xref: /netbsd-src/sys/net/if.c (revision aa73cae19608873cc4d1f712c4a0f8f8435f1ffa)
1 /*	$NetBSD: if.c,v 1.154 2005/02/26 22:45:09 perry 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 Studenmund 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. Neither the name of the University nor the names of its contributors
81  *    may be used to endorse or promote products derived from this software
82  *    without specific prior written permission.
83  *
84  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
85  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
86  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
87  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
88  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
89  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
90  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
91  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
92  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
93  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
94  * SUCH DAMAGE.
95  *
96  *	@(#)if.c	8.5 (Berkeley) 1/9/95
97  */
98 
99 #include <sys/cdefs.h>
100 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.154 2005/02/26 22:45:09 perry Exp $");
101 
102 #include "opt_inet.h"
103 
104 #include "opt_compat_linux.h"
105 #include "opt_compat_svr4.h"
106 #include "opt_compat_ultrix.h"
107 #include "opt_compat_43.h"
108 #include "opt_atalk.h"
109 #include "opt_ccitt.h"
110 #include "opt_natm.h"
111 #include "opt_pfil_hooks.h"
112 
113 #include <sys/param.h>
114 #include <sys/mbuf.h>
115 #include <sys/systm.h>
116 #include <sys/callout.h>
117 #include <sys/proc.h>
118 #include <sys/socket.h>
119 #include <sys/socketvar.h>
120 #include <sys/domain.h>
121 #include <sys/protosw.h>
122 #include <sys/kernel.h>
123 #include <sys/ioctl.h>
124 #include <sys/sysctl.h>
125 
126 #include <net/if.h>
127 #include <net/if_dl.h>
128 #include <net/if_ether.h>
129 #include <net/if_media.h>
130 #include <net80211/ieee80211.h>
131 #include <net80211/ieee80211_ioctl.h>
132 #include <net/if_types.h>
133 #include <net/radix.h>
134 #include <net/route.h>
135 #include <net/netisr.h>
136 #ifdef NETATALK
137 #include <netatalk/at_extern.h>
138 #include <netatalk/at.h>
139 #endif
140 #include <net/pfil.h>
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 #ifdef PFIL_HOOKS
165 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
166 #endif
167 
168 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
169     defined(ISO) || defined(CCITT) || defined(NATM)
170 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
171 #endif
172 
173 /*
174  * Network interface utility routines.
175  *
176  * Routines with ifa_ifwith* names take sockaddr *'s as
177  * parameters.
178  */
179 void
180 ifinit()
181 {
182 
183 	callout_init(&if_slowtimo_ch);
184 	if_slowtimo(NULL);
185 #ifdef PFIL_HOOKS
186 	if_pfil.ph_type = PFIL_TYPE_IFNET;
187 	if_pfil.ph_ifnet = NULL;
188 	if (pfil_head_register(&if_pfil) != 0)
189 		printf("WARNING: unable to register pfil hook\n");
190 #endif
191 }
192 
193 /*
194  * Null routines used while an interface is going away.  These routines
195  * just return an error.
196  */
197 
198 int
199 if_nulloutput(ifp, m, so, rt)
200 	struct ifnet *ifp;
201 	struct mbuf *m;
202 	struct sockaddr *so;
203 	struct rtentry *rt;
204 {
205 
206 	return (ENXIO);
207 }
208 
209 void
210 if_nullinput(ifp, m)
211 	struct ifnet *ifp;
212 	struct mbuf *m;
213 {
214 
215 	/* Nothing. */
216 }
217 
218 void
219 if_nullstart(ifp)
220 	struct ifnet *ifp;
221 {
222 
223 	/* Nothing. */
224 }
225 
226 int
227 if_nullioctl(ifp, cmd, data)
228 	struct ifnet *ifp;
229 	u_long cmd;
230 	caddr_t data;
231 {
232 
233 	return (ENXIO);
234 }
235 
236 int
237 if_nullinit(ifp)
238 	struct ifnet *ifp;
239 {
240 
241 	return (ENXIO);
242 }
243 
244 void
245 if_nullstop(ifp, disable)
246 	struct ifnet *ifp;
247 	int disable;
248 {
249 
250 	/* Nothing. */
251 }
252 
253 void
254 if_nullwatchdog(ifp)
255 	struct ifnet *ifp;
256 {
257 
258 	/* Nothing. */
259 }
260 
261 void
262 if_nulldrain(ifp)
263 	struct ifnet *ifp;
264 {
265 
266 	/* Nothing. */
267 }
268 
269 static u_int if_index = 1;
270 struct ifnet_head ifnet;
271 size_t if_indexlim = 0;
272 struct ifaddr **ifnet_addrs = NULL;
273 struct ifnet **ifindex2ifnet = NULL;
274 struct ifnet *lo0ifp;
275 
276 /*
277  * Allocate the link level name for the specified interface.  This
278  * is an attachment helper.  It must be called after ifp->if_addrlen
279  * is initialized, which may not be the case when if_attach() is
280  * called.
281  */
282 void
283 if_alloc_sadl(struct ifnet *ifp)
284 {
285 	unsigned socksize, ifasize;
286 	int namelen, masklen;
287 	struct sockaddr_dl *sdl;
288 	struct ifaddr *ifa;
289 
290 	/*
291 	 * If the interface already has a link name, release it
292 	 * now.  This is useful for interfaces that can change
293 	 * link types, and thus switch link names often.
294 	 */
295 	if (ifp->if_sadl != NULL)
296 		if_free_sadl(ifp);
297 
298 	namelen = strlen(ifp->if_xname);
299 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
300 	socksize = masklen + ifp->if_addrlen;
301 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
302 	if (socksize < sizeof(*sdl))
303 		socksize = sizeof(*sdl);
304 	socksize = ROUNDUP(socksize);
305 	ifasize = sizeof(*ifa) + 2 * socksize;
306 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
307 	memset((caddr_t)ifa, 0, ifasize);
308 	sdl = (struct sockaddr_dl *)(ifa + 1);
309 	sdl->sdl_len = socksize;
310 	sdl->sdl_family = AF_LINK;
311 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
312 	sdl->sdl_nlen = namelen;
313 	sdl->sdl_alen = ifp->if_addrlen;
314 	sdl->sdl_index = ifp->if_index;
315 	sdl->sdl_type = ifp->if_type;
316 	ifnet_addrs[ifp->if_index] = ifa;
317 	IFAREF(ifa);
318 	ifa->ifa_ifp = ifp;
319 	ifa->ifa_rtrequest = link_rtrequest;
320 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
321 	IFAREF(ifa);
322 	ifa->ifa_addr = (struct sockaddr *)sdl;
323 	ifp->if_sadl = sdl;
324 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
325 	ifa->ifa_netmask = (struct sockaddr *)sdl;
326 	sdl->sdl_len = masklen;
327 	while (namelen != 0)
328 		sdl->sdl_data[--namelen] = 0xff;
329 }
330 
331 /*
332  * Free the link level name for the specified interface.  This is
333  * a detach helper.  This is called from if_detach() or from
334  * link layer type specific detach functions.
