2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 * Copyright (C) 2001 WIDE Project. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)in.c 8.4 (Berkeley) 1/9/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
38 #define IN_HISTORICAL_NETS /* include class masks */
40 #include <sys/param.h>
41 #include <sys/eventhandler.h>
42 #include <sys/systm.h>
43 #include <sys/sockio.h>
44 #include <sys/malloc.h>
46 #include <sys/socket.h>
48 #include <sys/kernel.h>
51 #include <sys/rmlock.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
57 #include <net/if_var.h>
58 #include <net/if_arp.h>
59 #include <net/if_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/route.h>
63 #include <net/route/nhop.h>
64 #include <net/route/route_ctl.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/in.h>
69 #include <netinet/in_fib.h>
70 #include <netinet/in_var.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_carp.h>
74 #include <netinet/igmp_var.h>
75 #include <netinet/udp.h>
76 #include <netinet/udp_var.h>
78 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
79 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
80 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82 static void in_socktrim(struct sockaddr_in *);
83 static void in_purgemaddrs(struct ifnet *);
85 static bool ia_need_loopback_route(const struct in_ifaddr *);
87 VNET_DEFINE_STATIC(int, nosameprefix);
88 #define V_nosameprefix VNET(nosameprefix)
89 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
90 &VNET_NAME(nosameprefix), 0,
91 "Refuse to create same prefixes on different interfaces");
93 VNET_DEFINE_STATIC(bool, broadcast_lowest);
94 #define V_broadcast_lowest VNET(broadcast_lowest)
95 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
96 &VNET_NAME(broadcast_lowest), 0,
97 "Treat lowest address on a subnet (host 0) as broadcast");
99 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
100 #define V_ripcbinfo VNET(ripcbinfo)
102 static struct sx in_control_sx;
103 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
106 * Return 1 if an internet address is for a ``local'' host
107 * (one to which we have a connection).
110 in_localaddr(struct in_addr in)
112 struct rm_priotracker in_ifa_tracker;
113 u_long i = ntohl(in.s_addr);
114 struct in_ifaddr *ia;
116 IN_IFADDR_RLOCK(&in_ifa_tracker);
117 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
118 if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
119 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
123 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
128 * Return 1 if an internet address is for the local host and configured
129 * on one of its interfaces.
132 in_localip(struct in_addr in)
134 struct rm_priotracker in_ifa_tracker;
135 struct in_ifaddr *ia;
137 IN_IFADDR_RLOCK(&in_ifa_tracker);
138 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
139 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
140 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
144 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
149 * Return 1 if an internet address is configured on an interface.
152 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
155 struct in_ifaddr *ia;
159 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
160 if (ifa->ifa_addr->sa_family != AF_INET)
162 ia = (struct in_ifaddr *)ifa;
163 if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
171 * Return a reference to the interface address which is different to
172 * the supplied one but with same IP address value.
174 static struct in_ifaddr *
175 in_localip_more(struct in_ifaddr *original_ia)
177 struct rm_priotracker in_ifa_tracker;
178 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
179 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
180 struct in_ifaddr *ia;
182 IN_IFADDR_RLOCK(&in_ifa_tracker);
183 LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
184 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
185 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
186 if (!V_rt_add_addr_allfibs && (original_fib != fib))
188 if ((original_ia != ia) && (original_addr == addr)) {
189 ifa_ref(&ia->ia_ifa);
190 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
194 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
200 * Tries to find first IPv4 address in the provided fib.
201 * Prefers non-loopback addresses and return loopback IFF
202 * @loopback_ok is set.
204 * Returns ifa or NULL.
207 in_findlocal(uint32_t fibnum, bool loopback_ok)
209 struct rm_priotracker in_ifa_tracker;
210 struct in_ifaddr *ia = NULL, *ia_lo = NULL;
214 IN_IFADDR_RLOCK(&in_ifa_tracker);
215 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
216 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
217 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
220 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
225 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
234 * Determine whether an IP address is in a reserved set of addresses
235 * that may not be forwarded, or whether datagrams to that destination
239 in_canforward(struct in_addr in)
241 u_long i = ntohl(in.s_addr);
243 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
244 IN_ZERONET(i) || IN_LOOPBACK(i))
250 * Trim a mask in a sockaddr
253 in_socktrim(struct sockaddr_in *ap)
255 char *cplim = (char *) &ap->sin_addr;
256 char *cp = (char *) (&ap->sin_addr + 1);
259 while (--cp >= cplim)
261 (ap)->sin_len = cp - (char *) (ap) + 1;
267 * Generic internet control operations (ioctl's).
