xref: /netbsd-src/sys/net/if.c (revision cd22f25e6f6d1cc1f197fe8c5468a80f51d1c4e1)
1 /*	$NetBSD: if.c,v 1.222 2008/04/29 18:42:26 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000, 2001, 2008 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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the project nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  */
60 
61 /*
62  * Copyright (c) 1980, 1986, 1993
63  *	The Regents of the University of California.  All rights reserved.
64  *
65  * Redistribution and use in source and binary forms, with or without
66  * modification, are permitted provided that the following conditions
67  * are met:
68  * 1. Redistributions of source code must retain the above copyright
69  *    notice, this list of conditions and the following disclaimer.
70  * 2. Redistributions in binary form must reproduce the above copyright
71  *    notice, this list of conditions and the following disclaimer in the
72  *    documentation and/or other materials provided with the distribution.
73  * 3. Neither the name of the University nor the names of its contributors
74  *    may be used to endorse or promote products derived from this software
75  *    without specific prior written permission.
76  *
77  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
78  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
79  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
80  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
81  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
82  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
83  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
84  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
85  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
86  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
87  * SUCH DAMAGE.
88  *
89  *	@(#)if.c	8.5 (Berkeley) 1/9/95
90  */
91 
92 #include <sys/cdefs.h>
93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.222 2008/04/29 18:42:26 ad Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 #include "opt_pfil_hooks.h"
100 
101 #include <sys/param.h>
102 #include <sys/mbuf.h>
103 #include <sys/systm.h>
104 #include <sys/callout.h>
105 #include <sys/proc.h>
106 #include <sys/socket.h>
107 #include <sys/socketvar.h>
108 #include <sys/domain.h>
109 #include <sys/protosw.h>
110 #include <sys/kernel.h>
111 #include <sys/ioctl.h>
112 #include <sys/sysctl.h>
113 #include <sys/syslog.h>
114 #include <sys/kauth.h>
115 
116 #include <net/if.h>
117 #include <net/if_dl.h>
118 #include <net/if_ether.h>
119 #include <net/if_media.h>
120 #include <net80211/ieee80211.h>
121 #include <net80211/ieee80211_ioctl.h>
122 #include <net/if_types.h>
123 #include <net/radix.h>
124 #include <net/route.h>
125 #include <net/netisr.h>
126 #ifdef NETATALK
127 #include <netatalk/at_extern.h>
128 #include <netatalk/at.h>
129 #endif
130 #include <net/pfil.h>
131 
132 #ifdef INET6
133 #include <netinet/in.h>
134 #include <netinet6/in6_var.h>
135 #include <netinet6/nd6.h>
136 #endif
137 
138 #include "carp.h"
139 #if NCARP > 0
140 #include <netinet/ip_carp.h>
141 #endif
142 
143 #include <compat/sys/sockio.h>
144 #include <compat/sys/socket.h>
145 
146 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
147 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
148 
149 int	ifqmaxlen = IFQ_MAXLEN;
150 callout_t if_slowtimo_ch;
151 
152 int netisr;			/* scheduling bits for network */
153 
154 static int	if_rt_walktree(struct rtentry *, void *);
155 
156 static struct if_clone *if_clone_lookup(const char *, int *);
157 static int	if_clone_list(struct if_clonereq *);
158 
159 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
160 static int if_cloners_count;
161 
162 #ifdef PFIL_HOOKS
163 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
164 #endif
165 
166 static void if_detach_queues(struct ifnet *, struct ifqueue *);
167 
168 /*
169  * Network interface utility routines.
170  *
171  * Routines with ifa_ifwith* names take sockaddr *'s as
172  * parameters.
173  */
174 void
175 ifinit(void)
176 {
177 
178 	callout_init(&if_slowtimo_ch, 0);
179 	if_slowtimo(NULL);
180 #ifdef PFIL_HOOKS
181 	if_pfil.ph_type = PFIL_TYPE_IFNET;
182 	if_pfil.ph_ifnet = NULL;
183 	if (pfil_head_register(&if_pfil) != 0)
184 		printf("WARNING: unable to register pfil hook\n");
185 #endif
186 }
187 
188 /*
189  * Null routines used while an interface is going away.  These routines
190  * just return an error.
191  */
192 
193 int
194 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
195     const struct sockaddr *so, struct rtentry *rt)
196 {
197 
198 	return ENXIO;
199 }
200 
201 void
202 if_nullinput(struct ifnet *ifp, struct mbuf *m)
203 {
204 
205 	/* Nothing. */
206 }
207 
208 void
209 if_nullstart(struct ifnet *ifp)
210 {
211 
212 	/* Nothing. */
213 }
214 
215 int
216 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
217 {
218 
219 	return ENXIO;
220 }
221 
222 int
223 if_nullinit(struct ifnet *ifp)
224 {
225 
226 	return ENXIO;
227 }
228 
229 void
230 if_nullstop(struct ifnet *ifp, int disable)
231 {
232 
233 	/* Nothing. */
234 }
235 
236 void
237 if_nullwatchdog(struct ifnet *ifp)
238 {
239 
240 	/* Nothing. */
241 }
242 
243 void
244 if_nulldrain(struct ifnet *ifp)
245 {
246 
247 	/* Nothing. */
248 }
249 
250 static u_int if_index = 1;
251 struct ifnet_head ifnet;
252 size_t if_indexlim = 0;
253 struct ifaddr **ifnet_addrs = NULL;
254 struct ifnet **ifindex2ifnet = NULL;
255 struct ifnet *lo0ifp;
256 
257 void
258 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen)
259 {
260 	struct ifaddr *ifa;
261 	struct sockaddr_dl *sdl;
262 
263 	ifp->if_addrlen = addrlen;
264 	if_alloc_sadl(ifp);
265 	ifa = ifp->if_dl;
266 	sdl = satosdl(ifa->ifa_addr);
267 
268 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
269 }
270 
271 struct ifaddr *
272 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
273 {
274 	unsigned socksize, ifasize;
275 	int addrlen, namelen;
276 	struct sockaddr_dl *mask, *sdl;
277 	struct ifaddr *ifa;
278 
279 	namelen = strlen(ifp->if_xname);
280 	addrlen = ifp->if_addrlen;
281 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
282 	ifasize = sizeof(*ifa) + 2 * socksize;
283 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
284 
285 	sdl = (struct sockaddr_dl *)(ifa + 1);
286 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
287 
288 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
289 	    ifp->if_xname, namelen, NULL, addrlen);
290 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
291 	memset(&mask->sdl_data[0], 0xff, namelen);
292 	ifa->ifa_rtrequest = link_rtrequest;
293 	ifa->ifa_addr = (struct sockaddr *)sdl;
294 	ifa->ifa_netmask = (struct sockaddr *)mask;
295 
296 	*sdlp = sdl;
297 
298 	return ifa;
299 }
300 
301 /*
302  * Allocate the link level name for the specified interface.  This
303  * is an attachment helper.  It must be called after ifp->if_addrlen
304  * is initialized, which may not be the case when if_attach() is
305  * called.
