2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_inet6.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/callout.h>
44 #include <sys/malloc.h>
46 #include <sys/mutex.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
50 #include <sys/kernel.h>
51 #include <sys/protosw.h>
52 #include <sys/errno.h>
53 #include <sys/syslog.h>
54 #include <sys/rwlock.h>
55 #include <sys/queue.h>
57 #include <sys/sysctl.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <net/if_llatbl.h>
69 #include <netinet/if_ether.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet6/in6_ifattach.h>
76 #include <netinet/icmp6.h>
77 #include <netinet6/send.h>
79 #include <sys/limits.h>
81 #include <security/mac/mac_framework.h>
83 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
84 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
86 #define SIN6(s) ((const struct sockaddr_in6 *)(s))
88 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery");
91 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
92 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
93 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
94 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
95 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
97 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
100 /* preventing too many loops in ND option parsing */
101 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
103 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
105 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
107 #define V_nd6_maxndopt VNET(nd6_maxndopt)
108 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
111 VNET_DEFINE(int, nd6_debug) = 1;
113 VNET_DEFINE(int, nd6_debug) = 0;
116 static eventhandler_tag lle_event_eh, iflladdr_event_eh, ifnet_link_event_eh;
118 VNET_DEFINE(struct nd_drhead, nd_defrouter);
119 VNET_DEFINE(struct nd_prhead, nd_prefix);
120 VNET_DEFINE(struct rwlock, nd6_lock);
121 VNET_DEFINE(uint64_t, nd6_list_genid);
122 VNET_DEFINE(struct mtx, nd6_onlink_mtx);
124 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
125 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
127 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
129 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *,
131 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
132 static void nd6_slowtimo(void *);
133 static int regen_tmpaddr(struct in6_ifaddr *);
134 static void nd6_free(struct llentry **, int);
135 static void nd6_free_redirect(const struct llentry *);
136 static void nd6_llinfo_timer(void *);
137 static void nd6_llinfo_settimer_locked(struct llentry *, long);
138 static void clear_llinfo_pqueue(struct llentry *);
139 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
140 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *,
141 const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **);
142 static int nd6_need_cache(struct ifnet *);
145 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch);
146 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
148 VNET_DEFINE(struct callout, nd6_timer_ch);
149 #define V_nd6_timer_ch VNET(nd6_timer_ch)
152 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
154 struct rt_addrinfo rtinfo;
155 struct sockaddr_in6 dst;
156 struct sockaddr_dl gw;
161 LLE_WLOCK_ASSERT(lle);
163 if (lltable_get_af(lle->lle_tbl) != AF_INET6)
167 case LLENTRY_RESOLVED:
169 KASSERT(lle->la_flags & LLE_VALID,
170 ("%s: %p resolved but not valid?", __func__, lle));
172 case LLENTRY_EXPIRED:
179 ifp = lltable_get_ifp(lle->lle_tbl);
181 bzero(&dst, sizeof(dst));
182 bzero(&gw, sizeof(gw));
183 bzero(&rtinfo, sizeof(rtinfo));
184 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst);
185 dst.sin6_scope_id = in6_getscopezone(ifp,
186 in6_addrscope(&dst.sin6_addr));
187 gw.sdl_len = sizeof(struct sockaddr_dl);
188 gw.sdl_family = AF_LINK;
189 gw.sdl_alen = ifp->if_addrlen;
190 gw.sdl_index = ifp->if_index;
191 gw.sdl_type = ifp->if_type;
192 if (evt == LLENTRY_RESOLVED)
193 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
194 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
195 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
196 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
197 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib;
198 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | (
199 type == RTM_ADD ? RTF_UP: 0), 0, fibnum);
203 * A handler for interface link layer address change event.
206 nd6_iflladdr(void *arg __unused, struct ifnet *ifp)
209 lltable_update_ifaddr(LLTABLE6(ifp));
216 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF);
217 rw_init(&V_nd6_lock, "nd6 list");
219 LIST_INIT(&V_nd_prefix);
220 TAILQ_INIT(&V_nd_defrouter);
223 callout_init(&V_nd6_slowtimo_ch, 0);
224 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
225 nd6_slowtimo, curvnet);
227 callout_init(&V_nd6_timer_ch, 0);
228 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
231 if (IS_DEFAULT_VNET(curvnet)) {
232 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
233 NULL, EVENTHANDLER_PRI_ANY);
234 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event,
235 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
236 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event,
237 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY);
246 callout_drain(&V_nd6_slowtimo_ch);
247 callout_drain(&V_nd6_timer_ch);
248 if (IS_DEFAULT_VNET(curvnet)) {
249 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh);
250 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
251 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
253 rw_destroy(&V_nd6_lock);
254 mtx_destroy(&V_nd6_onlink_mtx);
259 nd6_ifattach(struct ifnet *ifp)
261 struct nd_ifinfo *nd;
263 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
266 nd->chlim = IPV6_DEFHLIM;
267 nd->basereachable = REACHABLE_TIME;
268 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
269 nd->retrans = RETRANS_TIMER;
271 nd->flags = ND6_IFF_PERFORMNUD;
273 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
274 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
275 * default regardless of the V_ip6_auto_linklocal configuration to
276 * give a reasonable default behavior.
278 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
279 (ifp->if_flags & IFF_LOOPBACK))
280 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
282 * A loopback interface does not need to accept RTADV.
283 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
284 * default regardless of the V_ip6_accept_rtadv configuration to
285 * prevent the interface from accepting RA messages arrived
286 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
288 if (V_ip6_accept_rtadv &&
289 !(ifp->if_flags & IFF_LOOPBACK) &&
290 (ifp->if_type != IFT_BRIDGE))
291 nd->flags |= ND6_IFF_ACCEPT_RTADV;
292 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
293 nd->flags |= ND6_IFF_NO_RADR;
295 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
296 nd6_setmtu0(ifp, nd);
302 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd)
304 struct epoch_tracker et;
305 struct ifaddr *ifa, *next;
308 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
309 if (ifa->ifa_addr->sa_family != AF_INET6)
312 /* stop DAD processing */
321 * Reset ND level link MTU. This function is called when the physical MTU
322 * changes, which means we might have to adjust the ND level MTU.
325 nd6_setmtu(struct ifnet *ifp)
327 if (ifp->if_afdata[AF_INET6] == NULL)
330 nd6_setmtu0(ifp, ND_IFINFO(ifp));
333 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
335 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
339 omaxmtu = ndi->maxmtu;
340 ndi->maxmtu = ifp->if_mtu;
343 * Decreasing the interface MTU under IPV6 minimum MTU may cause
344 * undesirable situation. We thus notify the operator of the change
345 * explicitly. The check for omaxmtu is necessary to restrict the
346 * log to the case of changing the MTU, not initializing it.
348 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
349 log(LOG_NOTICE, "nd6_setmtu0: "
350 "new link MTU on %s (%lu) is too small for IPv6\n",
351 if_name(ifp), (unsigned long)ndi->maxmtu);
354 if (ndi->maxmtu > V_in6_maxmtu)
355 in6_setmaxmtu(); /* check all interfaces just in case */
360 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
363 bzero(ndopts, sizeof(*ndopts));
364 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
366 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
369 ndopts->nd_opts_done = 1;
370 ndopts->nd_opts_search = NULL;
375 * Take one ND option.