270 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
273 struct ifreq *ifr = (struct ifreq *)data;
274 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
275 struct epoch_tracker et;
277 struct in_ifaddr *ia;
281 return (EADDRNOTAVAIL);
283 struct ucred *cred = (td != NULL) ? td->td_ucred : NULL;
286 * Filter out 4 ioctls we implement directly. Forward the rest
287 * to specific functions and ifp->if_ioctl().
296 sx_xlock(&in_control_sx);
297 error = in_gifaddr_ioctl(cmd, data, ifp, cred);
298 sx_xunlock(&in_control_sx);
301 sx_xlock(&in_control_sx);
302 error = in_difaddr_ioctl(cmd, data, ifp, cred);
303 sx_xunlock(&in_control_sx);
305 case OSIOCAIFADDR: /* 9.x compat */
307 sx_xlock(&in_control_sx);
308 error = in_aifaddr_ioctl(cmd, data, ifp, cred);
309 sx_xunlock(&in_control_sx);
315 /* We no longer support that old commands. */
318 if (ifp->if_ioctl == NULL)
320 return ((*ifp->if_ioctl)(ifp, cmd, data));
323 if (addr->sin_addr.s_addr != INADDR_ANY &&
324 prison_check_ip4(cred, &addr->sin_addr) != 0)
325 return (EADDRNOTAVAIL);
328 * Find address for this interface, if it exists. If an
329 * address was specified, find that one instead of the
330 * first one on the interface, if possible.
333 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
334 if (ifa->ifa_addr->sa_family != AF_INET)
336 ia = (struct in_ifaddr *)ifa;
337 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
341 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
342 if (ifa->ifa_addr->sa_family == AF_INET) {
343 ia = (struct in_ifaddr *)ifa;
344 if (prison_check_ip4(cred,
345 &ia->ia_addr.sin_addr) == 0)
351 return (EADDRNOTAVAIL);
361 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
365 *addr = ia->ia_broadaddr;
369 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
373 *addr = ia->ia_dstaddr;
377 *addr = ia->ia_sockmask;
387 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
389 const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
390 const struct sockaddr_in *addr = &ifra->ifra_addr;
391 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
392 const struct sockaddr_in *mask = &ifra->ifra_mask;
393 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
394 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
395 struct epoch_tracker et;
397 struct in_ifaddr *ia;
401 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
406 * ifra_addr must be present and be of INET family.
407 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
409 if (addr->sin_len != sizeof(struct sockaddr_in) ||
410 addr->sin_family != AF_INET)
412 if (broadaddr->sin_len != 0 &&
413 (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
414 broadaddr->sin_family != AF_INET))
416 if (mask->sin_len != 0 &&
417 (mask->sin_len != sizeof(struct sockaddr_in) ||
418 mask->sin_family != AF_INET))
420 if ((ifp->if_flags & IFF_POINTOPOINT) &&
421 (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
422 dstaddr->sin_addr.s_addr == INADDR_ANY))
423 return (EDESTADDRREQ);
424 if (vhid != 0 && carp_attach_p == NULL)
425 return (EPROTONOSUPPORT);
428 * See whether address already exist.
433 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
434 struct in_ifaddr *it;
436 if (ifa->ifa_addr->sa_family != AF_INET)
439 it = (struct in_ifaddr *)ifa;
440 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
441 prison_check_ip4(cred, &addr->sin_addr) == 0)
449 (void )in_difaddr_ioctl(cmd, data, ifp, cred);
451 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
452 ia = (struct in_ifaddr *)ifa;
453 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
454 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
455 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
456 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
457 CALLOUT_RETURNUNLOCKED);
461 if (mask->sin_len != 0) {
462 ia->ia_sockmask = *mask;
463 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
465 in_addr_t i = ntohl(addr->sin_addr.s_addr);
468 * If netmask isn't supplied, use historical default.
469 * This is deprecated for interfaces other than loopback
470 * or point-to-point; warn in other cases. In the future
471 * we should return an error rather than warning.
