122 if (ifa == NULL) {
2123 /*
2124 * This can still happen, when, for example, we receive an RA
2125 * containing a prefix with the L bit set and the A bit clear,
2126 * after removing all IPv6 addresses on the receiving
2127 * interface. This should, of course, be rare though.
2128 */
2129 nd6log((LOG_NOTICE,
2130 "%s: failed to find any ifaddr to add route for a "
2131 "prefix(%s/%d) on %s\n", __func__,
2132 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2133 pr->ndpr_plen, if_name(ifp)));
2134 error = 0;
2135 } else {
2136 error = nd6_prefix_onlink_rtrequest(pr, ifa);
2137 ifa_free(ifa);
2138 }
2139 NET_EPOCH_EXIT(et);
2140
2141 return (error);
2142 }
2143
2144 int
nd6_prefix_offlink(struct nd_prefix * pr)2145 nd6_prefix_offlink(struct nd_prefix *pr)
2146 {
2147 int error = 0;
2148 struct ifnet *ifp = pr->ndpr_ifp;
2149 struct nd_prefix *opr;
2150 char ip6buf[INET6_ADDRSTRLEN];
2151 uint64_t genid;
2152 int a_failure;
2153
2154 ND6_ONLINK_LOCK_ASSERT();
2155 ND6_UNLOCK_ASSERT();
2156
2157 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2158 return (EEXIST);
2159
2160 struct sockaddr_in6 mask6 = {
2161 .sin6_family = AF_INET6,
2162 .sin6_len = sizeof(struct sockaddr_in6),
2163 .sin6_addr = pr->ndpr_mask,
2164 };
2165 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2166
2167 error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE,
2168 &pr->ndpr_prefix, pmask6, ifp, NULL);
2169
2170 a_failure = 1;
2171 if (error == 0) {
2172 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2173
2174 /*
2175 * There might be the same prefix on another interface,
2176 * the prefix which could not be on-link just because we have
2177 * the interface route (see comments in nd6_prefix_onlink).
2178 * If there's one, try to make the prefix on-link on the
2179 * interface.
2180 */
2181 ND6_RLOCK();
2182 restart:
2183 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2184 /*
2185 * KAME specific: detached prefixes should not be
2186 * on-link.
2187 */
2188 if (opr == pr || (opr->ndpr_stateflags &
2189 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2190 continue;
2191
2192 if (opr->ndpr_plen == pr->ndpr_plen &&
2193 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2194 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2195 int e;
2196
2197 genid = V_nd6_list_genid;
2198 ND6_RUNLOCK();
2199 if ((e = nd6_prefix_onlink(opr)) != 0) {
2200 nd6log((LOG_ERR,
2201 "%s: failed to recover a prefix "
2202 "%s/%d from %s to %s (errno=%d)\n",
2203 __func__, ip6_sprintf(ip6buf,
2204 &opr->ndpr_prefix.sin6_addr),
2205 opr->ndpr_plen, if_name(ifp),
2206 if_name(opr->ndpr_ifp), e));
2207 } else
2208 a_failure = 0;
2209 ND6_RLOCK();
2210 if (genid != V_nd6_list_genid)
2211 goto restart;
2212 }
2213 }
2214 ND6_RUNLOCK();
2215 } else {
2216 /* XXX: can we still set the NDPRF_ONLINK flag? */
2217 nd6log((LOG_ERR,
2218 "%s: failed to delete route: %s/%d on %s (errno=%d)\n",
2219 __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2220 pr->ndpr_plen, if_name(ifp), error));
2221 }
2222
2223 if (a_failure)
2224 lltable_prefix_free(AF_INET6,
2225 (struct sockaddr *)&pr->ndpr_prefix,
2226 (struct sockaddr *)&mask6, LLE_STATIC);
2227
2228 return (error);
2229 }
2230
2231 /*
2232 * Get a randomized interface identifier for a temporary address
2233 * Based on RFC 8981, Section 3.3.1.
2234 */
2235 static int
in6_get_tmp_ifid(struct in6_aliasreq * ifra)2236 in6_get_tmp_ifid(struct in6_aliasreq *ifra)
2237 {
2238 struct in6_addr *addr;
2239
2240 if(!is_random_seeded()){
2241 return 1;
2242 }
2243
2244 addr = &(ifra->ifra_addr.sin6_addr);
2245 regen:
2246 ifra->ifra_addr.sin6_addr.s6_addr32[2] |=
2247 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[2]));
2248 ifra->ifra_addr.sin6_addr.s6_addr32[3] |=
2249 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[3]));
2250
2251 /*
2252 * Check if generated address is not inappropriate:
2253 *
2254 * - Reserved IPv6 Interface aIdentifers
2255 * (https://www.iana.org/assignments/ipv6-interface-ids/)
2256 */
2257
2258 /* Subnet-router anycast: 0000:0000:0000:0000 */
2259 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2260 goto regen;
2261
2262 /*
2263 * IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2264 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2265 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2266 */
2267 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2268 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2269 goto regen;
2270
2271 /* Reserved subnet anycast addresses */
2272 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2273 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2274 goto regen;
2275
2276 return 0;
2277 }
2278
2279 /*
2280 * ia0 - corresponding public address
2281 */
2282 int
in6_tmpifadd(const struct in6_ifaddr * ia0,int forcegen,int delay)2283 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2284 {
2285 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2286 struct in6_ifaddr *newia;
2287 struct in6_aliasreq ifra;
2288 int error;
2289 int trylimit = 3; /* XXX: adhoc value */
2290 int updateflags;
2291 time_t vltime0, pltime0;
2292
2293 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2294 &ia0->ia_prefixmask.sin6_addr);
2295
2296 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */
2297 /* clear the old IFID */
2298 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2299 &ifra.ifra_prefixmask.sin6_addr);
2300
2301 again:
2302 if (in6_get_tmp_ifid(&ifra) != 0) {
2303 nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n",
2304 __func__));
2305 return (EINVAL);
2306 }
2307
2308 /*
2309 * in6_get_tmpifid() quite likely provided a unique interface ID.