378 nd6_option(union nd_opts *ndopts)
380 struct nd_opt_hdr *nd_opt;
383 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
384 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
386 if (ndopts->nd_opts_search == NULL)
388 if (ndopts->nd_opts_done)
391 nd_opt = ndopts->nd_opts_search;
393 /* make sure nd_opt_len is inside the buffer */
394 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
395 bzero(ndopts, sizeof(*ndopts));
399 olen = nd_opt->nd_opt_len << 3;
402 * Message validation requires that all included
403 * options have a length that is greater than zero.
405 bzero(ndopts, sizeof(*ndopts));
409 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
410 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
411 /* option overruns the end of buffer, invalid */
412 bzero(ndopts, sizeof(*ndopts));
414 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
415 /* reached the end of options chain */
416 ndopts->nd_opts_done = 1;
417 ndopts->nd_opts_search = NULL;
423 * Parse multiple ND options.
424 * This function is much easier to use, for ND routines that do not need
425 * multiple options of the same type.
428 nd6_options(union nd_opts *ndopts)
430 struct nd_opt_hdr *nd_opt;
433 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
434 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
436 if (ndopts->nd_opts_search == NULL)
440 nd_opt = nd6_option(ndopts);
441 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
443 * Message validation requires that all included
444 * options have a length that is greater than zero.
446 ICMP6STAT_INC(icp6s_nd_badopt);
447 bzero(ndopts, sizeof(*ndopts));
454 switch (nd_opt->nd_opt_type) {
455 case ND_OPT_SOURCE_LINKADDR:
456 case ND_OPT_TARGET_LINKADDR:
458 case ND_OPT_REDIRECTED_HEADER:
460 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
462 "duplicated ND6 option found (type=%d)\n",
463 nd_opt->nd_opt_type));
466 ndopts->nd_opt_array[nd_opt->nd_opt_type]
470 case ND_OPT_PREFIX_INFORMATION:
471 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
472 ndopts->nd_opt_array[nd_opt->nd_opt_type]
475 ndopts->nd_opts_pi_end =
476 (struct nd_opt_prefix_info *)nd_opt;
478 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
479 case ND_OPT_RDNSS: /* RFC 6106 */
480 case ND_OPT_DNSSL: /* RFC 6106 */
482 * Silently ignore options we know and do not care about
488 * Unknown options must be silently ignored,
489 * to accommodate future extension to the protocol.
492 "nd6_options: unsupported option %d - "
493 "option ignored\n", nd_opt->nd_opt_type));
498 if (i > V_nd6_maxndopt) {
499 ICMP6STAT_INC(icp6s_nd_toomanyopt);
500 nd6log((LOG_INFO, "too many loop in nd opt\n"));
504 if (ndopts->nd_opts_done)
512 * ND6 timer routine to handle ND6 entries
515 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
519 LLE_WLOCK_ASSERT(ln);
524 canceled = callout_stop(&ln->lle_timer);
526 ln->la_expire = time_uptime + tick / hz;
528 if (tick > INT_MAX) {
529 ln->ln_ntick = tick - INT_MAX;
530 canceled = callout_reset(&ln->lle_timer, INT_MAX,
531 nd6_llinfo_timer, ln);
534 canceled = callout_reset(&ln->lle_timer, tick,
535 nd6_llinfo_timer, ln);
543 * Gets source address of the first packet in hold queue
544 * and stores it in @src.
545 * Returns pointer to @src (if hold queue is not empty) or NULL.
547 * Set noinline to be dtrace-friendly
549 static __noinline struct in6_addr *
550 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
555 if (ln->la_hold == NULL)
559 * assume every packet in la_hold has the same IP header
562 if (sizeof(hdr) > m->m_len)
565 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
572 * Checks if we need to switch from STALE state.
574 * RFC 4861 requires switching from STALE to DELAY state
575 * on first packet matching entry, waiting V_nd6_delay and
576 * transition to PROBE state (if upper layer confirmation was
579 * This code performs a bit differently:
580 * On packet hit we don't change state (but desired state
581 * can be guessed by control plane). However, after V_nd6_delay
582 * seconds code will transition to PROBE state (so DELAY state
583 * is kinda skipped in most situations).
585 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so
586 * we perform the following upon entering STALE state:
588 * 1) Arm timer to run each V_nd6_delay seconds to make sure that
589 * if packet was transmitted at the start of given interval, we
590 * would be able to switch to PROBE state in V_nd6_delay seconds
593 * 2) Reschedule timer until original V_nd6_gctimer expires keeping
594 * lle in STALE state (remaining timer value stored in lle_remtime).
596 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay
599 * Returns non-zero value if the entry is still STALE (storing
600 * the next timer interval in @pdelay).
602 * Returns zero value if original timer expired or we need to switch to
603 * PROBE (store that in @do_switch variable).
606 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch)
608 int nd_delay, nd_gctimer, r_skip_req;
613 nd_gctimer = V_nd6_gctimer;
614 nd_delay = V_nd6_delay;
617 r_skip_req = lle->r_skip_req;
618 lle_hittime = lle->lle_hittime;
621 if (r_skip_req > 0) {
624 * Nonzero r_skip_req value was set upon entering
625 * STALE state. Since value was not changed, no
626 * packets were passed using this lle. Ask for
627 * timer reschedule and keep STALE state.
629 delay = (long)(MIN(nd_gctimer, nd_delay));
631 if (lle->lle_remtime > delay)
632 lle->lle_remtime -= delay;
634 delay = lle->lle_remtime;
635 lle->lle_remtime = 0;
641 * The original ng6_gctime timeout ended,
642 * no more rescheduling.
652 * Packet received. Verify timestamp
654 delay = (long)(time_uptime - lle_hittime);
655 if (delay < nd_delay) {
658 * V_nd6_delay still not passed since the first
659 * hit in STALE state.
660 * Reshedule timer and return.
662 *pdelay = (long)(nd_delay - delay) * hz;
666 /* Request switching to probe */
673 * Switch @lle state to new state optionally arming timers.
675 * Set noinline to be dtrace-friendly
678 nd6_llinfo_setstate(struct llentry *lle, int newstate)
681 int nd_gctimer, nd_delay;
688 case ND6_LLINFO_INCOMPLETE:
689 ifp = lle->lle_tbl->llt_ifp;
690 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
692 case ND6_LLINFO_REACHABLE:
693 if (!ND6_LLINFO_PERMANENT(lle)) {
694 ifp = lle->lle_tbl->llt_ifp;
695 delay = (long)ND_IFINFO(ifp)->reachable * hz;
698 case ND6_LLINFO_STALE:
701 * Notify fast path that we want to know if any packet
702 * is transmitted by setting r_skip_req.
