xref: /netbsd-src/sys/net/if.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: if.c,v 1.120 2003/06/23 11:02:08 martin 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 1.120 2003/06/23 11:02:08 martin 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_ieee80211.h>
133 #include <net/if_types.h>
134 #include <net/radix.h>
135 #include <net/route.h>
136 #include <net/netisr.h>
137 #ifdef NETATALK
138 #include <netatalk/at_extern.h>
139 #include <netatalk/at.h>
140 #endif
141 
142 #ifdef INET6
143 #include <netinet/in.h>
144 #include <netinet6/in6_var.h>
145 #include <netinet6/nd6.h>
146 #endif
147 
148 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
149 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
150 
151 int	ifqmaxlen = IFQ_MAXLEN;
152 struct	callout if_slowtimo_ch;
153 
154 int netisr;			/* scheduling bits for network */
155 
156 int if_rt_walktree __P((struct radix_node *, void *));
157 
158 struct if_clone *if_clone_lookup __P((const char *, int *));
159 int if_clone_list __P((struct if_clonereq *));
160 
161 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
162 int if_cloners_count;
163 
164 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
165     defined(ISO) || defined(CCITT) || defined(NATM)
166 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
167 #endif
168 
169 /*
170  * Network interface utility routines.
171  *
172  * Routines with ifa_ifwith* names take sockaddr *'s as
173  * parameters.
174  */
175 void
176 ifinit()
177 {
178 
179 	callout_init(&if_slowtimo_ch);
180 	if_slowtimo(NULL);
181 }
182 
183 /*
184  * Null routines used while an interface is going away.  These routines
185  * just return an error.
186  */
187 
188 int
189 if_nulloutput(ifp, m, so, rt)
190 	struct ifnet *ifp;
191 	struct mbuf *m;
192 	struct sockaddr *so;
193 	struct rtentry *rt;
194 {
195 
196 	return (ENXIO);
197 }
198 
199 void
200 if_nullinput(ifp, m)
201 	struct ifnet *ifp;
202 	struct mbuf *m;
203 {
204 
205 	/* Nothing. */
206 }
207 
208 void
209 if_nullstart(ifp)
210 	struct ifnet *ifp;
211 {
212 
213 	/* Nothing. */
214 }
215 
216 int
217 if_nullioctl(ifp, cmd, data)
218 	struct ifnet *ifp;
219 	u_long cmd;
220 	caddr_t data;
221 {
222 
223 	return (ENXIO);
224 }
225 
226 int
227 if_nullinit(ifp)
228 	struct ifnet *ifp;
229 {
230 
231 	return (ENXIO);
232 }
233 
234 void
235 if_nullstop(ifp, disable)
236 	struct ifnet *ifp;
237 	int disable;
238 {
239 
240 	/* Nothing. */
241 }
242 
243 void
244 if_nullwatchdog(ifp)
245 	struct ifnet *ifp;
246 {
247 
248 	/* Nothing. */
249 }
250 
251 void
252 if_nulldrain(ifp)
253 	struct ifnet *ifp;
254 {
255 
256 	/* Nothing. */
257 }
258 
259 u_int if_index = 1;
260 struct ifnet_head ifnet;
261 struct ifaddr **ifnet_addrs = NULL;
262 struct ifnet **ifindex2ifnet = NULL;
263 
264 /*
265  * Allocate the link level name for the specified interface.  This
266  * is an attachment helper.  It must be called after ifp->if_addrlen
267  * is initialized, which may not be the case when if_attach() is
268  * called.
269  */
270 void
271 if_alloc_sadl(struct ifnet *ifp)
272 {
273 	unsigned socksize, ifasize;
274 	int namelen, masklen;
275 	struct sockaddr_dl *sdl;
276 	struct ifaddr *ifa;
277 
278 	/*
279 	 * If the interface already has a link name, release it
280 	 * now.  This is useful for interfaces that can change
281 	 * link types, and thus switch link names often.
282 	 */
283 	if (ifp->if_sadl != NULL)
284 		if_free_sadl(ifp);
285 
286 	namelen = strlen(ifp->if_xname);
287 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
288 	socksize = masklen + ifp->if_addrlen;
289 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
290 	if (socksize < sizeof(*sdl))
291 		socksize = sizeof(*sdl);
292 	socksize = ROUNDUP(socksize);
293 	ifasize = sizeof(*ifa) + 2 * socksize;
294 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
295 	memset((caddr_t)ifa, 0, ifasize);
296 	sdl = (struct sockaddr_dl *)(ifa + 1);
297 	sdl->sdl_len = socksize;
298 	sdl->sdl_family = AF_LINK;
299 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
300 	sdl->sdl_nlen = namelen;
301 	sdl->sdl_alen = ifp->if_addrlen;
302 	sdl->sdl_index = ifp->if_index;
303 	sdl->sdl_type = ifp->if_type;
304 	ifnet_addrs[ifp->if_index] = ifa;
305 	IFAREF(ifa);
306 	ifa->ifa_ifp = ifp;
307 	ifa->ifa_rtrequest = link_rtrequest;
308 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
309 	IFAREF(ifa);
310 	ifa->ifa_addr = (struct sockaddr *)sdl;
311 	ifp->if_sadl = sdl;
312 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
313 	ifa->ifa_netmask = (struct sockaddr *)sdl;
314 	sdl->sdl_len = masklen;
315 	while (namelen != 0)
316 		sdl->sdl_data[--namelen] = 0xff;
317 }
318 
319 /*
320  * Free the link level name for the specified interface.  This is
321  * a detach helper.  This is called from if_detach() or from
322  * link layer type specific detach functions.