335  */
336 void
337 if_free_sadl(struct ifnet *ifp)
338 {
339 	struct ifaddr *ifa;
340 	int s;
341 
342 	ifa = ifnet_addrs[ifp->if_index];
343 	if (ifa == NULL) {
344 		KASSERT(ifp->if_sadl == NULL);
345 		return;
346 	}
347 
348 	KASSERT(ifp->if_sadl != NULL);
349 
350 	s = splnet();
351 	rtinit(ifa, RTM_DELETE, 0);
352 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
353 	IFAFREE(ifa);
354 
355 	ifp->if_sadl = NULL;
356 
357 	ifnet_addrs[ifp->if_index] = NULL;
358 	IFAFREE(ifa);
359 	splx(s);
360 }
361 
362 /*
363  * Attach an interface to the
364  * list of "active" interfaces.
365  */
366 void
367 if_attach(ifp)
368 	struct ifnet *ifp;
369 {
370 	int indexlim = 0;
371 
372 	if (if_indexlim == 0) {
373 		TAILQ_INIT(&ifnet);
374 		if_indexlim = 8;
375 	}
376 	TAILQ_INIT(&ifp->if_addrlist);
377 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
378 	ifp->if_index = if_index;
379 	if (ifindex2ifnet == 0)
380 		if_index++;
381 	else
382 		while (ifp->if_index < if_indexlim &&
383 		    ifindex2ifnet[ifp->if_index] != NULL) {
384 			++if_index;
385 			if (if_index == 0)
386 				if_index = 1;
387 			/*
388 			 * If we hit USHRT_MAX, we skip back to 0 since
389 			 * there are a number of places where the value
390 			 * of if_index or if_index itself is compared
391 			 * to or stored in an unsigned short.  By
392 			 * jumping back, we won't botch those assignments
393 			 * or comparisons.
394 			 */
395 			else if (if_index == USHRT_MAX) {
396 				/*
397 				 * However, if we have to jump back to
398 				 * zero *twice* without finding an empty
399 				 * slot in ifindex2ifnet[], then there
400 				 * there are too many (>65535) interfaces.
401 				 */
402 				if (indexlim++)
403 					panic("too many interfaces");
404 				else
405 					if_index = 1;
406 			}
407 			ifp->if_index = if_index;
408 		}
409 
410 	/*
411 	 * We have some arrays that should be indexed by if_index.
412 	 * since if_index will grow dynamically, they should grow too.
413 	 *	struct ifadd **ifnet_addrs
414 	 *	struct ifnet **ifindex2ifnet
415 	 */
416 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
417 	    ifp->if_index >= if_indexlim) {
418 		size_t m, n, oldlim;
419 		caddr_t q;
420 
421 		oldlim = if_indexlim;
422 		while (ifp->if_index >= if_indexlim)
423 			if_indexlim <<= 1;
424 
425 		/* grow ifnet_addrs */
426 		m = oldlim * sizeof(struct ifaddr *);
427 		n = if_indexlim * sizeof(struct ifaddr *);
428 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
429 		memset(q, 0, n);
430 		if (ifnet_addrs) {
431 			bcopy((caddr_t)ifnet_addrs, q, m);
432 			free((caddr_t)ifnet_addrs, M_IFADDR);
433 		}
434 		ifnet_addrs = (struct ifaddr **)q;
435 
436 		/* grow ifindex2ifnet */
437 		m = oldlim * sizeof(struct ifnet *);
438 		n = if_indexlim * sizeof(struct ifnet *);
439 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
440 		memset(q, 0, n);
441 		if (ifindex2ifnet) {
442 			bcopy((caddr_t)ifindex2ifnet, q, m);
443 			free((caddr_t)ifindex2ifnet, M_IFADDR);
444 		}
445 		ifindex2ifnet = (struct ifnet **)q;
446 	}
447 
448 	ifindex2ifnet[ifp->if_index] = ifp;
449 
450 	/*
451 	 * Link level name is allocated later by a separate call to
452 	 * if_alloc_sadl().
453 	 */
454 
455 	if (ifp->if_snd.ifq_maxlen == 0)
456 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
457 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
458 
459 	ifp->if_link_state = LINK_STATE_UNKNOWN;
460 
461 	ifp->if_capenable = 0;
462 	ifp->if_csum_flags_tx = 0;
463 	ifp->if_csum_flags_rx = 0;
464 
465 #ifdef ALTQ
466 	ifp->if_snd.altq_type = 0;
467 	ifp->if_snd.altq_disc = NULL;
468 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
469 	ifp->if_snd.altq_tbr  = NULL;
470 	ifp->if_snd.altq_ifp  = ifp;
471 #endif
472 
473 #ifdef PFIL_HOOKS
474 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
475 	ifp->if_pfil.ph_ifnet = ifp;
476 	if (pfil_head_register(&ifp->if_pfil) != 0)
477 		printf("%s: WARNING: unable to register pfil hook\n",
478 		    ifp->if_xname);
479 	(void)pfil_run_hooks(&if_pfil,
480 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
481 #endif
482 
483 	if (!STAILQ_EMPTY(&domains))
484 		if_attachdomain1(ifp);
485 
486 	/* Announce the interface. */
487 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
488 }
489 
490 void
491 if_attachdomain()
492 {
493 	struct ifnet *ifp;
494 	int s;
495 
496 	s = splnet();
497 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
498 		if_attachdomain1(ifp);
499 	splx(s);
500 }
501 
502 void
503 if_attachdomain1(ifp)
504 	struct ifnet *ifp;
505 {
506 	struct domain *dp;
507 	int s;
508 
509 	s = splnet();
510 
511 	/* address family dependent data region */
512 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
513 	DOMAIN_FOREACH(dp) {
514 		if (dp->dom_ifattach)
515 			ifp->if_afdata[dp->dom_family] =
516 			    (*dp->dom_ifattach)(ifp);
517 	}
518 
519 	splx(s);
520 }
521 
522 /*
523  * Deactivate an interface.  This points all of the procedure
524  * handles at error stubs.  May be called from interrupt context.
