1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
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.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 *
32 *	@(#)in.c	8.4 (Berkeley) 1/9/95
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD$");
37
38#include "opt_mpath.h"
39
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>
45#include <sys/priv.h>
46#include <sys/socket.h>
47#include <sys/jail.h>
48#include <sys/kernel.h>
49#include <sys/lock.h>
50#include <sys/proc.h>
51#include <sys/rmlock.h>
52#include <sys/sysctl.h>
53#include <sys/syslog.h>
54#include <sys/sx.h>
55
56#include <net/if.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/vnet.h>
64
65#include <netinet/if_ether.h>
66#include <netinet/in.h>
67#include <netinet/in_var.h>
68#include <netinet/in_pcb.h>
69#include <netinet/ip_var.h>
70#include <netinet/ip_carp.h>
71#include <netinet/igmp_var.h>
72#include <netinet/udp.h>
73#include <netinet/udp_var.h>
74
75static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
76static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
77
78static void	in_socktrim(struct sockaddr_in *);
79static void	in_purgemaddrs(struct ifnet *);
80
81VNET_DEFINE_STATIC(int, nosameprefix);
82#define	V_nosameprefix			VNET(nosameprefix)
83SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
84	&VNET_NAME(nosameprefix), 0,
85	"Refuse to create same prefixes on different interfaces");
86
87VNET_DECLARE(struct inpcbinfo, ripcbinfo);
88#define	V_ripcbinfo			VNET(ripcbinfo)
89
90static struct sx in_control_sx;
91SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
92
93/*
94 * Return 1 if an internet address is for a ``local'' host
95 * (one to which we have a connection).
96 */
97int
98in_localaddr(struct in_addr in)
99{
100	struct rm_priotracker in_ifa_tracker;
101	u_long i = ntohl(in.s_addr);
102	struct in_ifaddr *ia;
103
104	IN_IFADDR_RLOCK(&in_ifa_tracker);
105	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
106		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
107			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
108			return (1);
109		}
110	}
111	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
112	return (0);
113}
114
115/*
116 * Return 1 if an internet address is for the local host and configured
117 * on one of its interfaces.
118 */
119int
120in_localip(struct in_addr in)
121{
122	struct rm_priotracker in_ifa_tracker;
123	struct in_ifaddr *ia;
124
125	IN_IFADDR_RLOCK(&in_ifa_tracker);
126	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
127		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
128			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
129			return (1);
130		}
131	}
132	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
133	return (0);
134}
135
136/*
137 * Return 1 if an internet address is configured on an interface.
138 */
139int
140in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
141{
142	struct ifaddr *ifa;
143	struct in_ifaddr *ia;
144
145	NET_EPOCH_ASSERT();
146
147	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
148		if (ifa->ifa_addr->sa_family != AF_INET)
149			continue;
150		ia = (struct in_ifaddr *)ifa;
151		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
152			return (1);
153	}
154
155	return (0);
156}
157
158/*
159 * Return a reference to the interface address which is different to
160 * the supplied one but with same IP address value.
161 */
162static struct in_ifaddr *
163in_localip_more(struct in_ifaddr *ia)
164{
165	struct rm_priotracker in_ifa_tracker;
166	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
167	struct in_ifaddr *it;
168
169	IN_IFADDR_RLOCK(&in_ifa_tracker);
170	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
171		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
172			ifa_ref(&it->ia_ifa);
173			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
174			return (it);
175		}
176	}
177	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
178
179	return (NULL);
180}
181
182/*
183 * Determine whether an IP address is in a reserved set of addresses
184 * that may not be forwarded, or whether datagrams to that destination
185 * may be forwarded.
186 */
187int
188in_canforward(struct in_addr in)
189{
190	u_long i = ntohl(in.s_addr);
191
192	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
193	    IN_ZERONET(i) || IN_LOOPBACK(i))
194		return (0);
195	return (1);
196}
197
198/*
199 * Trim a mask in a sockaddr
200 */
201static void
202in_socktrim(struct sockaddr_in *ap)
203{
204    char *cplim = (char *) &ap->sin_addr;
205    char *cp = (char *) (&ap->sin_addr + 1);
206
207    ap->sin_len = 0;
208    while (--cp >= cplim)
209	if (*cp) {
210	    (ap)->sin_len = cp - (char *) (ap) + 1;
211	    break;
212	}
213}
214
215/*
216 * Generic internet control operations (ioctl's).
217 */
218int
219in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
220    struct thread *td)
221{
222	struct ifreq *ifr = (struct ifreq *)data;
223	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
224	struct epoch_tracker et;
225	struct ifaddr *ifa;
226	struct in_ifaddr *ia;
227	int error;
228
229	if (ifp == NULL)
230		return (EADDRNOTAVAIL);
231
232	/*
233	 * Filter out 4 ioctls we implement directly.  Forward the rest
234	 * to specific functions and ifp->if_ioctl().
