xref: /dragonfly/sys/netinet/in.c (revision cc93b0eb)
1 /*
2  * Copyright (c) 1982, 1986, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)in.c	8.4 (Berkeley) 1/9/95
34  * $FreeBSD: src/sys/netinet/in.c,v 1.44.2.14 2002/11/08 00:45:50 suz Exp $
35  * $DragonFly: src/sys/netinet/in.c,v 1.40 2008/06/14 08:40:16 sephe Exp $
36  */
37 
38 #include "opt_bootp.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/sockio.h>
43 #include <sys/malloc.h>
44 #include <sys/proc.h>
45 #include <sys/msgport.h>
46 #include <sys/socket.h>
47 
48 #include <sys/kernel.h>
49 #include <sys/sysctl.h>
50 #include <sys/thread2.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/route.h>
55 #include <net/netmsg2.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/in_var.h>
59 #include <netinet/in_pcb.h>
60 
61 #include <netinet/igmp_var.h>
62 
63 MALLOC_DEFINE(M_IPMADDR, "in_multi", "internet multicast address");
64 
65 static int in_mask2len (struct in_addr *);
66 static void in_len2mask (struct in_addr *, int);
67 static int in_lifaddr_ioctl (struct socket *, u_long, caddr_t,
68 	struct ifnet *, struct thread *);
69 
70 static void	in_socktrim (struct sockaddr_in *);
71 static int	in_ifinit(struct ifnet *, struct in_ifaddr *,
72 		    const struct sockaddr_in *, int);
73 
74 static void	in_control_dispatch(struct netmsg *);
75 static int	in_control_internal(u_long, caddr_t, struct ifnet *,
76 		    struct thread *);
77 
78 static int subnetsarelocal = 0;
79 SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
80 	&subnetsarelocal, 0, "");
81 
82 struct in_multihead in_multihead; /* XXX BSS initialization */
83 
84 extern struct inpcbinfo ripcbinfo;
85 extern struct inpcbinfo udbinfo;
86 
87 /*
88  * Return 1 if an internet address is for a ``local'' host
89  * (one to which we have a connection).  If subnetsarelocal
90  * is true, this includes other subnets of the local net.
91  * Otherwise, it includes only the directly-connected (sub)nets.
92  */
93 int
94 in_localaddr(struct in_addr in)
95 {
96 	u_long i = ntohl(in.s_addr);
97 	struct in_ifaddr_container *iac;
98 	struct in_ifaddr *ia;
99 
100 	if (subnetsarelocal) {
101 		TAILQ_FOREACH(iac, &in_ifaddrheads[mycpuid], ia_link) {
102 			ia = iac->ia;
103 
104 			if ((i & ia->ia_netmask) == ia->ia_net)
105 				return (1);
106 		}
107 	} else {
108 		TAILQ_FOREACH(iac, &in_ifaddrheads[mycpuid], ia_link) {
109 			ia = iac->ia;
110 
111 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
112 				return (1);
113 		}
114 	}
115 	return (0);
116 }
117 
118 /*
119  * Determine whether an IP address is in a reserved set of addresses
120  * that may not be forwarded, or whether datagrams to that destination
121  * may be forwarded.
122  */
123 int
124 in_canforward(struct in_addr in)
125 {
126 	u_long i = ntohl(in.s_addr);
127 	u_long net;
128 
129 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i))
130 		return (0);
131 	if (IN_CLASSA(i)) {
132 		net = i & IN_CLASSA_NET;
133 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
134 			return (0);
135 	}
136 	return (1);
137 }
138 
139 /*
140  * Trim a mask in a sockaddr
141  */
142 static void
143 in_socktrim(struct sockaddr_in *ap)
144 {
145     char *cplim = (char *) &ap->sin_addr;
146     char *cp = (char *) (&ap->sin_addr + 1);
147 
148     ap->sin_len = 0;
149     while (--cp >= cplim)
150 	if (*cp) {
151 	    (ap)->sin_len = cp - (char *) (ap) + 1;
152 	    break;
153 	}
154 }
155 
156 static int
157 in_mask2len(struct in_addr *mask)
158 {
159 	int x, y;
160 	u_char *p;
161 
162 	p = (u_char *)mask;
163 	for (x = 0; x < sizeof *mask; x++) {
164 		if (p[x] != 0xff)
165 			break;
166 	}
167 	y = 0;
168 	if (x < sizeof *mask) {
169 		for (y = 0; y < 8; y++) {
170 			if ((p[x] & (0x80 >> y)) == 0)
171 				break;
172 		}
173 	}
174 	return x * 8 + y;
175 }
176 
177 static void
178 in_len2mask(struct in_addr *mask, int len)
179 {
180 	int i;
181 	u_char *p;
182 
183 	p = (u_char *)mask;
184 	bzero(mask, sizeof *mask);
185 	for (i = 0; i < len / 8; i++)
186 		p[i] = 0xff;
187 	if (len % 8)
188 		p[i] = (0xff00 >> (len % 8)) & 0xff;
189 }
190 
191 static int in_interfaces;	/* number of external internet interfaces */
192 
193 struct in_control_arg {
194 	u_long		cmd;
195 	caddr_t		data;
196 	struct ifnet	*ifp;
197 	struct thread	*td;
198 };
199 
200 static void
201 in_control_dispatch(struct netmsg *nmsg)
202 {
203 	struct lwkt_msg *msg = &nmsg->nm_lmsg;
204 	const struct in_control_arg *arg = msg->u.ms_resultp;
205 	int error;
206 
207 	error = in_control_internal(arg->cmd, arg->data, arg->ifp, arg->td);
208 	lwkt_replymsg(msg, error);
209 }
210 
211 /*
212  * Generic internet control operations (ioctl's).