525  */
526 void
527 if_deactivate(ifp)
528 	struct ifnet *ifp;
529 {
530 	int s;
531 
532 	s = splnet();
533 
534 	ifp->if_output	 = if_nulloutput;
535 	ifp->if_input	 = if_nullinput;
536 	ifp->if_start	 = if_nullstart;
537 	ifp->if_ioctl	 = if_nullioctl;
538 	ifp->if_init	 = if_nullinit;
539 	ifp->if_stop	 = if_nullstop;
540 	ifp->if_watchdog = if_nullwatchdog;
541 	ifp->if_drain	 = if_nulldrain;
542 
543 	/* No more packets may be enqueued. */
544 	ifp->if_snd.ifq_maxlen = 0;
545 
546 	splx(s);
547 }
548 
549 /*
550  * Detach an interface from the list of "active" interfaces,
551  * freeing any resources as we go along.
552  *
553  * NOTE: This routine must be called with a valid thread context,
554  * as it may block.
555  */
556 void
557 if_detach(ifp)
558 	struct ifnet *ifp;
559 {
560 	struct socket so;
561 	struct ifaddr *ifa, **ifap;
562 #ifdef IFAREF_DEBUG
563 	struct ifaddr *last_ifa = NULL;
564 #endif
565 	struct domain *dp;
566 	const struct protosw *pr;
567 	struct radix_node_head *rnh;
568 	int s, i, family, purged;
569 
570 	/*
571 	 * XXX It's kind of lame that we have to have the
572 	 * XXX socket structure...
573 	 */
574 	memset(&so, 0, sizeof(so));
575 
576 	s = splnet();
577 
578 	/*
579 	 * Do an if_down() to give protocols a chance to do something.
580 	 */
581 	if_down(ifp);
582 
583 #ifdef ALTQ
584 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
585 		altq_disable(&ifp->if_snd);
586 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
587 		altq_detach(&ifp->if_snd);
588 #endif
589 
590 #ifdef PFIL_HOOKS
591 	(void)pfil_run_hooks(&if_pfil,
592 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
593 	(void)pfil_head_unregister(&ifp->if_pfil);
594 #endif
595 
596 	/*
597 	 * Rip all the addresses off the interface.  This should make
598 	 * all of the routes go away.
599 	 */
600 	ifap = &TAILQ_FIRST(&ifp->if_addrlist); /* XXX abstraction violation */
601 	while ((ifa = *ifap)) {
602 		family = ifa->ifa_addr->sa_family;
603 #ifdef IFAREF_DEBUG
604 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
605 		    ifa, family, ifa->ifa_refcnt);
606 		if (last_ifa != NULL && ifa == last_ifa)
607 			panic("if_detach: loop detected");
608 		last_ifa = ifa;
609 #endif
610 		if (family == AF_LINK) {
611 			ifap = &TAILQ_NEXT(ifa, ifa_list);
612 			continue;
613 		}
614 		dp = pffinddomain(family);
615 #ifdef DIAGNOSTIC
616 		if (dp == NULL)
617 			panic("if_detach: no domain for AF %d",
618 			    family);
619 #endif
620 		purged = 0;
621 		for (pr = dp->dom_protosw;
622 		     pr < dp->dom_protoswNPROTOSW; pr++) {
623 			so.so_proto = pr;
624 			if (pr->pr_usrreq != NULL) {
625 				(void) (*pr->pr_usrreq)(&so,
626 				    PRU_PURGEIF, NULL, NULL,
627 				    (struct mbuf *) ifp, curproc);
628 				purged = 1;
629 			}
630 		}
631 		if (purged == 0) {
632 			/*
633 			 * XXX What's really the best thing to do
634 			 * XXX here?  --thorpej@NetBSD.org
635 			 */
636 			printf("if_detach: WARNING: AF %d not purged\n",
637 			    family);
638 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
639 		}
640 	}
641 
642 	if_free_sadl(ifp);
643 
644 	/* Walk the routing table looking for straglers. */
645 	for (i = 0; i <= AF_MAX; i++) {
646 		if ((rnh = rt_tables[i]) != NULL)
647 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
648 	}
649 
650 	DOMAIN_FOREACH(dp) {
651 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
652 			(*dp->dom_ifdetach)(ifp,
653 			    ifp->if_afdata[dp->dom_family]);
654 	}
655 
656 	/* Announce that the interface is gone. */
657 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
658 
659 	ifindex2ifnet[ifp->if_index] = NULL;
660 
661 	TAILQ_REMOVE(&ifnet, ifp, if_list);
662 
663 	/*
664 	 * remove packets came from ifp, from software interrupt queues.
665 	 * net/netisr_dispatch.h is not usable, as some of them use
666 	 * strange queue names.
667 	 */
668 #define IF_DETACH_QUEUES(x) \
669 do { \
670 	extern struct ifqueue x; \
671 	if_detach_queues(ifp, & x); \
672 } while (/*CONSTCOND*/ 0)
673 #ifdef INET
674 #if NARP > 0
675 	IF_DETACH_QUEUES(arpintrq);
676 #endif
677 	IF_DETACH_QUEUES(ipintrq);
678 #endif
679 #ifdef INET6
680 	IF_DETACH_QUEUES(ip6intrq);
681 #endif
682 #ifdef NETATALK
683 	IF_DETACH_QUEUES(atintrq1);
684 	IF_DETACH_QUEUES(atintrq2);
685 #endif
686 #ifdef NS
687 	IF_DETACH_QUEUES(nsintrq);
688 #endif
689 #ifdef ISO
690 	IF_DETACH_QUEUES(clnlintrq);
691 #endif
692 #ifdef CCITT
693 	IF_DETACH_QUEUES(llcintrq);
694 	IF_DETACH_QUEUES(hdintrq);
695 #endif
696 #ifdef NATM
697 	IF_DETACH_QUEUES(natmintrq);
698 #endif
699 #ifdef DECNET
700 	IF_DETACH_QUEUES(decnetintrq);
701 #endif
702 #undef IF_DETACH_QUEUES
703 
704 	splx(s);
705 }
706 
707 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
708     defined(ISO) || defined(CCITT) || defined(NATM) || defined(DECNET)
709 static void
710 if_detach_queues(ifp, q)
711 	struct ifnet *ifp;
712 	struct ifqueue *q;
713 {
714 	struct mbuf *m, *prev, *next;
715 
716 	prev = NULL;
717 	for (m = q->ifq_head; m; m = next) {
718 		next = m->m_nextpkt;
719 #ifdef DIAGNOSTIC
720 		if ((m->m_flags & M_PKTHDR) == 0) {
721 			prev = m;
722 			continue;
723 		}
724 #endif
725 		if (m->m_pkthdr.rcvif != ifp) {
726 			prev = m;
727 			continue;
728 		}
729 
730 		if (prev)
731 			prev->m_nextpkt = m->m_nextpkt;
732 		else
733 			q->ifq_head = m->m_nextpkt;
734 		if (q->ifq_tail == m)
735 			q->ifq_tail = prev;
736 		q->ifq_len--;
737 
738 		m->m_nextpkt = NULL;
739 		m_freem(m);
740 		IF_DROP(q);
741 	}
742 }
743 #endif /* defined(INET) || ... */
744 
745 /*
746  * Callback for a radix tree walk to delete all references to an
747  * ifnet.
