xref: /netbsd-src/sys/net/if.c (revision 212397c69a103ae7e5eafa8731ddfae671d2dee7)
1 /*	$NetBSD: if.c,v 1.322 2016/01/21 15:41:29 riastradh 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 1.322 2016/01/21 15:41:29 riastradh Exp $");
94 
95 #if defined(_KERNEL_OPT)
96 #include "opt_inet.h"
97 
98 #include "opt_atalk.h"
99 #include "opt_natm.h"
100 #include "opt_wlan.h"
101 #include "opt_net_mpsafe.h"
102 #endif
103 
104 #include <sys/param.h>
105 #include <sys/mbuf.h>
106 #include <sys/systm.h>
107 #include <sys/callout.h>
108 #include <sys/proc.h>
109 #include <sys/socket.h>
110 #include <sys/socketvar.h>
111 #include <sys/domain.h>
112 #include <sys/protosw.h>
113 #include <sys/kernel.h>
114 #include <sys/ioctl.h>
115 #include <sys/sysctl.h>
116 #include <sys/syslog.h>
117 #include <sys/kauth.h>
118 #include <sys/kmem.h>
119 #include <sys/xcall.h>
120 
121 #include <net/if.h>
122 #include <net/if_dl.h>
123 #include <net/if_ether.h>
124 #include <net/if_media.h>
125 #include <net80211/ieee80211.h>
126 #include <net80211/ieee80211_ioctl.h>
127 #include <net/if_types.h>
128 #include <net/radix.h>
129 #include <net/route.h>
130 #include <net/netisr.h>
131 #include <sys/module.h>
132 #ifdef NETATALK
133 #include <netatalk/at_extern.h>
134 #include <netatalk/at.h>
135 #endif
136 #include <net/pfil.h>
137 #include <netinet/in.h>
138 #include <netinet/in_var.h>
139 
140 #ifdef INET6
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 #include "ether.h"
146 #include "fddi.h"
147 #include "token.h"
148 
149 #include "carp.h"
150 #if NCARP > 0
151 #include <netinet/ip_carp.h>
152 #endif
153 
154 #include <compat/sys/sockio.h>
155 #include <compat/sys/socket.h>
156 
157 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
158 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
159 
160 /*
161  * Global list of interfaces.
162  */
163 struct ifnet_head		ifnet_list;
164 static ifnet_t **		ifindex2ifnet = NULL;
165 
166 static u_int			if_index = 1;
167 static size_t			if_indexlim = 0;
168 static uint64_t			index_gen;
169 static kmutex_t			index_gen_mtx;
170 static kmutex_t			if_clone_mtx;
171 
172 struct ifnet *lo0ifp;
173 int	ifqmaxlen = IFQ_MAXLEN;
174 
175 static int	if_rt_walktree(struct rtentry *, void *);
176 
177 static struct if_clone *if_clone_lookup(const char *, int *);
178 
179 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
180 static int if_cloners_count;
181 
182 /* Packet filtering hook for interfaces. */
183 pfil_head_t *	if_pfil;
184 
185 static kauth_listener_t if_listener;
186 
187 static int doifioctl(struct socket *, u_long, void *, struct lwp *);
188 static int ifioctl_attach(struct ifnet *);
189 static void ifioctl_detach(struct ifnet *);
190 static void ifnet_lock_enter(struct ifnet_lock *);
191 static void ifnet_lock_exit(struct ifnet_lock *);
192 static void if_detach_queues(struct ifnet *, struct ifqueue *);
193 static void sysctl_sndq_setup(struct sysctllog **, const char *,
194     struct ifaltq *);
195 static void if_slowtimo(void *);
196 static void if_free_sadl(struct ifnet *);
197 static void if_attachdomain1(struct ifnet *);
198 static int ifconf(u_long, void *);
199 static int if_clone_create(const char *);
200 static int if_clone_destroy(const char *);
201 
202 #if defined(INET) || defined(INET6)
203 static void sysctl_net_pktq_setup(struct sysctllog **, int);
204 #endif
205 
206 static int
207 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
208     void *arg0, void *arg1, void *arg2, void *arg3)
209 {
210 	int result;
211 	enum kauth_network_req req;
212 
213 	result = KAUTH_RESULT_DEFER;
214 	req = (enum kauth_network_req)arg1;
215 
216 	if (action != KAUTH_NETWORK_INTERFACE)
217 		return result;
218 
219 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
220 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
221 		result = KAUTH_RESULT_ALLOW;
222 
223 	return result;
224 }
225 
226 /*
227  * Network interface utility routines.
228  *
229  * Routines with ifa_ifwith* names take sockaddr *'s as
230  * parameters.
231  */
232 void
233 ifinit(void)
234 {
235 #if defined(INET)
236 	sysctl_net_pktq_setup(NULL, PF_INET);
237 #endif
238 #ifdef INET6
239 	if (in6_present)
240 		sysctl_net_pktq_setup(NULL, PF_INET6);
241 #endif
242 
243 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
244 	    if_listener_cb, NULL);
245 
246 	/* interfaces are available, inform socket code */
247 	ifioctl = doifioctl;
248 }
249 
250 /*
251  * XXX Initialization before configure().
252  * XXX hack to get pfil_add_hook working in autoconf.
253  */
254 void
255 ifinit1(void)
256 {
257 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
258 	mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE);
259 	TAILQ_INIT(&ifnet_list);
260 	if_indexlim = 8;
261 
262 	if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
263 	KASSERT(if_pfil != NULL);
264 
265 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN)
266 	etherinit();
267 #endif
268 }
269 
270 ifnet_t *
271 if_alloc(u_char type)
272 {
273 	return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP);
274 }
275 
276 void
277 if_free(ifnet_t *ifp)
278 {
279 	kmem_free(ifp, sizeof(ifnet_t));
280 }
281 
282 void
283 if_initname(struct ifnet *ifp, const char *name, int unit)
284 {
285 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
286 	    "%s%d", name, unit);
287 }
288 
289 /*
290  * Null routines used while an interface is going away.  These routines
291  * just return an error.
292  */
293 
294 int
295 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
296     const struct sockaddr *so, struct rtentry *rt)
297 {
298 
299 	return ENXIO;
300 }
301 
302 void
303 if_nullinput(struct ifnet *ifp, struct mbuf *m)
304 {
305 
306 	/* Nothing. */
307 }
308 
309 void
310 if_nullstart(struct ifnet *ifp)
311 {
312 
313 	/* Nothing. */
314 }
315 
316 int
317 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
318 {
319 
320 	/* Wake ifioctl_detach(), who may wait for all threads to
321 	 * quit the critical section.
