FreeBSD/Linux Kernel Cross Reference
sys/netinet6/nd6.c
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.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
32 */
33
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_route.h"
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/eventhandler.h>
44 #include <sys/callout.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/mutex.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/protosw.h>
54 #include <sys/errno.h>
55 #include <sys/syslog.h>
56 #include <sys/rwlock.h>
57 #include <sys/queue.h>
58 #include <sys/sdt.h>
59 #include <sys/sysctl.h>
60
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
66 #include <net/route/route_ctl.h>
67 #include <net/route/nhop.h>
68 #include <net/vnet.h>
69
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <net/if_llatbl.h>
73 #include <netinet/if_ether.h>
74 #include <netinet6/in6_fib.h>
75 #include <netinet6/in6_var.h>
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #include <netinet6/nd6.h>
80 #include <netinet6/in6_ifattach.h>
81 #include <netinet/icmp6.h>
82 #include <netinet6/send.h>
83
84 #include <sys/limits.h>
85
86 #include <security/mac/mac_framework.h>
87
88 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
89 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
90
91 #define SIN6(s) ((const struct sockaddr_in6 *)(s))
92
93 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery");
94
95 /* timer values */
96 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
97 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
98 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
99 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
100 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
101 * local traffic */
102 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
103 * collection timer */
104
105 /* preventing too many loops in ND option parsing */
106 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
107
108 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
109 * layer hints */
110 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 16; /* max pkts cached in unresolved
111 * ND entries */
112 #define V_nd6_maxndopt VNET(nd6_maxndopt)
113 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
114
115 #ifdef ND6_DEBUG
116 VNET_DEFINE(int, nd6_debug) = 1;
117 #else
118 VNET_DEFINE(int, nd6_debug) = 0;
119 #endif
120
121 static eventhandler_tag lle_event_eh, iflladdr_event_eh, ifnet_link_event_eh;
122
123 VNET_DEFINE(struct nd_prhead, nd_prefix);
124 VNET_DEFINE(struct rwlock, nd6_lock);
125 VNET_DEFINE(uint64_t, nd6_list_genid);
126 VNET_DEFINE(struct mtx, nd6_onlink_mtx);
127
128 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
129 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
130
131 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
132
133 static bool nd6_is_new_addr_neighbor(const struct sockaddr_in6 *,
134 struct ifnet *);
135 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
136 static void nd6_slowtimo(void *);
137 static int regen_tmpaddr(struct in6_ifaddr *);
138 static void nd6_free(struct llentry **, int);
139 static void nd6_free_redirect(const struct llentry *);
140 static void nd6_llinfo_timer(void *);
141 static void nd6_llinfo_settimer_locked(struct llentry *, long);
142 static int nd6_resolve_slow(struct ifnet *, int, int, struct mbuf *,
143 const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **);
144 static int nd6_need_cache(struct ifnet *);
145
146 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch);
147 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
148
149 VNET_DEFINE_STATIC(struct callout, nd6_timer_ch);
150 #define V_nd6_timer_ch VNET(nd6_timer_ch)
151
152 SYSCTL_DECL(_net_inet6_icmp6);
153
154 static void
155 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
156 {
157 struct rt_addrinfo rtinfo;
158 struct sockaddr_in6 dst;
159 struct sockaddr_dl gw;
160 struct ifnet *ifp;
161 int type;
162 int fibnum;
163
164 LLE_WLOCK_ASSERT(lle);
165
166 if (lltable_get_af(lle->lle_tbl) != AF_INET6)
167 return;
168
169 switch (evt) {
170 case LLENTRY_RESOLVED:
171 type = RTM_ADD;
172 KASSERT(lle->la_flags & LLE_VALID,
173 ("%s: %p resolved but not valid?", __func__, lle));
174 break;
175 case LLENTRY_EXPIRED:
176 type = RTM_DELETE;
177 break;
178 default:
179 return;
180 }
181
182 ifp = lltable_get_ifp(lle->lle_tbl);
183
184 bzero(&dst, sizeof(dst));
185 bzero(&gw, sizeof(gw));
186 bzero(&rtinfo, sizeof(rtinfo));
187 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst);
188 dst.sin6_scope_id = in6_getscopezone(ifp,
189 in6_addrscope(&dst.sin6_addr));
190 gw.sdl_len = sizeof(struct sockaddr_dl);
191 gw.sdl_family = AF_LINK;
192 gw.sdl_alen = ifp->if_addrlen;
193 gw.sdl_index = ifp->if_index;
194 gw.sdl_type = ifp->if_type;
195 if (evt == LLENTRY_RESOLVED)
196 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
197 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
198 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
199 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
200 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib;
201 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | (
202 type == RTM_ADD ? RTF_UP: 0), 0, fibnum);
203 }
204
205 /*
206 * A handler for interface link layer address change event.
207 */
208 static void
209 nd6_iflladdr(void *arg __unused, struct ifnet *ifp)
210 {
211 if (ifp->if_afdata[AF_INET6] == NULL)
212 return;
213
214 lltable_update_ifaddr(LLTABLE6(ifp));
215 }
216
217 void
218 nd6_init(void)
219 {
220
221 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF);
222 rw_init(&V_nd6_lock, "nd6 list");
223
224 LIST_INIT(&V_nd_prefix);
225 nd6_defrouter_init();
226
227 /* Start timers. */
228 callout_init(&V_nd6_slowtimo_ch, 1);
229 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
230 nd6_slowtimo, curvnet);
231
232 callout_init(&V_nd6_timer_ch, 1);
233 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
234
235 nd6_dad_init();
236 if (IS_DEFAULT_VNET(curvnet)) {
237 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
238 NULL, EVENTHANDLER_PRI_ANY);
239 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event,
240 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
241 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event,
242 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY);
243 }
244 }
245
246 #ifdef VIMAGE
247 void
248 nd6_destroy(void)
249 {
250
251 callout_drain(&V_nd6_slowtimo_ch);
252 callout_drain(&V_nd6_timer_ch);
253 if (IS_DEFAULT_VNET(curvnet)) {
254 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh);
255 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
256 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
257 }
258 rw_destroy(&V_nd6_lock);
259 mtx_destroy(&V_nd6_onlink_mtx);
260 }
261 #endif
262
263 struct nd_ifinfo *
264 nd6_ifattach(struct ifnet *ifp)
265 {
266 struct nd_ifinfo *nd;
267
268 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
269 nd->initialized = 1;
270
271 nd->chlim = IPV6_DEFHLIM;
272 nd->basereachable = REACHABLE_TIME;
273 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
274 nd->retrans = RETRANS_TIMER;
275
276 nd->flags = ND6_IFF_PERFORMNUD;
277
278 /* Set IPv6 disabled on all interfaces but loopback by default. */
279 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
280 nd->flags |= ND6_IFF_IFDISABLED;
281
282 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
283 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
284 * default regardless of the V_ip6_auto_linklocal configuration to
285 * give a reasonable default behavior.
286 */
287 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE &&
288 ifp->if_type != IFT_WIREGUARD) || (ifp->if_flags & IFF_LOOPBACK))
289 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
290 /*
291 * A loopback interface does not need to accept RTADV.
292 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
293 * default regardless of the V_ip6_accept_rtadv configuration to
294 * prevent the interface from accepting RA messages arrived
295 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
296 */
297 if (V_ip6_accept_rtadv &&
298 !(ifp->if_flags & IFF_LOOPBACK) &&
299 (ifp->if_type != IFT_BRIDGE)) {
300 nd->flags |= ND6_IFF_ACCEPT_RTADV;
301 /* If we globally accept rtadv, assume IPv6 on. */
302 nd->flags &= ~ND6_IFF_IFDISABLED;
303 }
304 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
305 nd->flags |= ND6_IFF_NO_RADR;
306
307 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
308 nd6_setmtu0(ifp, nd);
309
310 return nd;
311 }
312
313 void
314 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd)
315 {
316 struct epoch_tracker et;
317 struct ifaddr *ifa, *next;
318
319 NET_EPOCH_ENTER(et);
320 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
321 if (ifa->ifa_addr->sa_family != AF_INET6)
322 continue;
323
324 /* stop DAD processing */
325 nd6_dad_stop(ifa);
326 }
327 NET_EPOCH_EXIT(et);
328
329 free(nd, M_IP6NDP);
330 }
331
332 /*
333 * Reset ND level link MTU. This function is called when the physical MTU
334 * changes, which means we might have to adjust the ND level MTU.
335 */
336 void
337 nd6_setmtu(struct ifnet *ifp)
338 {
339 if (ifp->if_afdata[AF_INET6] == NULL)
340 return;
341
342 nd6_setmtu0(ifp, ND_IFINFO(ifp));
343 }
344
345 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
346 void
347 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
348 {
349 u_int32_t omaxmtu;
350
351 omaxmtu = ndi->maxmtu;
352 ndi->maxmtu = ifp->if_mtu;
353
354 /*
355 * Decreasing the interface MTU under IPV6 minimum MTU may cause
356 * undesirable situation. We thus notify the operator of the change
357 * explicitly. The check for omaxmtu is necessary to restrict the
358 * log to the case of changing the MTU, not initializing it.
359 */
360 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
361 log(LOG_NOTICE, "nd6_setmtu0: "
362 "new link MTU on %s (%lu) is too small for IPv6\n",
363 if_name(ifp), (unsigned long)ndi->maxmtu);
364 }
365
366 if (ndi->maxmtu > V_in6_maxmtu)
367 in6_setmaxmtu(); /* check all interfaces just in case */
368
369 }
370
371 void
372 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
373 {
374
375 bzero(ndopts, sizeof(*ndopts));
376 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
377 ndopts->nd_opts_last
378 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
379
380 if (icmp6len == 0) {
381 ndopts->nd_opts_done = 1;
382 ndopts->nd_opts_search = NULL;
383 }
384 }
385
386 /*
387 * Take one ND option.
388 */
389 struct nd_opt_hdr *
390 nd6_option(union nd_opts *ndopts)
391 {
392 struct nd_opt_hdr *nd_opt;
393 int olen;
394
395 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
396 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
397 __func__));
398 if (ndopts->nd_opts_search == NULL)
399 return NULL;
400 if (ndopts->nd_opts_done)
401 return NULL;
402
403 nd_opt = ndopts->nd_opts_search;
404
405 /* make sure nd_opt_len is inside the buffer */
406 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
407 bzero(ndopts, sizeof(*ndopts));
408 return NULL;
409 }
410
411 olen = nd_opt->nd_opt_len << 3;
412 if (olen == 0) {
413 /*
414 * Message validation requires that all included
415 * options have a length that is greater than zero.