306  */
307 void
308 if_alloc_sadl(struct ifnet *ifp)
309 {
310 	struct ifaddr *ifa;
311 	const struct sockaddr_dl *sdl;
312 
313 	/*
314 	 * If the interface already has a link name, release it
315 	 * now.  This is useful for interfaces that can change
316 	 * link types, and thus switch link names often.
317 	 */
318 	if (ifp->if_sadl != NULL)
319 		if_free_sadl(ifp);
320 
321 	ifa = if_dl_create(ifp, &sdl);
322 
323 	ifnet_addrs[ifp->if_index] = ifa;
324 	IFAREF(ifa);
325 	ifa_insert(ifp, ifa);
326 	ifp->if_dl = ifa;
327 	IFAREF(ifa);
328 	ifp->if_sadl = sdl;
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 		KASSERT(ifp->if_dl == NULL);
346 		return;
347 	}
348 
349 	KASSERT(ifp->if_sadl != NULL);
350 	KASSERT(ifp->if_dl != NULL);
351 
352 	s = splnet();
353 	rtinit(ifa, RTM_DELETE, 0);
354 	ifa_remove(ifp, ifa);
355 
356 	ifp->if_sadl = NULL;
357 
358 	ifnet_addrs[ifp->if_index] = NULL;
359 	IFAFREE(ifa);
360 	ifp->if_dl = NULL;
361 	IFAFREE(ifa);
362 	splx(s);
363 }
364 
365 /*
366  * Attach an interface to the
367  * list of "active" interfaces.
368  */
369 void
370 if_attach(struct ifnet *ifp)
371 {
372 	int indexlim = 0;
373 
374 	if (if_indexlim == 0) {
375 		TAILQ_INIT(&ifnet);
376 		if_indexlim = 8;
377 	}
378 	TAILQ_INIT(&ifp->if_addrlist);
379 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
380 	ifp->if_index = if_index;
381 	if (ifindex2ifnet == NULL)
382 		if_index++;
383 	else
384 		while (ifp->if_index < if_indexlim &&
385 		    ifindex2ifnet[ifp->if_index] != NULL) {
386 			++if_index;
387 			if (if_index == 0)
388 				if_index = 1;
389 			/*
390 			 * If we hit USHRT_MAX, we skip back to 0 since
391 			 * there are a number of places where the value
392 			 * of if_index or if_index itself is compared
393 			 * to or stored in an unsigned short.  By
394 			 * jumping back, we won't botch those assignments
395 			 * or comparisons.
396 			 */
397 			else if (if_index == USHRT_MAX) {
398 				/*
399 				 * However, if we have to jump back to
400 				 * zero *twice* without finding an empty
401 				 * slot in ifindex2ifnet[], then there
402 				 * there are too many (>65535) interfaces.
403 				 */
404 				if (indexlim++)
405 					panic("too many interfaces");
406 				else
407 					if_index = 1;
408 			}
409 			ifp->if_index = if_index;
410 		}
411 
412 	/*
413 	 * We have some arrays that should be indexed by if_index.
414 	 * since if_index will grow dynamically, they should grow too.
415 	 *	struct ifadd **ifnet_addrs
416 	 *	struct ifnet **ifindex2ifnet
417 	 */
418 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
419 	    ifp->if_index >= if_indexlim) {
420 		size_t m, n, oldlim;
421 		void *q;
422 
423 		oldlim = if_indexlim;
424 		while (ifp->if_index >= if_indexlim)
425 			if_indexlim <<= 1;
426 
427 		/* grow ifnet_addrs */
428 		m = oldlim * sizeof(struct ifaddr *);
429 		n = if_indexlim * sizeof(struct ifaddr *);
430 		q = (void *)malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
431 		if (ifnet_addrs != NULL) {
432 			memcpy(q, ifnet_addrs, m);
433 			free((void *)ifnet_addrs, M_IFADDR);
434 		}
435 		ifnet_addrs = (struct ifaddr **)q;
436 
437 		/* grow ifindex2ifnet */
438 		m = oldlim * sizeof(struct ifnet *);
439 		n = if_indexlim * sizeof(struct ifnet *);
440 		q = (void *)malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
441 		if (ifindex2ifnet != NULL) {
442 			memcpy(q, (void *)ifindex2ifnet, m);
443 			free((void *)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(void)
492 {
493 	struct ifnet *ifp;
494 	int s;
495 
496 	s = splnet();
497 	IFNET_FOREACH(ifp)
498 		if_attachdomain1(ifp);
499 	splx(s);
500 }
501 
502 void
503 if_attachdomain1(struct ifnet *ifp)
504 {
505 	struct domain *dp;
506 	int s;
507 
508 	s = splnet();
509 
510 	/* address family dependent data region */
511 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
512 	DOMAIN_FOREACH(dp) {
513 		if (dp->dom_ifattach != NULL)
514 			ifp->if_afdata[dp->dom_family] =
515 			    (*dp->dom_ifattach)(ifp);
516 	}
517 
518 	splx(s);
519 }
520 
521 /*
522  * Deactivate an interface.  This points all of the procedure
523  * handles at error stubs.  May be called from interrupt context.
524  */
525 void
526 if_deactivate(struct ifnet *ifp)
527 {
528 	int s;
529 
530 	s = splnet();
531 
532 	ifp->if_output	 = if_nulloutput;
533 	ifp->if_input	 = if_nullinput;
534 	ifp->if_start	 = if_nullstart;
535 	ifp->if_ioctl	 = if_nullioctl;
536 	ifp->if_init	 = if_nullinit;
537 	ifp->if_stop	 = if_nullstop;
538 	ifp->if_watchdog = if_nullwatchdog;
539 	ifp->if_drain	 = if_nulldrain;
540 
541 	/* No more packets may be enqueued. */
542 	ifp->if_snd.ifq_maxlen = 0;
543 
544 	splx(s);
545 }
546 
547 void
548 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
549 {
550 	struct ifaddr *ifa, *nifa;
551 
552 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
553 		nifa = IFADDR_NEXT(ifa);
554 		if (ifa->ifa_addr->sa_family != family)
555 			continue;
556 		(*purgeaddr)(ifa);
557 	}
558 }
559 
560 /*
561  * Detach an interface from the list of "active" interfaces,
562  * freeing any resources as we go along.
563  *
564  * NOTE: This routine must be called with a valid thread context,
565  * as it may block.
566  */
567 void
568 if_detach(struct ifnet *ifp)
569 {
570 	struct socket so;
571 	struct ifaddr *ifa;
572 #ifdef IFAREF_DEBUG
573 	struct ifaddr *last_ifa = NULL;
574 #endif
575 	struct domain *dp;
576 	const struct protosw *pr;
577 	int s, i, family, purged;
578 
579 	/*
580 	 * XXX It's kind of lame that we have to have the
581 	 * XXX socket structure...
582 	 */
583 	memset(&so, 0, sizeof(so));
584 
585 	s = splnet();
586 
587 	/*
588 	 * Do an if_down() to give protocols a chance to do something.
