in.c revision 194602
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 194602 2009-06-21 19:30:33Z rwatson $");
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_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
385			ia->ia_sockmask.sin_len = 8;
386			ia->ia_sockmask.sin_family = AF_INET;
387			if (ifp->if_flags & IFF_BROADCAST) {
388				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
389				ia->ia_broadaddr.sin_family = AF_INET;
390			}
391			ia->ia_ifp = ifp;
392
393			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
394			s = splnet();
395			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
396			splx(s);
397			iaIsNew = 1;
398		}
399		break;
400
401	case SIOCSIFBRDADDR:
402	case SIOCGIFADDR:
403	case SIOCGIFNETMASK:
404	case SIOCGIFDSTADDR:
405	case SIOCGIFBRDADDR:
406		if (ia == NULL) {
407			error = EADDRNOTAVAIL;
408			goto out_unlock;
409		}
410		break;
411	}
412
413	/*
414	 * Most paths in this switch return directly or via out_unlock.  Only
415	 * paths that remove the address break in order to hit common removal
416	 * code.
417	 *
418	 * XXXRW: We enter the switch with IF_ADDR_LOCK() held, but leave
419	 * without it.  This is a bug.
420	 */
421	IF_ADDR_LOCK_ASSERT(ifp);
422	switch (cmd) {
423	case SIOCGIFADDR:
424		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
425		goto out_unlock;
426
427	case SIOCGIFBRDADDR:
428		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
429			error = EINVAL;
430			goto out_unlock;
431		}
432		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
433		goto out_unlock;
434
435	case SIOCGIFDSTADDR:
436		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
437			error = EINVAL;
438			goto out_unlock;
439		}
440		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
441		goto out_unlock;
442
443	case SIOCGIFNETMASK:
444		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
445		goto out_unlock;
446
447	case SIOCSIFDSTADDR:
448		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
449			error = EINVAL;
450			goto out_unlock;
451		}
452		oldaddr = ia->ia_dstaddr;
453		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
454		IF_ADDR_UNLOCK(ifp);
455
456		/*
457		 * XXXRW: Locks dropped for if_ioctl and rtinit, but ia is
458		 * still being used.
459		 */
460		if (ifp->if_ioctl != NULL) {
461			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
462			    (caddr_t)ia);
463			if (error) {
464				ia->ia_dstaddr = oldaddr;
465				return (error);
466			}
467		}
468		if (ia->ia_flags & IFA_ROUTE) {
469			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
470			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
471			ia->ia_ifa.ifa_dstaddr =
472					(struct sockaddr *)&ia->ia_dstaddr;
473			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
474		}
475		return (0);
476
477	case SIOCSIFBRDADDR:
478		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
479			error = EINVAL;
480			goto out_unlock;
481		}
482		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
483		goto out_unlock;
484
485	case SIOCSIFADDR:
486		IF_ADDR_UNLOCK(ifp);
487
488		/*
489		 * XXXRW: Locks dropped for in_ifinit and in_joingroup, but ia
490		 * is still being used.
491		 */
492		error = in_ifinit(ifp, ia,
493		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
494		if (error != 0 && iaIsNew)
495			break;
496		if (error == 0) {
497			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
498			if (iaIsFirst &&
499			    (ifp->if_flags & IFF_MULTICAST) != 0) {
500				error = in_joingroup(ifp, &allhosts_addr,
501				    NULL, &ii->ii_allhosts);
502			}
503			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
504		}
505		return (0);
506
507	case SIOCSIFNETMASK:
508		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
509		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
510		goto out_unlock;
511
512	case SIOCAIFADDR:
513		maskIsNew = 0;
514		hostIsNew = 1;
515		error = 0;
516		if (ia->ia_addr.sin_family == AF_INET) {
517			if (ifra->ifra_addr.sin_len == 0) {
518				ifra->ifra_addr = ia->ia_addr;
519				hostIsNew = 0;
520			} else if (ifra->ifra_addr.sin_addr.s_addr ==
521					       ia->ia_addr.sin_addr.s_addr)
522				hostIsNew = 0;
523		}
524		IF_ADDR_UNLOCK(ifp);
525
526		/*
527		 * XXXRW: Locks dropped for in_ifscrub and in_ifinit, but ia
528		 * is still being used.