416 */
417 bzero(ndopts, sizeof(*ndopts));
418 return NULL;
419 }
420
421 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
422 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
423 /* option overruns the end of buffer, invalid */
424 bzero(ndopts, sizeof(*ndopts));
425 return NULL;
426 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
427 /* reached the end of options chain */
428 ndopts->nd_opts_done = 1;
429 ndopts->nd_opts_search = NULL;
430 }
431 return nd_opt;
432 }
433
434 /*
435 * Parse multiple ND options.
436 * This function is much easier to use, for ND routines that do not need
437 * multiple options of the same type.
438 */
439 int
440 nd6_options(union nd_opts *ndopts)
441 {
442 struct nd_opt_hdr *nd_opt;
443 int i = 0;
444
445 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
446 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
447 __func__));
448 if (ndopts->nd_opts_search == NULL)
449 return 0;
450
451 while (1) {
452 nd_opt = nd6_option(ndopts);
453 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
454 /*
455 * Message validation requires that all included
456 * options have a length that is greater than zero.
457 */
458 ICMP6STAT_INC(icp6s_nd_badopt);
459 bzero(ndopts, sizeof(*ndopts));
460 return -1;
461 }
462
463 if (nd_opt == NULL)
464 goto skip1;
465
466 switch (nd_opt->nd_opt_type) {
467 case ND_OPT_SOURCE_LINKADDR:
468 case ND_OPT_TARGET_LINKADDR:
469 case ND_OPT_MTU:
470 case ND_OPT_REDIRECTED_HEADER:
471 case ND_OPT_NONCE:
472 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
473 nd6log((LOG_INFO,
474 "duplicated ND6 option found (type=%d)\n",
475 nd_opt->nd_opt_type));
476 /* XXX bark? */
477 } else {
478 ndopts->nd_opt_array[nd_opt->nd_opt_type]
479 = nd_opt;
480 }
481 break;
482 case ND_OPT_PREFIX_INFORMATION:
483 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
484 ndopts->nd_opt_array[nd_opt->nd_opt_type]
485 = nd_opt;
486 }
487 ndopts->nd_opts_pi_end =
488 (struct nd_opt_prefix_info *)nd_opt;
489 break;
490 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
491 case ND_OPT_RDNSS: /* RFC 6106 */
492 case ND_OPT_DNSSL: /* RFC 6106 */
493 /*
494 * Silently ignore options we know and do not care about
495 * in the kernel.
496 */
497 break;
498 default:
499 /*
500 * Unknown options must be silently ignored,
501 * to accommodate future extension to the protocol.
502 */
503 nd6log((LOG_DEBUG,
504 "nd6_options: unsupported option %d - "
505 "option ignored\n", nd_opt->nd_opt_type));
506 }
507
508 skip1:
509 i++;
510 if (i > V_nd6_maxndopt) {
511 ICMP6STAT_INC(icp6s_nd_toomanyopt);
512 nd6log((LOG_INFO, "too many loop in nd opt\n"));
513 break;
514 }
515
516 if (ndopts->nd_opts_done)
517 break;
518 }
519
520 return 0;
521 }
522
523 /*
524 * ND6 timer routine to handle ND6 entries
525 */
526 static void
527 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
528 {
529 int canceled;
530
531 LLE_WLOCK_ASSERT(ln);
532
533 /* Do not schedule timers for child LLEs. */
534 if (ln->la_flags & LLE_CHILD)
535 return;
536
537 if (tick < 0) {
538 ln->la_expire = 0;
539 ln->ln_ntick = 0;
540 canceled = callout_stop(&ln->lle_timer);
541 } else {
542 ln->la_expire = time_uptime + tick / hz;
543 LLE_ADDREF(ln);
544 if (tick > INT_MAX) {
545 ln->ln_ntick = tick - INT_MAX;
546 canceled = callout_reset(&ln->lle_timer, INT_MAX,
547 nd6_llinfo_timer, ln);
548 } else {
549 ln->ln_ntick = 0;
550 canceled = callout_reset(&ln->lle_timer, tick,
551 nd6_llinfo_timer, ln);
552 }
553 }
554 if (canceled > 0)
555 LLE_REMREF(ln);
556 }
557
558 /*
559 * Gets source address of the first packet in hold queue
560 * and stores it in @src.
561 * Returns pointer to @src (if hold queue is not empty) or NULL.
562 *
563 * Set noinline to be dtrace-friendly
564 */
565 static __noinline struct in6_addr *
566 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
567 {
568 struct ip6_hdr hdr;
569 struct mbuf *m;
570
571 if (ln->la_hold == NULL)
572 return (NULL);
573
574 /*
575 * assume every packet in la_hold has the same IP header
576 */
577 m = ln->la_hold;
578 if (sizeof(hdr) > m->m_len)
579 return (NULL);
580
581 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
582 *src = hdr.ip6_src;
583
584 return (src);
585 }
586
587 /*
588 * Checks if we need to switch from STALE state.
589 *
590 * RFC 4861 requires switching from STALE to DELAY state
591 * on first packet matching entry, waiting V_nd6_delay and
592 * transition to PROBE state (if upper layer confirmation was
593 * not received).
594 *
595 * This code performs a bit differently:
596 * On packet hit we don't change state (but desired state
597 * can be guessed by control plane). However, after V_nd6_delay
598 * seconds code will transition to PROBE state (so DELAY state
599 * is kinda skipped in most situations).
600 *
601 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so
602 * we perform the following upon entering STALE state:
603 *
604 * 1) Arm timer to run each V_nd6_delay seconds to make sure that
605 * if packet was transmitted at the start of given interval, we
606 * would be able to switch to PROBE state in V_nd6_delay seconds
607 * as user expects.
608 *
609 * 2) Reschedule timer until original V_nd6_gctimer expires keeping
610 * lle in STALE state (remaining timer value stored in lle_remtime).
611 *
612 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay
613 * seconds ago.
614 *
615 * Returns non-zero value if the entry is still STALE (storing
616 * the next timer interval in @pdelay).
617 *
618 * Returns zero value if original timer expired or we need to switch to
619 * PROBE (store that in @do_switch variable).
620 */
621 static int
622 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch)
623 {
624 int nd_delay, nd_gctimer;
625 time_t lle_hittime;
626 long delay;
627
628 *do_switch = 0;
629 nd_gctimer = V_nd6_gctimer;
630 nd_delay = V_nd6_delay;
631
632 lle_hittime = llentry_get_hittime(lle);
633
634 if (lle_hittime == 0) {
635 /*
636 * Datapath feedback has been requested upon entering
637 * STALE state. No packets has been passed using this lle.
638 * Ask for the timer reschedule and keep STALE state.
639 */
640 delay = (long)(MIN(nd_gctimer, nd_delay));
641 delay *= hz;
642 if (lle->lle_remtime > delay)
643 lle->lle_remtime -= delay;
644 else {
645 delay = lle->lle_remtime;
646 lle->lle_remtime = 0;
647 }
648
649 if (delay == 0) {
650 /*
651 * The original ng6_gctime timeout ended,
652 * no more rescheduling.
653 */
654 return (0);
655 }
656
657 *pdelay = delay;
658 return (1);
659 }
660
661 /*
662 * Packet received. Verify timestamp
663 */
664 delay = (long)(time_uptime - lle_hittime);
665 if (delay < nd_delay) {
666 /*
667 * V_nd6_delay still not passed since the first
668 * hit in STALE state.
669 * Reschedule timer and return.
670 */
671 *pdelay = (long)(nd_delay - delay) * hz;
672 return (1);
673 }
674
675 /* Request switching to probe */
676 *do_switch = 1;
677 return (0);
678 }
679
680 /*
681 * Switch @lle state to new state optionally arming timers.
682 *
683 * Set noinline to be dtrace-friendly
684 */
685 __noinline void
686 nd6_llinfo_setstate(struct llentry *lle, int newstate)
687 {
688 struct ifnet *ifp;
689 int nd_gctimer, nd_delay;
690 long delay, remtime;
691
692 delay = 0;
693 remtime = 0;
694
695 switch (newstate) {
696 case ND6_LLINFO_INCOMPLETE:
697 ifp = lle->lle_tbl->llt_ifp;
698 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
699 break;
700 case ND6_LLINFO_REACHABLE:
701 if (!ND6_LLINFO_PERMANENT(lle)) {
702 ifp = lle->lle_tbl->llt_ifp;
703 delay = (long)ND_IFINFO(ifp)->reachable * hz;
704 }
705 break;
706 case ND6_LLINFO_STALE:
707
708 llentry_request_feedback(lle);
709 nd_delay = V_nd6_delay;
710 nd_gctimer = V_nd6_gctimer;
711
712 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz;
713 remtime = (long)nd_gctimer * hz - delay;
714 break;
715 case ND6_LLINFO_DELAY:
716 lle->la_asked = 0;
717 delay = (long)V_nd6_delay * hz;
718 break;
719 }
720
721 if (delay > 0)
722 nd6_llinfo_settimer_locked(lle, delay);
723
724 lle->lle_remtime = remtime;
725 lle->ln_state = newstate;
726 }
727
728 /*
729 * Timer-dependent part of nd state machine.
730 *
731 * Set noinline to be dtrace-friendly
732 */
733 static __noinline void
734 nd6_llinfo_timer(void *arg)
735 {
736 struct epoch_tracker et;
737 struct llentry *ln;
738 struct in6_addr *dst, *pdst, *psrc, src;
739 struct ifnet *ifp;
740 struct nd_ifinfo *ndi;
741 int do_switch, send_ns;
742 long delay;
743
744 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
745 ln = (struct llentry *)arg;
746 ifp = lltable_get_ifp(ln->lle_tbl);
747 CURVNET_SET(ifp->if_vnet);
748
749 ND6_RLOCK();
750 LLE_WLOCK(ln);
751 if (callout_pending(&ln->lle_timer)) {
752 /*
753 * Here we are a bit odd here in the treatment of
754 * active/pending. If the pending bit is set, it got
755 * rescheduled before I ran. The active
756 * bit we ignore, since if it was stopped
757 * in ll_tablefree() and was currently running
758 * it would have return 0 so the code would
759 * not have deleted it since the callout could
760 * not be stopped so we want to go through
761 * with the delete here now. If the callout
762 * was restarted, the pending bit will be back on and
763 * we just want to bail since the callout_reset would
764 * return 1 and our reference would have been removed
765 * by nd6_llinfo_settimer_locked above since canceled
766 * would have been 1.
