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;
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);
244 callout_drain(&V_nd6_slowtimo_ch);
245 callout_drain(&V_nd6_timer_ch);
246 if (IS_DEFAULT_VNET(curvnet)) {
247 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
248 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
250 rw_destroy(&V_nd6_lock);
251 mtx_destroy(&V_nd6_onlink_mtx);
256 nd6_ifattach(struct ifnet *ifp)
258 struct nd_ifinfo *nd;
260 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
263 nd->chlim = IPV6_DEFHLIM;
264 nd->basereachable = REACHABLE_TIME;
265 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
266 nd->retrans = RETRANS_TIMER;
268 nd->flags = ND6_IFF_PERFORMNUD;
270 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
271 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
272 * default regardless of the V_ip6_auto_linklocal configuration to
273 * give a reasonable default behavior.
275 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
276 (ifp->if_flags & IFF_LOOPBACK))
277 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
279 * A loopback interface does not need to accept RTADV.
280 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
281 * default regardless of the V_ip6_accept_rtadv configuration to
282 * prevent the interface from accepting RA messages arrived
283 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
285 if (V_ip6_accept_rtadv &&
286 !(ifp->if_flags & IFF_LOOPBACK) &&
287 (ifp->if_type != IFT_BRIDGE))
288 nd->flags |= ND6_IFF_ACCEPT_RTADV;
289 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
290 nd->flags |= ND6_IFF_NO_RADR;
292 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
293 nd6_setmtu0(ifp, nd);
299 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd)
301 struct ifaddr *ifa, *next;
304 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
305 if (ifa->ifa_addr->sa_family != AF_INET6)
308 /* stop DAD processing */
311 IF_ADDR_RUNLOCK(ifp);
317 * Reset ND level link MTU. This function is called when the physical MTU
318 * changes, which means we might have to adjust the ND level MTU.
321 nd6_setmtu(struct ifnet *ifp)
323 if (ifp->if_afdata[AF_INET6] == NULL)
326 nd6_setmtu0(ifp, ND_IFINFO(ifp));
329 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
331 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
335 omaxmtu = ndi->maxmtu;
336 ndi->maxmtu = ifp->if_mtu;
339 * Decreasing the interface MTU under IPV6 minimum MTU may cause
340 * undesirable situation. We thus notify the operator of the change
341 * explicitly. The check for omaxmtu is necessary to restrict the
342 * log to the case of changing the MTU, not initializing it.
344 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
345 log(LOG_NOTICE, "nd6_setmtu0: "
346 "new link MTU on %s (%lu) is too small for IPv6\n",
347 if_name(ifp), (unsigned long)ndi->maxmtu);
350 if (ndi->maxmtu > V_in6_maxmtu)
351 in6_setmaxmtu(); /* check all interfaces just in case */
356 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
359 bzero(ndopts, sizeof(*ndopts));
360 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
362 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
365 ndopts->nd_opts_done = 1;
366 ndopts->nd_opts_search = NULL;
371 * Take one ND option.
374 nd6_option(union nd_opts *ndopts)
376 struct nd_opt_hdr *nd_opt;
379 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
380 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
382 if (ndopts->nd_opts_search == NULL)
384 if (ndopts->nd_opts_done)
387 nd_opt = ndopts->nd_opts_search;
389 /* make sure nd_opt_len is inside the buffer */
390 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
391 bzero(ndopts, sizeof(*ndopts));
395 olen = nd_opt->nd_opt_len << 3;
398 * Message validation requires that all included
399 * options have a length that is greater than zero.
401 bzero(ndopts, sizeof(*ndopts));
405 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
406 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
407 /* option overruns the end of buffer, invalid */
408 bzero(ndopts, sizeof(*ndopts));
410 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
411 /* reached the end of options chain */
412 ndopts->nd_opts_done = 1;
413 ndopts->nd_opts_search = NULL;
419 * Parse multiple ND options.
420 * This function is much easier to use, for ND routines that do not need
421 * multiple options of the same type.
424 nd6_options(union nd_opts *ndopts)
426 struct nd_opt_hdr *nd_opt;
429 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
430 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
432 if (ndopts->nd_opts_search == NULL)
436 nd_opt = nd6_option(ndopts);
437 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
439 * Message validation requires that all included
440 * options have a length that is greater than zero.
442 ICMP6STAT_INC(icp6s_nd_badopt);
443 bzero(ndopts, sizeof(*ndopts));
450 switch (nd_opt->nd_opt_type) {
451 case ND_OPT_SOURCE_LINKADDR:
452 case ND_OPT_TARGET_LINKADDR:
454 case ND_OPT_REDIRECTED_HEADER:
456 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
458 "duplicated ND6 option found (type=%d)\n",
459 nd_opt->nd_opt_type));
462 ndopts->nd_opt_array[nd_opt->nd_opt_type]
466 case ND_OPT_PREFIX_INFORMATION:
467 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
468 ndopts->nd_opt_array[nd_opt->nd_opt_type]
471 ndopts->nd_opts_pi_end =
472 (struct nd_opt_prefix_info *)nd_opt;
474 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
475 case ND_OPT_RDNSS: /* RFC 6106 */
476 case ND_OPT_DNSSL: /* RFC 6106 */
478 * Silently ignore options we know and do not care about
484 * Unknown options must be silently ignored,
485 * to accommodate future extension to the protocol.
488 "nd6_options: unsupported option %d - "
489 "option ignored\n", nd_opt->nd_opt_type));
494 if (i > V_nd6_maxndopt) {
495 ICMP6STAT_INC(icp6s_nd_toomanyopt);
496 nd6log((LOG_INFO, "too many loop in nd opt\n"));
500 if (ndopts->nd_opts_done)
508 * ND6 timer routine to handle ND6 entries
511 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
515 LLE_WLOCK_ASSERT(ln);
520 canceled = callout_stop(&ln->lle_timer);
522 ln->la_expire = time_uptime + tick / hz;
524 if (tick > INT_MAX) {
525 ln->ln_ntick = tick - INT_MAX;
526 canceled = callout_reset(&ln->lle_timer, INT_MAX,
527 nd6_llinfo_timer, ln);
530 canceled = callout_reset(&ln->lle_timer, tick,
531 nd6_llinfo_timer, ln);
539 * Gets source address of the first packet in hold queue
540 * and stores it in @src.
541 * Returns pointer to @src (if hold queue is not empty) or NULL.
543 * Set noinline to be dtrace-friendly
545 static __noinline struct in6_addr *
546 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
551 if (ln->la_hold == NULL)
555 * assume every packet in la_hold has the same IP header
558 if (sizeof(hdr) > m->m_len)
561 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
568 * Checks if we need to switch from STALE state.
