xref: /netbsd-src/sys/net/if.c (revision b8c616269f5ebf18ab2e35cb8099d683130a177c)
1 /*	$NetBSD: if.c,v 1.117 2003/02/01 06:23:46 thorpej 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 1.117 2003/02/01 06:23:46 thorpej Exp $");
105 
106 #include "opt_inet.h"
107 
108 #include "opt_compat_linux.h"
109 #include "opt_compat_svr4.h"
110 #include "opt_compat_43.h"
111 #include "opt_atalk.h"
112 #include "opt_pfil_hooks.h"
113 
114 #include <sys/param.h>
115 #include <sys/mbuf.h>
116 #include <sys/systm.h>
117 #include <sys/callout.h>
118 #include <sys/proc.h>
119 #include <sys/socket.h>
120 #include <sys/socketvar.h>
121 #include <sys/domain.h>
122 #include <sys/protosw.h>
123 #include <sys/kernel.h>
124 #include <sys/ioctl.h>
125 
126 #include <net/if.h>
127 #include <net/if_dl.h>
128 #include <net/if_ether.h>
129 #include <net/if_ieee80211.h>
130 #include <net/if_types.h>
131 #include <net/radix.h>
132 #include <net/route.h>
133 #include <net/netisr.h>
134 #ifdef NETATALK
135 #include <netatalk/at_extern.h>
136 #include <netatalk/at.h>
137 #endif
138 
139 #ifdef INET6
140 #include <netinet/in.h>
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
146 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
147 
148 int	ifqmaxlen = IFQ_MAXLEN;
149 struct	callout if_slowtimo_ch;
150 
151 int netisr;			/* scheduling bits for network */
152 
153 int if_rt_walktree __P((struct radix_node *, void *));
154 
155 struct if_clone *if_clone_lookup __P((const char *, int *));
156 int if_clone_list __P((struct if_clonereq *));
157 
158 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
159 int if_cloners_count;
160 
161 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
162     defined(ISO) || defined(CCITT) || defined(NATM)
163 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
164 #endif
165 
166 /*
167  * Network interface utility routines.
168  *
169  * Routines with ifa_ifwith* names take sockaddr *'s as
170  * parameters.
171  */
172 void
173 ifinit()
174 {
175 
176 	callout_init(&if_slowtimo_ch);
177 	if_slowtimo(NULL);
178 }
179 
180 /*
181  * Null routines used while an interface is going away.  These routines
182  * just return an error.
183  */
184 
185 int
186 if_nulloutput(ifp, m, so, rt)
187 	struct ifnet *ifp;
188 	struct mbuf *m;
189 	struct sockaddr *so;
190 	struct rtentry *rt;
191 {
192 
193 	return (ENXIO);
194 }
195 
196 void
197 if_nullinput(ifp, m)
198 	struct ifnet *ifp;
199 	struct mbuf *m;
200 {
201 
202 	/* Nothing. */
203 }
204 
205 void
206 if_nullstart(ifp)
207 	struct ifnet *ifp;
208 {
209 
210 	/* Nothing. */
211 }
212 
213 int
214 if_nullioctl(ifp, cmd, data)
215 	struct ifnet *ifp;
216 	u_long cmd;
217 	caddr_t data;
218 {
219 
220 	return (ENXIO);
221 }
222 
223 int
224 if_nullinit(ifp)
225 	struct ifnet *ifp;
226 {
227 
228 	return (ENXIO);
229 }
230 
231 void
232 if_nullstop(ifp, disable)
233 	struct ifnet *ifp;
234 	int disable;
235 {
236 
237 	/* Nothing. */
238 }
239 
240 void
241 if_nullwatchdog(ifp)
242 	struct ifnet *ifp;
243 {
244 
245 	/* Nothing. */
246 }
247 
248 void
249 if_nulldrain(ifp)
250 	struct ifnet *ifp;
251 {
252 
253 	/* Nothing. */
254 }
255 
256 u_int if_index = 1;
257 struct ifnet_head ifnet;
258 struct ifaddr **ifnet_addrs = NULL;
259 struct ifnet **ifindex2ifnet = NULL;
260 
261 /*
262  * Allocate the link level name for the specified interface.  This
263  * is an attachment helper.  It must be called after ifp->if_addrlen
264  * is initialized, which may not be the case when if_attach() is
265  * called.
266  */
267 void
268 if_alloc_sadl(struct ifnet *ifp)
269 {
270 	unsigned socksize, ifasize;
271 	int namelen, masklen;
272 	struct sockaddr_dl *sdl;
273 	struct ifaddr *ifa;
274 
275 	/*
276 	 * If the interface already has a link name, release it
277 	 * now.  This is useful for interfaces that can change
278 	 * link types, and thus switch link names often.
279 	 */
280 	if (ifp->if_sadl != NULL)
281 		if_free_sadl(ifp);
282 
283 	namelen = strlen(ifp->if_xname);
284 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
285 	socksize = masklen + ifp->if_addrlen;
286 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
287 	if (socksize < sizeof(*sdl))
288 		socksize = sizeof(*sdl);
289 	socksize = ROUNDUP(socksize);
290 	ifasize = sizeof(*ifa) + 2 * socksize;
291 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
292 	memset((caddr_t)ifa, 0, ifasize);
293 	sdl = (struct sockaddr_dl *)(ifa + 1);
294 	sdl->sdl_len = socksize;
295 	sdl->sdl_family = AF_LINK;
296 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
297 	sdl->sdl_nlen = namelen;
298 	sdl->sdl_alen = ifp->if_addrlen;
299 	sdl->sdl_index = ifp->if_index;
300 	sdl->sdl_type = ifp->if_type;
301 	ifnet_addrs[ifp->if_index] = ifa;
302 	IFAREF(ifa);
303 	ifa->ifa_ifp = ifp;
304 	ifa->ifa_rtrequest = link_rtrequest;
305 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
306 	IFAREF(ifa);
307 	ifa->ifa_addr = (struct sockaddr *)sdl;
308 	ifp->if_sadl = sdl;
309 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
310 	ifa->ifa_netmask = (struct sockaddr *)sdl;
311 	sdl->sdl_len = masklen;
312 	while (namelen != 0)
313 		sdl->sdl_data[--namelen] = 0xff;
314 }
315 
316 /*
317  * Free the link level name for the specified interface.  This is
318  * a detach helper.  This is called from if_detach() or from
319  * link layer type specific detach functions.
