in.c revision 192262
1/*-
2 * Copyright (c) 1982, 1986, 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 * Copyright (C) 2001 WIDE Project.  All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 * 4. Neither the name of the University nor the names of its contributors
15 *    may be used to endorse or promote products derived from this software
16 *    without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 *
30 *	@(#)in.c	8.4 (Berkeley) 1/9/95
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: head/sys/netinet/in.c 192262 2009-05-17 20:53:10Z bz $");
35
36#include "opt_carp.h"
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/sockio.h>
41#include <sys/malloc.h>
42#include <sys/priv.h>
43#include <sys/socket.h>
44#include <sys/jail.h>
45#include <sys/kernel.h>
46#include <sys/proc.h>
47#include <sys/sysctl.h>
48#include <sys/syslog.h>
49#include <sys/vimage.h>
50
51#include <net/if.h>
52#include <net/if_dl.h>
53#include <net/if_llatbl.h>
54#include <net/if_types.h>
55#include <net/route.h>
56#include <net/vnet.h>
57
58#include <netinet/in.h>
59#include <netinet/in_var.h>
60#include <netinet/in_pcb.h>
61#include <netinet/ip_var.h>
62#include <netinet/vinet.h>
63#include <netinet/igmp_var.h>
64
65static int in_mask2len(struct in_addr *);
66static void in_len2mask(struct in_addr *, int);
67static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
68	struct ifnet *, struct thread *);
69
70static int	in_addprefix(struct in_ifaddr *, int);
71static int	in_scrubprefix(struct in_ifaddr *);
72static void	in_socktrim(struct sockaddr_in *);
73static int	in_ifinit(struct ifnet *,
74	    struct in_ifaddr *, struct sockaddr_in *, int);
75static void	in_purgemaddrs(struct ifnet *);
76
77#ifdef VIMAGE_GLOBALS
78static int subnetsarelocal;
79static int sameprefixcarponly;
80extern struct inpcbinfo ripcbinfo;
81#endif
82
83SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, subnets_are_local,
84	CTLFLAG_RW, subnetsarelocal, 0,
85	"Treat all subnets as directly connected");
86SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, same_prefix_carp_only,
87	CTLFLAG_RW, sameprefixcarponly, 0,
88	"Refuse to create same prefixes on different interfaces");
89
90/*
91 * Return 1 if an internet address is for a ``local'' host
92 * (one to which we have a connection).  If subnetsarelocal
93 * is true, this includes other subnets of the local net.
94 * Otherwise, it includes only the directly-connected (sub)nets.
95 */
96int
97in_localaddr(struct in_addr in)
98{
99	INIT_VNET_INET(curvnet);
100	register u_long i = ntohl(in.s_addr);
101	register struct in_ifaddr *ia;
102
103	if (V_subnetsarelocal) {
104		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
105			if ((i & ia->ia_netmask) == ia->ia_net)
106				return (1);
107	} else {
108		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
109			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
110				return (1);
111	}
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	INIT_VNET_INET(curvnet);
123	struct in_ifaddr *ia;
124
125	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
126		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
127			return (1);
128	}
129	return (0);
130}
131
132/*
133 * Determine whether an IP address is in a reserved set of addresses
134 * that may not be forwarded, or whether datagrams to that destination
135 * may be forwarded.
136 */
137int
138in_canforward(struct in_addr in)
139{
140	register u_long i = ntohl(in.s_addr);
141	register u_long net;
142
143	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
144		return (0);
145	if (IN_CLASSA(i)) {
146		net = i & IN_CLASSA_NET;
147		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
148			return (0);
149	}
150	return (1);
151}
152
153/*
154 * Trim a mask in a sockaddr
155 */
156static void
157in_socktrim(struct sockaddr_in *ap)
158{
159    register char *cplim = (char *) &ap->sin_addr;
160    register char *cp = (char *) (&ap->sin_addr + 1);
161
162    ap->sin_len = 0;
163    while (--cp >= cplim)
164	if (*cp) {
165	    (ap)->sin_len = cp - (char *) (ap) + 1;
166	    break;
167	}
168}
169
170static int
171in_mask2len(mask)
172	struct in_addr *mask;
173{
174	int x, y;
175	u_char *p;
176
177	p = (u_char *)mask;
178	for (x = 0; x < sizeof(*mask); x++) {
179		if (p[x] != 0xff)
180			break;
181	}
182	y = 0;
183	if (x < sizeof(*mask)) {
184		for (y = 0; y < 8; y++) {
185			if ((p[x] & (0x80 >> y)) == 0)
186				break;
187		}
188	}
189	return (x * 8 + y);
190}
191
192static void
193in_len2mask(struct in_addr *mask, int len)
194{
195	int i;
196	u_char *p;
197
198	p = (u_char *)mask;
199	bzero(mask, sizeof(*mask));
200	for (i = 0; i < len / 8; i++)
201		p[i] = 0xff;
202	if (len % 8)
203		p[i] = (0xff00 >> (len % 8)) & 0xff;
204}
205
206/*
207 * Generic internet control operations (ioctl's).
208 *
209 * ifp is NULL if not an interface-specific ioctl.