589 	 */
590 	if_down(ifp);
591 
592 #ifdef ALTQ
593 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
594 		altq_disable(&ifp->if_snd);
595 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
596 		altq_detach(&ifp->if_snd);
597 #endif
598 
599 
600 #if NCARP > 0
601 	/* Remove the interface from any carp group it is a part of.  */
602 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
603 		carp_ifdetach(ifp);
604 #endif
605 
606 	/*
607 	 * Rip all the addresses off the interface.  This should make
608 	 * all of the routes go away.
609 	 *
610 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
611 	 * from the list, including our "cursor", ifa.  For safety,
612 	 * and to honor the TAILQ abstraction, I just restart the
613 	 * loop after each removal.  Note that the loop will exit
614 	 * when all of the remaining ifaddrs belong to the AF_LINK
615 	 * family.  I am counting on the historical fact that at
616 	 * least one pr_usrreq in each address domain removes at
617 	 * least one ifaddr.
618 	 */
619 again:
620 	IFADDR_FOREACH(ifa, ifp) {
621 		family = ifa->ifa_addr->sa_family;
622 #ifdef IFAREF_DEBUG
623 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
624 		    ifa, family, ifa->ifa_refcnt);
625 		if (last_ifa != NULL && ifa == last_ifa)
626 			panic("if_detach: loop detected");
627 		last_ifa = ifa;
628 #endif
629 		if (family == AF_LINK)
630 			continue;
631 		dp = pffinddomain(family);
632 #ifdef DIAGNOSTIC
633 		if (dp == NULL)
634 			panic("if_detach: no domain for AF %d",
635 			    family);
636 #endif
637 		/*
638 		 * XXX These PURGEIF calls are redundant with the
639 		 * purge-all-families calls below, but are left in for
640 		 * now both to make a smaller change, and to avoid
641 		 * unplanned interactions with clearing of
642 		 * ifp->if_addrlist.
643 		 */
644 		purged = 0;
645 		for (pr = dp->dom_protosw;
646 		     pr < dp->dom_protoswNPROTOSW; pr++) {
647 			so.so_proto = pr;
648 			if (pr->pr_usrreq != NULL) {
649 				(void) (*pr->pr_usrreq)(&so,
650 				    PRU_PURGEIF, NULL, NULL,
651 				    (struct mbuf *) ifp, curlwp);
652 				purged = 1;
653 			}
654 		}
655 		if (purged == 0) {
656 			/*
657 			 * XXX What's really the best thing to do
658 			 * XXX here?  --thorpej@NetBSD.org
659 			 */
660 			printf("if_detach: WARNING: AF %d not purged\n",
661 			    family);
662 			ifa_remove(ifp, ifa);
663 		}
664 		goto again;
665 	}
666 
667 	if_free_sadl(ifp);
668 
669 	/* Walk the routing table looking for stragglers. */
670 	for (i = 0; i <= AF_MAX; i++)
671 		(void)rt_walktree(i, if_rt_walktree, ifp);
672 
673 	DOMAIN_FOREACH(dp) {
674 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
675 			(*dp->dom_ifdetach)(ifp,
676 			    ifp->if_afdata[dp->dom_family]);
677 
678 		/*
679 		 * One would expect multicast memberships (INET and
680 		 * INET6) on UDP sockets to be purged by the PURGEIF
681 		 * calls above, but if all addresses were removed from
682 		 * the interface prior to destruction, the calls will
683 		 * not be made (e.g. ppp, for which pppd(8) generally
684 		 * removes addresses before destroying the interface).
685 		 * Because there is no invariant that multicast
686 		 * memberships only exist for interfaces with IPv4
687 		 * addresses, we must call PURGEIF regardless of
688 		 * addresses.  (Protocols which might store ifnet
689 		 * pointers are marked with PR_PURGEIF.)
690 		 */
691 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
692 			so.so_proto = pr;
693 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
694 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
695 				    NULL, (struct mbuf *)ifp, curlwp);
696 		}
697 	}
698 
699 #ifdef PFIL_HOOKS
700 	(void)pfil_run_hooks(&if_pfil,
701 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
702 	(void)pfil_head_unregister(&ifp->if_pfil);
703 #endif
704 
705 	/* Announce that the interface is gone. */
706 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
707 
708 	ifindex2ifnet[ifp->if_index] = NULL;
709 
710 	TAILQ_REMOVE(&ifnet, ifp, if_list);
711 
712 	/*
713 	 * remove packets that came from ifp, from software interrupt queues.
714 	 */
715 	DOMAIN_FOREACH(dp) {
716 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
717 			if (dp->dom_ifqueues[i] == NULL)
718 				break;
719 			if_detach_queues(ifp, dp->dom_ifqueues[i]);
720 		}
721 	}
722 
723 	splx(s);
724 }
725 
726 static void
727 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
728 {
729 	struct mbuf *m, *prev, *next;
730 
731 	prev = NULL;
732 	for (m = q->ifq_head; m != NULL; m = next) {
733 		next = m->m_nextpkt;
734 #ifdef DIAGNOSTIC
735 		if ((m->m_flags & M_PKTHDR) == 0) {
736 			prev = m;
737 			continue;
738 		}
739 #endif
740 		if (m->m_pkthdr.rcvif != ifp) {
741 			prev = m;
742 			continue;
743 		}
744 
745 		if (prev != NULL)
746 			prev->m_nextpkt = m->m_nextpkt;
747 		else
748 			q->ifq_head = m->m_nextpkt;
749 		if (q->ifq_tail == m)
750 			q->ifq_tail = prev;
751 		q->ifq_len--;
752 
753 		m->m_nextpkt = NULL;
754 		m_freem(m);
755 		IF_DROP(q);
756 	}
757 }
758 
759 /*
760  * Callback for a radix tree walk to delete all references to an
761  * ifnet.
762  */
763 static int
764 if_rt_walktree(struct rtentry *rt, void *v)
765 {
766 	struct ifnet *ifp = (struct ifnet *)v;
767 	int error;
768 
769 	if (rt->rt_ifp != ifp)
770 		return 0;
771 
772 	/* Delete the entry. */
773 	++rt->rt_refcnt;
774 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
775 	    rt_mask(rt), rt->rt_flags, NULL);
776 	KASSERT((rt->rt_flags & RTF_UP) == 0);
777 	rt->rt_ifp = NULL;
778 	RTFREE(rt);
779 	if (error != 0)
780 		printf("%s: warning: unable to delete rtentry @ %p, "
781 		    "error = %d\n", ifp->if_xname, rt, error);
782 	return 0;
783 }
784 
785 /*
786  * Create a clone network interface.
787  */
788 int
789 if_clone_create(const char *name)
790 {
791 	struct if_clone *ifc;
792 	int unit;
793 
794 	ifc = if_clone_lookup(name, &unit);
795 	if (ifc == NULL)
796 		return EINVAL;
797 
798 	if (ifunit(name) != NULL)
799 		return EEXIST;
800 
801 	return (*ifc->ifc_create)(ifc, unit);
802 }
803 
804 /*
805  * Destroy a clone network interface.
