xref: /netbsd-src/sys/net/if.c (revision 001c68bd94f75ce9270b69227c4199fbf34ee396)
1 /*	$NetBSD: if.c,v 1.124 2003/07/06 07:54:43 dyoung Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000, 2001 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by William Studnemund and Jason R. Thorpe.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the NetBSD
21  *	Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 
39 /*
40  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
41  * All rights reserved.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. Neither the name of the project nor the names of its contributors
52  *    may be used to endorse or promote products derived from this software
53  *    without specific prior written permission.
54  *
55  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65  * SUCH DAMAGE.
66  */
67 
68 /*
69  * Copyright (c) 1980, 1986, 1993
70  *	The Regents of the University of California.  All rights reserved.
71  *
72  * Redistribution and use in source and binary forms, with or without
73  * modification, are permitted provided that the following conditions
74  * are met:
75  * 1. Redistributions of source code must retain the above copyright
76  *    notice, this list of conditions and the following disclaimer.
77  * 2. Redistributions in binary form must reproduce the above copyright
78  *    notice, this list of conditions and the following disclaimer in the
79  *    documentation and/or other materials provided with the distribution.
80  * 3. All advertising materials mentioning features or use of this software
81  *    must display the following acknowledgement:
82  *	This product includes software developed by the University of
83  *	California, Berkeley and its contributors.
84  * 4. Neither the name of the University nor the names of its contributors
85  *    may be used to endorse or promote products derived from this software
86  *    without specific prior written permission.
87  *
88  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
89  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
90  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
91  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
92  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
93  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
94  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
95  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
96  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
97  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
98  * SUCH DAMAGE.
99  *
100  *	@(#)if.c	8.5 (Berkeley) 1/9/95
101  */
102 
103 #include <sys/cdefs.h>
104 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.124 2003/07/06 07:54:43 dyoung Exp $");
105 
106 #include "opt_inet.h"
107 
108 #include "opt_compat_linux.h"
109 #include "opt_compat_svr4.h"
110 #include "opt_compat_ultrix.h"
111 #include "opt_compat_43.h"
112 #include "opt_atalk.h"
113 #include "opt_ccitt.h"
114 #include "opt_natm.h"
115 #include "opt_pfil_hooks.h"
116 
117 #include <sys/param.h>
118 #include <sys/mbuf.h>
119 #include <sys/systm.h>
120 #include <sys/callout.h>
121 #include <sys/proc.h>
122 #include <sys/socket.h>
123 #include <sys/socketvar.h>
124 #include <sys/domain.h>
125 #include <sys/protosw.h>
126 #include <sys/kernel.h>
127 #include <sys/ioctl.h>
128 
129 #include <net/if.h>
130 #include <net/if_dl.h>
131 #include <net/if_ether.h>
132 #include <net/if_media.h>
133 #include <net/if_ieee80211.h>
134 #include <net/if_types.h>
135 #include <net/radix.h>
136 #include <net/route.h>
137 #include <net/netisr.h>
138 #ifdef NETATALK
139 #include <netatalk/at_extern.h>
140 #include <netatalk/at.h>
141 #endif
142 
143 #ifdef INET6
144 #include <netinet/in.h>
145 #include <netinet6/in6_var.h>
146 #include <netinet6/nd6.h>
147 #endif
148 
149 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
150 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
151 
152 int	ifqmaxlen = IFQ_MAXLEN;
153 struct	callout if_slowtimo_ch;
154 
155 int netisr;			/* scheduling bits for network */
156 
157 int if_rt_walktree __P((struct radix_node *, void *));
158 
159 struct if_clone *if_clone_lookup __P((const char *, int *));
160 int if_clone_list __P((struct if_clonereq *));
161 
162 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
163 int if_cloners_count;
164 
165 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
166     defined(ISO) || defined(CCITT) || defined(NATM)
167 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
168 #endif
169 
170 /*
171  * Network interface utility routines.
172  *
173  * Routines with ifa_ifwith* names take sockaddr *'s as
174  * parameters.
175  */
176 void
177 ifinit()
178 {
179 
180 	callout_init(&if_slowtimo_ch);
181 	if_slowtimo(NULL);
182 }
183 
184 /*
185  * Null routines used while an interface is going away.  These routines
186  * just return an error.
187  */
188 
189 int
190 if_nulloutput(ifp, m, so, rt)
191 	struct ifnet *ifp;
192 	struct mbuf *m;
193 	struct sockaddr *so;
194 	struct rtentry *rt;
195 {
196 
197 	return (ENXIO);
198 }
199 
200 void
201 if_nullinput(ifp, m)
202 	struct ifnet *ifp;
203 	struct mbuf *m;
204 {
205 
206 	/* Nothing. */
207 }
208 
209 void
210 if_nullstart(ifp)
211 	struct ifnet *ifp;
212 {
213 
214 	/* Nothing. */
215 }
216 
217 int
218 if_nullioctl(ifp, cmd, data)
219 	struct ifnet *ifp;
220 	u_long cmd;
221 	caddr_t data;
222 {
223 
224 	return (ENXIO);
225 }
226 
227 int
228 if_nullinit(ifp)
229 	struct ifnet *ifp;
230 {
231 
232 	return (ENXIO);
233 }
234 
235 void
236 if_nullstop(ifp, disable)
237 	struct ifnet *ifp;
238 	int disable;
239 {
240 
241 	/* Nothing. */
242 }
243 
244 void
245 if_nullwatchdog(ifp)
246 	struct ifnet *ifp;
247 {
248 
249 	/* Nothing. */
250 }
251 
252 void
253 if_nulldrain(ifp)
254 	struct ifnet *ifp;
255 {
256 
257 	/* Nothing. */
258 }
259 
260 u_int if_index = 1;
261 struct ifnet_head ifnet;
262 struct ifaddr **ifnet_addrs = NULL;
263 struct ifnet **ifindex2ifnet = NULL;
264 
265 /*
266  * Allocate the link level name for the specified interface.  This
267  * is an attachment helper.  It must be called after ifp->if_addrlen
268  * is initialized, which may not be the case when if_attach() is
269  * called.