210 */
211/* ARGSUSED */
212int
213in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
214    struct thread *td)
215{
216	INIT_VNET_INET(curvnet); /* both so and ifp can be NULL here! */
217	register struct ifreq *ifr = (struct ifreq *)data;
218	register struct in_ifaddr *ia, *iap;
219	register struct ifaddr *ifa;
220	struct in_addr allhosts_addr;
221	struct in_addr dst;
222	struct in_ifaddr *oia;
223	struct in_ifinfo *ii;
224	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
225	struct sockaddr_in oldaddr;
226	int error, hostIsNew, iaIsNew, maskIsNew, s;
227	int iaIsFirst;
228
229	ia = NULL;
230	iaIsFirst = 0;
231	iaIsNew = 0;
232	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
233
234	/*
235	 * Filter out ioctls we implement directly; forward the rest on to
236	 * in_lifaddr_ioctl() and ifp->if_ioctl().
237	 */
238	switch (cmd) {
239	case SIOCAIFADDR:
240	case SIOCDIFADDR:
241	case SIOCGIFADDR:
242	case SIOCGIFBRDADDR:
243	case SIOCGIFDSTADDR:
244	case SIOCGIFNETMASK:
245	case SIOCSIFADDR:
246	case SIOCSIFBRDADDR:
247	case SIOCSIFDSTADDR:
248	case SIOCSIFNETMASK:
249		break;
250
251	case SIOCALIFADDR:
252		if (td != NULL) {
253			error = priv_check(td, PRIV_NET_ADDIFADDR);
254			if (error)
255				return (error);
256		}
257		if (ifp == NULL)
258			return (EINVAL);
259		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
260
261	case SIOCDLIFADDR:
262		if (td != NULL) {
263			error = priv_check(td, PRIV_NET_DELIFADDR);
264			if (error)
265				return (error);
266		}
267		if (ifp == NULL)
268			return (EINVAL);
269		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
270
271	case SIOCGLIFADDR:
272		if (ifp == NULL)
273			return (EINVAL);
274		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
275
276	default:
277		if (ifp == NULL || ifp->if_ioctl == NULL)
278			return (EOPNOTSUPP);
279		return ((*ifp->if_ioctl)(ifp, cmd, data));
280	}
281
282	if (ifp == NULL)
283		return (EADDRNOTAVAIL);
284
285	/*
286	 * Security checks before we get involved in any work.
287	 */
288	switch (cmd) {
289	case SIOCAIFADDR:
290	case SIOCSIFADDR:
291	case SIOCSIFBRDADDR:
292	case SIOCSIFNETMASK:
293	case SIOCSIFDSTADDR:
294		if (td != NULL) {
295			error = priv_check(td, PRIV_NET_ADDIFADDR);
296			if (error)
297				return (error);
298		}
299		break;
300
301	case SIOCDIFADDR:
302		if (td != NULL) {
303			error = priv_check(td, PRIV_NET_DELIFADDR);
304			if (error)
305				return (error);
306		}
307		break;
308	}
309
310	/*
311	 * Find address for this interface, if it exists.
312	 *
313	 * If an alias address was specified, find that one instead of the
314	 * first one on the interface, if possible.
315	 */
316	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
317	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
318		if (iap->ia_ifp == ifp &&
319		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
320			if (td == NULL || prison_check_ip4(td->td_ucred,
321			    &dst) == 0)
322				ia = iap;
323			break;
324		}
325	}
326	IF_ADDR_LOCK(ifp);
327	if (ia == NULL) {
328		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
329			iap = ifatoia(ifa);
330			if (iap->ia_addr.sin_family == AF_INET) {
331				if (td != NULL &&
332				    prison_check_ip4(td->td_ucred,
333				    &iap->ia_addr.sin_addr) != 0)
334					continue;
335				ia = iap;
336				break;
337			}
338		}
339	}
340	if (ia == NULL)
341		iaIsFirst = 1;
342
343	error = 0;
344	switch (cmd) {
345	case SIOCAIFADDR:
346	case SIOCDIFADDR:
347		if (ifra->ifra_addr.sin_family == AF_INET) {
348			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
349				if (ia->ia_ifp == ifp  &&
350				    ia->ia_addr.sin_addr.s_addr ==
351				    ifra->ifra_addr.sin_addr.s_addr)
352					break;
353			}
354			if ((ifp->if_flags & IFF_POINTOPOINT)
355			    && (cmd == SIOCAIFADDR)
356			    && (ifra->ifra_dstaddr.sin_addr.s_addr
357				== INADDR_ANY)) {
358				error = EDESTADDRREQ;
359				goto out_unlock;
360			}
361		}
362		if (cmd == SIOCDIFADDR && ia == NULL) {
363			error = EADDRNOTAVAIL;
364			goto out_unlock;
365		}
366		/* FALLTHROUGH */
367	case SIOCSIFADDR:
368	case SIOCSIFNETMASK:
369	case SIOCSIFDSTADDR:
370		if (ia == NULL) {
371			ia = (struct in_ifaddr *)
372				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
373				    M_ZERO);
374			if (ia == NULL) {
375				error = ENOBUFS;
376				goto out_unlock;
377			}
378
379			ifa = &ia->ia_ifa;
380			IFA_LOCK_INIT(ifa);
381			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
382			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
383			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
384			ifa->ifa_refcnt = 1;
385
386			ia->ia_sockmask.sin_len = 8;
387			ia->ia_sockmask.sin_family = AF_INET;
388			if (ifp->if_flags & IFF_BROADCAST) {
389				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
390				ia->ia_broadaddr.sin_family = AF_INET;
391			}
392			ia->ia_ifp = ifp;
393
394			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
395			s = splnet();
396			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
397			splx(s);
398			iaIsNew = 1;
399		}
400		break;
401
402	case SIOCSIFBRDADDR:
403	case SIOCGIFADDR:
404	case SIOCGIFNETMASK:
405	case SIOCGIFDSTADDR:
406	case SIOCGIFBRDADDR:
407		if (ia == NULL) {
408			error = EADDRNOTAVAIL;
409			goto out_unlock;
410		}
411		break;
412	}
413
414	/*
415	 * Most paths in this switch return directly or via out_unlock.  Only
416	 * paths that remove the address break in order to hit common removal
417	 * code.