806  */
807 int
808 if_clone_destroy(const char *name)
809 {
810 	struct if_clone *ifc;
811 	struct ifnet *ifp;
812 
813 	ifc = if_clone_lookup(name, NULL);
814 	if (ifc == NULL)
815 		return EINVAL;
816 
817 	ifp = ifunit(name);
818 	if (ifp == NULL)
819 		return ENXIO;
820 
821 	if (ifc->ifc_destroy == NULL)
822 		return EOPNOTSUPP;
823 
824 	return (*ifc->ifc_destroy)(ifp);
825 }
826 
827 /*
828  * Look up a network interface cloner.
829  */
830 static struct if_clone *
831 if_clone_lookup(const char *name, int *unitp)
832 {
833 	struct if_clone *ifc;
834 	const char *cp;
835 	int unit;
836 
837 	/* separate interface name from unit */
838 	for (cp = name;
839 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
840 	    cp++)
841 		continue;
842 
843 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
844 		return NULL;	/* No name or unit number */
845 
846 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
847 		if (strlen(ifc->ifc_name) == cp - name &&
848 		    strncmp(name, ifc->ifc_name, cp - name) == 0)
849 			break;
850 	}
851 
852 	if (ifc == NULL)
853 		return NULL;
854 
855 	unit = 0;
856 	while (cp - name < IFNAMSIZ && *cp) {
857 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
858 			/* Bogus unit number. */
859 			return NULL;
860 		}
861 		unit = (unit * 10) + (*cp++ - '0');
862 	}
863 
864 	if (unitp != NULL)
865 		*unitp = unit;
866 	return ifc;
867 }
868 
869 /*
870  * Register a network interface cloner.
871  */
872 void
873 if_clone_attach(struct if_clone *ifc)
874 {
875 
876 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
877 	if_cloners_count++;
878 }
879 
880 /*
881  * Unregister a network interface cloner.
882  */
883 void
884 if_clone_detach(struct if_clone *ifc)
885 {
886 
887 	LIST_REMOVE(ifc, ifc_list);
888 	if_cloners_count--;
889 }
890 
891 /*
892  * Provide list of interface cloners to userspace.
893  */
894 static int
895 if_clone_list(struct if_clonereq *ifcr)
896 {
897 	char outbuf[IFNAMSIZ], *dst;
898 	struct if_clone *ifc;
899 	int count, error = 0;
900 
901 	ifcr->ifcr_total = if_cloners_count;
902 	if ((dst = ifcr->ifcr_buffer) == NULL) {
903 		/* Just asking how many there are. */
904 		return 0;
905 	}
906 
907 	if (ifcr->ifcr_count < 0)
908 		return EINVAL;
909 
910 	count = (if_cloners_count < ifcr->ifcr_count) ?
911 	    if_cloners_count : ifcr->ifcr_count;
912 
913 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
914 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
915 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
916 		if (outbuf[sizeof(outbuf) - 1] != '\0')
917 			return ENAMETOOLONG;
918 		error = copyout(outbuf, dst, sizeof(outbuf));
919 		if (error != 0)
920 			break;
921 	}
922 
923 	return error;
924 }
925 
926 void
927 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
928 {
929 	ifa->ifa_ifp = ifp;
930 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
931 	IFAREF(ifa);
932 }
933 
934 void
935 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
936 {
937 	KASSERT(ifa->ifa_ifp == ifp);
938 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
939 	IFAFREE(ifa);
940 }
941 
942 static inline int
943 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
944 {
945 	return sockaddr_cmp(sa1, sa2) == 0;
946 }
947 
948 /*
949  * Locate an interface based on a complete address.
950  */
951 /*ARGSUSED*/
952 struct ifaddr *
953 ifa_ifwithaddr(const struct sockaddr *addr)
954 {
955 	struct ifnet *ifp;
956 	struct ifaddr *ifa;
957 
958 	IFNET_FOREACH(ifp) {
959 		if (ifp->if_output == if_nulloutput)
960 			continue;
961 		IFADDR_FOREACH(ifa, ifp) {
962 			if (ifa->ifa_addr->sa_family != addr->sa_family)
963 				continue;
964 			if (equal(addr, ifa->ifa_addr))
965 				return ifa;
966 			if ((ifp->if_flags & IFF_BROADCAST) &&
967 			    ifa->ifa_broadaddr &&
968 			    /* IP6 doesn't have broadcast */
969 			    ifa->ifa_broadaddr->sa_len != 0 &&
970 			    equal(ifa->ifa_broadaddr, addr))
971 				return ifa;
972 		}
973 	}
974 	return NULL;
975 }
976 
977 /*
978  * Locate the point to point interface with a given destination address.
979  */
980 /*ARGSUSED*/
981 struct ifaddr *
982 ifa_ifwithdstaddr(const struct sockaddr *addr)
983 {
984 	struct ifnet *ifp;
985 	struct ifaddr *ifa;
986 
987 	IFNET_FOREACH(ifp) {
988 		if (ifp->if_output == if_nulloutput)
989 			continue;
990 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
991 			continue;
992 		IFADDR_FOREACH(ifa, ifp) {
993 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
994 			    ifa->ifa_dstaddr == NULL)
995 				continue;
996 			if (equal(addr, ifa->ifa_dstaddr))
997 				return ifa;
998 		}
999 	}
1000 	return NULL;
1001 }
1002 
1003 /*
1004  * Find an interface on a specific network.  If many, choice
1005  * is most specific found.
1006  */
1007 struct ifaddr *
1008 ifa_ifwithnet(const struct sockaddr *addr)
1009 {
1010 	struct ifnet *ifp;
1011 	struct ifaddr *ifa;
1012 	const struct sockaddr_dl *sdl;
1013 	struct ifaddr *ifa_maybe = 0;
1014 	u_int af = addr->sa_family;
1015 	const char *addr_data = addr->sa_data, *cplim;
1016 
1017 	if (af == AF_LINK) {
1018 		sdl = satocsdl(addr);
1019 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1020 		    ifindex2ifnet[sdl->sdl_index] &&
1021 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1022 			return ifnet_addrs[sdl->sdl_index];
1023 	}
1024 #ifdef NETATALK
1025 	if (af == AF_APPLETALK) {
1026 		const struct sockaddr_at *sat, *sat2;
1027 		sat = (const struct sockaddr_at *)addr;
1028 		IFNET_FOREACH(ifp) {
1029 			if (ifp->if_output == if_nulloutput)
1030 				continue;
1031 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1032 			if (ifa == NULL)
1033 				continue;
1034 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1035 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1036 				return ifa; /* exact match */
1037 			if (ifa_maybe == NULL) {
1038 				/* else keep the if with the right range */
1039 				ifa_maybe = ifa;
1040 			}
1041 		}
1042 		return ifa_maybe;
1043 	}
1044 #endif
1045 	IFNET_FOREACH(ifp) {
1046 		if (ifp->if_output == if_nulloutput)
1047 			continue;
1048 		IFADDR_FOREACH(ifa, ifp) {
1049 			const char *cp, *cp2, *cp3;
1050 
1051 			if (ifa->ifa_addr->sa_family != af ||
1052 			    ifa->ifa_netmask == NULL)
1053  next:				continue;
1054 			cp = addr_data;
1055 			cp2 = ifa->ifa_addr->sa_data;
1056 			cp3 = ifa->ifa_netmask->sa_data;
1057 			cplim = (const char *)ifa->ifa_netmask +
1058 			    ifa->ifa_netmask->sa_len;
1059 			while (cp3 < cplim) {
1060 				if ((*cp++ ^ *cp2++) & *cp3++) {
1061 					/* want to continue for() loop */
1062 					goto next;
1063 				}
1064 			}
1065 			if (ifa_maybe == NULL ||
1066 			    rn_refines((void *)ifa->ifa_netmask,
1067 			    (void *)ifa_maybe->ifa_netmask))
1068 				ifa_maybe = ifa;
1069 		}
1070 	}
1071 	return ifa_maybe;
1072 }
1073 
1074 /*
1075  * Find the interface of the addresss.