320  */
321 void
322 if_free_sadl(struct ifnet *ifp)
323 {
324 	struct ifaddr *ifa;
325 	int s;
326 
327 	ifa = ifnet_addrs[ifp->if_index];
328 	if (ifa == NULL) {
329 		KASSERT(ifp->if_sadl == NULL);
330 		return;
331 	}
332 
333 	KASSERT(ifp->if_sadl != NULL);
334 
335 	s = splnet();
336 	rtinit(ifa, RTM_DELETE, 0);
337 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
338 	IFAFREE(ifa);
339 
340 	ifp->if_sadl = NULL;
341 
342 	ifnet_addrs[ifp->if_index] = NULL;
343 	IFAFREE(ifa);
344 	splx(s);
345 }
346 
347 /*
348  * Attach an interface to the
349  * list of "active" interfaces.
350  */
351 void
352 if_attach(ifp)
353 	struct ifnet *ifp;
354 {
355 	static size_t if_indexlim = 0;
356 	int indexlim = 0;
357 
358 	if (if_indexlim == 0) {
359 		TAILQ_INIT(&ifnet);
360 		if_indexlim = 8;
361 	}
362 	TAILQ_INIT(&ifp->if_addrlist);
363 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
364 	ifp->if_index = if_index;
365 	if (ifindex2ifnet == 0)
366 		if_index++;
367 	else
368 		while (ifindex2ifnet[ifp->if_index] != NULL) {
369 			++if_index;
370 			if (if_index == 0)
371 				if_index = 1;
372 			/*
373 			 * If we hit USHRT_MAX, we skip back to 0 since
374 			 * there are a number of places where the value
375 			 * of if_index or if_index itself is compared
376 			 * to or stored in an unsigned short.  By
377 			 * jumping back, we won't botch those assignments
378 			 * or comparisons.
379 			 */
380 			else if (if_index == USHRT_MAX) {
381 				/*
382 				 * However, if we have to jump back to
383 				 * zero *twice* without finding an empty
384 				 * slot in ifindex2ifnet[], then there
385 				 * there are too many (>65535) interfaces.
386 				 */
387 				if (indexlim++)
388 					panic("too many interfaces");
389 				else
390 					if_index = 1;
391 			}
392 			ifp->if_index = if_index;
393 		}
394 
395 	/*
396 	 * We have some arrays that should be indexed by if_index.
397 	 * since if_index will grow dynamically, they should grow too.
398 	 *	struct ifadd **ifnet_addrs
399 	 *	struct ifnet **ifindex2ifnet
400 	 */
401 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
402 	    ifp->if_index >= if_indexlim) {
403 		size_t n;
404 		caddr_t q;
405 
406 		while (ifp->if_index >= if_indexlim)
407 			if_indexlim <<= 1;
408 
409 		/* grow ifnet_addrs */
410 		n = if_indexlim * sizeof(struct ifaddr *);
411 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
412 		memset(q, 0, n);
413 		if (ifnet_addrs) {
414 			bcopy((caddr_t)ifnet_addrs, q, n/2);
415 			free((caddr_t)ifnet_addrs, M_IFADDR);
416 		}
417 		ifnet_addrs = (struct ifaddr **)q;
418 
419 		/* grow ifindex2ifnet */
420 		n = if_indexlim * sizeof(struct ifnet *);
421 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
422 		memset(q, 0, n);
423 		if (ifindex2ifnet) {
424 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
425 			free((caddr_t)ifindex2ifnet, M_IFADDR);
426 		}
427 		ifindex2ifnet = (struct ifnet **)q;
428 	}
429 
430 	ifindex2ifnet[ifp->if_index] = ifp;
431 
432 	/*
433 	 * Link level name is allocated later by a separate call to
434 	 * if_alloc_sadl().
435 	 */
436 
437 	if (ifp->if_snd.ifq_maxlen == 0)
438 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
439 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
440 
441 	ifp->if_link_state = LINK_STATE_UNKNOWN;
442 
443 	ifp->if_capenable = 0;
444 	ifp->if_csum_flags_tx = 0;
445 	ifp->if_csum_flags_rx = 0;
446 
447 #ifdef ALTQ
448 	ifp->if_snd.altq_type = 0;
449 	ifp->if_snd.altq_disc = NULL;
450 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
451 	ifp->if_snd.altq_tbr  = NULL;
452 	ifp->if_snd.altq_ifp  = ifp;
453 #endif
454 
455 #ifdef PFIL_HOOKS
456 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
457 	ifp->if_pfil.ph_ifnet = ifp;
458 	if (pfil_head_register(&ifp->if_pfil) != 0)
459 		printf("%s: WARNING: unable to register pfil hook\n",
460 		    ifp->if_xname);
461 #endif
462 
463 	if (domains)
464 		if_attachdomain1(ifp);
465 
466 	/* Announce the interface. */
467 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
468 }
469 
470 void
471 if_attachdomain()
472 {
473 	struct ifnet *ifp;
474 	int s;
475 
476 	s = splnet();
477 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
478 		if_attachdomain1(ifp);
479 	splx(s);
480 }
481 
482 void
483 if_attachdomain1(ifp)
484 	struct ifnet *ifp;
485 {
486 	struct domain *dp;
487 	int s;
488 
489 	s = splnet();
490 
491 	/* address family dependent data region */
492 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
493 	for (dp = domains; dp; dp = dp->dom_next) {
494 		if (dp->dom_ifattach)
495 			ifp->if_afdata[dp->dom_family] =
496 			    (*dp->dom_ifattach)(ifp);
497 	}
498 
499 	splx(s);
500 }
501 
502 /*
503  * Deactivate an interface.  This points all of the procedure
504  * handles at error stubs.  May be called from interrupt context.