707 nd_delay = V_nd6_delay;
708 nd_gctimer = V_nd6_gctimer;
710 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz;
711 remtime = (long)nd_gctimer * hz - delay;
713 case ND6_LLINFO_DELAY:
715 delay = (long)V_nd6_delay * hz;
720 nd6_llinfo_settimer_locked(lle, delay);
722 lle->lle_remtime = remtime;
723 lle->ln_state = newstate;
727 * Timer-dependent part of nd state machine.
729 * Set noinline to be dtrace-friendly
731 static __noinline void
732 nd6_llinfo_timer(void *arg)
735 struct in6_addr *dst, *pdst, *psrc, src;
737 struct nd_ifinfo *ndi;
738 int do_switch, send_ns;
741 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
742 ln = (struct llentry *)arg;
743 ifp = lltable_get_ifp(ln->lle_tbl);
744 CURVNET_SET(ifp->if_vnet);
748 if (callout_pending(&ln->lle_timer)) {
750 * Here we are a bit odd here in the treatment of
751 * active/pending. If the pending bit is set, it got
752 * rescheduled before I ran. The active
753 * bit we ignore, since if it was stopped
754 * in ll_tablefree() and was currently running
755 * it would have return 0 so the code would
756 * not have deleted it since the callout could
757 * not be stopped so we want to go through
758 * with the delete here now. If the callout
759 * was restarted, the pending bit will be back on and
760 * we just want to bail since the callout_reset would
761 * return 1 and our reference would have been removed
762 * by nd6_llinfo_settimer_locked above since canceled
770 ndi = ND_IFINFO(ifp);
772 dst = &ln->r_l3addr.addr6;
775 if (ln->ln_ntick > 0) {
776 if (ln->ln_ntick > INT_MAX) {
777 ln->ln_ntick -= INT_MAX;
778 nd6_llinfo_settimer_locked(ln, INT_MAX);
781 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
786 if (ln->la_flags & LLE_STATIC) {
790 if (ln->la_flags & LLE_DELETED) {
795 switch (ln->ln_state) {
796 case ND6_LLINFO_INCOMPLETE:
797 if (ln->la_asked < V_nd6_mmaxtries) {
800 /* Send NS to multicast address */
803 struct mbuf *m = ln->la_hold;
808 * assuming every packet in la_hold has the
809 * same IP header. Send error after unlock.
814 clear_llinfo_pqueue(ln);
818 icmp6_error2(m, ICMP6_DST_UNREACH,
819 ICMP6_DST_UNREACH_ADDR, 0, ifp);
822 case ND6_LLINFO_REACHABLE:
823 if (!ND6_LLINFO_PERMANENT(ln))
824 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
827 case ND6_LLINFO_STALE:
828 if (nd6_is_stale(ln, &delay, &do_switch) != 0) {
831 * No packet has used this entry and GC timeout
832 * has not been passed. Reshedule timer and
835 nd6_llinfo_settimer_locked(ln, delay);
839 if (do_switch == 0) {
842 * GC timer has ended and entry hasn't been used.
843 * Run Garbage collector (RFC 4861, 5.3)
845 if (!ND6_LLINFO_PERMANENT(ln))
850 /* Entry has been used AND delay timer has ended. */
854 case ND6_LLINFO_DELAY:
855 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
858 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
861 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
863 case ND6_LLINFO_PROBE:
864 if (ln->la_asked < V_nd6_umaxtries) {
872 panic("%s: paths in a dark night can be confusing: %d",
873 __func__, ln->ln_state);
879 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
880 psrc = nd6_llinfo_get_holdsrc(ln, &src);
883 nd6_ns_output(ifp, psrc, pdst, dst, NULL);
893 * ND6 timer routine to expire default route list and prefix list
898 CURVNET_SET((struct vnet *) arg);
899 struct nd_drhead drq;
900 struct nd_prhead prl;
901 struct nd_defrouter *dr, *ndr;
902 struct nd_prefix *pr, *npr;
904 struct in6_ifaddr *ia6, *nia6;
911 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr)
912 if (dr->expire && dr->expire < time_uptime)
913 defrouter_unlink(dr, &drq);
916 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
917 TAILQ_REMOVE(&drq, dr, dr_entry);
922 * expire interface addresses.
923 * in the past the loop was inside prefix expiry processing.
924 * However, from a stricter speci-confrmance standpoint, we should
925 * rather separate address lifetimes and prefix lifetimes.
927 * XXXRW: in6_ifaddrhead locking.
930 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
931 /* check address lifetime */
932 if (IFA6_IS_INVALID(ia6)) {
936 * If the expiring address is temporary, try
937 * regenerating a new one. This would be useful when
938 * we suspended a laptop PC, then turned it on after a
939 * period that could invalidate all temporary
940 * addresses. Although we may have to restart the
941 * loop (see below), it must be after purging the
942 * address. Otherwise, we'd see an infinite loop of
945 if (V_ip6_use_tempaddr &&
946 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
947 if (regen_tmpaddr(ia6) == 0)
951 in6_purgeaddr(&ia6->ia_ifa);
954 goto addrloop; /* XXX: see below */
955 } else if (IFA6_IS_DEPRECATED(ia6)) {
956 int oldflags = ia6->ia6_flags;
958 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
961 * If a temporary address has just become deprecated,
962 * regenerate a new one if possible.
964 if (V_ip6_use_tempaddr &&
965 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
966 (oldflags & IN6_IFF_DEPRECATED) == 0) {
968 if (regen_tmpaddr(ia6) == 0) {
970 * A new temporary address is
972 * XXX: this means the address chain
973 * has changed while we are still in
974 * the loop. Although the change
975 * would not cause disaster (because
976 * it's not a deletion, but an
977 * addition,) we'd rather restart the
978 * loop just for safety. Or does this
979 * significantly reduce performance??
984 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
986 * Schedule DAD for a tentative address. This happens
987 * if the interface was down or not running
988 * when the address was configured.
992 delay = arc4random() %
993 (MAX_RTR_SOLICITATION_DELAY * hz);
994 nd6_dad_start((struct ifaddr *)ia6, delay);
997 * Check status of the interface. If it is down,
998 * mark the address as tentative for future DAD.
1001 if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 &&
1002 ((ifp->if_flags & IFF_UP) == 0 ||
1003 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1004 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){
1005 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
1006 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
1010 * A new RA might have made a deprecated address
1013 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
1019 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1021 * Expire prefixes. Since the pltime is only used for
1022 * autoconfigured addresses, pltime processing for prefixes is
1025 * Only unlink after all derived addresses have expired. This
1026 * may not occur until two hours after the prefix has expired
1027 * per RFC 4862. If the prefix expires before its derived
1028 * addresses, mark it off-link. This will be done automatically
1029 * after unlinking if no address references remain.