748  */
749 int
750 if_rt_walktree(rn, v)
751 	struct radix_node *rn;
752 	void *v;
753 {
754 	struct ifnet *ifp = (struct ifnet *)v;
755 	struct rtentry *rt = (struct rtentry *)rn;
756 	int error;
757 
758 	if (rt->rt_ifp == ifp) {
759 		/* Delete the entry. */
760 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
761 		    rt_mask(rt), rt->rt_flags, NULL);
762 		if (error)
763 			printf("%s: warning: unable to delete rtentry @ %p, "
764 			    "error = %d\n", ifp->if_xname, rt, error);
765 	}
766 	return (0);
767 }
768 
769 /*
770  * Create a clone network interface.
771  */
772 int
773 if_clone_create(name)
774 	const char *name;
775 {
776 	struct if_clone *ifc;
777 	int unit;
778 
779 	ifc = if_clone_lookup(name, &unit);
780 	if (ifc == NULL)
781 		return (EINVAL);
782 
783 	if (ifunit(name) != NULL)
784 		return (EEXIST);
785 
786 	return ((*ifc->ifc_create)(ifc, unit));
787 }
788 
789 /*
790  * Destroy a clone network interface.
791  */
792 int
793 if_clone_destroy(name)
794 	const char *name;
795 {
796 	struct if_clone *ifc;
797 	struct ifnet *ifp;
798 
799 	ifc = if_clone_lookup(name, NULL);
800 	if (ifc == NULL)
801 		return (EINVAL);
802 
803 	ifp = ifunit(name);
804 	if (ifp == NULL)
805 		return (ENXIO);
806 
807 	if (ifc->ifc_destroy == NULL)
808 		return (EOPNOTSUPP);
809 
810 	return ((*ifc->ifc_destroy)(ifp));
811 }
812 
813 /*
814  * Look up a network interface cloner.
815  */
816 struct if_clone *
817 if_clone_lookup(name, unitp)
818 	const char *name;
819 	int *unitp;
820 {
821 	struct if_clone *ifc;
822 	const char *cp;
823 	int unit;
824 
825 	/* separate interface name from unit */
826 	for (cp = name;
827 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
828 	    cp++)
829 		continue;
830 
831 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
832 		return (NULL);	/* No name or unit number */
833 
834 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
835 		if (strlen(ifc->ifc_name) == cp - name &&
836 		    !strncmp(name, ifc->ifc_name, cp - name))
837 			break;
838 	}
839 
840 	if (ifc == NULL)
841 		return (NULL);
842 
843 	unit = 0;
844 	while (cp - name < IFNAMSIZ && *cp) {
845 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
846 			/* Bogus unit number. */
847 			return (NULL);
848 		}
849 		unit = (unit * 10) + (*cp++ - '0');
850 	}
851 
852 	if (unitp != NULL)
853 		*unitp = unit;
854 	return (ifc);
855 }
856 
857 /*
858  * Register a network interface cloner.
859  */
860 void
861 if_clone_attach(ifc)
862 	struct if_clone *ifc;
863 {
864 
865 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
866 	if_cloners_count++;
867 }
868 
869 /*
870  * Unregister a network interface cloner.
871  */
872 void
873 if_clone_detach(ifc)
874 	struct if_clone *ifc;
875 {
876 
877 	LIST_REMOVE(ifc, ifc_list);
878 	if_cloners_count--;
879 }
880 
881 /*
882  * Provide list of interface cloners to userspace.
883  */
884 int
885 if_clone_list(ifcr)
886 	struct if_clonereq *ifcr;
887 {
888 	char outbuf[IFNAMSIZ], *dst;
889 	struct if_clone *ifc;
890 	int count, error = 0;
891 
892 	ifcr->ifcr_total = if_cloners_count;
893 	if ((dst = ifcr->ifcr_buffer) == NULL) {
894 		/* Just asking how many there are. */
895 		return (0);
896 	}
897 
898 	if (ifcr->ifcr_count < 0)
899 		return (EINVAL);
900 
901 	count = (if_cloners_count < ifcr->ifcr_count) ?
902 	    if_cloners_count : ifcr->ifcr_count;
903 
904 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
905 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
906 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
907 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
908 		error = copyout(outbuf, dst, IFNAMSIZ);
909 		if (error)
910 			break;
911 	}
912 
913 	return (error);
914 }
915 
916 /*
917  * Locate an interface based on a complete address.
918  */
919 /*ARGSUSED*/
920 struct ifaddr *
921 ifa_ifwithaddr(addr)
922 	const struct sockaddr *addr;
923 {
924 	struct ifnet *ifp;
925 	struct ifaddr *ifa;
926 
927 #define	equal(a1, a2) \
928   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
929 
930 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
931 	     ifp = TAILQ_NEXT(ifp, if_list)) {
932 		if (ifp->if_output == if_nulloutput)
933 			continue;
934 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
935 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
936 			if (ifa->ifa_addr->sa_family != addr->sa_family)
937 				continue;
938 			if (equal(addr, ifa->ifa_addr))
939 				return (ifa);
940 			if ((ifp->if_flags & IFF_BROADCAST) &&
941 			    ifa->ifa_broadaddr &&
942 			    /* IP6 doesn't have broadcast */
943 			    ifa->ifa_broadaddr->sa_len != 0 &&
944 			    equal(ifa->ifa_broadaddr, addr))
945 				return (ifa);
946 		}
947 	}
948 	return (NULL);
949 }
950 
951 /*
952  * Locate the point to point interface with a given destination address.
953  */
954 /*ARGSUSED*/
955 struct ifaddr *
956 ifa_ifwithdstaddr(addr)
957 	const struct sockaddr *addr;
958 {
959 	struct ifnet *ifp;
960 	struct ifaddr *ifa;
961 
962 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
963 	     ifp = TAILQ_NEXT(ifp, if_list)) {
964 		if (ifp->if_output == if_nulloutput)
965 			continue;
966 		if (ifp->if_flags & IFF_POINTOPOINT) {
967 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
968 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
969 				if (ifa->ifa_addr->sa_family !=
970 				      addr->sa_family ||
971 				    ifa->ifa_dstaddr == NULL)
972 					continue;
973 				if (equal(addr, ifa->ifa_dstaddr))
974 					return (ifa);
975 			}
976 		}
977 	}
978 	return (NULL);
979 }
980 
981 /*
982  * Find an interface on a specific network.  If many, choice
983  * is most specific found.