505  */
506 void
507 if_deactivate(ifp)
508 	struct ifnet *ifp;
509 {
510 	int s;
511 
512 	s = splnet();
513 
514 	ifp->if_output	 = if_nulloutput;
515 	ifp->if_input	 = if_nullinput;
516 	ifp->if_start	 = if_nullstart;
517 	ifp->if_ioctl	 = if_nullioctl;
518 	ifp->if_init	 = if_nullinit;
519 	ifp->if_stop	 = if_nullstop;
520 	ifp->if_watchdog = if_nullwatchdog;
521 	ifp->if_drain	 = if_nulldrain;
522 
523 	/* No more packets may be enqueued. */
524 	ifp->if_snd.ifq_maxlen = 0;
525 
526 	splx(s);
527 }
528 
529 /*
530  * Detach an interface from the list of "active" interfaces,
531  * freeing any resources as we go along.
532  *
533  * NOTE: This routine must be called with a valid thread context,
534  * as it may block.
535  */
536 void
537 if_detach(ifp)
538 	struct ifnet *ifp;
539 {
540 	struct socket so;
541 	struct ifaddr *ifa;
542 #ifdef IFAREF_DEBUG
543 	struct ifaddr *last_ifa = NULL;
544 #endif
545 	struct domain *dp;
546 	struct protosw *pr;
547 	struct radix_node_head *rnh;
548 	int s, i, family, purged;
549 
550 	/*
551 	 * XXX It's kind of lame that we have to have the
552 	 * XXX socket structure...
553 	 */
554 	memset(&so, 0, sizeof(so));
555 
556 	s = splnet();
557 
558 	/*
559 	 * Do an if_down() to give protocols a chance to do something.
560 	 */
561 	if_down(ifp);
562 
563 #ifdef ALTQ
564 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
565 		altq_disable(&ifp->if_snd);
566 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
567 		altq_detach(&ifp->if_snd);
568 #endif
569 
570 #ifdef PFIL_HOOKS
571 	(void) pfil_head_unregister(&ifp->if_pfil);
572 #endif
573 
574 	if_free_sadl(ifp);
575 
576 	/*
577 	 * Rip all the addresses off the interface.  This should make
578 	 * all of the routes go away.
579 	 */
580 	while ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) != NULL) {
581 		family = ifa->ifa_addr->sa_family;
582 #ifdef IFAREF_DEBUG
583 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
584 		    ifa, family, ifa->ifa_refcnt);
585 		if (last_ifa != NULL && ifa == last_ifa)
586 			panic("if_detach: loop detected");
587 		last_ifa = ifa;
588 #endif
589 		if (family == AF_LINK) {
590 			/*
591 			 * XXX This case may now be obsolete by
592 			 * XXX the call to if_free_sadl().
593 			 */
594 			rtinit(ifa, RTM_DELETE, 0);
595 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
596 			IFAFREE(ifa);
597 		} else {
598 			dp = pffinddomain(family);
599 #ifdef DIAGNOSTIC
600 			if (dp == NULL)
601 				panic("if_detach: no domain for AF %d",
602 				    family);
603 #endif
604 			purged = 0;
605 			for (pr = dp->dom_protosw;
606 			     pr < dp->dom_protoswNPROTOSW; pr++) {
607 				so.so_proto = pr;
608 				if (pr->pr_usrreq != NULL) {
609 					(void) (*pr->pr_usrreq)(&so,
610 					    PRU_PURGEIF, NULL, NULL,
611 					    (struct mbuf *) ifp, curproc);
612 					purged = 1;
613 				}
614 			}
615 			if (purged == 0) {
616 				/*
617 				 * XXX What's really the best thing to do
618 				 * XXX here?  --thorpej@netbsd.org
619 				 */
620 				printf("if_detach: WARNING: AF %d not purged\n",
621 				    family);
622 			}
623 		}
624 	}
625 
626 	/* Walk the routing table looking for straglers. */
627 	for (i = 0; i <= AF_MAX; i++) {
628 		if ((rnh = rt_tables[i]) != NULL)
629 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
630 	}
631 
632 	for (dp = domains; dp; dp = dp->dom_next) {
633 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
634 			(*dp->dom_ifdetach)(ifp,
635 			    ifp->if_afdata[dp->dom_family]);
636 	}
637 
638 	/* Announce that the interface is gone. */
639 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
640 
641 	ifindex2ifnet[ifp->if_index] = NULL;
642 
643 	TAILQ_REMOVE(&ifnet, ifp, if_list);
644 
645 	/*
646 	 * remove packets came from ifp, from software interrupt queues.
647 	 * net/netisr_dispatch.h is not usable, as some of them use
648 	 * strange queue names.
649 	 */
650 #define IF_DETACH_QUEUES(x) \
651 do { \
652 	extern struct ifqueue x; \
653 	if_detach_queues(ifp, & x); \
654 } while (/*CONSTCOND*/ 0)
655 #ifdef INET
656 #if NARP > 0
657 	IF_DETACH_QUEUES(arpintrq);
658 #endif
659 	IF_DETACH_QUEUES(ipintrq);
660 #endif
661 #ifdef INET6
662 	IF_DETACH_QUEUES(ip6intrq);
663 #endif
664 #ifdef NETATALK
665 	IF_DETACH_QUEUES(atintrq1);
666 	IF_DETACH_QUEUES(atintrq2);
667 #endif
668 #ifdef NS
669 	IF_DETACH_QUEUES(nsintrq);
670 #endif
671 #ifdef ISO
672 	IF_DETACH_QUEUES(clnlintrq);
673 #endif
674 #ifdef CCITT
675 	IF_DETACH_QUEUES(llcintrq);
676 	IF_DETACH_QUEUES(hdintrq);
677 #endif
678 #ifdef NATM
679 	IF_DETACH_QUEUES(natmintrq);
680 #endif
681 #undef IF_DETACH_QUEUES
682 
683 	splx(s);
684 }
685 
686 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
687     defined(ISO) || defined(CCITT) || defined(NATM)
688 static void
689 if_detach_queues(ifp, q)
690 	struct ifnet *ifp;
691 	struct ifqueue *q;
692 {
693 	struct mbuf *m, *prev, *next;
694 
695 	prev = NULL;
696 	for (m = q->ifq_head; m; m = next) {
697 		next = m->m_nextpkt;
698 #ifdef DIAGNOSTIC
699 		if ((m->m_flags & M_PKTHDR) == 0) {
700 			prev = m;
701 			continue;
702 		}
703 #endif
704 		if (m->m_pkthdr.rcvif != ifp) {
705 			prev = m;
706 			continue;
707 		}
708 
709 		if (prev)
710 			prev->m_nextpkt = m->m_nextpkt;
711 		else
712 			q->ifq_head = m->m_nextpkt;
713 		if (q->ifq_tail == m)
714 			q->ifq_tail = prev;
715 		q->ifq_len--;
716 
717 		m->m_nextpkt = NULL;
718 		m_freem(m);
719 		IF_DROP(q);
720 	}
721 }
722 #endif /* defined(INET) || ... */
723 
724 /*
725  * Callback for a radix tree walk to delete all references to an
726  * ifnet.