1031 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME ||
1032 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime)
1035 if (pr->ndpr_addrcnt == 0) {
1036 nd6_prefix_unlink(pr, &prl);
1039 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1040 genid = V_nd6_list_genid;
1044 (void)nd6_prefix_offlink(pr);
1045 ND6_ONLINK_UNLOCK();
1047 nd6_prefix_rele(pr);
1048 if (genid != V_nd6_list_genid)
1054 while ((pr = LIST_FIRST(&prl)) != NULL) {
1055 LIST_REMOVE(pr, ndpr_entry);
1059 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
1060 nd6_timer, curvnet);
1066 * ia6 - deprecated/invalidated temporary address
1069 regen_tmpaddr(struct in6_ifaddr *ia6)
1071 struct epoch_tracker et;
1074 struct in6_ifaddr *public_ifa6 = NULL;
1076 ifp = ia6->ia_ifa.ifa_ifp;
1077 NET_EPOCH_ENTER(et);
1078 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1079 struct in6_ifaddr *it6;
1081 if (ifa->ifa_addr->sa_family != AF_INET6)
1084 it6 = (struct in6_ifaddr *)ifa;
1086 /* ignore no autoconf addresses. */
1087 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1090 /* ignore autoconf addresses with different prefixes. */
1091 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
1095 * Now we are looking at an autoconf address with the same
1096 * prefix as ours. If the address is temporary and is still
1097 * preferred, do not create another one. It would be rare, but
1098 * could happen, for example, when we resume a laptop PC after
1101 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1102 !IFA6_IS_DEPRECATED(it6)) {
1108 * This is a public autoconf address that has the same prefix
1109 * as ours. If it is preferred, keep it. We can't break the
1110 * loop here, because there may be a still-preferred temporary
1111 * address with the prefix.
1113 if (!IFA6_IS_DEPRECATED(it6))
1116 if (public_ifa6 != NULL)
1117 ifa_ref(&public_ifa6->ia_ifa);
1120 if (public_ifa6 != NULL) {
1123 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
1124 ifa_free(&public_ifa6->ia_ifa);
1125 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
1126 " tmp addr,errno=%d\n", e);
1129 ifa_free(&public_ifa6->ia_ifa);
1137 * Remove prefix and default router list entries corresponding to ifp. Neighbor
1138 * cache entries are freed in in6_domifdetach().
1141 nd6_purge(struct ifnet *ifp)
1143 struct nd_drhead drq;
1144 struct nd_prhead prl;
1145 struct nd_defrouter *dr, *ndr;
1146 struct nd_prefix *pr, *npr;
1152 * Nuke default router list entries toward ifp.
1153 * We defer removal of default router list entries that is installed
1154 * in the routing table, in order to keep additional side effects as
1155 * small as possible.
1158 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
1162 defrouter_unlink(dr, &drq);
1164 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
1168 defrouter_unlink(dr, &drq);
1172 * Remove prefixes on ifp. We should have already removed addresses on
1173 * this interface, so no addresses should be referencing these prefixes.
1175 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1176 if (pr->ndpr_ifp == ifp)
1177 nd6_prefix_unlink(pr, &prl);
1181 /* Delete the unlinked router and prefix objects. */
1182 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
1183 TAILQ_REMOVE(&drq, dr, dr_entry);
1186 while ((pr = LIST_FIRST(&prl)) != NULL) {
1187 LIST_REMOVE(pr, ndpr_entry);
1191 /* cancel default outgoing interface setting */
1192 if (V_nd6_defifindex == ifp->if_index)
1193 nd6_setdefaultiface(0);
1195 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1196 /* Refresh default router list. */
1197 defrouter_select_fib(ifp->if_fib);
1202 * the caller acquires and releases the lock on the lltbls
1203 * Returns the llentry locked
1206 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1208 struct sockaddr_in6 sin6;
1211 bzero(&sin6, sizeof(sin6));
1212 sin6.sin6_len = sizeof(struct sockaddr_in6);
1213 sin6.sin6_family = AF_INET6;
1214 sin6.sin6_addr = *addr6;
1216 IF_AFDATA_LOCK_ASSERT(ifp);
1218 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6);
1224 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1226 struct sockaddr_in6 sin6;
1229 bzero(&sin6, sizeof(sin6));
1230 sin6.sin6_len = sizeof(struct sockaddr_in6);
1231 sin6.sin6_family = AF_INET6;
1232 sin6.sin6_addr = *addr6;
1234 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
1236 ln->ln_state = ND6_LLINFO_NOSTATE;
1242 * Test whether a given IPv6 address is a neighbor or not, ignoring
1243 * the actual neighbor cache. The neighbor cache is ignored in order
1244 * to not reenter the routing code from within itself.
1247 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1249 struct nd_prefix *pr;
1251 struct rt_addrinfo info;
1252 struct sockaddr_in6 rt_key;
1253 const struct sockaddr *dst6;
1258 * A link-local address is always a neighbor.
1259 * XXX: a link does not necessarily specify a single interface.
1261 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
1262 struct sockaddr_in6 sin6_copy;
1266 * We need sin6_copy since sa6_recoverscope() may modify the
1270 if (sa6_recoverscope(&sin6_copy))
1271 return (0); /* XXX: should be impossible */
1272 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
1274 if (sin6_copy.sin6_scope_id == zone)
1280 bzero(&rt_key, sizeof(rt_key));
1281 bzero(&info, sizeof(info));
1282 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1285 * If the address matches one of our addresses,
1286 * it should be a neighbor.
1287 * If the address matches one of our on-link prefixes, it should be a
1292 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1293 if (pr->ndpr_ifp != ifp)
1296 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1297 dst6 = (const struct sockaddr *)&pr->ndpr_prefix;
1300 * We only need to check all FIBs if add_addr_allfibs
1301 * is unset. If set, checking any FIB will suffice.
1303 fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0;
1304 for (; fibnum < rt_numfibs; fibnum++) {
1305 genid = V_nd6_list_genid;
1309 * Restore length field before
1312 rt_key.sin6_len = sizeof(rt_key);
1313 error = rib_lookup_info(fibnum, dst6, 0, 0,
1317 if (genid != V_nd6_list_genid)
1326 * This is the case where multiple interfaces
1327 * have the same prefix, but only one is installed
1328 * into the routing table and that prefix entry
1329 * is not the one being examined here. In the case
1330 * where RADIX_MPATH is enabled, multiple route
1331 * entries (of the same rt_key value) will be
1332 * installed because the interface addresses all
1335 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1340 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1341 &addr->sin6_addr, &pr->ndpr_mask)) {
1349 * If the address is assigned on the node of the other side of
1350 * a p2p interface, the address should be a neighbor.
1352 if (ifp->if_flags & IFF_POINTOPOINT) {
1353 struct epoch_tracker et;
1355 NET_EPOCH_ENTER(et);
1356 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1357 if (ifa->ifa_addr->sa_family != addr->sin6_family)
1359 if (ifa->ifa_dstaddr != NULL &&
1360 sa_equal(addr, ifa->ifa_dstaddr)) {
1369 * If the default router list is empty, all addresses are regarded
1370 * as on-link, and thus, as a neighbor.
1372 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1373 TAILQ_EMPTY(&V_nd_defrouter) &&
1374 V_nd6_defifindex == ifp->if_index) {
1383 * Detect if a given IPv6 address identifies a neighbor on a given link.
1384 * XXX: should take care of the destination of a p2p link?