418	 *
419	 * XXXRW: We enter the switch with IF_ADDR_LOCK() held, but leave
420	 * without it.  This is a bug.
421	 */
422	IF_ADDR_LOCK_ASSERT(ifp);
423	switch (cmd) {
424	case SIOCGIFADDR:
425		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
426		goto out_unlock;
427
428	case SIOCGIFBRDADDR:
429		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
430			error = EINVAL;
431			goto out_unlock;
432		}
433		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
434		goto out_unlock;
435
436	case SIOCGIFDSTADDR:
437		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
438			error = EINVAL;
439			goto out_unlock;
440		}
441		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
442		goto out_unlock;
443
444	case SIOCGIFNETMASK:
445		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
446		goto out_unlock;
447
448	case SIOCSIFDSTADDR:
449		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
450			error = EINVAL;
451			goto out_unlock;
452		}
453		oldaddr = ia->ia_dstaddr;
454		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
455		IF_ADDR_UNLOCK(ifp);
456
457		/*
458		 * XXXRW: Locks dropped for if_ioctl and rtinit, but ia is
459		 * still being used.
460		 */
461		if (ifp->if_ioctl != NULL) {
462			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
463			    (caddr_t)ia);
464			if (error) {
465				ia->ia_dstaddr = oldaddr;
466				return (error);
467			}
468		}
469		if (ia->ia_flags & IFA_ROUTE) {
470			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
471			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
472			ia->ia_ifa.ifa_dstaddr =
473					(struct sockaddr *)&ia->ia_dstaddr;
474			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
475		}
476		return (0);
477
478	case SIOCSIFBRDADDR:
479		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
480			error = EINVAL;
481			goto out_unlock;
482		}
483		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
484		goto out_unlock;
485
486	case SIOCSIFADDR:
487		IF_ADDR_UNLOCK(ifp);
488
489		/*
490		 * XXXRW: Locks dropped for in_ifinit and in_joingroup, but ia
491		 * is still being used.
492		 */
493		error = in_ifinit(ifp, ia,
494		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
495		if (error != 0 && iaIsNew)
496			break;
497		if (error == 0) {
498			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
499			if (iaIsFirst &&
500			    (ifp->if_flags & IFF_MULTICAST) != 0) {
501				error = in_joingroup(ifp, &allhosts_addr,
502				    NULL, &ii->ii_allhosts);
503			}
504			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
505		}
506		return (0);
507
508	case SIOCSIFNETMASK:
509		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
510		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
511		goto out_unlock;
512
513	case SIOCAIFADDR:
514		maskIsNew = 0;
515		hostIsNew = 1;
516		error = 0;
517		if (ia->ia_addr.sin_family == AF_INET) {
518			if (ifra->ifra_addr.sin_len == 0) {
519				ifra->ifra_addr = ia->ia_addr;
520				hostIsNew = 0;
521			} else if (ifra->ifra_addr.sin_addr.s_addr ==
522					       ia->ia_addr.sin_addr.s_addr)
523				hostIsNew = 0;
524		}
525		IF_ADDR_UNLOCK(ifp);
526
527		/*
528		 * XXXRW: Locks dropped for in_ifscrub and in_ifinit, but ia
529		 * is still being used.
530		 */
531		if (ifra->ifra_mask.sin_len) {
532			in_ifscrub(ifp, ia);
533			ia->ia_sockmask = ifra->ifra_mask;
534			ia->ia_sockmask.sin_family = AF_INET;
535			ia->ia_subnetmask =
536			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
537			maskIsNew = 1;
538		}
539		if ((ifp->if_flags & IFF_POINTOPOINT) &&
540		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
541			in_ifscrub(ifp, ia);
542			ia->ia_dstaddr = ifra->ifra_dstaddr;
543			maskIsNew  = 1; /* We lie; but the effect's the same */
544		}
545		if (ifra->ifra_addr.sin_family == AF_INET &&
546		    (hostIsNew || maskIsNew))
547			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
548		if (error != 0 && iaIsNew)
549			break;
550
551		if ((ifp->if_flags & IFF_BROADCAST) &&
552		    (ifra->ifra_broadaddr.sin_family == AF_INET))
553			ia->ia_broadaddr = ifra->ifra_broadaddr;
554		if (error == 0) {
555			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
556			if (iaIsFirst &&
557			    (ifp->if_flags & IFF_MULTICAST) != 0) {
558				error = in_joingroup(ifp, &allhosts_addr,
559				    NULL, &ii->ii_allhosts);
560			}
561			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
562		}
563		return (error);
564
565	case SIOCDIFADDR:
566		IF_ADDR_UNLOCK(ifp);
567
568		/*
569		 * XXXRW: Locks dropped for in_ifscrub and in_ifadown, but ia
570		 * is still being used.
571		 *
572		 * in_ifscrub kills the interface route.