235	 */
236	switch (cmd) {
237	case SIOCGIFADDR:
238	case SIOCGIFBRDADDR:
239	case SIOCGIFDSTADDR:
240	case SIOCGIFNETMASK:
241		break;
242	case SIOCDIFADDR:
243		sx_xlock(&in_control_sx);
244		error = in_difaddr_ioctl(cmd, data, ifp, td);
245		sx_xunlock(&in_control_sx);
246		return (error);
247	case OSIOCAIFADDR:	/* 9.x compat */
248	case SIOCAIFADDR:
249		sx_xlock(&in_control_sx);
250		error = in_aifaddr_ioctl(cmd, data, ifp, td);
251		sx_xunlock(&in_control_sx);
252		return (error);
253	case SIOCSIFADDR:
254	case SIOCSIFBRDADDR:
255	case SIOCSIFDSTADDR:
256	case SIOCSIFNETMASK:
257		/* We no longer support that old commands. */
258		return (EINVAL);
259	default:
260		if (ifp->if_ioctl == NULL)
261			return (EOPNOTSUPP);
262		return ((*ifp->if_ioctl)(ifp, cmd, data));
263	}
264
265	if (addr->sin_addr.s_addr != INADDR_ANY &&
266	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
267		return (EADDRNOTAVAIL);
268
269	/*
270	 * Find address for this interface, if it exists.  If an
271	 * address was specified, find that one instead of the
272	 * first one on the interface, if possible.
273	 */
274	NET_EPOCH_ENTER(et);
275	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
276		if (ifa->ifa_addr->sa_family != AF_INET)
277			continue;
278		ia = (struct in_ifaddr *)ifa;
279		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
280			break;
281	}
282	if (ifa == NULL)
283		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
284			if (ifa->ifa_addr->sa_family == AF_INET) {
285				ia = (struct in_ifaddr *)ifa;
286				if (prison_check_ip4(td->td_ucred,
287				    &ia->ia_addr.sin_addr) == 0)
288					break;
289			}
290
291	if (ifa == NULL) {
292		NET_EPOCH_EXIT(et);
293		return (EADDRNOTAVAIL);
294	}
295
296	error = 0;
297	switch (cmd) {
298	case SIOCGIFADDR:
299		*addr = ia->ia_addr;
300		break;
301
302	case SIOCGIFBRDADDR:
303		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
304			error = EINVAL;
305			break;
306		}
307		*addr = ia->ia_broadaddr;
308		break;
309
310	case SIOCGIFDSTADDR:
311		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
312			error = EINVAL;
313			break;
314		}
315		*addr = ia->ia_dstaddr;
316		break;
317
318	case SIOCGIFNETMASK:
319		*addr = ia->ia_sockmask;
320		break;
321	}
322
323	NET_EPOCH_EXIT(et);
324
325	return (error);
326}
327
328static int
329in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
330{
331	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
332	const struct sockaddr_in *addr = &ifra->ifra_addr;
333	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
334	const struct sockaddr_in *mask = &ifra->ifra_mask;
335	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
336	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
337	struct epoch_tracker et;
338	struct ifaddr *ifa;
339	struct in_ifaddr *ia;
340	bool iaIsFirst;
341	int error = 0;
342
343	error = priv_check(td, PRIV_NET_ADDIFADDR);
344	if (error)
345		return (error);
346
347	/*
348	 * ifra_addr must be present and be of INET family.
349	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
350	 */
351	if (addr->sin_len != sizeof(struct sockaddr_in) ||
352	    addr->sin_family != AF_INET)
353		return (EINVAL);
354	if (broadaddr->sin_len != 0 &&
355	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
356	    broadaddr->sin_family != AF_INET))
357		return (EINVAL);
358	if (mask->sin_len != 0 &&
359	    (mask->sin_len != sizeof(struct sockaddr_in) ||
360	    mask->sin_family != AF_INET))
361		return (EINVAL);
362	if ((ifp->if_flags & IFF_POINTOPOINT) &&
363	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
364	     dstaddr->sin_addr.s_addr == INADDR_ANY))
365		return (EDESTADDRREQ);
366	if (vhid > 0 && carp_attach_p == NULL)
367		return (EPROTONOSUPPORT);
368
369	/*
370	 * See whether address already exist.
371	 */
372	iaIsFirst = true;
373	ia = NULL;
374	NET_EPOCH_ENTER(et);
375	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
376		struct in_ifaddr *it;
377
378		if (ifa->ifa_addr->sa_family != AF_INET)
379			continue;
380
381		it = (struct in_ifaddr *)ifa;
382		iaIsFirst = false;
383		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
384		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
385			ia = it;
386	}
387	NET_EPOCH_EXIT(et);
388
389	if (ia != NULL)
390		(void )in_difaddr_ioctl(cmd, data, ifp, td);
391
392	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
393	ia = (struct in_ifaddr *)ifa;
394	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
395	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
396	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
397	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
398	    CALLOUT_RETURNUNLOCKED);
399
400	ia->ia_ifp = ifp;
401	ia->ia_addr = *addr;
402	if (mask->sin_len != 0) {
403		ia->ia_sockmask = *mask;
404		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
405	} else {
406		in_addr_t i = ntohl(addr->sin_addr.s_addr);
407
408		/*
409	 	 * Be compatible with network classes, if netmask isn't
410		 * supplied, guess it based on classes.