270  */
271 void
272 if_alloc_sadl(struct ifnet *ifp)
273 {
274 	unsigned socksize, ifasize;
275 	int namelen, masklen;
276 	struct sockaddr_dl *sdl;
277 	struct ifaddr *ifa;
278 
279 	/*
280 	 * If the interface already has a link name, release it
281 	 * now.  This is useful for interfaces that can change
282 	 * link types, and thus switch link names often.
283 	 */
284 	if (ifp->if_sadl != NULL)
285 		if_free_sadl(ifp);
286 
287 	namelen = strlen(ifp->if_xname);
288 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
289 	socksize = masklen + ifp->if_addrlen;
290 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
291 	if (socksize < sizeof(*sdl))
292 		socksize = sizeof(*sdl);
293 	socksize = ROUNDUP(socksize);
294 	ifasize = sizeof(*ifa) + 2 * socksize;
295 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
296 	memset((caddr_t)ifa, 0, ifasize);
297 	sdl = (struct sockaddr_dl *)(ifa + 1);
298 	sdl->sdl_len = socksize;
299 	sdl->sdl_family = AF_LINK;
300 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
301 	sdl->sdl_nlen = namelen;
302 	sdl->sdl_alen = ifp->if_addrlen;
303 	sdl->sdl_index = ifp->if_index;
304 	sdl->sdl_type = ifp->if_type;
305 	ifnet_addrs[ifp->if_index] = ifa;
306 	IFAREF(ifa);
307 	ifa->ifa_ifp = ifp;
308 	ifa->ifa_rtrequest = link_rtrequest;
309 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
310 	IFAREF(ifa);
311 	ifa->ifa_addr = (struct sockaddr *)sdl;
312 	ifp->if_sadl = sdl;
313 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
314 	ifa->ifa_netmask = (struct sockaddr *)sdl;
315 	sdl->sdl_len = masklen;
316 	while (namelen != 0)
317 		sdl->sdl_data[--namelen] = 0xff;
318 }
319 
320 /*
321  * Free the link level name for the specified interface.  This is
322  * a detach helper.  This is called from if_detach() or from
323  * link layer type specific detach functions.
324  */
325 void
326 if_free_sadl(struct ifnet *ifp)
327 {
328 	struct ifaddr *ifa;
329 	int s;
330 
331 	ifa = ifnet_addrs[ifp->if_index];
332 	if (ifa == NULL) {
333 		KASSERT(ifp->if_sadl == NULL);
334 		return;
335 	}
336 
337 	KASSERT(ifp->if_sadl != NULL);
338 
339 	s = splnet();
340 	rtinit(ifa, RTM_DELETE, 0);
341 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
342 	IFAFREE(ifa);
343 
344 	ifp->if_sadl = NULL;
345 
346 	ifnet_addrs[ifp->if_index] = NULL;
347 	IFAFREE(ifa);
348 	splx(s);
349 }
350 
351 /*
352  * Attach an interface to the
353  * list of "active" interfaces.
354  */
355 void
356 if_attach(ifp)
357 	struct ifnet *ifp;
358 {
359 	static size_t if_indexlim = 0;
360 	int indexlim = 0;
361 
362 	if (if_indexlim == 0) {
363 		TAILQ_INIT(&ifnet);
364 		if_indexlim = 8;
365 	}
366 	TAILQ_INIT(&ifp->if_addrlist);
367 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
368 	ifp->if_index = if_index;
369 	if (ifindex2ifnet == 0)
370 		if_index++;
371 	else
372 		while (ifindex2ifnet[ifp->if_index] != NULL) {
373 			++if_index;
374 			if (if_index == 0)
375 				if_index = 1;
376 			/*
377 			 * If we hit USHRT_MAX, we skip back to 0 since
378 			 * there are a number of places where the value
379 			 * of if_index or if_index itself is compared
380 			 * to or stored in an unsigned short.  By
381 			 * jumping back, we won't botch those assignments
382 			 * or comparisons.
383 			 */
384 			else if (if_index == USHRT_MAX) {
385 				/*
386 				 * However, if we have to jump back to
387 				 * zero *twice* without finding an empty
388 				 * slot in ifindex2ifnet[], then there
389 				 * there are too many (>65535) interfaces.
390 				 */
391 				if (indexlim++)
392 					panic("too many interfaces");
393 				else
394 					if_index = 1;
395 			}
396 			ifp->if_index = if_index;
397 		}
398 
399 	/*
400 	 * We have some arrays that should be indexed by if_index.
401 	 * since if_index will grow dynamically, they should grow too.
402 	 *	struct ifadd **ifnet_addrs
403 	 *	struct ifnet **ifindex2ifnet
404 	 */
405 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
406 	    ifp->if_index >= if_indexlim) {
407 		size_t n;
408 		caddr_t q;
409 
410 		while (ifp->if_index >= if_indexlim)
411 			if_indexlim <<= 1;
412 
413 		/* grow ifnet_addrs */
414 		n = if_indexlim * sizeof(struct ifaddr *);
415 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
416 		memset(q, 0, n);
417 		if (ifnet_addrs) {
418 			bcopy((caddr_t)ifnet_addrs, q, n/2);
419 			free((caddr_t)ifnet_addrs, M_IFADDR);
420 		}
421 		ifnet_addrs = (struct ifaddr **)q;
422 
423 		/* grow ifindex2ifnet */
424 		n = if_indexlim * sizeof(struct ifnet *);
425 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
426 		memset(q, 0, n);
427 		if (ifindex2ifnet) {
428 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
429 			free((caddr_t)ifindex2ifnet, M_IFADDR);
430 		}
431 		ifindex2ifnet = (struct ifnet **)q;
432 	}
433 
434 	ifindex2ifnet[ifp->if_index] = ifp;
435 
436 	/*
437 	 * Link level name is allocated later by a separate call to
438 	 * if_alloc_sadl().