573		 */
574		in_ifscrub(ifp, ia);
575
576		/*
577		 * in_ifadown gets rid of all the rest of
578		 * the routes.  This is not quite the right
579		 * thing to do, but at least if we are running
580		 * a routing process they will come back.
581		 */
582		in_ifadown(&ia->ia_ifa, 1);
583		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
584		error = 0;
585		break;
586
587	default:
588		panic("in_control: unsupported ioctl");
589	}
590
591	/*
592	 * XXXRW: In a more ideal world, we would still be holding
593	 * IF_ADDR_LOCK here.
594	 */
595	IF_ADDR_LOCK(ifp);
596	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
597	IF_ADDR_UNLOCK(ifp);
598	s = splnet();
599	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
600	if (ia->ia_addr.sin_family == AF_INET) {
601		LIST_REMOVE(ia, ia_hash);
602		/*
603		 * If this is the last IPv4 address configured on this
604		 * interface, leave the all-hosts group.
605		 * No state-change report need be transmitted.
606		 */
607		oia = NULL;
608		IFP_TO_IA(ifp, oia);
609		if (oia == NULL) {
610			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
611			IN_MULTI_LOCK();
612			if (ii->ii_allhosts) {
613				(void)in_leavegroup_locked(ii->ii_allhosts,
614				    NULL);
615				ii->ii_allhosts = NULL;
616			}
617			IN_MULTI_UNLOCK();
618		}
619	}
620	IFAFREE(&ia->ia_ifa);
621	splx(s);
622
623	return (error);
624
625out_unlock:
626	IF_ADDR_UNLOCK(ifp);
627	return (error);
628}
629
630/*
631 * SIOC[GAD]LIFADDR.
632 *	SIOCGLIFADDR: get first address. (?!?)
633 *	SIOCGLIFADDR with IFLR_PREFIX:
634 *		get first address that matches the specified prefix.
635 *	SIOCALIFADDR: add the specified address.
636 *	SIOCALIFADDR with IFLR_PREFIX:
637 *		EINVAL since we can't deduce hostid part of the address.
638 *	SIOCDLIFADDR: delete the specified address.
639 *	SIOCDLIFADDR with IFLR_PREFIX:
640 *		delete the first address that matches the specified prefix.
641 * return values:
642 *	EINVAL on invalid parameters
643 *	EADDRNOTAVAIL on prefix match failed/specified address not found
644 *	other values may be returned from in_ioctl()
645 */
646static int
647in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
648    struct ifnet *ifp, struct thread *td)
649{
650	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
651	struct ifaddr *ifa;
652
653	/* sanity checks */
654	if (data == NULL || ifp == NULL) {
655		panic("invalid argument to in_lifaddr_ioctl");
656		/*NOTRECHED*/
657	}
658
659	switch (cmd) {
660	case SIOCGLIFADDR:
661		/* address must be specified on GET with IFLR_PREFIX */
662		if ((iflr->flags & IFLR_PREFIX) == 0)
663			break;
664		/*FALLTHROUGH*/
665	case SIOCALIFADDR:
666	case SIOCDLIFADDR:
667		/* address must be specified on ADD and DELETE */
668		if (iflr->addr.ss_family != AF_INET)
669			return (EINVAL);
670		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
671			return (EINVAL);
672		/* XXX need improvement */
673		if (iflr->dstaddr.ss_family
674		 && iflr->dstaddr.ss_family != AF_INET)
675			return (EINVAL);
676		if (iflr->dstaddr.ss_family
677		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
678			return (EINVAL);
679		break;
680	default: /*shouldn't happen*/
681		return (EOPNOTSUPP);
682	}
683	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
684		return (EINVAL);
685
686	switch (cmd) {
687	case SIOCALIFADDR:
688	    {
689		struct in_aliasreq ifra;
690
691		if (iflr->flags & IFLR_PREFIX)
692			return (EINVAL);
693
694		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
695		bzero(&ifra, sizeof(ifra));
696		bcopy(iflr->iflr_name, ifra.ifra_name,
697			sizeof(ifra.ifra_name));
698
699		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
700
701		if (iflr->dstaddr.ss_family) {	/*XXX*/
702			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
703				iflr->dstaddr.ss_len);
704		}
705
706		ifra.ifra_mask.sin_family = AF_INET;
707		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
708		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
709
710		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
711	    }
712	case SIOCGLIFADDR:
713	case SIOCDLIFADDR:
714	    {
715		struct in_ifaddr *ia;
716		struct in_addr mask, candidate, match;
717		struct sockaddr_in *sin;
718
719		bzero(&mask, sizeof(mask));
720		bzero(&match, sizeof(match));
721		if (iflr->flags & IFLR_PREFIX) {
722			/* lookup a prefix rather than address. */
723			in_len2mask(&mask, iflr->prefixlen);
724
725			sin = (struct sockaddr_in *)&iflr->addr;
726			match.s_addr = sin->sin_addr.s_addr;
727			match.s_addr &= mask.s_addr;
728
729			/* if you set extra bits, that's wrong */
730			if (match.s_addr != sin->sin_addr.s_addr)
731				return (EINVAL);
732
733		} else {
734			/* on getting an address, take the 1st match */
735			/* on deleting an address, do exact match */
736			if (cmd != SIOCGLIFADDR) {
737				in_len2mask(&mask, 32);
738				sin = (struct sockaddr_in *)&iflr->addr;
739				match.s_addr = sin->sin_addr.s_addr;
740			}
741		}
742
743		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
744			if (ifa->ifa_addr->sa_family != AF_INET6)
745				continue;
746			if (match.s_addr == 0)
747				break;
748			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
749			candidate.s_addr &= mask.s_addr;
750			if (candidate.s_addr == match.s_addr)
751				break;
752		}
753		if (ifa == NULL)
754			return (EADDRNOTAVAIL);
755		ia = (struct in_ifaddr *)ifa;
756
757		if (cmd == SIOCGLIFADDR) {
758			/* fill in the if_laddrreq structure */
759			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
760
761			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
762				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
763					ia->ia_dstaddr.sin_len);
764			} else
765				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
766
767			iflr->prefixlen =
768				in_mask2len(&ia->ia_sockmask.sin_addr);
769
770			iflr->flags = 0;	/*XXX*/
771
772			return (0);
773		} else {
774			struct in_aliasreq ifra;
775
776			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
777			bzero(&ifra, sizeof(ifra));
778			bcopy(iflr->iflr_name, ifra.ifra_name,
779				sizeof(ifra.ifra_name));
780
781			bcopy(&ia->ia_addr, &ifra.ifra_addr,
782				ia->ia_addr.sin_len);
783			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
784				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
785					ia->ia_dstaddr.sin_len);
786			}
787			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
788				ia->ia_sockmask.sin_len);
789
790			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
791			    ifp, td));
792		}
793	    }
794	}
795
796	return (EOPNOTSUPP);	/*just for safety*/
797}
798
799/*
800 * Delete any existing route for an interface.