411	 	 */
412		if (IN_CLASSA(i))
413			ia->ia_subnetmask = IN_CLASSA_NET;
414		else if (IN_CLASSB(i))
415			ia->ia_subnetmask = IN_CLASSB_NET;
416		else
417			ia->ia_subnetmask = IN_CLASSC_NET;
418		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
419	}
420	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
421	in_socktrim(&ia->ia_sockmask);
422
423	if (ifp->if_flags & IFF_BROADCAST) {
424		if (broadaddr->sin_len != 0) {
425			ia->ia_broadaddr = *broadaddr;
426		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
427			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
428			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
429			ia->ia_broadaddr.sin_family = AF_INET;
430		} else {
431			ia->ia_broadaddr.sin_addr.s_addr =
432			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
433			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
434			ia->ia_broadaddr.sin_family = AF_INET;
435		}
436	}
437
438	if (ifp->if_flags & IFF_POINTOPOINT)
439		ia->ia_dstaddr = *dstaddr;
440
441	/* XXXGL: rtinit() needs this strange assignment. */
442	if (ifp->if_flags & IFF_LOOPBACK)
443                ia->ia_dstaddr = ia->ia_addr;
444
445	if (vhid != 0) {
446		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
447		if (error)
448			return (error);
449	}
450
451	/* if_addrhead is already referenced by ifa_alloc() */
452	IF_ADDR_WLOCK(ifp);
453	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
454	IF_ADDR_WUNLOCK(ifp);
455
456	ifa_ref(ifa);			/* in_ifaddrhead */
457	IN_IFADDR_WLOCK();
458	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
459	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
460	IN_IFADDR_WUNLOCK();
461
462	/*
463	 * Give the interface a chance to initialize
464	 * if this is its first address,
465	 * and to validate the address if necessary.
466	 */
467	if (ifp->if_ioctl != NULL) {
468		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
469		if (error)
470			goto fail1;
471	}
472
473	/*
474	 * Add route for the network.
475	 */
476	if (vhid == 0) {
477		int flags = RTF_UP;
478
479		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
480			flags |= RTF_HOST;
481
482		error = in_addprefix(ia, flags);
483		if (error)
484			goto fail1;
485	}
486
487	/*
488	 * Add a loopback route to self.
489	 */
490	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
491	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
492	    !((ifp->if_flags & IFF_POINTOPOINT) &&
493	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
494		struct in_ifaddr *eia;
495
496		eia = in_localip_more(ia);
497
498		if (eia == NULL) {
499			error = ifa_add_loopback_route((struct ifaddr *)ia,
500			    (struct sockaddr *)&ia->ia_addr);
501			if (error)
502				goto fail2;
503		} else
504			ifa_free(&eia->ia_ifa);
505	}
506
507	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
508		struct in_addr allhosts_addr;
509		struct in_ifinfo *ii;
510
511		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
512		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
513
514		error = in_joingroup(ifp, &allhosts_addr, NULL,
515			&ii->ii_allhosts);
516	}
517
518	/*
519	 * Note: we don't need extra reference for ifa, since we called
520	 * with sx lock held, and ifaddr can not be deleted in concurrent
521	 * thread.
522	 */
523	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
524
525	return (error);
526
527fail2:
528	if (vhid == 0)
529		(void )in_scrubprefix(ia, LLE_STATIC);
530
531fail1:
532	if (ia->ia_ifa.ifa_carp)
533		(*carp_detach_p)(&ia->ia_ifa, false);
534
535	IF_ADDR_WLOCK(ifp);
536	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
537	IF_ADDR_WUNLOCK(ifp);
538	ifa_free(&ia->ia_ifa);		/* if_addrhead */
539
540	IN_IFADDR_WLOCK();
541	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
542	LIST_REMOVE(ia, ia_hash);
543	IN_IFADDR_WUNLOCK();
544	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
545
546	return (error);
547}
548
549static int
550in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
551{
552	const struct ifreq *ifr = (struct ifreq *)data;
553	const struct sockaddr_in *addr = (const struct sockaddr_in *)
554	    &ifr->ifr_addr;
555	struct ifaddr *ifa;
556	struct in_ifaddr *ia;
557	bool deleteAny, iaIsLast;
558	int error;
559
560	if (td != NULL) {
561		error = priv_check(td, PRIV_NET_DELIFADDR);
562		if (error)
563			return (error);
564	}
565
566	if (addr->sin_len != sizeof(struct sockaddr_in) ||
567	    addr->sin_family != AF_INET)
568		deleteAny = true;
569	else
570		deleteAny = false;
571
572	iaIsLast = true;
573	ia = NULL;
574	IF_ADDR_WLOCK(ifp);
575	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
576		struct in_ifaddr *it;
577
578		if (ifa->ifa_addr->sa_family != AF_INET)
579			continue;
580
581		it = (struct in_ifaddr *)ifa;
582		if (deleteAny && ia == NULL && (td == NULL ||
583		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
584			ia = it;
585
586		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
587		    (td == NULL || prison_check_ip4(td->td_ucred,
588		    &addr->sin_addr) == 0))
589			ia = it;
590
591		if (it != ia)
592			iaIsLast = false;
593	}
594
595	if (ia == NULL) {
596		IF_ADDR_WUNLOCK(ifp);
597		return (EADDRNOTAVAIL);
598	}
599
600	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
601	IF_ADDR_WUNLOCK(ifp);
602	ifa_free(&ia->ia_ifa);		/* if_addrhead */
603
604	IN_IFADDR_WLOCK();
605	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
606	LIST_REMOVE(ia, ia_hash);
607	IN_IFADDR_WUNLOCK();
608
609	/*
610	 * in_scrubprefix() kills the interface route.