439 	 */
440 
441 	if (ifp->if_snd.ifq_maxlen == 0)
442 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
443 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
444 
445 	ifp->if_link_state = LINK_STATE_UNKNOWN;
446 
447 	ifp->if_capenable = 0;
448 	ifp->if_csum_flags_tx = 0;
449 	ifp->if_csum_flags_rx = 0;
450 
451 #ifdef ALTQ
452 	ifp->if_snd.altq_type = 0;
453 	ifp->if_snd.altq_disc = NULL;
454 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
455 	ifp->if_snd.altq_tbr  = NULL;
456 	ifp->if_snd.altq_ifp  = ifp;
457 #endif
458 
459 #ifdef PFIL_HOOKS
460 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
461 	ifp->if_pfil.ph_ifnet = ifp;
462 	if (pfil_head_register(&ifp->if_pfil) != 0)
463 		printf("%s: WARNING: unable to register pfil hook\n",
464 		    ifp->if_xname);
465 #endif
466 
467 	if (domains)
468 		if_attachdomain1(ifp);
469 
470 	/* Announce the interface. */
471 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
472 }
473 
474 void
475 if_attachdomain()
476 {
477 	struct ifnet *ifp;
478 	int s;
479 
480 	s = splnet();
481 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
482 		if_attachdomain1(ifp);
483 	splx(s);
484 }
485 
486 void
487 if_attachdomain1(ifp)
488 	struct ifnet *ifp;
489 {
490 	struct domain *dp;
491 	int s;
492 
493 	s = splnet();
494 
495 	/* address family dependent data region */
496 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
497 	for (dp = domains; dp; dp = dp->dom_next) {
498 		if (dp->dom_ifattach)
499 			ifp->if_afdata[dp->dom_family] =
500 			    (*dp->dom_ifattach)(ifp);
501 	}
502 
503 	splx(s);
504 }
505 
506 /*
507  * Deactivate an interface.  This points all of the procedure
508  * handles at error stubs.  May be called from interrupt context.
509  */
510 void
511 if_deactivate(ifp)
512 	struct ifnet *ifp;
513 {
514 	int s;
515 
516 	s = splnet();
517 
518 	ifp->if_output	 = if_nulloutput;
519 	ifp->if_input	 = if_nullinput;
520 	ifp->if_start	 = if_nullstart;
521 	ifp->if_ioctl	 = if_nullioctl;
522 	ifp->if_init	 = if_nullinit;
523 	ifp->if_stop	 = if_nullstop;
524 	ifp->if_watchdog = if_nullwatchdog;
525 	ifp->if_drain	 = if_nulldrain;
526 
527 	/* No more packets may be enqueued. */
528 	ifp->if_snd.ifq_maxlen = 0;
529 
530 	splx(s);
531 }
532 
533 /*
534  * Detach an interface from the list of "active" interfaces,
535  * freeing any resources as we go along.
536  *
537  * NOTE: This routine must be called with a valid thread context,
538  * as it may block.
539  */
540 void
541 if_detach(ifp)
542 	struct ifnet *ifp;
543 {
544 	struct socket so;
545 	struct ifaddr *ifa, *next;
546 #ifdef IFAREF_DEBUG
547 	struct ifaddr *last_ifa = NULL;
548 #endif
549 	struct domain *dp;
550 	struct protosw *pr;
551 	struct radix_node_head *rnh;
552 	int s, i, family, purged;
553 
554 	/*
555 	 * XXX It's kind of lame that we have to have the
556 	 * XXX socket structure...
557 	 */
558 	memset(&so, 0, sizeof(so));
559 
560 	s = splnet();
561 
562 	/*
563 	 * Do an if_down() to give protocols a chance to do something.
564 	 */
565 	if_down(ifp);
566 
567 #ifdef ALTQ
568 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
569 		altq_disable(&ifp->if_snd);
570 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
571 		altq_detach(&ifp->if_snd);
572 #endif
573 
574 #ifdef PFIL_HOOKS
575 	(void) pfil_head_unregister(&ifp->if_pfil);
576 #endif
577 
578 	/*
579 	 * Rip all the addresses off the interface.  This should make
580 	 * all of the routes go away.
581 	 */
582 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa; ifa = next) {
583 		next = TAILQ_NEXT(ifa, ifa_list);
584 		family = ifa->ifa_addr->sa_family;
585 #ifdef IFAREF_DEBUG
586 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
587 		    ifa, family, ifa->ifa_refcnt);
588 		if (last_ifa != NULL && ifa == last_ifa)
589 			panic("if_detach: loop detected");
590 		last_ifa = ifa;
591 #endif
592 		if (family == AF_LINK)
593 			continue;
594 		dp = pffinddomain(family);
595 #ifdef DIAGNOSTIC
596 		if (dp == NULL)
597 			panic("if_detach: no domain for AF %d",
598 			    family);
599 #endif
600 		purged = 0;
601 		for (pr = dp->dom_protosw;
602 		     pr < dp->dom_protoswNPROTOSW; pr++) {
603 			so.so_proto = pr;
604 			if (pr->pr_usrreq != NULL) {
605 				(void) (*pr->pr_usrreq)(&so,
606 				    PRU_PURGEIF, NULL, NULL,
607 				    (struct mbuf *) ifp, curproc);
608 				purged = 1;
609 			}
610 		}
611 		if (purged == 0) {
612 			/*
613 			 * XXX What's really the best thing to do
614 			 * XXX here?  --thorpej@netbsd.org
615 			 */
616 			printf("if_detach: WARNING: AF %d not purged\n",
617 			    family);
618 		}
619 	}
620 
621 	if_free_sadl(ifp);
622 
623 	/* Walk the routing table looking for straglers. */
624 	for (i = 0; i <= AF_MAX; i++) {
625 		if ((rnh = rt_tables[i]) != NULL)
626 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
627 	}
628 
629 	for (dp = domains; dp; dp = dp->dom_next) {
630 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
631 			(*dp->dom_ifdetach)(ifp,
632 			    ifp->if_afdata[dp->dom_family]);
633 	}
634 
635 	/* Announce that the interface is gone. */
636 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
637 
638 	ifindex2ifnet[ifp->if_index] = NULL;
639 
640 	TAILQ_REMOVE(&ifnet, ifp, if_list);
641 
642 	/*
643 	 * remove packets came from ifp, from software interrupt queues.
644 	 * net/netisr_dispatch.h is not usable, as some of them use
645 	 * strange queue names.