322 	 */
323 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
324 	return ENXIO;
325 }
326 
327 int
328 if_nullinit(struct ifnet *ifp)
329 {
330 
331 	return ENXIO;
332 }
333 
334 void
335 if_nullstop(struct ifnet *ifp, int disable)
336 {
337 
338 	/* Nothing. */
339 }
340 
341 void
342 if_nullslowtimo(struct ifnet *ifp)
343 {
344 
345 	/* Nothing. */
346 }
347 
348 void
349 if_nulldrain(struct ifnet *ifp)
350 {
351 
352 	/* Nothing. */
353 }
354 
355 void
356 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
357 {
358 	struct ifaddr *ifa;
359 	struct sockaddr_dl *sdl;
360 
361 	ifp->if_addrlen = addrlen;
362 	if_alloc_sadl(ifp);
363 	ifa = ifp->if_dl;
364 	sdl = satosdl(ifa->ifa_addr);
365 
366 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
367 	if (factory) {
368 		ifp->if_hwdl = ifp->if_dl;
369 		ifaref(ifp->if_hwdl);
370 	}
371 	/* TBD routing socket */
372 }
373 
374 struct ifaddr *
375 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
376 {
377 	unsigned socksize, ifasize;
378 	int addrlen, namelen;
379 	struct sockaddr_dl *mask, *sdl;
380 	struct ifaddr *ifa;
381 
382 	namelen = strlen(ifp->if_xname);
383 	addrlen = ifp->if_addrlen;
384 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
385 	ifasize = sizeof(*ifa) + 2 * socksize;
386 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
387 
388 	sdl = (struct sockaddr_dl *)(ifa + 1);
389 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
390 
391 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
392 	    ifp->if_xname, namelen, NULL, addrlen);
393 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
394 	memset(&mask->sdl_data[0], 0xff, namelen);
395 	ifa->ifa_rtrequest = link_rtrequest;
396 	ifa->ifa_addr = (struct sockaddr *)sdl;
397 	ifa->ifa_netmask = (struct sockaddr *)mask;
398 
399 	*sdlp = sdl;
400 
401 	return ifa;
402 }
403 
404 static void
405 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
406 {
407 	const struct sockaddr_dl *sdl;
408 
409 	ifp->if_dl = ifa;
410 	ifaref(ifa);
411 	sdl = satosdl(ifa->ifa_addr);
412 	ifp->if_sadl = sdl;
413 }
414 
415 /*
416  * Allocate the link level name for the specified interface.  This
417  * is an attachment helper.  It must be called after ifp->if_addrlen
418  * is initialized, which may not be the case when if_attach() is
419  * called.
420  */
421 void
422 if_alloc_sadl(struct ifnet *ifp)
423 {
424 	struct ifaddr *ifa;
425 	const struct sockaddr_dl *sdl;
426 
427 	/*
428 	 * If the interface already has a link name, release it
429 	 * now.  This is useful for interfaces that can change
430 	 * link types, and thus switch link names often.
431 	 */
432 	if (ifp->if_sadl != NULL)
433 		if_free_sadl(ifp);
434 
435 	ifa = if_dl_create(ifp, &sdl);
436 
437 	ifa_insert(ifp, ifa);
438 	if_sadl_setrefs(ifp, ifa);
439 }
440 
441 static void
442 if_deactivate_sadl(struct ifnet *ifp)
443 {
444 	struct ifaddr *ifa;
445 
446 	KASSERT(ifp->if_dl != NULL);
447 
448 	ifa = ifp->if_dl;
449 
450 	ifp->if_sadl = NULL;
451 
452 	ifp->if_dl = NULL;
453 	ifafree(ifa);
454 }
455 
456 void
457 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
458     const struct sockaddr_dl *sdl)
459 {
460 	int s;
461 
462 	s = splnet();
463 
464 	if_deactivate_sadl(ifp);
465 
466 	if_sadl_setrefs(ifp, ifa);
467 	IFADDR_FOREACH(ifa, ifp)
468 		rtinit(ifa, RTM_LLINFO_UPD, 0);
469 	splx(s);
470 }
471 
472 /*
473  * Free the link level name for the specified interface.  This is
474  * a detach helper.  This is called from if_detach().
475  */
476 static void
477 if_free_sadl(struct ifnet *ifp)
478 {
479 	struct ifaddr *ifa;
480 	int s;
481 
482 	ifa = ifp->if_dl;
483 	if (ifa == NULL) {
484 		KASSERT(ifp->if_sadl == NULL);
485 		return;
486 	}
487 
488 	KASSERT(ifp->if_sadl != NULL);
489 
490 	s = splnet();
491 	rtinit(ifa, RTM_DELETE, 0);
492 	ifa_remove(ifp, ifa);
493 	if_deactivate_sadl(ifp);
494 	if (ifp->if_hwdl == ifa) {
495 		ifafree(ifa);
496 		ifp->if_hwdl = NULL;
497 	}
498 	splx(s);
499 }
500 
501 static void
502 if_getindex(ifnet_t *ifp)
503 {
504 	bool hitlimit = false;
505 
506 	mutex_enter(&index_gen_mtx);
507 	ifp->if_index_gen = index_gen++;
508 	mutex_exit(&index_gen_mtx);
509 
510 	ifp->if_index = if_index;
511 	if (ifindex2ifnet == NULL) {
512 		if_index++;
513 		goto skip;
514 	}
515 	while (if_byindex(ifp->if_index)) {
516 		/*
517 		 * If we hit USHRT_MAX, we skip back to 0 since
518 		 * there are a number of places where the value
519 		 * of if_index or if_index itself is compared
520 		 * to or stored in an unsigned short.  By
521 		 * jumping back, we won't botch those assignments
522 		 * or comparisons.
523 		 */
524 		if (++if_index == 0) {
525 			if_index = 1;
526 		} else if (if_index == USHRT_MAX) {
527 			/*
528 			 * However, if we have to jump back to
529 			 * zero *twice* without finding an empty
530 			 * slot in ifindex2ifnet[], then there
531 			 * there are too many (>65535) interfaces.
532 			 */
533 			if (hitlimit) {
534 				panic("too many interfaces");
535 			}
536 			hitlimit = true;
537 			if_index = 1;
538 		}
539 		ifp->if_index = if_index;
540 	}
541 skip:
542 	/*
543 	 * ifindex2ifnet is indexed by if_index. Since if_index will
544 	 * grow dynamically, it should grow too.
545 	 */
546 	if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) {
547 		size_t m, n, oldlim;
548 		void *q;
549 
550 		oldlim = if_indexlim;
551 		while (ifp->if_index >= if_indexlim)
552 			if_indexlim <<= 1;
553 
554 		/* grow ifindex2ifnet */
555 		m = oldlim * sizeof(struct ifnet *);
556 		n = if_indexlim * sizeof(struct ifnet *);
557 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
558 		if (ifindex2ifnet != NULL) {
559 			memcpy(q, ifindex2ifnet, m);
560 			free(ifindex2ifnet, M_IFADDR);
561 		}
562 		ifindex2ifnet = (struct ifnet **)q;
563 	}
564 	ifindex2ifnet[ifp->if_index] = ifp;
565 }
566 
567 /*
568  * Initialize an interface and assign an index for it.
569  *
570  * It must be called prior to a device specific attach routine
571  * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl,
572  * and be followed by if_register:
573  *
574  *     if_initialize(ifp);
575  *     ether_ifattach(ifp, enaddr);
576  *     if_register(ifp);
577  */
578 void
579 if_initialize(ifnet_t *ifp)
580 {
581 	KASSERT(if_indexlim > 0);
582 	TAILQ_INIT(&ifp->if_addrlist);
583 
584 	/*
585 	 * Link level name is allocated later by a separate call to
586 	 * if_alloc_sadl().