767 */
768 LLE_WUNLOCK(ln);
769 ND6_RUNLOCK();
770 CURVNET_RESTORE();
771 return;
772 }
773 NET_EPOCH_ENTER(et);
774 ndi = ND_IFINFO(ifp);
775 send_ns = 0;
776 dst = &ln->r_l3addr.addr6;
777 pdst = dst;
778
779 if (ln->ln_ntick > 0) {
780 if (ln->ln_ntick > INT_MAX) {
781 ln->ln_ntick -= INT_MAX;
782 nd6_llinfo_settimer_locked(ln, INT_MAX);
783 } else {
784 ln->ln_ntick = 0;
785 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
786 }
787 goto done;
788 }
789
790 if (ln->la_flags & LLE_STATIC) {
791 goto done;
792 }
793
794 if (ln->la_flags & LLE_DELETED) {
795 nd6_free(&ln, 0);
796 goto done;
797 }
798
799 switch (ln->ln_state) {
800 case ND6_LLINFO_INCOMPLETE:
801 if (ln->la_asked < V_nd6_mmaxtries) {
802 ln->la_asked++;
803 send_ns = 1;
804 /* Send NS to multicast address */
805 pdst = NULL;
806 } else {
807 struct mbuf *m;
808
809 ICMP6STAT_ADD(icp6s_dropped, ln->la_numheld);
810
811 m = ln->la_hold;
812 if (m != NULL) {
813 /*
814 * assuming every packet in la_hold has the
815 * same IP header. Send error after unlock.
816 */
817 ln->la_hold = m->m_nextpkt;
818 m->m_nextpkt = NULL;
819 ln->la_numheld--;
820 }
821 nd6_free(&ln, 0);
822 if (m != NULL) {
823 struct mbuf *n = m;
824
825 /*
826 * if there are any ummapped mbufs, we
827 * must free them, rather than using
828 * them for an ICMP, as they cannot be
829 * checksummed.
830 */
831 while ((n = n->m_next) != NULL) {
832 if (n->m_flags & M_EXTPG)
833 break;
834 }
835 if (n != NULL) {
836 m_freem(m);
837 m = NULL;
838 } else {
839 icmp6_error2(m, ICMP6_DST_UNREACH,
840 ICMP6_DST_UNREACH_ADDR, 0, ifp);
841 }
842 }
843 }
844 break;
845 case ND6_LLINFO_REACHABLE:
846 if (!ND6_LLINFO_PERMANENT(ln))
847 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
848 break;
849
850 case ND6_LLINFO_STALE:
851 if (nd6_is_stale(ln, &delay, &do_switch) != 0) {
852 /*
853 * No packet has used this entry and GC timeout
854 * has not been passed. Reschedule timer and
855 * return.
856 */
857 nd6_llinfo_settimer_locked(ln, delay);
858 break;
859 }
860
861 if (do_switch == 0) {
862 /*
863 * GC timer has ended and entry hasn't been used.
864 * Run Garbage collector (RFC 4861, 5.3)
865 */
866 if (!ND6_LLINFO_PERMANENT(ln))
867 nd6_free(&ln, 1);
868 break;
869 }
870
871 /* Entry has been used AND delay timer has ended. */
872
873 /* FALLTHROUGH */
874
875 case ND6_LLINFO_DELAY:
876 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
877 /* We need NUD */
878 ln->la_asked = 1;
879 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
880 send_ns = 1;
881 } else
882 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
883 break;
884 case ND6_LLINFO_PROBE:
885 if (ln->la_asked < V_nd6_umaxtries) {
886 ln->la_asked++;
887 send_ns = 1;
888 } else {
889 nd6_free(&ln, 0);
890 }
891 break;
892 default:
893 panic("%s: paths in a dark night can be confusing: %d",
894 __func__, ln->ln_state);
895 }
896 done:
897 if (ln != NULL)
898 ND6_RUNLOCK();
899 if (send_ns != 0) {
900 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
901 psrc = nd6_llinfo_get_holdsrc(ln, &src);
902 LLE_FREE_LOCKED(ln);
903 ln = NULL;
904 nd6_ns_output(ifp, psrc, pdst, dst, NULL);
905 }
906
907 if (ln != NULL)
908 LLE_FREE_LOCKED(ln);
909 NET_EPOCH_EXIT(et);
910 CURVNET_RESTORE();
911 }
912
913 /*
914 * ND6 timer routine to expire default route list and prefix list
915 */
916 void
917 nd6_timer(void *arg)
918 {
919 CURVNET_SET((struct vnet *) arg);
920 struct epoch_tracker et;
921 struct nd_prhead prl;
922 struct nd_prefix *pr, *npr;
923 struct ifnet *ifp;
924 struct in6_ifaddr *ia6, *nia6;
925 uint64_t genid;
926
927 LIST_INIT(&prl);
928
929 NET_EPOCH_ENTER(et);
930 nd6_defrouter_timer();
931
932 /*
933 * expire interface addresses.
934 * in the past the loop was inside prefix expiry processing.
935 * However, from a stricter speci-confrmance standpoint, we should
936 * rather separate address lifetimes and prefix lifetimes.
937 *
938 * XXXRW: in6_ifaddrhead locking.
939 */
940 addrloop:
941 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
942 /* check address lifetime */
943 if (IFA6_IS_INVALID(ia6)) {
944 int regen = 0;
945
946 /*
947 * If the expiring address is temporary, try
948 * regenerating a new one. This would be useful when
949 * we suspended a laptop PC, then turned it on after a
950 * period that could invalidate all temporary
951 * addresses. Although we may have to restart the
952 * loop (see below), it must be after purging the
953 * address. Otherwise, we'd see an infinite loop of
954 * regeneration.
955 */
956 if (V_ip6_use_tempaddr &&
957 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
958 if (regen_tmpaddr(ia6) == 0)
959 regen = 1;
960 }
961
962 in6_purgeaddr(&ia6->ia_ifa);
963
964 if (regen)
965 goto addrloop; /* XXX: see below */
966 } else if (IFA6_IS_DEPRECATED(ia6)) {
967 int oldflags = ia6->ia6_flags;
968
969 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
970
971 /*
972 * If a temporary address has just become deprecated,
973 * regenerate a new one if possible.
974 */
975 if (V_ip6_use_tempaddr &&
976 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
977 (oldflags & IN6_IFF_DEPRECATED) == 0) {
978 if (regen_tmpaddr(ia6) == 0) {
979 /*
980 * A new temporary address is
981 * generated.
982 * XXX: this means the address chain
983 * has changed while we are still in
984 * the loop. Although the change
985 * would not cause disaster (because
986 * it's not a deletion, but an
987 * addition,) we'd rather restart the
988 * loop just for safety. Or does this
989 * significantly reduce performance??
990 */
991 goto addrloop;
992 }
993 }
994 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
995 /*
996 * Schedule DAD for a tentative address. This happens
997 * if the interface was down or not running
998 * when the address was configured.
999 */
1000 int delay;
1001
1002 delay = arc4random() %
1003 (MAX_RTR_SOLICITATION_DELAY * hz);
1004 nd6_dad_start((struct ifaddr *)ia6, delay);
1005 } else {
1006 /*
1007 * Check status of the interface. If it is down,
1008 * mark the address as tentative for future DAD.
1009 */
1010 ifp = ia6->ia_ifp;
1011 if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 &&
1012 ((ifp->if_flags & IFF_UP) == 0 ||
1013 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1014 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){
1015 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
1016 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
1017 }
1018
1019 /*
1020 * A new RA might have made a deprecated address
1021 * preferred.
1022 */
1023 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
1024 }
1025 }
1026 NET_EPOCH_EXIT(et);
1027
1028 ND6_WLOCK();
1029 restart:
1030 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1031 /*
1032 * Expire prefixes. Since the pltime is only used for
1033 * autoconfigured addresses, pltime processing for prefixes is
1034 * not necessary.
1035 *
1036 * Only unlink after all derived addresses have expired. This
1037 * may not occur until two hours after the prefix has expired
1038 * per RFC 4862. If the prefix expires before its derived
1039 * addresses, mark it off-link. This will be done automatically
1040 * after unlinking if no address references remain.
1041 */
1042 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME ||
1043 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime)
1044 continue;
1045
1046 if (pr->ndpr_addrcnt == 0) {
1047 nd6_prefix_unlink(pr, &prl);
1048 continue;
1049 }
1050 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1051 genid = V_nd6_list_genid;
1052 nd6_prefix_ref(pr);
1053 ND6_WUNLOCK();
1054 ND6_ONLINK_LOCK();
1055 (void)nd6_prefix_offlink(pr);
1056 ND6_ONLINK_UNLOCK();
1057 ND6_WLOCK();
1058 nd6_prefix_rele(pr);
1059 if (genid != V_nd6_list_genid)
1060 goto restart;
1061 }
1062 }
1063 ND6_WUNLOCK();
1064
1065 while ((pr = LIST_FIRST(&prl)) != NULL) {
1066 LIST_REMOVE(pr, ndpr_entry);
1067 nd6_prefix_del(pr);
1068 }
1069
1070 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
1071 nd6_timer, curvnet);
1072
1073 CURVNET_RESTORE();
1074 }
1075
1076 /*
1077 * ia6 - deprecated/invalidated temporary address
1078 */
1079 static int
1080 regen_tmpaddr(struct in6_ifaddr *ia6)
1081 {
1082 struct ifaddr *ifa;
1083 struct ifnet *ifp;
1084 struct in6_ifaddr *public_ifa6 = NULL;
1085
1086 NET_EPOCH_ASSERT();
1087
1088 ifp = ia6->ia_ifa.ifa_ifp;
1089 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1090 struct in6_ifaddr *it6;
1091
1092 if (ifa->ifa_addr->sa_family != AF_INET6)
1093 continue;
1094
1095 it6 = (struct in6_ifaddr *)ifa;
1096
1097 /* ignore no autoconf addresses. */
1098 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1099 continue;
1100
1101 /* ignore autoconf addresses with different prefixes. */
1102 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
1103 continue;
1104
1105 /*
1106 * Now we are looking at an autoconf address with the same
1107 * prefix as ours. If the address is temporary and is still
1108 * preferred, do not create another one. It would be rare, but
1109 * could happen, for example, when we resume a laptop PC after
1110 * a long period.