984  */
985 struct ifaddr *
986 ifa_ifwithnet(addr)
987 	const struct sockaddr *addr;
988 {
989 	struct ifnet *ifp;
990 	struct ifaddr *ifa;
991 	const struct sockaddr_dl *sdl;
992 	struct ifaddr *ifa_maybe = 0;
993 	u_int af = addr->sa_family;
994 	char *addr_data = addr->sa_data, *cplim;
995 
996 	if (af == AF_LINK) {
997 		sdl = (struct sockaddr_dl *)addr;
998 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
999 		    ifindex2ifnet[sdl->sdl_index] &&
1000 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1001 			return (ifnet_addrs[sdl->sdl_index]);
1002 	}
1003 #ifdef NETATALK
1004 	if (af == AF_APPLETALK) {
1005 		const struct sockaddr_at *sat, *sat2;
1006 		sat = (struct sockaddr_at *)addr;
1007 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1008 		     ifp = TAILQ_NEXT(ifp, if_list)) {
1009 			if (ifp->if_output == if_nulloutput)
1010 				continue;
1011 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
1012 			if (ifa == NULL)
1013 				continue;
1014 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1015 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1016 				return (ifa); /* exact match */
1017 			if (ifa_maybe == NULL) {
1018 				/* else keep the if with the right range */
1019 				ifa_maybe = ifa;
1020 			}
1021 		}
1022 		return (ifa_maybe);
1023 	}
1024 #endif
1025 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1026 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1027 		if (ifp->if_output == if_nulloutput)
1028 			continue;
1029 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1030 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1031 			char *cp, *cp2, *cp3;
1032 
1033 			if (ifa->ifa_addr->sa_family != af ||
1034 			    ifa->ifa_netmask == 0)
1035  next:				continue;
1036 			cp = addr_data;
1037 			cp2 = ifa->ifa_addr->sa_data;
1038 			cp3 = ifa->ifa_netmask->sa_data;
1039 			cplim = (char *)ifa->ifa_netmask +
1040 			    ifa->ifa_netmask->sa_len;
1041 			while (cp3 < cplim) {
1042 				if ((*cp++ ^ *cp2++) & *cp3++) {
1043 					/* want to continue for() loop */
1044 					goto next;
1045 				}
1046 			}
1047 			if (ifa_maybe == 0 ||
1048 			    rn_refines((caddr_t)ifa->ifa_netmask,
1049 			    (caddr_t)ifa_maybe->ifa_netmask))
1050 				ifa_maybe = ifa;
1051 		}
1052 	}
1053 	return (ifa_maybe);
1054 }
1055 
1056 /*
1057  * Find the interface of the addresss.
1058  */
1059 struct ifaddr *
1060 ifa_ifwithladdr(addr)
1061 	const struct sockaddr *addr;
1062 {
1063 	struct ifaddr *ia;
1064 
1065 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1066 	    (ia = ifa_ifwithnet(addr)))
1067 		return (ia);
1068 	return (NULL);
1069 }
1070 
1071 /*
1072  * Find an interface using a specific address family
1073  */
1074 struct ifaddr *
1075 ifa_ifwithaf(af)
1076 	int af;
1077 {
1078 	struct ifnet *ifp;
1079 	struct ifaddr *ifa;
1080 
1081 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1082 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1083 		if (ifp->if_output == if_nulloutput)
1084 			continue;
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 				return (ifa);
1089 		}
1090 	}
1091 	return (NULL);
1092 }
1093 
1094 /*
1095  * Find an interface address specific to an interface best matching
1096  * a given address.
1097  */
1098 struct ifaddr *
1099 ifaof_ifpforaddr(addr, ifp)
1100 	const struct sockaddr *addr;
1101 	struct ifnet *ifp;
1102 {
1103 	struct ifaddr *ifa;
1104 	const char *cp, *cp2, *cp3;
1105 	const char *cplim;
1106 	struct ifaddr *ifa_maybe = 0;
1107 	u_int af = addr->sa_family;
1108 
1109 	if (ifp->if_output == if_nulloutput)
1110 		return (NULL);
1111 
1112 	if (af >= AF_MAX)
1113 		return (NULL);
1114 
1115 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1116 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1117 		if (ifa->ifa_addr->sa_family != af)
1118 			continue;
1119 		ifa_maybe = ifa;
1120 		if (ifa->ifa_netmask == 0) {
1121 			if (equal(addr, ifa->ifa_addr) ||
1122 			    (ifa->ifa_dstaddr &&
1123 			     equal(addr, ifa->ifa_dstaddr)))
1124 				return (ifa);
1125 			continue;
1126 		}
1127 		cp = addr->sa_data;
1128 		cp2 = ifa->ifa_addr->sa_data;
1129 		cp3 = ifa->ifa_netmask->sa_data;
1130 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1131 		for (; cp3 < cplim; cp3++) {
1132 			if ((*cp++ ^ *cp2++) & *cp3)
1133 				break;
1134 		}
1135 		if (cp3 == cplim)
1136 			return (ifa);
1137 	}
1138 	return (ifa_maybe);
1139 }
1140 
1141 /*
1142  * Default action when installing a route with a Link Level gateway.
1143  * Lookup an appropriate real ifa to point to.
1144  * This should be moved to /sys/net/link.c eventually.
1145  */
1146 void
1147 link_rtrequest(cmd, rt, info)
1148 	int cmd;
1149 	struct rtentry *rt;
1150 	struct rt_addrinfo *info;
1151 {
1152 	struct ifaddr *ifa;
1153 	struct sockaddr *dst;
1154 	struct ifnet *ifp;
1155 
1156 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1157 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1158 		return;
1159 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1160 		IFAFREE(rt->rt_ifa);
1161 		rt->rt_ifa = ifa;
1162 		IFAREF(ifa);
1163 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1164 			ifa->ifa_rtrequest(cmd, rt, info);
1165 	}
1166 }
1167 
1168 /*
1169  * Mark an interface down and notify protocols of
1170  * the transition.
1171  * NOTE: must be called at splsoftnet or equivalent.
1172  */
1173 void
1174 if_down(ifp)
1175 	struct ifnet *ifp;
1176 {
1177 	struct ifaddr *ifa;
1178 
1179 	ifp->if_flags &= ~IFF_UP;
1180 	microtime(&ifp->if_lastchange);
1181 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1182 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1183 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1184 	IFQ_PURGE(&ifp->if_snd);
1185 	rt_ifmsg(ifp);
1186 }
1187 
1188 /*
1189  * Mark an interface up and notify protocols of
1190  * the transition.
1191  * NOTE: must be called at splsoftnet or equivalent.