570 * RFC 4861 requires switching from STALE to DELAY state
571 * on first packet matching entry, waiting V_nd6_delay and
572 * transition to PROBE state (if upper layer confirmation was
575 * This code performs a bit differently:
576 * On packet hit we don't change state (but desired state
577 * can be guessed by control plane). However, after V_nd6_delay
578 * seconds code will transition to PROBE state (so DELAY state
579 * is kinda skipped in most situations).
581 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so
582 * we perform the following upon entering STALE state:
584 * 1) Arm timer to run each V_nd6_delay seconds to make sure that
585 * if packet was transmitted at the start of given interval, we
586 * would be able to switch to PROBE state in V_nd6_delay seconds
589 * 2) Reschedule timer until original V_nd6_gctimer expires keeping
590 * lle in STALE state (remaining timer value stored in lle_remtime).
592 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay
595 * Returns non-zero value if the entry is still STALE (storing
596 * the next timer interval in @pdelay).
598 * Returns zero value if original timer expired or we need to switch to
599 * PROBE (store that in @do_switch variable).
602 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch)
604 int nd_delay, nd_gctimer, r_skip_req;
609 nd_gctimer = V_nd6_gctimer;
610 nd_delay = V_nd6_delay;
613 r_skip_req = lle->r_skip_req;
614 lle_hittime = lle->lle_hittime;
617 if (r_skip_req > 0) {
620 * Nonzero r_skip_req value was set upon entering
621 * STALE state. Since value was not changed, no
622 * packets were passed using this lle. Ask for
623 * timer reschedule and keep STALE state.
625 delay = (long)(MIN(nd_gctimer, nd_delay));
627 if (lle->lle_remtime > delay)
628 lle->lle_remtime -= delay;
630 delay = lle->lle_remtime;
631 lle->lle_remtime = 0;
637 * The original ng6_gctime timeout ended,
638 * no more rescheduling.
648 * Packet received. Verify timestamp
650 delay = (long)(time_uptime - lle_hittime);
651 if (delay < nd_delay) {
654 * V_nd6_delay still not passed since the first
655 * hit in STALE state.
656 * Reshedule timer and return.
658 *pdelay = (long)(nd_delay - delay) * hz;
662 /* Request switching to probe */
669 * Switch @lle state to new state optionally arming timers.
671 * Set noinline to be dtrace-friendly
674 nd6_llinfo_setstate(struct llentry *lle, int newstate)
677 int nd_gctimer, nd_delay;
684 case ND6_LLINFO_INCOMPLETE:
685 ifp = lle->lle_tbl->llt_ifp;
686 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
688 case ND6_LLINFO_REACHABLE:
689 if (!ND6_LLINFO_PERMANENT(lle)) {
690 ifp = lle->lle_tbl->llt_ifp;
691 delay = (long)ND_IFINFO(ifp)->reachable * hz;
694 case ND6_LLINFO_STALE:
697 * Notify fast path that we want to know if any packet
698 * is transmitted by setting r_skip_req.
703 nd_delay = V_nd6_delay;
704 nd_gctimer = V_nd6_gctimer;
706 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz;
707 remtime = (long)nd_gctimer * hz - delay;
709 case ND6_LLINFO_DELAY:
711 delay = (long)V_nd6_delay * hz;
716 nd6_llinfo_settimer_locked(lle, delay);
718 lle->lle_remtime = remtime;
719 lle->ln_state = newstate;
723 * Timer-dependent part of nd state machine.
725 * Set noinline to be dtrace-friendly
727 static __noinline void
728 nd6_llinfo_timer(void *arg)
731 struct in6_addr *dst, *pdst, *psrc, src;
733 struct nd_ifinfo *ndi;
734 int do_switch, send_ns;
737 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
738 ln = (struct llentry *)arg;
739 ifp = lltable_get_ifp(ln->lle_tbl);
740 CURVNET_SET(ifp->if_vnet);
744 if (callout_pending(&ln->lle_timer)) {
746 * Here we are a bit odd here in the treatment of
747 * active/pending. If the pending bit is set, it got
748 * rescheduled before I ran. The active
749 * bit we ignore, since if it was stopped
750 * in ll_tablefree() and was currently running
751 * it would have return 0 so the code would
752 * not have deleted it since the callout could
753 * not be stopped so we want to go through
754 * with the delete here now. If the callout
755 * was restarted, the pending bit will be back on and
756 * we just want to bail since the callout_reset would
757 * return 1 and our reference would have been removed
758 * by nd6_llinfo_settimer_locked above since canceled
766 ndi = ND_IFINFO(ifp);
768 dst = &ln->r_l3addr.addr6;
771 if (ln->ln_ntick > 0) {
772 if (ln->ln_ntick > INT_MAX) {
773 ln->ln_ntick -= INT_MAX;
774 nd6_llinfo_settimer_locked(ln, INT_MAX);
777 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
782 if (ln->la_flags & LLE_STATIC) {
786 if (ln->la_flags & LLE_DELETED) {
791 switch (ln->ln_state) {
792 case ND6_LLINFO_INCOMPLETE:
793 if (ln->la_asked < V_nd6_mmaxtries) {
796 /* Send NS to multicast address */
799 struct mbuf *m = ln->la_hold;
804 * assuming every packet in la_hold has the
805 * same IP header. Send error after unlock.
810 clear_llinfo_pqueue(ln);
814 icmp6_error2(m, ICMP6_DST_UNREACH,
815 ICMP6_DST_UNREACH_ADDR, 0, ifp);
818 case ND6_LLINFO_REACHABLE:
819 if (!ND6_LLINFO_PERMANENT(ln))
820 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
823 case ND6_LLINFO_STALE:
824 if (nd6_is_stale(ln, &delay, &do_switch) != 0) {
827 * No packet has used this entry and GC timeout
828 * has not been passed. Reshedule timer and
831 nd6_llinfo_settimer_locked(ln, delay);
835 if (do_switch == 0) {
838 * GC timer has ended and entry hasn't been used.