587 	 */
588 
589 	if (ifp->if_snd.ifq_maxlen == 0)
590 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
591 
592 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
593 
594 	ifp->if_link_state = LINK_STATE_UNKNOWN;
595 
596 	ifp->if_capenable = 0;
597 	ifp->if_csum_flags_tx = 0;
598 	ifp->if_csum_flags_rx = 0;
599 
600 #ifdef ALTQ
601 	ifp->if_snd.altq_type = 0;
602 	ifp->if_snd.altq_disc = NULL;
603 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
604 	ifp->if_snd.altq_tbr  = NULL;
605 	ifp->if_snd.altq_ifp  = ifp;
606 #endif
607 
608 #ifdef NET_MPSAFE
609 	ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
610 #else
611 	ifp->if_snd.ifq_lock = NULL;
612 #endif
613 
614 	ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
615 	(void)pfil_run_hooks(if_pfil,
616 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
617 
618 	IF_AFDATA_LOCK_INIT(ifp);
619 
620 	if_getindex(ifp);
621 }
622 
623 /*
624  * Register an interface to the list of "active" interfaces.
625  */
626 void
627 if_register(ifnet_t *ifp)
628 {
629 	if (ifioctl_attach(ifp) != 0)
630 		panic("%s: ifioctl_attach() failed", __func__);
631 
632 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
633 
634 	if (!STAILQ_EMPTY(&domains))
635 		if_attachdomain1(ifp);
636 
637 	/* Announce the interface. */
638 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
639 
640 	if (ifp->if_slowtimo != NULL) {
641 		ifp->if_slowtimo_ch =
642 		    kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP);
643 		callout_init(ifp->if_slowtimo_ch, 0);
644 		callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp);
645 		if_slowtimo(ifp);
646 	}
647 
648 	TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list);
649 }
650 
651 /*
652  * Deprecated. Use if_initialize and if_register instead.
653  * See the above comment of if_initialize.
654  */
655 void
656 if_attach(ifnet_t *ifp)
657 {
658 	if_initialize(ifp);
659 	if_register(ifp);
660 }
661 
662 void
663 if_attachdomain(void)
664 {
665 	struct ifnet *ifp;
666 	int s;
667 
668 	s = splnet();
669 	IFNET_FOREACH(ifp)
670 		if_attachdomain1(ifp);
671 	splx(s);
672 }
673 
674 static void
675 if_attachdomain1(struct ifnet *ifp)
676 {
677 	struct domain *dp;
678 	int s;
679 
680 	s = splnet();
681 
682 	/* address family dependent data region */
683 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
684 	DOMAIN_FOREACH(dp) {
685 		if (dp->dom_ifattach != NULL)
686 			ifp->if_afdata[dp->dom_family] =
687 			    (*dp->dom_ifattach)(ifp);
688 	}
689 
690 	splx(s);
691 }
692 
693 /*
694  * Deactivate an interface.  This points all of the procedure
695  * handles at error stubs.  May be called from interrupt context.
696  */
697 void
698 if_deactivate(struct ifnet *ifp)
699 {
700 	int s;
701 
702 	s = splnet();
703 
704 	ifp->if_output	 = if_nulloutput;
705 	ifp->if_input	 = if_nullinput;
706 	ifp->if_start	 = if_nullstart;
707 	ifp->if_ioctl	 = if_nullioctl;
708 	ifp->if_init	 = if_nullinit;
709 	ifp->if_stop	 = if_nullstop;
710 	ifp->if_slowtimo = if_nullslowtimo;
711 	ifp->if_drain	 = if_nulldrain;
712 
713 	/* No more packets may be enqueued. */
714 	ifp->if_snd.ifq_maxlen = 0;
715 
716 	splx(s);
717 }
718 
719 void
720 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
721 {
722 	struct ifaddr *ifa, *nifa;
723 
724 	IFADDR_FOREACH_SAFE(ifa, ifp, nifa) {
725 		if (ifa->ifa_addr->sa_family != family)
726 			continue;
727 		(*purgeaddr)(ifa);
728 	}
729 }
730 
731 /*
732  * Detach an interface from the list of "active" interfaces,
733  * freeing any resources as we go along.
734  *
735  * NOTE: This routine must be called with a valid thread context,
736  * as it may block.
737  */
738 void
739 if_detach(struct ifnet *ifp)
740 {
741 	struct socket so;
742 	struct ifaddr *ifa;
743 #ifdef IFAREF_DEBUG
744 	struct ifaddr *last_ifa = NULL;
745 #endif
746 	struct domain *dp;
747 	const struct protosw *pr;
748 	int s, i, family, purged;
749 	uint64_t xc;
750 
751 	/*
752 	 * XXX It's kind of lame that we have to have the
753 	 * XXX socket structure...
754 	 */
755 	memset(&so, 0, sizeof(so));
756 
757 	s = splnet();
758 
759 	ifindex2ifnet[ifp->if_index] = NULL;
760 	TAILQ_REMOVE(&ifnet_list, ifp, if_list);
761 
762 	if (ifp->if_slowtimo != NULL) {
763 		ifp->if_slowtimo = NULL;
764 		callout_halt(ifp->if_slowtimo_ch, NULL);
765 		callout_destroy(ifp->if_slowtimo_ch);
766 		kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch));
767 	}
768 
769 	/*
770 	 * Do an if_down() to give protocols a chance to do something.
771 	 */
772 	if_down(ifp);
773 
774 #ifdef ALTQ
775 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
776 		altq_disable(&ifp->if_snd);
777 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
778 		altq_detach(&ifp->if_snd);
779 #endif
780 
781 	if (ifp->if_snd.ifq_lock)
782 		mutex_obj_free(ifp->if_snd.ifq_lock);
783 
784 	sysctl_teardown(&ifp->if_sysctl_log);
785 
786 #if NCARP > 0
787 	/* Remove the interface from any carp group it is a part of.  */
788 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
789 		carp_ifdetach(ifp);
790 #endif
791 
792 	/*
793 	 * Rip all the addresses off the interface.  This should make
794 	 * all of the routes go away.
795 	 *
796 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
797 	 * from the list, including our "cursor", ifa.  For safety,
798 	 * and to honor the TAILQ abstraction, I just restart the
799 	 * loop after each removal.  Note that the loop will exit
800 	 * when all of the remaining ifaddrs belong to the AF_LINK
801 	 * family.  I am counting on the historical fact that at
802 	 * least one pr_usrreq in each address domain removes at
803 	 * least one ifaddr.
804 	 */
805 again:
806 	IFADDR_FOREACH(ifa, ifp) {
807 		family = ifa->ifa_addr->sa_family;
808 #ifdef IFAREF_DEBUG
809 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
810 		    ifa, family, ifa->ifa_refcnt);
811 		if (last_ifa != NULL && ifa == last_ifa)
812 			panic("if_detach: loop detected");
813 		last_ifa = ifa;
814 #endif
815 		if (family == AF_LINK)
816 			continue;
817 		dp = pffinddomain(family);
818 #ifdef DIAGNOSTIC
819 		if (dp == NULL)
820 			panic("if_detach: no domain for AF %d",
821 			    family);
822 #endif
823 		/*
824 		 * XXX These PURGEIF calls are redundant with the
825 		 * purge-all-families calls below, but are left in for
826 		 * now both to make a smaller change, and to avoid
827 		 * unplanned interactions with clearing of
828 		 * ifp->if_addrlist.