2310 * However, we may still have a chance to see collision, because
2311 * there may be a time lag between generation of the ID and generation
2312 * of the address. So, we'll do one more sanity check.
2313 */
2314
2315 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2316 if (trylimit-- > 0) {
2317 forcegen = 1;
2318 goto again;
2319 }
2320
2321 /* Give up. Something strange should have happened. */
2322 nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n",
2323 __func__));
2324 return (EEXIST);
2325 }
2326
2327 /*
2328 * The Valid Lifetime is the lower of the Valid Lifetime of the
2329 * public address or TEMP_VALID_LIFETIME.
2330 * The Preferred Lifetime is the lower of the Preferred Lifetime
2331 * of the public address or TEMP_PREFERRED_LIFETIME -
2332 * DESYNC_FACTOR.
2333 */
2334 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2335 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2336 (ia0->ia6_lifetime.ia6t_vltime -
2337 (time_uptime - ia0->ia6_updatetime));
2338 if (vltime0 > V_ip6_temp_valid_lifetime)
2339 vltime0 = V_ip6_temp_valid_lifetime;
2340 } else
2341 vltime0 = V_ip6_temp_valid_lifetime;
2342 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2343 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2344 (ia0->ia6_lifetime.ia6t_pltime -
2345 (time_uptime - ia0->ia6_updatetime));
2346 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2347 pltime0 = V_ip6_temp_preferred_lifetime -
2348 V_ip6_desync_factor;
2349 }
2350 } else
2351 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2352 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2353 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2354
2355 /*
2356 * A temporary address is created only if this calculated Preferred
2357 * Lifetime is greater than REGEN_ADVANCE time units.
2358 */
2359 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2360 return (0);
2361
2362 /* XXX: scope zone ID? */
2363
2364 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2365
2366 /* allocate ifaddr structure, link into chain, etc. */
2367 updateflags = 0;
2368 if (delay)
2369 updateflags |= IN6_IFAUPDATE_DADDELAY;
2370 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2371 return (error);
2372
2373 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2374 if (newia == NULL) { /* XXX: can it happen? */
2375 nd6log((LOG_ERR,
2376 "%s: ifa update succeeded, but we got no ifaddr\n",
2377 __func__));
2378 return (EINVAL); /* XXX */
2379 }
2380 newia->ia6_ndpr = ia0->ia6_ndpr;
2381 newia->ia6_ndpr->ndpr_addrcnt++;
2382 ifa_free(&newia->ia_ifa);
2383
2384 /*
2385 * A newly added address might affect the status of other addresses.
2386 * XXX: when the temporary address is generated with a new public
2387 * address, the onlink check is redundant. However, it would be safe
2388 * to do the check explicitly everywhere a new address is generated,
2389 * and, in fact, we surely need the check when we create a new
2390 * temporary address due to deprecation of an old temporary address.
2391 */
2392 pfxlist_onlink_check();
2393
2394 return (0);
2395 }
2396
2397 static int
rt6_deleteroute(const struct rtentry * rt,const struct nhop_object * nh,void * arg)2398 rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh,
2399 void *arg)
2400 {
2401 struct in6_addr *gate = (struct in6_addr *)arg;
2402 int nh_rt_flags;
2403
2404 if (nh->gw_sa.sa_family != AF_INET6)
2405 return (0);
2406
2407 if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) {
2408 return (0);
2409 }
2410
2411 /*
2412 * Do not delete a static route.
2413 * XXX: this seems to be a bit ad-hoc. Should we consider the
2414 * 'cloned' bit instead?
2415 */
2416 nh_rt_flags = nhop_get_rtflags(nh);
2417 if ((nh_rt_flags & RTF_STATIC) != 0)
2418 return (0);
2419
2420 /*
2421 * We delete only host route. This means, in particular, we don't
2422 * delete default route.
2423 */
2424 if ((nh_rt_flags & RTF_HOST) == 0)
2425 return (0);
2426
2427 return (1);
2428 #undef SIN6
2429 }
2430
2431 /*
2432 * Delete all the routing table entries that use the specified gateway.
2433 * XXX: this function causes search through all entries of routing table, so
2434 * it shouldn't be called when acting as a router.
