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 #include <machine/atomic.h>
78
79 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
80 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
81 struct mbuf *, int);
82 static int nd6_prefix_onlink(struct nd_prefix *);
83 static int in6_get_tmp_ifid(struct in6_aliasreq *);
84
85 TAILQ_HEAD(nd6_drhead, nd_defrouter);
86 VNET_DEFINE_STATIC(struct nd6_drhead, nd6_defrouter);
87 #define V_nd6_defrouter VNET(nd6_defrouter)
88
89 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
90 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
91
92 VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp);
93 VNET_DEFINE(int, nd6_defifindex);
94 #define V_nd6_defifp VNET(nd6_defifp)
95
96 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
97 VNET_DEFINE(bool, ip6_use_stableaddr) = 0;
98
99 VNET_DEFINE(int, ip6_desync_factor);
100 VNET_DEFINE(uint32_t, ip6_temp_max_desync_factor) = TEMP_MAX_DESYNC_FACTOR_BASE;
101 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
102 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
103
104 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
105
106 #ifdef EXPERIMENTAL
107 VNET_DEFINE_STATIC(int, nd6_ignore_ipv6_only_ra) = 1;
108 #define V_nd6_ignore_ipv6_only_ra VNET(nd6_ignore_ipv6_only_ra)
109 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO,
110 nd6_ignore_ipv6_only_ra, CTLFLAG_VNET | CTLFLAG_RW,
111 &VNET_NAME(nd6_ignore_ipv6_only_ra), 0,
112 "Ignore the 'IPv6-Only flag' in RA messages in compliance with "
113 "draft-ietf-6man-ipv6only-flag");
114 #endif
115
116 /* RTPREF_MEDIUM has to be 0! */
117 #define RTPREF_HIGH 1
118 #define RTPREF_MEDIUM 0
119 #define RTPREF_LOW (-1)
120 #define RTPREF_RESERVED (-2)
121 #define RTPREF_INVALID (-3) /* internal */
122
123 static void
defrouter_ref(struct nd_defrouter * dr)124 defrouter_ref(struct nd_defrouter *dr)
125 {
126
127 refcount_acquire(&dr->refcnt);
128 }
129
130 void
defrouter_rele(struct nd_defrouter * dr)131 defrouter_rele(struct nd_defrouter *dr)
132 {
133
134 if (refcount_release(&dr->refcnt))
135 free(dr, M_IP6NDP);
136 }
137
138 /*
139 * Remove a router from the global list and optionally stash it in a
140 * caller-supplied queue.
141 */
142 static void
defrouter_unlink(struct nd_defrouter * dr,struct nd6_drhead * drq)143 defrouter_unlink(struct nd_defrouter *dr, struct nd6_drhead *drq)
144 {
145
146 ND6_WLOCK_ASSERT();
147
148 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
149 V_nd6_list_genid++;
150 if (drq != NULL)
151 TAILQ_INSERT_TAIL(drq, dr, dr_entry);
152 }
153
154 /*
155 * Receive Router Solicitation Message - just for routers.
156 * Router solicitation/advertisement is mostly managed by userland program
157 * (rtadvd) so here we have no function like nd6_ra_output().
158 *
159 * Based on RFC 2461
160 */
161 void
nd6_rs_input(struct mbuf * m,int off,int icmp6len)162 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
163 {
164 struct ifnet *ifp;
165 struct ip6_hdr *ip6;
166 struct nd_router_solicit *nd_rs;
167 struct in6_addr saddr6;
168 union nd_opts ndopts;
169 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
170 char *lladdr;
171 int lladdrlen;
172
173 ifp = m->m_pkthdr.rcvif;
174
175 /*
176 * Accept RS only when V_ip6_forwarding=1 and the interface has
177 * no ND6_IFF_ACCEPT_RTADV.
178 */
179 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
180 goto freeit;
181
182 /* RFC 6980: Nodes MUST silently ignore fragments */
183 if(m->m_flags & M_FRAGMENTED)
184 goto freeit;
185
186 /* Sanity checks */
187 ip6 = mtod(m, struct ip6_hdr *);
188 if (__predict_false(ip6->ip6_hlim != 255)) {
189 ICMP6STAT_INC(icp6s_invlhlim);
190 nd6log((LOG_ERR,
191 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
192 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
193 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
194 goto bad;
195 }
196
197 /*
198 * Don't update the neighbor cache, if src = ::.
199 * This indicates that the src has no IP address assigned yet.
200 */
201 saddr6 = ip6->ip6_src;
202 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
203 goto freeit;
204
205 if (m->m_len < off + icmp6len) {
206 m = m_pullup(m, off + icmp6len);
207 if (m == NULL) {
208 IP6STAT_INC(ip6s_exthdrtoolong);
209 return;
210 }
211 }
212 ip6 = mtod(m, struct ip6_hdr *);
213 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
214
215 icmp6len -= sizeof(*nd_rs);
216 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
217 if (nd6_options(&ndopts) < 0) {
218 nd6log((LOG_INFO,
219 "%s: invalid ND option, ignored\n", __func__));
220 /* nd6_options have incremented stats */
221 goto freeit;
222 }
223
224 lladdr = NULL;
225 lladdrlen = 0;
226 if (ndopts.nd_opts_src_lladdr) {
227 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
228 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
229 }
230
231 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
232 nd6log((LOG_INFO,
233 "%s: lladdrlen mismatch for %s (if %d, RS packet %d)\n",
234 __func__, ip6_sprintf(ip6bufs, &saddr6),
235 ifp->if_addrlen, lladdrlen - 2));
236 goto bad;
237 }
238
239 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
240
241 freeit:
242 m_freem(m);
243 return;
244
245 bad:
246 ICMP6STAT_INC(icp6s_badrs);
247 m_freem(m);
248 }
249
250 #ifdef EXPERIMENTAL
251 /*
252 * An initial update routine for draft-ietf-6man-ipv6only-flag.
253 * We need to iterate over all default routers for the given
254 * interface to see whether they are all advertising the "S"
255 * (IPv6-Only) flag. If they do set, otherwise unset, the
256 * interface flag we later use to filter on.
257 */
258 static void
defrtr_ipv6_only_ifp(struct ifnet * ifp)259 defrtr_ipv6_only_ifp(struct ifnet *ifp)
260 {
261 struct nd_defrouter *dr;
262 bool ipv6_only, ipv6_only_old;
263 #ifdef INET
264 struct epoch_tracker et;
265 struct ifaddr *ifa;
266 bool has_ipv4_addr;
267 #endif
268
269 if (V_nd6_ignore_ipv6_only_ra != 0)
270 return;
271
272 ipv6_only = true;
273 ND6_RLOCK();
274 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
275 if (dr->ifp == ifp &&
276 (dr->raflags & ND_RA_FLAG_IPV6_ONLY) == 0)
277 ipv6_only = false;
278 ND6_RUNLOCK();
279
280 IF_AFDATA_WLOCK(ifp);
281 ipv6_only_old = ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY;
282 IF_AFDATA_WUNLOCK(ifp);
283
284 /* If nothing changed, we have an early exit. */
285 if (ipv6_only == ipv6_only_old)
286 return;
287
288 #ifdef INET
289 /*
290 * Should we want to set the IPV6-ONLY flag, check if the
291 * interface has a non-0/0 and non-link-local IPv4 address
292 * configured on it. If it has we will assume working
293 * IPv4 operations and will clear the interface flag.
294 */
295 has_ipv4_addr = false;
296 if (ipv6_only) {
297 NET_EPOCH_ENTER(et);
298 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
299 if (ifa->ifa_addr->sa_family != AF_INET)
300 continue;
301 if (in_canforward(
302 satosin(ifa->ifa_addr)->sin_addr)) {
303 has_ipv4_addr = true;
304 break;
305 }
306 }
307 NET_EPOCH_EXIT(et);
308 }
309 if (ipv6_only && has_ipv4_addr) {
310 log(LOG_NOTICE, "%s rcvd RA w/ IPv6-Only flag set but has IPv4 "
311 "configured, ignoring IPv6-Only flag.\n", ifp->if_xname);
312 ipv6_only = false;
313 }
314 #endif
315
316 IF_AFDATA_WLOCK(ifp);
317 if (ipv6_only)
318 ND_IFINFO(ifp)->flags |= ND6_IFF_IPV6_ONLY;
319 else
320 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
321 IF_AFDATA_WUNLOCK(ifp);
322
323 #ifdef notyet
324 /* Send notification of flag change. */
325 #endif
326 }
327
328 static void
defrtr_ipv6_only_ipf_down(struct ifnet * ifp)329 defrtr_ipv6_only_ipf_down(struct ifnet *ifp)
330 {
331
332 IF_AFDATA_WLOCK(ifp);
333 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
334 IF_AFDATA_WUNLOCK(ifp);
335 }
336 #endif /* EXPERIMENTAL */
337
338 void
nd6_ifnet_link_event(void * arg __unused,struct ifnet * ifp,int linkstate)339 nd6_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate)
340 {
341
342 /*
343 * XXX-BZ we might want to trigger re-evaluation of our default router
344 * availability. E.g., on link down the default router might be
345 * unreachable but a different interface might still have connectivity.
346 */
347
348 #ifdef EXPERIMENTAL
349 if (linkstate == LINK_STATE_DOWN)
350 defrtr_ipv6_only_ipf_down(ifp);
351 #endif
352 }
353
354 /*
355 * Receive Router Advertisement Message.