829 		 */
830 		purged = 0;
831 		for (pr = dp->dom_protosw;
832 		     pr < dp->dom_protoswNPROTOSW; pr++) {
833 			so.so_proto = pr;
834 			if (pr->pr_usrreqs) {
835 				(void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
836 				purged = 1;
837 			}
838 		}
839 		if (purged == 0) {
840 			/*
841 			 * XXX What's really the best thing to do
842 			 * XXX here?  --thorpej@NetBSD.org
843 			 */
844 			printf("if_detach: WARNING: AF %d not purged\n",
845 			    family);
846 			ifa_remove(ifp, ifa);
847 		}
848 		goto again;
849 	}
850 
851 	if_free_sadl(ifp);
852 
853 	/* Walk the routing table looking for stragglers. */
854 	for (i = 0; i <= AF_MAX; i++) {
855 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
856 			continue;
857 	}
858 
859 	DOMAIN_FOREACH(dp) {
860 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
861 		{
862 			void *p = ifp->if_afdata[dp->dom_family];
863 			if (p) {
864 				ifp->if_afdata[dp->dom_family] = NULL;
865 				(*dp->dom_ifdetach)(ifp, p);
866 			}
867 		}
868 
869 		/*
870 		 * One would expect multicast memberships (INET and
871 		 * INET6) on UDP sockets to be purged by the PURGEIF
872 		 * calls above, but if all addresses were removed from
873 		 * the interface prior to destruction, the calls will
874 		 * not be made (e.g. ppp, for which pppd(8) generally
875 		 * removes addresses before destroying the interface).
876 		 * Because there is no invariant that multicast
877 		 * memberships only exist for interfaces with IPv4
878 		 * addresses, we must call PURGEIF regardless of
879 		 * addresses.  (Protocols which might store ifnet
880 		 * pointers are marked with PR_PURGEIF.)
881 		 */
882 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
883 			so.so_proto = pr;
884 			if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF)
885 				(void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
886 		}
887 	}
888 
889 	(void)pfil_run_hooks(if_pfil,
890 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
891 	(void)pfil_head_destroy(ifp->if_pfil);
892 
893 	/* Announce that the interface is gone. */
894 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
895 
896 	ifioctl_detach(ifp);
897 
898 	IF_AFDATA_LOCK_DESTROY(ifp);
899 
900 	/*
901 	 * remove packets that came from ifp, from software interrupt queues.
902 	 */
903 	DOMAIN_FOREACH(dp) {
904 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
905 			struct ifqueue *iq = dp->dom_ifqueues[i];
906 			if (iq == NULL)
907 				break;
908 			dp->dom_ifqueues[i] = NULL;
909 			if_detach_queues(ifp, iq);
910 		}
911 	}
912 
913 	/*
914 	 * IP queues have to be processed separately: net-queue barrier
915 	 * ensures that the packets are dequeued while a cross-call will
916 	 * ensure that the interrupts have completed. FIXME: not quite..
917 	 */
918 #ifdef INET
919 	pktq_barrier(ip_pktq);
920 #endif
921 #ifdef INET6
922 	if (in6_present)
923 		pktq_barrier(ip6_pktq);
924 #endif
925 	xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
926 	xc_wait(xc);
927 
928 	splx(s);
929 }
930 
931 static void
932 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
933 {
934 	struct mbuf *m, *prev, *next;
935 
936 	prev = NULL;
937 	for (m = q->ifq_head; m != NULL; m = next) {
938 		KASSERT((m->m_flags & M_PKTHDR) != 0);
939 
940 		next = m->m_nextpkt;
941 		if (m->m_pkthdr.rcvif != ifp) {
942 			prev = m;
943 			continue;
944 		}
945 
946 		if (prev != NULL)
947 			prev->m_nextpkt = m->m_nextpkt;
948 		else
949 			q->ifq_head = m->m_nextpkt;
950 		if (q->ifq_tail == m)
951 			q->ifq_tail = prev;
952 		q->ifq_len--;
953 
954 		m->m_nextpkt = NULL;
955 		m_freem(m);
956 		IF_DROP(q);
957 	}
958 }
959 
960 /*
961  * Callback for a radix tree walk to delete all references to an
962  * ifnet.
963  */
964 static int
965 if_rt_walktree(struct rtentry *rt, void *v)
966 {
967 	struct ifnet *ifp = (struct ifnet *)v;
968 	int error;
969 	struct rtentry *retrt;
970 
971 	if (rt->rt_ifp != ifp)
972 		return 0;
973 
974 	/* Delete the entry. */
975 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
976 	    rt_mask(rt), rt->rt_flags, &retrt);
977 	if (error == 0) {
978 		KASSERT(retrt == rt);
979 		KASSERT((retrt->rt_flags & RTF_UP) == 0);
980 		retrt->rt_ifp = NULL;
981 		rtfree(retrt);
982 	} else {
983 		printf("%s: warning: unable to delete rtentry @ %p, "
984 		    "error = %d\n", ifp->if_xname, rt, error);
985 	}
986 	return ERESTART;
987 }
988 
989 /*
990  * Create a clone network interface.
991  */
992 static int
993 if_clone_create(const char *name)
994 {
995 	struct if_clone *ifc;
996 	int unit;
997 
998 	ifc = if_clone_lookup(name, &unit);
999 	if (ifc == NULL)
1000 		return EINVAL;
1001 
1002 	if (ifunit(name) != NULL)
1003 		return EEXIST;
1004 
1005 	return (*ifc->ifc_create)(ifc, unit);
1006 }
1007 
1008 /*
1009  * Destroy a clone network interface.
1010  */
1011 static int
1012 if_clone_destroy(const char *name)
1013 {
1014 	struct if_clone *ifc;
1015 	struct ifnet *ifp;
1016 
1017 	ifc = if_clone_lookup(name, NULL);
1018 	if (ifc == NULL)
1019 		return EINVAL;
1020 
1021 	ifp = ifunit(name);
1022 	if (ifp == NULL)
1023 		return ENXIO;
1024 
1025 	if (ifc->ifc_destroy == NULL)
1026 		return EOPNOTSUPP;
1027 
1028 	return (*ifc->ifc_destroy)(ifp);
1029 }
1030 
1031 /*
1032  * Look up a network interface cloner.
1033  */
1034 static struct if_clone *
1035 if_clone_lookup(const char *name, int *unitp)
1036 {
1037 	struct if_clone *ifc;
1038 	const char *cp;
1039 	char *dp, ifname[IFNAMSIZ + 3];
1040 	int unit;
1041 
1042 	strcpy(ifname, "if_");
1043 	/* separate interface name from unit */
1044 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
1045 	    *cp && (*cp < '0' || *cp > '9');)
1046 		*dp++ = *cp++;
1047 
1048 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
1049 		return NULL;	/* No name or unit number */
1050 	*dp++ = '\0';
1051 
1052 again:
1053 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
1054 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
1055 			break;
1056 	}
1057 
1058 	if (ifc == NULL) {
1059 		if (*ifname == '\0' ||
1060 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
1061 			return NULL;
1062 		*ifname = '\0';
1063 		goto again;
1064 	}
1065 
1066 	unit = 0;
1067 	while (cp - name < IFNAMSIZ && *cp) {
1068 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1069 			/* Bogus unit number. */
1070 			return NULL;
1071 		}
1072 		unit = (unit * 10) + (*cp++ - '0');
1073 	}
1074 
1075 	if (unitp != NULL)
1076 		*unitp = unit;
1077 	return ifc;
1078 }
1079 
1080 /*
1081  * Register a network interface cloner.
1082  */
1083 void
1084 if_clone_attach(struct if_clone *ifc)
1085 {
1086 
1087 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1088 	if_cloners_count++;
1089 }
1090 
1091 /*
1092  * Unregister a network interface cloner.
1093  */
1094 void
1095 if_clone_detach(struct if_clone *ifc)
1096 {
1097 
1098 	LIST_REMOVE(ifc, ifc_list);
1099 	if_cloners_count--;
1100 }
1101 
1102 /*
1103  * Provide list of interface cloners to userspace.