213  * Ifp is 0 if not an interface-specific ioctl.
214  *
215  * NOTE! td might be NULL.
216  */
217 /* ARGSUSED */
218 int
219 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
220 	   struct thread *td)
221 {
222 	struct netmsg nmsg;
223 	struct in_control_arg arg;
224 	struct lwkt_msg *msg;
225 	int error;
226 
227 	switch (cmd) {
228 	case SIOCALIFADDR:
229 	case SIOCDLIFADDR:
230 		if (td && (error = suser(td)) != 0)
231 			return error;
232 		/* FALLTHROUGH */
233 	case SIOCGLIFADDR:
234 		if (!ifp)
235 			return EINVAL;
236 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
237 	}
238 
239 	KASSERT(cmd != SIOCALIFADDR && cmd != SIOCDLIFADDR,
240 		("recursive SIOC%cLIFADDR!\n",
241 		 cmd == SIOCDLIFADDR ? 'D' : 'A'));
242 
243 	/*
244 	 * IFADDR alterations are serialized by netisr0
245 	 */
246 	switch (cmd) {
247 	case SIOCSIFDSTADDR:
248 	case SIOCSIFBRDADDR:
249 	case SIOCSIFADDR:
250 	case SIOCSIFNETMASK:
251 	case SIOCAIFADDR:
252 	case SIOCDIFADDR:
253 		bzero(&arg, sizeof(arg));
254 		arg.cmd = cmd;
255 		arg.data = data;
256 		arg.ifp = ifp;
257 		arg.td = td;
258 
259 		netmsg_init(&nmsg, &curthread->td_msgport, 0,
260 			    in_control_dispatch);
261 		msg = &nmsg.nm_lmsg;
262 		msg->u.ms_resultp = &arg;
263 
264 		lwkt_domsg(cpu_portfn(0), msg, 0);
265 		return msg->ms_error;
266 	default:
267 		return in_control_internal(cmd, data, ifp, td);
268 	}
269 }
270 
271 static void
272 in_ialink_dispatch(struct netmsg *nmsg)
273 {
274 	struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
275 	struct in_ifaddr *ia = lmsg->u.ms_resultp;
276 	struct ifaddr_container *ifac;
277 	struct in_ifaddr_container *iac;
278 	int cpu = mycpuid;
279 
280 	crit_enter();
281 
282 	ifac = &ia->ia_ifa.ifa_containers[cpu];
283 	ASSERT_IFAC_VALID(ifac);
284 	KASSERT((ifac->ifa_listmask & IFA_LIST_IN_IFADDRHEAD) == 0,
285 		("ia is on in_ifaddrheads\n"));
286 
287 	ifac->ifa_listmask |= IFA_LIST_IN_IFADDRHEAD;
288 	iac = &ifac->ifa_proto_u.u_in_ifac;
289 	TAILQ_INSERT_TAIL(&in_ifaddrheads[cpu], iac, ia_link);
290 
291 	crit_exit();
292 
293 	ifa_forwardmsg(lmsg, cpu + 1);
294 }
295 
296 static void
297 in_iaunlink_dispatch(struct netmsg *nmsg)
298 {
299 	struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
300 	struct in_ifaddr *ia = lmsg->u.ms_resultp;
301 	struct ifaddr_container *ifac;
302 	struct in_ifaddr_container *iac;
303 	int cpu = mycpuid;
304 
305 	crit_enter();
306 
307 	ifac = &ia->ia_ifa.ifa_containers[cpu];
308 	ASSERT_IFAC_VALID(ifac);
309 	KASSERT(ifac->ifa_listmask & IFA_LIST_IN_IFADDRHEAD,
310 		("ia is not on in_ifaddrheads\n"));
311 
312 	iac = &ifac->ifa_proto_u.u_in_ifac;
313 	TAILQ_REMOVE(&in_ifaddrheads[cpu], iac, ia_link);
314 	ifac->ifa_listmask &= ~IFA_LIST_IN_IFADDRHEAD;
315 
316 	crit_exit();
317 
318 	ifa_forwardmsg(lmsg, cpu + 1);
319 }
320 
321 static void
322 in_iahashins_dispatch(struct netmsg *nmsg)
323 {
324 	struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
325 	struct in_ifaddr *ia = lmsg->u.ms_resultp;
326 	struct ifaddr_container *ifac;
327 	struct in_ifaddr_container *iac;
328 	int cpu = mycpuid;
329 
330 	crit_enter();
331 
332 	ifac = &ia->ia_ifa.ifa_containers[cpu];
333 	ASSERT_IFAC_VALID(ifac);