356 *
357 * Based on RFC 2461
358 * TODO: on-link bit on prefix information
359 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
360 */
361 void
nd6_ra_input(struct mbuf * m,int off,int icmp6len)362 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
363 {
364 struct ifnet *ifp;
365 struct nd_ifinfo *ndi;
366 struct ip6_hdr *ip6;
367 struct nd_router_advert *nd_ra;
368 struct in6_addr saddr6;
369 struct nd_defrouter *dr;
370 union nd_opts ndopts;
371 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
372 int mcast;
373
374 /*
375 * We only accept RAs only when the per-interface flag
376 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
377 */
378 ifp = m->m_pkthdr.rcvif;
379 ndi = ND_IFINFO(ifp);
380 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
381 goto freeit;
382
383 /* RFC 6980: Nodes MUST silently ignore fragments */
384 if(m->m_flags & M_FRAGMENTED)
385 goto freeit;
386
387 ip6 = mtod(m, struct ip6_hdr *);
388 if (__predict_false(ip6->ip6_hlim != 255)) {
389 ICMP6STAT_INC(icp6s_invlhlim);
390 nd6log((LOG_ERR,
391 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
392 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
393 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
394 goto bad;
395 }
396
397 saddr6 = ip6->ip6_src;
398 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
399 nd6log((LOG_ERR,
400 "%s: src %s is not link-local\n", __func__,
401 ip6_sprintf(ip6bufs, &saddr6)));
402 goto bad;
403 }
404
405 if (m->m_len < off + icmp6len) {
406 m = m_pullup(m, off + icmp6len);
407 if (m == NULL) {
408 IP6STAT_INC(ip6s_exthdrtoolong);
409 return;
410 }
411 }
412 ip6 = mtod(m, struct ip6_hdr *);
413 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
414
415 icmp6len -= sizeof(*nd_ra);
416 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
417 if (nd6_options(&ndopts) < 0) {
418 nd6log((LOG_INFO,
419 "%s: invalid ND option, ignored\n", __func__));
420 /* nd6_options have incremented stats */
421 goto freeit;
422 }
423
424 mcast = 0;
425 dr = NULL;
426 {
427 struct nd_defrouter dr0;
428 u_int32_t advreachable = nd_ra->nd_ra_reachable;
429
430 /* remember if this is a multicasted advertisement */
431 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
432 mcast = 1;
433
434 bzero(&dr0, sizeof(dr0));
435 dr0.rtaddr = saddr6;
436 dr0.raflags = nd_ra->nd_ra_flags_reserved;
437 /*
438 * Effectively-disable routes from RA messages when
439 * ND6_IFF_NO_RADR enabled on the receiving interface or
440 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
441 */
442 if (ndi->flags & ND6_IFF_NO_RADR)
443 dr0.rtlifetime = 0;
444 else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
445 dr0.rtlifetime = 0;
446 else
447 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
448 dr0.expire = time_uptime + dr0.rtlifetime;
449 dr0.ifp = ifp;
450 /* unspecified or not? (RFC 2461 6.3.4) */
451 if (advreachable) {
452 advreachable = ntohl(advreachable);
453 if (advreachable <= MAX_REACHABLE_TIME &&
454 ndi->basereachable != advreachable) {
455 ndi->basereachable = advreachable;
456 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
457 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
458 }
459 }
460 if (nd_ra->nd_ra_retransmit)
461 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
462 if (nd_ra->nd_ra_curhoplimit) {
463 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
464 ndi->chlim = nd_ra->nd_ra_curhoplimit;
465 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
466 log(LOG_ERR, "RA with a lower CurHopLimit sent from "
467 "%s on %s (current = %d, received = %d). "
468 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
469 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
470 }
471 }
472 dr = defrtrlist_update(&dr0);
473 #ifdef EXPERIMENTAL
474 defrtr_ipv6_only_ifp(ifp);
475 #endif
476 }
477
478 /*
479 * prefix
480 */
481 if (ndopts.nd_opts_pi) {
482 struct nd_opt_hdr *pt;
483 struct nd_opt_prefix_info *pi = NULL;
484 struct nd_prefixctl pr;
485
486 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
487 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
488 pt = (struct nd_opt_hdr *)((caddr_t)pt +
489 (pt->nd_opt_len << 3))) {
490 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
491 continue;
492 pi = (struct nd_opt_prefix_info *)pt;
493
494 if (pi->nd_opt_pi_len != 4) {
495 nd6log((LOG_INFO,
496 "%s: invalid option len %d for prefix "
497 "information option, ignored\n", __func__,
498 pi->nd_opt_pi_len));
499 continue;
500 }
501
502 if (128 < pi->nd_opt_pi_prefix_len) {
503 nd6log((LOG_INFO,
504 "%s: invalid prefix len %d for prefix "
505 "information option, ignored\n", __func__,
506 pi->nd_opt_pi_prefix_len));
507 continue;
508 }
509
510 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
511 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
512 nd6log((LOG_INFO,
513 "%s: invalid prefix %s, ignored\n",
514 __func__, ip6_sprintf(ip6bufs,
515 &pi->nd_opt_pi_prefix)));
516 continue;
517 }
518
519 bzero(&pr, sizeof(pr));
520 pr.ndpr_prefix.sin6_family = AF_INET6;
521 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
522 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
523 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
524
525 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
526 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
527 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
528 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
529 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
530 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
531 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
532 (void)prelist_update(&pr, dr, m, mcast);
533 }
534 }
535 if (dr != NULL) {
536 defrouter_rele(dr);
537 dr = NULL;
538 }
539
540 /*
541 * MTU
542 */
543 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
544 u_long mtu;
545 u_long maxmtu;
546
547 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
548
549 /* lower bound */
550 if (mtu < IPV6_MMTU) {
551 nd6log((LOG_INFO, "%s: bogus mtu option mtu=%lu sent "
552 "from %s, ignoring\n", __func__,
553 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
554 goto skip;
555 }
556
557 /* upper bound */
558 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
559 ? ndi->maxmtu : ifp->if_mtu;
560 if (mtu <= maxmtu) {
561 if (ndi->linkmtu != mtu) {
562 ndi->linkmtu = mtu;
563 rt_updatemtu(ifp);
564 }
565 } else {
566 nd6log((LOG_INFO, "%s: bogus mtu=%lu sent from %s; "
567 "exceeds maxmtu %lu, ignoring\n", __func__,
568 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
569 }
570 }
571
572 skip:
573
574 /*
575 * Source link layer address
576 */
577 {
578 char *lladdr = NULL;
579 int lladdrlen = 0;
580
581 if (ndopts.nd_opts_src_lladdr) {
582 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
583 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
584 }
585
586 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
587 nd6log((LOG_INFO,
588 "%s: lladdrlen mismatch for %s (if %d, RA packet %d)\n",
589 __func__, ip6_sprintf(ip6bufs, &saddr6),
590 ifp->if_addrlen, lladdrlen - 2));
591 goto bad;
592 }
593
594 nd6_cache_lladdr(ifp, &saddr6, lladdr,
595 lladdrlen, ND_ROUTER_ADVERT, 0);
596
597 /*
598 * Installing a link-layer address might change the state of the
599 * router's neighbor cache, which might also affect our on-link
600 * detection of adveritsed prefixes.
601 */
602 pfxlist_onlink_check();
603 }
604
605 freeit:
606 m_freem(m);
607 return;
608
609 bad:
610 ICMP6STAT_INC(icp6s_badra);
611 m_freem(m);
612 }
613
614 /* PFXRTR */
615 static struct nd_pfxrouter *
pfxrtr_lookup(struct nd_prefix * pr,struct nd_defrouter * dr)616 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
617 {
618 struct nd_pfxrouter *search;
619
620 ND6_LOCK_ASSERT();
621
622 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
623 if (search->router == dr)
624 break;
625 }
626 return (search);
627 }
628
629 static void
pfxrtr_add(struct nd_prefix * pr,struct nd_defrouter * dr)630 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
631 {
632 struct nd_pfxrouter *new;
633 bool update;
634
635 ND6_UNLOCK_ASSERT();
636
637 ND6_RLOCK();
638 if (pfxrtr_lookup(pr, dr) != NULL) {
639 ND6_RUNLOCK();
640 return;
641 }
642 ND6_RUNLOCK();
643
644 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
645 if (new == NULL)
646 return;
647 defrouter_ref(dr);
648 new->router = dr;
649
650 ND6_WLOCK();
651 if (pfxrtr_lookup(pr, dr) == NULL) {
652 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
653 update = true;
654 } else {
655 /* We lost a race to add the reference. */
656 defrouter_rele(dr);
657 free(new, M_IP6NDP);
658 update = false;
659 }
660 ND6_WUNLOCK();
661
662 if (update)
663 pfxlist_onlink_check();
664 }
665
666 static void
pfxrtr_del(struct nd_pfxrouter * pfr)667 pfxrtr_del(struct nd_pfxrouter *pfr)
668 {
669
670 ND6_WLOCK_ASSERT();
671
672 LIST_REMOVE(pfr, pfr_entry);
673 defrouter_rele(pfr->router);
674 free(pfr, M_IP6NDP);
675 }
676
677 /* Default router list processing sub routines. */
678 static void
defrouter_addreq(struct nd_defrouter * new)679 defrouter_addreq(struct nd_defrouter *new)
680 {
681 uint32_t fibnum = new->ifp->if_fib;
682 struct rib_cmd_info rc = {};
683 int error = 0;
684
685 NET_EPOCH_ASSERT();
686
687 struct sockaddr_in6 gw = {
688 .sin6_family = AF_INET6,
689 .sin6_len = sizeof(struct sockaddr_in6),
690 .sin6_addr = new->rtaddr,
691 };
692
693 error = rib_add_default_route(fibnum, AF_INET6, new->ifp,
694 (struct sockaddr *)&gw, &rc);
695
696 if (error == 0) {
697 struct nhop_object *nh = nhop_select_func(rc.rc_nh_new, 0);
698 rt_routemsg(RTM_ADD, rc.rc_rt, nh, fibnum);
699 new->installed = 1;
700 }
701 }
702
703 /*
704 * Remove the default route for a given router.
705 * This is just a subroutine function for defrouter_select_fib(), and
706 * should not be called from anywhere else.
707 */
708 static void
defrouter_delreq(struct nd_defrouter * dr)709 defrouter_delreq(struct nd_defrouter *dr)
710 {
711 uint32_t fibnum = dr->ifp->if_fib;
712 struct epoch_tracker et;
713 struct rib_cmd_info rc;
714 int error;
715
716 struct sockaddr_in6 dst = {
717 .sin6_family = AF_INET6,
718 .sin6_len = sizeof(struct sockaddr_in6),
719 };
720
721 struct sockaddr_in6 gw = {
722 .sin6_family = AF_INET6,
723 .sin6_len = sizeof(struct sockaddr_in6),
724 .sin6_addr = dr->rtaddr,
725 };
726
727 NET_EPOCH_ENTER(et);
728 error = rib_del_route_px(fibnum, (struct sockaddr *)&dst, 0,
729 rib_match_gw, (struct sockaddr *)&gw, 0, &rc);
730 if (error == 0) {
731 struct nhop_object *nh = nhop_select_func(rc.rc_nh_old, 0);
732 rt_routemsg(RTM_DELETE, rc.rc_rt, nh, fibnum);
733 }
734 NET_EPOCH_EXIT(et);
735
736 dr->installed = 0;
737 }
738
739 static void
defrouter_del(struct nd_defrouter * dr)740 defrouter_del(struct nd_defrouter *dr)
741 {
742 struct nd_defrouter *deldr = NULL;
743 struct nd_prefix *pr;
744 struct nd_pfxrouter *pfxrtr;
745
746 ND6_UNLOCK_ASSERT();
747
748 /*
749 * Flush all the routing table entries that use the router
750 * as a next hop.