727  */
728 int
729 if_rt_walktree(rn, v)
730 	struct radix_node *rn;
731 	void *v;
732 {
733 	struct ifnet *ifp = (struct ifnet *)v;
734 	struct rtentry *rt = (struct rtentry *)rn;
735 	int error;
736 
737 	if (rt->rt_ifp == ifp) {
738 		/* Delete the entry. */
739 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
740 		    rt_mask(rt), rt->rt_flags, NULL);
741 		if (error)
742 			printf("%s: warning: unable to delete rtentry @ %p, "
743 			    "error = %d\n", ifp->if_xname, rt, error);
744 	}
745 	return (0);
746 }
747 
748 /*
749  * Create a clone network interface.
750  */
751 int
752 if_clone_create(name)
753 	const char *name;
754 {
755 	struct if_clone *ifc;
756 	int unit;
757 
758 	ifc = if_clone_lookup(name, &unit);
759 	if (ifc == NULL)
760 		return (EINVAL);
761 
762 	if (ifunit(name) != NULL)
763 		return (EEXIST);
764 
765 	return ((*ifc->ifc_create)(ifc, unit));
766 }
767 
768 /*
769  * Destroy a clone network interface.
770  */
771 int
772 if_clone_destroy(name)
773 	const char *name;
774 {
775 	struct if_clone *ifc;
776 	struct ifnet *ifp;
777 
778 	ifc = if_clone_lookup(name, NULL);
779 	if (ifc == NULL)
780 		return (EINVAL);
781 
782 	ifp = ifunit(name);
783 	if (ifp == NULL)
784 		return (ENXIO);
785 
786 	if (ifc->ifc_destroy == NULL)
787 		return (EOPNOTSUPP);
788 
789 	(*ifc->ifc_destroy)(ifp);
790 	return (0);
791 }
792 
793 /*
794  * Look up a network interface cloner.
795  */
796 struct if_clone *
797 if_clone_lookup(name, unitp)
798 	const char *name;
799 	int *unitp;
800 {
801 	struct if_clone *ifc;
802 	const char *cp;
803 	size_t i;
804 
805 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
806 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
807 			if (ifc->ifc_name[i] != *cp)
808 				goto next_ifc;
809 		}
810 		goto found_name;
811  next_ifc:
812 		ifc = LIST_NEXT(ifc, ifc_list);
813 	}
814 
815 	/* No match. */
816 	return (NULL);
817 
818  found_name:
819 	for (i = 0; *cp != '\0'; cp++) {
820 		if (*cp < '0' || *cp > '9') {
821 			/* Bogus unit number. */
822 			return (NULL);
823 		}
824 		i = (i * 10) + (*cp - '0');
825 	}
826 
827 	if (unitp != NULL)
828 		*unitp = i;
829 	return (ifc);
830 }
831 
832 /*
833  * Register a network interface cloner.
834  */
835 void
836 if_clone_attach(ifc)
837 	struct if_clone *ifc;
838 {
839 
840 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
841 	if_cloners_count++;
842 }
843 
844 /*
845  * Unregister a network interface cloner.
846  */
847 void
848 if_clone_detach(ifc)
849 	struct if_clone *ifc;
850 {
851 
852 	LIST_REMOVE(ifc, ifc_list);
853 	if_cloners_count--;
854 }
855 
856 /*
857  * Provide list of interface cloners to userspace.
858  */
859 int
860 if_clone_list(ifcr)
861 	struct if_clonereq *ifcr;
862 {
863 	char outbuf[IFNAMSIZ], *dst;
864 	struct if_clone *ifc;
865 	int count, error = 0;
866 
867 	ifcr->ifcr_total = if_cloners_count;
868 	if ((dst = ifcr->ifcr_buffer) == NULL) {
869 		/* Just asking how many there are. */
870 		return (0);
871 	}
872 
873 	if (ifcr->ifcr_count < 0)
874 		return (EINVAL);
875 
876 	count = (if_cloners_count < ifcr->ifcr_count) ?
877 	    if_cloners_count : ifcr->ifcr_count;
878 
879 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
880 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
881 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
882 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
883 		error = copyout(outbuf, dst, IFNAMSIZ);
884 		if (error)
885 			break;
886 	}
887 
888 	return (error);
889 }
890 
891 /*
892  * Locate an interface based on a complete address.
893  */
894 /*ARGSUSED*/
895 struct ifaddr *
896 ifa_ifwithaddr(addr)
897 	struct sockaddr *addr;
898 {
899 	struct ifnet *ifp;
900 	struct ifaddr *ifa;
901 
902 #define	equal(a1, a2) \
903   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
904 
905 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
906 	     ifp = TAILQ_NEXT(ifp, if_list)) {
907 		if (ifp->if_output == if_nulloutput)
908 			continue;
909 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
910 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
911 			if (ifa->ifa_addr->sa_family != addr->sa_family)
912 				continue;
913 			if (equal(addr, ifa->ifa_addr))
914 				return (ifa);
915 			if ((ifp->if_flags & IFF_BROADCAST) &&
916 			    ifa->ifa_broadaddr &&
917 			    /* IP6 doesn't have broadcast */
918 			    ifa->ifa_broadaddr->sa_len != 0 &&
919 			    equal(ifa->ifa_broadaddr, addr))
920 				return (ifa);
921 		}
922 	}
923 	return (NULL);
924 }
925 
926 /*
927  * Locate the point to point interface with a given destination address.