646 	 */
647 #define IF_DETACH_QUEUES(x) \
648 do { \
649 	extern struct ifqueue x; \
650 	if_detach_queues(ifp, & x); \
651 } while (/*CONSTCOND*/ 0)
652 #ifdef INET
653 #if NARP > 0
654 	IF_DETACH_QUEUES(arpintrq);
655 #endif
656 	IF_DETACH_QUEUES(ipintrq);
657 #endif
658 #ifdef INET6
659 	IF_DETACH_QUEUES(ip6intrq);
660 #endif
661 #ifdef NETATALK
662 	IF_DETACH_QUEUES(atintrq1);
663 	IF_DETACH_QUEUES(atintrq2);
664 #endif
665 #ifdef NS
666 	IF_DETACH_QUEUES(nsintrq);
667 #endif
668 #ifdef ISO
669 	IF_DETACH_QUEUES(clnlintrq);
670 #endif
671 #ifdef CCITT
672 	IF_DETACH_QUEUES(llcintrq);
673 	IF_DETACH_QUEUES(hdintrq);
674 #endif
675 #ifdef NATM
676 	IF_DETACH_QUEUES(natmintrq);
677 #endif
678 #undef IF_DETACH_QUEUES
679 
680 	splx(s);
681 }
682 
683 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
684     defined(ISO) || defined(CCITT) || defined(NATM)
685 static void
686 if_detach_queues(ifp, q)
687 	struct ifnet *ifp;
688 	struct ifqueue *q;
689 {
690 	struct mbuf *m, *prev, *next;
691 
692 	prev = NULL;
693 	for (m = q->ifq_head; m; m = next) {
694 		next = m->m_nextpkt;
695 #ifdef DIAGNOSTIC
696 		if ((m->m_flags & M_PKTHDR) == 0) {
697 			prev = m;
698 			continue;
699 		}
700 #endif
701 		if (m->m_pkthdr.rcvif != ifp) {
702 			prev = m;
703 			continue;
704 		}
705 
706 		if (prev)
707 			prev->m_nextpkt = m->m_nextpkt;
708 		else
709 			q->ifq_head = m->m_nextpkt;
710 		if (q->ifq_tail == m)
711 			q->ifq_tail = prev;
712 		q->ifq_len--;
713 
714 		m->m_nextpkt = NULL;
715 		m_freem(m);
716 		IF_DROP(q);
717 	}
718 }
719 #endif /* defined(INET) || ... */
720 
721 /*
722  * Callback for a radix tree walk to delete all references to an
723  * ifnet.
724  */
725 int
726 if_rt_walktree(rn, v)
727 	struct radix_node *rn;
728 	void *v;
729 {
730 	struct ifnet *ifp = (struct ifnet *)v;
731 	struct rtentry *rt = (struct rtentry *)rn;
732 	int error;
733 
734 	if (rt->rt_ifp == ifp) {
735 		/* Delete the entry. */
736 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
737 		    rt_mask(rt), rt->rt_flags, NULL);
738 		if (error)
739 			printf("%s: warning: unable to delete rtentry @ %p, "
740 			    "error = %d\n", ifp->if_xname, rt, error);
741 	}
742 	return (0);
743 }
744 
745 /*
746  * Create a clone network interface.
747  */
748 int
749 if_clone_create(name)
750 	const char *name;
751 {
752 	struct if_clone *ifc;
753 	int unit;
754 
755 	ifc = if_clone_lookup(name, &unit);
756 	if (ifc == NULL)
757 		return (EINVAL);
758 
759 	if (ifunit(name) != NULL)
760 		return (EEXIST);
761 
762 	return ((*ifc->ifc_create)(ifc, unit));
763 }
764 
765 /*
766  * Destroy a clone network interface.
767  */
768 int
769 if_clone_destroy(name)
770 	const char *name;
771 {
772 	struct if_clone *ifc;
773 	struct ifnet *ifp;
774 
775 	ifc = if_clone_lookup(name, NULL);
776 	if (ifc == NULL)
777 		return (EINVAL);
778 
779 	ifp = ifunit(name);
780 	if (ifp == NULL)
781 		return (ENXIO);
782 
783 	if (ifc->ifc_destroy == NULL)
784 		return (EOPNOTSUPP);
785 
786 	(*ifc->ifc_destroy)(ifp);
787 	return (0);
788 }
789 
790 /*
791  * Look up a network interface cloner.
792  */
793 struct if_clone *
794 if_clone_lookup(name, unitp)
795 	const char *name;
796 	int *unitp;
797 {
798 	struct if_clone *ifc;
799 	const char *cp;
800 	size_t i;
801 
802 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
803 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
804 			if (ifc->ifc_name[i] != *cp)
805 				goto next_ifc;
806 		}
807 		goto found_name;
808  next_ifc:
809 		ifc = LIST_NEXT(ifc, ifc_list);
810 	}
811 
812 	/* No match. */
813 	return (NULL);
814 
815  found_name:
816 	for (i = 0; *cp != '\0'; cp++) {
817 		if (*cp < '0' || *cp > '9') {
818 			/* Bogus unit number. */
819 			return (NULL);
820 		}
821 		i = (i * 10) + (*cp - '0');
822 	}
823 
824 	if (unitp != NULL)
825 		*unitp = i;
826 	return (ifc);
827 }
828 
829 /*
830  * Register a network interface cloner.
831  */
832 void
833 if_clone_attach(ifc)
834 	struct if_clone *ifc;
835 {
836 
837 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
838 	if_cloners_count++;
839 }
840 
841 /*
842  * Unregister a network interface cloner.
843  */
844 void
845 if_clone_detach(ifc)
846 	struct if_clone *ifc;
847 {
848 
849 	LIST_REMOVE(ifc, ifc_list);
850 	if_cloners_count--;
851 }
852 
853 /*
854  * Provide list of interface cloners to userspace.
855  */
856 int
857 if_clone_list(ifcr)
858 	struct if_clonereq *ifcr;
859 {
860 	char outbuf[IFNAMSIZ], *dst;
861 	struct if_clone *ifc;
862 	int count, error = 0;
863 
864 	ifcr->ifcr_total = if_cloners_count;
865 	if ((dst = ifcr->ifcr_buffer) == NULL) {
866 		/* Just asking how many there are. */
867 		return (0);
868 	}
869 
870 	if (ifcr->ifcr_count < 0)
871 		return (EINVAL);
872 
873 	count = (if_cloners_count < ifcr->ifcr_count) ?
874 	    if_cloners_count : ifcr->ifcr_count;
875 
876 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
877 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
878 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
879 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
880 		error = copyout(outbuf, dst, IFNAMSIZ);
881 		if (error)
882 			break;
883 	}
884 
885 	return (error);
886 }
887 
888 /*
889  * Locate an interface based on a complete address.