334 	KASSERT((ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH) == 0,
335 		("ia is on in_ifaddrhashtbls\n"));
336 
337 	ifac->ifa_listmask |= IFA_LIST_IN_IFADDRHASH;
338 	iac = &ifac->ifa_proto_u.u_in_ifac;
339 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
340 			 iac, ia_hash);
341 
342 	crit_exit();
343 
344 	ifa_forwardmsg(lmsg, cpu + 1);
345 }
346 
347 static void
348 in_iahashrem_dispatch(struct netmsg *nmsg)
349 {
350 	struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
351 	struct in_ifaddr *ia = lmsg->u.ms_resultp;
352 	struct ifaddr_container *ifac;
353 	struct in_ifaddr_container *iac;
354 	int cpu = mycpuid;
355 
356 	crit_enter();
357 
358 	ifac = &ia->ia_ifa.ifa_containers[cpu];
359 	ASSERT_IFAC_VALID(ifac);
360 	KASSERT(ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH,
361 		("ia is not on in_ifaddrhashtbls\n"));
362 
363 	iac = &ifac->ifa_proto_u.u_in_ifac;
364 	LIST_REMOVE(iac, ia_hash);
365 	ifac->ifa_listmask &= ~IFA_LIST_IN_IFADDRHASH;
366 
367 	crit_exit();
368 
369 	ifa_forwardmsg(lmsg, cpu + 1);
370 }
371 
372 static void
373 in_ialink(struct in_ifaddr *ia)
374 {
375 	struct netmsg nmsg;
376 	struct lwkt_msg *lmsg;
377 
378 	netmsg_init(&nmsg, &curthread->td_msgport, 0, in_ialink_dispatch);
379 	lmsg = &nmsg.nm_lmsg;
380 	lmsg->u.ms_resultp = ia;
381 
382 	ifa_domsg(lmsg);
383 }
384 
385 void
386 in_iaunlink(struct in_ifaddr *ia)
387 {
388 	struct netmsg nmsg;
389 	struct lwkt_msg *lmsg;
390 
391 	netmsg_init(&nmsg, &curthread->td_msgport, 0, in_iaunlink_dispatch);
392 	lmsg = &nmsg.nm_lmsg;
393 	lmsg->u.ms_resultp = ia;
394 
395 	ifa_domsg(lmsg);
396 }
397 
398 void
399 in_iahash_insert(struct in_ifaddr *ia)
400 {
401 	struct netmsg nmsg;
402 	struct lwkt_msg *lmsg;
403 
404 	netmsg_init(&nmsg, &curthread->td_msgport, 0, in_iahashins_dispatch);
405 	lmsg = &nmsg.nm_lmsg;
406 	lmsg->u.ms_resultp = ia;
407 
408 	ifa_domsg(lmsg);
409 }
410 
411 void
412 in_iahash_remove(struct in_ifaddr *ia)
413 {
414 	struct netmsg nmsg;
415 	struct lwkt_msg *lmsg;
416 
417 	netmsg_init(&nmsg, &curthread->td_msgport, 0, in_iahashrem_dispatch);
418 	lmsg = &nmsg.nm_lmsg;
419 	lmsg->u.ms_resultp = ia;
420 
421 	ifa_domsg(lmsg);
422 }
423 
424 static __inline struct in_ifaddr *
425 in_ianext(struct in_ifaddr *oia)
426 {
427 	struct ifaddr_container *ifac;
428 	struct in_ifaddr_container *iac;
429 
430 	ifac = &oia->ia_ifa.ifa_containers[mycpuid];
431 	ASSERT_IFAC_VALID(ifac);
432 	KASSERT(ifac->ifa_listmask & IFA_LIST_IN_IFADDRHEAD,
433 		("ia is not on in_ifaddrheads\n"));
434 
435 	iac = &ifac->ifa_proto_u.u_in_ifac;
436 	iac = TAILQ_NEXT(iac, ia_link);
437 	if (iac != NULL)
438 		return iac->ia;
439 	else
440 		return NULL;
441 }
442 
443 static int
444 in_control_internal(u_long cmd, caddr_t data, struct ifnet *ifp,
445 		    struct thread *td)
446 {
447 	struct ifreq *ifr = (struct ifreq *)data;
448 	struct in_ifaddr *ia = NULL;
449 	struct in_addr dst;
450 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
451 	struct ifaddr_container *ifac;
452 	struct in_ifaddr_container *iac;
453 	struct sockaddr_in oldaddr;
454 	int hostIsNew, iaIsNew, maskIsNew;
455 	int error = 0;
456 
457 	iaIsNew = 0;
458 
459 	/*
460 	 * Find address for this interface, if it exists.