529		 */
530		if (ifra->ifra_mask.sin_len) {
531			in_ifscrub(ifp, ia);
532			ia->ia_sockmask = ifra->ifra_mask;
533			ia->ia_sockmask.sin_family = AF_INET;
534			ia->ia_subnetmask =
535			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
536			maskIsNew = 1;
537		}
538		if ((ifp->if_flags & IFF_POINTOPOINT) &&
539		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
540			in_ifscrub(ifp, ia);
541			ia->ia_dstaddr = ifra->ifra_dstaddr;
542			maskIsNew  = 1; /* We lie; but the effect's the same */
543		}
544		if (ifra->ifra_addr.sin_family == AF_INET &&
545		    (hostIsNew || maskIsNew))
546			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
547		if (error != 0 && iaIsNew)
548			break;
549
550		if ((ifp->if_flags & IFF_BROADCAST) &&
551		    (ifra->ifra_broadaddr.sin_family == AF_INET))
552			ia->ia_broadaddr = ifra->ifra_broadaddr;
553		if (error == 0) {
554			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
555			if (iaIsFirst &&
556			    (ifp->if_flags & IFF_MULTICAST) != 0) {
557				error = in_joingroup(ifp, &allhosts_addr,
558				    NULL, &ii->ii_allhosts);
559			}
560			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
561		}
562		return (error);
563
564	case SIOCDIFADDR:
565		IF_ADDR_UNLOCK(ifp);
566
567		/*
568		 * XXXRW: Locks dropped for in_ifscrub and in_ifadown, but ia
569		 * is still being used.
570		 *
571		 * in_ifscrub kills the interface route.
572		 */
573		in_ifscrub(ifp, ia);
574
575		/*
576		 * in_ifadown gets rid of all the rest of
577		 * the routes.  This is not quite the right
578		 * thing to do, but at least if we are running
579		 * a routing process they will come back.
580		 */
581		in_ifadown(&ia->ia_ifa, 1);
582		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
583		error = 0;
584		break;
585
586	default:
587		panic("in_control: unsupported ioctl");
588	}
589
590	/*
591	 * XXXRW: In a more ideal world, we would still be holding
592	 * IF_ADDR_LOCK here.
593	 */
594	IF_ADDR_LOCK(ifp);
595	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
596	IF_ADDR_UNLOCK(ifp);
597	s = splnet();
598	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
599	if (ia->ia_addr.sin_family == AF_INET) {
600		LIST_REMOVE(ia, ia_hash);
601		/*
602		 * If this is the last IPv4 address configured on this
603		 * interface, leave the all-hosts group.
604		 * No state-change report need be transmitted.
605		 */
606		oia = NULL;
607		IFP_TO_IA(ifp, oia);
608		if (oia == NULL) {
609			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
610			IN_MULTI_LOCK();
611			if (ii->ii_allhosts) {
612				(void)in_leavegroup_locked(ii->ii_allhosts,
613				    NULL);
614				ii->ii_allhosts = NULL;
615			}
616			IN_MULTI_UNLOCK();
617		}
618	}
619	ifa_free(&ia->ia_ifa);
620	splx(s);
621
622	return (error);
623
624out_unlock:
625	IF_ADDR_UNLOCK(ifp);
626	return (error);
627}
628
629/*
630 * SIOC[GAD]LIFADDR.
631 *	SIOCGLIFADDR: get first address. (?!?)
632 *	SIOCGLIFADDR with IFLR_PREFIX:
633 *		get first address that matches the specified prefix.
634 *	SIOCALIFADDR: add the specified address.
635 *	SIOCALIFADDR with IFLR_PREFIX:
636 *		EINVAL since we can't deduce hostid part of the address.
637 *	SIOCDLIFADDR: delete the specified address.