839 * Run Garbage collector (RFC 4861, 5.3)
841 if (!ND6_LLINFO_PERMANENT(ln))
846 /* Entry has been used AND delay timer has ended. */
850 case ND6_LLINFO_DELAY:
851 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
854 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
857 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
859 case ND6_LLINFO_PROBE:
860 if (ln->la_asked < V_nd6_umaxtries) {
868 panic("%s: paths in a dark night can be confusing: %d",
869 __func__, ln->ln_state);
875 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
876 psrc = nd6_llinfo_get_holdsrc(ln, &src);
879 nd6_ns_output(ifp, psrc, pdst, dst, NULL);
889 * ND6 timer routine to expire default route list and prefix list
894 CURVNET_SET((struct vnet *) arg);
895 struct nd_drhead drq;
896 struct nd_prhead prl;
897 struct nd_defrouter *dr, *ndr;
898 struct nd_prefix *pr, *npr;
900 struct in6_ifaddr *ia6, *nia6;
907 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr)
908 if (dr->expire && dr->expire < time_uptime)
909 defrouter_unlink(dr, &drq);
912 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
913 TAILQ_REMOVE(&drq, dr, dr_entry);
918 * expire interface addresses.
919 * in the past the loop was inside prefix expiry processing.
920 * However, from a stricter speci-confrmance standpoint, we should
921 * rather separate address lifetimes and prefix lifetimes.
923 * XXXRW: in6_ifaddrhead locking.
926 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
927 /* check address lifetime */
928 if (IFA6_IS_INVALID(ia6)) {
932 * If the expiring address is temporary, try
933 * regenerating a new one. This would be useful when
934 * we suspended a laptop PC, then turned it on after a
935 * period that could invalidate all temporary
936 * addresses. Although we may have to restart the
937 * loop (see below), it must be after purging the
938 * address. Otherwise, we'd see an infinite loop of
941 if (V_ip6_use_tempaddr &&
942 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
943 if (regen_tmpaddr(ia6) == 0)
947 in6_purgeaddr(&ia6->ia_ifa);
950 goto addrloop; /* XXX: see below */
951 } else if (IFA6_IS_DEPRECATED(ia6)) {
952 int oldflags = ia6->ia6_flags;
954 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
957 * If a temporary address has just become deprecated,
958 * regenerate a new one if possible.
960 if (V_ip6_use_tempaddr &&
961 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
962 (oldflags & IN6_IFF_DEPRECATED) == 0) {
964 if (regen_tmpaddr(ia6) == 0) {
966 * A new temporary address is
968 * XXX: this means the address chain
969 * has changed while we are still in
970 * the loop. Although the change
971 * would not cause disaster (because
972 * it's not a deletion, but an
973 * addition,) we'd rather restart the
974 * loop just for safety. Or does this
975 * significantly reduce performance??
980 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
982 * Schedule DAD for a tentative address. This happens
983 * if the interface was down or not running
984 * when the address was configured.
988 delay = arc4random() %
989 (MAX_RTR_SOLICITATION_DELAY * hz);
990 nd6_dad_start((struct ifaddr *)ia6, delay);
993 * Check status of the interface. If it is down,
994 * mark the address as tentative for future DAD.
997 if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 &&
998 ((ifp->if_flags & IFF_UP) == 0 ||
999 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1000 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){
1001 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
1002 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
1006 * A new RA might have made a deprecated address
1009 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
1015 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1017 * Expire prefixes. Since the pltime is only used for
1018 * autoconfigured addresses, pltime processing for prefixes is
1021 * Only unlink after all derived addresses have expired. This
1022 * may not occur until two hours after the prefix has expired
1023 * per RFC 4862. If the prefix expires before its derived
1024 * addresses, mark it off-link. This will be done automatically
1025 * after unlinking if no address references remain.
1027 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME ||
1028 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime)
1031 if (pr->ndpr_addrcnt == 0) {
1032 nd6_prefix_unlink(pr, &prl);
1035 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1036 genid = V_nd6_list_genid;
1040 (void)nd6_prefix_offlink(pr);
1041 ND6_ONLINK_UNLOCK();
1043 nd6_prefix_rele(pr);
1044 if (genid != V_nd6_list_genid)
1050 while ((pr = LIST_FIRST(&prl)) != NULL) {
1051 LIST_REMOVE(pr, ndpr_entry);
1055 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
1056 nd6_timer, curvnet);
1062 * ia6 - deprecated/invalidated temporary address
1065 regen_tmpaddr(struct in6_ifaddr *ia6)
1069 struct in6_ifaddr *public_ifa6 = NULL;
1071 ifp = ia6->ia_ifa.ifa_ifp;
1073 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1074 struct in6_ifaddr *it6;
1076 if (ifa->ifa_addr->sa_family != AF_INET6)
1079 it6 = (struct in6_ifaddr *)ifa;
1081 /* ignore no autoconf addresses. */
1082 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1085 /* ignore autoconf addresses with different prefixes. */
1086 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
1090 * Now we are looking at an autoconf address with the same
1091 * prefix as ours. If the address is temporary and is still
1092 * preferred, do not create another one. It would be rare, but
1093 * could happen, for example, when we resume a laptop PC after
1096 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1097 !IFA6_IS_DEPRECATED(it6)) {
1103 * This is a public autoconf address that has the same prefix
1104 * as ours. If it is preferred, keep it. We can't break the
1105 * loop here, because there may be a still-preferred temporary
1106 * address with the prefix.
1108 if (!IFA6_IS_DEPRECATED(it6))
1111 if (public_ifa6 != NULL)
1112 ifa_ref(&public_ifa6->ia_ifa);
1113 IF_ADDR_RUNLOCK(ifp);
1115 if (public_ifa6 != NULL) {
1118 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
1119 ifa_free(&public_ifa6->ia_ifa);
1120 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
1121 " tmp addr,errno=%d\n", e);
1124 ifa_free(&public_ifa6->ia_ifa);
1132 * Remove prefix and default router list entries corresponding to ifp. Neighbor
1133 * cache entries are freed in in6_domifdetach().
1136 nd6_purge(struct ifnet *ifp)
1138 struct nd_drhead drq;
1139 struct nd_prhead prl;
1140 struct nd_defrouter *dr, *ndr;
1141 struct nd_prefix *pr, *npr;
1147 * Nuke default router list entries toward ifp.
1148 * We defer removal of default router list entries that is installed
1149 * in the routing table, in order to keep additional side effects as
1150 * small as possible.
1153 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
1157 defrouter_unlink(dr, &drq);
1159 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
1163 defrouter_unlink(dr, &drq);
1167 * Remove prefixes on ifp. We should have already removed addresses on
1168 * this interface, so no addresses should be referencing these prefixes.