473 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
474 printf("%s: set address: WARNING: network mask "
475 "should be specified; using historical default\n",
478 ia->ia_subnetmask = IN_CLASSA_NET;
479 else if (IN_CLASSB(i))
480 ia->ia_subnetmask = IN_CLASSB_NET;
482 ia->ia_subnetmask = IN_CLASSC_NET;
483 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
485 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
486 in_socktrim(&ia->ia_sockmask);
488 if (ifp->if_flags & IFF_BROADCAST) {
489 if (broadaddr->sin_len != 0) {
490 ia->ia_broadaddr = *broadaddr;
491 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
492 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
493 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
494 ia->ia_broadaddr.sin_family = AF_INET;
496 ia->ia_broadaddr.sin_addr.s_addr =
497 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
498 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
499 ia->ia_broadaddr.sin_family = AF_INET;
503 if (ifp->if_flags & IFF_POINTOPOINT)
504 ia->ia_dstaddr = *dstaddr;
507 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
512 /* if_addrhead is already referenced by ifa_alloc() */
514 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
515 IF_ADDR_WUNLOCK(ifp);
517 ifa_ref(ifa); /* in_ifaddrhead */
519 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
520 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
524 * Give the interface a chance to initialize
525 * if this is its first address,
526 * and to validate the address if necessary.
528 if (ifp->if_ioctl != NULL) {
529 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
535 * Add route for the network.
538 error = in_addprefix(ia);
544 * Add a loopback route to self.
546 if (vhid == 0 && ia_need_loopback_route(ia)) {
547 struct in_ifaddr *eia;
549 eia = in_localip_more(ia);
552 error = ifa_add_loopback_route((struct ifaddr *)ia,
553 (struct sockaddr *)&ia->ia_addr);
557 ifa_free(&eia->ia_ifa);
560 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
561 struct in_addr allhosts_addr;
562 struct in_ifinfo *ii;
564 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
565 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
567 error = in_joingroup(ifp, &allhosts_addr, NULL,
572 * Note: we don't need extra reference for ifa, since we called
573 * with sx lock held, and ifaddr can not be deleted in concurrent
576 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
582 (void )in_scrubprefix(ia, LLE_STATIC);
585 if (ia->ia_ifa.ifa_carp)
586 (*carp_detach_p)(&ia->ia_ifa, false);
589 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
590 IF_ADDR_WUNLOCK(ifp);
591 ifa_free(&ia->ia_ifa); /* if_addrhead */
594 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
595 LIST_REMOVE(ia, ia_hash);
597 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
603 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
605 const struct ifreq *ifr = (struct ifreq *)data;
606 const struct sockaddr_in *addr = (const struct sockaddr_in *)
609 struct in_ifaddr *ia;
610 bool deleteAny, iaIsLast;
614 error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
619 if (addr->sin_len != sizeof(struct sockaddr_in) ||
620 addr->sin_family != AF_INET)
628 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
629 struct in_ifaddr *it;
631 if (ifa->ifa_addr->sa_family != AF_INET)
634 it = (struct in_ifaddr *)ifa;
635 if (deleteAny && ia == NULL && (cred == NULL ||
636 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
639 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
640 (cred == NULL || prison_check_ip4(cred,
641 &addr->sin_addr) == 0))
649 IF_ADDR_WUNLOCK(ifp);
650 return (EADDRNOTAVAIL);
653 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
654 IF_ADDR_WUNLOCK(ifp);
655 ifa_free(&ia->ia_ifa); /* if_addrhead */
658 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
659 LIST_REMOVE(ia, ia_hash);
663 * in_scrubprefix() kills the interface route.
665 in_scrubprefix(ia, LLE_STATIC);
668 * in_ifadown gets rid of all the rest of
669 * the routes. This is not quite the right
670 * thing to do, but at least if we are running
671 * a routing process they will come back.
673 in_ifadown(&ia->ia_ifa, 1);
675 if (ia->ia_ifa.ifa_carp)
676 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
679 * If this is the last IPv4 address configured on this
680 * interface, leave the all-hosts group.
681 * No state-change report need be transmitted.
683 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
684 struct in_ifinfo *ii;
686 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
687 if (ii->ii_allhosts) {
688 (void)in_leavegroup(ii->ii_allhosts, NULL);
689 ii->ii_allhosts = NULL;
694 if (callout_stop(&ia->ia_garp_timer) == 1) {
695 ifa_free(&ia->ia_ifa);
697 IF_ADDR_WUNLOCK(ifp);
699 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
701 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
707 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
709 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
710 const struct sockaddr_in *addr = &ifra->ifra_addr;
711 struct epoch_tracker et;
713 struct in_ifaddr *ia;
716 * ifra_addr must be present and be of INET family.
718 if (addr->sin_len != sizeof(struct sockaddr_in) ||
719 addr->sin_family != AF_INET)
723 * See whether address exist.