323  */
324 void
325 if_free_sadl(struct ifnet *ifp)
326 {
327 	struct ifaddr *ifa;
328 	int s;
329 
330 	ifa = ifnet_addrs[ifp->if_index];
331 	if (ifa == NULL) {
332 		KASSERT(ifp->if_sadl == NULL);
333 		return;
334 	}
335 
336 	KASSERT(ifp->if_sadl != NULL);
337 
338 	s = splnet();
339 	rtinit(ifa, RTM_DELETE, 0);
340 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
341 	IFAFREE(ifa);
342 
343 	ifp->if_sadl = NULL;
344 
345 	ifnet_addrs[ifp->if_index] = NULL;
346 	IFAFREE(ifa);
347 	splx(s);
348 }
349 
350 /*
351  * Attach an interface to the
352  * list of "active" interfaces.
353  */
354 void
355 if_attach(ifp)
356 	struct ifnet *ifp;
357 {
358 	static size_t if_indexlim = 0;
359 	int indexlim = 0;
360 
361 	if (if_indexlim == 0) {
362 		TAILQ_INIT(&ifnet);
363 		if_indexlim = 8;
364 	}
365 	TAILQ_INIT(&ifp->if_addrlist);
366 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
367 	ifp->if_index = if_index;
368 	if (ifindex2ifnet == 0)
369 		if_index++;
370 	else
371 		while (ifindex2ifnet[ifp->if_index] != NULL) {
372 			++if_index;
373 			if (if_index == 0)
374 				if_index = 1;
375 			/*
376 			 * If we hit USHRT_MAX, we skip back to 0 since
377 			 * there are a number of places where the value
378 			 * of if_index or if_index itself is compared
379 			 * to or stored in an unsigned short.  By
380 			 * jumping back, we won't botch those assignments
381 			 * or comparisons.
382 			 */
383 			else if (if_index == USHRT_MAX) {
384 				/*
385 				 * However, if we have to jump back to
386 				 * zero *twice* without finding an empty
387 				 * slot in ifindex2ifnet[], then there
388 				 * there are too many (>65535) interfaces.
389 				 */
390 				if (indexlim++)
391 					panic("too many interfaces");
392 				else
393 					if_index = 1;
394 			}
395 			ifp->if_index = if_index;
396 		}
397 
398 	/*
399 	 * We have some arrays that should be indexed by if_index.
400 	 * since if_index will grow dynamically, they should grow too.
401 	 *	struct ifadd **ifnet_addrs
402 	 *	struct ifnet **ifindex2ifnet
403 	 */
404 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
405 	    ifp->if_index >= if_indexlim) {
406 		size_t n;
407 		caddr_t q;
408 
409 		while (ifp->if_index >= if_indexlim)
410 			if_indexlim <<= 1;
411 
412 		/* grow ifnet_addrs */
413 		n = if_indexlim * sizeof(struct ifaddr *);
414 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
415 		memset(q, 0, n);
416 		if (ifnet_addrs) {
417 			bcopy((caddr_t)ifnet_addrs, q, n/2);
418 			free((caddr_t)ifnet_addrs, M_IFADDR);
419 		}
420 		ifnet_addrs = (struct ifaddr **)q;
421 
422 		/* grow ifindex2ifnet */
423 		n = if_indexlim * sizeof(struct ifnet *);
424 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
425 		memset(q, 0, n);
426 		if (ifindex2ifnet) {
427 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
428 			free((caddr_t)ifindex2ifnet, M_IFADDR);
429 		}
430 		ifindex2ifnet = (struct ifnet **)q;
431 	}
432 
433 	ifindex2ifnet[ifp->if_index] = ifp;
434 
435 	/*
436 	 * Link level name is allocated later by a separate call to
437 	 * if_alloc_sadl().
438 	 */
439 
440 	if (ifp->if_snd.ifq_maxlen == 0)
441 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
442 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
443 
444 	ifp->if_link_state = LINK_STATE_UNKNOWN;
445 
446 	ifp->if_capenable = 0;
447 	ifp->if_csum_flags_tx = 0;
448 	ifp->if_csum_flags_rx = 0;
449 
450 #ifdef ALTQ
451 	ifp->if_snd.altq_type = 0;
452 	ifp->if_snd.altq_disc = NULL;
453 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
454 	ifp->if_snd.altq_tbr  = NULL;
455 	ifp->if_snd.altq_ifp  = ifp;
456 #endif
457 
458 #ifdef PFIL_HOOKS
459 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
460 	ifp->if_pfil.ph_ifnet = ifp;
461 	if (pfil_head_register(&ifp->if_pfil) != 0)
462 		printf("%s: WARNING: unable to register pfil hook\n",
463 		    ifp->if_xname);
464 #endif
465 
466 	if (domains)
467 		if_attachdomain1(ifp);
468 
469 	/* Announce the interface. */
470 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
471 }
472 
473 void
474 if_attachdomain()
475 {
476 	struct ifnet *ifp;
477 	int s;
478 
479 	s = splnet();
480 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
481 		if_attachdomain1(ifp);
482 	splx(s);
483 }
484 
485 void
486 if_attachdomain1(ifp)
487 	struct ifnet *ifp;
488 {
489 	struct domain *dp;
490 	int s;
491 
492 	s = splnet();
493 
494 	/* address family dependent data region */
495 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
496 	for (dp = domains; dp; dp = dp->dom_next) {
497 		if (dp->dom_ifattach)
498 			ifp->if_afdata[dp->dom_family] =
499 			    (*dp->dom_ifattach)(ifp);
500 	}
501 
502 	splx(s);
503 }
504 
505 /*
506  * Deactivate an interface.  This points all of the procedure
507  * handles at error stubs.  May be called from interrupt context.
508  */
509 void
510 if_deactivate(ifp)
511 	struct ifnet *ifp;
512 {
513 	int s;
514 
515 	s = splnet();
516 
517 	ifp->if_output	 = if_nulloutput;
518 	ifp->if_input	 = if_nullinput;
519 	ifp->if_start	 = if_nullstart;
520 	ifp->if_ioctl	 = if_nullioctl;
521 	ifp->if_init	 = if_nullinit;
522 	ifp->if_stop	 = if_nullstop;
523 	ifp->if_watchdog = if_nullwatchdog;
524 	ifp->if_drain	 = if_nulldrain;
525 
526 	/* No more packets may be enqueued. */
527 	ifp->if_snd.ifq_maxlen = 0;
528 
529 	splx(s);
530 }
531 
532 /*
533  * Detach an interface from the list of "active" interfaces,
534  * freeing any resources as we go along.