1170 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
1171 if (pr->ndpr_ifp == ifp)
1172 nd6_prefix_unlink(pr, &prl);
1176 /* Delete the unlinked router and prefix objects. */
1177 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
1178 TAILQ_REMOVE(&drq, dr, dr_entry);
1181 while ((pr = LIST_FIRST(&prl)) != NULL) {
1182 LIST_REMOVE(pr, ndpr_entry);
1186 /* cancel default outgoing interface setting */
1187 if (V_nd6_defifindex == ifp->if_index)
1188 nd6_setdefaultiface(0);
1190 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1191 /* Refresh default router list. */
1192 defrouter_select_fib(ifp->if_fib);
1197 * the caller acquires and releases the lock on the lltbls
1198 * Returns the llentry locked
1201 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1203 struct sockaddr_in6 sin6;
1206 bzero(&sin6, sizeof(sin6));
1207 sin6.sin6_len = sizeof(struct sockaddr_in6);
1208 sin6.sin6_family = AF_INET6;
1209 sin6.sin6_addr = *addr6;
1211 IF_AFDATA_LOCK_ASSERT(ifp);
1213 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6);
1219 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1221 struct sockaddr_in6 sin6;
1224 bzero(&sin6, sizeof(sin6));
1225 sin6.sin6_len = sizeof(struct sockaddr_in6);
1226 sin6.sin6_family = AF_INET6;
1227 sin6.sin6_addr = *addr6;
1229 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
1231 ln->ln_state = ND6_LLINFO_NOSTATE;
1237 * Test whether a given IPv6 address is a neighbor or not, ignoring
1238 * the actual neighbor cache. The neighbor cache is ignored in order
1239 * to not reenter the routing code from within itself.
1242 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1244 struct nd_prefix *pr;
1246 struct rt_addrinfo info;
1247 struct sockaddr_in6 rt_key;
1248 const struct sockaddr *dst6;
1253 * A link-local address is always a neighbor.
1254 * XXX: a link does not necessarily specify a single interface.
1256 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
1257 struct sockaddr_in6 sin6_copy;
1261 * We need sin6_copy since sa6_recoverscope() may modify the
1265 if (sa6_recoverscope(&sin6_copy))
1266 return (0); /* XXX: should be impossible */
1267 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
1269 if (sin6_copy.sin6_scope_id == zone)
1275 bzero(&rt_key, sizeof(rt_key));
1276 bzero(&info, sizeof(info));
1277 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1280 * If the address matches one of our addresses,
1281 * it should be a neighbor.
1282 * If the address matches one of our on-link prefixes, it should be a
1287 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1288 if (pr->ndpr_ifp != ifp)
1291 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1292 dst6 = (const struct sockaddr *)&pr->ndpr_prefix;
1295 * We only need to check all FIBs if add_addr_allfibs
1296 * is unset. If set, checking any FIB will suffice.
1298 fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0;
1299 for (; fibnum < rt_numfibs; fibnum++) {
1300 genid = V_nd6_list_genid;
1304 * Restore length field before
1307 rt_key.sin6_len = sizeof(rt_key);
1308 error = rib_lookup_info(fibnum, dst6, 0, 0,
1312 if (genid != V_nd6_list_genid)
1321 * This is the case where multiple interfaces
1322 * have the same prefix, but only one is installed
1323 * into the routing table and that prefix entry
1324 * is not the one being examined here. In the case
1325 * where RADIX_MPATH is enabled, multiple route
1326 * entries (of the same rt_key value) will be
1327 * installed because the interface addresses all
1330 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1335 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1336 &addr->sin6_addr, &pr->ndpr_mask)) {
1344 * If the address is assigned on the node of the other side of
1345 * a p2p interface, the address should be a neighbor.
1347 if (ifp->if_flags & IFF_POINTOPOINT) {
1349 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1350 if (ifa->ifa_addr->sa_family != addr->sin6_family)
1352 if (ifa->ifa_dstaddr != NULL &&
1353 sa_equal(addr, ifa->ifa_dstaddr)) {
1354 IF_ADDR_RUNLOCK(ifp);
1358 IF_ADDR_RUNLOCK(ifp);
1362 * If the default router list is empty, all addresses are regarded
1363 * as on-link, and thus, as a neighbor.
1365 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1366 TAILQ_EMPTY(&V_nd_defrouter) &&
1367 V_nd6_defifindex == ifp->if_index) {
1376 * Detect if a given IPv6 address identifies a neighbor on a given link.
1377 * XXX: should take care of the destination of a p2p link?
1380 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1382 struct llentry *lle;
1385 IF_AFDATA_UNLOCK_ASSERT(ifp);
1386 if (nd6_is_new_addr_neighbor(addr, ifp))
1390 * Even if the address matches none of our addresses, it might be
1391 * in the neighbor cache.
1393 IF_AFDATA_RLOCK(ifp);
1394 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1398 IF_AFDATA_RUNLOCK(ifp);
1403 * Free an nd6 llinfo entry.
1404 * Since the function would cause significant changes in the kernel, DO NOT
1405 * make it global, unless you have a strong reason for the change, and are sure
1406 * that the change is safe.
1408 * Set noinline to be dtrace-friendly
1410 static __noinline void
1411 nd6_free(struct llentry **lnp, int gc)
1415 struct nd_defrouter *dr;
1420 LLE_WLOCK_ASSERT(ln);
1423 ifp = lltable_get_ifp(ln->lle_tbl);
1424 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0)
1425 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp);
1430 if ((ln->la_flags & LLE_DELETED) == 0)
1431 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
1434 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1435 * even though it is not harmful, it was not really necessary.
1439 nd6_llinfo_settimer_locked(ln, -1);
1441 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1442 if (dr != NULL && dr->expire &&
1443 ln->ln_state == ND6_LLINFO_STALE && gc) {
1445 * If the reason for the deletion is just garbage
1446 * collection, and the neighbor is an active default
1447 * router, do not delete it. Instead, reset the GC
1448 * timer using the router's lifetime.
1449 * Simply deleting the entry would affect default
1450 * router selection, which is not necessarily a good
1451 * thing, especially when we're using router preference
1453 * XXX: the check for ln_state would be redundant,
1454 * but we intentionally keep it just in case.
1456 if (dr->expire > time_uptime)
1457 nd6_llinfo_settimer_locked(ln,
1458 (dr->expire - time_uptime) * hz);
1460 nd6_llinfo_settimer_locked(ln,
1461 (long)V_nd6_gctimer * hz);
1471 * Unreachablity of a router might affect the default
1472 * router selection and on-link detection of advertised
1477 * Temporarily fake the state to choose a new default
1478 * router and to perform on-link determination of
1479 * prefixes correctly.
1480 * Below the state will be set correctly,
1481 * or the entry itself will be deleted.
1483 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1486 if (ln->ln_router || dr) {
1489 * We need to unlock to avoid a LOR with rt6_flush() with the
1490 * rnh and for the calls to pfxlist_onlink_check() and
1491 * defrouter_select_fib() in the block further down for calls
1492 * into nd6_lookup(). We still hold a ref.