611	 */
612	in_scrubprefix(ia, LLE_STATIC);
613
614	/*
615	 * in_ifadown gets rid of all the rest of
616	 * the routes.  This is not quite the right
617	 * thing to do, but at least if we are running
618	 * a routing process they will come back.
619	 */
620	in_ifadown(&ia->ia_ifa, 1);
621
622	if (ia->ia_ifa.ifa_carp)
623		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
624
625	/*
626	 * If this is the last IPv4 address configured on this
627	 * interface, leave the all-hosts group.
628	 * No state-change report need be transmitted.
629	 */
630	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
631		struct in_ifinfo *ii;
632
633		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
634		if (ii->ii_allhosts) {
635			(void)in_leavegroup(ii->ii_allhosts, NULL);
636			ii->ii_allhosts = NULL;
637		}
638	}
639
640	IF_ADDR_WLOCK(ifp);
641	if (callout_stop(&ia->ia_garp_timer) == 1) {
642		ifa_free(&ia->ia_ifa);
643	}
644	IF_ADDR_WUNLOCK(ifp);
645
646	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
647	    IFADDR_EVENT_DEL);
648	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
649
650	return (0);
651}
652
653#define rtinitflags(x) \
654	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
655	    ? RTF_HOST : 0)
656
657/*
658 * Check if we have a route for the given prefix already or add one accordingly.
659 */
660int
661in_addprefix(struct in_ifaddr *target, int flags)
662{
663	struct rm_priotracker in_ifa_tracker;
664	struct in_ifaddr *ia;
665	struct in_addr prefix, mask, p, m;
666	int error;
667
668	if ((flags & RTF_HOST) != 0) {
669		prefix = target->ia_dstaddr.sin_addr;
670		mask.s_addr = 0;
671	} else {
672		prefix = target->ia_addr.sin_addr;
673		mask = target->ia_sockmask.sin_addr;
674		prefix.s_addr &= mask.s_addr;
675	}
676
677	IN_IFADDR_RLOCK(&in_ifa_tracker);
678	/* Look for an existing address with the same prefix, mask, and fib */
679	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
680		if (rtinitflags(ia)) {
681			p = ia->ia_dstaddr.sin_addr;
682
683			if (prefix.s_addr != p.s_addr)
684				continue;
685		} else {
686			p = ia->ia_addr.sin_addr;
687			m = ia->ia_sockmask.sin_addr;
688			p.s_addr &= m.s_addr;
689
690			if (prefix.s_addr != p.s_addr ||
691			    mask.s_addr != m.s_addr)
692				continue;
693		}
694		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
695			continue;
696
697		/*
698		 * If we got a matching prefix route inserted by other
699		 * interface address, we are done here.
700		 */
701		if (ia->ia_flags & IFA_ROUTE) {
702#ifdef RADIX_MPATH
703			if (ia->ia_addr.sin_addr.s_addr ==
704			    target->ia_addr.sin_addr.s_addr) {
705				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
706				return (EEXIST);
707			} else
708				break;
709#endif
710			if (V_nosameprefix) {
711				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
712				return (EEXIST);
713			} else {
714				int fibnum;
715
716				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
717					target->ia_ifp->if_fib;
718				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
719				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
720				return (0);
721			}
722		}
723	}
724	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
725
726	/*
727	 * No-one seem to have this prefix route, so we try to insert it.
728	 */
729	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
730	if (!error)
731		target->ia_flags |= IFA_ROUTE;
732	return (error);
733}
734
735/*
736 * Removes either all lle entries for given @ia, or lle
737 * corresponding to @ia address.
738 */
739static void
740in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
741{
742	struct sockaddr_in addr, mask;
743	struct sockaddr *saddr, *smask;
744	struct ifnet *ifp;
745
746	saddr = (struct sockaddr *)&addr;
747	bzero(&addr, sizeof(addr));
748	addr.sin_len = sizeof(addr);
749	addr.sin_family = AF_INET;
750	smask = (struct sockaddr *)&mask;
751	bzero(&mask, sizeof(mask));
752	mask.sin_len = sizeof(mask);
753	mask.sin_family = AF_INET;
754	mask.sin_addr.s_addr = ia->ia_subnetmask;
755	ifp = ia->ia_ifp;
756
757	if (all) {
758
759		/*
760		 * Remove all L2 entries matching given prefix.
761		 * Convert address to host representation to avoid
762		 * doing this on every callback. ia_subnetmask is already
763		 * stored in host representation.
764		 */
765		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
766		lltable_prefix_free(AF_INET, saddr, smask, flags);
767	} else {
768		/* Remove interface address only */
769		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
770		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
771	}
772}
773
774/*
775 * If there is no other address in the system that can serve a route to the
776 * same prefix, remove the route.  Hand over the route to the new address
777 * otherwise.
778 */
779int
780in_scrubprefix(struct in_ifaddr *target, u_int flags)
781{
782	struct rm_priotracker in_ifa_tracker;
783	struct in_ifaddr *ia;
784	struct in_addr prefix, mask, p, m;
785	int error = 0;
786
787	/*
788	 * Remove the loopback route to the interface address.