727 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
728 struct in_ifaddr *it;
730 if (ifa->ifa_addr->sa_family != AF_INET)
733 it = (struct in_ifaddr *)ifa;
734 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
735 prison_check_ip4(cred, &addr->sin_addr) == 0) {
742 return (EADDRNOTAVAIL);
745 ifra->ifra_mask = ia->ia_sockmask;
746 if ((ifp->if_flags & IFF_POINTOPOINT) &&
747 ia->ia_dstaddr.sin_family == AF_INET)
748 ifra->ifra_dstaddr = ia->ia_dstaddr;
749 else if ((ifp->if_flags & IFF_BROADCAST) &&
750 ia->ia_broadaddr.sin_family == AF_INET)
751 ifra->ifra_broadaddr = ia->ia_broadaddr;
753 memset(&ifra->ifra_broadaddr, 0,
754 sizeof(ifra->ifra_broadaddr));
761 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
764 if (nh->nh_ifa == (struct ifaddr *)arg)
771 in_handle_prefix_route(uint32_t fibnum, int cmd,
772 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
778 /* Prepare gateway */
779 struct sockaddr_dl_short sdl = {
780 .sdl_family = AF_LINK,
781 .sdl_len = sizeof(struct sockaddr_dl_short),
782 .sdl_type = ifa->ifa_ifp->if_type,
783 .sdl_index = ifa->ifa_ifp->if_index,
786 struct rt_addrinfo info = {
789 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
791 [RTAX_DST] = (struct sockaddr *)dst,
792 [RTAX_NETMASK] = (struct sockaddr *)netmask,
793 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
795 /* Ensure we delete the prefix IFF prefix ifa matches */
796 .rti_filter = in_match_ifaddr,
797 .rti_filterdata = ifa,
800 return (rib_handle_ifaddr_info(fibnum, cmd, &info));
804 * Routing table interaction with interface addresses.
806 * In general, two types of routes needs to be installed:
807 * a) "interface" or "prefix" route, telling user that the addresses
808 * behind the ifa prefix are reached directly.
809 * b) "loopback" route installed for the ifa address, telling user that
810 * the address belongs to local system.
812 * Handling for (a) and (b) differs in multi-fib aspects, hence they
813 * are implemented in different functions below.
815 * The cases above may intersect - /32 interface aliases results in
816 * the same prefix produced by (a) and (b). This blurs the definition
817 * of the "loopback" route and complicate interactions. The interaction
818 * table is defined below. The case numbers are used in the multiple
819 * functions below to refer to the particular test case.
821 * There can be multiple options:
822 * 1) Adding address with prefix on non-p2p/non-loopback interface.
823 * Example: 192.0.2.1/24. Action:
824 * * add "prefix" route towards 192.0.2.0/24 via @ia interface,
825 * using @ia as an address source.
826 * * add "loopback" route towards 192.0.2.1 via V_loif, saving
827 * @ia ifp in the gateway and using @ia as an address source.
829 * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
830 * Example: 192.0.2.2/32. Action:
831 * * add "prefix" host route via V_loif, using @ia as an address source.
833 * 3) Adding address with or without prefix to p2p interface.
834 * Example: 10.0.0.1/24->10.0.0.2. Action:
835 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
836 * as an address source. Note: no sense in installing full /24 as the interface
838 * * add "loopback" route towards 10.0.9.1 via V_loif, saving
839 * @ia ifp in the gateway and using @ia as an address source.
841 * 4) Adding address with or without prefix to loopback interface.
842 * Example: 192.0.2.1/24. Action:
843 * * add "prefix" host route via @ia interface, using @ia as an address source.
844 * Note: Skip installing /24 prefix as it would introduce TTL loop
845 * for the traffic destined to these addresses.
849 * Checks if @ia needs to install loopback route to @ia address via
850 * ifa_maintain_loopback_route().
852 * Return true on success.
855 ia_need_loopback_route(const struct in_ifaddr *ia)
857 struct ifnet *ifp = ia->ia_ifp;
859 /* Case 4: Skip loopback interfaces */
860 if ((ifp->if_flags & IFF_LOOPBACK) ||
861 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
864 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */
865 if ((ifp->if_flags & IFF_POINTOPOINT) &&
866 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
869 /* Case 2: skip /32 prefixes */
870 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
871 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
878 * Calculate "prefix" route corresponding to @ia.
881 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
884 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
885 /* Case 3: return host route for dstaddr */
886 *prefix = ia->ia_dstaddr.sin_addr;
887 mask->s_addr = INADDR_BROADCAST;
888 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
889 /* Case 4: return host route for ifaddr */
890 *prefix = ia->ia_addr.sin_addr;
891 mask->s_addr = INADDR_BROADCAST;
893 /* Cases 1,2: return actual ia prefix */
894 *prefix = ia->ia_addr.sin_addr;
895 *mask = ia->ia_sockmask.sin_addr;
896 prefix->s_addr &= mask->s_addr;
901 * Adds or delete interface "prefix" route corresponding to @ifa.