1497 * rt6_flush must be called whether or not the neighbor
1498 * is in the Default Router List.
1499 * See a corresponding comment in nd6_na_input().
1501 rt6_flush(&ln->r_l3addr.addr6, ifp);
1506 * Since defrouter_select_fib() does not affect the
1507 * on-link determination and MIP6 needs the check
1508 * before the default router selection, we perform
1511 pfxlist_onlink_check();
1514 * Refresh default router list.
1516 defrouter_select_fib(dr->ifp->if_fib);
1520 * If this entry was added by an on-link redirect, remove the
1521 * corresponding host route.
1523 if (ln->la_flags & LLE_REDIRECT)
1524 nd6_free_redirect(ln);
1526 if (ln->ln_router || dr)
1531 * Save to unlock. We still hold an extra reference and will not
1532 * free(9) in llentry_free() if someone else holds one as well.
1535 IF_AFDATA_LOCK(ifp);
1537 /* Guard against race with other llentry_free(). */
1538 if (ln->la_flags & LLE_LINKED) {
1539 /* Remove callout reference */
1541 lltable_unlink_entry(ln->lle_tbl, ln);
1543 IF_AFDATA_UNLOCK(ifp);
1551 nd6_isdynrte(const struct rtentry *rt, void *xap)
1554 if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC))
1560 * Remove the rtentry for the given llentry,
1561 * both of which were installed by a redirect.
1564 nd6_free_redirect(const struct llentry *ln)
1567 struct sockaddr_in6 sin6;
1568 struct rt_addrinfo info;
1570 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
1571 memset(&info, 0, sizeof(info));
1572 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6;
1573 info.rti_filter = nd6_isdynrte;
1575 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
1576 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum);
1580 * Rejuvenate this function for routing operations related
1584 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1586 struct sockaddr_in6 *gateway;
1587 struct nd_defrouter *dr;
1590 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1601 * Only indirect routes are interesting.
1603 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1606 * check for default route
1608 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1609 &SIN6(rt_key(rt))->sin6_addr)) {
1610 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1622 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
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;
1629 if (ifp->if_afdata[AF_INET6] == NULL)
1630 return (EPFNOSUPPORT);
1632 case OSIOCGIFINFO_IN6:
1634 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1635 bzero(&ND, sizeof(ND));
1636 ND.linkmtu = IN6_LINKMTU(ifp);
1637 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1638 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1639 ND.reachable = ND_IFINFO(ifp)->reachable;
1640 ND.retrans = ND_IFINFO(ifp)->retrans;
1641 ND.flags = ND_IFINFO(ifp)->flags;
1642 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1643 ND.chlim = ND_IFINFO(ifp)->chlim;
1645 case SIOCGIFINFO_IN6:
1646 ND = *ND_IFINFO(ifp);
1648 case SIOCSIFINFO_IN6:
1650 * used to change host variables from userland.
1651 * intended for a use on router to reflect RA configurations.
1653 /* 0 means 'unspecified' */
1654 if (ND.linkmtu != 0) {
1655 if (ND.linkmtu < IPV6_MMTU ||
1656 ND.linkmtu > IN6_LINKMTU(ifp)) {
1660 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1663 if (ND.basereachable != 0) {
1664 int obasereachable = ND_IFINFO(ifp)->basereachable;
1666 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1667 if (ND.basereachable != obasereachable)
1668 ND_IFINFO(ifp)->reachable =
1669 ND_COMPUTE_RTIME(ND.basereachable);
1671 if (ND.retrans != 0)
1672 ND_IFINFO(ifp)->retrans = ND.retrans;
1674 ND_IFINFO(ifp)->chlim = ND.chlim;
1676 case SIOCSIFINFO_FLAGS:
1679 struct in6_ifaddr *ia;
1681 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1682 !(ND.flags & ND6_IFF_IFDISABLED)) {
1683 /* ifdisabled 1->0 transision */
1686 * If the interface is marked as ND6_IFF_IFDISABLED and
1687 * has an link-local address with IN6_IFF_DUPLICATED,
1688 * do not clear ND6_IFF_IFDISABLED.
1689 * See RFC 4862, Section 5.4.5.
1692 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1693 if (ifa->ifa_addr->sa_family != AF_INET6)
1695 ia = (struct in6_ifaddr *)ifa;
1696 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1697 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1700 IF_ADDR_RUNLOCK(ifp);
1703 /* LLA is duplicated. */
1704 ND.flags |= ND6_IFF_IFDISABLED;
1705 log(LOG_ERR, "Cannot enable an interface"
1706 " with a link-local address marked"
1709 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1710 if (ifp->if_flags & IFF_UP)
1713 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1714 (ND.flags & ND6_IFF_IFDISABLED)) {
1715 /* ifdisabled 0->1 transision */
1716 /* Mark all IPv6 address as tentative. */
1718 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1719 if (V_ip6_dad_count > 0 &&
1720 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1722 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1724 if (ifa->ifa_addr->sa_family !=
1727 ia = (struct in6_ifaddr *)ifa;
1728 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1730 IF_ADDR_RUNLOCK(ifp);
1734 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1735 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1736 /* auto_linklocal 0->1 transision */
1738 /* If no link-local address on ifp, configure */
1739 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1740 in6_ifattach(ifp, NULL);
1741 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1742 ifp->if_flags & IFF_UP) {
1744 * When the IF already has
1745 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1746 * address is assigned, and IFF_UP, try to
1750 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
1752 if (ifa->ifa_addr->sa_family !=
1755 ia = (struct in6_ifaddr *)ifa;
1756 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1759 IF_ADDR_RUNLOCK(ifp);
1761 /* No LLA is configured. */
1762 in6_ifattach(ifp, NULL);
1766 ND_IFINFO(ifp)->flags = ND.flags;
1769 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1770 /* sync kernel routing table with the default router list */
1774 case SIOCSPFXFLUSH_IN6:
1776 /* flush all the prefix advertised by routers */
1777 struct in6_ifaddr *ia, *ia_next;
1778 struct nd_prefix *pr, *next;
1779 struct nd_prhead prl;
1784 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1785 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1787 nd6_prefix_unlink(pr, &prl);
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,
1796 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1799 if (ia->ia6_ndpr == pr)
1800 in6_purgeaddr(&ia->ia_ifa);
1806 case SIOCSRTRFLUSH_IN6:
1808 /* flush all the default routers */
1809 struct nd_drhead drq;
1810 struct nd_defrouter *dr;
1817 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL)
1818 defrouter_unlink(dr, &drq);
1820 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
1821 TAILQ_REMOVE(&drq, dr, dr_entry);
1828 case SIOCGNBRINFO_IN6:
1831 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1833 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1836 IF_AFDATA_RLOCK(ifp);
1837 ln = nd6_lookup(&nb_addr, 0, ifp);
1838 IF_AFDATA_RUNLOCK(ifp);
1844 nbi->state = ln->ln_state;
1845 nbi->asked = ln->la_asked;
1846 nbi->isrouter = ln->ln_router;
1847 if (ln->la_expire == 0)
1850 nbi->expire = ln->la_expire + ln->lle_remtime / hz +
1851 (time_second - time_uptime);
1855 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1856 ndif->ifindex = V_nd6_defifindex;
1858 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1859 return (nd6_setdefaultiface(ndif->ifindex));
1865 * Calculates new isRouter value based on provided parameters and
1869 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
1874 * ICMP6 type dependent behavior.