789	 */
790	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
791	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
792	    (flags & LLE_STATIC)) {
793		struct in_ifaddr *eia;
794
795		/*
796		 * XXXME: add fib-aware in_localip.
797		 * We definitely don't want to switch between
798		 * prefixes in different fibs.
799		 */
800		eia = in_localip_more(target);
801
802		if (eia != NULL) {
803			error = ifa_switch_loopback_route((struct ifaddr *)eia,
804			    (struct sockaddr *)&target->ia_addr);
805			ifa_free(&eia->ia_ifa);
806		} else {
807			error = ifa_del_loopback_route((struct ifaddr *)target,
808			    (struct sockaddr *)&target->ia_addr);
809		}
810	}
811
812	if (rtinitflags(target)) {
813		prefix = target->ia_dstaddr.sin_addr;
814		mask.s_addr = 0;
815	} else {
816		prefix = target->ia_addr.sin_addr;
817		mask = target->ia_sockmask.sin_addr;
818		prefix.s_addr &= mask.s_addr;
819	}
820
821	if ((target->ia_flags & IFA_ROUTE) == 0) {
822		int fibnum;
823
824		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
825			target->ia_ifp->if_fib;
826		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
827
828		/*
829		 * Removing address from !IFF_UP interface or
830		 * prefix which exists on other interface (along with route).
831		 * No entries should exist here except target addr.
832		 * Given that, delete this entry only.
833		 */
834		in_scrubprefixlle(target, 0, flags);
835		return (0);
836	}
837
838	IN_IFADDR_RLOCK(&in_ifa_tracker);
839	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
840		if (rtinitflags(ia)) {
841			p = ia->ia_dstaddr.sin_addr;
842
843			if (prefix.s_addr != p.s_addr)
844				continue;
845		} else {
846			p = ia->ia_addr.sin_addr;
847			m = ia->ia_sockmask.sin_addr;
848			p.s_addr &= m.s_addr;
849
850			if (prefix.s_addr != p.s_addr ||
851			    mask.s_addr != m.s_addr)
852				continue;
853		}
854
855		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
856			continue;
857
858		/*
859		 * If we got a matching prefix address, move IFA_ROUTE and
860		 * the route itself to it.  Make sure that routing daemons
861		 * get a heads-up.
862		 */
863		if ((ia->ia_flags & IFA_ROUTE) == 0) {
864			ifa_ref(&ia->ia_ifa);
865			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
866			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
867			    rtinitflags(target));
868			if (error == 0)
869				target->ia_flags &= ~IFA_ROUTE;
870			else
871				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
872					error);
873			/* Scrub all entries IFF interface is different */
874			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
875			    flags);
876			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
877			    rtinitflags(ia) | RTF_UP);
878			if (error == 0)
879				ia->ia_flags |= IFA_ROUTE;
880			else
881				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
882					error);
883			ifa_free(&ia->ia_ifa);
884			return (error);
885		}
886	}
887	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
888
889	/*
890	 * remove all L2 entries on the given prefix
891	 */
892	in_scrubprefixlle(target, 1, flags);
893
894	/*
895	 * As no-one seem to have this prefix, we can remove the route.
896	 */
897	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
898	if (error == 0)
899		target->ia_flags &= ~IFA_ROUTE;
900	else
901		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
902	return (error);
903}
904
905#undef rtinitflags
906
907void
908in_ifscrub_all(void)
909{
910	struct ifnet *ifp;
911	struct ifaddr *ifa, *nifa;
912	struct ifaliasreq ifr;
913
914	IFNET_RLOCK();
915	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
916		/* Cannot lock here - lock recursion. */
917		/* NET_EPOCH_ENTER(et); */
918		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
919			if (ifa->ifa_addr->sa_family != AF_INET)
920				continue;
921
922			/*
923			 * This is ugly but the only way for legacy IP to
924			 * cleanly remove addresses and everything attached.
925			 */
926			bzero(&ifr, sizeof(ifr));
927			ifr.ifra_addr = *ifa->ifa_addr;
928			if (ifa->ifa_dstaddr)
929			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
930			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
931			    ifp, NULL);
932		}
933		/* NET_EPOCH_EXIT(et); */
934		in_purgemaddrs(ifp);
935		igmp_domifdetach(ifp);
936	}
937	IFNET_RUNLOCK();
938}
939
940int
941in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
942{
943
944	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
945	     /*
946	      * Check for old-style (host 0) broadcast, but
947	      * taking into account that RFC 3021 obsoletes it.
948	      */
949	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
950	    ntohl(in.s_addr) == ia->ia_subnet)) &&
951	     /*
952	      * Check for an all one subnetmask. These
953	      * only exist when an interface gets a secondary
954	      * address.
955	      */
956	    ia->ia_subnetmask != (u_long)0xffffffff);
957}
958
959/*
960 * Return 1 if the address might be a local broadcast address.
961 */
962int
963in_broadcast(struct in_addr in, struct ifnet *ifp)
964{
965	struct ifaddr *ifa;
966	int found;
967
968	NET_EPOCH_ASSERT();
969
970	if (in.s_addr == INADDR_BROADCAST ||
971	    in.s_addr == INADDR_ANY)
972		return (1);
973	if ((ifp->if_flags & IFF_BROADCAST) == 0)
974		return (0);
975	found = 0;
976	/*
977	 * Look through the list of addresses for a match
978	 * with a broadcast address.