1104  */
1105 int
1106 if_clone_list(int buf_count, char *buffer, int *total)
1107 {
1108 	char outbuf[IFNAMSIZ], *dst;
1109 	struct if_clone *ifc;
1110 	int count, error = 0;
1111 
1112 	*total = if_cloners_count;
1113 	if ((dst = buffer) == NULL) {
1114 		/* Just asking how many there are. */
1115 		return 0;
1116 	}
1117 
1118 	if (buf_count < 0)
1119 		return EINVAL;
1120 
1121 	count = (if_cloners_count < buf_count) ?
1122 	    if_cloners_count : buf_count;
1123 
1124 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1125 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1126 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1127 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1128 			return ENAMETOOLONG;
1129 		error = copyout(outbuf, dst, sizeof(outbuf));
1130 		if (error != 0)
1131 			break;
1132 	}
1133 
1134 	return error;
1135 }
1136 
1137 void
1138 ifaref(struct ifaddr *ifa)
1139 {
1140 	ifa->ifa_refcnt++;
1141 }
1142 
1143 void
1144 ifafree(struct ifaddr *ifa)
1145 {
1146 	KASSERT(ifa != NULL);
1147 	KASSERT(ifa->ifa_refcnt > 0);
1148 
1149 	if (--ifa->ifa_refcnt == 0) {
1150 		free(ifa, M_IFADDR);
1151 	}
1152 }
1153 
1154 void
1155 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1156 {
1157 	ifa->ifa_ifp = ifp;
1158 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1159 	ifaref(ifa);
1160 }
1161 
1162 void
1163 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1164 {
1165 	KASSERT(ifa->ifa_ifp == ifp);
1166 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1167 	ifafree(ifa);
1168 }
1169 
1170 static inline int
1171 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1172 {
1173 	return sockaddr_cmp(sa1, sa2) == 0;
1174 }
1175 
1176 /*
1177  * Locate an interface based on a complete address.
1178  */
1179 /*ARGSUSED*/
1180 struct ifaddr *
1181 ifa_ifwithaddr(const struct sockaddr *addr)
1182 {
1183 	struct ifnet *ifp;
1184 	struct ifaddr *ifa;
1185 
1186 	IFNET_FOREACH(ifp) {
1187 		if (ifp->if_output == if_nulloutput)
1188 			continue;
1189 		IFADDR_FOREACH(ifa, ifp) {
1190 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1191 				continue;
1192 			if (equal(addr, ifa->ifa_addr))
1193 				return ifa;
1194 			if ((ifp->if_flags & IFF_BROADCAST) &&
1195 			    ifa->ifa_broadaddr &&
1196 			    /* IP6 doesn't have broadcast */
1197 			    ifa->ifa_broadaddr->sa_len != 0 &&
1198 			    equal(ifa->ifa_broadaddr, addr))
1199 				return ifa;
1200 		}
1201 	}
1202 	return NULL;
1203 }
1204 
1205 /*
1206  * Locate the point to point interface with a given destination address.
1207  */
1208 /*ARGSUSED*/
1209 struct ifaddr *
1210 ifa_ifwithdstaddr(const struct sockaddr *addr)
1211 {
1212 	struct ifnet *ifp;
1213 	struct ifaddr *ifa;
1214 
1215 	IFNET_FOREACH(ifp) {
1216 		if (ifp->if_output == if_nulloutput)
1217 			continue;
1218 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1219 			continue;
1220 		IFADDR_FOREACH(ifa, ifp) {
1221 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1222 			    ifa->ifa_dstaddr == NULL)
1223 				continue;
1224 			if (equal(addr, ifa->ifa_dstaddr))
1225 				return ifa;
1226 		}
1227 	}
1228 	return NULL;
1229 }
1230 
1231 /*
1232  * Find an interface on a specific network.  If many, choice
1233  * is most specific found.
1234  */
1235 struct ifaddr *
1236 ifa_ifwithnet(const struct sockaddr *addr)
1237 {
1238 	struct ifnet *ifp;
1239 	struct ifaddr *ifa;
1240 	const struct sockaddr_dl *sdl;
1241 	struct ifaddr *ifa_maybe = 0;
1242 	u_int af = addr->sa_family;
1243 	const char *addr_data = addr->sa_data, *cplim;
1244 
1245 	if (af == AF_LINK) {
1246 		sdl = satocsdl(addr);
1247 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1248 		    ifindex2ifnet[sdl->sdl_index] &&
1249 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) {
1250 			return ifindex2ifnet[sdl->sdl_index]->if_dl;
1251 		}
1252 	}
1253 #ifdef NETATALK
1254 	if (af == AF_APPLETALK) {
1255 		const struct sockaddr_at *sat, *sat2;
1256 		sat = (const struct sockaddr_at *)addr;
1257 		IFNET_FOREACH(ifp) {
1258 			if (ifp->if_output == if_nulloutput)
1259 				continue;
1260 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1261 			if (ifa == NULL)
1262 				continue;
1263 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1264 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1265 				return ifa; /* exact match */
1266 			if (ifa_maybe == NULL) {
1267 				/* else keep the if with the right range */
1268 				ifa_maybe = ifa;
1269 			}
1270 		}
1271 		return ifa_maybe;
1272 	}
1273 #endif
1274 	IFNET_FOREACH(ifp) {
1275 		if (ifp->if_output == if_nulloutput)
1276 			continue;
1277 		IFADDR_FOREACH(ifa, ifp) {
1278 			const char *cp, *cp2, *cp3;
1279 
1280 			if (ifa->ifa_addr->sa_family != af ||
1281 			    ifa->ifa_netmask == NULL)
1282  next:				continue;
1283 			cp = addr_data;
1284 			cp2 = ifa->ifa_addr->sa_data;
1285 			cp3 = ifa->ifa_netmask->sa_data;
1286 			cplim = (const char *)ifa->ifa_netmask +
1287 			    ifa->ifa_netmask->sa_len;
1288 			while (cp3 < cplim) {
1289 				if ((*cp++ ^ *cp2++) & *cp3++) {
1290 					/* want to continue for() loop */
1291 					goto next;
1292 				}
1293 			}
1294 			if (ifa_maybe == NULL ||
1295 			    rn_refines((void *)ifa->ifa_netmask,
1296 			    (void *)ifa_maybe->ifa_netmask))
1297 				ifa_maybe = ifa;
1298 		}
1299 	}
1300 	return ifa_maybe;
1301 }
1302 
1303 /*
1304  * Find the interface of the addresss.
1305  */
1306 struct ifaddr *
1307 ifa_ifwithladdr(const struct sockaddr *addr)
1308 {
1309 	struct ifaddr *ia;
1310 
1311 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1312 	    (ia = ifa_ifwithnet(addr)))
1313 		return ia;
1314 	return NULL;
1315 }
1316 
1317 /*
1318  * Find an interface using a specific address family
1319  */
1320 struct ifaddr *
1321 ifa_ifwithaf(int af)
1322 {
1323 	struct ifnet *ifp;
1324 	struct ifaddr *ifa;
1325 
1326 	IFNET_FOREACH(ifp) {
1327 		if (ifp->if_output == if_nulloutput)
1328 			continue;
1329 		IFADDR_FOREACH(ifa, ifp) {
1330 			if (ifa->ifa_addr->sa_family == af)
1331 				return ifa;
1332 		}
1333 	}
1334 	return NULL;
1335 }
1336 
1337 /*
1338  * Find an interface address specific to an interface best matching
1339  * a given address.