902 * Returns 0 on success or errno.
905 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
907 struct ifaddr *ifa = &ia->ia_ifa;
908 struct in_addr daddr, maddr;
909 struct sockaddr_in *pmask;
910 struct epoch_tracker et;
913 ia_getrtprefix(ia, &daddr, &maddr);
915 struct sockaddr_in mask = {
916 .sin_family = AF_INET,
917 .sin_len = sizeof(struct sockaddr_in),
921 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
923 struct sockaddr_in dst = {
924 .sin_family = AF_INET,
925 .sin_len = sizeof(struct sockaddr_in),
926 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
929 struct ifnet *ifp = ia->ia_ifp;
931 if ((maddr.s_addr == INADDR_BROADCAST) &&
932 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
933 /* Case 2: host route on broadcast interface */
937 uint32_t fibnum = ifa->ifa_ifp->if_fib;
939 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
946 * Check if we have a route for the given prefix already.
949 in_hasrtprefix(struct in_ifaddr *target)
951 struct rm_priotracker in_ifa_tracker;
952 struct in_ifaddr *ia;
953 struct in_addr prefix, mask, p, m;
956 ia_getrtprefix(target, &prefix, &mask);
958 IN_IFADDR_RLOCK(&in_ifa_tracker);
959 /* Look for an existing address with the same prefix, mask, and fib */
960 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
961 ia_getrtprefix(ia, &p, &m);
963 if (prefix.s_addr != p.s_addr ||
964 mask.s_addr != m.s_addr)
967 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
971 * If we got a matching prefix route inserted by other
972 * interface address, we are done here.
974 if (ia->ia_flags & IFA_ROUTE) {
979 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
985 in_addprefix(struct in_ifaddr *target)
989 if (in_hasrtprefix(target)) {
993 rt_addrmsg(RTM_ADD, &target->ia_ifa,
994 target->ia_ifp->if_fib);
1000 * No-one seem to have this prefix route, so we try to insert it.
1002 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1003 error = in_handle_ifaddr_route(RTM_ADD, target);
1005 target->ia_flags |= IFA_ROUTE;
1010 * Removes either all lle entries for given @ia, or lle
1011 * corresponding to @ia address.
1014 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1016 struct sockaddr_in addr, mask;
1017 struct sockaddr *saddr, *smask;
1020 saddr = (struct sockaddr *)&addr;
1021 bzero(&addr, sizeof(addr));
1022 addr.sin_len = sizeof(addr);
1023 addr.sin_family = AF_INET;
1024 smask = (struct sockaddr *)&mask;
1025 bzero(&mask, sizeof(mask));
1026 mask.sin_len = sizeof(mask);
1027 mask.sin_family = AF_INET;
1028 mask.sin_addr.s_addr = ia->ia_subnetmask;
1033 * Remove all L2 entries matching given prefix.
1034 * Convert address to host representation to avoid
1035 * doing this on every callback. ia_subnetmask is already
1036 * stored in host representation.
1038 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1039 lltable_prefix_free(AF_INET, saddr, smask, flags);
1041 /* Remove interface address only */
1042 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1043 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1048 * If there is no other address in the system that can serve a route to the
1049 * same prefix, remove the route. Hand over the route to the new address
1053 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1055 struct rm_priotracker in_ifa_tracker;
1056 struct in_ifaddr *ia;
1057 struct in_addr prefix, mask, p, m;
1061 * Remove the loopback route to the interface address.
1063 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1064 struct in_ifaddr *eia;
1066 eia = in_localip_more(target);
1069 error = ifa_switch_loopback_route((struct ifaddr *)eia,
1070 (struct sockaddr *)&target->ia_addr);
1071 ifa_free(&eia->ia_ifa);
1073 error = ifa_del_loopback_route((struct ifaddr *)target,
1074 (struct sockaddr *)&target->ia_addr);
1078 ia_getrtprefix(target, &prefix, &mask);
1080 if ((target->ia_flags & IFA_ROUTE) == 0) {
1081 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1084 * Removing address from !IFF_UP interface or
1085 * prefix which exists on other interface (along with route).
1086 * No entries should exist here except target addr.
1087 * Given that, delete this entry only.