979	 */
980	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
981		if (ifa->ifa_addr->sa_family == AF_INET &&
982		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
983			found = 1;
984			break;
985		}
986	return (found);
987}
988
989/*
990 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
991 */
992void
993in_ifdetach(struct ifnet *ifp)
994{
995	IN_MULTI_LOCK();
996	in_pcbpurgeif0(&V_ripcbinfo, ifp);
997	in_pcbpurgeif0(&V_udbinfo, ifp);
998	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
999	in_purgemaddrs(ifp);
1000	IN_MULTI_UNLOCK();
1001}
1002
1003/*
1004 * Delete all IPv4 multicast address records, and associated link-layer
1005 * multicast address records, associated with ifp.
1006 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1007 * XXX This should not race with ifma_protospec being set during
1008 * a new allocation, if it does, we have bigger problems.
1009 */
1010static void
1011in_purgemaddrs(struct ifnet *ifp)
1012{
1013	struct in_multi_head purgeinms;
1014	struct in_multi		*inm;
1015	struct ifmultiaddr	*ifma, *next;
1016
1017	SLIST_INIT(&purgeinms);
1018	IN_MULTI_LIST_LOCK();
1019
1020	/*
1021	 * Extract list of in_multi associated with the detaching ifp
1022	 * which the PF_INET layer is about to release.
1023	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1024	 * by code further down.
1025	 */
1026	IF_ADDR_WLOCK(ifp);
1027 restart:
1028	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1029		if (ifma->ifma_addr->sa_family != AF_INET ||
1030		    ifma->ifma_protospec == NULL)
1031			continue;
1032		inm = (struct in_multi *)ifma->ifma_protospec;
1033		inm_rele_locked(&purgeinms, inm);
1034		if (__predict_false(ifma_restart)) {
1035			ifma_restart = true;
1036			goto restart;
1037		}
1038	}
1039	IF_ADDR_WUNLOCK(ifp);
1040
1041	inm_release_list_deferred(&purgeinms);
1042	igmp_ifdetach(ifp);
1043	IN_MULTI_LIST_UNLOCK();
1044}
1045
1046struct in_llentry {
1047	struct llentry		base;
1048};
1049
1050#define	IN_LLTBL_DEFAULT_HSIZE	32
1051#define	IN_LLTBL_HASH(k, h) \
1052	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1053
1054/*
1055 * Do actual deallocation of @lle.
1056 */
1057static void
1058in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1059{
1060	struct llentry *lle;
1061
1062	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1063	LLE_LOCK_DESTROY(lle);
1064	LLE_REQ_DESTROY(lle);
1065	free(lle, M_LLTABLE);
1066}
1067
1068/*
1069 * Called by the datapath to indicate that
1070 * the entry was used.
1071 */
1072static void
1073in_lltable_mark_used(struct llentry *lle)
1074{
1075
1076	LLE_REQ_LOCK(lle);
1077	lle->r_skip_req = 0;
1078	LLE_REQ_UNLOCK(lle);
1079}
1080
1081/*
1082 * Called by LLE_FREE_LOCKED when number of references
1083 * drops to zero.
1084 */
1085static void
1086in_lltable_destroy_lle(struct llentry *lle)
1087{
1088
1089	LLE_WUNLOCK(lle);
1090	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1091}
1092
1093static struct llentry *
1094in_lltable_new(struct in_addr addr4, u_int flags)
1095{
1096	struct in_llentry *lle;
1097
1098	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1099	if (lle == NULL)		/* NB: caller generates msg */
1100		return NULL;
1101
1102	/*
1103	 * For IPv4 this will trigger "arpresolve" to generate
1104	 * an ARP request.
1105	 */
1106	lle->base.la_expire = time_uptime; /* mark expired */
1107	lle->base.r_l3addr.addr4 = addr4;
1108	lle->base.lle_refcnt = 1;
1109	lle->base.lle_free = in_lltable_destroy_lle;
1110	LLE_LOCK_INIT(&lle->base);
1111	LLE_REQ_INIT(&lle->base);
1112	callout_init(&lle->base.lle_timer, 1);
1113
1114	return (&lle->base);
1115}
1116
1117#define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1118	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1119
1120static int
1121in_lltable_match_prefix(const struct sockaddr *saddr,
1122    const struct sockaddr *smask, u_int flags, struct llentry *lle)
1123{
1124	struct in_addr addr, mask, lle_addr;
1125
1126	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1127	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1128	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1129
1130	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1131		return (0);
1132
1133	if (lle->la_flags & LLE_IFADDR) {
1134
1135		/*
1136		 * Delete LLE_IFADDR records IFF address & flag matches.
1137		 * Note that addr is the interface address within prefix
1138		 * being matched.
1139		 * Note also we should handle 'ifdown' cases without removing
1140		 * ifaddr macs.