638 *	SIOCDLIFADDR with IFLR_PREFIX:
639 *		delete the first address that matches the specified prefix.
640 * return values:
641 *	EINVAL on invalid parameters
642 *	EADDRNOTAVAIL on prefix match failed/specified address not found
643 *	other values may be returned from in_ioctl()
644 */
645static int
646in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
647    struct ifnet *ifp, struct thread *td)
648{
649	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
650	struct ifaddr *ifa;
651
652	/* sanity checks */
653	if (data == NULL || ifp == NULL) {
654		panic("invalid argument to in_lifaddr_ioctl");
655		/*NOTRECHED*/
656	}
657
658	switch (cmd) {
659	case SIOCGLIFADDR:
660		/* address must be specified on GET with IFLR_PREFIX */
661		if ((iflr->flags & IFLR_PREFIX) == 0)
662			break;
663		/*FALLTHROUGH*/
664	case SIOCALIFADDR:
665	case SIOCDLIFADDR:
666		/* address must be specified on ADD and DELETE */
667		if (iflr->addr.ss_family != AF_INET)
668			return (EINVAL);
669		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
670			return (EINVAL);
671		/* XXX need improvement */
672		if (iflr->dstaddr.ss_family
673		 && iflr->dstaddr.ss_family != AF_INET)
674			return (EINVAL);
675		if (iflr->dstaddr.ss_family
676		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
677			return (EINVAL);
678		break;
679	default: /*shouldn't happen*/
680		return (EOPNOTSUPP);
681	}
682	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
683		return (EINVAL);
684
685	switch (cmd) {
686	case SIOCALIFADDR:
687	    {
688		struct in_aliasreq ifra;
689
690		if (iflr->flags & IFLR_PREFIX)
691			return (EINVAL);
692
693		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
694		bzero(&ifra, sizeof(ifra));
695		bcopy(iflr->iflr_name, ifra.ifra_name,
696			sizeof(ifra.ifra_name));
697
698		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
699
700		if (iflr->dstaddr.ss_family) {	/*XXX*/
701			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
702				iflr->dstaddr.ss_len);
703		}
704
705		ifra.ifra_mask.sin_family = AF_INET;
706		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
707		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
708
709		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
710	    }
711	case SIOCGLIFADDR:
712	case SIOCDLIFADDR:
713	    {
714		struct in_ifaddr *ia;
715		struct in_addr mask, candidate, match;
716		struct sockaddr_in *sin;
717
718		bzero(&mask, sizeof(mask));
719		bzero(&match, sizeof(match));
720		if (iflr->flags & IFLR_PREFIX) {
721			/* lookup a prefix rather than address. */
722			in_len2mask(&mask, iflr->prefixlen);
723
724			sin = (struct sockaddr_in *)&iflr->addr;
725			match.s_addr = sin->sin_addr.s_addr;
726			match.s_addr &= mask.s_addr;
727
728			/* if you set extra bits, that's wrong */
729			if (match.s_addr != sin->sin_addr.s_addr)
730				return (EINVAL);
731
732		} else {
733			/* on getting an address, take the 1st match */
734			/* on deleting an address, do exact match */
735			if (cmd != SIOCGLIFADDR) {
736				in_len2mask(&mask, 32);
737				sin = (struct sockaddr_in *)&iflr->addr;
738				match.s_addr = sin->sin_addr.s_addr;
739			}
740		}
741
742		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
743			if (ifa->ifa_addr->sa_family != AF_INET6)
744				continue;
745			if (match.s_addr == 0)
746				break;
747			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
748			candidate.s_addr &= mask.s_addr;
749			if (candidate.s_addr == match.s_addr)
750				break;
751		}
752		if (ifa == NULL)
753			return (EADDRNOTAVAIL);
754		ia = (struct in_ifaddr *)ifa;
755
756		if (cmd == SIOCGLIFADDR) {
757			/* fill in the if_laddrreq structure */
758			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
759
760			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
761				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
762					ia->ia_dstaddr.sin_len);
763			} else
764				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
765
766			iflr->prefixlen =
767				in_mask2len(&ia->ia_sockmask.sin_addr);
768
769			iflr->flags = 0;	/*XXX*/
770
771			return (0);
772		} else {
773			struct in_aliasreq ifra;
774
775			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
776			bzero(&ifra, sizeof(ifra));
777			bcopy(iflr->iflr_name, ifra.ifra_name,
778				sizeof(ifra.ifra_name));
779
780			bcopy(&ia->ia_addr, &ifra.ifra_addr,
781				ia->ia_addr.sin_len);
782			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
783				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
784					ia->ia_dstaddr.sin_len);
785			}
786			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
787				ia->ia_sockmask.sin_len);
788
789			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
790			    ifp, td));
791		}
792	    }
793	}
794
795	return (EOPNOTSUPP);	/*just for safety*/
796}
797
798/*
799 * Delete any existing route for an interface.