535  *
536  * NOTE: This routine must be called with a valid thread context,
537  * as it may block.
538  */
539 void
540 if_detach(ifp)
541 	struct ifnet *ifp;
542 {
543 	struct socket so;
544 	struct ifaddr *ifa, *next;
545 #ifdef IFAREF_DEBUG
546 	struct ifaddr *last_ifa = NULL;
547 #endif
548 	struct domain *dp;
549 	struct protosw *pr;
550 	struct radix_node_head *rnh;
551 	int s, i, family, purged;
552 
553 	/*
554 	 * XXX It's kind of lame that we have to have the
555 	 * XXX socket structure...
556 	 */
557 	memset(&so, 0, sizeof(so));
558 
559 	s = splnet();
560 
561 	/*
562 	 * Do an if_down() to give protocols a chance to do something.
563 	 */
564 	if_down(ifp);
565 
566 #ifdef ALTQ
567 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
568 		altq_disable(&ifp->if_snd);
569 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
570 		altq_detach(&ifp->if_snd);
571 #endif
572 
573 #ifdef PFIL_HOOKS
574 	(void) pfil_head_unregister(&ifp->if_pfil);
575 #endif
576 
577 	/*
578 	 * Rip all the addresses off the interface.  This should make
579 	 * all of the routes go away.
580 	 */
581 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa; ifa = next) {
582 		next = TAILQ_NEXT(ifa, ifa_list);
583 		family = ifa->ifa_addr->sa_family;
584 #ifdef IFAREF_DEBUG
585 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
586 		    ifa, family, ifa->ifa_refcnt);
587 		if (last_ifa != NULL && ifa == last_ifa)
588 			panic("if_detach: loop detected");
589 		last_ifa = ifa;
590 #endif
591 		if (family == AF_LINK)
592 			continue;
593 		dp = pffinddomain(family);
594 #ifdef DIAGNOSTIC
595 		if (dp == NULL)
596 			panic("if_detach: no domain for AF %d",
597 			    family);
598 #endif
599 		purged = 0;
600 		for (pr = dp->dom_protosw;
601 		     pr < dp->dom_protoswNPROTOSW; pr++) {
602 			so.so_proto = pr;
603 			if (pr->pr_usrreq != NULL) {
604 				(void) (*pr->pr_usrreq)(&so,
605 				    PRU_PURGEIF, NULL, NULL,
606 				    (struct mbuf *) ifp, curproc);
607 				purged = 1;
608 			}
609 		}
610 		if (purged == 0) {
611 			/*
612 			 * XXX What's really the best thing to do
613 			 * XXX here?  --thorpej@netbsd.org
614 			 */
615 			printf("if_detach: WARNING: AF %d not purged\n",
616 			    family);
617 		}
618 	}
619 
620 	if_free_sadl(ifp);
621 
622 	/* Walk the routing table looking for straglers. */
623 	for (i = 0; i <= AF_MAX; i++) {
624 		if ((rnh = rt_tables[i]) != NULL)
625 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
626 	}
627 
628 	for (dp = domains; dp; dp = dp->dom_next) {
629 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
630 			(*dp->dom_ifdetach)(ifp,
631 			    ifp->if_afdata[dp->dom_family]);
632 	}
633 
634 	/* Announce that the interface is gone. */
635 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
636 
637 	ifindex2ifnet[ifp->if_index] = NULL;
638 
639 	TAILQ_REMOVE(&ifnet, ifp, if_list);
640 
641 	/*
642 	 * remove packets came from ifp, from software interrupt queues.
643 	 * net/netisr_dispatch.h is not usable, as some of them use
644 	 * strange queue names.
645 	 */
646 #define IF_DETACH_QUEUES(x) \
647 do { \
648 	extern struct ifqueue x; \
649 	if_detach_queues(ifp, & x); \
650 } while (/*CONSTCOND*/ 0)
651 #ifdef INET
652 #if NARP > 0
653 	IF_DETACH_QUEUES(arpintrq);
654 #endif
655 	IF_DETACH_QUEUES(ipintrq);
656 #endif
657 #ifdef INET6
658 	IF_DETACH_QUEUES(ip6intrq);
659 #endif
660 #ifdef NETATALK
661 	IF_DETACH_QUEUES(atintrq1);
662 	IF_DETACH_QUEUES(atintrq2);
663 #endif
664 #ifdef NS
665 	IF_DETACH_QUEUES(nsintrq);
666 #endif
667 #ifdef ISO
668 	IF_DETACH_QUEUES(clnlintrq);
669 #endif
670 #ifdef CCITT
671 	IF_DETACH_QUEUES(llcintrq);
672 	IF_DETACH_QUEUES(hdintrq);
673 #endif
674 #ifdef NATM
675 	IF_DETACH_QUEUES(natmintrq);
676 #endif
677 #undef IF_DETACH_QUEUES
678 
679 	splx(s);
680 }
681 
682 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
683     defined(ISO) || defined(CCITT) || defined(NATM)
684 static void
685 if_detach_queues(ifp, q)
686 	struct ifnet *ifp;
687 	struct ifqueue *q;
688 {
689 	struct mbuf *m, *prev, *next;
690 
691 	prev = NULL;
692 	for (m = q->ifq_head; m; m = next) {
693 		next = m->m_nextpkt;
694 #ifdef DIAGNOSTIC
695 		if ((m->m_flags & M_PKTHDR) == 0) {
696 			prev = m;
697 			continue;
698 		}
699 #endif
700 		if (m->m_pkthdr.rcvif != ifp) {
701 			prev = m;
702 			continue;
703 		}
704 
705 		if (prev)
706 			prev->m_nextpkt = m->m_nextpkt;
707 		else
708 			q->ifq_head = m->m_nextpkt;
709 		if (q->ifq_tail == m)
710 			q->ifq_tail = prev;
711 		q->ifq_len--;
712 
713 		m->m_nextpkt = NULL;
714 		m_freem(m);
715 		IF_DROP(q);
716 	}
717 }
718 #endif /* defined(INET) || ... */
719 
720 /*
721  * Callback for a radix tree walk to delete all references to an
722  * ifnet.