801 */
802void
803in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
804{
805
806	in_scrubprefix(ia);
807}
808
809/*
810 * Initialize an interface's internet address
811 * and routing table entry.
812 */
813static int
814in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
815    int scrub)
816{
817	INIT_VNET_NET(ifp->if_vnet);
818	INIT_VNET_INET(ifp->if_vnet);
819	register u_long i = ntohl(sin->sin_addr.s_addr);
820	struct sockaddr_in oldaddr;
821	struct rtentry *rt = NULL;
822	struct rt_addrinfo info;
823	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
824	int s = splimp(), flags = RTF_UP, error = 0;
825
826	oldaddr = ia->ia_addr;
827	if (oldaddr.sin_family == AF_INET)
828		LIST_REMOVE(ia, ia_hash);
829	ia->ia_addr = *sin;
830	if (ia->ia_addr.sin_family == AF_INET)
831		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
832		    ia, ia_hash);
833	/*
834	 * Give the interface a chance to initialize
835	 * if this is its first address,
836	 * and to validate the address if necessary.
837	 */
838	if (ifp->if_ioctl != NULL) {
839		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
840		if (error) {
841			splx(s);
842			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
843			ia->ia_addr = oldaddr;
844			if (ia->ia_addr.sin_family == AF_INET)
845				LIST_INSERT_HEAD(INADDR_HASH(
846				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
847			else
848				/*
849				 * If oldaddr family is not AF_INET (e.g.
850				 * interface has been just created) in_control
851				 * does not call LIST_REMOVE, and we end up
852				 * with bogus ia entries in hash
853				 */
854				LIST_REMOVE(ia, ia_hash);
855			return (error);
856		}
857	}
858	splx(s);
859	if (scrub) {
860		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
861		in_ifscrub(ifp, ia);
862		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
863	}
864	if (IN_CLASSA(i))
865		ia->ia_netmask = IN_CLASSA_NET;
866	else if (IN_CLASSB(i))
867		ia->ia_netmask = IN_CLASSB_NET;
868	else
869		ia->ia_netmask = IN_CLASSC_NET;
870	/*
871	 * The subnet mask usually includes at least the standard network part,
872	 * but may may be smaller in the case of supernetting.
873	 * If it is set, we believe it.
874	 */
875	if (ia->ia_subnetmask == 0) {
876		ia->ia_subnetmask = ia->ia_netmask;
877		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
878	} else
879		ia->ia_netmask &= ia->ia_subnetmask;
880	ia->ia_net = i & ia->ia_netmask;
881	ia->ia_subnet = i & ia->ia_subnetmask;
882	in_socktrim(&ia->ia_sockmask);
883#ifdef DEV_CARP
884	/*
885	 * XXX: carp(4) does not have interface route
886	 */
887	if (ifp->if_type == IFT_CARP)
888		return (0);
889#endif
890	/*
891	 * Add route for the network.
892	 */
893	ia->ia_ifa.ifa_metric = ifp->if_metric;
894	if (ifp->if_flags & IFF_BROADCAST) {
895		ia->ia_broadaddr.sin_addr.s_addr =
896			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
897		ia->ia_netbroadcast.s_addr =
898			htonl(ia->ia_net | ~ ia->ia_netmask);
899	} else if (ifp->if_flags & IFF_LOOPBACK) {
900		ia->ia_dstaddr = ia->ia_addr;
901		flags |= RTF_HOST;
902	} else if (ifp->if_flags & IFF_POINTOPOINT) {
903		if (ia->ia_dstaddr.sin_family != AF_INET)
904			return (0);
905		flags |= RTF_HOST;
906	}
907	if ((error = in_addprefix(ia, flags)) != 0)
908		return (error);
909
910	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
911		return (0);
912
913	/*
914	 * add a loopback route to self
915	 */
916	if (!(ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT))) {
917		bzero(&info, sizeof(info));
918		info.rti_ifp = V_loif;
919		info.rti_flags = ia->ia_flags | RTF_HOST | RTF_STATIC;
920		info.rti_info[RTAX_DST] = (struct sockaddr *)&ia->ia_addr;
921		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
922		error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
923
924		if (error == 0 && rt != NULL) {
925			RT_LOCK(rt);
926			((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
927				rt->rt_ifp->if_type;
928			((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
929				rt->rt_ifp->if_index;
930			RT_REMREF(rt);
931			RT_UNLOCK(rt);
932		} else if (error != 0)
933			log(LOG_INFO, "in_ifinit: insertion failed\n");
934	}
935
936	return (error);
937}
938
939#define rtinitflags(x) \
940	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
941	    ? RTF_HOST : 0)
942/*
943 * Check if we have a route for the given prefix already or add one accordingly.