751 */
752 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
753 rt6_flush(&dr->rtaddr, dr->ifp);
754
755 #ifdef EXPERIMENTAL
756 defrtr_ipv6_only_ifp(dr->ifp);
757 #endif
758
759 if (dr->installed) {
760 deldr = dr;
761 defrouter_delreq(dr);
762 }
763
764 /*
765 * Also delete all the pointers to the router in each prefix lists.
766 */
767 ND6_WLOCK();
768 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
769 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
770 pfxrtr_del(pfxrtr);
771 }
772 ND6_WUNLOCK();
773
774 pfxlist_onlink_check();
775
776 /*
777 * If the router is the primary one, choose a new one.
778 * Note that defrouter_select_fib() will remove the current
779 * gateway from the routing table.
780 */
781 if (deldr)
782 defrouter_select_fib(deldr->ifp->if_fib);
783
784 /*
785 * Release the list reference.
786 */
787 defrouter_rele(dr);
788 }
789
790 struct nd_defrouter *
defrouter_lookup_locked(const struct in6_addr * addr,struct ifnet * ifp)791 defrouter_lookup_locked(const struct in6_addr *addr, struct ifnet *ifp)
792 {
793 struct nd_defrouter *dr;
794
795 ND6_LOCK_ASSERT();
796 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
797 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
798 defrouter_ref(dr);
799 return (dr);
800 }
801 return (NULL);
802 }
803
804 struct nd_defrouter *
defrouter_lookup(const struct in6_addr * addr,struct ifnet * ifp)805 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
806 {
807 struct nd_defrouter *dr;
808
809 ND6_RLOCK();
810 dr = defrouter_lookup_locked(addr, ifp);
811 ND6_RUNLOCK();
812 return (dr);
813 }
814
815 /*
816 * Remove all default routes from default router list.
817 */
818 void
defrouter_reset(void)819 defrouter_reset(void)
820 {
821 struct nd_defrouter *dr, **dra;
822 int count, i;
823
824 count = i = 0;
825
826 /*
827 * We can't delete routes with the ND lock held, so make a copy of the
828 * current default router list and use that when deleting routes.
829 */
830 ND6_RLOCK();
831 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
832 count++;
833 ND6_RUNLOCK();
834
835 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
836
837 ND6_RLOCK();
838 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
839 if (i == count)
840 break;
841 defrouter_ref(dr);
842 dra[i++] = dr;
843 }
844 ND6_RUNLOCK();
845
846 for (i = 0; i < count && dra[i] != NULL; i++) {
847 defrouter_delreq(dra[i]);
848 defrouter_rele(dra[i]);
849 }
850 free(dra, M_TEMP);
851
852 /*
853 * XXX should we also nuke any default routers in the kernel, by
854 * going through them by rtalloc1()?
855 */
856 }
857
858 /*
859 * Look up a matching default router list entry and remove it. Returns true if a
860 * matching entry was found, false otherwise.
861 */
862 bool
defrouter_remove(struct in6_addr * addr,struct ifnet * ifp)863 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
864 {
865 struct nd_defrouter *dr;
866
867 ND6_WLOCK();
868 dr = defrouter_lookup_locked(addr, ifp);
869 if (dr == NULL) {
870 ND6_WUNLOCK();
871 return (false);
872 }
873
874 defrouter_unlink(dr, NULL);
875 ND6_WUNLOCK();
876 defrouter_del(dr);
877 defrouter_rele(dr);
878 return (true);
879 }
880
881 /*
882 * for default router selection
883 * regards router-preference field as a 2-bit signed integer
884 */
885 static int
rtpref(struct nd_defrouter * dr)886 rtpref(struct nd_defrouter *dr)
887 {
888 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
889 case ND_RA_FLAG_RTPREF_HIGH:
890 return (RTPREF_HIGH);
891 case ND_RA_FLAG_RTPREF_MEDIUM:
892 case ND_RA_FLAG_RTPREF_RSV:
893 return (RTPREF_MEDIUM);
894 case ND_RA_FLAG_RTPREF_LOW:
895 return (RTPREF_LOW);
896 default:
897 /*
898 * This case should never happen. If it did, it would mean a
899 * serious bug of kernel internal. We thus always bark here.
900 * Or, can we even panic?
901 */
902 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
903 return (RTPREF_INVALID);
904 }
905 /* NOTREACHED */
906 }
907
908 static bool
is_dr_reachable(const struct nd_defrouter * dr)909 is_dr_reachable(const struct nd_defrouter *dr) {
910 struct llentry *ln = NULL;
911
912 ln = nd6_lookup(&dr->rtaddr, LLE_SF(AF_INET6, 0), dr->ifp);
913 if (ln == NULL)
914 return (false);
915 bool reachable = ND6_IS_LLINFO_PROBREACH(ln);
916 LLE_RUNLOCK(ln);
917 return reachable;
918 }
919
920 /*
921 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
922 * draft-ietf-ipngwg-router-selection:
923 * 1) Routers that are reachable or probably reachable should be preferred.
924 * If we have more than one (probably) reachable router, prefer ones
925 * with the highest router preference.
926 * 2) When no routers on the list are known to be reachable or
927 * probably reachable, routers SHOULD be selected in a round-robin
928 * fashion, regardless of router preference values.
929 * 3) If the Default Router List is empty, assume that all
930 * destinations are on-link.
931 *
932 * We assume nd_defrouter is sorted by router preference value.
933 * Since the code below covers both with and without router preference cases,
934 * we do not need to classify the cases by ifdef.
935 *
936 * At this moment, we do not try to install more than one default router,
937 * even when the multipath routing is available, because we're not sure about
938 * the benefits for stub hosts comparing to the risk of making the code
939 * complicated and the possibility of introducing bugs.
940 *
941 * We maintain a single list of routers for multiple FIBs, only considering one
942 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
943 * we do the whole thing multiple times.
944 */
945 void
defrouter_select_fib(int fibnum)946 defrouter_select_fib(int fibnum)
947 {
948 struct epoch_tracker et;
949 struct nd_defrouter *dr, *selected_dr, *installed_dr;
950
951 if (fibnum == RT_ALL_FIBS) {
952 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
953 defrouter_select_fib(fibnum);
954 }
955 return;
956 }
957
958 ND6_RLOCK();
959 /*
960 * Let's handle easy case (3) first:
961 * If default router list is empty, there's nothing to be done.
962 */
963 if (TAILQ_EMPTY(&V_nd6_defrouter)) {
964 ND6_RUNLOCK();
965 return;
966 }
967
968 /*
969 * Search for a (probably) reachable router from the list.
970 * We just pick up the first reachable one (if any), assuming that
971 * the ordering rule of the list described in defrtrlist_update().
972 */
973 selected_dr = installed_dr = NULL;
974 NET_EPOCH_ENTER(et);
975 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
976 if (dr->ifp->if_fib != fibnum)
977 continue;
978
979 if (selected_dr == NULL && is_dr_reachable(dr)) {
980 selected_dr = dr;
981 defrouter_ref(selected_dr);
982 }
983
984 if (dr->installed) {
985 if (installed_dr == NULL) {
986 installed_dr = dr;
987 defrouter_ref(installed_dr);
988 } else {
989 /*
990 * this should not happen.
991 * warn for diagnosis.
992 */
993 log(LOG_ERR, "defrouter_select_fib: more than "
994 "one router is installed\n");
995 }
996 }
997 }
998
999 /*
1000 * If none of the default routers was found to be reachable,
1001 * round-robin the list regardless of preference.
1002 * Otherwise, if we have an installed router, check if the selected
1003 * (reachable) router should really be preferred to the installed one.
1004 * We only prefer the new router when the old one is not reachable
1005 * or when the new one has a really higher preference value.
1006 */
1007 if (selected_dr == NULL) {
1008 if (installed_dr == NULL ||
1009 TAILQ_NEXT(installed_dr, dr_entry) == NULL)
1010 dr = TAILQ_FIRST(&V_nd6_defrouter);
1011 else
1012 dr = TAILQ_NEXT(installed_dr, dr_entry);
1013
1014 /* Ensure we select a router for this FIB. */
1015 TAILQ_FOREACH_FROM(dr, &V_nd6_defrouter, dr_entry) {
1016 if (dr->ifp->if_fib == fibnum) {
1017 selected_dr = dr;
1018 defrouter_ref(selected_dr);
1019 break;
1020 }
1021 }
1022 } else if (installed_dr != NULL) {
1023 if (is_dr_reachable(installed_dr) &&
1024 rtpref(selected_dr) <= rtpref(installed_dr)) {
1025 defrouter_rele(selected_dr);
1026 selected_dr = installed_dr;
1027 }
1028 }
1029 ND6_RUNLOCK();
1030
1031 /*
1032 * If we selected a router for this FIB and it's different
1033 * than the installed one, remove the installed router and
1034 * install the selected one in its place.
1035 */
1036 if (installed_dr != selected_dr) {
1037 if (installed_dr != NULL) {
1038 defrouter_delreq(installed_dr);
1039 defrouter_rele(installed_dr);
1040 }
1041 if (selected_dr != NULL)
1042 defrouter_addreq(selected_dr);
1043 }
1044 if (selected_dr != NULL)
1045 defrouter_rele(selected_dr);
1046 NET_EPOCH_EXIT(et);
1047 }
1048
1049 static struct nd_defrouter *
defrtrlist_update(struct nd_defrouter * new)1050 defrtrlist_update(struct nd_defrouter *new)
1051 {
1052 struct nd_defrouter *dr, *n;
1053 uint64_t genid;
1054 int oldpref;
1055 bool writelocked;
1056
1057 if (new->rtlifetime == 0) {
1058 defrouter_remove(&new->rtaddr, new->ifp);
1059 return (NULL);
1060 }
1061
1062 ND6_RLOCK();
1063 writelocked = false;
1064 restart:
1065 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
1066 if (dr != NULL) {
1067 oldpref = rtpref(dr);
1068
1069 /* override */
1070 dr->raflags = new->raflags; /* XXX flag check */
1071 dr->rtlifetime = new->rtlifetime;
1072 dr->expire = new->expire;
1073
1074 /*
1075 * If the preference does not change, there's no need
1076 * to sort the entries. Also make sure the selected
1077 * router is still installed in the kernel.