723  */
724 int
725 if_rt_walktree(rn, v)
726 	struct radix_node *rn;
727 	void *v;
728 {
729 	struct ifnet *ifp = (struct ifnet *)v;
730 	struct rtentry *rt = (struct rtentry *)rn;
731 	int error;
732 
733 	if (rt->rt_ifp == ifp) {
734 		/* Delete the entry. */
735 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
736 		    rt_mask(rt), rt->rt_flags, NULL);
737 		if (error)
738 			printf("%s: warning: unable to delete rtentry @ %p, "
739 			    "error = %d\n", ifp->if_xname, rt, error);
740 	}
741 	return (0);
742 }
743 
744 /*
745  * Create a clone network interface.
746  */
747 int
748 if_clone_create(name)
749 	const char *name;
750 {
751 	struct if_clone *ifc;
752 	int unit;
753 
754 	ifc = if_clone_lookup(name, &unit);
755 	if (ifc == NULL)
756 		return (EINVAL);
757 
758 	if (ifunit(name) != NULL)
759 		return (EEXIST);
760 
761 	return ((*ifc->ifc_create)(ifc, unit));
762 }
763 
764 /*
765  * Destroy a clone network interface.
766  */
767 int
768 if_clone_destroy(name)
769 	const char *name;
770 {
771 	struct if_clone *ifc;
772 	struct ifnet *ifp;
773 
774 	ifc = if_clone_lookup(name, NULL);
775 	if (ifc == NULL)
776 		return (EINVAL);
777 
778 	ifp = ifunit(name);
779 	if (ifp == NULL)
780 		return (ENXIO);
781 
782 	if (ifc->ifc_destroy == NULL)
783 		return (EOPNOTSUPP);
784 
785 	(*ifc->ifc_destroy)(ifp);
786 	return (0);
787 }
788 
789 /*
790  * Look up a network interface cloner.
791  */
792 struct if_clone *
793 if_clone_lookup(name, unitp)
794 	const char *name;
795 	int *unitp;
796 {
797 	struct if_clone *ifc;
798 	const char *cp;
799 	size_t i;
800 
801 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
802 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
803 			if (ifc->ifc_name[i] != *cp)
804 				goto next_ifc;
805 		}
806 		goto found_name;
807  next_ifc:
808 		ifc = LIST_NEXT(ifc, ifc_list);
809 	}
810 
811 	/* No match. */
812 	return (NULL);
813 
814  found_name:
815 	for (i = 0; *cp != '\0'; cp++) {
816 		if (*cp < '0' || *cp > '9') {
817 			/* Bogus unit number. */
818 			return (NULL);
819 		}
820 		i = (i * 10) + (*cp - '0');
821 	}
822 
823 	if (unitp != NULL)
824 		*unitp = i;
825 	return (ifc);
826 }
827 
828 /*
829  * Register a network interface cloner.
830  */
831 void
832 if_clone_attach(ifc)
833 	struct if_clone *ifc;
834 {
835 
836 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
837 	if_cloners_count++;
838 }
839 
840 /*
841  * Unregister a network interface cloner.
842  */
843 void
844 if_clone_detach(ifc)
845 	struct if_clone *ifc;
846 {
847 
848 	LIST_REMOVE(ifc, ifc_list);
849 	if_cloners_count--;
850 }
851 
852 /*
853  * Provide list of interface cloners to userspace.
854  */
855 int
856 if_clone_list(ifcr)
857 	struct if_clonereq *ifcr;
858 {
859 	char outbuf[IFNAMSIZ], *dst;
860 	struct if_clone *ifc;
861 	int count, error = 0;
862 
863 	ifcr->ifcr_total = if_cloners_count;
864 	if ((dst = ifcr->ifcr_buffer) == NULL) {
865 		/* Just asking how many there are. */
866 		return (0);
867 	}
868 
869 	if (ifcr->ifcr_count < 0)
870 		return (EINVAL);
871 
872 	count = (if_cloners_count < ifcr->ifcr_count) ?
873 	    if_cloners_count : ifcr->ifcr_count;
874 
875 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
876 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
877 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
878 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
879 		error = copyout(outbuf, dst, IFNAMSIZ);
880 		if (error)
881 			break;
882 	}
883 
884 	return (error);
885 }
886 
887 /*
888  * Locate an interface based on a complete address.
889  */
890 /*ARGSUSED*/
891 struct ifaddr *
892 ifa_ifwithaddr(addr)
893 	struct sockaddr *addr;
894 {
895 	struct ifnet *ifp;
896 	struct ifaddr *ifa;
897 
898 #define	equal(a1, a2) \
899   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
900 
901 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
902 	     ifp = TAILQ_NEXT(ifp, if_list)) {
903 		if (ifp->if_output == if_nulloutput)
904 			continue;
905 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
906 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
907 			if (ifa->ifa_addr->sa_family != addr->sa_family)
908 				continue;
909 			if (equal(addr, ifa->ifa_addr))
910 				return (ifa);
911 			if ((ifp->if_flags & IFF_BROADCAST) &&
912 			    ifa->ifa_broadaddr &&
913 			    /* IP6 doesn't have broadcast */
914 			    ifa->ifa_broadaddr->sa_len != 0 &&
915 			    equal(ifa->ifa_broadaddr, addr))
916 				return (ifa);
917 		}
918 	}
919 	return (NULL);
920 }
921 
922 /*
923  * Locate the point to point interface with a given destination address.