944 */
945static int
946in_addprefix(struct in_ifaddr *target, int flags)
947{
948	INIT_VNET_INET(curvnet);
949	struct in_ifaddr *ia;
950	struct in_addr prefix, mask, p, m;
951	int error;
952
953	if ((flags & RTF_HOST) != 0) {
954		prefix = target->ia_dstaddr.sin_addr;
955		mask.s_addr = 0;
956	} else {
957		prefix = target->ia_addr.sin_addr;
958		mask = target->ia_sockmask.sin_addr;
959		prefix.s_addr &= mask.s_addr;
960	}
961
962	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
963		if (rtinitflags(ia)) {
964			p = ia->ia_addr.sin_addr;
965
966			if (prefix.s_addr != p.s_addr)
967				continue;
968		} else {
969			p = ia->ia_addr.sin_addr;
970			m = ia->ia_sockmask.sin_addr;
971			p.s_addr &= m.s_addr;
972
973			if (prefix.s_addr != p.s_addr ||
974			    mask.s_addr != m.s_addr)
975				continue;
976		}
977
978		/*
979		 * If we got a matching prefix route inserted by other
980		 * interface address, we are done here.
981		 */
982		if (ia->ia_flags & IFA_ROUTE) {
983			if (V_sameprefixcarponly &&
984			    target->ia_ifp->if_type != IFT_CARP &&
985			    ia->ia_ifp->if_type != IFT_CARP)
986				return (EEXIST);
987			else
988				return (0);
989		}
990	}
991
992	/*
993	 * No-one seem to have this prefix route, so we try to insert it.
994	 */
995	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
996	if (!error)
997		target->ia_flags |= IFA_ROUTE;
998	return (error);
999}
1000
1001extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1002
1003/*
1004 * If there is no other address in the system that can serve a route to the
1005 * same prefix, remove the route.  Hand over the route to the new address
1006 * otherwise.
1007 */
1008static int
1009in_scrubprefix(struct in_ifaddr *target)
1010{
1011	INIT_VNET_NET(curvnet);
1012	INIT_VNET_INET(curvnet);
1013	struct in_ifaddr *ia;
1014	struct in_addr prefix, mask, p;
1015	int error;
1016	struct rt_addrinfo info;
1017	struct sockaddr_dl null_sdl;
1018
1019	if ((target->ia_flags & IFA_ROUTE) == 0)
1020		return (0);
1021
1022	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1023	    !(target->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT))) {
1024		bzero(&null_sdl, sizeof(null_sdl));
1025		null_sdl.sdl_len = sizeof(null_sdl);
1026		null_sdl.sdl_family = AF_LINK;
1027		null_sdl.sdl_type = V_loif->if_type;
1028		null_sdl.sdl_index = V_loif->if_index;
1029		bzero(&info, sizeof(info));
1030		info.rti_flags = target->ia_flags | RTF_HOST | RTF_STATIC;
1031		info.rti_info[RTAX_DST] = (struct sockaddr *)&target->ia_addr;
1032		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1033		error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1034
1035		if (error != 0)
1036			log(LOG_INFO, "in_scrubprefix: deletion failed\n");
1037	}
1038
1039	if (rtinitflags(target))
1040		prefix = target->ia_dstaddr.sin_addr;
1041	else {
1042		prefix = target->ia_addr.sin_addr;
1043		mask = target->ia_sockmask.sin_addr;
1044		prefix.s_addr &= mask.s_addr;
1045		/* remove arp cache */
1046		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1047	}
1048
1049	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1050		if (rtinitflags(ia))
1051			p = ia->ia_dstaddr.sin_addr;
1052		else {
1053			p = ia->ia_addr.sin_addr;
1054			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1055		}
1056
1057		if (prefix.s_addr != p.s_addr)
1058			continue;
1059
1060		/*
1061		 * If we got a matching prefix address, move IFA_ROUTE and
1062		 * the route itself to it.  Make sure that routing daemons
1063		 * get a heads-up.
1064		 *
1065		 * XXX: a special case for carp(4) interface
1066		 */
1067		if ((ia->ia_flags & IFA_ROUTE) == 0
1068#ifdef DEV_CARP
1069		    && (ia->ia_ifp->if_type != IFT_CARP)
1070#endif
1071							) {
1072			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1073			    rtinitflags(target));
1074			target->ia_flags &= ~IFA_ROUTE;
1075
1076			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1077			    rtinitflags(ia) | RTF_UP);
1078			if (error == 0)
1079				ia->ia_flags |= IFA_ROUTE;
1080			return (error);
1081		}
1082	}
1083
1084	/*
1085	 * As no-one seem to have this prefix, we can remove the route.