800 */
801void
802in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
803{
804
805	in_scrubprefix(ia);
806}
807
808/*
809 * Initialize an interface's internet address
810 * and routing table entry.
811 */
812static int
813in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
814    int scrub)
815{
816	INIT_VNET_NET(ifp->if_vnet);
817	INIT_VNET_INET(ifp->if_vnet);
818	register u_long i = ntohl(sin->sin_addr.s_addr);
819	struct sockaddr_in oldaddr;
820	struct rtentry *rt = NULL;
821	struct rt_addrinfo info;
822	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
823	int s = splimp(), flags = RTF_UP, error = 0;
824
825	oldaddr = ia->ia_addr;
826	if (oldaddr.sin_family == AF_INET)
827		LIST_REMOVE(ia, ia_hash);
828	ia->ia_addr = *sin;
829	if (ia->ia_addr.sin_family == AF_INET)
830		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
831		    ia, ia_hash);
832	/*
833	 * Give the interface a chance to initialize
834	 * if this is its first address,
835	 * and to validate the address if necessary.
836	 */
837	if (ifp->if_ioctl != NULL) {
838		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
839		if (error) {
840			splx(s);
841			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
842			ia->ia_addr = oldaddr;
843			if (ia->ia_addr.sin_family == AF_INET)
844				LIST_INSERT_HEAD(INADDR_HASH(
845				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
846			else
847				/*
848				 * If oldaddr family is not AF_INET (e.g.
849				 * interface has been just created) in_control
850				 * does not call LIST_REMOVE, and we end up
851				 * with bogus ia entries in hash
852				 */
853				LIST_REMOVE(ia, ia_hash);
854			return (error);
855		}
856	}
857	splx(s);
858	if (scrub) {
859		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
860		in_ifscrub(ifp, ia);
861		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
862	}
863	if (IN_CLASSA(i))
864		ia->ia_netmask = IN_CLASSA_NET;
865	else if (IN_CLASSB(i))
866		ia->ia_netmask = IN_CLASSB_NET;
867	else
868		ia->ia_netmask = IN_CLASSC_NET;
869	/*
870	 * The subnet mask usually includes at least the standard network part,
871	 * but may may be smaller in the case of supernetting.
872	 * If it is set, we believe it.
873	 */
874	if (ia->ia_subnetmask == 0) {
875		ia->ia_subnetmask = ia->ia_netmask;
876		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
877	} else
878		ia->ia_netmask &= ia->ia_subnetmask;
879	ia->ia_net = i & ia->ia_netmask;
880	ia->ia_subnet = i & ia->ia_subnetmask;
881	in_socktrim(&ia->ia_sockmask);
882#ifdef DEV_CARP
883	/*
884	 * XXX: carp(4) does not have interface route
885	 */
886	if (ifp->if_type == IFT_CARP)
887		return (0);
888#endif
889	/*
890	 * Add route for the network.