928  */
929 /*ARGSUSED*/
930 struct ifaddr *
931 ifa_ifwithdstaddr(addr)
932 	struct sockaddr *addr;
933 {
934 	struct ifnet *ifp;
935 	struct ifaddr *ifa;
936 
937 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
938 	     ifp = TAILQ_NEXT(ifp, if_list)) {
939 		if (ifp->if_output == if_nulloutput)
940 			continue;
941 		if (ifp->if_flags & IFF_POINTOPOINT) {
942 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
943 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
944 				if (ifa->ifa_addr->sa_family !=
945 				      addr->sa_family ||
946 				    ifa->ifa_dstaddr == NULL)
947 					continue;
948 				if (equal(addr, ifa->ifa_dstaddr))
949 					return (ifa);
950 			}
951 		}
952 	}
953 	return (NULL);
954 }
955 
956 /*
957  * Find an interface on a specific network.  If many, choice
958  * is most specific found.
959  */
960 struct ifaddr *
961 ifa_ifwithnet(addr)
962 	struct sockaddr *addr;
963 {
964 	struct ifnet *ifp;
965 	struct ifaddr *ifa;
966 	struct sockaddr_dl *sdl;
967 	struct ifaddr *ifa_maybe = 0;
968 	u_int af = addr->sa_family;
969 	char *addr_data = addr->sa_data, *cplim;
970 
971 	if (af == AF_LINK) {
972 		sdl = (struct sockaddr_dl *)addr;
973 		if (sdl->sdl_index && sdl->sdl_index <= if_index &&
974 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
975 			return (ifnet_addrs[sdl->sdl_index]);
976 	}
977 #ifdef NETATALK
978 	if (af == AF_APPLETALK) {
979 		struct sockaddr_at *sat, *sat2;
980 		sat = (struct sockaddr_at *)addr;
981 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
982 		     ifp = TAILQ_NEXT(ifp, if_list)) {
983 			if (ifp->if_output == if_nulloutput)
984 				continue;
985 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
986 			if (ifa == NULL)
987 				continue;
988 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
989 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
990 				return (ifa); /* exact match */
991 			if (ifa_maybe == NULL) {
992 				/* else keep the if with the right range */
993 				ifa_maybe = ifa;
994 			}
995 		}
996 		return (ifa_maybe);
997 	}
998 #endif
999 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1000 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1001 		if (ifp->if_output == if_nulloutput)
1002 			continue;
1003 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1004 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1005 			char *cp, *cp2, *cp3;
1006 
1007 			if (ifa->ifa_addr->sa_family != af ||
1008 			    ifa->ifa_netmask == 0)
1009  next:				continue;
1010 			cp = addr_data;
1011 			cp2 = ifa->ifa_addr->sa_data;
1012 			cp3 = ifa->ifa_netmask->sa_data;
1013 			cplim = (char *)ifa->ifa_netmask +
1014 			    ifa->ifa_netmask->sa_len;
1015 			while (cp3 < cplim) {
1016 				if ((*cp++ ^ *cp2++) & *cp3++) {
1017 					/* want to continue for() loop */
1018 					goto next;
1019 				}
1020 			}
1021 			if (ifa_maybe == 0 ||
1022 			    rn_refines((caddr_t)ifa->ifa_netmask,
1023 			    (caddr_t)ifa_maybe->ifa_netmask))
1024 				ifa_maybe = ifa;
1025 		}
1026 	}
1027 	return (ifa_maybe);
1028 }
1029 
1030 /*
1031  * Find the interface of the addresss.
1032  */
1033 struct ifaddr *
1034 ifa_ifwithladdr(addr)
1035 	struct sockaddr *addr;
1036 {
1037 	struct ifaddr *ia;
1038 
1039 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1040 	    (ia = ifa_ifwithnet(addr)))
1041 		return (ia);
1042 	return (NULL);
1043 }
1044 
1045 /*
1046  * Find an interface using a specific address family
1047  */
1048 struct ifaddr *
1049 ifa_ifwithaf(af)
1050 	int af;
1051 {
1052 	struct ifnet *ifp;
1053 	struct ifaddr *ifa;
1054 
1055 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1056 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1057 		if (ifp->if_output == if_nulloutput)
1058 			continue;
1059 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1060 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1061 			if (ifa->ifa_addr->sa_family == af)
1062 				return (ifa);
1063 		}
1064 	}
1065 	return (NULL);
1066 }
1067 
1068 /*
1069  * Find an interface address specific to an interface best matching
1070  * a given address.
1071  */
1072 struct ifaddr *
1073 ifaof_ifpforaddr(addr, ifp)
1074 	struct sockaddr *addr;
1075 	struct ifnet *ifp;
1076 {
1077 	struct ifaddr *ifa;
1078 	char *cp, *cp2, *cp3;
1079 	char *cplim;
1080 	struct ifaddr *ifa_maybe = 0;
1081 	u_int af = addr->sa_family;
1082 
1083 	if (ifp->if_output == if_nulloutput)
1084 		return (NULL);
1085 
1086 	if (af >= AF_MAX)
1087 		return (NULL);
1088 
1089 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1090 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1091 		if (ifa->ifa_addr->sa_family != af)
1092 			continue;
1093 		ifa_maybe = ifa;
1094 		if (ifa->ifa_netmask == 0) {
1095 			if (equal(addr, ifa->ifa_addr) ||
1096 			    (ifa->ifa_dstaddr &&
1097 			     equal(addr, ifa->ifa_dstaddr)))
1098 				return (ifa);
1099 			continue;
1100 		}
1101 		cp = addr->sa_data;
1102 		cp2 = ifa->ifa_addr->sa_data;
1103 		cp3 = ifa->ifa_netmask->sa_data;
1104 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1105 		for (; cp3 < cplim; cp3++) {
1106 			if ((*cp++ ^ *cp2++) & *cp3)
1107 				break;
1108 		}
1109 		if (cp3 == cplim)
1110 			return (ifa);
1111 	}
1112 	return (ifa_maybe);
1113 }
1114 
1115 /*
1116  * Default action when installing a route with a Link Level gateway.