1340  */
1341 struct ifaddr *
1342 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1343 {
1344 	struct ifaddr *ifa;
1345 	const char *cp, *cp2, *cp3;
1346 	const char *cplim;
1347 	struct ifaddr *ifa_maybe = 0;
1348 	u_int af = addr->sa_family;
1349 
1350 	if (ifp->if_output == if_nulloutput)
1351 		return NULL;
1352 
1353 	if (af >= AF_MAX)
1354 		return NULL;
1355 
1356 	IFADDR_FOREACH(ifa, ifp) {
1357 		if (ifa->ifa_addr->sa_family != af)
1358 			continue;
1359 		ifa_maybe = ifa;
1360 		if (ifa->ifa_netmask == NULL) {
1361 			if (equal(addr, ifa->ifa_addr) ||
1362 			    (ifa->ifa_dstaddr &&
1363 			     equal(addr, ifa->ifa_dstaddr)))
1364 				return ifa;
1365 			continue;
1366 		}
1367 		cp = addr->sa_data;
1368 		cp2 = ifa->ifa_addr->sa_data;
1369 		cp3 = ifa->ifa_netmask->sa_data;
1370 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1371 		for (; cp3 < cplim; cp3++) {
1372 			if ((*cp++ ^ *cp2++) & *cp3)
1373 				break;
1374 		}
1375 		if (cp3 == cplim)
1376 			return ifa;
1377 	}
1378 	return ifa_maybe;
1379 }
1380 
1381 /*
1382  * Default action when installing a route with a Link Level gateway.
1383  * Lookup an appropriate real ifa to point to.
1384  * This should be moved to /sys/net/link.c eventually.
1385  */
1386 void
1387 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1388 {
1389 	struct ifaddr *ifa;
1390 	const struct sockaddr *dst;
1391 	struct ifnet *ifp;
1392 
1393 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1394 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1395 		return;
1396 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1397 		rt_replace_ifa(rt, ifa);
1398 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1399 			ifa->ifa_rtrequest(cmd, rt, info);
1400 	}
1401 }
1402 
1403 /*
1404  * Handle a change in the interface link state.
1405  * XXX: We should listen to the routing socket in-kernel rather
1406  * than calling in6_if_link_* functions directly from here.
1407  */
1408 void
1409 if_link_state_change(struct ifnet *ifp, int link_state)
1410 {
1411 	int s;
1412 	int old_link_state;
1413 	struct domain *dp;
1414 
1415 	s = splnet();
1416 	if (ifp->if_link_state == link_state) {
1417 		splx(s);
1418 		return;
1419 	}
1420 
1421 	old_link_state = ifp->if_link_state;
1422 	ifp->if_link_state = link_state;
1423 #ifdef DEBUG
1424 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
1425 		link_state == LINK_STATE_UP ? "UP" :
1426 		link_state == LINK_STATE_DOWN ? "DOWN" :
1427 		"UNKNOWN",
1428 		 old_link_state == LINK_STATE_UP ? "UP" :
1429 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
1430 		"UNKNOWN");
1431 #endif
1432 
1433 	/*
1434 	 * When going from UNKNOWN to UP, we need to mark existing
1435 	 * addresses as tentative and restart DAD as we may have
1436 	 * erroneously not found a duplicate.
1437 	 *
1438 	 * This needs to happen before rt_ifmsg to avoid a race where
1439 	 * listeners would have an address and expect it to work right
1440 	 * away.
1441 	 */
1442 	if (link_state == LINK_STATE_UP &&
1443 	    old_link_state == LINK_STATE_UNKNOWN)
1444 	{
1445 		DOMAIN_FOREACH(dp) {
1446 			if (dp->dom_if_link_state_change != NULL)
1447 				dp->dom_if_link_state_change(ifp,
1448 				    LINK_STATE_DOWN);
1449 		}
1450 	}
1451 
1452 	/* Notify that the link state has changed. */
1453 	rt_ifmsg(ifp);
1454 
1455 #if NCARP > 0
1456 	if (ifp->if_carp)
1457 		carp_carpdev_state(ifp);
1458 #endif
1459 
1460 	DOMAIN_FOREACH(dp) {
1461 		if (dp->dom_if_link_state_change != NULL)
1462 			dp->dom_if_link_state_change(ifp, link_state);
1463 	}
1464 
1465 	splx(s);
1466 }
1467 
1468 /*
1469  * Default action when installing a local route on a point-to-point
1470  * interface.
1471  */
1472 void
1473 p2p_rtrequest(int req, struct rtentry *rt,
1474     __unused const struct rt_addrinfo *info)
1475 {
1476 	struct ifnet *ifp = rt->rt_ifp;
1477 	struct ifaddr *ifa, *lo0ifa;
1478 
1479 	switch (req) {
1480 	case RTM_ADD:
1481 		if ((rt->rt_flags & RTF_LOCAL) == 0)
1482 			break;
1483 
1484 		IFADDR_FOREACH(ifa, ifp) {
1485 			if (equal(rt_getkey(rt), ifa->ifa_addr))
1486 				break;
1487 		}
1488 		if (ifa == NULL)
1489 			break;
1490 
1491 		/*
1492 		 * Ensure lo0 has an address of the same family.
1493 		 */
1494 		IFADDR_FOREACH(lo0ifa, lo0ifp) {
1495 			if (lo0ifa->ifa_addr->sa_family ==
1496 			    ifa->ifa_addr->sa_family)
1497 				break;
1498 		}
1499 		if (lo0ifa == NULL)
1500 			break;
1501 
1502 		rt->rt_ifp = lo0ifp;
1503 		rt->rt_flags &= ~RTF_LLINFO;
1504 
1505 		/*
1506 		 * Make sure to set rt->rt_ifa to the interface
1507 		 * address we are using, otherwise we will have trouble
1508 		 * with source address selection.
1509 		 */
1510 		if (ifa != rt->rt_ifa)
1511 			rt_replace_ifa(rt, ifa);
1512 		break;
1513 	case RTM_DELETE:
1514 	case RTM_RESOLVE:
1515 	default:
1516 		break;
1517 	}
1518 }
1519 
1520 /*
1521  * Mark an interface down and notify protocols of
1522  * the transition.
1523  * NOTE: must be called at splsoftnet or equivalent.
1524  */
1525 void
1526 if_down(struct ifnet *ifp)
1527 {
1528 	struct ifaddr *ifa;
1529 	struct domain *dp;
1530 
1531 	ifp->if_flags &= ~IFF_UP;
1532 	nanotime(&ifp->if_lastchange);
1533 	IFADDR_FOREACH(ifa, ifp)
1534 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1535 	IFQ_PURGE(&ifp->if_snd);
1536 #if NCARP > 0
1537 	if (ifp->if_carp)
1538 		carp_carpdev_state(ifp);
1539 #endif
1540 	rt_ifmsg(ifp);
1541 	DOMAIN_FOREACH(dp) {
1542 		if (dp->dom_if_down)
1543 			dp->dom_if_down(ifp);
1544 	}
1545 }
1546 
1547 /*
1548  * Mark an interface up and notify protocols of
1549  * the transition.
1550  * NOTE: must be called at splsoftnet or equivalent.