1078 */
1079 if (dr->installed && rtpref(new) == oldpref) {
1080 if (writelocked)
1081 ND6_WUNLOCK();
1082 else
1083 ND6_RUNLOCK();
1084 return (dr);
1085 }
1086 }
1087
1088 /*
1089 * The router needs to be reinserted into the default router
1090 * list, so upgrade to a write lock. If that fails and the list
1091 * has potentially changed while the lock was dropped, we'll
1092 * redo the lookup with the write lock held.
1093 */
1094 if (!writelocked) {
1095 writelocked = true;
1096 if (!ND6_TRY_UPGRADE()) {
1097 genid = V_nd6_list_genid;
1098 ND6_RUNLOCK();
1099 ND6_WLOCK();
1100 if (genid != V_nd6_list_genid)
1101 goto restart;
1102 }
1103 }
1104
1105 if (dr != NULL) {
1106 /*
1107 * The preferred router may have changed, so relocate this
1108 * router.
1109 */
1110 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
1111 n = dr;
1112 } else {
1113 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
1114 if (n == NULL) {
1115 ND6_WUNLOCK();
1116 return (NULL);
1117 }
1118 memcpy(n, new, sizeof(*n));
1119 /* Initialize with an extra reference for the caller. */
1120 refcount_init(&n->refcnt, 2);
1121 }
1122
1123 /*
1124 * Insert the new router in the Default Router List;
1125 * The Default Router List should be in the descending order
1126 * of router-preferece. Routers with the same preference are
1127 * sorted in the arriving time order.
1128 */
1129
1130 /* insert at the end of the group */
1131 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1132 if (rtpref(n) > rtpref(dr))
1133 break;
1134 }
1135 if (dr != NULL)
1136 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
1137 else
1138 TAILQ_INSERT_TAIL(&V_nd6_defrouter, n, dr_entry);
1139 V_nd6_list_genid++;
1140 ND6_WUNLOCK();
1141
1142 defrouter_select_fib(new->ifp->if_fib);
1143
1144 return (n);
1145 }
1146
1147 static int
in6_init_prefix_ltimes(struct nd_prefix * ndpr)1148 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1149 {
1150 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1151 ndpr->ndpr_preferred = 0;
1152 else
1153 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
1154 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1155 ndpr->ndpr_expire = 0;
1156 else
1157 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
1158
1159 return 0;
1160 }
1161
1162 static void
in6_init_address_ltimes(struct nd_prefix * new,struct in6_addrlifetime * lt6)1163 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1164 {
1165 /* init ia6t_expire */
1166 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1167 lt6->ia6t_expire = 0;
1168 else {
1169 lt6->ia6t_expire = time_uptime;
1170 lt6->ia6t_expire += lt6->ia6t_vltime;
1171 }
1172
1173 /* init ia6t_preferred */
1174 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1175 lt6->ia6t_preferred = 0;
1176 else {
1177 lt6->ia6t_preferred = time_uptime;
1178 lt6->ia6t_preferred += lt6->ia6t_pltime;
1179 }
1180 }
1181
1182 static struct in6_ifaddr *
in6_ifadd(struct nd_prefixctl * pr,int mcast)1183 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1184 {
1185 struct ifnet *ifp = pr->ndpr_ifp;
1186 struct ifaddr *ifa;
1187 struct in6_aliasreq ifra;
1188 struct in6_ifaddr *ia = NULL, *ib = NULL;
1189 int error, plen0;
1190 struct in6_addr *ifid_addr = NULL, mask, newaddr;
1191 int prefixlen = pr->ndpr_plen;
1192 int updateflags;
1193 char ip6buf[INET6_ADDRSTRLEN];
1194
1195 in6_prefixlen2mask(&mask, prefixlen);
1196
1197 /*
1198 * find a link-local address (will be interface ID).
1199 * Is it really mandatory? Theoretically, a global or a site-local
1200 * address can be configured without a link-local address, if we
1201 * have a unique interface identifier...
1202 *
1203 * it is not mandatory to have a link-local address, we can generate
1204 * interface identifier on the fly. we do this because:
1205 * (1) it should be the easiest way to find interface identifier.
1206 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1207 * for multiple addresses on a single interface, and possible shortcut
1208 * of DAD. we omitted DAD for this reason in the past.
1209 * (3) a user can prevent autoconfiguration of global address
1210 * by removing link-local address by hand (this is partly because we
1211 * don't have other way to control the use of IPv6 on an interface.
1212 * this has been our design choice - cf. NRL's "ifconfig auto").
1213 * (4) it is easier to manage when an interface has addresses
1214 * with the same interface identifier, than to have multiple addresses
1215 * with different interface identifiers.
1216 *
1217 * If using stable privacy generation, generate a new address with
1218 * the algorithm specified in RFC 7217 section 5
1219 */
1220
1221 /* make ifaddr */
1222 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1223
1224 if (ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR) {
1225 memcpy(&newaddr, &pr->ndpr_prefix.sin6_addr, sizeof(pr->ndpr_prefix.sin6_addr));
1226
1227 if(!in6_get_stableifid(ifp, &newaddr, prefixlen))
1228 return NULL;
1229 } else {
1230 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1231 if (ifa) {
1232 ib = (struct in6_ifaddr *)ifa;
1233 ifid_addr = &ib->ia_addr.sin6_addr;
1234
1235 /* prefixlen + ifidlen must be equal to 128 */
1236 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1237 if (prefixlen != plen0) {
1238 ifa_free(ifa);
1239 ifid_addr = NULL;
1240 nd6log((LOG_DEBUG,
1241 "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n",
1242 __func__, if_name(ifp), prefixlen, 128 - plen0));
1243 }
1244 }
1245
1246 /* No suitable LL address, get the ifid directly */
1247 if (ifid_addr == NULL) {
1248 ifa = ifa_alloc(sizeof(struct in6_ifaddr), M_NOWAIT);
1249 if (ifa != NULL) {
1250 ib = (struct in6_ifaddr *)ifa;
1251 ifid_addr = &ib->ia_addr.sin6_addr;
1252 if(in6_get_ifid(ifp, NULL, ifid_addr) != 0) {
1253 nd6log((LOG_DEBUG,
1254 "%s: failed to get ifid for %s\n",
1255 __func__, if_name(ifp)));
1256 ifa_free(ifa);
1257 ifid_addr = NULL;
1258 }
1259 }
1260 }
1261
1262 if (ifid_addr == NULL) {
1263 nd6log((LOG_INFO,
1264 "%s: could not determine ifid for %s\n",
1265 __func__, if_name(ifp)));
1266 return NULL;
1267 }
1268
1269 memcpy(&newaddr, &ib->ia_addr.sin6_addr, sizeof(ib->ia_addr.sin6_addr));
1270 ifa_free(ifa);
1271 }
1272
1273 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1274 /* interface ID */
1275 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= (newaddr.s6_addr32[0] & ~mask.s6_addr32[0]);
1276 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= (newaddr.s6_addr32[1] & ~mask.s6_addr32[1]);
1277 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= (newaddr.s6_addr32[2] & ~mask.s6_addr32[2]);
1278 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= (newaddr.s6_addr32[3] & ~mask.s6_addr32[3]);
1279
1280 /* lifetimes. */
1281 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1282 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1283
1284 /* XXX: scope zone ID? */
1285
1286 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1287
1288 /*
1289 * Make sure that we do not have this address already. This should
1290 * usually not happen, but we can still see this case, e.g., if we
1291 * have manually configured the exact address to be configured.
1292 */
1293 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1294 &ifra.ifra_addr.sin6_addr);
1295 if (ifa != NULL) {
1296 ifa_free(ifa);
1297 /* this should be rare enough to make an explicit log */
1298 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1299 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1300 return (NULL);
1301 }
1302
1303 /*
1304 * Allocate ifaddr structure, link into chain, etc.
1305 * If we are going to create a new address upon receiving a multicasted
1306 * RA, we need to impose a random delay before starting DAD.
1307 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1308 */
1309 updateflags = 0;
1310 if (mcast)
1311 updateflags |= IN6_IFAUPDATE_DADDELAY;
1312 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1313 nd6log((LOG_ERR,
1314 "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__,
1315 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1316 if_name(ifp), error));
1317 return (NULL); /* ifaddr must not have been allocated. */
1318 }
1319
1320 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1321 /*
1322 * XXXRW: Assumption of non-NULLness here might not be true with
1323 * fine-grained locking -- should we validate it? Or just return
1324 * earlier ifa rather than looking it up again?