461 	 *
462 	 * If an alias address was specified, find that one instead of
463 	 * the first one on the interface, if possible
464 	 */
465 	if (ifp) {
466 		struct in_ifaddr *iap;
467 
468 		dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
469 		LIST_FOREACH(iac, INADDR_HASH(dst.s_addr), ia_hash) {
470 			iap = iac->ia;
471 			if (iap->ia_ifp == ifp &&
472 			    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
473 				ia = iap;
474 				break;
475 			}
476 		}
477 		if (ia == NULL) {
478 			TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid],
479 				      ifa_link) {
480 				iap = ifatoia(ifac->ifa);
481 				if (iap->ia_addr.sin_family == AF_INET) {
482 					ia = iap;
483 					break;
484 				}
485 			}
486 		}
487 	}
488 
489 	switch (cmd) {
490 	case SIOCAIFADDR:
491 	case SIOCDIFADDR:
492 		if (ifp == NULL)
493 			return (EADDRNOTAVAIL);
494 		if (ifra->ifra_addr.sin_family == AF_INET) {
495 			while (ia != NULL) {
496 				if (ia->ia_ifp == ifp  &&
497 				    ia->ia_addr.sin_addr.s_addr ==
498 				    ifra->ifra_addr.sin_addr.s_addr)
499 					break;
500 				ia = in_ianext(ia);
501 			}
502 			if ((ifp->if_flags & IFF_POINTOPOINT) &&
503 			    cmd == SIOCAIFADDR &&
504 			    ifra->ifra_dstaddr.sin_addr.s_addr == INADDR_ANY) {
505 				return EDESTADDRREQ;
506 			}
507 		}
508 		if (cmd == SIOCDIFADDR && ia == NULL)
509 			return (EADDRNOTAVAIL);
510 		/* FALLTHROUGH */
511 	case SIOCSIFADDR:
512 	case SIOCSIFNETMASK:
513 	case SIOCSIFDSTADDR:
514 		if (td && (error = suser(td)) != 0)
515 			return error;
516 
517 		if (ifp == NULL)
518 			return (EADDRNOTAVAIL);
519 
520 		if (cmd == SIOCSIFDSTADDR &&
521 		    (ifp->if_flags & IFF_POINTOPOINT) == 0)
522 			return (EINVAL);
523 
524 		if (ia == NULL) {
525 			struct ifaddr *ifa;
526 			int i;
527 
528 			ia = ifa_create(sizeof(*ia), M_WAITOK);
529 			ifa = &ia->ia_ifa;
530 
531 			/*
532 			 * Setup per-CPU information
533 			 */
534 			for (i = 0; i < ncpus; ++i) {
535 				ifac = &ifa->ifa_containers[i];
536 				iac = &ifac->ifa_proto_u.u_in_ifac;
537 				iac->ia = ia;
538 				iac->ia_ifac = ifac;
539 			}
540 
541 			/*
542 			 * Protect from NETISR_IP traversing address list
543 			 * while we're modifying it.
544 			 */
545 			crit_enter();
546 
547 			in_ialink(ia);
548 			ifa_iflink(ifa, ifp, 1);
549 
550 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
551 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
552 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
553 			ia->ia_sockmask.sin_len = 8;
554 			ia->ia_sockmask.sin_family = AF_INET;
555 			if (ifp->if_flags & IFF_BROADCAST) {
556 				ia->ia_broadaddr.sin_len = sizeof ia->ia_addr;
557 				ia->ia_broadaddr.sin_family = AF_INET;
558 			}
559 			ia->ia_ifp = ifp;
560 			if (!(ifp->if_flags & IFF_LOOPBACK))
561 				in_interfaces++;
562 			iaIsNew = 1;
563 
564 			crit_exit();
565 		}
566 		break;
567 
568 	case SIOCSIFBRDADDR:
569 		if (td && (error = suser(td)) != 0)
570 			return error;
571 		/* FALLTHROUGH */
572 
573 	case SIOCGIFADDR:
574 	case SIOCGIFNETMASK:
575 	case SIOCGIFDSTADDR:
576 	case SIOCGIFBRDADDR:
577 		if (ia == NULL)
578 			return (EADDRNOTAVAIL);
579 		break;
580 	}
581 
582 	switch (cmd) {
583 	case SIOCGIFADDR:
584 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
585 		return (0);
586 
587 	case SIOCGIFBRDADDR:
588 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
589 			return (EINVAL);
590 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
591 		return (0);
592 
593 	case SIOCGIFDSTADDR:
594 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
595 			return (EINVAL);
596 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
597 		return (0);
598 
599 	case SIOCGIFNETMASK:
600 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
601 		return (0);
602 
603 	case SIOCSIFDSTADDR:
604 		KKASSERT(ifp->if_flags & IFF_POINTOPOINT);
605 
606 		oldaddr = ia->ia_dstaddr;
607 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
608 		if (ifp->if_ioctl != NULL) {
609 			lwkt_serialize_enter(ifp->if_serializer);
610 			error = ifp->if_ioctl(ifp, SIOCSIFDSTADDR, (caddr_t)ia,
611 					      td->td_proc->p_ucred);
612 			lwkt_serialize_exit(ifp->if_serializer);
613 			if (error) {
614 				ia->ia_dstaddr = oldaddr;
615 				return (error);
616 			}
617 		}
618 		if (ia->ia_flags & IFA_ROUTE) {
619 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
620 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