1387 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1389 struct epoch_tracker et;
1390 struct llentry *lle;
1393 IF_AFDATA_UNLOCK_ASSERT(ifp);
1394 if (nd6_is_new_addr_neighbor(addr, ifp))
1398 * Even if the address matches none of our addresses, it might be
1399 * in the neighbor cache.
1401 NET_EPOCH_ENTER(et);
1402 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1411 * Free an nd6 llinfo entry.
1412 * Since the function would cause significant changes in the kernel, DO NOT
1413 * make it global, unless you have a strong reason for the change, and are sure
1414 * that the change is safe.
1416 * Set noinline to be dtrace-friendly
1418 static __noinline void
1419 nd6_free(struct llentry **lnp, int gc)
1423 struct nd_defrouter *dr;
1428 LLE_WLOCK_ASSERT(ln);
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);
1438 if ((ln->la_flags & LLE_DELETED) == 0)
1439 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
1442 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1443 * even though it is not harmful, it was not really necessary.
1447 nd6_llinfo_settimer_locked(ln, -1);
1449 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1450 if (dr != NULL && dr->expire &&
1451 ln->ln_state == ND6_LLINFO_STALE && gc) {
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
1461 * XXX: the check for ln_state would be redundant,
1462 * but we intentionally keep it just in case.
1464 if (dr->expire > time_uptime)
1465 nd6_llinfo_settimer_locked(ln,
1466 (dr->expire - time_uptime) * hz);
1468 nd6_llinfo_settimer_locked(ln,
1469 (long)V_nd6_gctimer * hz);
1479 * Unreachablity of a router might affect the default
1480 * router selection and on-link detection of advertised
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.
1491 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1494 if (ln->ln_router || dr) {
1497 * We need to unlock to avoid a LOR with rt6_flush() with the
1498 * rnh and for the calls to pfxlist_onlink_check() and
1499 * defrouter_select_fib() in the block further down for calls
1500 * into nd6_lookup(). We still hold a ref.
1505 * rt6_flush must be called whether or not the neighbor
1506 * is in the Default Router List.
1507 * See a corresponding comment in nd6_na_input().
1509 rt6_flush(&ln->r_l3addr.addr6, ifp);
1514 * Since defrouter_select_fib() does not affect the
1515 * on-link determination and MIP6 needs the check
1516 * before the default router selection, we perform
1519 pfxlist_onlink_check();
1522 * Refresh default router list.
1524 defrouter_select_fib(dr->ifp->if_fib);
1528 * If this entry was added by an on-link redirect, remove the
1529 * corresponding host route.
1531 if (ln->la_flags & LLE_REDIRECT)
1532 nd6_free_redirect(ln);
1534 if (ln->ln_router || dr)
1539 * Save to unlock. We still hold an extra reference and will not
1540 * free(9) in llentry_free() if someone else holds one as well.
1543 IF_AFDATA_LOCK(ifp);
1545 /* Guard against race with other llentry_free(). */
1546 if (ln->la_flags & LLE_LINKED) {
1547 /* Remove callout reference */
1549 lltable_unlink_entry(ln->lle_tbl, ln);
1551 IF_AFDATA_UNLOCK(ifp);
1559 nd6_isdynrte(const struct rtentry *rt, void *xap)
1562 if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC))
1568 * Remove the rtentry for the given llentry,
1569 * both of which were installed by a redirect.
1572 nd6_free_redirect(const struct llentry *ln)
1575 struct sockaddr_in6 sin6;
1576 struct rt_addrinfo info;
1578 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
1579 memset(&info, 0, sizeof(info));
1580 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6;
1581 info.rti_filter = nd6_isdynrte;
1583 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
1584 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum);
1588 * Rejuvenate this function for routing operations related
1592 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1594 struct sockaddr_in6 *gateway;
1595 struct nd_defrouter *dr;
1598 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1609 * Only indirect routes are interesting.
1611 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1614 * check for default route
1616 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1617 &SIN6(rt_key(rt))->sin6_addr)) {
1618 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1630 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1632 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1633 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1634 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1635 struct epoch_tracker et;
1638 if (ifp->if_afdata[AF_INET6] == NULL)
1639 return (EPFNOSUPPORT);
1641 case OSIOCGIFINFO_IN6:
1643 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1644 bzero(&ND, sizeof(ND));
1645 ND.linkmtu = IN6_LINKMTU(ifp);
1646 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1647 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1648 ND.reachable = ND_IFINFO(ifp)->reachable;
1649 ND.retrans = ND_IFINFO(ifp)->retrans;
1650 ND.flags = ND_IFINFO(ifp)->flags;
1651 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1652 ND.chlim = ND_IFINFO(ifp)->chlim;
1654 case SIOCGIFINFO_IN6:
1655 ND = *ND_IFINFO(ifp);
1657 case SIOCSIFINFO_IN6:
1659 * used to change host variables from userland.
1660 * intended for a use on router to reflect RA configurations.
1662 /* 0 means 'unspecified' */
1663 if (ND.linkmtu != 0) {
1664 if (ND.linkmtu < IPV6_MMTU ||
1665 ND.linkmtu > IN6_LINKMTU(ifp)) {
1669 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1672 if (ND.basereachable != 0) {
1673 int obasereachable = ND_IFINFO(ifp)->basereachable;
1675 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1676 if (ND.basereachable != obasereachable)
1677 ND_IFINFO(ifp)->reachable =
1678 ND_COMPUTE_RTIME(ND.basereachable);
1680 if (ND.retrans != 0)
1681 ND_IFINFO(ifp)->retrans = ND.retrans;
1683 ND_IFINFO(ifp)->chlim = ND.chlim;
1685 case SIOCSIFINFO_FLAGS:
1688 struct in6_ifaddr *ia;
1690 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1691 !(ND.flags & ND6_IFF_IFDISABLED)) {
1692 /* ifdisabled 1->0 transision */
1695 * If the interface is marked as ND6_IFF_IFDISABLED and
1696 * has an link-local address with IN6_IFF_DUPLICATED,
1697 * do not clear ND6_IFF_IFDISABLED.
1698 * See RFC 4862, Section 5.4.5.