1325 */
1326 return (ia); /* this is always non-NULL and referenced. */
1327 }
1328
1329 static struct nd_prefix *
nd6_prefix_lookup_locked(struct nd_prefixctl * key)1330 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1331 {
1332 struct nd_prefix *search;
1333
1334 ND6_LOCK_ASSERT();
1335
1336 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1337 if (key->ndpr_ifp == search->ndpr_ifp &&
1338 key->ndpr_plen == search->ndpr_plen &&
1339 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1340 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1341 nd6_prefix_ref(search);
1342 break;
1343 }
1344 }
1345 return (search);
1346 }
1347
1348 struct nd_prefix *
nd6_prefix_lookup(struct nd_prefixctl * key)1349 nd6_prefix_lookup(struct nd_prefixctl *key)
1350 {
1351 struct nd_prefix *search;
1352
1353 ND6_RLOCK();
1354 search = nd6_prefix_lookup_locked(key);
1355 ND6_RUNLOCK();
1356 return (search);
1357 }
1358
1359 void
nd6_prefix_ref(struct nd_prefix * pr)1360 nd6_prefix_ref(struct nd_prefix *pr)
1361 {
1362
1363 refcount_acquire(&pr->ndpr_refcnt);
1364 }
1365
1366 void
nd6_prefix_rele(struct nd_prefix * pr)1367 nd6_prefix_rele(struct nd_prefix *pr)
1368 {
1369
1370 if (refcount_release(&pr->ndpr_refcnt)) {
1371 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1372 ("prefix %p has advertising routers", pr));
1373 free(pr, M_IP6NDP);
1374 }
1375 }
1376
1377 int
nd6_prelist_add(struct nd_prefixctl * pr,struct nd_defrouter * dr,struct nd_prefix ** newp)1378 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1379 struct nd_prefix **newp)
1380 {
1381 struct nd_prefix *new;
1382 char ip6buf[INET6_ADDRSTRLEN];
1383 int error;
1384
1385 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1386 if (new == NULL)
1387 return (ENOMEM);
1388 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1389 new->ndpr_ifp = pr->ndpr_ifp;
1390 new->ndpr_prefix = pr->ndpr_prefix;
1391 new->ndpr_plen = pr->ndpr_plen;
1392 new->ndpr_vltime = pr->ndpr_vltime;
1393 new->ndpr_pltime = pr->ndpr_pltime;
1394 new->ndpr_flags = pr->ndpr_flags;
1395 if ((error = in6_init_prefix_ltimes(new)) != 0) {
1396 free(new, M_IP6NDP);
1397 return (error);
1398 }
1399 new->ndpr_lastupdate = time_uptime;
1400
1401 /* initialization */
1402 LIST_INIT(&new->ndpr_advrtrs);
1403 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1404 /* make prefix in the canonical form */
1405 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1406
1407 ND6_WLOCK();
1408 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1409 V_nd6_list_genid++;
1410 ND6_WUNLOCK();
1411
1412 /* ND_OPT_PI_FLAG_ONLINK processing */
1413 if (new->ndpr_raf_onlink) {
1414 struct epoch_tracker et;
1415
1416 ND6_ONLINK_LOCK();
1417 NET_EPOCH_ENTER(et);
1418 if ((error = nd6_prefix_onlink(new)) != 0) {
1419 nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d "
1420 "on-link on %s (errno=%d)\n", __func__,
1421 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1422 pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1423 /* proceed anyway. XXX: is it correct? */
1424 }
1425 NET_EPOCH_EXIT(et);
1426 ND6_ONLINK_UNLOCK();
1427 }
1428
1429 if (dr != NULL)
1430 pfxrtr_add(new, dr);
1431 if (newp != NULL)
1432 *newp = new;
1433 return (0);
1434 }
1435
1436 /*
1437 * Remove a prefix from the prefix list and optionally stash it in a
1438 * caller-provided list.
1439 *
1440 * The ND6 lock must be held.
1441 */
1442 void
nd6_prefix_unlink(struct nd_prefix * pr,struct nd_prhead * list)1443 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1444 {
1445
1446 ND6_WLOCK_ASSERT();
1447
1448 LIST_REMOVE(pr, ndpr_entry);
1449 V_nd6_list_genid++;
1450 if (list != NULL)
1451 LIST_INSERT_HEAD(list, pr, ndpr_entry);
1452 }
1453
1454 /*
1455 * Free an unlinked prefix, first marking it off-link if necessary.
1456 */
1457 void
nd6_prefix_del(struct nd_prefix * pr)1458 nd6_prefix_del(struct nd_prefix *pr)
1459 {
1460 struct nd_pfxrouter *pfr, *next;
1461 int e;
1462 char ip6buf[INET6_ADDRSTRLEN];
1463
1464 KASSERT(pr->ndpr_addrcnt == 0,
1465 ("prefix %p has referencing addresses", pr));
1466 ND6_UNLOCK_ASSERT();
1467
1468 /*
1469 * Though these flags are now meaningless, we'd rather keep the value
1470 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1471 * when executing "ndp -p".
1472 */
1473 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1474 ND6_ONLINK_LOCK();
1475 if ((e = nd6_prefix_offlink(pr)) != 0) {
1476 nd6log((LOG_ERR,
1477 "%s: failed to make the prefix %s/%d offlink on %s "
1478 "(errno=%d)\n", __func__,
1479 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1480 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1481 /* what should we do? */
1482 }
1483 ND6_ONLINK_UNLOCK();
1484 }
1485
1486 /* Release references to routers that have advertised this prefix. */
1487 ND6_WLOCK();
1488 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1489 pfxrtr_del(pfr);
1490 ND6_WUNLOCK();
1491
1492 nd6_prefix_rele(pr);
1493
1494 pfxlist_onlink_check();
1495 }
1496
1497 static int
prelist_update(struct nd_prefixctl * new,struct nd_defrouter * dr,struct mbuf * m,int mcast)1498 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1499 struct mbuf *m, int mcast)
1500 {
1501 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1502 struct ifaddr *ifa;
1503 struct ifnet *ifp = new->ndpr_ifp;
1504 struct nd_prefix *pr;
1505 int error = 0;
1506 int auth;
1507 struct in6_addrlifetime lt6_tmp;
1508 char ip6buf[INET6_ADDRSTRLEN];
1509 bool has_temporary = false;
1510
1511 NET_EPOCH_ASSERT();
1512
1513 auth = 0;
1514 if (m) {
1515 /*
1516 * Authenticity for NA consists authentication for
1517 * both IP header and IP datagrams, doesn't it ?
1518 */
1519 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1520 auth = ((m->m_flags & M_AUTHIPHDR) &&
1521 (m->m_flags & M_AUTHIPDGM));
1522 #endif
1523 }
1524
1525 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1526 /*
1527 * nd6_prefix_lookup() ensures that pr and new have the same
1528 * prefix on a same interface.
1529 */
1530
1531 /*
1532 * Update prefix information. Note that the on-link (L) bit
1533 * and the autonomous (A) bit should NOT be changed from 1
1534 * to 0.
1535 */
1536 if (new->ndpr_raf_onlink == 1)
1537 pr->ndpr_raf_onlink = 1;
1538 if (new->ndpr_raf_auto == 1)
1539 pr->ndpr_raf_auto = 1;
1540 if (new->ndpr_raf_onlink) {
1541 pr->ndpr_vltime = new->ndpr_vltime;
1542 pr->ndpr_pltime = new->ndpr_pltime;
1543 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1544 pr->ndpr_lastupdate = time_uptime;
1545 }
1546
1547 if (new->ndpr_raf_onlink &&
1548 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1549 ND6_ONLINK_LOCK();
1550 if ((error = nd6_prefix_onlink(pr)) != 0) {
1551 nd6log((LOG_ERR,
1552 "%s: failed to make the prefix %s/%d "
1553 "on-link on %s (errno=%d)\n", __func__,
1554 ip6_sprintf(ip6buf,
1555 &pr->ndpr_prefix.sin6_addr),
1556 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1557 error));
1558 /* proceed anyway. XXX: is it correct? */
1559 }
1560 ND6_ONLINK_UNLOCK();
1561 }
1562
1563 if (dr != NULL)
1564 pfxrtr_add(pr, dr);
1565 } else {
1566 if (new->ndpr_vltime == 0)
1567 goto end;
1568 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1569 goto end;
1570
1571 error = nd6_prelist_add(new, dr, &pr);
1572 if (error != 0) {
1573 nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for "
1574 "the prefix %s/%d on %s (errno=%d)\n", __func__,
1575 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1576 new->ndpr_plen, if_name(new->ndpr_ifp), error));
1577 goto end; /* we should just give up in this case. */
1578 }
1579
1580 /*
1581 * XXX: from the ND point of view, we can ignore a prefix
1582 * with the on-link bit being zero. However, we need a
1583 * prefix structure for references from autoconfigured
1584 * addresses. Thus, we explicitly make sure that the prefix
1585 * itself expires now.
1586 */
1587 if (pr->ndpr_raf_onlink == 0) {
1588 pr->ndpr_vltime = 0;
1589 pr->ndpr_pltime = 0;
1590 in6_init_prefix_ltimes(pr);
1591 }
1592 }
1593
1594 /*
1595 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1596 * Note that pr must be non NULL at this point.
1597 */
1598
1599 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1600 if (!new->ndpr_raf_auto)
1601 goto end;
1602
1603 /*
1604 * 5.5.3 (b). the link-local prefix should have been ignored in
1605 * nd6_ra_input.
1606 */
1607
1608 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1609 if (new->ndpr_pltime > new->ndpr_vltime) {
1610 error = EINVAL; /* XXX: won't be used */
1611 goto end;
1612 }
1613
1614 /*
1615 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1616 * an address configured by stateless autoconfiguration already in the
1617 * list of addresses associated with the interface, and the Valid
1618 * Lifetime is not 0, form an address. We first check if we have
1619 * a matching prefix.
1620 * Note: we apply a clarification in rfc2462bis-02 here. We only
1621 * consider autoconfigured addresses while RFC2462 simply said
1622 * "address".
1623 */
1624 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1625 struct in6_ifaddr *ifa6;
1626 u_int32_t remaininglifetime;
1627
1628 if (ifa->ifa_addr->sa_family != AF_INET6)
1629 continue;
1630
1631 ifa6 = (struct in6_ifaddr *)ifa;
1632
1633 /*
1634 * We only consider autoconfigured addresses as per rfc2462bis.
1635 */
1636 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1637 continue;
1638
1639 /*
1640 * Spec is not clear here, but I believe we should concentrate
1641 * on unicast (i.e. not anycast) addresses.
1642 * XXX: other ia6_flags? detached or duplicated?
1643 */
1644 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1645 continue;
1646
1647 /*
1648 * Ignore the address if it is not associated with a prefix
1649 * or is associated with a prefix that is different from this
1650 * one. (pr is never NULL here)
1651 */
1652 if (ifa6->ia6_ndpr != pr)
1653 continue;
1654
1655 /*
1656 * An already autoconfigured address matched. Now that we
1657 * are sure there is at least one matched address, we can
1658 * proceed to 5.5.3. (e): update the lifetimes according to the
1659 * "two hours" rule and the privacy extension.
1660 * We apply some clarifications in rfc2462bis:
1661 * - use remaininglifetime instead of storedlifetime as a
1662 * variable name
1663 * - remove the dead code in the "two-hour" rule
1664 */
1665 #define TWOHOUR (120*60)
1666 lt6_tmp = ifa6->ia6_lifetime;
1667
1668 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1669 remaininglifetime = ND6_INFINITE_LIFETIME;
1670 else if (time_uptime - ifa6->ia6_updatetime >
1671 lt6_tmp.ia6t_vltime) {
1672 /*
1673 * The case of "invalid" address. We should usually
1674 * not see this case.