1876 * NS: clear IsRouter if new entry
1877 * RS: clear IsRouter
1878 * RA: set IsRouter if there's lladdr
1879 * redir: clear IsRouter if new entry
1882 * The spec says that we must set IsRouter in the following cases:
1883 * - If lladdr exist, set IsRouter. This means (1-5).
1884 * - If it is old entry (!newentry), set IsRouter. This means (7).
1885 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1886 * A quetion arises for (1) case. (1) case has no lladdr in the
1887 * neighbor cache, this is similar to (6).
1888 * This case is rare but we figured that we MUST NOT set IsRouter.
1890 * is_new old_addr new_addr NS RS RA redir
1897 * 1 -- n (6) c c c s
1898 * 1 -- y (7) c c s c s
1902 switch (type & 0xff) {
1903 case ND_NEIGHBOR_SOLICIT:
1905 * New entry must have is_router flag cleared.
1907 if (is_new) /* (6-7) */
1912 * If the icmp is a redirect to a better router, always set the
1913 * is_router flag. Otherwise, if the entry is newly created,
1914 * clear the flag. [RFC 2461, sec 8.3]
1916 if (code == ND_REDIRECT_ROUTER)
1919 if (is_new) /* (6-7) */
1923 case ND_ROUTER_SOLICIT:
1925 * is_router flag must always be cleared.
1929 case ND_ROUTER_ADVERT:
1931 * Mark an entry with lladdr as a router.
1933 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */
1934 (is_new && new_addr)) { /* (7) */
1944 * Create neighbor cache entry and cache link-layer address,
1945 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1948 * code - type dependent information
1952 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1953 int lladdrlen, int type, int code)
1955 struct llentry *ln = NULL, *ln_tmp;
1961 uint16_t router = 0;
1962 struct sockaddr_in6 sin6;
1963 struct mbuf *chain = NULL;
1964 u_char linkhdr[LLE_MAX_LINKHDR];
1968 IF_AFDATA_UNLOCK_ASSERT(ifp);
1970 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1971 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1973 /* nothing must be updated for unspecified address */
1974 if (IN6_IS_ADDR_UNSPECIFIED(from))
1978 * Validation about ifp->if_addrlen and lladdrlen must be done in
1981 * XXX If the link does not have link-layer adderss, what should
1982 * we do? (ifp->if_addrlen == 0)
1983 * Spec says nothing in sections for RA, RS and NA. There's small
1984 * description on it in NS section (RFC 2461 7.2.3).
1986 flags = lladdr ? LLE_EXCLUSIVE : 0;
1987 IF_AFDATA_RLOCK(ifp);
1988 ln = nd6_lookup(from, flags, ifp);
1989 IF_AFDATA_RUNLOCK(ifp);
1992 flags |= LLE_EXCLUSIVE;
1993 ln = nd6_alloc(from, 0, ifp);
1998 * Since we already know all the data for the new entry,
1999 * fill it before insertion.
2001 if (lladdr != NULL) {
2002 linkhdrsize = sizeof(linkhdr);
2003 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
2004 linkhdr, &linkhdrsize, &lladdr_off) != 0)
2006 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
2010 IF_AFDATA_WLOCK(ifp);
2012 /* Prefer any existing lle over newly-created one */
2013 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp);
2015 lltable_link_entry(LLTABLE6(ifp), ln);
2016 IF_AFDATA_WUNLOCK(ifp);
2017 if (ln_tmp == NULL) {
2018 /* No existing lle, mark as new entry (6,7) */
2020 if (lladdr != NULL) { /* (7) */
2021 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2022 EVENTHANDLER_INVOKE(lle_event, ln,
2026 lltable_free_entry(LLTABLE6(ifp), ln);
2031 /* do nothing if static ndp is set */
2032 if ((ln->la_flags & LLE_STATIC)) {
2033 if (flags & LLE_EXCLUSIVE)
2040 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
2041 if (olladdr && lladdr) {
2042 llchange = bcmp(lladdr, ln->ll_addr,
2044 } else if (!olladdr && lladdr)
2050 * newentry olladdr lladdr llchange (*=record)
2053 * 0 n y y (3) * STALE
2055 * 0 y y y (5) * STALE
2056 * 1 -- n -- (6) NOSTATE(= PASSIVE)
2057 * 1 -- y -- (7) * STALE
2061 if (is_newentry == 0 && llchange != 0) {
2062 do_update = 1; /* (3,5) */
2065 * Record source link-layer address
2066 * XXX is it dependent to ifp->if_type?
2068 linkhdrsize = sizeof(linkhdr);
2069 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
2070 linkhdr, &linkhdrsize, &lladdr_off) != 0)
2073 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
2075 /* Entry was deleted */
2079 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
2081 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2083 if (ln->la_hold != NULL)
2084 nd6_grab_holdchain(ln, &chain, &sin6);
2087 /* Calculates new router status */
2088 router = nd6_is_router(type, code, is_newentry, olladdr,
2089 lladdr != NULL ? 1 : 0, ln->ln_router);
2091 ln->ln_router = router;
2092 /* Mark non-router redirects with special flag */
2093 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
2094 ln->la_flags |= LLE_REDIRECT;
2096 if (flags & LLE_EXCLUSIVE)
2102 nd6_flush_holdchain(ifp, chain, &sin6);
2105 * When the link-layer address of a router changes, select the
2106 * best router again. In particular, when the neighbor entry is newly
2107 * created, it might affect the selection policy.