891	 */
892	ia->ia_ifa.ifa_metric = ifp->if_metric;
893	if (ifp->if_flags & IFF_BROADCAST) {
894		ia->ia_broadaddr.sin_addr.s_addr =
895			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
896		ia->ia_netbroadcast.s_addr =
897			htonl(ia->ia_net | ~ ia->ia_netmask);
898	} else if (ifp->if_flags & IFF_LOOPBACK) {
899		ia->ia_dstaddr = ia->ia_addr;
900		flags |= RTF_HOST;
901	} else if (ifp->if_flags & IFF_POINTOPOINT) {
902		if (ia->ia_dstaddr.sin_family != AF_INET)
903			return (0);
904		flags |= RTF_HOST;
905	}
906	if ((error = in_addprefix(ia, flags)) != 0)
907		return (error);
908
909	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
910		return (0);
911
912	/*
913	 * add a loopback route to self
914	 */
915	if (!(ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT))) {
916		bzero(&info, sizeof(info));
917		info.rti_ifp = V_loif;
918		info.rti_flags = ia->ia_flags | RTF_HOST | RTF_STATIC;
919		info.rti_info[RTAX_DST] = (struct sockaddr *)&ia->ia_addr;
920		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
921		error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
922
923		if (error == 0 && rt != NULL) {
924			RT_LOCK(rt);
925			((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
926				rt->rt_ifp->if_type;
927			((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
928				rt->rt_ifp->if_index;
929			RT_REMREF(rt);
930			RT_UNLOCK(rt);
931		} else if (error != 0)
932			log(LOG_INFO, "in_ifinit: insertion failed\n");
933	}
934
935	return (error);
936}
937
938#define rtinitflags(x) \
939	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
940	    ? RTF_HOST : 0)
941/*
942 * Check if we have a route for the given prefix already or add one accordingly.
943 */
944static int
945in_addprefix(struct in_ifaddr *target, int flags)
946{
947	INIT_VNET_INET(curvnet);
948	struct in_ifaddr *ia;
949	struct in_addr prefix, mask, p, m;
950	int error;
951
952	if ((flags & RTF_HOST) != 0) {
953		prefix = target->ia_dstaddr.sin_addr;
954		mask.s_addr = 0;
955	} else {
956		prefix = target->ia_addr.sin_addr;
957		mask = target->ia_sockmask.sin_addr;
958		prefix.s_addr &= mask.s_addr;
959	}
960
961	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
962		if (rtinitflags(ia)) {
963			p = ia->ia_addr.sin_addr;
964
965			if (prefix.s_addr != p.s_addr)
966				continue;
967		} else {
968			p = ia->ia_addr.sin_addr;
969			m = ia->ia_sockmask.sin_addr;
970			p.s_addr &= m.s_addr;
971
972			if (prefix.s_addr != p.s_addr ||
973			    mask.s_addr != m.s_addr)
974				continue;
975		}
976
977		/*
978		 * If we got a matching prefix route inserted by other
979		 * interface address, we are done here.
980		 */
981		if (ia->ia_flags & IFA_ROUTE) {
982			if (V_sameprefixcarponly &&
983			    target->ia_ifp->if_type != IFT_CARP &&
984			    ia->ia_ifp->if_type != IFT_CARP)
985				return (EEXIST);
986			else
987				return (0);
988		}
989	}
990
991	/*
992	 * No-one seem to have this prefix route, so we try to insert it.
993	 */
994	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
995	if (!error)
996		target->ia_flags |= IFA_ROUTE;
997	return (error);
998}
999
1000extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1001
1002/*
1003 * If there is no other address in the system that can serve a route to the
1004 * same prefix, remove the route.  Hand over the route to the new address
1005 * otherwise.
1006 */
1007static int
1008in_scrubprefix(struct in_ifaddr *target)
1009{
1010	INIT_VNET_NET(curvnet);
1011	INIT_VNET_INET(curvnet);
1012	struct in_ifaddr *ia;
1013	struct in_addr prefix, mask, p;
1014	int error;
1015	struct sockaddr_in prefix0, mask0;
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	 * remove all L2 entries on the given prefix
1086	 */
1087	bzero(&prefix0, sizeof(prefix0));
1088	prefix0.sin_len = sizeof(prefix0);
1089	prefix0.sin_family = AF_INET;
1090	prefix0.sin_addr.s_addr = target->ia_subnet;
1091	bzero(&mask0, sizeof(mask0));
1092	mask0.sin_len = sizeof(mask0);
1093	mask0.sin_family = AF_INET;
1094	mask0.sin_addr.s_addr = target->ia_subnetmask;
1095	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1096			    (struct sockaddr *)&mask0);
1097
1098	/*
1099	 * As no-one seem to have this prefix, we can remove the route.