1076  */
1077 struct ifaddr *
1078 ifa_ifwithladdr(const struct sockaddr *addr)
1079 {
1080 	struct ifaddr *ia;
1081 
1082 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1083 	    (ia = ifa_ifwithnet(addr)))
1084 		return ia;
1085 	return NULL;
1086 }
1087 
1088 /*
1089  * Find an interface using a specific address family
1090  */
1091 struct ifaddr *
1092 ifa_ifwithaf(int af)
1093 {
1094 	struct ifnet *ifp;
1095 	struct ifaddr *ifa;
1096 
1097 	IFNET_FOREACH(ifp) {
1098 		if (ifp->if_output == if_nulloutput)
1099 			continue;
1100 		IFADDR_FOREACH(ifa, ifp) {
1101 			if (ifa->ifa_addr->sa_family == af)
1102 				return ifa;
1103 		}
1104 	}
1105 	return NULL;
1106 }
1107 
1108 /*
1109  * Find an interface address specific to an interface best matching
1110  * a given address.
1111  */
1112 struct ifaddr *
1113 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1114 {
1115 	struct ifaddr *ifa;
1116 	const char *cp, *cp2, *cp3;
1117 	const char *cplim;
1118 	struct ifaddr *ifa_maybe = 0;
1119 	u_int af = addr->sa_family;
1120 
1121 	if (ifp->if_output == if_nulloutput)
1122 		return NULL;
1123 
1124 	if (af >= AF_MAX)
1125 		return NULL;
1126 
1127 	IFADDR_FOREACH(ifa, ifp) {
1128 		if (ifa->ifa_addr->sa_family != af)
1129 			continue;
1130 		ifa_maybe = ifa;
1131 		if (ifa->ifa_netmask == NULL) {
1132 			if (equal(addr, ifa->ifa_addr) ||
1133 			    (ifa->ifa_dstaddr &&
1134 			     equal(addr, ifa->ifa_dstaddr)))
1135 				return ifa;
1136 			continue;
1137 		}
1138 		cp = addr->sa_data;
1139 		cp2 = ifa->ifa_addr->sa_data;
1140 		cp3 = ifa->ifa_netmask->sa_data;
1141 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1142 		for (; cp3 < cplim; cp3++) {
1143 			if ((*cp++ ^ *cp2++) & *cp3)
1144 				break;
1145 		}
1146 		if (cp3 == cplim)
1147 			return ifa;
1148 	}
1149 	return ifa_maybe;
1150 }
1151 
1152 /*
1153  * Default action when installing a route with a Link Level gateway.
1154  * Lookup an appropriate real ifa to point to.
1155  * This should be moved to /sys/net/link.c eventually.
1156  */
1157 void
1158 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1159 {
1160 	struct ifaddr *ifa;
1161 	const struct sockaddr *dst;
1162 	struct ifnet *ifp;
1163 
1164 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
1165 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_getkey(rt)) == NULL))
1166 		return;
1167 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1168 		rt_replace_ifa(rt, ifa);
1169 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1170 			ifa->ifa_rtrequest(cmd, rt, info);
1171 	}
1172 }
1173 
1174 /*
1175  * Handle a change in the interface link state.
1176  */
1177 void
1178 if_link_state_change(struct ifnet *ifp, int link_state)
1179 {
1180 	if (ifp->if_link_state == link_state)
1181 		return;
1182 	ifp->if_link_state = link_state;
1183 	/* Notify that the link state has changed. */
1184 	rt_ifmsg(ifp);
1185 #if NCARP > 0
1186 	if (ifp->if_carp)
1187 		carp_carpdev_state(ifp);
1188 #endif
1189 }
1190 
1191 /*
1192  * Mark an interface down and notify protocols of
1193  * the transition.
1194  * NOTE: must be called at splsoftnet or equivalent.
1195  */
1196 void
1197 if_down(struct ifnet *ifp)
1198 {
1199 	struct ifaddr *ifa;
1200 
1201 	ifp->if_flags &= ~IFF_UP;
1202 	microtime(&ifp->if_lastchange);
1203 	IFADDR_FOREACH(ifa, ifp)
1204 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1205 	IFQ_PURGE(&ifp->if_snd);
1206 #if NCARP > 0
1207 	if (ifp->if_carp)
1208 		carp_carpdev_state(ifp);
1209 #endif
1210 	rt_ifmsg(ifp);
1211 }
1212 
1213 /*
1214  * Mark an interface up and notify protocols of
1215  * the transition.
1216  * NOTE: must be called at splsoftnet or equivalent.
1217  */
1218 void
1219 if_up(struct ifnet *ifp)
1220 {
1221 #ifdef notyet
1222 	struct ifaddr *ifa;
1223 #endif
1224 
1225 	ifp->if_flags |= IFF_UP;
1226 	microtime(&ifp->if_lastchange);
1227 #ifdef notyet
1228 	/* this has no effect on IP, and will kill all ISO connections XXX */
1229 	IFADDR_FOREACH(ifa, ifp)
1230 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1231 #endif
1232 #if NCARP > 0
1233 	if (ifp->if_carp)
1234 		carp_carpdev_state(ifp);
1235 #endif
1236 	rt_ifmsg(ifp);
1237 #ifdef INET6
1238 	in6_if_up(ifp);
1239 #endif
1240 }
1241 
1242 /*
1243  * Handle interface watchdog timer routines.  Called
1244  * from softclock, we decrement timers (if set) and
1245  * call the appropriate interface routine on expiration.
1246  */
1247 void
1248 if_slowtimo(void *arg)
1249 {
1250 	struct ifnet *ifp;
1251 	int s = splnet();
1252 
1253 	IFNET_FOREACH(ifp) {
1254 		if (ifp->if_timer == 0 || --ifp->if_timer)
1255 			continue;
1256 		if (ifp->if_watchdog != NULL)
1257 			(*ifp->if_watchdog)(ifp);
1258 	}
1259 	splx(s);
1260 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1261 }
1262 
1263 /*
1264  * Set/clear promiscuous mode on interface ifp based on the truth value
1265  * of pswitch.  The calls are reference counted so that only the first
1266  * "on" request actually has an effect, as does the final "off" request.
1267  * Results are undefined if the "off" and "on" requests are not matched.