1089 in_scrubprefixlle(target, 0, flags);
1093 IN_IFADDR_RLOCK(&in_ifa_tracker);
1094 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1095 ia_getrtprefix(ia, &p, &m);
1097 if (prefix.s_addr != p.s_addr ||
1098 mask.s_addr != m.s_addr)
1101 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1105 * If we got a matching prefix address, move IFA_ROUTE and
1106 * the route itself to it. Make sure that routing daemons
1109 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1110 ifa_ref(&ia->ia_ifa);
1111 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
1112 error = in_handle_ifaddr_route(RTM_DELETE, target);
1114 target->ia_flags &= ~IFA_ROUTE;
1116 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1118 /* Scrub all entries IFF interface is different */
1119 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1121 error = in_handle_ifaddr_route(RTM_ADD, ia);
1123 ia->ia_flags |= IFA_ROUTE;
1125 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1127 ifa_free(&ia->ia_ifa);
1131 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
1134 * remove all L2 entries on the given prefix
1136 in_scrubprefixlle(target, 1, flags);
1139 * As no-one seem to have this prefix, we can remove the route.
1141 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1142 error = in_handle_ifaddr_route(RTM_DELETE, target);
1144 target->ia_flags &= ~IFA_ROUTE;
1146 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1151 in_ifscrub_all(void)
1154 struct ifaddr *ifa, *nifa;
1155 struct ifaliasreq ifr;
1158 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1159 /* Cannot lock here - lock recursion. */
1160 /* NET_EPOCH_ENTER(et); */
1161 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1162 if (ifa->ifa_addr->sa_family != AF_INET)
1166 * This is ugly but the only way for legacy IP to
1167 * cleanly remove addresses and everything attached.
1169 bzero(&ifr, sizeof(ifr));
1170 ifr.ifra_addr = *ifa->ifa_addr;
1171 if (ifa->ifa_dstaddr)
1172 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1173 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1176 /* NET_EPOCH_EXIT(et); */
1177 in_purgemaddrs(ifp);
1178 igmp_domifdetach(ifp);
1184 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1187 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1189 * Optionally check for old-style (host 0) broadcast, but
1190 * taking into account that RFC 3021 obsoletes it.
1192 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK &&
1193 ntohl(in.s_addr) == ia->ia_subnet)) &&
1195 * Check for an all one subnetmask. These
1196 * only exist when an interface gets a secondary
1199 ia->ia_subnetmask != (u_long)0xffffffff);
1203 * Return 1 if the address might be a local broadcast address.
1206 in_broadcast(struct in_addr in, struct ifnet *ifp)
1213 if (in.s_addr == INADDR_BROADCAST ||
1214 in.s_addr == INADDR_ANY)
1216 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1220 * Look through the list of addresses for a match
1221 * with a broadcast address.
1223 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1224 if (ifa->ifa_addr->sa_family == AF_INET &&
1225 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
1233 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1236 in_ifdetach(struct ifnet *ifp)
1239 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1240 in_pcbpurgeif0(&V_udbinfo, ifp);
1241 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1242 in_purgemaddrs(ifp);
1246 * Make sure all multicast deletions invoking if_ioctl() are
1247 * completed before returning. Else we risk accessing a freed
1248 * ifnet structure pointer.
1250 inm_release_wait(NULL);
1254 * Delete all IPv4 multicast address records, and associated link-layer
1255 * multicast address records, associated with ifp.
1256 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1257 * XXX This should not race with ifma_protospec being set during
1258 * a new allocation, if it does, we have bigger problems.
1261 in_purgemaddrs(struct ifnet *ifp)
1263 struct epoch_tracker et;
1264 struct in_multi_head purgeinms;
1265 struct in_multi *inm;
1266 struct ifmultiaddr *ifma;
1268 SLIST_INIT(&purgeinms);
1269 IN_MULTI_LIST_LOCK();
1272 * Extract list of in_multi associated with the detaching ifp
1273 * which the PF_INET layer is about to release.
1274 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1275 * by code further down.
1278 NET_EPOCH_ENTER(et);
1279 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1280 inm = inm_ifmultiaddr_get_inm(ifma);
1283 inm_rele_locked(&purgeinms, inm);
1286 IF_ADDR_WUNLOCK(ifp);
1288 inm_release_list_deferred(&purgeinms);
1290 IN_MULTI_LIST_UNLOCK();
1294 struct llentry base;
1297 #define IN_LLTBL_DEFAULT_HSIZE 32
1298 #define IN_LLTBL_HASH(k, h) \
1299 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1302 * Do actual deallocation of @lle.