1100	 */
1101	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1102	target->ia_flags &= ~IFA_ROUTE;
1103	return (0);
1104}
1105
1106#undef rtinitflags
1107
1108/*
1109 * Return 1 if the address might be a local broadcast address.
1110 */
1111int
1112in_broadcast(struct in_addr in, struct ifnet *ifp)
1113{
1114	register struct ifaddr *ifa;
1115	u_long t;
1116
1117	if (in.s_addr == INADDR_BROADCAST ||
1118	    in.s_addr == INADDR_ANY)
1119		return (1);
1120	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1121		return (0);
1122	t = ntohl(in.s_addr);
1123	/*
1124	 * Look through the list of addresses for a match
1125	 * with a broadcast address.
1126	 */
1127#define ia ((struct in_ifaddr *)ifa)
1128	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1129		if (ifa->ifa_addr->sa_family == AF_INET &&
1130		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1131		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1132		     /*
1133		      * Check for old-style (host 0) broadcast.
1134		      */
1135		     t == ia->ia_subnet || t == ia->ia_net) &&
1136		     /*
1137		      * Check for an all one subnetmask. These
1138		      * only exist when an interface gets a secondary
1139		      * address.
1140		      */
1141		     ia->ia_subnetmask != (u_long)0xffffffff)
1142			    return (1);
1143	return (0);
1144#undef ia
1145}
1146
1147/*
1148 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1149 */
1150void
1151in_ifdetach(struct ifnet *ifp)
1152{
1153	INIT_VNET_INET(ifp->if_vnet);
1154
1155	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1156	in_pcbpurgeif0(&V_udbinfo, ifp);
1157	in_purgemaddrs(ifp);
1158}
1159
1160/*
1161 * Delete all IPv4 multicast address records, and associated link-layer
1162 * multicast address records, associated with ifp.
1163 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1164 * XXX This should not race with ifma_protospec being set during
1165 * a new allocation, if it does, we have bigger problems.
1166 */
1167static void
1168in_purgemaddrs(struct ifnet *ifp)
1169{
1170	LIST_HEAD(,in_multi) purgeinms;
1171	struct in_multi		*inm, *tinm;
1172	struct ifmultiaddr	*ifma;
1173
1174	LIST_INIT(&purgeinms);
1175	IN_MULTI_LOCK();
1176
1177	/*
1178	 * Extract list of in_multi associated with the detaching ifp
1179	 * which the PF_INET layer is about to release.
1180	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1181	 * by code further down.
1182	 */
1183	IF_ADDR_LOCK(ifp);
1184	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1185		if (ifma->ifma_addr->sa_family != AF_INET ||
1186		    ifma->ifma_protospec == NULL)
1187			continue;
1188#if 0
1189		KASSERT(ifma->ifma_protospec != NULL,
1190		    ("%s: ifma_protospec is NULL", __func__));
1191#endif
1192		inm = (struct in_multi *)ifma->ifma_protospec;
1193		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1194	}
1195	IF_ADDR_UNLOCK(ifp);
1196
1197	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1198		LIST_REMOVE(inm, inm_link);
1199		inm_release_locked(inm);
1200	}
1201	igmp_ifdetach(ifp);
1202
1203	IN_MULTI_UNLOCK();
1204}
1205
1206#include <net/if_dl.h>
1207#include <netinet/if_ether.h>
1208
1209struct in_llentry {
1210	struct llentry		base;
1211	struct sockaddr_in	l3_addr4;
1212};
1213
1214static struct llentry *
1215in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1216{
1217	struct in_llentry *lle;
1218
1219	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1220	if (lle == NULL)		/* NB: caller generates msg */
1221		return NULL;
1222
1223	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1224	/*
1225	 * For IPv4 this will trigger "arpresolve" to generate
1226	 * an ARP request.