924  */
925 /*ARGSUSED*/
926 struct ifaddr *
927 ifa_ifwithdstaddr(addr)
928 	struct sockaddr *addr;
929 {
930 	struct ifnet *ifp;
931 	struct ifaddr *ifa;
932 
933 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
934 	     ifp = TAILQ_NEXT(ifp, if_list)) {
935 		if (ifp->if_output == if_nulloutput)
936 			continue;
937 		if (ifp->if_flags & IFF_POINTOPOINT) {
938 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
939 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
940 				if (ifa->ifa_addr->sa_family !=
941 				      addr->sa_family ||
942 				    ifa->ifa_dstaddr == NULL)
943 					continue;
944 				if (equal(addr, ifa->ifa_dstaddr))
945 					return (ifa);
946 			}
947 		}
948 	}
949 	return (NULL);
950 }
951 
952 /*
953  * Find an interface on a specific network.  If many, choice
954  * is most specific found.
955  */
956 struct ifaddr *
957 ifa_ifwithnet(addr)
958 	struct sockaddr *addr;
959 {
960 	struct ifnet *ifp;
961 	struct ifaddr *ifa;
962 	struct sockaddr_dl *sdl;
963 	struct ifaddr *ifa_maybe = 0;
964 	u_int af = addr->sa_family;
965 	char *addr_data = addr->sa_data, *cplim;
966 
967 	if (af == AF_LINK) {
968 		sdl = (struct sockaddr_dl *)addr;
969 		if (sdl->sdl_index && sdl->sdl_index <= if_index &&
970 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
971 			return (ifnet_addrs[sdl->sdl_index]);
972 	}
973 #ifdef NETATALK
974 	if (af == AF_APPLETALK) {
975 		struct sockaddr_at *sat, *sat2;
976 		sat = (struct sockaddr_at *)addr;
977 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
978 		     ifp = TAILQ_NEXT(ifp, if_list)) {
979 			if (ifp->if_output == if_nulloutput)
980 				continue;
981 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
982 			if (ifa == NULL)
983 				continue;
984 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
985 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
986 				return (ifa); /* exact match */
987 			if (ifa_maybe == NULL) {
988 				/* else keep the if with the right range */
989 				ifa_maybe = ifa;
990 			}
991 		}
992 		return (ifa_maybe);
993 	}
994 #endif
995 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
996 	     ifp = TAILQ_NEXT(ifp, if_list)) {
997 		if (ifp->if_output == if_nulloutput)
998 			continue;
999 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1000 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1001 			char *cp, *cp2, *cp3;
1002 
1003 			if (ifa->ifa_addr->sa_family != af ||
1004 			    ifa->ifa_netmask == 0)
1005  next:				continue;
1006 			cp = addr_data;
1007 			cp2 = ifa->ifa_addr->sa_data;
1008 			cp3 = ifa->ifa_netmask->sa_data;
1009 			cplim = (char *)ifa->ifa_netmask +
1010 			    ifa->ifa_netmask->sa_len;
1011 			while (cp3 < cplim) {
1012 				if ((*cp++ ^ *cp2++) & *cp3++) {
1013 					/* want to continue for() loop */
1014 					goto next;
1015 				}
1016 			}
1017 			if (ifa_maybe == 0 ||
1018 			    rn_refines((caddr_t)ifa->ifa_netmask,
1019 			    (caddr_t)ifa_maybe->ifa_netmask))
1020 				ifa_maybe = ifa;
1021 		}
1022 	}
1023 	return (ifa_maybe);
1024 }
1025 
1026 /*
1027  * Find the interface of the addresss.
1028  */
1029 struct ifaddr *
1030 ifa_ifwithladdr(addr)
1031 	struct sockaddr *addr;
1032 {
1033 	struct ifaddr *ia;
1034 
1035 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1036 	    (ia = ifa_ifwithnet(addr)))
1037 		return (ia);
1038 	return (NULL);
1039 }
1040 
1041 /*
1042  * Find an interface using a specific address family
1043  */
1044 struct ifaddr *
1045 ifa_ifwithaf(af)
1046 	int af;
1047 {
1048 	struct ifnet *ifp;
1049 	struct ifaddr *ifa;
1050 
1051 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1052 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1053 		if (ifp->if_output == if_nulloutput)
1054 			continue;
1055 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1056 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1057 			if (ifa->ifa_addr->sa_family == af)
1058 				return (ifa);
1059 		}
1060 	}
1061 	return (NULL);
1062 }
1063 
1064 /*
1065  * Find an interface address specific to an interface best matching
1066  * a given address.
1067  */
1068 struct ifaddr *
1069 ifaof_ifpforaddr(addr, ifp)
1070 	struct sockaddr *addr;
1071 	struct ifnet *ifp;
1072 {
1073 	struct ifaddr *ifa;
1074 	char *cp, *cp2, *cp3;
1075 	char *cplim;
1076 	struct ifaddr *ifa_maybe = 0;
1077 	u_int af = addr->sa_family;
1078 
1079 	if (ifp->if_output == if_nulloutput)
1080 		return (NULL);
1081 
1082 	if (af >= AF_MAX)
1083 		return (NULL);
1084 
1085 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1086 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1087 		if (ifa->ifa_addr->sa_family != af)
1088 			continue;
1089 		ifa_maybe = ifa;
1090 		if (ifa->ifa_netmask == 0) {
1091 			if (equal(addr, ifa->ifa_addr) ||
1092 			    (ifa->ifa_dstaddr &&
1093 			     equal(addr, ifa->ifa_dstaddr)))
1094 				return (ifa);
1095 			continue;
1096 		}
1097 		cp = addr->sa_data;
1098 		cp2 = ifa->ifa_addr->sa_data;
1099 		cp3 = ifa->ifa_netmask->sa_data;
1100 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1101 		for (; cp3 < cplim; cp3++) {
1102 			if ((*cp++ ^ *cp2++) & *cp3)
1103 				break;
1104 		}
1105 		if (cp3 == cplim)
1106 			return (ifa);
1107 	}
1108 	return (ifa_maybe);
1109 }
1110 
1111 /*
1112  * Default action when installing a route with a Link Level gateway.