1305 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1307 struct llentry *lle;
1309 lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1310 LLE_LOCK_DESTROY(lle);
1311 LLE_REQ_DESTROY(lle);
1312 free(lle, M_LLTABLE);
1316 * Called by LLE_FREE_LOCKED when number of references
1320 in_lltable_destroy_lle(struct llentry *lle)
1324 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1327 static struct llentry *
1328 in_lltable_new(struct in_addr addr4, u_int flags)
1330 struct in_llentry *lle;
1332 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1333 if (lle == NULL) /* NB: caller generates msg */
1337 * For IPv4 this will trigger "arpresolve" to generate
1340 lle->base.la_expire = time_uptime; /* mark expired */
1341 lle->base.r_l3addr.addr4 = addr4;
1342 lle->base.lle_refcnt = 1;
1343 lle->base.lle_free = in_lltable_destroy_lle;
1344 LLE_LOCK_INIT(&lle->base);
1345 LLE_REQ_INIT(&lle->base);
1346 callout_init(&lle->base.lle_timer, 1);
1348 return (&lle->base);
1351 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \
1352 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1355 in_lltable_match_prefix(const struct sockaddr *saddr,
1356 const struct sockaddr *smask, u_int flags, struct llentry *lle)
1358 struct in_addr addr, mask, lle_addr;
1360 addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1361 mask = ((const struct sockaddr_in *)smask)->sin_addr;
1362 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1364 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1367 if (lle->la_flags & LLE_IFADDR) {
1369 * Delete LLE_IFADDR records IFF address & flag matches.
1370 * Note that addr is the interface address within prefix
1372 * Note also we should handle 'ifdown' cases without removing
1375 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1380 /* flags & LLE_STATIC means deleting both dynamic and static entries */
1381 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1388 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1390 size_t pkts_dropped;
1392 LLE_WLOCK_ASSERT(lle);
1393 KASSERT(llt != NULL, ("lltable is NULL"));
1395 /* Unlink entry from table if not already */
1396 if ((lle->la_flags & LLE_LINKED) != 0) {
1397 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1398 lltable_unlink_entry(llt, lle);
1401 /* Drop hold queue */
1402 pkts_dropped = llentry_free(lle);
1403 ARPSTAT_ADD(dropped, pkts_dropped);
1407 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1409 struct nhop_object *nh;
1410 struct in_addr addr;
1412 KASSERT(l3addr->sa_family == AF_INET,
1413 ("sin_family %d", l3addr->sa_family));
1415 addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1417 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1422 * If the gateway for an existing host route matches the target L3
1423 * address, which is a special route inserted by some implementation
1424 * such as MANET, and the interface is of the correct type, then
1425 * allow for ARP to proceed.
1427 if (nh->nh_flags & NHF_GATEWAY) {
1428 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1429 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1430 memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1431 sizeof(in_addr_t)) != 0) {
1437 * Make sure that at least the destination address is covered
1438 * by the route. This is for handling the case where 2 or more
1439 * interfaces have the same prefix. An incoming packet arrives
1440 * on one interface and the corresponding outgoing packet leaves
1441 * another interface.
1443 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1444 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1445 struct in_addr dst_addr, mask_addr;
1451 * ifaof_ifpforaddr() returns _best matching_ IFA.
1452 * It is possible that ifa prefix does not cover our address.
1453 * Explicitly verify and fail if that's the case.
1455 dst_addr = IA_SIN(ia)->sin_addr;
1456 mask_addr.s_addr = htonl(ia->ia_subnetmask);
1458 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1465 static inline uint32_t
1466 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1469 return (IN_LLTBL_HASH(dst.s_addr, hsize));
1473 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1476 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1480 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1482 struct sockaddr_in *sin;
1484 sin = (struct sockaddr_in *)sa;
1485 bzero(sin, sizeof(*sin));
1486 sin->sin_family = AF_INET;
1487 sin->sin_len = sizeof(*sin);
1488 sin->sin_addr = lle->r_l3addr.addr4;
1491 static inline struct llentry *
1492 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1494 struct llentry *lle;
1495 struct llentries *lleh;
1498 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1499 lleh = &llt->lle_head[hashidx];
1500 CK_LIST_FOREACH(lle, lleh, lle_next) {
1501 if (lle->la_flags & LLE_DELETED)
1503 if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1511 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1514 lle->la_flags |= LLE_DELETED;
1515 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1517 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1522 static struct llentry *
1523 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1525 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1526 struct ifnet *ifp = llt->llt_ifp;
1527 struct llentry *lle;
1528 char linkhdr[LLE_MAX_LINKHDR];
1532 KASSERT(l3addr->sa_family == AF_INET,
1533 ("sin_family %d", l3addr->sa_family));
1536 * A route that covers the given address must have
1537 * been installed 1st because we are doing a resolution,
1540 if (!(flags & LLE_IFADDR) &&
1541 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1544 lle = in_lltable_new(sin->sin_addr, flags);
1546 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1549 lle->la_flags = flags;
1550 if (flags & LLE_STATIC)
1551 lle->r_flags |= RLLE_VALID;
1552 if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1553 linkhdrsize = LLE_MAX_LINKHDR;
1554 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1555 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1556 in_lltable_free_entry(llt, lle);
1559 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1561 lle->la_flags |= LLE_STATIC;
1562 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1569 * Return NULL if not found or marked for deletion.