1227	 */
1228	lle->base.la_expire = time_second; /* mark expired */
1229	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1230	lle->base.lle_refcnt = 1;
1231	LLE_LOCK_INIT(&lle->base);
1232	return &lle->base;
1233}
1234
1235/*
1236 * Deletes an address from the address table.
1237 * This function is called by the timer functions
1238 * such as arptimer() and nd6_llinfo_timer(), and
1239 * the caller does the locking.
1240 */
1241static void
1242in_lltable_free(struct lltable *llt, struct llentry *lle)
1243{
1244	LLE_WUNLOCK(lle);
1245	LLE_LOCK_DESTROY(lle);
1246	free(lle, M_LLTABLE);
1247}
1248
1249
1250#define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1251	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1252
1253static void
1254in_lltable_prefix_free(struct lltable *llt,
1255		       const struct sockaddr *prefix,
1256		       const struct sockaddr *mask)
1257{
1258	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1259	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1260	struct llentry *lle, *next;
1261	register int i;
1262
1263	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1264		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1265
1266			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1267						     pfx, msk)) {
1268				callout_drain(&lle->la_timer);
1269				LLE_WLOCK(lle);
1270				llentry_free(lle);
1271			}
1272		}
1273	}
1274}
1275
1276
1277static int
1278in_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
1279{
1280	struct rtentry *rt;
1281
1282	KASSERT(l3addr->sa_family == AF_INET,
1283	    ("sin_family %d", l3addr->sa_family));
1284
1285	/* XXX rtalloc1 should take a const param */
1286	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1287	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
1288		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1289		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1290		if (rt != NULL)
1291			RTFREE_LOCKED(rt);
1292		return (EINVAL);
1293	}
1294	RTFREE_LOCKED(rt);
1295	return 0;
1296}
1297
1298/*
1299 * Return NULL if not found or marked for deletion.
1300 * If found return lle read locked.
1301 */
1302static struct llentry *
1303in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1304{
1305	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1306	struct ifnet *ifp = llt->llt_ifp;
1307	struct llentry *lle;
1308	struct llentries *lleh;
1309	u_int hashkey;
1310
1311	IF_AFDATA_LOCK_ASSERT(ifp);
1312	KASSERT(l3addr->sa_family == AF_INET,
1313	    ("sin_family %d", l3addr->sa_family));
1314
1315	hashkey = sin->sin_addr.s_addr;
1316	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1317	LIST_FOREACH(lle, lleh, lle_next) {
1318		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1319		if (lle->la_flags & LLE_DELETED)
1320			continue;
1321		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1322			break;
1323	}
1324	if (lle == NULL) {
1325#ifdef DIAGNOSTICS
1326		if (flags & LLE_DELETE)
1327			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1328#endif
1329		if (!(flags & LLE_CREATE))
1330			return (NULL);
1331		/*
1332		 * A route that covers the given address must have
1333		 * been installed 1st because we are doing a resolution,
1334		 * verify this.