1111 */
1112 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1113 !IFA6_IS_DEPRECATED(it6)) {
1114 public_ifa6 = NULL;
1115 break;
1116 }
1117
1118 /*
1119 * This is a public autoconf address that has the same prefix
1120 * as ours. If it is preferred, keep it. We can't break the
1121 * loop here, because there may be a still-preferred temporary
1122 * address with the prefix.
1123 */
1124 if (!IFA6_IS_DEPRECATED(it6))
1125 public_ifa6 = it6;
1126 }
1127 if (public_ifa6 != NULL)
1128 ifa_ref(&public_ifa6->ia_ifa);
1129
1130 if (public_ifa6 != NULL) {
1131 int e;
1132
1133 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
1134 ifa_free(&public_ifa6->ia_ifa);
1135 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
1136 " tmp addr,errno=%d\n", e);
1137 return (-1);
1138 }
1139 ifa_free(&public_ifa6->ia_ifa);
1140 return (0);
1141 }
1142
1143 return (-1);
1144 }
1145
1146 /*
1147 * Remove prefix and default router list entries corresponding to ifp. Neighbor
1148 * cache entries are freed in in6_domifdetach().
1149 */
1150 void
1151 nd6_purge(struct ifnet *ifp)
1152 {
1153 struct nd_prhead prl;
1154 struct nd_prefix *pr, *npr;
1155
1156 LIST_INIT(&prl);
1157
1158 /* Purge default router list entries toward ifp. */
1159 nd6_defrouter_purge(ifp);
1160
1161 ND6_WLOCK();
1162 /*
1163 * Remove prefixes on ifp. We should have already removed addresses on
1164 * this interface, so no addresses should be referencing these prefixes.
1165 */
1166 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1167 if (pr->ndpr_ifp == ifp)
1168 nd6_prefix_unlink(pr, &prl);
1169 }
1170 ND6_WUNLOCK();
1171
1172 /* Delete the unlinked prefix objects. */
1173 while ((pr = LIST_FIRST(&prl)) != NULL) {
1174 LIST_REMOVE(pr, ndpr_entry);
1175 nd6_prefix_del(pr);
1176 }
1177
1178 /* cancel default outgoing interface setting */
1179 if (V_nd6_defifindex == ifp->if_index)
1180 nd6_setdefaultiface(0);
1181
1182 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1183 /* Refresh default router list. */
1184 defrouter_select_fib(ifp->if_fib);
1185 }
1186 }
1187
1188 /*
1189 * the caller acquires and releases the lock on the lltbls
1190 * Returns the llentry locked
1191 */
1192 struct llentry *
1193 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1194 {
1195 struct sockaddr_in6 sin6;
1196 struct llentry *ln;
1197
1198 bzero(&sin6, sizeof(sin6));
1199 sin6.sin6_len = sizeof(struct sockaddr_in6);
1200 sin6.sin6_family = AF_INET6;
1201 sin6.sin6_addr = *addr6;
1202
1203 IF_AFDATA_LOCK_ASSERT(ifp);
1204
1205 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6);
1206
1207 return (ln);
1208 }
1209
1210 static struct llentry *
1211 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1212 {
1213 struct sockaddr_in6 sin6;
1214 struct llentry *ln;
1215
1216 bzero(&sin6, sizeof(sin6));
1217 sin6.sin6_len = sizeof(struct sockaddr_in6);
1218 sin6.sin6_family = AF_INET6;
1219 sin6.sin6_addr = *addr6;
1220
1221 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
1222 if (ln != NULL)
1223 ln->ln_state = ND6_LLINFO_NOSTATE;
1224
1225 return (ln);
1226 }
1227
1228 /*
1229 * Test whether a given IPv6 address can be a neighbor.
1230 */
1231 static bool
1232 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1233 {
1234
1235 /*
1236 * A link-local address is always a neighbor.
1237 * XXX: a link does not necessarily specify a single interface.
1238 */
1239 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
1240 struct sockaddr_in6 sin6_copy;
1241 u_int32_t zone;
1242
1243 /*
1244 * We need sin6_copy since sa6_recoverscope() may modify the
1245 * content (XXX).
1246 */
1247 sin6_copy = *addr;
1248 if (sa6_recoverscope(&sin6_copy))
1249 return (0); /* XXX: should be impossible */
1250 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
1251 return (0);
1252 if (sin6_copy.sin6_scope_id == zone)
1253 return (1);
1254 else
1255 return (0);
1256 }
1257 /* Checking global unicast */
1258
1259 /* If an address is directly reachable, it is a neigbor */
1260 struct nhop_object *nh;
1261 nh = fib6_lookup(ifp->if_fib, &addr->sin6_addr, 0, NHR_NONE, 0);
1262 if (nh != NULL && nh->nh_aifp == ifp && (nh->nh_flags & NHF_GATEWAY) == 0)
1263 return (true);
1264
1265 /*
1266 * Check prefixes with desired on-link state, as some may be not
1267 * installed in the routing table.
1268 */
1269 bool matched = false;
1270 struct nd_prefix *pr;
1271 ND6_RLOCK();
1272 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1273 if (pr->ndpr_ifp != ifp)
1274 continue;
1275 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1276 continue;
1277 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1278 &addr->sin6_addr, &pr->ndpr_mask)) {
1279 matched = true;
1280 break;
1281 }
1282 }
1283 ND6_RUNLOCK();
1284 if (matched)
1285 return (true);
1286
1287 /*
1288 * If the address is assigned on the node of the other side of
1289 * a p2p interface, the address should be a neighbor.
1290 */
1291 if (ifp->if_flags & IFF_POINTOPOINT) {
1292 struct ifaddr *ifa;
1293
1294 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1295 if (ifa->ifa_addr->sa_family != addr->sin6_family)
1296 continue;
1297 if (ifa->ifa_dstaddr != NULL &&
1298 sa_equal(addr, ifa->ifa_dstaddr)) {
1299 return (true);
1300 }
1301 }
1302 }
1303
1304 /*
1305 * If the default router list is empty, all addresses are regarded
1306 * as on-link, and thus, as a neighbor.
1307 */
1308 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1309 nd6_defrouter_list_empty() &&
1310 V_nd6_defifindex == ifp->if_index) {
1311 return (1);
1312 }
1313
1314 return (0);
1315 }
1316
1317 /*
1318 * Detect if a given IPv6 address identifies a neighbor on a given link.
1319 * XXX: should take care of the destination of a p2p link?
1320 */
1321 int
1322 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1323 {
1324 struct llentry *lle;
1325 int rc = 0;
1326
1327 NET_EPOCH_ASSERT();
1328 IF_AFDATA_UNLOCK_ASSERT(ifp);
1329 if (nd6_is_new_addr_neighbor(addr, ifp))
1330 return (1);
1331
1332 /*
1333 * Even if the address matches none of our addresses, it might be
1334 * in the neighbor cache.
1335 */
1336 if ((lle = nd6_lookup(&addr->sin6_addr, LLE_SF(AF_INET6, 0), ifp)) != NULL) {
1337 LLE_RUNLOCK(lle);
1338 rc = 1;
1339 }
1340 return (rc);
1341 }
1342
1343 static __noinline void
1344 nd6_free_children(struct llentry *lle)
1345 {
1346 struct llentry *child_lle;
1347
1348 NET_EPOCH_ASSERT();
1349 LLE_WLOCK_ASSERT(lle);
1350
1351 while ((child_lle = CK_SLIST_FIRST(&lle->lle_children)) != NULL) {
1352 LLE_WLOCK(child_lle);
1353 lltable_unlink_child_entry(child_lle);
1354 llentry_free(child_lle);
1355 }
1356 }
1357
1358 /*
1359 * Tries to update @lle address/prepend data with new @lladdr.
1360 *
1361 * Returns true on success.
1362 * In any case, @lle is returned wlocked.
1363 */
1364 static __noinline bool
1365 nd6_try_set_entry_addr_locked(struct ifnet *ifp, struct llentry *lle, char *lladdr)
1366 {
1367 u_char buf[LLE_MAX_LINKHDR];
1368 int fam, off;
1369 size_t sz;
1370
1371 sz = sizeof(buf);
1372 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, buf, &sz, &off) != 0)
1373 return (false);
1374
1375 /* Update data */
1376 lltable_set_entry_addr(ifp, lle, buf, sz, off);
1377
1378 struct llentry *child_lle;
1379 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) {
1380 LLE_WLOCK(child_lle);
1381 fam = child_lle->r_family;
1382 sz = sizeof(buf);
1383 if (lltable_calc_llheader(ifp, fam, lladdr, buf, &sz, &off) == 0) {
1384 /* success */
1385 lltable_set_entry_addr(ifp, child_lle, buf, sz, off);
1386 child_lle->ln_state = ND6_LLINFO_REACHABLE;
1387 }
1388 LLE_WUNLOCK(child_lle);
1389 }
1390
1391 return (true);
1392 }
1393
1394 bool
1395 nd6_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle, char *lladdr)
1396 {
1397 NET_EPOCH_ASSERT();
1398 LLE_WLOCK_ASSERT(lle);
1399
1400 if (!lltable_acquire_wlock(ifp, lle))
1401 return (false);
1402 bool ret = nd6_try_set_entry_addr_locked(ifp, lle, lladdr);
1403 IF_AFDATA_WUNLOCK(ifp);
1404
1405 return (ret);
1406 }
1407
1408 /*
1409 * Free an nd6 llinfo entry.
1410 * Since the function would cause significant changes in the kernel, DO NOT
1411 * make it global, unless you have a strong reason for the change, and are sure
1412 * that the change is safe.
1413 *
1414 * Set noinline to be dtrace-friendly
1415 */
1416 static __noinline void
1417 nd6_free(struct llentry **lnp, int gc)
1418 {
1419 struct ifnet *ifp;
1420 struct llentry *ln;
1421 struct nd_defrouter *dr;
1422
1423 ln = *lnp;
1424 *lnp = NULL;
1425
1426 LLE_WLOCK_ASSERT(ln);
1427 ND6_RLOCK_ASSERT();
1428
1429 KASSERT((ln->la_flags & LLE_CHILD) == 0, ("child lle"));
1430
1431 ifp = lltable_get_ifp(ln->lle_tbl);
1432 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0)
1433 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp);
1434 else
1435 dr = NULL;
1436 ND6_RUNLOCK();
1437
1438 if ((ln->la_flags & LLE_DELETED) == 0)
1439 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
1440
1441 /*
1442 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1443 * even though it is not harmful, it was not really necessary.