1086	 */
1087	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1088	target->ia_flags &= ~IFA_ROUTE;
1089	return (0);
1090}
1091
1092#undef rtinitflags
1093
1094/*
1095 * Return 1 if the address might be a local broadcast address.
1096 */
1097int
1098in_broadcast(struct in_addr in, struct ifnet *ifp)
1099{
1100	register struct ifaddr *ifa;
1101	u_long t;
1102
1103	if (in.s_addr == INADDR_BROADCAST ||
1104	    in.s_addr == INADDR_ANY)
1105		return (1);
1106	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1107		return (0);
1108	t = ntohl(in.s_addr);
1109	/*
1110	 * Look through the list of addresses for a match
1111	 * with a broadcast address.
1112	 */
1113#define ia ((struct in_ifaddr *)ifa)
1114	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1115		if (ifa->ifa_addr->sa_family == AF_INET &&
1116		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1117		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1118		     /*
1119		      * Check for old-style (host 0) broadcast.
1120		      */
1121		     t == ia->ia_subnet || t == ia->ia_net) &&
1122		     /*
1123		      * Check for an all one subnetmask. These
1124		      * only exist when an interface gets a secondary
1125		      * address.
1126		      */
1127		     ia->ia_subnetmask != (u_long)0xffffffff)
1128			    return (1);
1129	return (0);
1130#undef ia
1131}
1132
1133/*
1134 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1135 */
1136void
1137in_ifdetach(struct ifnet *ifp)
1138{
1139	INIT_VNET_INET(ifp->if_vnet);
1140
1141	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1142	in_pcbpurgeif0(&V_udbinfo, ifp);
1143	in_purgemaddrs(ifp);
1144}
1145
1146/*
1147 * Delete all IPv4 multicast address records, and associated link-layer
1148 * multicast address records, associated with ifp.
1149 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1150 * XXX This should not race with ifma_protospec being set during
1151 * a new allocation, if it does, we have bigger problems.
1152 */
1153static void
1154in_purgemaddrs(struct ifnet *ifp)
1155{
1156	LIST_HEAD(,in_multi) purgeinms;
1157	struct in_multi		*inm, *tinm;
1158	struct ifmultiaddr	*ifma;
1159
1160	LIST_INIT(&purgeinms);
1161	IN_MULTI_LOCK();
1162
1163	/*
1164	 * Extract list of in_multi associated with the detaching ifp
1165	 * which the PF_INET layer is about to release.
1166	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1167	 * by code further down.
1168	 */
1169	IF_ADDR_LOCK(ifp);
1170	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1171		if (ifma->ifma_addr->sa_family != AF_INET ||
1172		    ifma->ifma_protospec == NULL)
1173			continue;
1174#if 0
1175		KASSERT(ifma->ifma_protospec != NULL,
1176		    ("%s: ifma_protospec is NULL", __func__));
1177#endif
1178		inm = (struct in_multi *)ifma->ifma_protospec;
1179		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1180	}
1181	IF_ADDR_UNLOCK(ifp);
1182
1183	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1184		LIST_REMOVE(inm, inm_link);
1185		inm_release_locked(inm);
1186	}
1187	igmp_ifdetach(ifp);
1188
1189	IN_MULTI_UNLOCK();
1190}
1191
1192#include <net/if_dl.h>
1193#include <netinet/if_ether.h>
1194
1195struct in_llentry {
1196	struct llentry		base;
1197	struct sockaddr_in	l3_addr4;
1198};
1199
1200static struct llentry *
1201in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1202{
1203	struct in_llentry *lle;
1204
1205	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1206	if (lle == NULL)		/* NB: caller generates msg */
1207		return NULL;
1208
1209	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1210	/*
1211	 * For IPv4 this will trigger "arpresolve" to generate
1212	 * an ARP request.
1213	 */
1214	lle->base.la_expire = time_second; /* mark expired */
1215	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1216	lle->base.lle_refcnt = 1;
1217	LLE_LOCK_INIT(&lle->base);
1218	return &lle->base;
1219}
1220
1221/*
1222 * Deletes an address from the address table.
1223 * This function is called by the timer functions
1224 * such as arptimer() and nd6_llinfo_timer(), and
1225 * the caller does the locking.
1226 */
1227static void
1228in_lltable_free(struct lltable *llt, struct llentry *lle)
1229{
1230	LLE_WUNLOCK(lle);
1231	LLE_LOCK_DESTROY(lle);
1232	free(lle, M_LLTABLE);
1233}
1234
1235static int
1236in_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
1237{
1238	struct rtentry *rt;
1239
1240	KASSERT(l3addr->sa_family == AF_INET,
1241	    ("sin_family %d", l3addr->sa_family));
1242
1243	/* XXX rtalloc1 should take a const param */
1244	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1245	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
1246		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1247		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1248		if (rt != NULL)
1249			RTFREE_LOCKED(rt);
1250		return (EINVAL);
1251	}
1252	RTFREE_LOCKED(rt);
1253	return 0;
1254}
1255
1256/*
1257 * Return NULL if not found or marked for deletion.
1258 * If found return lle read locked.