890  */
891 /*ARGSUSED*/
892 struct ifaddr *
893 ifa_ifwithaddr(addr)
894 	struct sockaddr *addr;
895 {
896 	struct ifnet *ifp;
897 	struct ifaddr *ifa;
898 
899 #define	equal(a1, a2) \
900   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
901 
902 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
903 	     ifp = TAILQ_NEXT(ifp, if_list)) {
904 		if (ifp->if_output == if_nulloutput)
905 			continue;
906 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
907 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
908 			if (ifa->ifa_addr->sa_family != addr->sa_family)
909 				continue;
910 			if (equal(addr, ifa->ifa_addr))
911 				return (ifa);
912 			if ((ifp->if_flags & IFF_BROADCAST) &&
913 			    ifa->ifa_broadaddr &&
914 			    /* IP6 doesn't have broadcast */
915 			    ifa->ifa_broadaddr->sa_len != 0 &&
916 			    equal(ifa->ifa_broadaddr, addr))
917 				return (ifa);
918 		}
919 	}
920 	return (NULL);
921 }
922 
923 /*
924  * Locate the point to point interface with a given destination address.
925  */
926 /*ARGSUSED*/
927 struct ifaddr *
928 ifa_ifwithdstaddr(addr)
929 	struct sockaddr *addr;
930 {
931 	struct ifnet *ifp;
932 	struct ifaddr *ifa;
933 
934 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
935 	     ifp = TAILQ_NEXT(ifp, if_list)) {
936 		if (ifp->if_output == if_nulloutput)
937 			continue;
938 		if (ifp->if_flags & IFF_POINTOPOINT) {
939 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
940 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
941 				if (ifa->ifa_addr->sa_family !=
942 				      addr->sa_family ||
943 				    ifa->ifa_dstaddr == NULL)
944 					continue;
945 				if (equal(addr, ifa->ifa_dstaddr))
946 					return (ifa);
947 			}
948 		}
949 	}
950 	return (NULL);
951 }
952 
953 /*
954  * Find an interface on a specific network.  If many, choice
955  * is most specific found.
956  */
957 struct ifaddr *
958 ifa_ifwithnet(addr)
959 	struct sockaddr *addr;
960 {
961 	struct ifnet *ifp;
962 	struct ifaddr *ifa;
963 	struct sockaddr_dl *sdl;
964 	struct ifaddr *ifa_maybe = 0;
965 	u_int af = addr->sa_family;
966 	char *addr_data = addr->sa_data, *cplim;
967 
968 	if (af == AF_LINK) {
969 		sdl = (struct sockaddr_dl *)addr;
970 		if (sdl->sdl_index && sdl->sdl_index <= if_index &&
971 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
972 			return (ifnet_addrs[sdl->sdl_index]);
973 	}
974 #ifdef NETATALK
975 	if (af == AF_APPLETALK) {
976 		struct sockaddr_at *sat, *sat2;
977 		sat = (struct sockaddr_at *)addr;
978 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
979 		     ifp = TAILQ_NEXT(ifp, if_list)) {
980 			if (ifp->if_output == if_nulloutput)
981 				continue;
982 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
983 			if (ifa == NULL)
984 				continue;
985 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
986 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
987 				return (ifa); /* exact match */
988 			if (ifa_maybe == NULL) {
989 				/* else keep the if with the right range */
990 				ifa_maybe = ifa;
991 			}
992 		}
993 		return (ifa_maybe);
994 	}
995 #endif
996 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
997 	     ifp = TAILQ_NEXT(ifp, if_list)) {
998 		if (ifp->if_output == if_nulloutput)
999 			continue;
1000 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1001 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1002 			char *cp, *cp2, *cp3;
1003 
1004 			if (ifa->ifa_addr->sa_family != af ||
1005 			    ifa->ifa_netmask == 0)
1006  next:				continue;
1007 			cp = addr_data;
1008 			cp2 = ifa->ifa_addr->sa_data;
1009 			cp3 = ifa->ifa_netmask->sa_data;
1010 			cplim = (char *)ifa->ifa_netmask +
1011 			    ifa->ifa_netmask->sa_len;
1012 			while (cp3 < cplim) {
1013 				if ((*cp++ ^ *cp2++) & *cp3++) {
1014 					/* want to continue for() loop */
1015 					goto next;
1016 				}
1017 			}
1018 			if (ifa_maybe == 0 ||
1019 			    rn_refines((caddr_t)ifa->ifa_netmask,
1020 			    (caddr_t)ifa_maybe->ifa_netmask))
1021 				ifa_maybe = ifa;
1022 		}
1023 	}
1024 	return (ifa_maybe);
1025 }
1026 
1027 /*
1028  * Find the interface of the addresss.
1029  */
1030 struct ifaddr *
1031 ifa_ifwithladdr(addr)
1032 	struct sockaddr *addr;
1033 {
1034 	struct ifaddr *ia;
1035 
1036 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1037 	    (ia = ifa_ifwithnet(addr)))
1038 		return (ia);
1039 	return (NULL);
1040 }
1041 
1042 /*
1043  * Find an interface using a specific address family
1044  */
1045 struct ifaddr *
1046 ifa_ifwithaf(af)
1047 	int af;
1048 {
1049 	struct ifnet *ifp;
1050 	struct ifaddr *ifa;
1051 
1052 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1053 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1054 		if (ifp->if_output == if_nulloutput)
1055 			continue;
1056 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1057 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1058 			if (ifa->ifa_addr->sa_family == af)
1059 				return (ifa);
1060 		}
1061 	}
1062 	return (NULL);
1063 }
1064 
1065 /*
1066  * Find an interface address specific to an interface best matching
1067  * a given address.