1268  */
1269 int
1270 ifpromisc(struct ifnet *ifp, int pswitch)
1271 {
1272 	int pcount, ret;
1273 	short flags;
1274 	struct ifreq ifr;
1275 
1276 	pcount = ifp->if_pcount;
1277 	flags = ifp->if_flags;
1278 	if (pswitch) {
1279 		/*
1280 		 * Allow the device to be "placed" into promiscuous
1281 		 * mode even if it is not configured up.  It will
1282 		 * consult IFF_PROMISC when it is is brought up.
1283 		 */
1284 		if (ifp->if_pcount++ != 0)
1285 			return 0;
1286 		ifp->if_flags |= IFF_PROMISC;
1287 		if ((ifp->if_flags & IFF_UP) == 0)
1288 			return 0;
1289 	} else {
1290 		if (--ifp->if_pcount > 0)
1291 			return 0;
1292 		ifp->if_flags &= ~IFF_PROMISC;
1293 		/*
1294 		 * If the device is not configured up, we should not need to
1295 		 * turn off promiscuous mode (device should have turned it
1296 		 * off when interface went down; and will look at IFF_PROMISC
1297 		 * again next time interface comes up).
1298 		 */
1299 		if ((ifp->if_flags & IFF_UP) == 0)
1300 			return 0;
1301 	}
1302 	memset(&ifr, 0, sizeof(ifr));
1303 	ifr.ifr_flags = ifp->if_flags;
1304 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (void *) &ifr);
1305 	/* Restore interface state if not successful. */
1306 	if (ret != 0) {
1307 		ifp->if_pcount = pcount;
1308 		ifp->if_flags = flags;
1309 	}
1310 	return ret;
1311 }
1312 
1313 /*
1314  * Map interface name to
1315  * interface structure pointer.
1316  */
1317 struct ifnet *
1318 ifunit(const char *name)
1319 {
1320 	struct ifnet *ifp;
1321 	const char *cp = name;
1322 	u_int unit = 0;
1323 	u_int i;
1324 
1325 	/*
1326 	 * If the entire name is a number, treat it as an ifindex.
1327 	 */
1328 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1329 		unit = unit * 10 + (*cp - '0');
1330 	}
1331 
1332 	/*
1333 	 * If the number took all of the name, then it's a valid ifindex.
1334 	 */
1335 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1336 		if (unit >= if_indexlim)
1337 			return NULL;
1338 		ifp = ifindex2ifnet[unit];
1339 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1340 			return NULL;
1341 		return ifp;
1342 	}
1343 
1344 	IFNET_FOREACH(ifp) {
1345 		if (ifp->if_output == if_nulloutput)
1346 			continue;
1347 	 	if (strcmp(ifp->if_xname, name) == 0)
1348 			return ifp;
1349 	}
1350 	return NULL;
1351 }
1352 
1353 /* common */
1354 int
1355 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1356 {
1357 	int s;
1358 	struct ifreq *ifr;
1359 	struct ifcapreq *ifcr;
1360 	struct ifdatareq *ifdr;
1361 
1362 	switch (cmd) {
1363 	case SIOCSIFCAP:
1364 		ifcr = data;
1365 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1366 			return EINVAL;
1367 
1368 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1369 			return 0;
1370 
1371 		ifp->if_capenable = ifcr->ifcr_capenable;
1372 
1373 		/* Pre-compute the checksum flags mask. */
1374 		ifp->if_csum_flags_tx = 0;
1375 		ifp->if_csum_flags_rx = 0;
1376 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1377 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1378 		}
1379 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1380 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1381 		}
1382 
1383 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1384 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1385 		}
1386 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1387 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1388 		}
1389 
1390 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1391 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1392 		}
1393 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1394 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1395 		}
1396 
1397 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1398 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1399 		}
1400 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1401 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1402 		}
1403 
1404 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1405 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1406 		}
1407 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1408 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1409 		}
1410 		if (ifp->if_flags & IFF_UP)
1411 			return ENETRESET;
1412 		return 0;
1413 	case SIOCSIFFLAGS:
1414 		ifr = data;
1415 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1416 			s = splnet();
1417 			if_down(ifp);
1418 			splx(s);
1419 		}
1420 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1421 			s = splnet();
1422 			if_up(ifp);
1423 			splx(s);
1424 		}
1425 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1426 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1427 		break;
1428 	case SIOCGIFFLAGS:
1429 		ifr = data;
1430 		ifr->ifr_flags = ifp->if_flags;
1431 		break;
1432 
1433 	case SIOCGIFMETRIC:
1434 		ifr = data;
1435 		ifr->ifr_metric = ifp->if_metric;
1436 		break;
1437 
1438 	case SIOCGIFMTU:
1439 		ifr = data;
1440 		ifr->ifr_mtu = ifp->if_mtu;
1441 		break;
1442 
1443 	case SIOCGIFDLT:
1444 		ifr = data;
1445 		ifr->ifr_dlt = ifp->if_dlt;
1446 		break;
1447 
1448 	case SIOCGIFCAP:
1449 		ifcr = data;
1450 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1451 		ifcr->ifcr_capenable = ifp->if_capenable;
1452 		break;
1453 
1454 	case SIOCSIFMETRIC:
1455 		ifr = data;
1456 		ifp->if_metric = ifr->ifr_metric;
1457 		break;
1458 
1459 	case SIOCGIFDATA:
1460 		ifdr = data;
1461 		ifdr->ifdr_data = ifp->if_data;
1462 		break;
1463 
1464 	case SIOCZIFDATA:
1465 		ifdr = data;
1466 		ifdr->ifdr_data = ifp->if_data;
1467 		/*
1468 		 * Assumes that the volatile counters that can be
1469 		 * zero'ed are at the end of if_data.
1470 		 */
1471 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1472 		    offsetof(struct if_data, ifi_ipackets));
1473 		break;
1474 	case SIOCSIFMTU:
1475 		ifr = data;
1476 		if (ifp->if_mtu == ifr->ifr_mtu)
1477 			break;
1478 		ifp->if_mtu = ifr->ifr_mtu;
1479 		/*
1480 		 * If the link MTU changed, do network layer specific procedure.
1481 		 */
1482 #ifdef INET6
1483 		nd6_setmtu(ifp);
1484 #endif
1485 		return ENETRESET;
1486 	default:
1487 		return EOPNOTSUPP;
1488 	}
1489 	return 0;
1490 }
1491 
1492 /*
1493  * Interface ioctls.