2108 * Question: can we restrict the first condition to the "is_newentry"
2110 * XXX: when we hear an RA from a new router with the link-layer
2111 * address option, defrouter_select_fib() is called twice, since
2112 * defrtrlist_update called the function as well. However, I believe
2113 * we can compromise the overhead, since it only happens the first
2115 * XXX: although defrouter_select_fib() should not have a bad effect
2116 * for those are not autoconfigured hosts, we explicitly avoid such
2119 if ((do_update || is_newentry) && router &&
2120 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
2122 * guaranteed recursion
2124 defrouter_select_fib(ifp->if_fib);
2129 nd6_slowtimo(void *arg)
2131 CURVNET_SET((struct vnet *) arg);
2132 struct nd_ifinfo *nd6if;
2135 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
2136 nd6_slowtimo, curvnet);
2137 IFNET_RLOCK_NOSLEEP();
2138 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2139 if (ifp->if_afdata[AF_INET6] == NULL)
2141 nd6if = ND_IFINFO(ifp);
2142 if (nd6if->basereachable && /* already initialized */
2143 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
2145 * Since reachable time rarely changes by router
2146 * advertisements, we SHOULD insure that a new random
2147 * value gets recomputed at least once every few hours.
2150 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
2151 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
2154 IFNET_RUNLOCK_NOSLEEP();
2159 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
2160 struct sockaddr_in6 *sin6)
2163 LLE_WLOCK_ASSERT(ln);
2165 *chain = ln->la_hold;
2167 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6);
2169 if (ln->ln_state == ND6_LLINFO_STALE) {
2172 * The first time we send a packet to a
2173 * neighbor whose entry is STALE, we have
2174 * to change the state to DELAY and a sets
2175 * a timer to expire in DELAY_FIRST_PROBE_TIME
2176 * seconds to ensure do neighbor unreachability
2177 * detection on expiration.
2180 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
2185 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
2186 struct sockaddr_in6 *dst, struct route *ro)
2190 struct ip6_hdr *ip6;
2194 mac_netinet6_nd6_send(ifp, m);
2198 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
2199 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
2200 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
2201 * to be diverted to user space. When re-injected into the kernel,
2202 * send_output() will directly dispatch them to the outgoing interface.
2204 if (send_sendso_input_hook != NULL) {
2205 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
2207 ip6 = mtod(m, struct ip6_hdr *);
2208 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2209 /* Use the SEND socket */
2210 error = send_sendso_input_hook(m, ifp, SND_OUT,
2212 /* -1 == no app on SEND socket */
2213 if (error == 0 || error != -1)
2218 m_clrprotoflags(m); /* Avoid confusing lower layers. */
2219 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
2220 mtod(m, struct ip6_hdr *));
2222 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2225 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro);
2230 * Lookup link headerfor @sa_dst address. Stores found
2231 * data in @desten buffer. Copy of lle ln_flags can be also
2232 * saved in @pflags if @pflags is non-NULL.
2234 * If destination LLE does not exists or lle state modification
2235 * is required, call "slow" version.
2238 * - 0 on success (address copied to buffer).
2239 * - EWOULDBLOCK (no local error, but address is still unresolved)
2240 * - other errors (alloc failure, etc)
2243 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
2244 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags,
2245 struct llentry **plle)
2247 struct llentry *ln = NULL;
2248 const struct sockaddr_in6 *dst6;
2253 dst6 = (const struct sockaddr_in6 *)sa_dst;
2255 /* discard the packet if IPv6 operation is disabled on the interface */
2256 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2258 return (ENETDOWN); /* better error? */
2261 if (m != NULL && m->m_flags & M_MCAST) {
2262 switch (ifp->if_type) {
2266 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
2271 return (EAFNOSUPPORT);
2275 IF_AFDATA_RLOCK(ifp);
2276 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED,
2278 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
2279 /* Entry found, let's copy lle info */
2280 bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
2282 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
2283 /* Check if we have feedback request from nd6 timer */
2284 if (ln->r_skip_req != 0) {
2286 ln->r_skip_req = 0; /* Notify that entry was used */
2287 ln->lle_hittime = time_uptime;
2295 IF_AFDATA_RUNLOCK(ifp);
2297 } else if (plle && ln)
2299 IF_AFDATA_RUNLOCK(ifp);
2301 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle));
2306 * Do L2 address resolution for @sa_dst address. Stores found
2307 * address in @desten buffer. Copy of lle ln_flags can be also
2308 * saved in @pflags if @pflags is non-NULL.
2311 * Function assume that destination LLE does not exist,
2312 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired.
2314 * Set noinline to be dtrace-friendly
2316 static __noinline int
2317 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m,
2318 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags,
2319 struct llentry **plle)
2321 struct llentry *lle = NULL, *lle_tmp;
2322 struct in6_addr *psrc, src;
2323 int send_ns, ll_len;
2327 * Address resolution or Neighbor Unreachability Detection
2329 * At this point, the destination of the packet must be a unicast
2330 * or an anycast address(i.e. not a multicast).
2333 IF_AFDATA_RLOCK(ifp);
2334 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
2335 IF_AFDATA_RUNLOCK(ifp);
2336 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
2338 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2339 * the condition below is not very efficient. But we believe
2340 * it is tolerable, because this should be a rare case.
2342 lle = nd6_alloc(&dst->sin6_addr, 0, ifp);
2344 char ip6buf[INET6_ADDRSTRLEN];
2346 "nd6_output: can't allocate llinfo for %s "
2348 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
2353 IF_AFDATA_WLOCK(ifp);
2355 /* Prefer any existing entry over newly-created one */
2356 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
2357 if (lle_tmp == NULL)
2358 lltable_link_entry(LLTABLE6(ifp), lle);
2359 IF_AFDATA_WUNLOCK(ifp);
2360 if (lle_tmp != NULL) {
2361 lltable_free_entry(LLTABLE6(ifp), lle);
2368 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2378 LLE_WLOCK_ASSERT(lle);
2381 * The first time we send a packet to a neighbor whose entry is
2382 * STALE, we have to change the state to DELAY and a sets a timer to
2383 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2384 * neighbor unreachability detection on expiration.
2387 if (lle->ln_state == ND6_LLINFO_STALE)
2388 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
2391 * If the neighbor cache entry has a state other than INCOMPLETE
2392 * (i.e. its link-layer address is already resolved), just
2395 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
2396 if (flags & LLE_ADDRONLY) {
2397 lladdr = lle->ll_addr;
2398 ll_len = ifp->if_addrlen;
2400 lladdr = lle->r_linkdata;
2401 ll_len = lle->r_hdrlen;
2403 bcopy(lladdr, desten, ll_len);
2405 *pflags = lle->la_flags;
2415 * There is a neighbor cache entry, but no ethernet address
2416 * response yet. Append this latest packet to the end of the
2417 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen,
2418 * the oldest packet in the queue will be removed.