1675 */
1676 remaininglifetime = 0;
1677 } else
1678 remaininglifetime = lt6_tmp.ia6t_vltime -
1679 (time_uptime - ifa6->ia6_updatetime);
1680
1681 /* when not updating, keep the current stored lifetime. */
1682 lt6_tmp.ia6t_vltime = remaininglifetime;
1683
1684 if (TWOHOUR < new->ndpr_vltime ||
1685 remaininglifetime < new->ndpr_vltime) {
1686 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1687 } else if (remaininglifetime <= TWOHOUR) {
1688 if (auth) {
1689 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1690 }
1691 } else {
1692 /*
1693 * new->ndpr_vltime <= TWOHOUR &&
1694 * TWOHOUR < remaininglifetime
1695 */
1696 lt6_tmp.ia6t_vltime = TWOHOUR;
1697 }
1698
1699 /* The 2 hour rule is not imposed for preferred lifetime. */
1700 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1701
1702 in6_init_address_ltimes(pr, <6_tmp);
1703
1704 /*
1705 * We need to treat lifetimes for temporary addresses
1706 * differently, according to
1707 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1708 * we only update the lifetimes when they are in the maximum
1709 * intervals.
1710 */
1711 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1712 u_int32_t maxvltime, maxpltime;
1713
1714 /*
1715 * if stable addresses (RFC 7217) are enabled, mark that a temporary address has been found
1716 * to avoid generating uneeded extra ones.
1717 */
1718 if (ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR)
1719 has_temporary = true;
1720
1721 if (V_ip6_temp_valid_lifetime >
1722 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1723 V_ip6_desync_factor)) {
1724 maxvltime = V_ip6_temp_valid_lifetime -
1725 (time_uptime - ifa6->ia6_createtime) -
1726 V_ip6_desync_factor;
1727 } else
1728 maxvltime = 0;
1729 if (V_ip6_temp_preferred_lifetime >
1730 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1731 V_ip6_desync_factor)) {
1732 maxpltime = V_ip6_temp_preferred_lifetime -
1733 (time_uptime - ifa6->ia6_createtime) -
1734 V_ip6_desync_factor;
1735 } else
1736 maxpltime = 0;
1737
1738 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1739 lt6_tmp.ia6t_vltime > maxvltime) {
1740 lt6_tmp.ia6t_vltime = maxvltime;
1741 }
1742 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1743 lt6_tmp.ia6t_pltime > maxpltime) {
1744 lt6_tmp.ia6t_pltime = maxpltime;
1745 }
1746 }
1747 ifa6->ia6_lifetime = lt6_tmp;
1748 ifa6->ia6_updatetime = time_uptime;
1749
1750 /*
1751 * If using stable addresses (RFC 7217) and we still have retries to perform, ignore
1752 * addresses already marked as duplicated, since a new one will be generated.
1753 * Also ignore addresses marked as temporary, since their generation is orthogonal to
1754 * opaque stable ones.
1755 *
1756 * There is a small race condition, in that the dad_counter could be incremented
1757 * between here and when a new address is generated, but this will cause that generation
1758 * to fail and no further retries should happen.
1759 */
1760 if (ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR &&
1761 atomic_load_int(&DAD_FAILURES(ifp)) <= V_ip6_stableaddr_maxretries &&
1762 ifa6->ia6_flags & (IN6_IFF_DUPLICATED | IN6_IFF_TEMPORARY))
1763 continue;
1764
1765 if (ia6_match == NULL) /* remember the first one */
1766 ia6_match = ifa6;
1767 }
1768 if (ia6_match == NULL && new->ndpr_vltime) {
1769 int ifidlen;
1770
1771 /*
1772 * 5.5.3 (d) (continued)
1773 * No address matched and the valid lifetime is non-zero.
1774 * Create a new address.
1775 */
1776
1777 /*
1778 * Prefix Length check:
1779 * If the sum of the prefix length and interface identifier
1780 * length does not equal 128 bits, the Prefix Information
1781 * option MUST be ignored. The length of the interface
1782 * identifier is defined in a separate link-type specific
1783 * document.
1784 */
1785 ifidlen = in6_if2idlen(ifp);
1786 if (ifidlen < 0) {
1787 /* this should not happen, so we always log it. */
1788 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1789 if_name(ifp));
1790 goto end;
1791 }
1792 if (ifidlen + pr->ndpr_plen != 128) {
1793 nd6log((LOG_INFO,
1794 "%s: invalid prefixlen %d for %s, ignored\n",
1795 __func__, pr->ndpr_plen, if_name(ifp)));
1796 goto end;
1797 }
1798
1799 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1800 /*
1801 * note that we should use pr (not new) for reference.
1802 */
1803 pr->ndpr_addrcnt++;
1804 ia6->ia6_ndpr = pr;
1805
1806 /*
1807 * RFC 3041 3.3 (2).
1808 * When a new public address is created as described
1809 * in RFC2462, also create a new temporary address.
1810 *
1811 * RFC 3041 3.5.
1812 * When an interface connects to a new link, a new
1813 * randomized interface identifier should be generated
1814 * immediately together with a new set of temporary
1815 * addresses. Thus, we specifiy 1 as the 2nd arg of
1816 * in6_tmpifadd().
1817 *
1818 * Skip this if a temporary address has been marked as
1819 * found (happens only if stable addresses (RFC 7217) is in use)
1820 */
1821 if (V_ip6_use_tempaddr && !has_temporary) {
1822 int e;
1823 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1824 nd6log((LOG_NOTICE, "%s: failed to "
1825 "create a temporary address "
1826 "(errno=%d)\n", __func__, e));
1827 }
1828 }
1829 ifa_free(&ia6->ia_ifa);
1830
1831 /*
1832 * A newly added address might affect the status
1833 * of other addresses, so we check and update it.
1834 * XXX: what if address duplication happens?
1835 */
1836 pfxlist_onlink_check();
1837 } else {
1838 /* just set an error. do not bark here. */
1839 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1840 }
1841 }
1842
1843 end:
1844 if (pr != NULL)
1845 nd6_prefix_rele(pr);
1846 return (error);
1847 }
1848
1849 /*
1850 * A supplement function used in the on-link detection below;
1851 * detect if a given prefix has a (probably) reachable advertising router.
1852 * XXX: lengthy function name...
1853 */
1854 static struct nd_pfxrouter *
find_pfxlist_reachable_router(struct nd_prefix * pr)1855 find_pfxlist_reachable_router(struct nd_prefix *pr)
1856 {
1857 struct epoch_tracker et;
1858 struct nd_pfxrouter *pfxrtr;
1859
1860 ND6_LOCK_ASSERT();
1861
1862 NET_EPOCH_ENTER(et);
1863 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1864 if (is_dr_reachable(pfxrtr->router))
1865 break;
1866 }
1867 NET_EPOCH_EXIT(et);
1868 return (pfxrtr);
1869 }
1870
1871 /*
1872 * Check if each prefix in the prefix list has at least one available router
1873 * that advertised the prefix (a router is "available" if its neighbor cache
1874 * entry is reachable or probably reachable).
1875 * If the check fails, the prefix may be off-link, because, for example,
1876 * we have moved from the network but the lifetime of the prefix has not
1877 * expired yet. So we should not use the prefix if there is another prefix
1878 * that has an available router.
1879 * But, if there is no prefix that has an available router, we still regard
1880 * all the prefixes as on-link. This is because we can't tell if all the
1881 * routers are simply dead or if we really moved from the network and there
1882 * is no router around us.
1883 */
1884 void
pfxlist_onlink_check(void)1885 pfxlist_onlink_check(void)
1886 {
1887 struct nd_prefix *pr;
1888 struct in6_ifaddr *ifa;
1889 struct nd_defrouter *dr;
1890 struct nd_pfxrouter *pfxrtr = NULL;
1891 struct rm_priotracker in6_ifa_tracker;
1892 uint64_t genid;
1893 uint32_t flags;
1894
1895 ND6_ONLINK_LOCK();
1896 ND6_RLOCK();
1897
1898 /*
1899 * Check if there is a prefix that has a reachable advertising
1900 * router.
1901 */
1902 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1903 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1904 break;
1905 }
1906
1907 /*
1908 * If we have no such prefix, check whether we still have a router
1909 * that does not advertise any prefixes.
1910 */
1911 if (pr == NULL) {
1912 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1913 struct nd_prefix *pr0;
1914
1915 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1916 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1917 break;
1918 }
1919 if (pfxrtr != NULL)
1920 break;
1921 }
1922 }
1923 if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) {
1924 /*
1925 * There is at least one prefix that has a reachable router,
1926 * or at least a router which probably does not advertise
1927 * any prefixes. The latter would be the case when we move
1928 * to a new link where we have a router that does not provide
1929 * prefixes and we configure an address by hand.
1930 * Detach prefixes which have no reachable advertising
1931 * router, and attach other prefixes.
1932 */
1933 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1934 /* XXX: a link-local prefix should never be detached */
1935 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1936 pr->ndpr_raf_onlink == 0 ||
1937 pr->ndpr_raf_auto == 0)
1938 continue;
1939
1940 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1941 find_pfxlist_reachable_router(pr) == NULL)
1942 pr->ndpr_stateflags |= NDPRF_DETACHED;
1943 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1944 find_pfxlist_reachable_router(pr) != NULL)
1945 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1946 }
1947 } else {
1948 /* there is no prefix that has a reachable router */
1949 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1950 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1951 pr->ndpr_raf_onlink == 0 ||
1952 pr->ndpr_raf_auto == 0)
1953 continue;
1954 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1955 }
1956 }
1957
1958 /*
1959 * Remove each interface route associated with a (just) detached
1960 * prefix, and reinstall the interface route for a (just) attached
1961 * prefix. Note that all attempt of reinstallation does not
1962 * necessarily success, when a same prefix is shared among multiple
1963 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1964 * so we don't have to care about them.
1965 */
1966 restart:
1967 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1968 char ip6buf[INET6_ADDRSTRLEN];
1969 int e;
1970
1971 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1972 pr->ndpr_raf_onlink == 0 ||
1973 pr->ndpr_raf_auto == 0)
1974 continue;
1975
1976 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1977 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1978 genid = V_nd6_list_genid;
1979 ND6_RUNLOCK();
1980 if ((flags & NDPRF_ONLINK) != 0 &&
1981 (e = nd6_prefix_offlink(pr)) != 0) {
1982 nd6log((LOG_ERR,
1983 "%s: failed to make %s/%d offlink "
1984 "(errno=%d)\n", __func__,
1985 ip6_sprintf(ip6buf,
1986 &pr->ndpr_prefix.sin6_addr),
1987 pr->ndpr_plen, e));
1988 } else if ((flags & NDPRF_ONLINK) == 0 &&
1989 (e = nd6_prefix_onlink(pr)) != 0) {
1990 nd6log((LOG_ERR,
1991 "%s: failed to make %s/%d onlink "
1992 "(errno=%d)\n", __func__,
1993 ip6_sprintf(ip6buf,
1994 &pr->ndpr_prefix.sin6_addr),
1995 pr->ndpr_plen, e));
1996 }
1997 ND6_RLOCK();
1998 if (genid != V_nd6_list_genid)
1999 goto restart;
2000 }
2001 }
2002
2003 /*
2004 * Changes on the prefix status might affect address status as well.