1259 */
1260static struct llentry *
1261in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1262{
1263	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1264	struct ifnet *ifp = llt->llt_ifp;
1265	struct llentry *lle;
1266	struct llentries *lleh;
1267	u_int hashkey;
1268
1269	IF_AFDATA_LOCK_ASSERT(ifp);
1270	KASSERT(l3addr->sa_family == AF_INET,
1271	    ("sin_family %d", l3addr->sa_family));
1272
1273	hashkey = sin->sin_addr.s_addr;
1274	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1275	LIST_FOREACH(lle, lleh, lle_next) {
1276		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1277		if (lle->la_flags & LLE_DELETED)
1278			continue;
1279		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1280			break;
1281	}
1282	if (lle == NULL) {
1283#ifdef DIAGNOSTICS
1284		if (flags & LLE_DELETE)
1285			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1286#endif
1287		if (!(flags & LLE_CREATE))
1288			return (NULL);
1289		/*
1290		 * A route that covers the given address must have
1291		 * been installed 1st because we are doing a resolution,
1292		 * verify this.
1293		 */
1294		if (!(flags & LLE_IFADDR) &&
1295		    in_lltable_rtcheck(ifp, l3addr) != 0)
1296			goto done;
1297
1298		lle = in_lltable_new(l3addr, flags);
1299		if (lle == NULL) {
1300			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1301			goto done;
1302		}
1303		lle->la_flags = flags & ~LLE_CREATE;
1304		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1305			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1306			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1307		}
1308
1309		lle->lle_tbl  = llt;
1310		lle->lle_head = lleh;
1311		LIST_INSERT_HEAD(lleh, lle, lle_next);
1312	} else if (flags & LLE_DELETE) {
1313		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1314			LLE_WLOCK(lle);
1315			lle->la_flags = LLE_DELETED;
1316			LLE_WUNLOCK(lle);
1317#ifdef DIAGNOSTICS
1318			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1319#endif
1320		}
1321		lle = (void *)-1;
1322
1323	}
1324	if (LLE_IS_VALID(lle)) {
1325		if (flags & LLE_EXCLUSIVE)
1326			LLE_WLOCK(lle);
1327		else
1328			LLE_RLOCK(lle);
1329	}
1330done:
1331	return (lle);
1332}
1333
1334static int
1335in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1336{
1337#define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1338	struct ifnet *ifp = llt->llt_ifp;
1339	struct llentry *lle;
1340	/* XXX stack use */
1341	struct {
1342		struct rt_msghdr	rtm;
1343		struct sockaddr_inarp	sin;
1344		struct sockaddr_dl	sdl;
1345	} arpc;
1346	int error, i;
1347
1348	/* XXXXX
1349	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
1350	 * so it is okay to use this ASSERT, change it when
1351	 * IFNET lock is finalized
1352	 */
1353	IFNET_WLOCK_ASSERT();
1354
1355	error = 0;
1356	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1357		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1358			struct sockaddr_dl *sdl;
1359
1360			/* skip deleted entries */
1361			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
1362				continue;
1363			/* Skip if jailed and not a valid IP of the prison. */
1364			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1365				continue;
1366			/*
1367			 * produce a msg made of:
1368			 *  struct rt_msghdr;
1369			 *  struct sockaddr_inarp; (IPv4)
1370			 *  struct sockaddr_dl;
1371			 */
1372			bzero(&arpc, sizeof(arpc));
1373			arpc.rtm.rtm_msglen = sizeof(arpc);
1374			arpc.rtm.rtm_version = RTM_VERSION;
1375			arpc.rtm.rtm_type = RTM_GET;
1376			arpc.rtm.rtm_flags = RTF_UP;
1377			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1378			arpc.sin.sin_family = AF_INET;
1379			arpc.sin.sin_len = sizeof(arpc.sin);
1380			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1381
1382			/* publish */
1383			if (lle->la_flags & LLE_PUB) {
1384				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1385				/* proxy only */
1386				if (lle->la_flags & LLE_PROXY)
1387					arpc.sin.sin_other = SIN_PROXY;
1388			}
1389
1390			sdl = &arpc.sdl;
1391			sdl->sdl_family = AF_LINK;
1392			sdl->sdl_len = sizeof(*sdl);
1393			sdl->sdl_alen = ifp->if_addrlen;
1394			sdl->sdl_index = ifp->if_index;
1395			sdl->sdl_type = ifp->if_type;
1396			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1397
1398			arpc.rtm.rtm_rmx.rmx_expire =
1399			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1400			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1401			if (lle->la_flags & LLE_STATIC)
1402				arpc.rtm.rtm_flags |= RTF_STATIC;
1403			arpc.rtm.rtm_index = ifp->if_index;
1404			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1405			if (error)
1406				break;
1407		}
1408	}
1409	return error;
1410#undef SIN
1411}
1412
1413void *
1414in_domifattach(struct ifnet *ifp)
1415{
1416	struct in_ifinfo *ii;
1417	struct lltable *llt;
1418
1419	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1420
1421	llt = lltable_init(ifp, AF_INET);
1422	if (llt != NULL) {
1423		llt->llt_new = in_lltable_new;
1424		llt->llt_free = in_lltable_free;
1425		llt->llt_rtcheck = in_lltable_rtcheck;
1426		llt->llt_lookup = in_lltable_lookup;
1427		llt->llt_dump = in_lltable_dump;
1428	}
1429	ii->ii_llt = llt;
1430
1431	ii->ii_igmp = igmp_domifattach(ifp);
1432
1433	return ii;
1434}
1435
1436void
1437in_domifdetach(struct ifnet *ifp, void *aux)
1438{
1439	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1440
1441	igmp_domifdetach(ifp);
1442	lltable_free(ii->ii_llt);
1443	free(ii, M_IFADDR);
1444}
1445