1192  */
1193 void
1194 if_up(ifp)
1195 	struct ifnet *ifp;
1196 {
1197 #ifdef notyet
1198 	struct ifaddr *ifa;
1199 #endif
1200 
1201 	ifp->if_flags |= IFF_UP;
1202 	microtime(&ifp->if_lastchange);
1203 #ifdef notyet
1204 	/* this has no effect on IP, and will kill all ISO connections XXX */
1205 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1206 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1207 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1208 #endif
1209 	rt_ifmsg(ifp);
1210 #ifdef INET6
1211 	in6_if_up(ifp);
1212 #endif
1213 }
1214 
1215 /*
1216  * Handle interface watchdog timer routines.  Called
1217  * from softclock, we decrement timers (if set) and
1218  * call the appropriate interface routine on expiration.
1219  */
1220 void
1221 if_slowtimo(arg)
1222 	void *arg;
1223 {
1224 	struct ifnet *ifp;
1225 	int s = splnet();
1226 
1227 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1228 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1229 		if (ifp->if_timer == 0 || --ifp->if_timer)
1230 			continue;
1231 		if (ifp->if_watchdog)
1232 			(*ifp->if_watchdog)(ifp);
1233 	}
1234 	splx(s);
1235 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1236 	    if_slowtimo, NULL);
1237 }
1238 
1239 /*
1240  * Set/clear promiscuous mode on interface ifp based on the truth value
1241  * of pswitch.  The calls are reference counted so that only the first
1242  * "on" request actually has an effect, as does the final "off" request.
1243  * Results are undefined if the "off" and "on" requests are not matched.
1244  */
1245 int
1246 ifpromisc(ifp, pswitch)
1247 	struct ifnet *ifp;
1248 	int pswitch;
1249 {
1250 	int pcount, ret;
1251 	short flags;
1252 	struct ifreq ifr;
1253 
1254 	pcount = ifp->if_pcount;
1255 	flags = ifp->if_flags;
1256 	if (pswitch) {
1257 		/*
1258 		 * Allow the device to be "placed" into promiscuous
1259 		 * mode even if it is not configured up.  It will
1260 		 * consult IFF_PROMISC when it is is brought up.
1261 		 */
1262 		if (ifp->if_pcount++ != 0)
1263 			return (0);
1264 		ifp->if_flags |= IFF_PROMISC;
1265 		if ((ifp->if_flags & IFF_UP) == 0)
1266 			return (0);
1267 	} else {
1268 		if (--ifp->if_pcount > 0)
1269 			return (0);
1270 		ifp->if_flags &= ~IFF_PROMISC;
1271 		/*
1272 		 * If the device is not configured up, we should not need to
1273 		 * turn off promiscuous mode (device should have turned it
1274 		 * off when interface went down; and will look at IFF_PROMISC
1275 		 * again next time interface comes up).
1276 		 */
1277 		if ((ifp->if_flags & IFF_UP) == 0)
1278 			return (0);
1279 	}
1280 	memset(&ifr, 0, sizeof(ifr));
1281 	ifr.ifr_flags = ifp->if_flags;
1282 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1283 	/* Restore interface state if not successful. */
1284 	if (ret != 0) {
1285 		ifp->if_pcount = pcount;
1286 		ifp->if_flags = flags;
1287 	}
1288 	return (ret);
1289 }
1290 
1291 /*
1292  * Map interface name to
1293  * interface structure pointer.
1294  */
1295 struct ifnet *
1296 ifunit(name)
1297 	const char *name;
1298 {
1299 	struct ifnet *ifp;
1300 	const char *cp = name;
1301 	u_int unit = 0;
1302 	u_int i;
1303 
1304 	/*
1305 	 * If the entire name is a number, treat it as an ifindex.
1306 	 */
1307 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1308 		unit = unit * 10 + (*cp - '0');
1309 	}
1310 
1311 	/*
1312 	 * If the number took all of the name, then it's a valid ifindex.
1313 	 */
1314 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1315 		if (unit >= if_indexlim)
1316 			return (NULL);
1317 		ifp = ifindex2ifnet[unit];
1318 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1319 			return (NULL);
1320 		return (ifp);
1321 	}
1322 
1323 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1324 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1325 		if (ifp->if_output == if_nulloutput)
1326 			continue;
1327 	 	if (strcmp(ifp->if_xname, name) == 0)
1328 			return (ifp);
1329 	}
1330 	return (NULL);
1331 }
1332 
1333 /*
1334  * Interface ioctls.
1335  */
1336 int
1337 ifioctl(so, cmd, data, p)
1338 	struct socket *so;
1339 	u_long cmd;
1340 	caddr_t data;
1341 	struct proc *p;
1342 {
1343 	struct ifnet *ifp;
1344 	struct ifreq *ifr;
1345 	struct ifcapreq *ifcr;
1346 	struct ifdatareq *ifdr;
1347 	int s, error = 0;
1348 	short oif_flags;
1349 
1350 	switch (cmd) {
1351 
1352 	case SIOCGIFCONF:
1353 	case OSIOCGIFCONF:
1354 		return (ifconf(cmd, data));
1355 	}
1356 	ifr = (struct ifreq *)data;
1357 	ifcr = (struct ifcapreq *)data;
1358 	ifdr = (struct ifdatareq *)data;
1359 
1360 	switch (cmd) {
1361 	case SIOCIFCREATE:
1362 	case SIOCIFDESTROY:
1363 		if (p) {
1364 			error = suser(p->p_ucred, &p->p_acflag);
1365 			if (error)
1366 				return error;
1367 		}
1368 		return ((cmd == SIOCIFCREATE) ?