2005 * Make sure that all addresses derived from an attached prefix are
2006 * attached, and that all addresses derived from a detached prefix are
2007 * detached. Note, however, that a manually configured address should
2008 * always be attached.
2009 * The precise detection logic is same as the one for prefixes.
2010 */
2011 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
2012 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
2013 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
2014 continue;
2015
2016 if (ifa->ia6_ndpr == NULL) {
2017 /*
2018 * This can happen when we first configure the address
2019 * (i.e. the address exists, but the prefix does not).
2020 * XXX: complicated relationships...
2021 */
2022 continue;
2023 }
2024
2025 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
2026 break;
2027 }
2028 if (ifa) {
2029 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
2030 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
2031 continue;
2032
2033 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
2034 continue;
2035
2036 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
2037 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
2038 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
2039 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
2040 nd6_dad_start((struct ifaddr *)ifa, 0);
2041 }
2042 } else {
2043 ifa->ia6_flags |= IN6_IFF_DETACHED;
2044 }
2045 }
2046 } else {
2047 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
2048 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
2049 continue;
2050
2051 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
2052 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
2053 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
2054 /* Do we need a delay in this case? */
2055 nd6_dad_start((struct ifaddr *)ifa, 0);
2056 }
2057 }
2058 }
2059 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
2060 ND6_RUNLOCK();
2061 ND6_ONLINK_UNLOCK();
2062 }
2063
2064 /*
2065 * Add or remove interface route specified by @dst, @netmask and @ifp.
2066 * ifa can be NULL.
2067 * Returns 0 on success
2068 */
2069 static int
nd6_prefix_rtrequest(uint32_t fibnum,int cmd,struct sockaddr_in6 * dst,struct sockaddr_in6 * netmask,struct ifnet * ifp,struct ifaddr * ifa)2070 nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst,
2071 struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa)
2072 {
2073 struct epoch_tracker et;
2074 int error;
2075
2076 /* Prepare gateway */
2077 struct sockaddr_dl_short sdl = {
2078 .sdl_family = AF_LINK,
2079 .sdl_len = sizeof(struct sockaddr_dl_short),
2080 .sdl_type = ifp->if_type,
2081 .sdl_index = ifp->if_index,
2082 };
2083
2084 struct rt_addrinfo info = {
2085 .rti_ifa = ifa,
2086 .rti_ifp = ifp,
2087 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
2088 .rti_info = {
2089 [RTAX_DST] = (struct sockaddr *)dst,
2090 [RTAX_NETMASK] = (struct sockaddr *)netmask,
2091 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
2092 },
2093 };
2094 /* Don't set additional per-gw filters on removal */
2095
2096 NET_EPOCH_ENTER(et);
2097 error = rib_handle_ifaddr_info(fibnum, cmd, &info);
2098 NET_EPOCH_EXIT(et);
2099 return (error);
2100 }
2101
2102 static int
nd6_prefix_onlink_rtrequest(struct nd_prefix * pr,struct ifaddr * ifa)2103 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
2104 {
2105 int error;
2106
2107 struct sockaddr_in6 mask6 = {
2108 .sin6_family = AF_INET6,
2109 .sin6_len = sizeof(struct sockaddr_in6),
2110 .sin6_addr = pr->ndpr_mask,
2111 };
2112 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2113
2114 error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD,
2115 &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa);
2116 if (error == 0)
2117 pr->ndpr_stateflags |= NDPRF_ONLINK;
2118
2119 return (error);
2120 }
2121
2122 static int
nd6_prefix_onlink(struct nd_prefix * pr)2123 nd6_prefix_onlink(struct nd_prefix *pr)
2124 {
2125 struct epoch_tracker et;
2126 struct ifaddr *ifa;
2127 struct ifnet *ifp = pr->ndpr_ifp;
2128 struct nd_prefix *opr;
2129 char ip6buf[INET6_ADDRSTRLEN];
2130 int error;
2131
2132 ND6_ONLINK_LOCK_ASSERT();
2133 ND6_UNLOCK_ASSERT();
2134
2135 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
2136 return (EEXIST);
2137
2138 /*
2139 * Add the interface route associated with the prefix. Before
2140 * installing the route, check if there's the same prefix on another
2141 * interface, and the prefix has already installed the interface route.
2142 * Although such a configuration is expected to be rare, we explicitly
2143 * allow it.
2144 */
2145 ND6_RLOCK();
2146 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2147 if (opr == pr)
2148 continue;
2149
2150 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2151 continue;
2152
2153 if (!V_rt_add_addr_allfibs &&
2154 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
2155 continue;
2156
2157 if (opr->ndpr_plen == pr->ndpr_plen &&
2158 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2159 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2160 ND6_RUNLOCK();
2161 return (0);
2162 }
2163 }
2164 ND6_RUNLOCK();
2165
2166 /*
2167 * We prefer link-local addresses as the associated interface address.
2168 */
2169 /* search for a link-local addr */
2170 NET_EPOCH_ENTER(et);
2171 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
2172 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
2173 if (ifa == NULL) {
2174 /* XXX: freebsd does not have ifa_ifwithaf */
2175 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2176 if (ifa->ifa_addr->sa_family == AF_INET6) {
2177 ifa_ref(ifa);
2178 break;
2179 }
2180 }
2181 /* should we care about ia6_flags? */
2182 }
2183 if (ifa == NULL) {
2184 /*
2185 * This can still happen, when, for example, we receive an RA
2186 * containing a prefix with the L bit set and the A bit clear,
2187 * after removing all IPv6 addresses on the receiving
2188 * interface. This should, of course, be rare though.
2189 */
2190 nd6log((LOG_NOTICE,
2191 "%s: failed to find any ifaddr to add route for a "
2192 "prefix(%s/%d) on %s\n", __func__,
2193 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2194 pr->ndpr_plen, if_name(ifp)));
2195 error = 0;
2196 } else {
2197 error = nd6_prefix_onlink_rtrequest(pr, ifa);
2198 ifa_free(ifa);
2199 }
2200 NET_EPOCH_EXIT(et);
2201
2202 return (error);
2203 }
2204
2205 int
nd6_prefix_offlink(struct nd_prefix * pr)2206 nd6_prefix_offlink(struct nd_prefix *pr)
2207 {
2208 int error = 0;
2209 struct ifnet *ifp = pr->ndpr_ifp;
2210 struct nd_prefix *opr;
2211 char ip6buf[INET6_ADDRSTRLEN];
2212 uint64_t genid;
2213 int a_failure;
2214
2215 ND6_ONLINK_LOCK_ASSERT();
2216 ND6_UNLOCK_ASSERT();
2217
2218 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2219 return (EEXIST);
2220
2221 struct sockaddr_in6 mask6 = {
2222 .sin6_family = AF_INET6,
2223 .sin6_len = sizeof(struct sockaddr_in6),
2224 .sin6_addr = pr->ndpr_mask,
2225 };
2226 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2227
2228 error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE,
2229 &pr->ndpr_prefix, pmask6, ifp, NULL);
2230
2231 a_failure = 1;
2232 if (error == 0) {
2233 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2234
2235 /*
2236 * There might be the same prefix on another interface,
2237 * the prefix which could not be on-link just because we have
2238 * the interface route (see comments in nd6_prefix_onlink).
2239 * If there's one, try to make the prefix on-link on the
2240 * interface.
2241 */
2242 ND6_RLOCK();
2243 restart:
2244 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2245 /*
2246 * KAME specific: detached prefixes should not be
2247 * on-link.
2248 */
2249 if (opr == pr || (opr->ndpr_stateflags &
2250 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2251 continue;
2252
2253 if (opr->ndpr_plen == pr->ndpr_plen &&
2254 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2255 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2256 int e;
2257
2258 genid = V_nd6_list_genid;
2259 ND6_RUNLOCK();
2260 if ((e = nd6_prefix_onlink(opr)) != 0) {
2261 nd6log((LOG_ERR,
2262 "%s: failed to recover a prefix "
2263 "%s/%d from %s to %s (errno=%d)\n",
2264 __func__, ip6_sprintf(ip6buf,
2265 &opr->ndpr_prefix.sin6_addr),
2266 opr->ndpr_plen, if_name(ifp),
2267 if_name(opr->ndpr_ifp), e));
2268 } else
2269 a_failure = 0;
2270 ND6_RLOCK();
2271 if (genid != V_nd6_list_genid)
2272 goto restart;
2273 }
2274 }
2275 ND6_RUNLOCK();
2276 } else {
2277 /* XXX: can we still set the NDPRF_ONLINK flag? */
2278 nd6log((LOG_ERR,
2279 "%s: failed to delete route: %s/%d on %s (errno=%d)\n",
2280 __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2281 pr->ndpr_plen, if_name(ifp), error));
2282 }
2283
2284 if (a_failure)
2285 lltable_prefix_free(AF_INET6,
2286 (struct sockaddr *)&pr->ndpr_prefix,
2287 (struct sockaddr *)&mask6, LLE_STATIC);
2288
2289 return (error);
2290 }
2291
2292 /*
2293 * Get a randomized interface identifier for a temporary address
2294 * Based on RFC 8981, Section 3.3.1.