621 			ia->ia_ifa.ifa_dstaddr =
622 					(struct sockaddr *)&ia->ia_dstaddr;
623 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST | RTF_UP);
624 		}
625 		return (0);
626 
627 	case SIOCSIFBRDADDR:
628 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
629 			return (EINVAL);
630 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
631 		return (0);
632 
633 	case SIOCSIFADDR:
634 		error = in_ifinit(ifp, ia,
635 		    (const struct sockaddr_in *)&ifr->ifr_addr, 1);
636 		if (error != 0 && iaIsNew)
637 			break;
638 		if (error == 0)
639 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
640 		return (0);
641 
642 	case SIOCSIFNETMASK:
643 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
644 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
645 		return (0);
646 
647 	case SIOCAIFADDR:
648 		maskIsNew = 0;
649 		hostIsNew = 1;
650 		error = 0;
651 		if (ia->ia_addr.sin_family == AF_INET) {
652 			if (ifra->ifra_addr.sin_len == 0) {
653 				ifra->ifra_addr = ia->ia_addr;
654 				hostIsNew = 0;
655 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
656 				   ia->ia_addr.sin_addr.s_addr) {
657 				hostIsNew = 0;
658 			}
659 		}
660 		if (ifra->ifra_mask.sin_len) {
661 			in_ifscrub(ifp, ia);
662 			ia->ia_sockmask = ifra->ifra_mask;
663 			ia->ia_sockmask.sin_family = AF_INET;
664 			ia->ia_subnetmask =
665 			    ntohl(ia->ia_sockmask.sin_addr.s_addr);
666 			maskIsNew = 1;
667 		}
668 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
669 		    ifra->ifra_dstaddr.sin_family == AF_INET) {
670 			in_ifscrub(ifp, ia);
671 			ia->ia_dstaddr = ifra->ifra_dstaddr;
672 			maskIsNew  = 1; /* We lie; but the effect's the same */
673 		}
674 		if (ifra->ifra_addr.sin_family == AF_INET &&
675 		    (hostIsNew || maskIsNew))
676 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
677 
678 		if (error != 0 && iaIsNew)
679 			break;
680 
681 		if ((ifp->if_flags & IFF_BROADCAST) &&
682 		    ifra->ifra_broadaddr.sin_family == AF_INET)
683 			ia->ia_broadaddr = ifra->ifra_broadaddr;
684 		if (error == 0)
685 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
686 		return (error);
687 
688 	case SIOCDIFADDR:
689 		/*
690 		 * in_ifscrub kills the interface route.
691 		 */
692 		in_ifscrub(ifp, ia);
693 		/*
694 		 * in_ifadown gets rid of all the rest of
695 		 * the routes.  This is not quite the right
696 		 * thing to do, but at least if we are running
697 		 * a routing process they will come back.
698 		 */
699 		in_ifadown(&ia->ia_ifa, 1);
700 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
701 		error = 0;
702 		break;
703 
704 	default:
705 		if (ifp == NULL || ifp->if_ioctl == NULL)
706 			return (EOPNOTSUPP);
707 		lwkt_serialize_enter(ifp->if_serializer);
708 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
709 		lwkt_serialize_exit(ifp->if_serializer);
710 		return (error);
711 	}
712 
713 	KKASSERT(cmd == SIOCDIFADDR ||
714 		 ((cmd == SIOCAIFADDR || cmd == SIOCSIFADDR) && iaIsNew));
715 
716 	ifa_ifunlink(&ia->ia_ifa, ifp);
717 	in_iaunlink(ia);
718 
719 	if (cmd == SIOCDIFADDR) {
720 		ifac = &ia->ia_ifa.ifa_containers[mycpuid];
721 		if (ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH)
722 			in_iahash_remove(ia);
723 	}
724 #ifdef INVARIANTS
725 	else {
726 		/*
727 		 * If cmd is SIOCSIFADDR or SIOCAIFADDR, in_ifinit() has
728 		 * already taken care of the deletion from hash table
729 		 */
730 		ifac = &ia->ia_ifa.ifa_containers[mycpuid];
731 		KASSERT((ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH) == 0,
732 			("SIOC%cIFADDR failed on new ia, "
733 			 "but the new ia is still in hash table\n",
734 			 cmd == SIOCSIFADDR ? 'S' : 'A'));
735 	}
736 #endif
737 
738 	ifa_destroy(&ia->ia_ifa);
739 
740 	return (error);
741 }
742 
743 /*
744  * SIOC[GAD]LIFADDR.
745  *	SIOCGLIFADDR: get first address. (?!?)
746  *	SIOCGLIFADDR with IFLR_PREFIX:
747  *		get first address that matches the specified prefix.
748  *	SIOCALIFADDR: add the specified address.
749  *	SIOCALIFADDR with IFLR_PREFIX:
750  *		EINVAL since we can't deduce hostid part of the address.
751  *	SIOCDLIFADDR: delete the specified address.
752  *	SIOCDLIFADDR with IFLR_PREFIX:
753  *		delete the first address that matches the specified prefix.
754  * return values:
755  *	EINVAL on invalid parameters
756  *	EADDRNOTAVAIL on prefix match failed/specified address not found
757  *	other values may be returned from in_ioctl()
758  *
759  * NOTE! td might be NULL.