1369 			if_clone_create(ifr->ifr_name) :
1370 			if_clone_destroy(ifr->ifr_name));
1371 
1372 	case SIOCIFGCLONERS:
1373 		return (if_clone_list((struct if_clonereq *)data));
1374 	}
1375 
1376 	ifp = ifunit(ifr->ifr_name);
1377 	if (ifp == 0)
1378 		return (ENXIO);
1379 
1380 	switch (cmd) {
1381 	case SIOCSIFFLAGS:
1382 	case SIOCSIFCAP:
1383 	case SIOCSIFMETRIC:
1384 	case SIOCZIFDATA:
1385 	case SIOCSIFMTU:
1386 	case SIOCSIFPHYADDR:
1387 	case SIOCDIFPHYADDR:
1388 #ifdef INET6
1389 	case SIOCSIFPHYADDR_IN6:
1390 #endif
1391 	case SIOCSLIFPHYADDR:
1392 	case SIOCADDMULTI:
1393 	case SIOCDELMULTI:
1394 	case SIOCSIFMEDIA:
1395 	case SIOCSDRVSPEC:
1396 	case SIOCS80211NWID:
1397 	case SIOCS80211NWKEY:
1398 	case SIOCS80211POWER:
1399 	case SIOCS80211BSSID:
1400 	case SIOCS80211CHANNEL:
1401 		if (p) {
1402 			error = suser(p->p_ucred, &p->p_acflag);
1403 			if (error)
1404 				return error;
1405 		}
1406 	}
1407 
1408 	oif_flags = ifp->if_flags;
1409 	switch (cmd) {
1410 
1411 	case SIOCGIFFLAGS:
1412 		ifr->ifr_flags = ifp->if_flags;
1413 		break;
1414 
1415 	case SIOCGIFMETRIC:
1416 		ifr->ifr_metric = ifp->if_metric;
1417 		break;
1418 
1419 	case SIOCGIFMTU:
1420 		ifr->ifr_mtu = ifp->if_mtu;
1421 		break;
1422 
1423 	case SIOCGIFDLT:
1424 		ifr->ifr_dlt = ifp->if_dlt;
1425 		break;
1426 
1427 	case SIOCSIFFLAGS:
1428 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1429 			s = splnet();
1430 			if_down(ifp);
1431 			splx(s);
1432 		}
1433 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1434 			s = splnet();
1435 			if_up(ifp);
1436 			splx(s);
1437 		}
1438 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1439 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1440 		if (ifp->if_ioctl)
1441 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1442 		break;
1443 
1444 	case SIOCGIFCAP:
1445 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1446 		ifcr->ifcr_capenable = ifp->if_capenable;
1447 		break;
1448 
1449 	case SIOCSIFCAP:
1450 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1451 			return (EINVAL);
1452 		if (ifp->if_ioctl == NULL)
1453 			return (EOPNOTSUPP);
1454 
1455 		/* Must prevent race with packet reception here. */
1456 		s = splnet();
1457 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1458 			struct ifreq ifrq;
1459 
1460 			ifrq.ifr_flags = ifp->if_flags;
1461 			ifp->if_capenable = ifcr->ifcr_capenable;
1462 
1463 			/* Pre-compute the checksum flags mask. */
1464 			ifp->if_csum_flags_tx = 0;
1465 			ifp->if_csum_flags_rx = 0;
1466 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1467 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1468 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1469 			}
1470 
1471 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1472 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1473 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1474 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1475 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1476 
1477 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1478 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1479 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1480 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1481 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1482 
1483 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1484 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1485 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1486 			}
1487 
1488 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1489 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1490 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1491 			}
1492 
1493 			/*
1494 			 * Only kick the interface if it's up.  If it's
1495 			 * not up now, it will notice the cap enables
1496 			 * when it is brought up later.
1497 			 */
1498 			if (ifp->if_flags & IFF_UP)
1499 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1500 				    (caddr_t) &ifrq);
1501 		}
1502 		splx(s);
1503 		break;
1504 
1505 	case SIOCSIFMETRIC:
1506 		ifp->if_metric = ifr->ifr_metric;
1507 		break;
1508 
1509 	case SIOCGIFDATA:
1510 		ifdr->ifdr_data = ifp->if_data;
1511 		break;
1512 
1513 	case SIOCZIFDATA:
1514 		ifdr->ifdr_data = ifp->if_data;
1515 		/*
1516 		 * Assumes that the volatile counters that can be
1517 		 * zero'ed are at the end of if_data.
1518 		 */
1519 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1520 		    offsetof(struct if_data, ifi_ipackets));
1521 		break;
1522 
1523 	case SIOCSIFMTU:
1524 	{
1525 		u_long oldmtu = ifp->if_mtu;
1526 
1527 		if (ifp->if_ioctl == NULL)
1528 			return (EOPNOTSUPP);
1529 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1530 
1531 		/*
1532 		 * If the link MTU changed, do network layer specific procedure.
1533 		 */
1534 		if (ifp->if_mtu != oldmtu) {
1535 #ifdef INET6
1536 			nd6_setmtu(ifp);
1537 #endif
1538 		}
1539 		break;
1540 	}
1541 	case SIOCSIFPHYADDR:
1542 	case SIOCDIFPHYADDR:
1543 #ifdef INET6
1544 	case SIOCSIFPHYADDR_IN6:
1545 #endif
1546 	case SIOCSLIFPHYADDR:
1547 	case SIOCADDMULTI:
1548 	case SIOCDELMULTI:
1549 	case SIOCSIFMEDIA:
1550 	case SIOCGIFPSRCADDR:
1551 	case SIOCGIFPDSTADDR:
1552 	case SIOCGLIFPHYADDR:
1553 	case SIOCGIFMEDIA:
1554 		if (ifp->if_ioctl == 0)
1555 			return (EOPNOTSUPP);
1556 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1557 		break;
1558 
1559 	case SIOCSDRVSPEC:
1560 	case SIOCS80211NWID:
1561 	case SIOCS80211NWKEY:
1562 	case SIOCS80211POWER:
1563 	case SIOCS80211BSSID:
1564 	case SIOCS80211CHANNEL:
1565 	default:
1566 		if (so->so_proto == 0)
1567 			return (EOPNOTSUPP);
1568 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(COMPAT_ULTRIX) && !defined(LKM)
1569 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1570 		    (struct mbuf *)cmd, (struct mbuf *)data,
1571 		    (struct mbuf *)ifp, p));
1572 #else
1573 	    {
1574 		int ocmd = cmd;
1575 
1576 		switch (cmd) {
1577 
1578 		case SIOCSIFADDR:
1579 		case SIOCSIFDSTADDR:
1580 		case SIOCSIFBRDADDR:
1581 		case SIOCSIFNETMASK:
1582 #if BYTE_ORDER != BIG_ENDIAN
1583 			if (ifr->ifr_addr.sa_family == 0 &&
1584 			    ifr->ifr_addr.sa_len < 16) {
1585 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1586 				ifr->ifr_addr.sa_len = 16;
1587 			}
1588 #else
1589 			if (ifr->ifr_addr.sa_len == 0)
1590 				ifr->ifr_addr.sa_len = 16;
1591 #endif
1592 			break;
1593 
1594 		case OSIOCGIFADDR:
1595 			cmd = SIOCGIFADDR;
1596 			break;
1597 
1598 		case OSIOCGIFDSTADDR:
1599 			cmd = SIOCGIFDSTADDR;
1600 			break;
1601 
1602 		case OSIOCGIFBRDADDR:
1603 			cmd = SIOCGIFBRDADDR;
1604 			break;
1605 
1606 		case OSIOCGIFNETMASK:
1607 			cmd = SIOCGIFNETMASK;
1608 		}
1609 
1610 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1611 		    (struct mbuf *)cmd, (struct mbuf *)data,
1612 		    (struct mbuf *)ifp, p));
1613 
1614 		switch (ocmd) {
1615 		case OSIOCGIFADDR:
1616 		case OSIOCGIFDSTADDR:
1617 		case OSIOCGIFBRDADDR:
1618 		case OSIOCGIFNETMASK:
1619 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1620 		}
1621 	    }
1622 #endif /* COMPAT_43 */
1623 		break;
1624 	}
1625 
1626 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1627 #ifdef INET6
1628 		if ((ifp->if_flags & IFF_UP) != 0) {
1629 			s = splnet();
1630 			in6_if_up(ifp);
1631 			splx(s);
1632 		}
1633 #endif
1634 	}
1635 
1636 	return (error);
1637 }
1638 
1639 /*
1640  * Return interface configuration
1641  * of system.  List may be used
1642  * in later ioctl's (above) to get
1643  * other information.