1068  */
1069 struct ifaddr *
1070 ifaof_ifpforaddr(addr, ifp)
1071 	struct sockaddr *addr;
1072 	struct ifnet *ifp;
1073 {
1074 	struct ifaddr *ifa;
1075 	char *cp, *cp2, *cp3;
1076 	char *cplim;
1077 	struct ifaddr *ifa_maybe = 0;
1078 	u_int af = addr->sa_family;
1079 
1080 	if (ifp->if_output == if_nulloutput)
1081 		return (NULL);
1082 
1083 	if (af >= AF_MAX)
1084 		return (NULL);
1085 
1086 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1087 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1088 		if (ifa->ifa_addr->sa_family != af)
1089 			continue;
1090 		ifa_maybe = ifa;
1091 		if (ifa->ifa_netmask == 0) {
1092 			if (equal(addr, ifa->ifa_addr) ||
1093 			    (ifa->ifa_dstaddr &&
1094 			     equal(addr, ifa->ifa_dstaddr)))
1095 				return (ifa);
1096 			continue;
1097 		}
1098 		cp = addr->sa_data;
1099 		cp2 = ifa->ifa_addr->sa_data;
1100 		cp3 = ifa->ifa_netmask->sa_data;
1101 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1102 		for (; cp3 < cplim; cp3++) {
1103 			if ((*cp++ ^ *cp2++) & *cp3)
1104 				break;
1105 		}
1106 		if (cp3 == cplim)
1107 			return (ifa);
1108 	}
1109 	return (ifa_maybe);
1110 }
1111 
1112 /*
1113  * Default action when installing a route with a Link Level gateway.
1114  * Lookup an appropriate real ifa to point to.
1115  * This should be moved to /sys/net/link.c eventually.
1116  */
1117 void
1118 link_rtrequest(cmd, rt, info)
1119 	int cmd;
1120 	struct rtentry *rt;
1121 	struct rt_addrinfo *info;
1122 {
1123 	struct ifaddr *ifa;
1124 	struct sockaddr *dst;
1125 	struct ifnet *ifp;
1126 
1127 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1128 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1129 		return;
1130 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1131 		IFAFREE(rt->rt_ifa);
1132 		rt->rt_ifa = ifa;
1133 		IFAREF(ifa);
1134 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1135 			ifa->ifa_rtrequest(cmd, rt, info);
1136 	}
1137 }
1138 
1139 /*
1140  * Mark an interface down and notify protocols of
1141  * the transition.
1142  * NOTE: must be called at splsoftnet or equivalent.
1143  */
1144 void
1145 if_down(ifp)
1146 	struct ifnet *ifp;
1147 {
1148 	struct ifaddr *ifa;
1149 
1150 	ifp->if_flags &= ~IFF_UP;
1151 	microtime(&ifp->if_lastchange);
1152 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1153 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1154 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1155 	IFQ_PURGE(&ifp->if_snd);
1156 	rt_ifmsg(ifp);
1157 }
1158 
1159 /*
1160  * Mark an interface up and notify protocols of
1161  * the transition.
1162  * NOTE: must be called at splsoftnet or equivalent.
1163  */
1164 void
1165 if_up(ifp)
1166 	struct ifnet *ifp;
1167 {
1168 #ifdef notyet
1169 	struct ifaddr *ifa;
1170 #endif
1171 
1172 	ifp->if_flags |= IFF_UP;
1173 	microtime(&ifp->if_lastchange);
1174 #ifdef notyet
1175 	/* this has no effect on IP, and will kill all ISO connections XXX */
1176 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1177 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1178 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1179 #endif
1180 	rt_ifmsg(ifp);
1181 #ifdef INET6
1182 	in6_if_up(ifp);
1183 #endif
1184 }
1185 
1186 /*
1187  * Handle interface watchdog timer routines.  Called
1188  * from softclock, we decrement timers (if set) and
1189  * call the appropriate interface routine on expiration.
1190  */
1191 void
1192 if_slowtimo(arg)
1193 	void *arg;
1194 {
1195 	struct ifnet *ifp;
1196 	int s = splnet();
1197 
1198 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1199 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1200 		if (ifp->if_timer == 0 || --ifp->if_timer)
1201 			continue;
1202 		if (ifp->if_watchdog)
1203 			(*ifp->if_watchdog)(ifp);
1204 	}
1205 	splx(s);
1206 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1207 	    if_slowtimo, NULL);
1208 }
1209 
1210 /*
1211  * Set/clear promiscuous mode on interface ifp based on the truth value
1212  * of pswitch.  The calls are reference counted so that only the first
1213  * "on" request actually has an effect, as does the final "off" request.
1214  * Results are undefined if the "off" and "on" requests are not matched.
1215  */
1216 int
1217 ifpromisc(ifp, pswitch)
1218 	struct ifnet *ifp;
1219 	int pswitch;
1220 {
1221 	int pcount, ret;
1222 	short flags;
1223 	struct ifreq ifr;
1224 
1225 	pcount = ifp->if_pcount;
1226 	flags = ifp->if_flags;
1227 	if (pswitch) {
1228 		/*
1229 		 * Allow the device to be "placed" into promiscuous
1230 		 * mode even if it is not configured up.  It will
1231 		 * consult IFF_PROMISC when it is is brought up.
1232 		 */
1233 		if (ifp->if_pcount++ != 0)
1234 			return (0);
1235 		ifp->if_flags |= IFF_PROMISC;
1236 		if ((ifp->if_flags & IFF_UP) == 0)
1237 			return (0);
1238 	} else {
1239 		if (--ifp->if_pcount > 0)
1240 			return (0);
1241 		ifp->if_flags &= ~IFF_PROMISC;
1242 		/*
1243 		 * If the device is not configured up, we should not need to
1244 		 * turn off promiscuous mode (device should have turned it
1245 		 * off when interface went down; and will look at IFF_PROMISC
1246 		 * again next time interface comes up).
1247 		 */
1248 		if ((ifp->if_flags & IFF_UP) == 0)
1249 			return (0);
1250 	}
1251 	memset(&ifr, 0, sizeof(ifr));
1252 	ifr.ifr_flags = ifp->if_flags;
1253 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1254 	/* Restore interface state if not successful. */
1255 	if (ret != 0) {
1256 		ifp->if_pcount = pcount;
1257 		ifp->if_flags = flags;
1258 	}
1259 	return (ret);
1260 }
1261 
1262 /*
1263  * Map interface name to
1264  * interface structure pointer.
1265  */
1266 struct ifnet *
1267 ifunit(name)
1268 	const char *name;
1269 {
1270 	struct ifnet *ifp;
1271 	const char *cp = name;
1272 	u_int unit = 0;
1273 	u_int i;
1274 
1275 	/*
1276 	 * If the entire name is a number, treat it as an ifindex.