1444 */
1445
1446 /* cancel timer */
1447 nd6_llinfo_settimer_locked(ln, -1);
1448
1449 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1450 if (dr != NULL && dr->expire &&
1451 ln->ln_state == ND6_LLINFO_STALE && gc) {
1452 /*
1453 * If the reason for the deletion is just garbage
1454 * collection, and the neighbor is an active default
1455 * router, do not delete it. Instead, reset the GC
1456 * timer using the router's lifetime.
1457 * Simply deleting the entry would affect default
1458 * router selection, which is not necessarily a good
1459 * thing, especially when we're using router preference
1460 * values.
1461 * XXX: the check for ln_state would be redundant,
1462 * but we intentionally keep it just in case.
1463 */
1464 if (dr->expire > time_uptime)
1465 nd6_llinfo_settimer_locked(ln,
1466 (dr->expire - time_uptime) * hz);
1467 else
1468 nd6_llinfo_settimer_locked(ln,
1469 (long)V_nd6_gctimer * hz);
1470
1471 LLE_REMREF(ln);
1472 LLE_WUNLOCK(ln);
1473 defrouter_rele(dr);
1474 return;
1475 }
1476
1477 if (dr) {
1478 /*
1479 * Unreachability of a router might affect the default
1480 * router selection and on-link detection of advertised
1481 * prefixes.
1482 */
1483
1484 /*
1485 * Temporarily fake the state to choose a new default
1486 * router and to perform on-link determination of
1487 * prefixes correctly.
1488 * Below the state will be set correctly,
1489 * or the entry itself will be deleted.
1490 */
1491 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1492 }
1493
1494 if (ln->ln_router || dr) {
1495 /*
1496 * We need to unlock to avoid a LOR with rt6_flush() with the
1497 * rnh and for the calls to pfxlist_onlink_check() and
1498 * defrouter_select_fib() in the block further down for calls
1499 * into nd6_lookup(). We still hold a ref.
1500 */
1501 LLE_WUNLOCK(ln);
1502
1503 /*
1504 * rt6_flush must be called whether or not the neighbor
1505 * is in the Default Router List.
1506 * See a corresponding comment in nd6_na_input().
1507 */
1508 rt6_flush(&ln->r_l3addr.addr6, ifp);
1509 }
1510
1511 if (dr) {
1512 /*
1513 * Since defrouter_select_fib() does not affect the
1514 * on-link determination and MIP6 needs the check
1515 * before the default router selection, we perform
1516 * the check now.
1517 */
1518 pfxlist_onlink_check();
1519
1520 /*
1521 * Refresh default router list.
1522 */
1523 defrouter_select_fib(dr->ifp->if_fib);
1524 }
1525
1526 /*
1527 * If this entry was added by an on-link redirect, remove the
1528 * corresponding host route.
1529 */
1530 if (ln->la_flags & LLE_REDIRECT)
1531 nd6_free_redirect(ln);
1532
1533 if (ln->ln_router || dr)
1534 LLE_WLOCK(ln);
1535 }
1536
1537 /*
1538 * Save to unlock. We still hold an extra reference and will not
1539 * free(9) in llentry_free() if someone else holds one as well.
1540 */
1541 LLE_WUNLOCK(ln);
1542 IF_AFDATA_LOCK(ifp);
1543 LLE_WLOCK(ln);
1544 /* Guard against race with other llentry_free(). */
1545 if (ln->la_flags & LLE_LINKED) {
1546 /* Remove callout reference */
1547 LLE_REMREF(ln);
1548 lltable_unlink_entry(ln->lle_tbl, ln);
1549 }
1550 IF_AFDATA_UNLOCK(ifp);
1551
1552 nd6_free_children(ln);
1553
1554 llentry_free(ln);
1555 if (dr != NULL)
1556 defrouter_rele(dr);
1557 }
1558
1559 static int
1560 nd6_isdynrte(const struct rtentry *rt, const struct nhop_object *nh, void *xap)
1561 {
1562
1563 if (nh->nh_flags & NHF_REDIRECT)
1564 return (1);
1565
1566 return (0);
1567 }
1568
1569 /*
1570 * Remove the rtentry for the given llentry,
1571 * both of which were installed by a redirect.
1572 */
1573 static void
1574 nd6_free_redirect(const struct llentry *ln)
1575 {
1576 int fibnum;
1577 struct sockaddr_in6 sin6;
1578 struct rib_cmd_info rc;
1579 struct epoch_tracker et;
1580
1581 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
1582
1583 NET_EPOCH_ENTER(et);
1584 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
1585 rib_del_route_px(fibnum, (struct sockaddr *)&sin6, 128,
1586 nd6_isdynrte, NULL, 0, &rc);
1587 NET_EPOCH_EXIT(et);
1588 }
1589
1590 /*
1591 * Updates status of the default router route.
1592 */
1593 static void
1594 check_release_defrouter(const struct rib_cmd_info *rc, void *_cbdata)
1595 {
1596 struct nd_defrouter *dr;
1597 struct nhop_object *nh;
1598
1599 nh = rc->rc_nh_old;
1600
1601 if ((nh != NULL) && (nh->nh_flags & NHF_DEFAULT)) {
1602 dr = defrouter_lookup(&nh->gw6_sa.sin6_addr, nh->nh_ifp);
1603 if (dr != NULL) {
1604 dr->installed = 0;
1605 defrouter_rele(dr);
1606 }
1607 }
1608 }
1609
1610 void
1611 nd6_subscription_cb(struct rib_head *rnh, struct rib_cmd_info *rc, void *arg)
1612 {
1613
1614 #ifdef ROUTE_MPATH
1615 rib_decompose_notification(rc, check_release_defrouter, NULL);
1616 #else
1617 check_release_defrouter(rc, NULL);
1618 #endif
1619 }
1620
1621 int
1622 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1623 {
1624 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1625 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1626 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1627 struct epoch_tracker et;
1628 int error = 0;
1629
1630 if (ifp->if_afdata[AF_INET6] == NULL)
1631 return (EPFNOSUPPORT);
1632 switch (cmd) {
1633 case OSIOCGIFINFO_IN6:
1634 #define ND ndi->ndi
1635 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1636 bzero(&ND, sizeof(ND));
1637 ND.linkmtu = IN6_LINKMTU(ifp);
1638 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1639 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1640 ND.reachable = ND_IFINFO(ifp)->reachable;
1641 ND.retrans = ND_IFINFO(ifp)->retrans;
1642 ND.flags = ND_IFINFO(ifp)->flags;
1643 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1644 ND.chlim = ND_IFINFO(ifp)->chlim;
1645 break;
1646 case SIOCGIFINFO_IN6:
1647 ND = *ND_IFINFO(ifp);
1648 break;
1649 case SIOCSIFINFO_IN6:
1650 /*
1651 * used to change host variables from userland.
1652 * intended for a use on router to reflect RA configurations.
1653 */
1654 /* 0 means 'unspecified' */
1655 if (ND.linkmtu != 0) {
1656 if (ND.linkmtu < IPV6_MMTU ||
1657 ND.linkmtu > IN6_LINKMTU(ifp)) {
1658 error = EINVAL;
1659 break;
1660 }
1661 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1662 }
1663
1664 if (ND.basereachable != 0) {
1665 int obasereachable = ND_IFINFO(ifp)->basereachable;
1666
1667 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1668 if (ND.basereachable != obasereachable)
1669 ND_IFINFO(ifp)->reachable =
1670 ND_COMPUTE_RTIME(ND.basereachable);
1671 }
1672 if (ND.retrans != 0)
1673 ND_IFINFO(ifp)->retrans = ND.retrans;
1674 if (ND.chlim != 0)
1675 ND_IFINFO(ifp)->chlim = ND.chlim;
1676 /* FALLTHROUGH */
1677 case SIOCSIFINFO_FLAGS:
1678 {
1679 struct ifaddr *ifa;
1680 struct in6_ifaddr *ia;
1681
1682 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1683 !(ND.flags & ND6_IFF_IFDISABLED)) {
1684 /* ifdisabled 1->0 transision */
1685
1686 /*
1687 * If the interface is marked as ND6_IFF_IFDISABLED and
1688 * has an link-local address with IN6_IFF_DUPLICATED,
1689 * do not clear ND6_IFF_IFDISABLED.
1690 * See RFC 4862, Section 5.4.5.
1691 */
1692 NET_EPOCH_ENTER(et);
1693 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1694 if (ifa->ifa_addr->sa_family != AF_INET6)
1695 continue;
1696 ia = (struct in6_ifaddr *)ifa;
1697 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1698 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1699 break;
1700 }
1701 NET_EPOCH_EXIT(et);
1702
1703 if (ifa != NULL) {
1704 /* LLA is duplicated. */
1705 ND.flags |= ND6_IFF_IFDISABLED;
1706 log(LOG_ERR, "Cannot enable an interface"
1707 " with a link-local address marked"
1708 " duplicate.\n");
1709 } else {
1710 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1711 if (ifp->if_flags & IFF_UP)
1712 in6_if_up(ifp);
1713 }
1714 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1715 (ND.flags & ND6_IFF_IFDISABLED)) {
1716 /* ifdisabled 0->1 transision */
1717 /* Mark all IPv6 address as tentative. */
1718
1719 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1720 if (V_ip6_dad_count > 0 &&
1721 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1722 NET_EPOCH_ENTER(et);
1723 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1724 ifa_link) {
1725 if (ifa->ifa_addr->sa_family !=
1726 AF_INET6)
1727 continue;
1728 ia = (struct in6_ifaddr *)ifa;
1729 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1730 }
1731 NET_EPOCH_EXIT(et);
1732 }
1733 }
1734
1735 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1736 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1737 /* auto_linklocal 0->1 transision */
1738
1739 /* If no link-local address on ifp, configure */
1740 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1741 in6_ifattach(ifp, NULL);
1742 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1743 ifp->if_flags & IFF_UP) {
1744 /*
1745 * When the IF already has
1746 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1747 * address is assigned, and IFF_UP, try to
1748 * assign one.