1494  */
1495 int
1496 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1497 {
1498 	struct ifnet *ifp;
1499 	struct ifreq *ifr;
1500 	struct ifcapreq *ifcr;
1501 	struct ifdatareq *ifdr;
1502 	int error = 0;
1503 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1504 	u_long ocmd = cmd;
1505 #endif
1506 	short oif_flags;
1507 #ifdef COMPAT_OIFREQ
1508 	struct ifreq ifrb;
1509 	struct oifreq *oifr = NULL;
1510 #endif
1511 
1512 	switch (cmd) {
1513 #ifdef COMPAT_OIFREQ
1514 	case OSIOCGIFCONF:
1515 	case OOSIOCGIFCONF:
1516 		return compat_ifconf(cmd, data);
1517 #endif
1518 	case SIOCGIFCONF:
1519 		return ifconf(cmd, data);
1520 	}
1521 
1522 #ifdef COMPAT_OIFREQ
1523 	cmd = compat_cvtcmd(cmd);
1524 	if (cmd != ocmd) {
1525 		oifr = data;
1526 		data = ifr = &ifrb;
1527 		ifreqo2n(oifr, ifr);
1528 	} else
1529 #endif
1530 		ifr = data;
1531 	ifcr = data;
1532 	ifdr = data;
1533 
1534 	ifp = ifunit(ifr->ifr_name);
1535 
1536 	switch (cmd) {
1537 	case SIOCIFCREATE:
1538 	case SIOCIFDESTROY:
1539 		if (l != NULL) {
1540 			error = kauth_authorize_network(l->l_cred,
1541 			    KAUTH_NETWORK_INTERFACE,
1542 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1543 			    (void *)cmd, NULL);
1544 			if (error != 0)
1545 				return error;
1546 		}
1547 		return (cmd == SIOCIFCREATE) ?
1548 			if_clone_create(ifr->ifr_name) :
1549 			if_clone_destroy(ifr->ifr_name);
1550 
1551 	case SIOCIFGCLONERS:
1552 		return if_clone_list((struct if_clonereq *)data);
1553 	}
1554 
1555 	if (ifp == NULL)
1556 		return ENXIO;
1557 
1558 	switch (cmd) {
1559 	case SIOCSIFFLAGS:
1560 	case SIOCSIFCAP:
1561 	case SIOCSIFMETRIC:
1562 	case SIOCZIFDATA:
1563 	case SIOCSIFMTU:
1564 	case SIOCSIFPHYADDR:
1565 	case SIOCDIFPHYADDR:
1566 #ifdef INET6
1567 	case SIOCSIFPHYADDR_IN6:
1568 #endif
1569 	case SIOCSLIFPHYADDR:
1570 	case SIOCADDMULTI:
1571 	case SIOCDELMULTI:
1572 	case SIOCSIFMEDIA:
1573 	case SIOCSDRVSPEC:
1574 	case SIOCG80211:
1575 	case SIOCS80211:
1576 	case SIOCS80211NWID:
1577 	case SIOCS80211NWKEY:
1578 	case SIOCS80211POWER:
1579 	case SIOCS80211BSSID:
1580 	case SIOCS80211CHANNEL:
1581 		if (l != NULL) {
1582 			error = kauth_authorize_network(l->l_cred,
1583 			    KAUTH_NETWORK_INTERFACE,
1584 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1585 			    (void *)cmd, NULL);
1586 			if (error != 0)
1587 				return error;
1588 		}
1589 	}
1590 
1591 	oif_flags = ifp->if_flags;
1592 	switch (cmd) {
1593 
1594 	case SIOCSIFFLAGS:
1595 		ifioctl_common(ifp, cmd, data);
1596 		if (ifp->if_ioctl)
1597 			(void)(*ifp->if_ioctl)(ifp, cmd, data);
1598 		break;
1599 
1600 	case SIOCSIFPHYADDR:
1601 	case SIOCDIFPHYADDR:
1602 #ifdef INET6
1603 	case SIOCSIFPHYADDR_IN6:
1604 #endif
1605 	case SIOCSLIFPHYADDR:
1606 	case SIOCADDMULTI:
1607 	case SIOCDELMULTI:
1608 	case SIOCSIFMEDIA:
1609 	case SIOCGIFPSRCADDR:
1610 	case SIOCGIFPDSTADDR:
1611 	case SIOCGLIFPHYADDR:
1612 	case SIOCGIFMEDIA:
1613 	case SIOCG80211:
1614 	case SIOCS80211:
1615 	case SIOCS80211NWID:
1616 	case SIOCS80211NWKEY:
1617 	case SIOCS80211POWER:
1618 	case SIOCS80211BSSID:
1619 	case SIOCS80211CHANNEL:
1620 	case SIOCSIFCAP:
1621 	case SIOCSIFMTU:
1622 		if (ifp->if_ioctl == NULL)
1623 			return EOPNOTSUPP;
1624 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1625 		break;
1626 
1627 	default:
1628 		error = ifioctl_common(ifp, cmd, data);
1629 		if (error != EOPNOTSUPP)
1630 			break;
1631 		if (so->so_proto == NULL)
1632 			return EOPNOTSUPP;
1633 #ifdef COMPAT_OSOCK
1634 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1635 #else
1636 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1637 		    (struct mbuf *)cmd, (struct mbuf *)data,
1638 		    (struct mbuf *)ifp, l);
1639 #endif
1640 		break;
1641 	}
1642 
1643 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1644 #ifdef INET6
1645 		if ((ifp->if_flags & IFF_UP) != 0) {
1646 			int s = splnet();
1647 			in6_if_up(ifp);
1648 			splx(s);
1649 		}
1650 #endif
1651 	}
1652 #ifdef COMPAT_OIFREQ
1653 	if (cmd != ocmd)
1654 		ifreqn2o(oifr, ifr);
1655 #endif
1656 
1657 	return error;
1658 }
1659 
1660 /*
1661  * Return interface configuration
1662  * of system.  List may be used
1663  * in later ioctl's (above) to get
1664  * other information.
1665  *
1666  * Each record is a struct ifreq.  Before the addition of
1667  * sockaddr_storage, the API rule was that sockaddr flavors that did
1668  * not fit would extend beyond the struct ifreq, with the next struct
1669  * ifreq starting sa_len beyond the struct sockaddr.  Because the
1670  * union in struct ifreq includes struct sockaddr_storage, every kind
1671  * of sockaddr must fit.  Thus, there are no longer any overlength
1672  * records.
1673  *
1674  * Records are added to the user buffer if they fit, and ifc_len is
1675  * adjusted to the length that was written.  Thus, the user is only
1676  * assured of getting the complete list if ifc_len on return is at
1677  * least sizeof(struct ifreq) less than it was on entry.
1678  *
1679  * If the user buffer pointer is NULL, this routine copies no data and
1680  * returns the amount of space that would be needed.
1681  *
1682  * Invariants:
1683  * ifrp points to the next part of the user's buffer to be used.  If
1684  * ifrp != NULL, space holds the number of bytes remaining that we may
1685  * write at ifrp.  Otherwise, space holds the number of bytes that
1686  * would have been written had there been adequate space.