1277 	 */
1278 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1279 		unit = unit * 10 + (*cp - '0');
1280 	}
1281 
1282 	/*
1283 	 * If the number took all of the name, then it's a valid ifindex.
1284 	 */
1285 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1286 		if (unit >= if_index)
1287 			return (NULL);
1288 		ifp = ifindex2ifnet[unit];
1289 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1290 			return (NULL);
1291 		return (ifp);
1292 	}
1293 
1294 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1295 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1296 		if (ifp->if_output == if_nulloutput)
1297 			continue;
1298 	 	if (strcmp(ifp->if_xname, name) == 0)
1299 			return (ifp);
1300 	}
1301 	return (NULL);
1302 }
1303 
1304 /*
1305  * Interface ioctls.
1306  */
1307 int
1308 ifioctl(so, cmd, data, p)
1309 	struct socket *so;
1310 	u_long cmd;
1311 	caddr_t data;
1312 	struct proc *p;
1313 {
1314 	struct ifnet *ifp;
1315 	struct ifreq *ifr;
1316 	struct ifcapreq *ifcr;
1317 	struct ifdatareq *ifdr;
1318 	int s, error = 0;
1319 	short oif_flags;
1320 
1321 	switch (cmd) {
1322 
1323 	case SIOCGIFCONF:
1324 	case OSIOCGIFCONF:
1325 		return (ifconf(cmd, data));
1326 	}
1327 	ifr = (struct ifreq *)data;
1328 	ifcr = (struct ifcapreq *)data;
1329 	ifdr = (struct ifdatareq *)data;
1330 
1331 	switch (cmd) {
1332 	case SIOCIFCREATE:
1333 	case SIOCIFDESTROY:
1334 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1335 			return (error);
1336 		return ((cmd == SIOCIFCREATE) ?
1337 			if_clone_create(ifr->ifr_name) :
1338 			if_clone_destroy(ifr->ifr_name));
1339 
1340 	case SIOCIFGCLONERS:
1341 		return (if_clone_list((struct if_clonereq *)data));
1342 	}
1343 
1344 	ifp = ifunit(ifr->ifr_name);
1345 	if (ifp == 0)
1346 		return (ENXIO);
1347 	oif_flags = ifp->if_flags;
1348 	switch (cmd) {
1349 
1350 	case SIOCGIFFLAGS:
1351 		ifr->ifr_flags = ifp->if_flags;
1352 		break;
1353 
1354 	case SIOCGIFMETRIC:
1355 		ifr->ifr_metric = ifp->if_metric;
1356 		break;
1357 
1358 	case SIOCGIFMTU:
1359 		ifr->ifr_mtu = ifp->if_mtu;
1360 		break;
1361 
1362 	case SIOCGIFDLT:
1363 		ifr->ifr_dlt = ifp->if_dlt;
1364 		break;
1365 
1366 	case SIOCSIFFLAGS:
1367 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1368 			return (error);
1369 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1370 			s = splnet();
1371 			if_down(ifp);
1372 			splx(s);
1373 		}
1374 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1375 			s = splnet();
1376 			if_up(ifp);
1377 			splx(s);
1378 		}
1379 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1380 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1381 		if (ifp->if_ioctl)
1382 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1383 		break;
1384 
1385 	case SIOCGIFCAP:
1386 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1387 		ifcr->ifcr_capenable = ifp->if_capenable;
1388 		break;
1389 
1390 	case SIOCSIFCAP:
1391 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1392 			return (error);
1393 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1394 			return (EINVAL);
1395 		if (ifp->if_ioctl == NULL)
1396 			return (EOPNOTSUPP);
1397 
1398 		/* Must prevent race with packet reception here. */
1399 		s = splnet();
1400 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1401 			struct ifreq ifrq;
1402 
1403 			ifrq.ifr_flags = ifp->if_flags;
1404 			ifp->if_capenable = ifcr->ifcr_capenable;
1405 
1406 			/* Pre-compute the checksum flags mask. */
1407 			ifp->if_csum_flags_tx = 0;
1408 			ifp->if_csum_flags_rx = 0;
1409 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1410 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1411 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1412 			}
1413 
1414 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1415 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1416 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1417 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1418 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1419 
1420 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1421 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1422 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1423 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1424 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1425 
1426 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1427 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1428 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1429 			}
1430 
1431 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1432 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1433 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1434 			}
1435 
1436 			/*
1437 			 * Only kick the interface if it's up.  If it's
1438 			 * not up now, it will notice the cap enables
1439 			 * when it is brought up later.
1440 			 */
1441 			if (ifp->if_flags & IFF_UP)
1442 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1443 				    (caddr_t) &ifrq);
1444 		}
1445 		splx(s);
1446 		break;
1447 
1448 	case SIOCSIFMETRIC:
1449 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1450 			return (error);
1451 		ifp->if_metric = ifr->ifr_metric;
1452 		break;
1453 
1454 	case SIOCGIFDATA:
1455 		ifdr->ifdr_data = ifp->if_data;
1456 		break;
1457 
1458 	case SIOCZIFDATA:
1459 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1460 			return (error);
1461 		ifdr->ifdr_data = ifp->if_data;
1462 		/*
1463 		 * Assumes that the volatile counters that can be
1464 		 * zero'ed are at the end of if_data.
1465 		 */
1466 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1467 		    offsetof(struct if_data, ifi_ipackets));
1468 		break;
1469 
1470 	case SIOCSIFMTU:
1471 	{
1472 		u_long oldmtu = ifp->if_mtu;
1473 
1474 		error = suser(p->p_ucred, &p->p_acflag);
1475 		if (error)
1476 			return (error);
1477 		if (ifp->if_ioctl == NULL)
1478 			return (EOPNOTSUPP);
1479 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1480 
1481 		/*
1482 		 * If the link MTU changed, do network layer specific procedure.