1749 */
1750 NET_EPOCH_ENTER(et);
1751 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1752 ifa_link) {
1753 if (ifa->ifa_addr->sa_family !=
1754 AF_INET6)
1755 continue;
1756 ia = (struct in6_ifaddr *)ifa;
1757 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1758 break;
1759 }
1760 NET_EPOCH_EXIT(et);
1761 if (ifa != NULL)
1762 /* No LLA is configured. */
1763 in6_ifattach(ifp, NULL);
1764 }
1765 }
1766 ND_IFINFO(ifp)->flags = ND.flags;
1767 break;
1768 }
1769 #undef ND
1770 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1771 /* sync kernel routing table with the default router list */
1772 defrouter_reset();
1773 defrouter_select_fib(RT_ALL_FIBS);
1774 break;
1775 case SIOCSPFXFLUSH_IN6:
1776 {
1777 /* flush all the prefix advertised by routers */
1778 struct in6_ifaddr *ia, *ia_next;
1779 struct nd_prefix *pr, *next;
1780 struct nd_prhead prl;
1781
1782 LIST_INIT(&prl);
1783
1784 ND6_WLOCK();
1785 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1786 if (pr->ndpr_raf_ra_derived)
1787 nd6_prefix_unlink(pr, &prl);
1788 }
1789 ND6_WUNLOCK();
1790
1791 while ((pr = LIST_FIRST(&prl)) != NULL) {
1792 LIST_REMOVE(pr, ndpr_entry);
1793 /* XXXRW: in6_ifaddrhead locking. */
1794 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1795 ia_next) {
1796 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1797 continue;
1798
1799 if (ia->ia6_ndpr == pr)
1800 in6_purgeaddr(&ia->ia_ifa);
1801 }
1802 nd6_prefix_del(pr);
1803 }
1804 break;
1805 }
1806 case SIOCSRTRFLUSH_IN6:
1807 {
1808 /* flush all the default routers */
1809
1810 defrouter_reset();
1811 nd6_defrouter_flush_all();
1812 defrouter_select_fib(RT_ALL_FIBS);
1813 break;
1814 }
1815 case SIOCGNBRINFO_IN6:
1816 {
1817 struct llentry *ln;
1818 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1819
1820 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1821 return (error);
1822
1823 NET_EPOCH_ENTER(et);
1824 ln = nd6_lookup(&nb_addr, LLE_SF(AF_INET6, 0), ifp);
1825 NET_EPOCH_EXIT(et);
1826
1827 if (ln == NULL) {
1828 error = EINVAL;
1829 break;
1830 }
1831 nbi->state = ln->ln_state;
1832 nbi->asked = ln->la_asked;
1833 nbi->isrouter = ln->ln_router;
1834 if (ln->la_expire == 0)
1835 nbi->expire = 0;
1836 else
1837 nbi->expire = ln->la_expire + ln->lle_remtime / hz +
1838 (time_second - time_uptime);
1839 LLE_RUNLOCK(ln);
1840 break;
1841 }
1842 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1843 ndif->ifindex = V_nd6_defifindex;
1844 break;
1845 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1846 return (nd6_setdefaultiface(ndif->ifindex));
1847 }
1848 return (error);
1849 }
1850
1851 /*
1852 * Calculates new isRouter value based on provided parameters and
1853 * returns it.
1854 */
1855 static int
1856 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
1857 int ln_router)
1858 {
1859
1860 /*
1861 * ICMP6 type dependent behavior.
1862 *
1863 * NS: clear IsRouter if new entry
1864 * RS: clear IsRouter
1865 * RA: set IsRouter if there's lladdr
1866 * redir: clear IsRouter if new entry
1867 *
1868 * RA case, (1):
1869 * The spec says that we must set IsRouter in the following cases:
1870 * - If lladdr exist, set IsRouter. This means (1-5).
1871 * - If it is old entry (!newentry), set IsRouter. This means (7).
1872 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1873 * A quetion arises for (1) case. (1) case has no lladdr in the
1874 * neighbor cache, this is similar to (6).
1875 * This case is rare but we figured that we MUST NOT set IsRouter.
1876 *
1877 * is_new old_addr new_addr NS RS RA redir
1878 * D R
1879 * 0 n n (1) c ? s
1880 * 0 y n (2) c s s
1881 * 0 n y (3) c s s
1882 * 0 y y (4) c s s
1883 * 0 y y (5) c s s
1884 * 1 -- n (6) c c c s
1885 * 1 -- y (7) c c s c s
1886 *
1887 * (c=clear s=set)
1888 */
1889 switch (type & 0xff) {
1890 case ND_NEIGHBOR_SOLICIT:
1891 /*
1892 * New entry must have is_router flag cleared.
1893 */
1894 if (is_new) /* (6-7) */
1895 ln_router = 0;
1896 break;
1897 case ND_REDIRECT:
1898 /*
1899 * If the icmp is a redirect to a better router, always set the
1900 * is_router flag. Otherwise, if the entry is newly created,
1901 * clear the flag. [RFC 2461, sec 8.3]
1902 */
1903 if (code == ND_REDIRECT_ROUTER)
1904 ln_router = 1;
1905 else {
1906 if (is_new) /* (6-7) */
1907 ln_router = 0;
1908 }
1909 break;
1910 case ND_ROUTER_SOLICIT:
1911 /*
1912 * is_router flag must always be cleared.
1913 */
1914 ln_router = 0;
1915 break;
1916 case ND_ROUTER_ADVERT:
1917 /*
1918 * Mark an entry with lladdr as a router.
1919 */
1920 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */
1921 (is_new && new_addr)) { /* (7) */
1922 ln_router = 1;
1923 }
1924 break;
1925 }
1926
1927 return (ln_router);
1928 }
1929
1930 /*
1931 * Create neighbor cache entry and cache link-layer address,
1932 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1933 *
1934 * type - ICMP6 type
1935 * code - type dependent information
1936 *
1937 */
1938 void
1939 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1940 int lladdrlen, int type, int code)
1941 {
1942 struct llentry *ln = NULL, *ln_tmp;
1943 int is_newentry;
1944 int do_update;
1945 int olladdr;
1946 int llchange;
1947 int flags;
1948 uint16_t router = 0;
1949 struct mbuf *chain = NULL;
1950 u_char linkhdr[LLE_MAX_LINKHDR];
1951 size_t linkhdrsize;
1952 int lladdr_off;
1953
1954 NET_EPOCH_ASSERT();
1955 IF_AFDATA_UNLOCK_ASSERT(ifp);
1956
1957 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1958 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1959
1960 /* nothing must be updated for unspecified address */
1961 if (IN6_IS_ADDR_UNSPECIFIED(from))
1962 return;
1963
1964 /*
1965 * Validation about ifp->if_addrlen and lladdrlen must be done in
1966 * the caller.
1967 *
1968 * XXX If the link does not have link-layer adderss, what should
1969 * we do? (ifp->if_addrlen == 0)
1970 * Spec says nothing in sections for RA, RS and NA. There's small
1971 * description on it in NS section (RFC 2461 7.2.3).
1972 */
1973 flags = lladdr ? LLE_EXCLUSIVE : 0;
1974 ln = nd6_lookup(from, LLE_SF(AF_INET6, flags), ifp);
1975 is_newentry = 0;
1976 if (ln == NULL) {
1977 flags |= LLE_EXCLUSIVE;
1978 ln = nd6_alloc(from, 0, ifp);
1979 if (ln == NULL)
1980 return;
1981
1982 /*
1983 * Since we already know all the data for the new entry,
1984 * fill it before insertion.
1985 */
1986 if (lladdr != NULL) {
1987 linkhdrsize = sizeof(linkhdr);
1988 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
1989 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1990 lltable_free_entry(LLTABLE6(ifp), ln);
1991 return;
1992 }
1993 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
1994 lladdr_off);
1995 }
1996
1997 IF_AFDATA_WLOCK(ifp);
1998 LLE_WLOCK(ln);
1999 /* Prefer any existing lle over newly-created one */
2000 ln_tmp = nd6_lookup(from, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
2001 if (ln_tmp == NULL)
2002 lltable_link_entry(LLTABLE6(ifp), ln);
2003 IF_AFDATA_WUNLOCK(ifp);
2004 if (ln_tmp == NULL) {
2005 /* No existing lle, mark as new entry (6,7) */
2006 is_newentry = 1;
2007 if (lladdr != NULL) { /* (7) */
2008 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2009 EVENTHANDLER_INVOKE(lle_event, ln,
2010 LLENTRY_RESOLVED);
2011 }
2012 } else {
2013 lltable_free_entry(LLTABLE6(ifp), ln);
2014 ln = ln_tmp;
2015 ln_tmp = NULL;
2016 }
2017 }
2018 /* do nothing if static ndp is set */
2019 if ((ln->la_flags & LLE_STATIC)) {
2020 if (flags & LLE_EXCLUSIVE)
2021 LLE_WUNLOCK(ln);
2022 else
2023 LLE_RUNLOCK(ln);
2024 return;
2025 }
2026
2027 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
2028 if (olladdr && lladdr) {
2029 llchange = bcmp(lladdr, ln->ll_addr,
2030 ifp->if_addrlen);
2031 } else if (!olladdr && lladdr)
2032 llchange = 1;
2033 else
2034 llchange = 0;
2035
2036 /*
2037 * newentry olladdr lladdr llchange (*=record)
2038 * 0 n n -- (1)
2039 * 0 y n -- (2)
2040 * 0 n y y (3) * STALE
2041 * 0 y y n (4) *
2042 * 0 y y y (5) * STALE
2043 * 1 -- n -- (6) NOSTATE(= PASSIVE)
2044 * 1 -- y -- (7) * STALE
2045 */
2046
2047 do_update = 0;
2048 if (is_newentry == 0 && llchange != 0) {
2049 do_update = 1; /* (3,5) */
2050
2051 /*
2052 * Record source link-layer address
2053 * XXX is it dependent to ifp->if_type?
2054 */
2055 if (!nd6_try_set_entry_addr(ifp, ln, lladdr)) {
2056 /* Entry was deleted */
2057 LLE_WUNLOCK(ln);
2058 return;
2059 }
2060
2061 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2062
2063 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2064
2065 if (ln->la_hold != NULL)
2066 chain = nd6_grab_holdchain(ln);
2067 }
2068
2069 /* Calculates new router status */
2070 router = nd6_is_router(type, code, is_newentry, olladdr,
2071 lladdr != NULL ? 1 : 0, ln->ln_router);
2072
2073 ln->ln_router = router;
2074 /* Mark non-router redirects with special flag */
2075 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
2076 ln->la_flags |= LLE_REDIRECT;
2077
2078 if (flags & LLE_EXCLUSIVE)
2079 LLE_WUNLOCK(ln);
2080 else
2081 LLE_RUNLOCK(ln);
2082
2083 if (chain != NULL)
2084 nd6_flush_holdchain(ifp, ln, chain);
2085 if (do_update)
2086 nd6_flush_children_holdchain(ifp, ln);
2087
2088 /*
2089 * When the link-layer address of a router changes, select the
2090 * best router again. In particular, when the neighbor entry is newly
2091 * created, it might affect the selection policy.