2435 */
2436 void
rt6_flush(struct in6_addr * gateway,struct ifnet * ifp)2437 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2438 {
2439
2440 /* We'll care only link-local addresses */
2441 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2442 return;
2443
2444 /* XXX Do we really need to walk any but the default FIB? */
2445 rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2446 }
2447
2448 int
nd6_setdefaultiface(int ifindex)2449 nd6_setdefaultiface(int ifindex)
2450 {
2451
2452 if (V_nd6_defifindex != ifindex) {
2453 V_nd6_defifindex = ifindex;
2454 if (V_nd6_defifindex != 0) {
2455 struct epoch_tracker et;
2456
2457 /*
2458 * XXXGL: this function should use ifnet_byindex_ref!
2459 */
2460 NET_EPOCH_ENTER(et);
2461 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2462 NET_EPOCH_EXIT(et);
2463 if (V_nd6_defifp == NULL)
2464 return (EINVAL);
2465 } else
2466 V_nd6_defifp = NULL;
2467
2468 /*
2469 * Our current implementation assumes one-to-one mapping between
2470 * interfaces and links, so it would be natural to use the
2471 * default interface as the default link.
2472 */
2473 scope6_setdefault(V_nd6_defifp);
2474 }
2475
2476 return (0);
2477 }
2478
2479 bool
nd6_defrouter_list_empty(void)2480 nd6_defrouter_list_empty(void)
2481 {
2482
2483 return (TAILQ_EMPTY(&V_nd6_defrouter));
2484 }
2485
2486 void
nd6_defrouter_timer(void)2487 nd6_defrouter_timer(void)
2488 {
2489 struct nd_defrouter *dr, *ndr;
2490 struct nd6_drhead drq;
2491
2492 TAILQ_INIT(&drq);
2493
2494 ND6_WLOCK();
2495 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr)
2496 if (dr->expire && dr->expire < time_uptime)
2497 defrouter_unlink(dr, &drq);
2498 ND6_WUNLOCK();
2499
2500 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2501 TAILQ_REMOVE(&drq, dr, dr_entry);
2502 defrouter_del(dr);
2503 }
2504 }
2505
2506 /*
2507 * Nuke default router list entries toward ifp.
2508 * We defer removal of default router list entries that is installed in the
2509 * routing table, in order to keep additional side effects as small as possible.
2510 */
2511 void
nd6_defrouter_purge(struct ifnet * ifp)2512 nd6_defrouter_purge(struct ifnet *ifp)
2513 {
2514 struct nd_defrouter *dr, *ndr;
2515 struct nd6_drhead drq;
2516
2517 TAILQ_INIT(&drq);
2518
2519 ND6_WLOCK();
2520 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2521 if (dr->installed)
2522 continue;
2523 if (dr->ifp == ifp)
2524 defrouter_unlink(dr, &drq);
2525 }
2526 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2527 if (!dr->installed)
2528 continue;
2529 if (dr->ifp == ifp)
2530 defrouter_unlink(dr, &drq);
2531 }
2532 ND6_WUNLOCK();
2533
2534 /* Delete the unlinked router objects. */
2535 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2536 TAILQ_REMOVE(&drq, dr, dr_entry);
2537 defrouter_del(dr);
2538 }
2539 }
2540
2541 void
nd6_defrouter_flush_all(void)2542 nd6_defrouter_flush_all(void)
2543 {
2544 struct nd_defrouter *dr;
2545 struct nd6_drhead drq;
2546
2547 TAILQ_INIT(&drq);
2548
2549 ND6_WLOCK();
2550 while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL)
2551 defrouter_unlink(dr, &drq);
2552 ND6_WUNLOCK();
2553
2554 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2555 TAILQ_REMOVE(&drq, dr, dr_entry);
2556 defrouter_del(dr);
2557 }
2558 }
2559
2560 void
nd6_defrouter_init(void)2561 nd6_defrouter_init(void)
2562 {
2563
2564 TAILQ_INIT(&V_nd6_defrouter);
2565 }
2566
2567 static int
nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)2568 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2569 {
2570 struct in6_defrouter d;
2571 struct nd_defrouter *dr;
2572 int error;
2573
2574 if (req->newptr != NULL)
2575 return (EPERM);
2576
2577 error = sysctl_wire_old_buffer(req, 0);
2578 if (error != 0)
2579 return (error);
2580
2581 bzero(&d, sizeof(d));
2582 d.rtaddr.sin6_family = AF_INET6;
2583 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2584
2585 ND6_RLOCK();
2586 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
2587 d.rtaddr.sin6_addr = dr->rtaddr;
2588 error = sa6_recoverscope(&d.rtaddr);
2589 if (error != 0)
2590 break;
2591 d.flags = dr->raflags;
2592 d.rtlifetime = dr->rtlifetime;
2593 d.expire = dr->expire + (time_second - time_uptime);
2594 d.if_index = dr->ifp->if_index;
2595 error = SYSCTL_OUT(req, &d, sizeof(d));
2596 if (error != 0)
2597 break;
2598 }
2599 ND6_RUNLOCK();
2600 return (error);
2601 }
2602 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2603 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2604 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2605 "NDP default router list");
2606