1700 NET_EPOCH_ENTER(et);
1701 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1702 if (ifa->ifa_addr->sa_family != AF_INET6)
1704 ia = (struct in6_ifaddr *)ifa;
1705 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1706 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1712 /* LLA is duplicated. */
1713 ND.flags |= ND6_IFF_IFDISABLED;
1714 log(LOG_ERR, "Cannot enable an interface"
1715 " with a link-local address marked"
1718 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1719 if (ifp->if_flags & IFF_UP)
1722 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1723 (ND.flags & ND6_IFF_IFDISABLED)) {
1724 /* ifdisabled 0->1 transision */
1725 /* Mark all IPv6 address as tentative. */
1727 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1728 if (V_ip6_dad_count > 0 &&
1729 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1730 NET_EPOCH_ENTER(et);
1731 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1733 if (ifa->ifa_addr->sa_family !=
1736 ia = (struct in6_ifaddr *)ifa;
1737 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1743 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1744 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1745 /* auto_linklocal 0->1 transision */
1747 /* If no link-local address on ifp, configure */
1748 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1749 in6_ifattach(ifp, NULL);
1750 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1751 ifp->if_flags & IFF_UP) {
1753 * When the IF already has
1754 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1755 * address is assigned, and IFF_UP, try to
1758 NET_EPOCH_ENTER(et);
1759 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1761 if (ifa->ifa_addr->sa_family !=
1764 ia = (struct in6_ifaddr *)ifa;
1765 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1770 /* No LLA is configured. */
1771 in6_ifattach(ifp, NULL);
1775 ND_IFINFO(ifp)->flags = ND.flags;
1778 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1779 /* sync kernel routing table with the default router list */
1783 case SIOCSPFXFLUSH_IN6:
1785 /* flush all the prefix advertised by routers */
1786 struct in6_ifaddr *ia, *ia_next;
1787 struct nd_prefix *pr, *next;
1788 struct nd_prhead prl;
1793 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1794 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1796 nd6_prefix_unlink(pr, &prl);
1800 while ((pr = LIST_FIRST(&prl)) != NULL) {
1801 LIST_REMOVE(pr, ndpr_entry);
1802 /* XXXRW: in6_ifaddrhead locking. */
1803 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1805 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1808 if (ia->ia6_ndpr == pr)
1809 in6_purgeaddr(&ia->ia_ifa);
1815 case SIOCSRTRFLUSH_IN6:
1817 /* flush all the default routers */
1818 struct nd_drhead drq;
1819 struct nd_defrouter *dr;
1826 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL)
1827 defrouter_unlink(dr, &drq);
1829 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
1830 TAILQ_REMOVE(&drq, dr, dr_entry);
1837 case SIOCGNBRINFO_IN6:
1840 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1842 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1845 NET_EPOCH_ENTER(et);
1846 ln = nd6_lookup(&nb_addr, 0, ifp);
1853 nbi->state = ln->ln_state;
1854 nbi->asked = ln->la_asked;
1855 nbi->isrouter = ln->ln_router;
1856 if (ln->la_expire == 0)
1859 nbi->expire = ln->la_expire + ln->lle_remtime / hz +
1860 (time_second - time_uptime);
1864 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1865 ndif->ifindex = V_nd6_defifindex;
1867 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1868 return (nd6_setdefaultiface(ndif->ifindex));
1874 * Calculates new isRouter value based on provided parameters and
1878 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
1883 * ICMP6 type dependent behavior.
1885 * NS: clear IsRouter if new entry
1886 * RS: clear IsRouter
1887 * RA: set IsRouter if there's lladdr
1888 * redir: clear IsRouter if new entry
1891 * The spec says that we must set IsRouter in the following cases:
1892 * - If lladdr exist, set IsRouter. This means (1-5).
1893 * - If it is old entry (!newentry), set IsRouter. This means (7).
1894 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1895 * A quetion arises for (1) case. (1) case has no lladdr in the
1896 * neighbor cache, this is similar to (6).
1897 * This case is rare but we figured that we MUST NOT set IsRouter.
1899 * is_new old_addr new_addr NS RS RA redir
1906 * 1 -- n (6) c c c s
1907 * 1 -- y (7) c c s c s
1911 switch (type & 0xff) {
1912 case ND_NEIGHBOR_SOLICIT:
1914 * New entry must have is_router flag cleared.
1916 if (is_new) /* (6-7) */
1921 * If the icmp is a redirect to a better router, always set the
1922 * is_router flag. Otherwise, if the entry is newly created,
1923 * clear the flag. [RFC 2461, sec 8.3]
1925 if (code == ND_REDIRECT_ROUTER)
1928 if (is_new) /* (6-7) */
1932 case ND_ROUTER_SOLICIT:
1934 * is_router flag must always be cleared.
1938 case ND_ROUTER_ADVERT:
1940 * Mark an entry with lladdr as a router.
1942 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */
1943 (is_new && new_addr)) { /* (7) */
1953 * Create neighbor cache entry and cache link-layer address,
1954 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1957 * code - type dependent information
1961 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1962 int lladdrlen, int type, int code)
1964 struct llentry *ln = NULL, *ln_tmp;
1970 uint16_t router = 0;
1971 struct sockaddr_in6 sin6;
1972 struct epoch_tracker et;
1973 struct mbuf *chain = NULL;
1974 u_char linkhdr[LLE_MAX_LINKHDR];
1978 IF_AFDATA_UNLOCK_ASSERT(ifp);
1980 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1981 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1983 /* nothing must be updated for unspecified address */
1984 if (IN6_IS_ADDR_UNSPECIFIED(from))
1988 * Validation about ifp->if_addrlen and lladdrlen must be done in
1991 * XXX If the link does not have link-layer adderss, what should
1992 * we do? (ifp->if_addrlen == 0)
1993 * Spec says nothing in sections for RA, RS and NA. There's small
1994 * description on it in NS section (RFC 2461 7.2.3).
1996 flags = lladdr ? LLE_EXCLUSIVE : 0;
1997 NET_EPOCH_ENTER(et);
1998 ln = nd6_lookup(from, flags, ifp);
2002 flags |= LLE_EXCLUSIVE;
2003 ln = nd6_alloc(from, 0, ifp);
2008 * Since we already know all the data for the new entry,
2009 * fill it before insertion.
2011 if (lladdr != NULL) {
2012 linkhdrsize = sizeof(linkhdr);
2013 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
2014 linkhdr, &linkhdrsize, &lladdr_off) != 0)
2016 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
2020 IF_AFDATA_WLOCK(ifp);
2022 /* Prefer any existing lle over newly-created one */
2023 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp);
2025 lltable_link_entry(LLTABLE6(ifp), ln);
2026 IF_AFDATA_WUNLOCK(ifp);
2027 if (ln_tmp == NULL) {
2028 /* No existing lle, mark as new entry (6,7) */
2030 if (lladdr != NULL) { /* (7) */
2031 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2032 EVENTHANDLER_INVOKE(lle_event, ln,
2036 lltable_free_entry(LLTABLE6(ifp), ln);
2041 /* do nothing if static ndp is set */
2042 if ((ln->la_flags & LLE_STATIC)) {
2043 if (flags & LLE_EXCLUSIVE)
2050 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
2051 if (olladdr && lladdr) {
2052 llchange = bcmp(lladdr, ln->ll_addr,
2054 } else if (!olladdr && lladdr)
2060 * newentry olladdr lladdr llchange (*=record)
2063 * 0 n y y (3) * STALE
2065 * 0 y y y (5) * STALE
2066 * 1 -- n -- (6) NOSTATE(= PASSIVE)
2067 * 1 -- y -- (7) * STALE
2071 if (is_newentry == 0 && llchange != 0) {
2072 do_update = 1; /* (3,5) */
2075 * Record source link-layer address
2076 * XXX is it dependent to ifp->if_type?