1141		 */
1142		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1143			return (1);
1144		return (0);
1145	}
1146
1147	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1148	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1149		return (1);
1150
1151	return (0);
1152}
1153
1154static void
1155in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1156{
1157	size_t pkts_dropped;
1158
1159	LLE_WLOCK_ASSERT(lle);
1160	KASSERT(llt != NULL, ("lltable is NULL"));
1161
1162	/* Unlink entry from table if not already */
1163	if ((lle->la_flags & LLE_LINKED) != 0) {
1164		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1165		lltable_unlink_entry(llt, lle);
1166	}
1167
1168	/* Drop hold queue */
1169	pkts_dropped = llentry_free(lle);
1170	ARPSTAT_ADD(dropped, pkts_dropped);
1171}
1172
1173static int
1174in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1175{
1176	struct rt_addrinfo info;
1177	struct sockaddr_in rt_key, rt_mask;
1178	struct sockaddr rt_gateway;
1179	int rt_flags;
1180
1181	KASSERT(l3addr->sa_family == AF_INET,
1182	    ("sin_family %d", l3addr->sa_family));
1183
1184	bzero(&rt_key, sizeof(rt_key));
1185	rt_key.sin_len = sizeof(rt_key);
1186	bzero(&rt_mask, sizeof(rt_mask));
1187	rt_mask.sin_len = sizeof(rt_mask);
1188	bzero(&rt_gateway, sizeof(rt_gateway));
1189	rt_gateway.sa_len = sizeof(rt_gateway);
1190
1191	bzero(&info, sizeof(info));
1192	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1193	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1194	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1195
1196	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1197		return (EINVAL);
1198
1199	rt_flags = info.rti_flags;
1200
1201	/*
1202	 * If the gateway for an existing host route matches the target L3
1203	 * address, which is a special route inserted by some implementation
1204	 * such as MANET, and the interface is of the correct type, then
1205	 * allow for ARP to proceed.
1206	 */
1207	if (rt_flags & RTF_GATEWAY) {
1208		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1209		    info.rti_ifp->if_type != IFT_ETHER ||
1210		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1211		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1212		    sizeof(in_addr_t)) != 0) {
1213			rib_free_info(&info);
1214			return (EINVAL);
1215		}
1216	}
1217	rib_free_info(&info);
1218
1219	/*
1220	 * Make sure that at least the destination address is covered
1221	 * by the route. This is for handling the case where 2 or more
1222	 * interfaces have the same prefix. An incoming packet arrives
1223	 * on one interface and the corresponding outgoing packet leaves
1224	 * another interface.
1225	 */
1226	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1227		const char *sa, *mask, *addr, *lim;
1228		const struct sockaddr_in *l3sin;
1229
1230		mask = (const char *)&rt_mask;
1231		/*
1232		 * Just being extra cautious to avoid some custom
1233		 * code getting into trouble.
1234		 */
1235		if ((info.rti_addrs & RTA_NETMASK) == 0)
1236			return (EINVAL);
1237
1238		sa = (const char *)&rt_key;
1239		addr = (const char *)l3addr;
1240		l3sin = (const struct sockaddr_in *)l3addr;
1241		lim = addr + l3sin->sin_len;
1242
1243		for ( ; addr < lim; sa++, mask++, addr++) {
1244			if ((*sa ^ *addr) & *mask) {
1245#ifdef DIAGNOSTIC
1246				char addrbuf[INET_ADDRSTRLEN];
1247
1248				log(LOG_INFO, "IPv4 address: \"%s\" "
1249				    "is not on the network\n",
1250				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1251#endif
1252				return (EINVAL);
1253			}
1254		}
1255	}
1256
1257	return (0);
1258}
1259
1260static inline uint32_t
1261in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1262{
1263
1264	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1265}
1266
1267static uint32_t
1268in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1269{
1270
1271	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1272}
1273
1274static void
1275in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1276{
1277	struct sockaddr_in *sin;
1278
1279	sin = (struct sockaddr_in *)sa;
1280	bzero(sin, sizeof(*sin));
1281	sin->sin_family = AF_INET;
1282	sin->sin_len = sizeof(*sin);
1283	sin->sin_addr = lle->r_l3addr.addr4;
1284}
1285
1286static inline struct llentry *
1287in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1288{
1289	struct llentry *lle;
1290	struct llentries *lleh;
1291	u_int hashidx;
1292
1293	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1294	lleh = &llt->lle_head[hashidx];
1295	CK_LIST_FOREACH(lle, lleh, lle_next) {
1296		if (lle->la_flags & LLE_DELETED)
1297			continue;
1298		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1299			break;
1300	}
1301
1302	return (lle);
1303}
1304
1305static void
1306in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1307{
1308
1309	lle->la_flags |= LLE_DELETED;
1310	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1311#ifdef DIAGNOSTIC
1312	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1313#endif
1314	llentry_free(lle);
1315}
1316
1317static struct llentry *
1318in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1319{
1320	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1321	struct ifnet *ifp = llt->llt_ifp;
1322	struct llentry *lle;
1323	char linkhdr[LLE_MAX_LINKHDR];
1324	size_t linkhdrsize;
1325	int lladdr_off;
1326
1327	KASSERT(l3addr->sa_family == AF_INET,
1328	    ("sin_family %d", l3addr->sa_family));
1329
1330	/*
1331	 * A route that covers the given address must have
1332	 * been installed 1st because we are doing a resolution,
1333	 * verify this.