1483 		 */
1484 		if (ifp->if_mtu != oldmtu) {
1485 #ifdef INET6
1486 			nd6_setmtu(ifp);
1487 #endif
1488 		}
1489 		break;
1490 	}
1491 	case SIOCSIFPHYADDR:
1492 	case SIOCDIFPHYADDR:
1493 #ifdef INET6
1494 	case SIOCSIFPHYADDR_IN6:
1495 #endif
1496 	case SIOCSLIFPHYADDR:
1497 	case SIOCADDMULTI:
1498 	case SIOCDELMULTI:
1499 	case SIOCSIFMEDIA:
1500 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1501 			return (error);
1502 		/* FALLTHROUGH */
1503 	case SIOCGIFPSRCADDR:
1504 	case SIOCGIFPDSTADDR:
1505 	case SIOCGLIFPHYADDR:
1506 	case SIOCGIFMEDIA:
1507 		if (ifp->if_ioctl == 0)
1508 			return (EOPNOTSUPP);
1509 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1510 		break;
1511 
1512 	case SIOCSDRVSPEC:
1513 	case SIOCS80211NWID:
1514 	case SIOCS80211NWKEY:
1515 	case SIOCS80211POWER:
1516 	case SIOCS80211BSSID:
1517 	case SIOCS80211CHANNEL:
1518 		/* XXX:  need to pass proc pointer through to driver... */
1519 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1520 			return (error);
1521 	/* FALLTHROUGH */
1522 	default:
1523 		if (so->so_proto == 0)
1524 			return (EOPNOTSUPP);
1525 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(COMPAT_ULTRIX) && !defined(LKM)
1526 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1527 		    (struct mbuf *)cmd, (struct mbuf *)data,
1528 		    (struct mbuf *)ifp, p));
1529 #else
1530 	    {
1531 		int ocmd = cmd;
1532 
1533 		switch (cmd) {
1534 
1535 		case SIOCSIFADDR:
1536 		case SIOCSIFDSTADDR:
1537 		case SIOCSIFBRDADDR:
1538 		case SIOCSIFNETMASK:
1539 #if BYTE_ORDER != BIG_ENDIAN
1540 			if (ifr->ifr_addr.sa_family == 0 &&
1541 			    ifr->ifr_addr.sa_len < 16) {
1542 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1543 				ifr->ifr_addr.sa_len = 16;
1544 			}
1545 #else
1546 			if (ifr->ifr_addr.sa_len == 0)
1547 				ifr->ifr_addr.sa_len = 16;
1548 #endif
1549 			break;
1550 
1551 		case OSIOCGIFADDR:
1552 			cmd = SIOCGIFADDR;
1553 			break;
1554 
1555 		case OSIOCGIFDSTADDR:
1556 			cmd = SIOCGIFDSTADDR;
1557 			break;
1558 
1559 		case OSIOCGIFBRDADDR:
1560 			cmd = SIOCGIFBRDADDR;
1561 			break;
1562 
1563 		case OSIOCGIFNETMASK:
1564 			cmd = SIOCGIFNETMASK;
1565 		}
1566 
1567 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1568 		    (struct mbuf *)cmd, (struct mbuf *)data,
1569 		    (struct mbuf *)ifp, p));
1570 
1571 		switch (ocmd) {
1572 		case OSIOCGIFADDR:
1573 		case OSIOCGIFDSTADDR:
1574 		case OSIOCGIFBRDADDR:
1575 		case OSIOCGIFNETMASK:
1576 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1577 		}
1578 	    }
1579 #endif /* COMPAT_43 */
1580 		break;
1581 	}
1582 
1583 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1584 #ifdef INET6
1585 		if ((ifp->if_flags & IFF_UP) != 0) {
1586 			s = splnet();
1587 			in6_if_up(ifp);
1588 			splx(s);
1589 		}
1590 #endif
1591 	}
1592 
1593 	return (error);
1594 }
1595 
1596 /*
1597  * Return interface configuration
1598  * of system.  List may be used
1599  * in later ioctl's (above) to get
1600  * other information.
1601  */
1602 /*ARGSUSED*/
1603 int
1604 ifconf(cmd, data)
1605 	u_long cmd;
1606 	caddr_t data;
1607 {
1608 	struct ifconf *ifc = (struct ifconf *)data;
1609 	struct ifnet *ifp;
1610 	struct ifaddr *ifa;
1611 	struct ifreq ifr, *ifrp;
1612 	int space = ifc->ifc_len, error = 0;
1613 
1614 	ifrp = ifc->ifc_req;
1615 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1616 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1617 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1618 			memset((caddr_t)&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1619 			if (space >= (int)sizeof (ifr)) {
1620 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1621 						sizeof(ifr));
1622 				if (error)
1623 					break;
1624 			}
1625 			space -= sizeof (ifr), ifrp++;
1626 		} else
1627 		    for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1628 			struct sockaddr *sa = ifa->ifa_addr;
1629 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX)
1630 			if (cmd == OSIOCGIFCONF) {
1631 				struct osockaddr *osa =
1632 					 (struct osockaddr *)&ifr.ifr_addr;
1633 				ifr.ifr_addr = *sa;
1634 				osa->sa_family = sa->sa_family;
1635 				if (space >= (int)sizeof (ifr)) {
1636 					error = copyout((caddr_t)&ifr,
1637 							(caddr_t)ifrp,
1638 							sizeof (ifr));
1639 					ifrp++;
1640 				}
1641 			} else
1642 #endif
1643 			if (sa->sa_len <= sizeof(*sa)) {
1644 				ifr.ifr_addr = *sa;
1645 				if (space >= (int)sizeof (ifr)) {
1646 					error = copyout((caddr_t)&ifr,
1647 							(caddr_t)ifrp,
1648 							sizeof (ifr));
1649 					ifrp++;
1650 				}
1651 			} else {
1652 				space -= sa->sa_len - sizeof(*sa);
1653 				if (space >= (int)sizeof (ifr)) {
1654 					error = copyout((caddr_t)&ifr,
1655 							(caddr_t)ifrp,
1656 							sizeof (ifr.ifr_name));
1657 					if (error == 0) {
1658 						error = copyout((caddr_t)sa,
1659 						  (caddr_t)&ifrp->ifr_addr,
1660 						  sa->sa_len);
1661 					}
1662 					ifrp = (struct ifreq *)
1663 						(sa->sa_len +
1664 						 (caddr_t)&ifrp->ifr_addr);
1665 				}
1666 			}
1667 			if (error)
1668 				break;
1669 			space -= sizeof (ifr);
1670 		}
1671 	}
1672 	ifc->ifc_len -= space;
1673 	return (error);
1674 }
1675