1117  * Lookup an appropriate real ifa to point to.
1118  * This should be moved to /sys/net/link.c eventually.
1119  */
1120 void
1121 link_rtrequest(cmd, rt, info)
1122 	int cmd;
1123 	struct rtentry *rt;
1124 	struct rt_addrinfo *info;
1125 {
1126 	struct ifaddr *ifa;
1127 	struct sockaddr *dst;
1128 	struct ifnet *ifp;
1129 
1130 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1131 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1132 		return;
1133 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1134 		IFAFREE(rt->rt_ifa);
1135 		rt->rt_ifa = ifa;
1136 		IFAREF(ifa);
1137 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1138 			ifa->ifa_rtrequest(cmd, rt, info);
1139 	}
1140 }
1141 
1142 /*
1143  * Mark an interface down and notify protocols of
1144  * the transition.
1145  * NOTE: must be called at splsoftnet or equivalent.
1146  */
1147 void
1148 if_down(ifp)
1149 	struct ifnet *ifp;
1150 {
1151 	struct ifaddr *ifa;
1152 
1153 	ifp->if_flags &= ~IFF_UP;
1154 	microtime(&ifp->if_lastchange);
1155 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1156 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1157 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1158 	IFQ_PURGE(&ifp->if_snd);
1159 	rt_ifmsg(ifp);
1160 }
1161 
1162 /*
1163  * Mark an interface up and notify protocols of
1164  * the transition.
1165  * NOTE: must be called at splsoftnet or equivalent.
1166  */
1167 void
1168 if_up(ifp)
1169 	struct ifnet *ifp;
1170 {
1171 #ifdef notyet
1172 	struct ifaddr *ifa;
1173 #endif
1174 
1175 	ifp->if_flags |= IFF_UP;
1176 	microtime(&ifp->if_lastchange);
1177 #ifdef notyet
1178 	/* this has no effect on IP, and will kill all ISO connections XXX */
1179 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1180 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1181 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1182 #endif
1183 	rt_ifmsg(ifp);
1184 #ifdef INET6
1185 	in6_if_up(ifp);
1186 #endif
1187 }
1188 
1189 /*
1190  * Handle interface watchdog timer routines.  Called
1191  * from softclock, we decrement timers (if set) and
1192  * call the appropriate interface routine on expiration.
1193  */
1194 void
1195 if_slowtimo(arg)
1196 	void *arg;
1197 {
1198 	struct ifnet *ifp;
1199 	int s = splnet();
1200 
1201 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1202 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1203 		if (ifp->if_timer == 0 || --ifp->if_timer)
1204 			continue;
1205 		if (ifp->if_watchdog)
1206 			(*ifp->if_watchdog)(ifp);
1207 	}
1208 	splx(s);
1209 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1210 	    if_slowtimo, NULL);
1211 }
1212 
1213 /*
1214  * Set/clear promiscuous mode on interface ifp based on the truth value
1215  * of pswitch.  The calls are reference counted so that only the first
1216  * "on" request actually has an effect, as does the final "off" request.
1217  * Results are undefined if the "off" and "on" requests are not matched.
1218  */
1219 int
1220 ifpromisc(ifp, pswitch)
1221 	struct ifnet *ifp;
1222 	int pswitch;
1223 {
1224 	int pcount, ret;
1225 	short flags;
1226 	struct ifreq ifr;
1227 
1228 	pcount = ifp->if_pcount;
1229 	flags = ifp->if_flags;
1230 	if (pswitch) {
1231 		/*
1232 		 * Allow the device to be "placed" into promiscuous
1233 		 * mode even if it is not configured up.  It will
1234 		 * consult IFF_PROMISC when it is is brought up.
1235 		 */
1236 		if (ifp->if_pcount++ != 0)
1237 			return (0);
1238 		ifp->if_flags |= IFF_PROMISC;
1239 		if ((ifp->if_flags & IFF_UP) == 0)
1240 			return (0);
1241 	} else {
1242 		if (--ifp->if_pcount > 0)
1243 			return (0);
1244 		ifp->if_flags &= ~IFF_PROMISC;
1245 		/*
1246 		 * If the device is not configured up, we should not need to
1247 		 * turn off promiscuous mode (device should have turned it
1248 		 * off when interface went down; and will look at IFF_PROMISC
1249 		 * again next time interface comes up).
1250 		 */
1251 		if ((ifp->if_flags & IFF_UP) == 0)
1252 			return (0);
1253 	}
1254 	memset(&ifr, 0, sizeof(ifr));
1255 	ifr.ifr_flags = ifp->if_flags;
1256 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1257 	/* Restore interface state if not successful. */
1258 	if (ret != 0) {
1259 		ifp->if_pcount = pcount;
1260 		ifp->if_flags = flags;
1261 	}
1262 	return (ret);
1263 }
1264 
1265 /*
1266  * Map interface name to
1267  * interface structure pointer.
1268  */
1269 struct ifnet *
1270 ifunit(name)
1271 	const char *name;
1272 {
1273 	struct ifnet *ifp;
1274 	const char *cp = name;
1275 	u_int unit = 0;
1276 	u_int i;
1277 
1278 	/*
1279 	 * If the entire name is a number, treat it as an ifindex.
1280 	 */
1281 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1282 		unit = unit * 10 + (*cp - '0');
1283 	}
1284 
1285 	/*
1286 	 * If the number took all of the name, then it's a valid ifindex.