1335		 */
1336		if (!(flags & LLE_IFADDR) &&
1337		    in_lltable_rtcheck(ifp, l3addr) != 0)
1338			goto done;
1339
1340		lle = in_lltable_new(l3addr, flags);
1341		if (lle == NULL) {
1342			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1343			goto done;
1344		}
1345		lle->la_flags = flags & ~LLE_CREATE;
1346		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1347			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1348			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1349		}
1350
1351		lle->lle_tbl  = llt;
1352		lle->lle_head = lleh;
1353		LIST_INSERT_HEAD(lleh, lle, lle_next);
1354	} else if (flags & LLE_DELETE) {
1355		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1356			LLE_WLOCK(lle);
1357			lle->la_flags = LLE_DELETED;
1358			LLE_WUNLOCK(lle);
1359#ifdef DIAGNOSTICS
1360			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1361#endif
1362		}
1363		lle = (void *)-1;
1364
1365	}
1366	if (LLE_IS_VALID(lle)) {
1367		if (flags & LLE_EXCLUSIVE)
1368			LLE_WLOCK(lle);
1369		else
1370			LLE_RLOCK(lle);
1371	}
1372done:
1373	return (lle);
1374}
1375
1376static int
1377in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1378{
1379#define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1380	struct ifnet *ifp = llt->llt_ifp;
1381	struct llentry *lle;
1382	/* XXX stack use */
1383	struct {
1384		struct rt_msghdr	rtm;
1385		struct sockaddr_inarp	sin;
1386		struct sockaddr_dl	sdl;
1387	} arpc;
1388	int error, i;
1389
1390	/* XXXXX
1391	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
1392	 * so it is okay to use this ASSERT, change it when
1393	 * IFNET lock is finalized
1394	 */
1395	IFNET_WLOCK_ASSERT();
1396
1397	error = 0;
1398	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1399		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1400			struct sockaddr_dl *sdl;
1401
1402			/* skip deleted entries */
1403			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
1404				continue;
1405			/* Skip if jailed and not a valid IP of the prison. */
1406			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1407				continue;
1408			/*
1409			 * produce a msg made of:
1410			 *  struct rt_msghdr;
1411			 *  struct sockaddr_inarp; (IPv4)
1412			 *  struct sockaddr_dl;
1413			 */
1414			bzero(&arpc, sizeof(arpc));
1415			arpc.rtm.rtm_msglen = sizeof(arpc);
1416			arpc.rtm.rtm_version = RTM_VERSION;
1417			arpc.rtm.rtm_type = RTM_GET;
1418			arpc.rtm.rtm_flags = RTF_UP;
1419			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1420			arpc.sin.sin_family = AF_INET;
1421			arpc.sin.sin_len = sizeof(arpc.sin);
1422			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1423
1424			/* publish */
1425			if (lle->la_flags & LLE_PUB) {
1426				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1427				/* proxy only */
1428				if (lle->la_flags & LLE_PROXY)
1429					arpc.sin.sin_other = SIN_PROXY;
1430			}
1431
1432			sdl = &arpc.sdl;
1433			sdl->sdl_family = AF_LINK;
1434			sdl->sdl_len = sizeof(*sdl);
1435			sdl->sdl_alen = ifp->if_addrlen;
1436			sdl->sdl_index = ifp->if_index;
1437			sdl->sdl_type = ifp->if_type;
1438			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1439
1440			arpc.rtm.rtm_rmx.rmx_expire =
1441			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1442			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1443			if (lle->la_flags & LLE_STATIC)
1444				arpc.rtm.rtm_flags |= RTF_STATIC;
1445			arpc.rtm.rtm_index = ifp->if_index;
1446			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1447			if (error)
1448				break;
1449		}
1450	}
1451	return error;
1452#undef SIN
1453}
1454
1455void *
1456in_domifattach(struct ifnet *ifp)
1457{
1458	struct in_ifinfo *ii;
1459	struct lltable *llt;
1460
1461	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1462
1463	llt = lltable_init(ifp, AF_INET);
1464	if (llt != NULL) {
1465		llt->llt_new = in_lltable_new;
1466		llt->llt_free = in_lltable_free;
1467		llt->llt_prefix_free = in_lltable_prefix_free;
1468		llt->llt_rtcheck = in_lltable_rtcheck;
1469		llt->llt_lookup = in_lltable_lookup;
1470		llt->llt_dump = in_lltable_dump;
1471	}
1472	ii->ii_llt = llt;
1473
1474	ii->ii_igmp = igmp_domifattach(ifp);
1475
1476	return ii;
1477}
1478
1479void
1480in_domifdetach(struct ifnet *ifp, void *aux)
1481{
1482	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1483
1484	igmp_domifdetach(ifp);
1485	lltable_free(ii->ii_llt);
1486	free(ii, M_IFADDR);
1487}
1488