760  */
761 static int
762 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
763 		 struct thread *td)
764 {
765 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
766 
767 	/* sanity checks */
768 	if (!data || !ifp) {
769 		panic("invalid argument to in_lifaddr_ioctl");
770 		/*NOTRECHED*/
771 	}
772 
773 	switch (cmd) {
774 	case SIOCGLIFADDR:
775 		/* address must be specified on GET with IFLR_PREFIX */
776 		if ((iflr->flags & IFLR_PREFIX) == 0)
777 			break;
778 		/*FALLTHROUGH*/
779 	case SIOCALIFADDR:
780 	case SIOCDLIFADDR:
781 		/* address must be specified on ADD and DELETE */
782 		if (iflr->addr.ss_family != AF_INET)
783 			return EINVAL;
784 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
785 			return EINVAL;
786 		/* XXX need improvement */
787 		if (iflr->dstaddr.ss_family
788 		 && iflr->dstaddr.ss_family != AF_INET)
789 			return EINVAL;
790 		if (iflr->dstaddr.ss_family
791 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
792 			return EINVAL;
793 		break;
794 	default: /*shouldn't happen*/
795 		return EOPNOTSUPP;
796 	}
797 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
798 		return EINVAL;
799 
800 	switch (cmd) {
801 	case SIOCALIFADDR:
802 	    {
803 		struct in_aliasreq ifra;
804 
805 		if (iflr->flags & IFLR_PREFIX)
806 			return EINVAL;
807 
808 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
809 		bzero(&ifra, sizeof ifra);
810 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof ifra.ifra_name);
811 
812 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
813 
814 		if (iflr->dstaddr.ss_family) {	/*XXX*/
815 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
816 				iflr->dstaddr.ss_len);
817 		}
818 
819 		ifra.ifra_mask.sin_family = AF_INET;
820 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
821 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
822 
823 		return in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td);
824 	    }
825 	case SIOCGLIFADDR:
826 	case SIOCDLIFADDR:
827 	    {
828 		struct ifaddr_container *ifac;
829 		struct in_ifaddr *ia;
830 		struct in_addr mask, candidate, match;
831 		struct sockaddr_in *sin;
832 		int cmp;
833 
834 		bzero(&mask, sizeof mask);
835 		if (iflr->flags & IFLR_PREFIX) {
836 			/* lookup a prefix rather than address. */
837 			in_len2mask(&mask, iflr->prefixlen);
838 
839 			sin = (struct sockaddr_in *)&iflr->addr;
840 			match.s_addr = sin->sin_addr.s_addr;
841 			match.s_addr &= mask.s_addr;
842 
843 			/* if you set extra bits, that's wrong */
844 			if (match.s_addr != sin->sin_addr.s_addr)
845 				return EINVAL;
846 
847 			cmp = 1;
848 		} else {
849 			if (cmd == SIOCGLIFADDR) {
850 				/* on getting an address, take the 1st match */
851 				match.s_addr = 0; /* gcc4 warning */
852 				cmp = 0;	/*XXX*/
853 			} else {
854 				/* on deleting an address, do exact match */
855 				in_len2mask(&mask, 32);
856 				sin = (struct sockaddr_in *)&iflr->addr;
857 				match.s_addr = sin->sin_addr.s_addr;
858 
859 				cmp = 1;
860 			}
861 		}
862 
863 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
864 			struct ifaddr *ifa = ifac->ifa;
865 
866 			if (ifa->ifa_addr->sa_family != AF_INET6)
867 				continue;
868 			if (!cmp)
869 				break;
870 			candidate.s_addr =
871 			((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
872 			candidate.s_addr &= mask.s_addr;
873 			if (candidate.s_addr == match.s_addr)
874 				break;
875 		}
876 		if (ifac == NULL)
877 			return EADDRNOTAVAIL;
878 		ia = (struct in_ifaddr *)(ifac->ifa);
879 
880 		if (cmd == SIOCGLIFADDR) {
881 			/* fill in the if_laddrreq structure */
882 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
883 
884 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
885 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
886 					ia->ia_dstaddr.sin_len);
887 			} else
888 				bzero(&iflr->dstaddr, sizeof iflr->dstaddr);
889 
890 			iflr->prefixlen =
891 				in_mask2len(&ia->ia_sockmask.sin_addr);
892 
893 			iflr->flags = 0;	/*XXX*/
894 
895 			return 0;
896 		} else {
897 			struct in_aliasreq ifra;
898 
899 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
900 			bzero(&ifra, sizeof ifra);
901 			bcopy(iflr->iflr_name, ifra.ifra_name,
902 				sizeof ifra.ifra_name);
903 
904 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
905 				ia->ia_addr.sin_len);
906 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
907 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
908 					ia->ia_dstaddr.sin_len);
909 			}
910 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
911 				ia->ia_sockmask.sin_len);
912 
913 			return in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
914 					  ifp, td);
915 		}
916 	    }
917 	}
918 
919 	return EOPNOTSUPP;	/*just for safety*/
920 }
921 
922 /*
923  * Delete any existing route for an interface.