1287 	 */
1288 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1289 		if (unit >= if_index)
1290 			return (NULL);
1291 		ifp = ifindex2ifnet[unit];
1292 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1293 			return (NULL);
1294 		return (ifp);
1295 	}
1296 
1297 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1298 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1299 		if (ifp->if_output == if_nulloutput)
1300 			continue;
1301 	 	if (strcmp(ifp->if_xname, name) == 0)
1302 			return (ifp);
1303 	}
1304 	return (NULL);
1305 }
1306 
1307 /*
1308  * Interface ioctls.
1309  */
1310 int
1311 ifioctl(so, cmd, data, p)
1312 	struct socket *so;
1313 	u_long cmd;
1314 	caddr_t data;
1315 	struct proc *p;
1316 {
1317 	struct ifnet *ifp;
1318 	struct ifreq *ifr;
1319 	struct ifcapreq *ifcr;
1320 	struct ifdatareq *ifdr;
1321 	int s, error = 0;
1322 	short oif_flags;
1323 
1324 	switch (cmd) {
1325 
1326 	case SIOCGIFCONF:
1327 	case OSIOCGIFCONF:
1328 		return (ifconf(cmd, data));
1329 	}
1330 	ifr = (struct ifreq *)data;
1331 	ifcr = (struct ifcapreq *)data;
1332 	ifdr = (struct ifdatareq *)data;
1333 
1334 	switch (cmd) {
1335 	case SIOCIFCREATE:
1336 	case SIOCIFDESTROY:
1337 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1338 			return (error);
1339 		return ((cmd == SIOCIFCREATE) ?
1340 			if_clone_create(ifr->ifr_name) :
1341 			if_clone_destroy(ifr->ifr_name));
1342 
1343 	case SIOCIFGCLONERS:
1344 		return (if_clone_list((struct if_clonereq *)data));
1345 	}
1346 
1347 	ifp = ifunit(ifr->ifr_name);
1348 	if (ifp == 0)
1349 		return (ENXIO);
1350 	oif_flags = ifp->if_flags;
1351 	switch (cmd) {
1352 
1353 	case SIOCGIFFLAGS:
1354 		ifr->ifr_flags = ifp->if_flags;
1355 		break;
1356 
1357 	case SIOCGIFMETRIC:
1358 		ifr->ifr_metric = ifp->if_metric;
1359 		break;
1360 
1361 	case SIOCGIFMTU:
1362 		ifr->ifr_mtu = ifp->if_mtu;
1363 		break;
1364 
1365 	case SIOCGIFDLT:
1366 		ifr->ifr_dlt = ifp->if_dlt;
1367 		break;
1368 
1369 	case SIOCSIFFLAGS:
1370 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1371 			return (error);
1372 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1373 			s = splnet();
1374 			if_down(ifp);
1375 			splx(s);
1376 		}
1377 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1378 			s = splnet();
1379 			if_up(ifp);
1380 			splx(s);
1381 		}
1382 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1383 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1384 		if (ifp->if_ioctl)
1385 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1386 		break;
1387 
1388 	case SIOCGIFCAP:
1389 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1390 		ifcr->ifcr_capenable = ifp->if_capenable;
1391 		break;
1392 
1393 	case SIOCSIFCAP:
1394 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1395 			return (error);
1396 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1397 			return (EINVAL);
1398 		if (ifp->if_ioctl == NULL)
1399 			return (EOPNOTSUPP);
1400 
1401 		/* Must prevent race with packet reception here. */
1402 		s = splnet();
1403 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1404 			struct ifreq ifrq;
1405 
1406 			ifrq.ifr_flags = ifp->if_flags;
1407 			ifp->if_capenable = ifcr->ifcr_capenable;
1408 
1409 			/* Pre-compute the checksum flags mask. */
1410 			ifp->if_csum_flags_tx = 0;
1411 			ifp->if_csum_flags_rx = 0;
1412 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1413 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1414 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1415 			}
1416 
1417 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1418 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1419 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1420 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1421 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1422 
1423 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1424 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1425 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1426 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1427 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1428 
1429 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1430 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1431 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1432 			}
1433 
1434 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1435 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1436 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1437 			}
1438 
1439 			/*
1440 			 * Only kick the interface if it's up.  If it's
1441 			 * not up now, it will notice the cap enables
1442 			 * when it is brought up later.
1443 			 */
1444 			if (ifp->if_flags & IFF_UP)
1445 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1446 				    (caddr_t) &ifrq);
1447 		}
1448 		splx(s);
1449 		break;
1450 
1451 	case SIOCSIFMETRIC:
1452 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1453 			return (error);
1454 		ifp->if_metric = ifr->ifr_metric;
1455 		break;
1456 
1457 	case SIOCGIFDATA:
1458 		ifdr->ifdr_data = ifp->if_data;
1459 		break;
1460 
1461 	case SIOCZIFDATA:
1462 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1463 			return (error);
1464 		ifdr->ifdr_data = ifp->if_data;
1465 		/*
1466 		 * Assumes that the volatile counters that can be
1467 		 * zero'ed are at the end of if_data.
1468 		 */
1469 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1470 		    offsetof(struct if_data, ifi_ipackets));
1471 		break;
1472 
1473 	case SIOCSIFMTU:
1474 	{
1475 		u_long oldmtu = ifp->if_mtu;
1476 
1477 		error = suser(p->p_ucred, &p->p_acflag);
1478 		if (error)
1479 			return (error);
1480 		if (ifp->if_ioctl == NULL)
1481 			return (EOPNOTSUPP);
1482 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1483 
1484 		/*
1485 		 * If the link MTU changed, do network layer specific procedure.