1644  */
1645 /*ARGSUSED*/
1646 int
1647 ifconf(cmd, data)
1648 	u_long cmd;
1649 	caddr_t data;
1650 {
1651 	struct ifconf *ifc = (struct ifconf *)data;
1652 	struct ifnet *ifp;
1653 	struct ifaddr *ifa;
1654 	struct ifreq ifr, *ifrp;
1655 	int space = ifc->ifc_len, error = 0;
1656 	const int sz = (int)sizeof(ifr);
1657 	int sign;
1658 
1659 	if ((ifrp = ifc->ifc_req) == NULL) {
1660 		space = 0;
1661 		sign = -1;
1662 	} else {
1663 		sign = 1;
1664 	}
1665 	IFNET_FOREACH(ifp) {
1666 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1667 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1668 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1669 			if (ifrp != NULL && space >= sz) {
1670 				error = copyout(&ifr, ifrp, sz);
1671 				if (error)
1672 					break;
1673 				ifrp++;
1674 			}
1675 			space -= sizeof(ifr) * sign;
1676 			continue;
1677 		}
1678 
1679 		for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1680 			struct sockaddr *sa = ifa->ifa_addr;
1681 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX)
1682 			if (cmd == OSIOCGIFCONF) {
1683 				struct osockaddr *osa =
1684 					 (struct osockaddr *)&ifr.ifr_addr;
1685 				/*
1686 				 * If it does not fit, we don't bother with it
1687 				 */
1688 				if (sa->sa_len > sizeof(*osa))
1689 					continue;
1690 				ifr.ifr_addr = *sa;
1691 				osa->sa_family = sa->sa_family;
1692 				if (ifrp != NULL && space >= sz) {
1693 					error = copyout(&ifr, ifrp, sz);
1694 					ifrp++;
1695 				}
1696 			} else
1697 #endif
1698 			if (sa->sa_len <= sizeof(*sa)) {
1699 				ifr.ifr_addr = *sa;
1700 				if (ifrp != NULL && space >= sz) {
1701 					error = copyout(&ifr, ifrp, sz);
1702 					ifrp++;
1703 				}
1704 			} else {
1705 				space -= (sa->sa_len - sizeof(*sa)) * sign;
1706 				if (ifrp != NULL && space >= sz) {
1707 					error = copyout(&ifr, ifrp,
1708 					    sizeof(ifr.ifr_name));
1709 					if (error == 0) {
1710 						error = copyout(sa,
1711 						    &ifrp->ifr_addr,
1712 						    sa->sa_len);
1713 					}
1714 					ifrp = (struct ifreq *)
1715 						(sa->sa_len +
1716 						 (caddr_t)&ifrp->ifr_addr);
1717 				}
1718 			}
1719 			if (error)
1720 				break;
1721 			space -= sz * sign;
1722 		}
1723 	}
1724 	if (ifrp != NULL)
1725 		ifc->ifc_len -= space;
1726 	else
1727 		ifc->ifc_len = space;
1728 	return (error);
1729 }
1730 
1731 #if defined(INET) || defined(INET6)
1732 static void
1733 sysctl_net_ifq_setup(struct sysctllog **clog,
1734 		     int pf, const char *pfname,
1735 		     int ipn, const char *ipname,
1736 		     int qid, struct ifqueue *ifq)
1737 {
1738 
1739 	sysctl_createv(clog, 0, NULL, NULL,
1740 		       CTLFLAG_PERMANENT,
1741 		       CTLTYPE_NODE, "net", NULL,
1742 		       NULL, 0, NULL, 0,
1743 		       CTL_NET, CTL_EOL);
1744 	sysctl_createv(clog, 0, NULL, NULL,
1745 		       CTLFLAG_PERMANENT,
1746 		       CTLTYPE_NODE, pfname, NULL,
1747 		       NULL, 0, NULL, 0,
1748 		       CTL_NET, pf, CTL_EOL);
1749 	sysctl_createv(clog, 0, NULL, NULL,
1750 		       CTLFLAG_PERMANENT,
1751 		       CTLTYPE_NODE, ipname, NULL,
1752 		       NULL, 0, NULL, 0,
1753 		       CTL_NET, pf, ipn, CTL_EOL);
1754 	sysctl_createv(clog, 0, NULL, NULL,
1755 		       CTLFLAG_PERMANENT,
1756 		       CTLTYPE_NODE, "ifq",
1757 		       SYSCTL_DESCR("Protocol input queue controls"),
1758 		       NULL, 0, NULL, 0,
1759 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1760 
1761 	sysctl_createv(clog, 0, NULL, NULL,
1762 		       CTLFLAG_PERMANENT,
1763 		       CTLTYPE_INT, "len",
1764 		       SYSCTL_DESCR("Current input queue length"),
1765 		       NULL, 0, &ifq->ifq_len, 0,
1766 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1767 	sysctl_createv(clog, 0, NULL, NULL,
1768 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1769 		       CTLTYPE_INT, "maxlen",
1770 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1771 		       NULL, 0, &ifq->ifq_maxlen, 0,
1772 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1773 #ifdef notyet
1774 	sysctl_createv(clog, 0, NULL, NULL,
1775 		       CTLFLAG_PERMANENT,
1776 		       CTLTYPE_INT, "peak",
1777 		       SYSCTL_DESCR("Highest input queue length"),
1778 		       NULL, 0, &ifq->ifq_peak, 0,
1779 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1780 #endif
1781 	sysctl_createv(clog, 0, NULL, NULL,
1782 		       CTLFLAG_PERMANENT,
1783 		       CTLTYPE_INT, "drops",
1784 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1785 		       NULL, 0, &ifq->ifq_drops, 0,
1786 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1787 }
1788 
1789 #ifdef INET
1790 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1791 	     "sysctl net.inet.ip.ifq subtree setup")
1792 {
1793 	extern struct ifqueue ipintrq;
1794 
1795 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1796 			     IPCTL_IFQ, &ipintrq);
1797 }
1798 #endif /* INET */
1799 
1800 #ifdef INET6
1801 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1802 	     "sysctl net.inet6.ip6.ifq subtree setup")
1803 {
1804 	extern struct ifqueue ip6intrq;
1805 
1806 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1807 			     IPV6CTL_IFQ, &ip6intrq);
1808 }
1809 #endif /* INET6 */
1810 #endif /* INET || INET6 */
1811