924  */
925 void
926 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
927 {
928 
929 	if ((ia->ia_flags & IFA_ROUTE) == 0)
930 		return;
931 	if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
932 		rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
933 	else
934 		rtinit(&ia->ia_ifa, RTM_DELETE, 0);
935 	ia->ia_flags &= ~IFA_ROUTE;
936 }
937 
938 /*
939  * Initialize an interface's internet address
940  * and routing table entry.
941  */
942 static int
943 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
944 	  const struct sockaddr_in *sin, int scrub)
945 {
946 	u_long i = ntohl(sin->sin_addr.s_addr);
947 	struct sockaddr_in oldaddr;
948 	struct ifaddr_container *ifac;
949 	int flags = RTF_UP, error = 0;
950 	int was_hash = 0;
951 
952 	ifac = &ia->ia_ifa.ifa_containers[mycpuid];
953 	oldaddr = ia->ia_addr;
954 
955 	if (ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH) {
956 		was_hash = 1;
957 		in_iahash_remove(ia);
958 	}
959 
960 	ia->ia_addr = *sin;
961 	if (ia->ia_addr.sin_family == AF_INET)
962 		in_iahash_insert(ia);
963 
964 	/*
965 	 * Give the interface a chance to initialize
966 	 * if this is its first address,
967 	 * and to validate the address if necessary.
968 	 */
969 	if (ifp->if_ioctl != NULL) {
970 		lwkt_serialize_enter(ifp->if_serializer);
971 		error = ifp->if_ioctl(ifp, SIOCSIFADDR, (caddr_t)ia, NULL);
972 		lwkt_serialize_exit(ifp->if_serializer);
973 		if (error)
974 			goto fail;
975 	}
976 
977 	/*
978 	 * Delete old route, if requested.
979 	 */
980 	if (scrub) {
981 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
982 		in_ifscrub(ifp, ia);
983 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
984 	}
985 
986 	/*
987 	 * Calculate netmask/subnetmask.
988 	 */
989 	if (IN_CLASSA(i))
990 		ia->ia_netmask = IN_CLASSA_NET;
991 	else if (IN_CLASSB(i))
992 		ia->ia_netmask = IN_CLASSB_NET;
993 	else
994 		ia->ia_netmask = IN_CLASSC_NET;
995 	/*
996 	 * The subnet mask usually includes at least the standard network part,
997 	 * but may may be smaller in the case of supernetting.
998 	 * If it is set, we believe it.
999 	 */
1000 	if (ia->ia_subnetmask == 0) {
1001 		ia->ia_subnetmask = ia->ia_netmask;
1002 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
1003 	} else {
1004 		ia->ia_netmask &= ia->ia_subnetmask;
1005 	}
1006 	ia->ia_net = i & ia->ia_netmask;
1007 	ia->ia_subnet = i & ia->ia_subnetmask;
1008 	in_socktrim(&ia->ia_sockmask);
1009 
1010 	/*
1011 	 * Add route for the network.
1012 	 */
1013 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1014 	if (ifp->if_flags & IFF_BROADCAST) {
1015 		ia->ia_broadaddr.sin_addr.s_addr =
1016 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
1017 		ia->ia_netbroadcast.s_addr =
1018 			htonl(ia->ia_net | ~ ia->ia_netmask);
1019 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1020 		ia->ia_ifa.ifa_dstaddr = ia->ia_ifa.ifa_addr;
1021 		flags |= RTF_HOST;
1022 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1023 		if (ia->ia_dstaddr.sin_family != AF_INET)
1024 			return (0);
1025 		flags |= RTF_HOST;
1026 	}
1027 
1028 	/*-
1029 	 * Don't add host routes for interface addresses of
1030 	 * 0.0.0.0 --> 0.255.255.255 netmask 255.0.0.0.  This makes it
1031 	 * possible to assign several such address pairs with consistent
1032 	 * results (no host route) and is required by BOOTP.
1033 	 *
1034 	 * XXX: This is ugly !  There should be a way for the caller to
1035 	 *      say that they don't want a host route.
1036 	 */
1037 	if (ia->ia_addr.sin_addr.s_addr != INADDR_ANY ||
1038 	    ia->ia_netmask != IN_CLASSA_NET ||
1039 	    ia->ia_dstaddr.sin_addr.s_addr != htonl(IN_CLASSA_HOST)) {
1040 		if ((error = rtinit(&ia->ia_ifa, RTM_ADD, flags)) != 0)
1041 			goto fail;
1042 		ia->ia_flags |= IFA_ROUTE;
1043 	}
1044 
1045 	/*
1046 	 * If the interface supports multicast, join the "all hosts"
1047 	 * multicast group on that interface.
1048 	 */
1049 	if (ifp->if_flags & IFF_MULTICAST) {
1050 		struct in_addr addr;
1051 
1052 		addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
1053 		in_addmulti(&addr, ifp);
1054 	}
1055 	return (0);
1056 fail:
1057 	if (ifac->ifa_listmask & IFA_LIST_IN_IFADDRHASH)
1058 		in_iahash_remove(ia);
1059 
1060 	ia->ia_addr = oldaddr;
1061 	if (was_hash)
1062 		in_iahash_insert(ia);
1063 	return (error);
1064 }
1065 
1066 
1067 /*
1068  * Return 1 if the address might be a local broadcast address.