2295 */
2296 static int
in6_get_tmp_ifid(struct in6_aliasreq * ifra)2297 in6_get_tmp_ifid(struct in6_aliasreq *ifra)
2298 {
2299 struct in6_addr *addr;
2300
2301 if(!is_random_seeded()){
2302 return 1;
2303 }
2304
2305 addr = &(ifra->ifra_addr.sin6_addr);
2306 regen:
2307 ifra->ifra_addr.sin6_addr.s6_addr32[2] |=
2308 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[2]));
2309 ifra->ifra_addr.sin6_addr.s6_addr32[3] |=
2310 (arc4random() & ~(ifra->ifra_prefixmask.sin6_addr.s6_addr32[3]));
2311
2312 /*
2313 * Check if generated address is not inappropriate:
2314 *
2315 * - Reserved IPv6 Interface aIdentifers
2316 * (https://www.iana.org/assignments/ipv6-interface-ids/)
2317 */
2318
2319 /* Subnet-router anycast: 0000:0000:0000:0000 */
2320 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2321 goto regen;
2322
2323 /*
2324 * IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2325 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2326 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2327 */
2328 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2329 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2330 goto regen;
2331
2332 /* Reserved subnet anycast addresses */
2333 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2334 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2335 goto regen;
2336
2337 return 0;
2338 }
2339
2340 /*
2341 * ia0 - corresponding public address
2342 */
2343 int
in6_tmpifadd(const struct in6_ifaddr * ia0,int forcegen,int delay)2344 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2345 {
2346 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2347 struct in6_ifaddr *newia;
2348 struct in6_aliasreq ifra;
2349 int error;
2350 int trylimit = 3; /* XXX: adhoc value */
2351 int updateflags;
2352 time_t vltime0, pltime0;
2353
2354 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2355 &ia0->ia_prefixmask.sin6_addr);
2356
2357 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */
2358 /* clear the old IFID */
2359 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2360 &ifra.ifra_prefixmask.sin6_addr);
2361
2362 again:
2363 if (in6_get_tmp_ifid(&ifra) != 0) {
2364 nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n",
2365 __func__));
2366 return (EINVAL);
2367 }
2368
2369 /*
2370 * in6_get_tmpifid() quite likely provided a unique interface ID.
2371 * However, we may still have a chance to see collision, because
2372 * there may be a time lag between generation of the ID and generation
2373 * of the address. So, we'll do one more sanity check.
2374 */
2375
2376 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2377 if (trylimit-- > 0) {
2378 forcegen = 1;
2379 goto again;
2380 }
2381
2382 /* Give up. Something strange should have happened. */
2383 nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n",
2384 __func__));
2385 return (EEXIST);
2386 }
2387
2388 /*
2389 * The Valid Lifetime is the lower of the Valid Lifetime of the
2390 * public address or TEMP_VALID_LIFETIME.
2391 * The Preferred Lifetime is the lower of the Preferred Lifetime
2392 * of the public address or TEMP_PREFERRED_LIFETIME -
2393 * DESYNC_FACTOR.
2394 */
2395 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2396 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2397 (ia0->ia6_lifetime.ia6t_vltime -
2398 (time_uptime - ia0->ia6_updatetime));
2399 if (vltime0 > V_ip6_temp_valid_lifetime)
2400 vltime0 = V_ip6_temp_valid_lifetime;
2401 } else
2402 vltime0 = V_ip6_temp_valid_lifetime;
2403 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2404 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2405 (ia0->ia6_lifetime.ia6t_pltime -
2406 (time_uptime - ia0->ia6_updatetime));
2407 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2408 pltime0 = V_ip6_temp_preferred_lifetime -
2409 V_ip6_desync_factor;
2410 }
2411 } else
2412 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2413 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2414 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2415
2416 /*
2417 * A temporary address is created only if this calculated Preferred
2418 * Lifetime is greater than REGEN_ADVANCE time units.
2419 */
2420 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2421 return (0);
2422
2423 /* XXX: scope zone ID? */
2424
2425 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2426
2427 /* allocate ifaddr structure, link into chain, etc. */
2428 updateflags = 0;
2429 if (delay)
2430 updateflags |= IN6_IFAUPDATE_DADDELAY;
2431 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2432 return (error);
2433
2434 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2435 if (newia == NULL) { /* XXX: can it happen? */
2436 nd6log((LOG_ERR,
2437 "%s: ifa update succeeded, but we got no ifaddr\n",
2438 __func__));
2439 return (EINVAL); /* XXX */
2440 }
2441 newia->ia6_ndpr = ia0->ia6_ndpr;
2442 newia->ia6_ndpr->ndpr_addrcnt++;
2443 ifa_free(&newia->ia_ifa);
2444
2445 /*
2446 * A newly added address might affect the status of other addresses.
2447 * XXX: when the temporary address is generated with a new public
2448 * address, the onlink check is redundant. However, it would be safe
2449 * to do the check explicitly everywhere a new address is generated,
2450 * and, in fact, we surely need the check when we create a new
2451 * temporary address due to deprecation of an old temporary address.
2452 */
2453 pfxlist_onlink_check();
2454
2455 return (0);
2456 }
2457
2458 static int
rt6_deleteroute(const struct rtentry * rt,const struct nhop_object * nh,void * arg)2459 rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh,
2460 void *arg)
2461 {
2462 struct in6_addr *gate = (struct in6_addr *)arg;
2463 int nh_rt_flags;
2464
2465 if (nh->gw_sa.sa_family != AF_INET6)
2466 return (0);
2467
2468 if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) {
2469 return (0);
2470 }
2471
2472 /*
2473 * Do not delete a static route.
2474 * XXX: this seems to be a bit ad-hoc. Should we consider the
2475 * 'cloned' bit instead?
2476 */
2477 nh_rt_flags = nhop_get_rtflags(nh);
2478 if ((nh_rt_flags & RTF_STATIC) != 0)
2479 return (0);
2480
2481 /*
2482 * We delete only host route. This means, in particular, we don't
2483 * delete default route.
2484 */
2485 if ((nh_rt_flags & RTF_HOST) == 0)
2486 return (0);
2487
2488 return (1);
2489 #undef SIN6
2490 }
2491
2492 /*
2493 * Delete all the routing table entries that use the specified gateway.
2494 * XXX: this function causes search through all entries of routing table, so
2495 * it shouldn't be called when acting as a router.
2496 */
2497 void
rt6_flush(struct in6_addr * gateway,struct ifnet * ifp)2498 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2499 {
2500
2501 /* We'll care only link-local addresses */
2502 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2503 return;
2504
2505 /* XXX Do we really need to walk any but the default FIB? */
2506 rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2507 }
2508
2509 int
nd6_setdefaultiface(int ifindex)2510 nd6_setdefaultiface(int ifindex)
2511 {
2512
2513 if (V_nd6_defifindex != ifindex) {
2514 V_nd6_defifindex = ifindex;
2515 if (V_nd6_defifindex != 0) {
2516 struct epoch_tracker et;
2517
2518 /*
2519 * XXXGL: this function should use ifnet_byindex_ref!
2520 */
2521 NET_EPOCH_ENTER(et);
2522 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2523 NET_EPOCH_EXIT(et);
2524 if (V_nd6_defifp == NULL)
2525 return (EINVAL);
2526 } else
2527 V_nd6_defifp = NULL;
2528
2529 /*
2530 * Our current implementation assumes one-to-one mapping between
2531 * interfaces and links, so it would be natural to use the
2532 * default interface as the default link.
2533 */
2534 scope6_setdefault(V_nd6_defifp);
2535 }
2536
2537 return (0);
2538 }
2539
2540 bool
nd6_defrouter_list_empty(void)2541 nd6_defrouter_list_empty(void)
2542 {
2543
2544 return (TAILQ_EMPTY(&V_nd6_defrouter));
2545 }
2546
2547 void
nd6_defrouter_timer(void)2548 nd6_defrouter_timer(void)
2549 {
2550 struct nd_defrouter *dr, *ndr;
2551 struct nd6_drhead drq;
2552
2553 TAILQ_INIT(&drq);
2554
2555 ND6_WLOCK();
2556 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr)
2557 if (dr->expire && dr->expire < time_uptime)
2558 defrouter_unlink(dr, &drq);
2559 ND6_WUNLOCK();
2560
2561 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2562 TAILQ_REMOVE(&drq, dr, dr_entry);
2563 defrouter_del(dr);
2564 }
2565 }
2566
2567 /*
2568 * Nuke default router list entries toward ifp.
2569 * We defer removal of default router list entries that is installed in the
2570 * routing table, in order to keep additional side effects as small as possible.
2571 */
2572 void
nd6_defrouter_purge(struct ifnet * ifp)2573 nd6_defrouter_purge(struct ifnet *ifp)
2574 {
2575 struct nd_defrouter *dr, *ndr;
2576 struct nd6_drhead drq;
2577
2578 TAILQ_INIT(&drq);
2579
2580 ND6_WLOCK();
2581 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2582 if (dr->installed)
2583 continue;
2584 if (dr->ifp == ifp)
2585 defrouter_unlink(dr, &drq);
2586 }
2587 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2588 if (!dr->installed)
2589 continue;
2590 if (dr->ifp == ifp)
2591 defrouter_unlink(dr, &drq);
2592 }
2593 ND6_WUNLOCK();
2594
2595 /* Delete the unlinked router objects. */
2596 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2597 TAILQ_REMOVE(&drq, dr, dr_entry);
2598 defrouter_del(dr);
2599 }
2600 }
2601
2602 void
nd6_defrouter_flush_all(void)2603 nd6_defrouter_flush_all(void)
2604 {
2605 struct nd_defrouter *dr;
2606 struct nd6_drhead drq;
2607
2608 TAILQ_INIT(&drq);
2609
2610 ND6_WLOCK();
2611 while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL)
2612 defrouter_unlink(dr, &drq);
2613 ND6_WUNLOCK();
2614
2615 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2616 TAILQ_REMOVE(&drq, dr, dr_entry);
2617 defrouter_del(dr);
2618 }
2619 }
2620
2621 void
nd6_defrouter_init(void)2622 nd6_defrouter_init(void)
2623 {
2624
2625 TAILQ_INIT(&V_nd6_defrouter);
2626 }
2627
2628 static int
nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)2629 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2630 {
2631 struct in6_defrouter d;
2632 struct nd_defrouter *dr;
2633 int error;
2634
2635 if (req->newptr != NULL)
2636 return (EPERM);
2637
2638 error = sysctl_wire_old_buffer(req, 0);
2639 if (error != 0)
2640 return (error);
2641
2642 bzero(&d, sizeof(d));
2643 d.rtaddr.sin6_family = AF_INET6;
2644 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2645
2646 ND6_RLOCK();
2647 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
2648 d.rtaddr.sin6_addr = dr->rtaddr;
2649 error = sa6_recoverscope(&d.rtaddr);
2650 if (error != 0)
2651 break;
2652 d.flags = dr->raflags;
2653 d.rtlifetime = dr->rtlifetime;
2654 d.expire = dr->expire + (time_second - time_uptime);
2655 d.if_index = dr->ifp->if_index;
2656 error = SYSCTL_OUT(req, &d, sizeof(d));
2657 if (error != 0)
2658 break;
2659 }
2660 ND6_RUNLOCK();
2661 return (error);
2662 }
2663 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2664 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2665 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2666 "NDP default router list");
2667