1551  */
1552 void
1553 if_up(struct ifnet *ifp)
1554 {
1555 #ifdef notyet
1556 	struct ifaddr *ifa;
1557 #endif
1558 	struct domain *dp;
1559 
1560 	ifp->if_flags |= IFF_UP;
1561 	nanotime(&ifp->if_lastchange);
1562 #ifdef notyet
1563 	/* this has no effect on IP, and will kill all ISO connections XXX */
1564 	IFADDR_FOREACH(ifa, ifp)
1565 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1566 #endif
1567 #if NCARP > 0
1568 	if (ifp->if_carp)
1569 		carp_carpdev_state(ifp);
1570 #endif
1571 	rt_ifmsg(ifp);
1572 	DOMAIN_FOREACH(dp) {
1573 		if (dp->dom_if_up)
1574 			dp->dom_if_up(ifp);
1575 	}
1576 }
1577 
1578 /*
1579  * Handle interface slowtimo timer routine.  Called
1580  * from softclock, we decrement timer (if set) and
1581  * call the appropriate interface routine on expiration.
1582  */
1583 static void
1584 if_slowtimo(void *arg)
1585 {
1586 	void (*slowtimo)(struct ifnet *);
1587 	struct ifnet *ifp = arg;
1588 	int s;
1589 
1590 	slowtimo = ifp->if_slowtimo;
1591 	if (__predict_false(slowtimo == NULL))
1592 		return;
1593 
1594 	s = splnet();
1595 	if (ifp->if_timer != 0 && --ifp->if_timer == 0)
1596 		(*slowtimo)(ifp);
1597 
1598 	splx(s);
1599 
1600 	if (__predict_true(ifp->if_slowtimo != NULL))
1601 		callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ);
1602 }
1603 
1604 /*
1605  * Set/clear promiscuous mode on interface ifp based on the truth value
1606  * of pswitch.  The calls are reference counted so that only the first
1607  * "on" request actually has an effect, as does the final "off" request.
1608  * Results are undefined if the "off" and "on" requests are not matched.
1609  */
1610 int
1611 ifpromisc(struct ifnet *ifp, int pswitch)
1612 {
1613 	int pcount, ret;
1614 	short nflags;
1615 
1616 	pcount = ifp->if_pcount;
1617 	if (pswitch) {
1618 		/*
1619 		 * Allow the device to be "placed" into promiscuous
1620 		 * mode even if it is not configured up.  It will
1621 		 * consult IFF_PROMISC when it is brought up.
1622 		 */
1623 		if (ifp->if_pcount++ != 0)
1624 			return 0;
1625 		nflags = ifp->if_flags | IFF_PROMISC;
1626 	} else {
1627 		if (--ifp->if_pcount > 0)
1628 			return 0;
1629 		nflags = ifp->if_flags & ~IFF_PROMISC;
1630 	}
1631 	ret = if_flags_set(ifp, nflags);
1632 	/* Restore interface state if not successful. */
1633 	if (ret != 0) {
1634 		ifp->if_pcount = pcount;
1635 	}
1636 	return ret;
1637 }
1638 
1639 /*
1640  * Map interface name to
1641  * interface structure pointer.
1642  */
1643 struct ifnet *
1644 ifunit(const char *name)
1645 {
1646 	struct ifnet *ifp;
1647 	const char *cp = name;
1648 	u_int unit = 0;
1649 	u_int i;
1650 
1651 	/*
1652 	 * If the entire name is a number, treat it as an ifindex.
1653 	 */
1654 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1655 		unit = unit * 10 + (*cp - '0');
1656 	}
1657 
1658 	/*
1659 	 * If the number took all of the name, then it's a valid ifindex.
1660 	 */
1661 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1662 		if (unit >= if_indexlim)
1663 			return NULL;
1664 		ifp = ifindex2ifnet[unit];
1665 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1666 			return NULL;
1667 		return ifp;
1668 	}
1669 
1670 	IFNET_FOREACH(ifp) {
1671 		if (ifp->if_output == if_nulloutput)
1672 			continue;
1673 	 	if (strcmp(ifp->if_xname, name) == 0)
1674 			return ifp;
1675 	}
1676 	return NULL;
1677 }
1678 
1679 ifnet_t *
1680 if_byindex(u_int idx)
1681 {
1682 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1683 }
1684 
1685 /* common */
1686 int
1687 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1688 {
1689 	int s;
1690 	struct ifreq *ifr;
1691 	struct ifcapreq *ifcr;
1692 	struct ifdatareq *ifdr;
1693 
1694 	switch (cmd) {
1695 	case SIOCSIFCAP:
1696 		ifcr = data;
1697 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1698 			return EINVAL;
1699 
1700 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1701 			return 0;
1702 
1703 		ifp->if_capenable = ifcr->ifcr_capenable;
1704 
1705 		/* Pre-compute the checksum flags mask. */
1706 		ifp->if_csum_flags_tx = 0;
1707 		ifp->if_csum_flags_rx = 0;
1708 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1709 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1710 		}
1711 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1712 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1713 		}
1714 
1715 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1716 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1717 		}
1718 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1719 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1720 		}
1721 
1722 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1723 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1724 		}
1725 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1726 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1727 		}
1728 
1729 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1730 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1731 		}
1732 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1733 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1734 		}
1735 
1736 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1737 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1738 		}
1739 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1740 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1741 		}
1742 		if (ifp->if_flags & IFF_UP)
1743 			return ENETRESET;
1744 		return 0;
1745 	case SIOCSIFFLAGS:
1746 		ifr = data;
1747 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1748 			s = splnet();
1749 			if_down(ifp);
1750 			splx(s);
1751 		}
1752 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1753 			s = splnet();
1754 			if_up(ifp);
1755 			splx(s);
1756 		}
1757 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1758 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1759 		break;
1760 	case SIOCGIFFLAGS:
1761 		ifr = data;
1762 		ifr->ifr_flags = ifp->if_flags;
1763 		break;
1764 
1765 	case SIOCGIFMETRIC:
1766 		ifr = data;
1767 		ifr->ifr_metric = ifp->if_metric;
1768 		break;
1769 
1770 	case SIOCGIFMTU:
1771 		ifr = data;
1772 		ifr->ifr_mtu = ifp->if_mtu;
1773 		break;
1774 
1775 	case SIOCGIFDLT:
1776 		ifr = data;
1777 		ifr->ifr_dlt = ifp->if_dlt;
1778 		break;
1779 
1780 	case SIOCGIFCAP:
1781 		ifcr = data;
1782 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1783 		ifcr->ifcr_capenable = ifp->if_capenable;
1784 		break;
1785 
1786 	case SIOCSIFMETRIC:
1787 		ifr = data;
1788 		ifp->if_metric = ifr->ifr_metric;
1789 		break;
1790 
1791 	case SIOCGIFDATA:
1792 		ifdr = data;
1793 		ifdr->ifdr_data = ifp->if_data;
1794 		break;
1795 
1796 	case SIOCGIFINDEX:
1797 		ifr = data;
1798 		ifr->ifr_index = ifp->if_index;
1799 		break;
1800 
1801 	case SIOCZIFDATA:
1802 		ifdr = data;
1803 		ifdr->ifdr_data = ifp->if_data;
1804 		/*
1805 		 * Assumes that the volatile counters that can be
1806 		 * zero'ed are at the end of if_data.
1807 		 */
1808 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1809 		    offsetof(struct if_data, ifi_ipackets));
1810 		/*
1811 		 * The memset() clears to the bottm of if_data. In the area,
1812 		 * if_lastchange is included. Please be careful if new entry
1813 		 * will be added into if_data or rewite this.