2092 * Question: can we restrict the first condition to the "is_newentry"
2093 * case?
2094 * XXX: when we hear an RA from a new router with the link-layer
2095 * address option, defrouter_select_fib() is called twice, since
2096 * defrtrlist_update called the function as well. However, I believe
2097 * we can compromise the overhead, since it only happens the first
2098 * time.
2099 * XXX: although defrouter_select_fib() should not have a bad effect
2100 * for those are not autoconfigured hosts, we explicitly avoid such
2101 * cases for safety.
2102 */
2103 if ((do_update || is_newentry) && router &&
2104 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
2105 /*
2106 * guaranteed recursion
2107 */
2108 defrouter_select_fib(ifp->if_fib);
2109 }
2110 }
2111
2112 static void
2113 nd6_slowtimo(void *arg)
2114 {
2115 struct epoch_tracker et;
2116 CURVNET_SET((struct vnet *) arg);
2117 struct nd_ifinfo *nd6if;
2118 struct ifnet *ifp;
2119
2120 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
2121 nd6_slowtimo, curvnet);
2122 NET_EPOCH_ENTER(et);
2123 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2124 if (ifp->if_afdata[AF_INET6] == NULL)
2125 continue;
2126 nd6if = ND_IFINFO(ifp);
2127 if (nd6if->basereachable && /* already initialized */
2128 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
2129 /*
2130 * Since reachable time rarely changes by router
2131 * advertisements, we SHOULD insure that a new random
2132 * value gets recomputed at least once every few hours.
2133 * (RFC 2461, 6.3.4)
2134 */
2135 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
2136 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
2137 }
2138 }
2139 NET_EPOCH_EXIT(et);
2140 CURVNET_RESTORE();
2141 }
2142
2143 struct mbuf *
2144 nd6_grab_holdchain(struct llentry *ln)
2145 {
2146 struct mbuf *chain;
2147
2148 LLE_WLOCK_ASSERT(ln);
2149
2150 chain = ln->la_hold;
2151 ln->la_hold = NULL;
2152 ln->la_numheld = 0;
2153
2154 if (ln->ln_state == ND6_LLINFO_STALE) {
2155 /*
2156 * The first time we send a packet to a
2157 * neighbor whose entry is STALE, we have
2158 * to change the state to DELAY and a sets
2159 * a timer to expire in DELAY_FIRST_PROBE_TIME
2160 * seconds to ensure do neighbor unreachability
2161 * detection on expiration.
2162 * (RFC 2461 7.3.3)
2163 */
2164 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
2165 }
2166
2167 return (chain);
2168 }
2169
2170 int
2171 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
2172 struct sockaddr_in6 *dst, struct route *ro)
2173 {
2174 int error;
2175 int ip6len;
2176 struct ip6_hdr *ip6;
2177 struct m_tag *mtag;
2178
2179 #ifdef MAC
2180 mac_netinet6_nd6_send(ifp, m);
2181 #endif
2182
2183 /*
2184 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
2185 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
2186 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
2187 * to be diverted to user space. When re-injected into the kernel,
2188 * send_output() will directly dispatch them to the outgoing interface.
2189 */
2190 if (send_sendso_input_hook != NULL) {
2191 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
2192 if (mtag != NULL) {
2193 ip6 = mtod(m, struct ip6_hdr *);
2194 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2195 /* Use the SEND socket */
2196 error = send_sendso_input_hook(m, ifp, SND_OUT,
2197 ip6len);
2198 /* -1 == no app on SEND socket */
2199 if (error == 0 || error != -1)
2200 return (error);
2201 }
2202 }
2203
2204 m_clrprotoflags(m); /* Avoid confusing lower layers. */
2205 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
2206 mtod(m, struct ip6_hdr *));
2207
2208 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2209 origifp = ifp;
2210
2211 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro);
2212 return (error);
2213 }
2214
2215 /*
2216 * Lookup link headerfor @sa_dst address. Stores found
2217 * data in @desten buffer. Copy of lle ln_flags can be also
2218 * saved in @pflags if @pflags is non-NULL.
2219 *
2220 * If destination LLE does not exists or lle state modification
2221 * is required, call "slow" version.
2222 *
2223 * Return values:
2224 * - 0 on success (address copied to buffer).
2225 * - EWOULDBLOCK (no local error, but address is still unresolved)
2226 * - other errors (alloc failure, etc)
2227 */
2228 int
2229 nd6_resolve(struct ifnet *ifp, int gw_flags, struct mbuf *m,
2230 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags,
2231 struct llentry **plle)
2232 {
2233 struct llentry *ln = NULL;
2234 const struct sockaddr_in6 *dst6;
2235
2236 NET_EPOCH_ASSERT();
2237
2238 if (pflags != NULL)
2239 *pflags = 0;
2240
2241 dst6 = (const struct sockaddr_in6 *)sa_dst;
2242
2243 /* discard the packet if IPv6 operation is disabled on the interface */
2244 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2245 m_freem(m);
2246 return (ENETDOWN); /* better error? */
2247 }
2248
2249 if (m != NULL && m->m_flags & M_MCAST) {
2250 switch (ifp->if_type) {
2251 case IFT_ETHER:
2252 case IFT_L2VLAN:
2253 case IFT_BRIDGE:
2254 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
2255 desten);
2256 return (0);
2257 default:
2258 m_freem(m);
2259 return (EAFNOSUPPORT);
2260 }
2261 }
2262
2263 int family = gw_flags >> 16;
2264 int lookup_flags = plle ? LLE_EXCLUSIVE : LLE_UNLOCKED;
2265 ln = nd6_lookup(&dst6->sin6_addr, LLE_SF(family, lookup_flags), ifp);
2266 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
2267 /* Entry found, let's copy lle info */
2268 bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
2269 if (pflags != NULL)
2270 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
2271 llentry_provide_feedback(ln);
2272 if (plle) {
2273 LLE_ADDREF(ln);
2274 *plle = ln;
2275 LLE_WUNLOCK(ln);
2276 }
2277 return (0);
2278 } else if (plle && ln)
2279 LLE_WUNLOCK(ln);
2280
2281 return (nd6_resolve_slow(ifp, family, 0, m, dst6, desten, pflags, plle));
2282 }
2283
2284 /*
2285 * Finds or creates a new llentry for @addr and @family.
2286 * Returns wlocked llentry or NULL.
2287 *
2288 *
2289 * Child LLEs.
2290 *
2291 * Do not have their own state machine (gets marked as static)
2292 * settimer bails out for child LLEs just in case.
2293 *
2294 * Locking order: parent lle gets locked first, chen goes the child.
2295 */
2296 static __noinline struct llentry *
2297 nd6_get_llentry(struct ifnet *ifp, const struct in6_addr *addr, int family)
2298 {
2299 struct llentry *child_lle = NULL;
2300 struct llentry *lle, *lle_tmp;
2301
2302 lle = nd6_alloc(addr, 0, ifp);
2303 if (lle != NULL && family != AF_INET6) {
2304 child_lle = nd6_alloc(addr, 0, ifp);
2305 if (child_lle == NULL) {
2306 lltable_free_entry(LLTABLE6(ifp), lle);
2307 return (NULL);
2308 }
2309 child_lle->r_family = family;
2310 child_lle->la_flags |= LLE_CHILD | LLE_STATIC;
2311 child_lle->ln_state = ND6_LLINFO_INCOMPLETE;
2312 }
2313
2314 if (lle == NULL) {
2315 char ip6buf[INET6_ADDRSTRLEN];
2316 log(LOG_DEBUG,
2317 "nd6_get_llentry: can't allocate llinfo for %s "
2318 "(ln=%p)\n",
2319 ip6_sprintf(ip6buf, addr), lle);
2320 return (NULL);
2321 }
2322
2323 IF_AFDATA_WLOCK(ifp);
2324 LLE_WLOCK(lle);
2325 /* Prefer any existing entry over newly-created one */
2326 lle_tmp = nd6_lookup(addr, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
2327 if (lle_tmp == NULL)
2328 lltable_link_entry(LLTABLE6(ifp), lle);
2329 else {
2330 lltable_free_entry(LLTABLE6(ifp), lle);
2331 lle = lle_tmp;
2332 }
2333 if (child_lle != NULL) {
2334 /* Check if child lle for the same family exists */
2335 lle_tmp = llentry_lookup_family(lle, child_lle->r_family);
2336 LLE_WLOCK(child_lle);
2337 if (lle_tmp == NULL) {
2338 /* Attach */
2339 lltable_link_child_entry(lle, child_lle);
2340 } else {
2341 /* child lle already exists, free newly-created one */
2342 lltable_free_entry(LLTABLE6(ifp), child_lle);
2343 child_lle = lle_tmp;
2344 }
2345 LLE_WUNLOCK(lle);
2346 lle = child_lle;
2347 }
2348 IF_AFDATA_WUNLOCK(ifp);
2349 return (lle);
2350 }
2351
2352 /*
2353 * Do L2 address resolution for @sa_dst address. Stores found
2354 * address in @desten buffer. Copy of lle ln_flags can be also
2355 * saved in @pflags if @pflags is non-NULL.
2356 *
2357 * Heavy version.
2358 * Function assume that destination LLE does not exist,
2359 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired.
2360 *
2361 * Set noinline to be dtrace-friendly
2362 */
2363 static __noinline int
2364 nd6_resolve_slow(struct ifnet *ifp, int family, int flags, struct mbuf *m,
2365 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags,
2366 struct llentry **plle)
2367 {
2368 struct llentry *lle = NULL;
2369 struct in6_addr *psrc, src;
2370 int send_ns, ll_len;
2371 char *lladdr;
2372
2373 NET_EPOCH_ASSERT();
2374
2375 /*
2376 * Address resolution or Neighbor Unreachability Detection
2377 * for the next hop.
2378 * At this point, the destination of the packet must be a unicast
2379 * or an anycast address(i.e. not a multicast).
2380 */
2381 lle = nd6_lookup(&dst->sin6_addr, LLE_SF(family, LLE_EXCLUSIVE), ifp);
2382 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
2383 /*
2384 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2385 * the condition below is not very efficient. But we believe
2386 * it is tolerable, because this should be a rare case.
2387 */
2388 lle = nd6_get_llentry(ifp, &dst->sin6_addr, family);
2389 }
2390
2391 if (lle == NULL) {
2392 m_freem(m);
2393 return (ENOBUFS);
2394 }
2395
2396 LLE_WLOCK_ASSERT(lle);
2397
2398 /*
2399 * The first time we send a packet to a neighbor whose entry is
2400 * STALE, we have to change the state to DELAY and a sets a timer to
2401 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2402 * neighbor unreachability detection on expiration.