1334	 */
1335	if (!(flags & LLE_IFADDR) &&
1336	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1337		return (NULL);
1338
1339	lle = in_lltable_new(sin->sin_addr, flags);
1340	if (lle == NULL) {
1341		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1342		return (NULL);
1343	}
1344	lle->la_flags = flags;
1345	if (flags & LLE_STATIC)
1346		lle->r_flags |= RLLE_VALID;
1347	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1348		linkhdrsize = LLE_MAX_LINKHDR;
1349		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1350		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1351			NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1352			return (NULL);
1353		}
1354		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1355		    lladdr_off);
1356		lle->la_flags |= LLE_STATIC;
1357		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1358	}
1359
1360	return (lle);
1361}
1362
1363/*
1364 * Return NULL if not found or marked for deletion.
1365 * If found return lle read locked.
1366 */
1367static struct llentry *
1368in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1369{
1370	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1371	struct llentry *lle;
1372
1373	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1374	KASSERT(l3addr->sa_family == AF_INET,
1375	    ("sin_family %d", l3addr->sa_family));
1376	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1377	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1378	    ("wrong lle request flags: %#x", flags));
1379
1380	lle = in_lltable_find_dst(llt, sin->sin_addr);
1381	if (lle == NULL)
1382		return (NULL);
1383	if (flags & LLE_UNLOCKED)
1384		return (lle);
1385
1386	if (flags & LLE_EXCLUSIVE)
1387		LLE_WLOCK(lle);
1388	else
1389		LLE_RLOCK(lle);
1390
1391	/*
1392	 * If the afdata lock is not held, the LLE may have been unlinked while
1393	 * we were blocked on the LLE lock.  Check for this case.
1394	 */
1395	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1396		if (flags & LLE_EXCLUSIVE)
1397			LLE_WUNLOCK(lle);
1398		else
1399			LLE_RUNLOCK(lle);
1400		return (NULL);
1401	}
1402	return (lle);
1403}
1404
1405static int
1406in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1407    struct sysctl_req *wr)
1408{
1409	struct ifnet *ifp = llt->llt_ifp;
1410	/* XXX stack use */
1411	struct {
1412		struct rt_msghdr	rtm;
1413		struct sockaddr_in	sin;
1414		struct sockaddr_dl	sdl;
1415	} arpc;
1416	struct sockaddr_dl *sdl;
1417	int error;
1418
1419	bzero(&arpc, sizeof(arpc));
1420	/* skip deleted entries */
1421	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1422		return (0);
1423	/* Skip if jailed and not a valid IP of the prison. */
1424	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1425	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1426		return (0);
1427	/*
1428	 * produce a msg made of:
1429	 *  struct rt_msghdr;
1430	 *  struct sockaddr_in; (IPv4)
1431	 *  struct sockaddr_dl;
1432	 */
1433	arpc.rtm.rtm_msglen = sizeof(arpc);
1434	arpc.rtm.rtm_version = RTM_VERSION;
1435	arpc.rtm.rtm_type = RTM_GET;
1436	arpc.rtm.rtm_flags = RTF_UP;
1437	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1438
1439	/* publish */
1440	if (lle->la_flags & LLE_PUB)
1441		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1442
1443	sdl = &arpc.sdl;
1444	sdl->sdl_family = AF_LINK;
1445	sdl->sdl_len = sizeof(*sdl);
1446	sdl->sdl_index = ifp->if_index;
1447	sdl->sdl_type = ifp->if_type;
1448	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1449		sdl->sdl_alen = ifp->if_addrlen;
1450		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1451	} else {
1452		sdl->sdl_alen = 0;
1453		bzero(LLADDR(sdl), ifp->if_addrlen);
1454	}
1455
1456	arpc.rtm.rtm_rmx.rmx_expire =
1457	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1458	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1459	if (lle->la_flags & LLE_STATIC)
1460		arpc.rtm.rtm_flags |= RTF_STATIC;
1461	if (lle->la_flags & LLE_IFADDR)
1462		arpc.rtm.rtm_flags |= RTF_PINNED;
1463	arpc.rtm.rtm_index = ifp->if_index;
1464	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1465
1466	return (error);
1467}
1468
1469static struct lltable *
1470in_lltattach(struct ifnet *ifp)
1471{
1472	struct lltable *llt;
1473
1474	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1475 	llt->llt_af = AF_INET;
1476 	llt->llt_ifp = ifp;
1477
1478	llt->llt_lookup = in_lltable_lookup;
1479	llt->llt_alloc_entry = in_lltable_alloc;
1480	llt->llt_delete_entry = in_lltable_delete_entry;
1481	llt->llt_dump_entry = in_lltable_dump_entry;
1482	llt->llt_hash = in_lltable_hash;
1483	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1484	llt->llt_free_entry = in_lltable_free_entry;
1485	llt->llt_match_prefix = in_lltable_match_prefix;
1486	llt->llt_mark_used = in_lltable_mark_used;
1487 	lltable_link(llt);
1488
1489	return (llt);
1490}
1491
1492void *
1493in_domifattach(struct ifnet *ifp)
1494{
1495	struct in_ifinfo *ii;
1496
1497	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1498
1499	ii->ii_llt = in_lltattach(ifp);
1500	ii->ii_igmp = igmp_domifattach(ifp);
1501
1502	return (ii);
1503}
1504
1505void
1506in_domifdetach(struct ifnet *ifp, void *aux)
1507{
1508	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1509
1510	igmp_domifdetach(ifp);
1511	lltable_free(ii->ii_llt);
1512	free(ii, M_IFADDR);
1513}
1514