1113  * Lookup an appropriate real ifa to point to.
1114  * This should be moved to /sys/net/link.c eventually.
1115  */
1116 void
1117 link_rtrequest(cmd, rt, info)
1118 	int cmd;
1119 	struct rtentry *rt;
1120 	struct rt_addrinfo *info;
1121 {
1122 	struct ifaddr *ifa;
1123 	struct sockaddr *dst;
1124 	struct ifnet *ifp;
1125 
1126 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1127 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1128 		return;
1129 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1130 		IFAFREE(rt->rt_ifa);
1131 		rt->rt_ifa = ifa;
1132 		IFAREF(ifa);
1133 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1134 			ifa->ifa_rtrequest(cmd, rt, info);
1135 	}
1136 }
1137 
1138 /*
1139  * Mark an interface down and notify protocols of
1140  * the transition.
1141  * NOTE: must be called at splsoftnet or equivalent.
1142  */
1143 void
1144 if_down(ifp)
1145 	struct ifnet *ifp;
1146 {
1147 	struct ifaddr *ifa;
1148 
1149 	ifp->if_flags &= ~IFF_UP;
1150 	microtime(&ifp->if_lastchange);
1151 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1152 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1153 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1154 	IFQ_PURGE(&ifp->if_snd);
1155 	rt_ifmsg(ifp);
1156 }
1157 
1158 /*
1159  * Mark an interface up and notify protocols of
1160  * the transition.
1161  * NOTE: must be called at splsoftnet or equivalent.
1162  */
1163 void
1164 if_up(ifp)
1165 	struct ifnet *ifp;
1166 {
1167 #ifdef notyet
1168 	struct ifaddr *ifa;
1169 #endif
1170 
1171 	ifp->if_flags |= IFF_UP;
1172 	microtime(&ifp->if_lastchange);
1173 #ifdef notyet
1174 	/* this has no effect on IP, and will kill all ISO connections XXX */
1175 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1176 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1177 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1178 #endif
1179 	rt_ifmsg(ifp);
1180 #ifdef INET6
1181 	in6_if_up(ifp);
1182 #endif
1183 }
1184 
1185 /*
1186  * Handle interface watchdog timer routines.  Called
1187  * from softclock, we decrement timers (if set) and
1188  * call the appropriate interface routine on expiration.
1189  */
1190 void
1191 if_slowtimo(arg)
1192 	void *arg;
1193 {
1194 	struct ifnet *ifp;
1195 	int s = splnet();
1196 
1197 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1198 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1199 		if (ifp->if_timer == 0 || --ifp->if_timer)
1200 			continue;
1201 		if (ifp->if_watchdog)
1202 			(*ifp->if_watchdog)(ifp);
1203 	}
1204 	splx(s);
1205 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1206 	    if_slowtimo, NULL);
1207 }
1208 
1209 /*
1210  * Set/clear promiscuous mode on interface ifp based on the truth value
1211  * of pswitch.  The calls are reference counted so that only the first
1212  * "on" request actually has an effect, as does the final "off" request.
1213  * Results are undefined if the "off" and "on" requests are not matched.
1214  */
1215 int
1216 ifpromisc(ifp, pswitch)
1217 	struct ifnet *ifp;
1218 	int pswitch;
1219 {
1220 	int pcount, ret;
1221 	short flags;
1222 	struct ifreq ifr;
1223 
1224 	pcount = ifp->if_pcount;
1225 	flags = ifp->if_flags;
1226 	if (pswitch) {
1227 		/*
1228 		 * Allow the device to be "placed" into promiscuous
1229 		 * mode even if it is not configured up.  It will
1230 		 * consult IFF_PROMISC when it is is brought up.
1231 		 */
1232 		if (ifp->if_pcount++ != 0)
1233 			return (0);
1234 		ifp->if_flags |= IFF_PROMISC;
1235 		if ((ifp->if_flags & IFF_UP) == 0)
1236 			return (0);
1237 	} else {
1238 		if (--ifp->if_pcount > 0)
1239 			return (0);
1240 		ifp->if_flags &= ~IFF_PROMISC;
1241 		/*
1242 		 * If the device is not configured up, we should not need to
1243 		 * turn off promiscuous mode (device should have turned it
1244 		 * off when interface went down; and will look at IFF_PROMISC
1245 		 * again next time interface comes up).
1246 		 */
1247 		if ((ifp->if_flags & IFF_UP) == 0)
1248 			return (0);
1249 	}
1250 	memset(&ifr, 0, sizeof(ifr));
1251 	ifr.ifr_flags = ifp->if_flags;
1252 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1253 	/* Restore interface state if not successful. */
1254 	if (ret != 0) {
1255 		ifp->if_pcount = pcount;
1256 		ifp->if_flags = flags;
1257 	}
1258 	return (ret);
1259 }
1260 
1261 /*
1262  * Map interface name to
1263  * interface structure pointer.
1264  */
1265 struct ifnet *
1266 ifunit(name)
1267 	const char *name;
1268 {
1269 	struct ifnet *ifp;
1270 	const char *cp = name;
1271 	u_int unit = 0;
1272 	u_int i;
1273 
1274 	/*
1275 	 * If the entire name is a number, treat it as an ifindex.
1276 	 */
1277 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1278 		unit = unit * 10 + (*cp - '0');
1279 	}
1280 
1281 	/*
1282 	 * If the number took all of the name, then it's a valid ifindex.
1283 	 */
1284 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1285 		if (unit >= if_index)
1286 			return (NULL);
1287 		ifp = ifindex2ifnet[unit];
1288 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1289 			return (NULL);
1290 		return (ifp);
1291 	}
1292 
1293 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1294 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1295 		if (ifp->if_output == if_nulloutput)
1296 			continue;
1297 	 	if (strcmp(ifp->if_xname, name) == 0)
1298 			return (ifp);
1299 	}
1300 	return (NULL);
1301 }
1302 
1303 /*
1304  * Interface ioctls.