2403 * (RFC 2461 7.3.3)
2404 */
2405 if ((!(lle->la_flags & LLE_CHILD)) && (lle->ln_state == ND6_LLINFO_STALE))
2406 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
2407
2408 /*
2409 * If the neighbor cache entry has a state other than INCOMPLETE
2410 * (i.e. its link-layer address is already resolved), just
2411 * send the packet.
2412 */
2413 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
2414 if (flags & LLE_ADDRONLY) {
2415 lladdr = lle->ll_addr;
2416 ll_len = ifp->if_addrlen;
2417 } else {
2418 lladdr = lle->r_linkdata;
2419 ll_len = lle->r_hdrlen;
2420 }
2421 bcopy(lladdr, desten, ll_len);
2422 if (pflags != NULL)
2423 *pflags = lle->la_flags;
2424 if (plle) {
2425 LLE_ADDREF(lle);
2426 *plle = lle;
2427 }
2428 LLE_WUNLOCK(lle);
2429 return (0);
2430 }
2431
2432 /*
2433 * There is a neighbor cache entry, but no ethernet address
2434 * response yet. Append this latest packet to the end of the
2435 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen,
2436 * the oldest packet in the queue will be removed.
2437 */
2438 if (m != NULL) {
2439 size_t dropped;
2440
2441 dropped = lltable_append_entry_queue(lle, m, V_nd6_maxqueuelen);
2442 ICMP6STAT_ADD(icp6s_dropped, dropped);
2443 }
2444
2445 /*
2446 * If there has been no NS for the neighbor after entering the
2447 * INCOMPLETE state, send the first solicitation.
2448 * Note that for newly-created lle la_asked will be 0,
2449 * so we will transition from ND6_LLINFO_NOSTATE to
2450 * ND6_LLINFO_INCOMPLETE state here.
2451 */
2452 psrc = NULL;
2453 send_ns = 0;
2454
2455 /* If we have child lle, switch to the parent to send NS */
2456 if (lle->la_flags & LLE_CHILD) {
2457 struct llentry *lle_parent = lle->lle_parent;
2458 LLE_WUNLOCK(lle);
2459 lle = lle_parent;
2460 LLE_WLOCK(lle);
2461 }
2462 if (lle->la_asked == 0) {
2463 lle->la_asked++;
2464 send_ns = 1;
2465 psrc = nd6_llinfo_get_holdsrc(lle, &src);
2466
2467 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
2468 }
2469 LLE_WUNLOCK(lle);
2470 if (send_ns != 0)
2471 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
2472
2473 return (EWOULDBLOCK);
2474 }
2475
2476 /*
2477 * Do L2 address resolution for @sa_dst address. Stores found
2478 * address in @desten buffer. Copy of lle ln_flags can be also
2479 * saved in @pflags if @pflags is non-NULL.
2480 *
2481 * Return values:
2482 * - 0 on success (address copied to buffer).
2483 * - EWOULDBLOCK (no local error, but address is still unresolved)
2484 * - other errors (alloc failure, etc)
2485 */
2486 int
2487 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
2488 char *desten, uint32_t *pflags)
2489 {
2490 int error;
2491
2492 flags |= LLE_ADDRONLY;
2493 error = nd6_resolve_slow(ifp, AF_INET6, flags, NULL,
2494 (const struct sockaddr_in6 *)dst, desten, pflags, NULL);
2495 return (error);
2496 }
2497
2498 int
2499 nd6_flush_holdchain(struct ifnet *ifp, struct llentry *lle, struct mbuf *chain)
2500 {
2501 struct mbuf *m, *m_head;
2502 struct sockaddr_in6 dst6;
2503 int error = 0;
2504
2505 NET_EPOCH_ASSERT();
2506
2507 struct route_in6 ro = {
2508 .ro_prepend = lle->r_linkdata,
2509 .ro_plen = lle->r_hdrlen,
2510 };
2511
2512 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst6);
2513 m_head = chain;
2514
2515 while (m_head) {
2516 m = m_head;
2517 m_head = m_head->m_nextpkt;
2518 m->m_nextpkt = NULL;
2519 error = nd6_output_ifp(ifp, ifp, m, &dst6, (struct route *)&ro);
2520 }
2521
2522 /*
2523 * XXX
2524 * note that intermediate errors are blindly ignored
2525 */
2526 return (error);
2527 }
2528
2529 __noinline void
2530 nd6_flush_children_holdchain(struct ifnet *ifp, struct llentry *lle)
2531 {
2532 struct llentry *child_lle;
2533 struct mbuf *chain;
2534
2535 NET_EPOCH_ASSERT();
2536
2537 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) {
2538 LLE_WLOCK(child_lle);
2539 chain = nd6_grab_holdchain(child_lle);
2540 LLE_WUNLOCK(child_lle);
2541 nd6_flush_holdchain(ifp, child_lle, chain);
2542 }
2543 }
2544
2545 static int
2546 nd6_need_cache(struct ifnet *ifp)
2547 {
2548 /*
2549 * XXX: we currently do not make neighbor cache on any interface
2550 * other than Ethernet and GIF.
2551 *
2552 * RFC2893 says:
2553 * - unidirectional tunnels needs no ND
2554 */
2555 switch (ifp->if_type) {
2556 case IFT_ETHER:
2557 case IFT_IEEE1394:
2558 case IFT_L2VLAN:
2559 case IFT_INFINIBAND:
2560 case IFT_BRIDGE:
2561 case IFT_PROPVIRTUAL:
2562 return (1);
2563 default:
2564 return (0);
2565 }
2566 }
2567
2568 /*
2569 * Add pernament ND6 link-layer record for given
2570 * interface address.
2571 *
2572 * Very similar to IPv4 arp_ifinit(), but:
2573 * 1) IPv6 DAD is performed in different place
2574 * 2) It is called by IPv6 protocol stack in contrast to
2575 * arp_ifinit() which is typically called in SIOCSIFADDR
2576 * driver ioctl handler.
2577 *
2578 */
2579 int
2580 nd6_add_ifa_lle(struct in6_ifaddr *ia)
2581 {
2582 struct ifnet *ifp;
2583 struct llentry *ln, *ln_tmp;
2584 struct sockaddr *dst;
2585
2586 ifp = ia->ia_ifa.ifa_ifp;
2587 if (nd6_need_cache(ifp) == 0)
2588 return (0);
2589
2590 dst = (struct sockaddr *)&ia->ia_addr;
2591 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
2592 if (ln == NULL)
2593 return (ENOBUFS);
2594
2595 IF_AFDATA_WLOCK(ifp);
2596 LLE_WLOCK(ln);
2597 /* Unlink any entry if exists */
2598 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_SF(AF_INET6, LLE_EXCLUSIVE), dst);
2599 if (ln_tmp != NULL)
2600 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
2601 lltable_link_entry(LLTABLE6(ifp), ln);
2602 IF_AFDATA_WUNLOCK(ifp);
2603
2604 if (ln_tmp != NULL)
2605 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
2606 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2607
2608 LLE_WUNLOCK(ln);
2609 if (ln_tmp != NULL)
2610 llentry_free(ln_tmp);
2611
2612 return (0);
2613 }
2614
2615 /*
2616 * Removes either all lle entries for given @ia, or lle
2617 * corresponding to @ia address.
2618 */
2619 void
2620 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
2621 {
2622 struct sockaddr_in6 mask, addr;
2623 struct sockaddr *saddr, *smask;
2624 struct ifnet *ifp;
2625
2626 ifp = ia->ia_ifa.ifa_ifp;
2627 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
2628 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
2629 saddr = (struct sockaddr *)&addr;
2630 smask = (struct sockaddr *)&mask;
2631
2632 if (all != 0)
2633 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
2634 else
2635 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
2636 }
2637
2638 static int
2639 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2640 {
2641 struct in6_prefix p;
2642 struct sockaddr_in6 s6;
2643 struct nd_prefix *pr;
2644 struct nd_pfxrouter *pfr;
2645 time_t maxexpire;
2646 int error;
2647 char ip6buf[INET6_ADDRSTRLEN];
2648
2649 if (req->newptr)
2650 return (EPERM);
2651
2652 error = sysctl_wire_old_buffer(req, 0);
2653 if (error != 0)
2654 return (error);
2655
2656 bzero(&p, sizeof(p));
2657 p.origin = PR_ORIG_RA;
2658 bzero(&s6, sizeof(s6));
2659 s6.sin6_family = AF_INET6;
2660 s6.sin6_len = sizeof(s6);
2661
2662 ND6_RLOCK();
2663 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2664 if (!pr->ndpr_raf_ra_derived)
2665 continue;
2666 p.prefix = pr->ndpr_prefix;
2667 if (sa6_recoverscope(&p.prefix)) {
2668 log(LOG_ERR, "scope error in prefix list (%s)\n",
2669 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2670 /* XXX: press on... */
2671 }
2672 p.raflags = pr->ndpr_raf;
2673 p.prefixlen = pr->ndpr_plen;
2674 p.vltime = pr->ndpr_vltime;
2675 p.pltime = pr->ndpr_pltime;
2676 p.if_index = pr->ndpr_ifp->if_index;
2677 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2678 p.expire = 0;
2679 else {
2680 /* XXX: we assume time_t is signed. */
2681 maxexpire = (-1) &
2682 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2683 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2684 p.expire = pr->ndpr_lastupdate +
2685 pr->ndpr_vltime +
2686 (time_second - time_uptime);
2687 else
2688 p.expire = maxexpire;
2689 }
2690 p.refcnt = pr->ndpr_addrcnt;
2691 p.flags = pr->ndpr_stateflags;
2692 p.advrtrs = 0;
2693 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2694 p.advrtrs++;
2695 error = SYSCTL_OUT(req, &p, sizeof(p));
2696 if (error != 0)
2697 break;
2698 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2699 s6.sin6_addr = pfr->router->rtaddr;
2700 if (sa6_recoverscope(&s6))
2701 log(LOG_ERR,
2702 "scope error in prefix list (%s)\n",
2703 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2704 error = SYSCTL_OUT(req, &s6, sizeof(s6));
2705 if (error != 0)
2706 goto out;
2707 }
2708 }
2709 out:
2710 ND6_RUNLOCK();
2711 return (error);
2712 }
2713 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2714 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2715 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix",
2716 "NDP prefix list");
2717 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2718 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2719 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2720 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
Cache object: d49796097b5833869354fe9778ef1c90
|