1570 * If found return lle read locked.
1572 static struct llentry *
1573 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1575 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1576 struct llentry *lle;
1578 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1579 KASSERT(l3addr->sa_family == AF_INET,
1580 ("sin_family %d", l3addr->sa_family));
1581 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1582 (LLE_UNLOCKED | LLE_EXCLUSIVE),
1583 ("wrong lle request flags: %#x", flags));
1585 lle = in_lltable_find_dst(llt, sin->sin_addr);
1588 if (flags & LLE_UNLOCKED)
1591 if (flags & LLE_EXCLUSIVE)
1597 * If the afdata lock is not held, the LLE may have been unlinked while
1598 * we were blocked on the LLE lock. Check for this case.
1600 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1601 if (flags & LLE_EXCLUSIVE)
1611 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1612 struct sysctl_req *wr)
1614 struct ifnet *ifp = llt->llt_ifp;
1617 struct rt_msghdr rtm;
1618 struct sockaddr_in sin;
1619 struct sockaddr_dl sdl;
1621 struct sockaddr_dl *sdl;
1624 bzero(&arpc, sizeof(arpc));
1625 /* skip deleted entries */
1626 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1628 /* Skip if jailed and not a valid IP of the prison. */
1629 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1630 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1633 * produce a msg made of:
1635 * struct sockaddr_in; (IPv4)
1636 * struct sockaddr_dl;
1638 arpc.rtm.rtm_msglen = sizeof(arpc);
1639 arpc.rtm.rtm_version = RTM_VERSION;
1640 arpc.rtm.rtm_type = RTM_GET;
1641 arpc.rtm.rtm_flags = RTF_UP;
1642 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1645 if (lle->la_flags & LLE_PUB)
1646 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1649 sdl->sdl_family = AF_LINK;
1650 sdl->sdl_len = sizeof(*sdl);
1651 sdl->sdl_index = ifp->if_index;
1652 sdl->sdl_type = ifp->if_type;
1653 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1654 sdl->sdl_alen = ifp->if_addrlen;
1655 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1658 bzero(LLADDR(sdl), ifp->if_addrlen);
1661 arpc.rtm.rtm_rmx.rmx_expire =
1662 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1663 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1664 if (lle->la_flags & LLE_STATIC)
1665 arpc.rtm.rtm_flags |= RTF_STATIC;
1666 if (lle->la_flags & LLE_IFADDR)
1667 arpc.rtm.rtm_flags |= RTF_PINNED;
1668 arpc.rtm.rtm_index = ifp->if_index;
1669 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1674 static struct lltable *
1675 in_lltattach(struct ifnet *ifp)
1677 struct lltable *llt;
1679 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1680 llt->llt_af = AF_INET;
1683 llt->llt_lookup = in_lltable_lookup;
1684 llt->llt_alloc_entry = in_lltable_alloc;
1685 llt->llt_delete_entry = in_lltable_delete_entry;
1686 llt->llt_dump_entry = in_lltable_dump_entry;
1687 llt->llt_hash = in_lltable_hash;
1688 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1689 llt->llt_free_entry = in_lltable_free_entry;
1690 llt->llt_match_prefix = in_lltable_match_prefix;
1691 llt->llt_mark_used = llentry_mark_used;
1698 in_lltable_get(struct ifnet *ifp)
1700 struct lltable *llt = NULL;
1702 void *afdata_ptr = ifp->if_afdata[AF_INET];
1703 if (afdata_ptr != NULL)
1704 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1709 in_domifattach(struct ifnet *ifp)
1711 struct in_ifinfo *ii;
1713 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1715 ii->ii_llt = in_lltattach(ifp);
1716 ii->ii_igmp = igmp_domifattach(ifp);
1722 in_domifdetach(struct ifnet *ifp, void *aux)
1724 struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1726 igmp_domifdetach(ifp);
1727 lltable_free(ii->ii_llt);