1486 		 */
1487 		if (ifp->if_mtu != oldmtu) {
1488 #ifdef INET6
1489 			nd6_setmtu(ifp);
1490 #endif
1491 		}
1492 		break;
1493 	}
1494 	case SIOCSIFPHYADDR:
1495 	case SIOCDIFPHYADDR:
1496 #ifdef INET6
1497 	case SIOCSIFPHYADDR_IN6:
1498 #endif
1499 	case SIOCSLIFPHYADDR:
1500 	case SIOCADDMULTI:
1501 	case SIOCDELMULTI:
1502 	case SIOCSIFMEDIA:
1503 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1504 			return (error);
1505 		/* FALLTHROUGH */
1506 	case SIOCGIFPSRCADDR:
1507 	case SIOCGIFPDSTADDR:
1508 	case SIOCGLIFPHYADDR:
1509 	case SIOCGIFMEDIA:
1510 		if (ifp->if_ioctl == 0)
1511 			return (EOPNOTSUPP);
1512 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1513 		break;
1514 
1515 	case SIOCSDRVSPEC:
1516 	case SIOCS80211NWID:
1517 	case SIOCS80211NWKEY:
1518 	case SIOCS80211POWER:
1519 	case SIOCS80211BSSID:
1520 	case SIOCS80211CHANNEL:
1521 		/* XXX:  need to pass proc pointer through to driver... */
1522 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1523 			return (error);
1524 	/* FALLTHROUGH */
1525 	default:
1526 		if (so->so_proto == 0)
1527 			return (EOPNOTSUPP);
1528 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(LKM)
1529 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1530 		    (struct mbuf *)cmd, (struct mbuf *)data,
1531 		    (struct mbuf *)ifp, p));
1532 #else
1533 	    {
1534 		int ocmd = cmd;
1535 
1536 		switch (cmd) {
1537 
1538 		case SIOCSIFADDR:
1539 		case SIOCSIFDSTADDR:
1540 		case SIOCSIFBRDADDR:
1541 		case SIOCSIFNETMASK:
1542 #if BYTE_ORDER != BIG_ENDIAN
1543 			if (ifr->ifr_addr.sa_family == 0 &&
1544 			    ifr->ifr_addr.sa_len < 16) {
1545 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1546 				ifr->ifr_addr.sa_len = 16;
1547 			}
1548 #else
1549 			if (ifr->ifr_addr.sa_len == 0)
1550 				ifr->ifr_addr.sa_len = 16;
1551 #endif
1552 			break;
1553 
1554 		case OSIOCGIFADDR:
1555 			cmd = SIOCGIFADDR;
1556 			break;
1557 
1558 		case OSIOCGIFDSTADDR:
1559 			cmd = SIOCGIFDSTADDR;
1560 			break;
1561 
1562 		case OSIOCGIFBRDADDR:
1563 			cmd = SIOCGIFBRDADDR;
1564 			break;
1565 
1566 		case OSIOCGIFNETMASK:
1567 			cmd = SIOCGIFNETMASK;
1568 		}
1569 
1570 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1571 		    (struct mbuf *)cmd, (struct mbuf *)data,
1572 		    (struct mbuf *)ifp, p));
1573 
1574 		switch (ocmd) {
1575 		case OSIOCGIFADDR:
1576 		case OSIOCGIFDSTADDR:
1577 		case OSIOCGIFBRDADDR:
1578 		case OSIOCGIFNETMASK:
1579 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1580 		}
1581 	    }
1582 #endif /* COMPAT_43 */
1583 		break;
1584 	}
1585 
1586 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1587 #ifdef INET6
1588 		if ((ifp->if_flags & IFF_UP) != 0) {
1589 			s = splnet();
1590 			in6_if_up(ifp);
1591 			splx(s);
1592 		}
1593 #endif
1594 	}
1595 
1596 	return (error);
1597 }
1598 
1599 /*
1600  * Return interface configuration
1601  * of system.  List may be used
1602  * in later ioctl's (above) to get
1603  * other information.
1604  */
1605 /*ARGSUSED*/
1606 int
1607 ifconf(cmd, data)
1608 	u_long cmd;
1609 	caddr_t data;
1610 {
1611 	struct ifconf *ifc = (struct ifconf *)data;
1612 	struct ifnet *ifp;
1613 	struct ifaddr *ifa;
1614 	struct ifreq ifr, *ifrp;
1615 	int space = ifc->ifc_len, error = 0;
1616 
1617 	ifrp = ifc->ifc_req;
1618 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1619 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1620 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1621 			memset((caddr_t)&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1622 			if (space >= (int)sizeof (ifr)) {
1623 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1624 						sizeof(ifr));
1625 				if (error)
1626 					break;
1627 			}
1628 			space -= sizeof (ifr), ifrp++;
1629 		} else
1630 		    for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1631 			struct sockaddr *sa = ifa->ifa_addr;
1632 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4)
1633 			if (cmd == OSIOCGIFCONF) {
1634 				struct osockaddr *osa =
1635 					 (struct osockaddr *)&ifr.ifr_addr;
1636 				ifr.ifr_addr = *sa;
1637 				osa->sa_family = sa->sa_family;
1638 				if (space >= (int)sizeof (ifr)) {
1639 					error = copyout((caddr_t)&ifr,
1640 							(caddr_t)ifrp,
1641 							sizeof (ifr));
1642 					ifrp++;
1643 				}
1644 			} else
1645 #endif
1646 			if (sa->sa_len <= sizeof(*sa)) {
1647 				ifr.ifr_addr = *sa;
1648 				if (space >= (int)sizeof (ifr)) {
1649 					error = copyout((caddr_t)&ifr,
1650 							(caddr_t)ifrp,
1651 							sizeof (ifr));
1652 					ifrp++;
1653 				}
1654 			} else {
1655 				space -= sa->sa_len - sizeof(*sa);
1656 				if (space >= (int)sizeof (ifr)) {
1657 					error = copyout((caddr_t)&ifr,
1658 							(caddr_t)ifrp,
1659 							sizeof (ifr.ifr_name));
1660 					if (error == 0) {
1661 						error = copyout((caddr_t)sa,
1662 						  (caddr_t)&ifrp->ifr_addr,
1663 						  sa->sa_len);
1664 					}
1665 					ifrp = (struct ifreq *)
1666 						(sa->sa_len +
1667 						 (caddr_t)&ifrp->ifr_addr);
1668 				}
1669 			}
1670 			if (error)
1671 				break;
1672 			space -= sizeof (ifr);
1673 		}
1674 	}
1675 	ifc->ifc_len -= space;
1676 	return (error);
1677 }
1678