1305  */
1306 int
1307 ifioctl(so, cmd, data, p)
1308 	struct socket *so;
1309 	u_long cmd;
1310 	caddr_t data;
1311 	struct proc *p;
1312 {
1313 	struct ifnet *ifp;
1314 	struct ifreq *ifr;
1315 	struct ifcapreq *ifcr;
1316 	struct ifdatareq *ifdr;
1317 	int s, error = 0;
1318 	short oif_flags;
1319 
1320 	switch (cmd) {
1321 
1322 	case SIOCGIFCONF:
1323 	case OSIOCGIFCONF:
1324 		return (ifconf(cmd, data));
1325 	}
1326 	ifr = (struct ifreq *)data;
1327 	ifcr = (struct ifcapreq *)data;
1328 	ifdr = (struct ifdatareq *)data;
1329 
1330 	switch (cmd) {
1331 	case SIOCIFCREATE:
1332 	case SIOCIFDESTROY:
1333 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1334 			return (error);
1335 		return ((cmd == SIOCIFCREATE) ?
1336 			if_clone_create(ifr->ifr_name) :
1337 			if_clone_destroy(ifr->ifr_name));
1338 
1339 	case SIOCIFGCLONERS:
1340 		return (if_clone_list((struct if_clonereq *)data));
1341 	}
1342 
1343 	ifp = ifunit(ifr->ifr_name);
1344 	if (ifp == 0)
1345 		return (ENXIO);
1346 	oif_flags = ifp->if_flags;
1347 	switch (cmd) {
1348 
1349 	case SIOCGIFFLAGS:
1350 		ifr->ifr_flags = ifp->if_flags;
1351 		break;
1352 
1353 	case SIOCGIFMETRIC:
1354 		ifr->ifr_metric = ifp->if_metric;
1355 		break;
1356 
1357 	case SIOCGIFMTU:
1358 		ifr->ifr_mtu = ifp->if_mtu;
1359 		break;
1360 
1361 	case SIOCGIFDLT:
1362 		ifr->ifr_dlt = ifp->if_dlt;
1363 		break;
1364 
1365 	case SIOCSIFFLAGS:
1366 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1367 			return (error);
1368 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1369 			s = splnet();
1370 			if_down(ifp);
1371 			splx(s);
1372 		}
1373 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1374 			s = splnet();
1375 			if_up(ifp);
1376 			splx(s);
1377 		}
1378 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1379 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1380 		if (ifp->if_ioctl)
1381 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1382 		break;
1383 
1384 	case SIOCGIFCAP:
1385 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1386 		ifcr->ifcr_capenable = ifp->if_capenable;
1387 		break;
1388 
1389 	case SIOCSIFCAP:
1390 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1391 			return (error);
1392 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1393 			return (EINVAL);
1394 		if (ifp->if_ioctl == NULL)
1395 			return (EOPNOTSUPP);
1396 
1397 		/* Must prevent race with packet reception here. */
1398 		s = splnet();
1399 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1400 			struct ifreq ifrq;
1401 
1402 			ifrq.ifr_flags = ifp->if_flags;
1403 			ifp->if_capenable = ifcr->ifcr_capenable;
1404 
1405 			/* Pre-compute the checksum flags mask. */
1406 			ifp->if_csum_flags_tx = 0;
1407 			ifp->if_csum_flags_rx = 0;
1408 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1409 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1410 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1411 			}
1412 
1413 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1414 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1415 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1416 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1417 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1418 
1419 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1420 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1421 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1422 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1423 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1424 
1425 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1426 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1427 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1428 			}
1429 
1430 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1431 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1432 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1433 			}
1434 
1435 			/*
1436 			 * Only kick the interface if it's up.  If it's
1437 			 * not up now, it will notice the cap enables
1438 			 * when it is brought up later.
1439 			 */
1440 			if (ifp->if_flags & IFF_UP)
1441 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1442 				    (caddr_t) &ifrq);
1443 		}
1444 		splx(s);
1445 		break;
1446 
1447 	case SIOCSIFMETRIC:
1448 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1449 			return (error);
1450 		ifp->if_metric = ifr->ifr_metric;
1451 		break;
1452 
1453 	case SIOCGIFDATA:
1454 		ifdr->ifdr_data = ifp->if_data;
1455 		break;
1456 
1457 	case SIOCZIFDATA:
1458 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1459 			return (error);
1460 		ifdr->ifdr_data = ifp->if_data;
1461 		/*
1462 		 * Assumes that the volatile counters that can be
1463 		 * zero'ed are at the end of if_data.
1464 		 */
1465 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1466 		    offsetof(struct if_data, ifi_ipackets));
1467 		break;
1468 
1469 	case SIOCSIFMTU:
1470 	{
1471 		u_long oldmtu = ifp->if_mtu;
1472 
1473 		error = suser(p->p_ucred, &p->p_acflag);
1474 		if (error)
1475 			return (error);
1476 		if (ifp->if_ioctl == NULL)
1477 			return (EOPNOTSUPP);
1478 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1479 
1480 		/*
1481 		 * If the link MTU changed, do network layer specific procedure.