1069  */
1070 int
1071 in_broadcast(struct in_addr in, struct ifnet *ifp)
1072 {
1073 	struct ifaddr_container *ifac;
1074 	u_long t;
1075 
1076 	if (in.s_addr == INADDR_BROADCAST ||
1077 	    in.s_addr == INADDR_ANY)
1078 		return 1;
1079 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1080 		return 0;
1081 	t = ntohl(in.s_addr);
1082 	/*
1083 	 * Look through the list of addresses for a match
1084 	 * with a broadcast address.
1085 	 */
1086 #define ia ((struct in_ifaddr *)ifa)
1087 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1088 		struct ifaddr *ifa = ifac->ifa;
1089 
1090 		if (ifa->ifa_addr->sa_family == AF_INET &&
1091 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1092 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1093 		     /*
1094 		      * Check for old-style (host 0) broadcast.
1095 		      */
1096 		     t == ia->ia_subnet || t == ia->ia_net) &&
1097 		     /*
1098 		      * Check for an all one subnetmask. These
1099 		      * only exist when an interface gets a secondary
1100 		      * address.
1101 		      */
1102 		     ia->ia_subnetmask != (u_long)0xffffffff)
1103 			    return 1;
1104 	}
1105 	return (0);
1106 #undef ia
1107 }
1108 /*
1109  * Add an address to the list of IP multicast addresses for a given interface.
1110  */
1111 struct in_multi *
1112 in_addmulti(struct in_addr *ap, struct ifnet *ifp)
1113 {
1114 	struct in_multi *inm;
1115 	int error;
1116 	struct sockaddr_in sin;
1117 	struct ifmultiaddr *ifma;
1118 
1119 	/*
1120 	 * Call generic routine to add membership or increment
1121 	 * refcount.  It wants addresses in the form of a sockaddr,
1122 	 * so we build one here (being careful to zero the unused bytes).
1123 	 */
1124 	bzero(&sin, sizeof sin);
1125 	sin.sin_family = AF_INET;
1126 	sin.sin_len = sizeof sin;
1127 	sin.sin_addr = *ap;
1128 	crit_enter();
1129 	error = if_addmulti(ifp, (struct sockaddr *)&sin, &ifma);
1130 	if (error) {
1131 		crit_exit();
1132 		return 0;
1133 	}
1134 
1135 	/*
1136 	 * If ifma->ifma_protospec is null, then if_addmulti() created
1137 	 * a new record.  Otherwise, we are done.
1138 	 */
1139 	if (ifma->ifma_protospec != 0) {
1140 		crit_exit();
1141 		return ifma->ifma_protospec;
1142 	}
1143 
1144 	/* XXX - if_addmulti uses M_WAITOK.  Can this really be called
1145 	   at interrupt time?  If so, need to fix if_addmulti. XXX */
1146 	inm = kmalloc(sizeof *inm, M_IPMADDR, M_WAITOK | M_ZERO);
1147 	inm->inm_addr = *ap;
1148 	inm->inm_ifp = ifp;
1149 	inm->inm_ifma = ifma;
1150 	ifma->ifma_protospec = inm;
1151 	LIST_INSERT_HEAD(&in_multihead, inm, inm_link);
1152 
1153 	/*
1154 	 * Let IGMP know that we have joined a new IP multicast group.
1155 	 */
1156 	igmp_joingroup(inm);
1157 	crit_exit();
1158 	return (inm);
1159 }
1160 
1161 /*
1162  * Delete a multicast address record.
1163  */
1164 void
1165 in_delmulti(struct in_multi *inm)
1166 {
1167 	struct ifmultiaddr *ifma;
1168 	struct in_multi my_inm;
1169 
1170 	crit_enter();
1171 	ifma = inm->inm_ifma;
1172 	my_inm.inm_ifp = NULL ; /* don't send the leave msg */
1173 	if (ifma->ifma_refcount == 1) {
1174 		/*
1175 		 * No remaining claims to this record; let IGMP know that
1176 		 * we are leaving the multicast group.
1177 		 * But do it after the if_delmulti() which might reset
1178 		 * the interface and nuke the packet.
1179 		 */
1180 		my_inm = *inm ;
1181 		ifma->ifma_protospec = 0;
1182 		LIST_REMOVE(inm, inm_link);
1183 		kfree(inm, M_IPMADDR);
1184 	}
1185 	/* XXX - should be separate API for when we have an ifma? */
1186 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1187 	if (my_inm.inm_ifp != NULL)
1188 		igmp_leavegroup(&my_inm);
1189 	crit_exit();
1190 }
1191 
1192 void
1193 in_ifdetach(struct ifnet *ifp)
1194 {
1195 	in_pcbpurgeif0(LIST_FIRST(&ripcbinfo.pcblisthead), ifp);
1196 	in_pcbpurgeif0(LIST_FIRST(&udbinfo.pcblisthead), ifp);
1197 }
1198