2078 linkhdrsize = sizeof(linkhdr);
2079 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
2080 linkhdr, &linkhdrsize, &lladdr_off) != 0)
2083 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
2085 /* Entry was deleted */
2089 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2091 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2093 if (ln->la_hold != NULL)
2094 nd6_grab_holdchain(ln, &chain, &sin6);
2097 /* Calculates new router status */
2098 router = nd6_is_router(type, code, is_newentry, olladdr,
2099 lladdr != NULL ? 1 : 0, ln->ln_router);
2101 ln->ln_router = router;
2102 /* Mark non-router redirects with special flag */
2103 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
2104 ln->la_flags |= LLE_REDIRECT;
2106 if (flags & LLE_EXCLUSIVE)
2112 nd6_flush_holdchain(ifp, chain, &sin6);
2115 * When the link-layer address of a router changes, select the
2116 * best router again. In particular, when the neighbor entry is newly
2117 * created, it might affect the selection policy.
2118 * Question: can we restrict the first condition to the "is_newentry"
2120 * XXX: when we hear an RA from a new router with the link-layer
2121 * address option, defrouter_select_fib() is called twice, since
2122 * defrtrlist_update called the function as well. However, I believe
2123 * we can compromise the overhead, since it only happens the first
2125 * XXX: although defrouter_select_fib() should not have a bad effect
2126 * for those are not autoconfigured hosts, we explicitly avoid such
2129 if ((do_update || is_newentry) && router &&
2130 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
2132 * guaranteed recursion
2134 defrouter_select_fib(ifp->if_fib);
2139 nd6_slowtimo(void *arg)
2141 struct epoch_tracker et;
2142 CURVNET_SET((struct vnet *) arg);
2143 struct nd_ifinfo *nd6if;
2146 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
2147 nd6_slowtimo, curvnet);
2148 NET_EPOCH_ENTER(et);
2149 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2150 if (ifp->if_afdata[AF_INET6] == NULL)
2152 nd6if = ND_IFINFO(ifp);
2153 if (nd6if->basereachable && /* already initialized */
2154 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
2156 * Since reachable time rarely changes by router
2157 * advertisements, we SHOULD insure that a new random
2158 * value gets recomputed at least once every few hours.
2161 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
2162 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
2170 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
2171 struct sockaddr_in6 *sin6)
2174 LLE_WLOCK_ASSERT(ln);
2176 *chain = ln->la_hold;
2178 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6);
2180 if (ln->ln_state == ND6_LLINFO_STALE) {
2183 * The first time we send a packet to a
2184 * neighbor whose entry is STALE, we have
2185 * to change the state to DELAY and a sets
2186 * a timer to expire in DELAY_FIRST_PROBE_TIME
2187 * seconds to ensure do neighbor unreachability
2188 * detection on expiration.
2191 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
2196 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
2197 struct sockaddr_in6 *dst, struct route *ro)
2201 struct ip6_hdr *ip6;
2205 mac_netinet6_nd6_send(ifp, m);
2209 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
2210 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
2211 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
2212 * to be diverted to user space. When re-injected into the kernel,
2213 * send_output() will directly dispatch them to the outgoing interface.
2215 if (send_sendso_input_hook != NULL) {
2216 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
2218 ip6 = mtod(m, struct ip6_hdr *);
2219 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2220 /* Use the SEND socket */
2221 error = send_sendso_input_hook(m, ifp, SND_OUT,
2223 /* -1 == no app on SEND socket */
2224 if (error == 0 || error != -1)
2229 m_clrprotoflags(m); /* Avoid confusing lower layers. */
2230 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
2231 mtod(m, struct ip6_hdr *));
2233 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2236 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro);
2241 * Lookup link headerfor @sa_dst address. Stores found
2242 * data in @desten buffer. Copy of lle ln_flags can be also
2243 * saved in @pflags if @pflags is non-NULL.
2245 * If destination LLE does not exists or lle state modification
2246 * is required, call "slow" version.
2249 * - 0 on success (address copied to buffer).
2250 * - EWOULDBLOCK (no local error, but address is still unresolved)
2251 * - other errors (alloc failure, etc)
2254 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
2255 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags,
2256 struct llentry **plle)
2258 struct epoch_tracker et;
2259 struct llentry *ln = NULL;
2260 const struct sockaddr_in6 *dst6;
2265 dst6 = (const struct sockaddr_in6 *)sa_dst;
2267 /* discard the packet if IPv6 operation is disabled on the interface */
2268 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2270 return (ENETDOWN); /* better error? */
2273 if (m != NULL && m->m_flags & M_MCAST) {
2274 switch (ifp->if_type) {
2278 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
2283 return (EAFNOSUPPORT);
2287 NET_EPOCH_ENTER(et);
2288 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED,
2290 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
2291 /* Entry found, let's copy lle info */
2292 bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
2294 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
2295 /* Check if we have feedback request from nd6 timer */
2296 if (ln->r_skip_req != 0) {
2298 ln->r_skip_req = 0; /* Notify that entry was used */
2299 ln->lle_hittime = time_uptime;
2309 } else if (plle && ln)
2313 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle));
2318 * Do L2 address resolution for @sa_dst address. Stores found
2319 * address in @desten buffer. Copy of lle ln_flags can be also
2320 * saved in @pflags if @pflags is non-NULL.
2323 * Function assume that destination LLE does not exist,
2324 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired.
2326 * Set noinline to be dtrace-friendly
2328 static __noinline int
2329 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m,
2330 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags,
2331 struct llentry **plle)
2333 struct llentry *lle = NULL, *lle_tmp;
2334 struct in6_addr *psrc, src;
2335 int send_ns, ll_len;
2339 * Address resolution or Neighbor Unreachability Detection
2341 * At this point, the destination of the packet must be a unicast
2342 * or an anycast address(i.e. not a multicast).
2345 struct epoch_tracker et;
2347 NET_EPOCH_ENTER(et);
2348 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
2350 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
2352 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2353 * the condition below is not very efficient. But we believe
2354 * it is tolerable, because this should be a rare case.
2356 lle = nd6_alloc(&dst->sin6_addr, 0, ifp);
2358 char ip6buf[INET6_ADDRSTRLEN];
2360 "nd6_output: can't allocate llinfo for %s "
2362 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
2367 IF_AFDATA_WLOCK(ifp);
2369 /* Prefer any existing entry over newly-created one */
2370 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
2371 if (lle_tmp == NULL)
2372 lltable_link_entry(LLTABLE6(ifp), lle);
2373 IF_AFDATA_WUNLOCK(ifp);
2374 if (lle_tmp != NULL) {
2375 lltable_free_entry(LLTABLE6(ifp), lle);
2382 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2392 LLE_WLOCK_ASSERT(lle);
2395 * The first time we send a packet to a neighbor whose entry is
2396 * STALE, we have to change the state to DELAY and a sets a timer to
2397 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2398 * neighbor unreachability detection on expiration.