1482 		 */
1483 		if (ifp->if_mtu != oldmtu) {
1484 #ifdef INET6
1485 			nd6_setmtu(ifp);
1486 #endif
1487 		}
1488 		break;
1489 	}
1490 	case SIOCSIFPHYADDR:
1491 	case SIOCDIFPHYADDR:
1492 #ifdef INET6
1493 	case SIOCSIFPHYADDR_IN6:
1494 #endif
1495 	case SIOCSLIFPHYADDR:
1496 	case SIOCADDMULTI:
1497 	case SIOCDELMULTI:
1498 	case SIOCSIFMEDIA:
1499 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1500 			return (error);
1501 		/* FALLTHROUGH */
1502 	case SIOCGIFPSRCADDR:
1503 	case SIOCGIFPDSTADDR:
1504 	case SIOCGLIFPHYADDR:
1505 	case SIOCGIFMEDIA:
1506 		if (ifp->if_ioctl == 0)
1507 			return (EOPNOTSUPP);
1508 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1509 		break;
1510 
1511 	case SIOCSDRVSPEC:
1512 	case SIOCS80211NWID:
1513 	case SIOCS80211NWKEY:
1514 	case SIOCS80211POWER:
1515 	case SIOCS80211BSSID:
1516 	case SIOCS80211CHANNEL:
1517 		/* XXX:  need to pass proc pointer through to driver... */
1518 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1519 			return (error);
1520 	/* FALLTHROUGH */
1521 	default:
1522 		if (so->so_proto == 0)
1523 			return (EOPNOTSUPP);
1524 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(COMPAT_ULTRIX) && !defined(LKM)
1525 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1526 		    (struct mbuf *)cmd, (struct mbuf *)data,
1527 		    (struct mbuf *)ifp, p));
1528 #else
1529 	    {
1530 		int ocmd = cmd;
1531 
1532 		switch (cmd) {
1533 
1534 		case SIOCSIFADDR:
1535 		case SIOCSIFDSTADDR:
1536 		case SIOCSIFBRDADDR:
1537 		case SIOCSIFNETMASK:
1538 #if BYTE_ORDER != BIG_ENDIAN
1539 			if (ifr->ifr_addr.sa_family == 0 &&
1540 			    ifr->ifr_addr.sa_len < 16) {
1541 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1542 				ifr->ifr_addr.sa_len = 16;
1543 			}
1544 #else
1545 			if (ifr->ifr_addr.sa_len == 0)
1546 				ifr->ifr_addr.sa_len = 16;
1547 #endif
1548 			break;
1549 
1550 		case OSIOCGIFADDR:
1551 			cmd = SIOCGIFADDR;
1552 			break;
1553 
1554 		case OSIOCGIFDSTADDR:
1555 			cmd = SIOCGIFDSTADDR;
1556 			break;
1557 
1558 		case OSIOCGIFBRDADDR:
1559 			cmd = SIOCGIFBRDADDR;
1560 			break;
1561 
1562 		case OSIOCGIFNETMASK:
1563 			cmd = SIOCGIFNETMASK;
1564 		}
1565 
1566 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1567 		    (struct mbuf *)cmd, (struct mbuf *)data,
1568 		    (struct mbuf *)ifp, p));
1569 
1570 		switch (ocmd) {
1571 		case OSIOCGIFADDR:
1572 		case OSIOCGIFDSTADDR:
1573 		case OSIOCGIFBRDADDR:
1574 		case OSIOCGIFNETMASK:
1575 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1576 		}
1577 	    }
1578 #endif /* COMPAT_43 */
1579 		break;
1580 	}
1581 
1582 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1583 #ifdef INET6
1584 		if ((ifp->if_flags & IFF_UP) != 0) {
1585 			s = splnet();
1586 			in6_if_up(ifp);
1587 			splx(s);
1588 		}
1589 #endif
1590 	}
1591 
1592 	return (error);
1593 }
1594 
1595 /*
1596  * Return interface configuration
1597  * of system.  List may be used
1598  * in later ioctl's (above) to get
1599  * other information.
1600  */
1601 /*ARGSUSED*/
1602 int
1603 ifconf(cmd, data)
1604 	u_long cmd;
1605 	caddr_t data;
1606 {
1607 	struct ifconf *ifc = (struct ifconf *)data;
1608 	struct ifnet *ifp;
1609 	struct ifaddr *ifa;
1610 	struct ifreq ifr, *ifrp;
1611 	int space = ifc->ifc_len, error = 0;
1612 
1613 	ifrp = ifc->ifc_req;
1614 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1615 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1616 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1617 			memset((caddr_t)&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1618 			if (space >= (int)sizeof (ifr)) {
1619 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1620 						sizeof(ifr));
1621 				if (error)
1622 					break;
1623 			}
1624 			space -= sizeof (ifr), ifrp++;
1625 		} else
1626 		    for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1627 			struct sockaddr *sa = ifa->ifa_addr;
1628 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX)
1629 			if (cmd == OSIOCGIFCONF) {
1630 				struct osockaddr *osa =
1631 					 (struct osockaddr *)&ifr.ifr_addr;
1632 				ifr.ifr_addr = *sa;
1633 				osa->sa_family = sa->sa_family;
1634 				if (space >= (int)sizeof (ifr)) {
1635 					error = copyout((caddr_t)&ifr,
1636 							(caddr_t)ifrp,
1637 							sizeof (ifr));
1638 					ifrp++;
1639 				}
1640 			} else
1641 #endif
1642 			if (sa->sa_len <= sizeof(*sa)) {
1643 				ifr.ifr_addr = *sa;
1644 				if (space >= (int)sizeof (ifr)) {
1645 					error = copyout((caddr_t)&ifr,
1646 							(caddr_t)ifrp,
1647 							sizeof (ifr));
1648 					ifrp++;
1649 				}
1650 			} else {
1651 				space -= sa->sa_len - sizeof(*sa);
1652 				if (space >= (int)sizeof (ifr)) {
1653 					error = copyout((caddr_t)&ifr,
1654 							(caddr_t)ifrp,
1655 							sizeof (ifr.ifr_name));
1656 					if (error == 0) {
1657 						error = copyout((caddr_t)sa,
1658 						  (caddr_t)&ifrp->ifr_addr,
1659 						  sa->sa_len);
1660 					}
1661 					ifrp = (struct ifreq *)
1662 						(sa->sa_len +
1663 						 (caddr_t)&ifrp->ifr_addr);
1664 				}
1665 			}
1666 			if (error)
1667 				break;
1668 			space -= sizeof (ifr);
1669 		}
1670 	}
1671 	ifc->ifc_len -= space;
1672 	return (error);
1673 }
1674