2421 if (lle->la_hold != NULL) {
2422 struct mbuf *m_hold;
2426 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2428 if (m_hold->m_nextpkt == NULL) {
2429 m_hold->m_nextpkt = m;
2433 while (i >= V_nd6_maxqueuelen) {
2434 m_hold = lle->la_hold;
2435 lle->la_hold = lle->la_hold->m_nextpkt;
2444 * If there has been no NS for the neighbor after entering the
2445 * INCOMPLETE state, send the first solicitation.
2446 * Note that for newly-created lle la_asked will be 0,
2447 * so we will transition from ND6_LLINFO_NOSTATE to
2448 * ND6_LLINFO_INCOMPLETE state here.
2452 if (lle->la_asked == 0) {
2455 psrc = nd6_llinfo_get_holdsrc(lle, &src);
2457 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
2461 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
2463 return (EWOULDBLOCK);
2467 * Do L2 address resolution for @sa_dst address. Stores found
2468 * address in @desten buffer. Copy of lle ln_flags can be also
2469 * saved in @pflags if @pflags is non-NULL.
2472 * - 0 on success (address copied to buffer).
2473 * - EWOULDBLOCK (no local error, but address is still unresolved)
2474 * - other errors (alloc failure, etc)
2477 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
2478 char *desten, uint32_t *pflags)
2482 flags |= LLE_ADDRONLY;
2483 error = nd6_resolve_slow(ifp, flags, NULL,
2484 (const struct sockaddr_in6 *)dst, desten, pflags, NULL);
2489 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain,
2490 struct sockaddr_in6 *dst)
2492 struct mbuf *m, *m_head;
2499 m_head = m_head->m_nextpkt;
2500 error = nd6_output_ifp(ifp, ifp, m, dst, NULL);
2505 * note that intermediate errors are blindly ignored
2511 nd6_need_cache(struct ifnet *ifp)
2514 * XXX: we currently do not make neighbor cache on any interface
2515 * other than Ethernet and GIF.
2518 * - unidirectional tunnels needs no ND
2520 switch (ifp->if_type) {
2524 case IFT_INFINIBAND:
2526 case IFT_PROPVIRTUAL:
2534 * Add pernament ND6 link-layer record for given
2535 * interface address.
2537 * Very similar to IPv4 arp_ifinit(), but:
2538 * 1) IPv6 DAD is performed in different place
2539 * 2) It is called by IPv6 protocol stack in contrast to
2540 * arp_ifinit() which is typically called in SIOCSIFADDR
2541 * driver ioctl handler.
2545 nd6_add_ifa_lle(struct in6_ifaddr *ia)
2548 struct llentry *ln, *ln_tmp;
2549 struct sockaddr *dst;
2551 ifp = ia->ia_ifa.ifa_ifp;
2552 if (nd6_need_cache(ifp) == 0)
2555 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
2556 dst = (struct sockaddr *)&ia->ia_addr;
2557 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
2561 IF_AFDATA_WLOCK(ifp);
2563 /* Unlink any entry if exists */
2564 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst);
2566 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
2567 lltable_link_entry(LLTABLE6(ifp), ln);
2568 IF_AFDATA_WUNLOCK(ifp);
2571 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
2572 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2576 llentry_free(ln_tmp);
2582 * Removes either all lle entries for given @ia, or lle
2583 * corresponding to @ia address.
2586 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
2588 struct sockaddr_in6 mask, addr;
2589 struct sockaddr *saddr, *smask;
2592 ifp = ia->ia_ifa.ifa_ifp;
2593 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
2594 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
2595 saddr = (struct sockaddr *)&addr;
2596 smask = (struct sockaddr *)&mask;
2599 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
2601 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
2605 clear_llinfo_pqueue(struct llentry *ln)
2607 struct mbuf *m_hold, *m_hold_next;
2609 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2610 m_hold_next = m_hold->m_nextpkt;
2617 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2618 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2620 SYSCTL_DECL(_net_inet6_icmp6);
2621 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2622 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2623 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2624 "NDP default router list");
2625 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2626 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2627 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix",
2629 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2630 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2631 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2632 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2635 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2637 struct in6_defrouter d;
2638 struct nd_defrouter *dr;
2641 if (req->newptr != NULL)
2644 error = sysctl_wire_old_buffer(req, 0);
2648 bzero(&d, sizeof(d));
2649 d.rtaddr.sin6_family = AF_INET6;
2650 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2653 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2654 d.rtaddr.sin6_addr = dr->rtaddr;
2655 error = sa6_recoverscope(&d.rtaddr);
2658 d.flags = dr->raflags;
2659 d.rtlifetime = dr->rtlifetime;
2660 d.expire = dr->expire + (time_second - time_uptime);
2661 d.if_index = dr->ifp->if_index;
2662 error = SYSCTL_OUT(req, &d, sizeof(d));
2671 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2673 struct in6_prefix p;
2674 struct sockaddr_in6 s6;
2675 struct nd_prefix *pr;
2676 struct nd_pfxrouter *pfr;
2679 char ip6buf[INET6_ADDRSTRLEN];
2684 error = sysctl_wire_old_buffer(req, 0);
2688 bzero(&p, sizeof(p));
2689 p.origin = PR_ORIG_RA;
2690 bzero(&s6, sizeof(s6));
2691 s6.sin6_family = AF_INET6;
2692 s6.sin6_len = sizeof(s6);
2695 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2696 p.prefix = pr->ndpr_prefix;
2697 if (sa6_recoverscope(&p.prefix)) {
2698 log(LOG_ERR, "scope error in prefix list (%s)\n",
2699 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2700 /* XXX: press on... */
2702 p.raflags = pr->ndpr_raf;
2703 p.prefixlen = pr->ndpr_plen;
2704 p.vltime = pr->ndpr_vltime;
2705 p.pltime = pr->ndpr_pltime;
2706 p.if_index = pr->ndpr_ifp->if_index;
2707 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2710 /* XXX: we assume time_t is signed. */
2712 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2713 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2714 p.expire = pr->ndpr_lastupdate +
2716 (time_second - time_uptime);
2718 p.expire = maxexpire;
2720 p.refcnt = pr->ndpr_addrcnt;
2721 p.flags = pr->ndpr_stateflags;
2723 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2725 error = SYSCTL_OUT(req, &p, sizeof(p));
2728 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2729 s6.sin6_addr = pfr->router->rtaddr;
2730 if (sa6_recoverscope(&s6))
2732 "scope error in prefix list (%s)\n",
2733 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2734 error = SYSCTL_OUT(req, &s6, sizeof(s6));