2401 if (lle->ln_state == ND6_LLINFO_STALE)
2402 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
2405 * If the neighbor cache entry has a state other than INCOMPLETE
2406 * (i.e. its link-layer address is already resolved), just
2409 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
2410 if (flags & LLE_ADDRONLY) {
2411 lladdr = lle->ll_addr;
2412 ll_len = ifp->if_addrlen;
2414 lladdr = lle->r_linkdata;
2415 ll_len = lle->r_hdrlen;
2417 bcopy(lladdr, desten, ll_len);
2419 *pflags = lle->la_flags;
2429 * There is a neighbor cache entry, but no ethernet address
2430 * response yet. Append this latest packet to the end of the
2431 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen,
2432 * the oldest packet in the queue will be removed.
2435 if (lle->la_hold != NULL) {
2436 struct mbuf *m_hold;
2440 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2442 if (m_hold->m_nextpkt == NULL) {
2443 m_hold->m_nextpkt = m;
2447 while (i >= V_nd6_maxqueuelen) {
2448 m_hold = lle->la_hold;
2449 lle->la_hold = lle->la_hold->m_nextpkt;
2458 * If there has been no NS for the neighbor after entering the
2459 * INCOMPLETE state, send the first solicitation.
2460 * Note that for newly-created lle la_asked will be 0,
2461 * so we will transition from ND6_LLINFO_NOSTATE to
2462 * ND6_LLINFO_INCOMPLETE state here.
2466 if (lle->la_asked == 0) {
2469 psrc = nd6_llinfo_get_holdsrc(lle, &src);
2471 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
2475 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
2477 return (EWOULDBLOCK);
2481 * Do L2 address resolution for @sa_dst address. Stores found
2482 * address in @desten buffer. Copy of lle ln_flags can be also
2483 * saved in @pflags if @pflags is non-NULL.
2486 * - 0 on success (address copied to buffer).
2487 * - EWOULDBLOCK (no local error, but address is still unresolved)
2488 * - other errors (alloc failure, etc)
2491 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
2492 char *desten, uint32_t *pflags)
2496 flags |= LLE_ADDRONLY;
2497 error = nd6_resolve_slow(ifp, flags, NULL,
2498 (const struct sockaddr_in6 *)dst, desten, pflags, NULL);
2503 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain,
2504 struct sockaddr_in6 *dst)
2506 struct mbuf *m, *m_head;
2513 m_head = m_head->m_nextpkt;
2514 error = nd6_output_ifp(ifp, ifp, m, dst, NULL);
2519 * note that intermediate errors are blindly ignored
2525 nd6_need_cache(struct ifnet *ifp)
2528 * XXX: we currently do not make neighbor cache on any interface
2529 * other than Ethernet and GIF.
2532 * - unidirectional tunnels needs no ND
2534 switch (ifp->if_type) {
2538 case IFT_INFINIBAND:
2540 case IFT_PROPVIRTUAL:
2548 * Add pernament ND6 link-layer record for given
2549 * interface address.
2551 * Very similar to IPv4 arp_ifinit(), but:
2552 * 1) IPv6 DAD is performed in different place
2553 * 2) It is called by IPv6 protocol stack in contrast to
2554 * arp_ifinit() which is typically called in SIOCSIFADDR
2555 * driver ioctl handler.
2559 nd6_add_ifa_lle(struct in6_ifaddr *ia)
2562 struct llentry *ln, *ln_tmp;
2563 struct sockaddr *dst;
2565 ifp = ia->ia_ifa.ifa_ifp;
2566 if (nd6_need_cache(ifp) == 0)
2569 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
2570 dst = (struct sockaddr *)&ia->ia_addr;
2571 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
2575 IF_AFDATA_WLOCK(ifp);
2577 /* Unlink any entry if exists */
2578 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst);
2580 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
2581 lltable_link_entry(LLTABLE6(ifp), ln);
2582 IF_AFDATA_WUNLOCK(ifp);
2585 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
2586 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2590 llentry_free(ln_tmp);
2596 * Removes either all lle entries for given @ia, or lle
2597 * corresponding to @ia address.
2600 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
2602 struct sockaddr_in6 mask, addr;
2603 struct sockaddr *saddr, *smask;
2606 ifp = ia->ia_ifa.ifa_ifp;
2607 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
2608 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
2609 saddr = (struct sockaddr *)&addr;
2610 smask = (struct sockaddr *)&mask;
2613 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
2615 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
2619 clear_llinfo_pqueue(struct llentry *ln)
2621 struct mbuf *m_hold, *m_hold_next;
2623 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2624 m_hold_next = m_hold->m_nextpkt;
2631 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2632 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2634 SYSCTL_DECL(_net_inet6_icmp6);
2635 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2636 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2637 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2638 "NDP default router list");
2639 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2640 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2641 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix",
2643 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2644 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2645 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2646 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2649 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2651 struct in6_defrouter d;
2652 struct nd_defrouter *dr;
2655 if (req->newptr != NULL)
2658 error = sysctl_wire_old_buffer(req, 0);
2662 bzero(&d, sizeof(d));
2663 d.rtaddr.sin6_family = AF_INET6;
2664 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2667 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2668 d.rtaddr.sin6_addr = dr->rtaddr;
2669 error = sa6_recoverscope(&d.rtaddr);
2672 d.flags = dr->raflags;
2673 d.rtlifetime = dr->rtlifetime;
2674 d.expire = dr->expire + (time_second - time_uptime);
2675 d.if_index = dr->ifp->if_index;
2676 error = SYSCTL_OUT(req, &d, sizeof(d));
2685 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2687 struct in6_prefix p;
2688 struct sockaddr_in6 s6;
2689 struct nd_prefix *pr;
2690 struct nd_pfxrouter *pfr;
2693 char ip6buf[INET6_ADDRSTRLEN];
2698 error = sysctl_wire_old_buffer(req, 0);
2702 bzero(&p, sizeof(p));
2703 p.origin = PR_ORIG_RA;
2704 bzero(&s6, sizeof(s6));
2705 s6.sin6_family = AF_INET6;
2706 s6.sin6_len = sizeof(s6);
2709 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2710 p.prefix = pr->ndpr_prefix;
2711 if (sa6_recoverscope(&p.prefix)) {
2712 log(LOG_ERR, "scope error in prefix list (%s)\n",
2713 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2714 /* XXX: press on... */
2716 p.raflags = pr->ndpr_raf;
2717 p.prefixlen = pr->ndpr_plen;
2718 p.vltime = pr->ndpr_vltime;
2719 p.pltime = pr->ndpr_pltime;
2720 p.if_index = pr->ndpr_ifp->if_index;
2721 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2724 /* XXX: we assume time_t is signed. */
2726 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2727 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2728 p.expire = pr->ndpr_lastupdate +
2730 (time_second - time_uptime);
2732 p.expire = maxexpire;
2734 p.refcnt = pr->ndpr_addrcnt;
2735 p.flags = pr->ndpr_stateflags;
2737 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2739 error = SYSCTL_OUT(req, &p, sizeof(p));
2742 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2743 s6.sin6_addr = pfr->router->rtaddr;
2744 if (sa6_recoverscope(&s6))
2746 "scope error in prefix list (%s)\n",
2747 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2748 error = SYSCTL_OUT(req, &s6, sizeof(s6));