1687  */
1688 /*ARGSUSED*/
1689 int
1690 ifconf(u_long cmd, void *data)
1691 {
1692 	struct ifconf *ifc = (struct ifconf *)data;
1693 	struct ifnet *ifp;
1694 	struct ifaddr *ifa;
1695 	struct ifreq ifr, *ifrp;
1696 	int space, error = 0;
1697 	const int sz = (int)sizeof(struct ifreq);
1698 
1699 	if ((ifrp = ifc->ifc_req) == NULL)
1700 		space = 0;
1701 	else
1702 		space = ifc->ifc_len;
1703 	IFNET_FOREACH(ifp) {
1704 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1705 		    sizeof(ifr.ifr_name));
1706 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1707 			return ENAMETOOLONG;
1708 		if (IFADDR_EMPTY(ifp)) {
1709 			/* Interface with no addresses - send zero sockaddr. */
1710 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1711 			if (ifrp == NULL) {
1712 				space += sz;
1713 				continue;
1714 			}
1715 			if (space >= sz) {
1716 				error = copyout(&ifr, ifrp, sz);
1717 				if (error != 0)
1718 					return error;
1719 				ifrp++;
1720 				space -= sz;
1721 			}
1722 		}
1723 
1724 		IFADDR_FOREACH(ifa, ifp) {
1725 			struct sockaddr *sa = ifa->ifa_addr;
1726 			/* all sockaddrs must fit in sockaddr_storage */
1727 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
1728 
1729 			if (ifrp == NULL) {
1730 				space += sz;
1731 				continue;
1732 			}
1733 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
1734 			if (space >= sz) {
1735 				error = copyout(&ifr, ifrp, sz);
1736 				if (error != 0)
1737 					return (error);
1738 				ifrp++; space -= sz;
1739 			}
1740 		}
1741 	}
1742 	if (ifrp != NULL) {
1743 		KASSERT(0 <= space && space <= ifc->ifc_len);
1744 		ifc->ifc_len -= space;
1745 	} else {
1746 		KASSERT(space >= 0);
1747 		ifc->ifc_len = space;
1748 	}
1749 	return (0);
1750 }
1751 
1752 int
1753 ifreq_setaddr(const u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
1754 {
1755 	uint8_t len;
1756 	u_long ncmd;
1757 	const uint8_t osockspace = sizeof(ifr->ifr_addr);
1758 	const uint8_t sockspace = sizeof(ifr->ifr_ifru.ifru_space);
1759 
1760 #ifdef INET6
1761 	if (cmd == SIOCGIFPSRCADDR_IN6 || cmd == SIOCGIFPDSTADDR_IN6)
1762 		len = MIN(sizeof(struct sockaddr_in6), sa->sa_len);
1763 	else
1764 #endif /* INET6 */
1765 	if ((ncmd = compat_cvtcmd(cmd)) != cmd)
1766 		len = MIN(osockspace, sa->sa_len);
1767 	else
1768 		len = MIN(sockspace, sa->sa_len);
1769 	if (len < sa->sa_len)
1770 		return EFBIG;
1771 	sockaddr_copy(&ifr->ifr_addr, len, sa);
1772 	return 0;
1773 }
1774 
1775 /*
1776  * Queue message on interface, and start output if interface
1777  * not yet active.
1778  */
1779 int
1780 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
1781     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1782 {
1783 	int len = m->m_pkthdr.len;
1784 	int mflags = m->m_flags;
1785 	int s = splnet();
1786 	int error;
1787 
1788 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1789 	if (error != 0)
1790 		goto out;
1791 	ifp->if_obytes += len;
1792 	if (mflags & M_MCAST)
1793 		ifp->if_omcasts++;
1794 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
1795 		(*ifp->if_start)(ifp);
1796 out:
1797 	splx(s);
1798 	return error;
1799 }
1800 
1801 /*
1802  * Queue message on interface, possibly using a second fast queue
1803  */
1804 int
1805 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
1806     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1807 {
1808 	int error = 0;
1809 
1810 	if (ifq != NULL
1811 #ifdef ALTQ
1812 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
1813 #endif
1814 	    ) {
1815 		if (IF_QFULL(ifq)) {
1816 			IF_DROP(&ifp->if_snd);
1817 			m_freem(m);
1818 			if (error == 0)
1819 				error = ENOBUFS;
1820 		} else
1821 			IF_ENQUEUE(ifq, m);
1822 	} else
1823 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1824 	if (error != 0) {
1825 		++ifp->if_oerrors;
1826 		return error;
1827 	}
1828 	return 0;
1829 }
1830 
1831 
1832 #if defined(INET) || defined(INET6)
1833 static void
1834 sysctl_net_ifq_setup(struct sysctllog **clog,
1835 		     int pf, const char *pfname,
1836 		     int ipn, const char *ipname,
1837 		     int qid, struct ifqueue *ifq)
1838 {
1839 
1840 	sysctl_createv(clog, 0, NULL, NULL,
1841 		       CTLFLAG_PERMANENT,
1842 		       CTLTYPE_NODE, "net", NULL,
1843 		       NULL, 0, NULL, 0,
1844 		       CTL_NET, CTL_EOL);
1845 	sysctl_createv(clog, 0, NULL, NULL,
1846 		       CTLFLAG_PERMANENT,
1847 		       CTLTYPE_NODE, pfname, NULL,
1848 		       NULL, 0, NULL, 0,
1849 		       CTL_NET, pf, CTL_EOL);
1850 	sysctl_createv(clog, 0, NULL, NULL,
1851 		       CTLFLAG_PERMANENT,
1852 		       CTLTYPE_NODE, ipname, NULL,
1853 		       NULL, 0, NULL, 0,
1854 		       CTL_NET, pf, ipn, CTL_EOL);
1855 	sysctl_createv(clog, 0, NULL, NULL,
1856 		       CTLFLAG_PERMANENT,
1857 		       CTLTYPE_NODE, "ifq",
1858 		       SYSCTL_DESCR("Protocol input queue controls"),
1859 		       NULL, 0, NULL, 0,
1860 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1861 
1862 	sysctl_createv(clog, 0, NULL, NULL,
1863 		       CTLFLAG_PERMANENT,
1864 		       CTLTYPE_INT, "len",
1865 		       SYSCTL_DESCR("Current input queue length"),
1866 		       NULL, 0, &ifq->ifq_len, 0,
1867 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1868 	sysctl_createv(clog, 0, NULL, NULL,
1869 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1870 		       CTLTYPE_INT, "maxlen",
1871 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1872 		       NULL, 0, &ifq->ifq_maxlen, 0,
1873 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1874 #ifdef notyet
1875 	sysctl_createv(clog, 0, NULL, NULL,
1876 		       CTLFLAG_PERMANENT,
1877 		       CTLTYPE_INT, "peak",
1878 		       SYSCTL_DESCR("Highest input queue length"),
1879 		       NULL, 0, &ifq->ifq_peak, 0,
1880 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1881 #endif
1882 	sysctl_createv(clog, 0, NULL, NULL,
1883 		       CTLFLAG_PERMANENT,
1884 		       CTLTYPE_INT, "drops",
1885 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1886 		       NULL, 0, &ifq->ifq_drops, 0,
1887 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1888 }
1889 
1890 #ifdef INET
1891 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1892 	     "sysctl net.inet.ip.ifq subtree setup")
1893 {
1894 	extern struct ifqueue ipintrq;
1895 
1896 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1897 			     IPCTL_IFQ, &ipintrq);
1898 }
1899 #endif /* INET */
1900 
1901 #ifdef INET6
1902 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1903 	     "sysctl net.inet6.ip6.ifq subtree setup")
1904 {
1905 	extern struct ifqueue ip6intrq;
1906 
1907 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1908 			     IPV6CTL_IFQ, &ip6intrq);
1909 }
1910 #endif /* INET6 */
1911 #endif /* INET || INET6 */
1912