1814 		 *
1815 		 * And also, update if_lastchnage.
1816 		 */
1817 		getnanotime(&ifp->if_lastchange);
1818 		break;
1819 	case SIOCSIFMTU:
1820 		ifr = data;
1821 		if (ifp->if_mtu == ifr->ifr_mtu)
1822 			break;
1823 		ifp->if_mtu = ifr->ifr_mtu;
1824 		/*
1825 		 * If the link MTU changed, do network layer specific procedure.
1826 		 */
1827 #ifdef INET6
1828 		if (in6_present)
1829 			nd6_setmtu(ifp);
1830 #endif
1831 		return ENETRESET;
1832 	default:
1833 		return ENOTTY;
1834 	}
1835 	return 0;
1836 }
1837 
1838 int
1839 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
1840 {
1841 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1842 	struct ifaddr *ifa;
1843 	const struct sockaddr *any, *sa;
1844 	union {
1845 		struct sockaddr sa;
1846 		struct sockaddr_storage ss;
1847 	} u, v;
1848 
1849 	switch (cmd) {
1850 	case SIOCSIFADDRPREF:
1851 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
1852 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1853 		    NULL) != 0)
1854 			return EPERM;
1855 	case SIOCGIFADDRPREF:
1856 		break;
1857 	default:
1858 		return EOPNOTSUPP;
1859 	}
1860 
1861 	/* sanity checks */
1862 	if (data == NULL || ifp == NULL) {
1863 		panic("invalid argument to %s", __func__);
1864 		/*NOTREACHED*/
1865 	}
1866 
1867 	/* address must be specified on ADD and DELETE */
1868 	sa = sstocsa(&ifap->ifap_addr);
1869 	if (sa->sa_family != sofamily(so))
1870 		return EINVAL;
1871 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1872 		return EINVAL;
1873 
1874 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1875 
1876 	IFADDR_FOREACH(ifa, ifp) {
1877 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1878 			continue;
1879 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1880 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1881 			break;
1882 	}
1883 	if (ifa == NULL)
1884 		return EADDRNOTAVAIL;
1885 
1886 	switch (cmd) {
1887 	case SIOCSIFADDRPREF:
1888 		ifa->ifa_preference = ifap->ifap_preference;
1889 		return 0;
1890 	case SIOCGIFADDRPREF:
1891 		/* fill in the if_laddrreq structure */
1892 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1893 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1894 		ifap->ifap_preference = ifa->ifa_preference;
1895 		return 0;
1896 	default:
1897 		return EOPNOTSUPP;
1898 	}
1899 }
1900 
1901 static void
1902 ifnet_lock_enter(struct ifnet_lock *il)
1903 {
1904 	uint64_t *nenter;
1905 
1906 	/* Before trying to acquire the mutex, increase the count of threads
1907 	 * who have entered or who wait to enter the critical section.
1908 	 * Avoid one costly locked memory transaction by keeping a count for
1909 	 * each CPU.
1910 	 */
1911 	nenter = percpu_getref(il->il_nenter);
1912 	(*nenter)++;
1913 	percpu_putref(il->il_nenter);
1914 	mutex_enter(&il->il_lock);
1915 }
1916 
1917 static void
1918 ifnet_lock_exit(struct ifnet_lock *il)
1919 {
1920 	/* Increase the count of threads who have exited the critical
1921 	 * section.  Increase while we still hold the lock.
1922 	 */
1923 	il->il_nexit++;
1924 	mutex_exit(&il->il_lock);
1925 }
1926 
1927 /*
1928  * Interface ioctls.
1929  */
1930 static int
1931 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1932 {
1933 	struct ifnet *ifp;
1934 	struct ifreq *ifr;
1935 	int error = 0;
1936 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1937 	u_long ocmd = cmd;
1938 #endif
1939 	short oif_flags;
1940 #ifdef COMPAT_OIFREQ
1941 	struct ifreq ifrb;
1942 	struct oifreq *oifr = NULL;
1943 #endif
1944 	int r;
1945 
1946 	switch (cmd) {
1947 #ifdef COMPAT_OIFREQ
1948 	case OSIOCGIFCONF:
1949 	case OOSIOCGIFCONF:
1950 		return compat_ifconf(cmd, data);
1951 #endif
1952 #ifdef COMPAT_OIFDATA
1953 	case OSIOCGIFDATA:
1954 	case OSIOCZIFDATA:
1955 		return compat_ifdatareq(l, cmd, data);
1956 #endif
1957 	case SIOCGIFCONF:
1958 		return ifconf(cmd, data);
1959 	case SIOCINITIFADDR:
1960 		return EPERM;
1961 	}
1962 
1963 #ifdef COMPAT_OIFREQ
1964 	cmd = compat_cvtcmd(cmd);
1965 	if (cmd != ocmd) {
1966 		oifr = data;
1967 		data = ifr = &ifrb;
1968 		ifreqo2n(oifr, ifr);
1969 	} else
1970 #endif
1971 		ifr = data;
1972 
1973 	ifp = ifunit(ifr->ifr_name);
1974 
1975 	switch (cmd) {
1976 	case SIOCIFCREATE:
1977 	case SIOCIFDESTROY:
1978 		if (l != NULL) {
1979 			error = kauth_authorize_network(l->l_cred,
1980 			    KAUTH_NETWORK_INTERFACE,
1981 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1982 			    (void *)cmd, NULL);
1983 			if (error != 0)
1984 				return error;
1985 		}
1986 		mutex_enter(&if_clone_mtx);
1987 		r = (cmd == SIOCIFCREATE) ?
1988 			if_clone_create(ifr->ifr_name) :
1989 			if_clone_destroy(ifr->ifr_name);
1990 		mutex_exit(&if_clone_mtx);
1991 		return r;
1992 
1993 	case SIOCIFGCLONERS:
1994 		{
1995 			struct if_clonereq *req = (struct if_clonereq *)data;
1996 			return if_clone_list(req->ifcr_count, req->ifcr_buffer,
1997 			    &req->ifcr_total);
1998 		}
1999 	}
2000 
2001 	if (ifp == NULL)
2002 		return ENXIO;
2003 
2004 	switch (cmd) {
2005 	case SIOCALIFADDR:
2006 	case SIOCDLIFADDR:
2007 	case SIOCSIFADDRPREF:
2008 	case SIOCSIFFLAGS:
2009 	case SIOCSIFCAP:
2010 	case SIOCSIFMETRIC:
2011 	case SIOCZIFDATA:
2012 	case SIOCSIFMTU:
2013 	case SIOCSIFPHYADDR:
2014 	case SIOCDIFPHYADDR:
2015 #ifdef INET6
2016 	case SIOCSIFPHYADDR_IN6:
2017 #endif
2018 	case SIOCSLIFPHYADDR:
2019 	case SIOCADDMULTI:
2020 	case SIOCDELMULTI:
2021 	case SIOCSIFMEDIA:
2022 	case SIOCSDRVSPEC:
2023 	case SIOCG80211:
2024 	case SIOCS80211:
2025 	case SIOCS80211NWID:
2026 	case SIOCS80211NWKEY:
2027 	case SIOCS80211POWER:
2028 	case SIOCS80211BSSID:
2029 	case SIOCS80211CHANNEL:
2030 	case SIOCSLINKSTR:
2031 		if (l != NULL) {
2032 			error = kauth_authorize_network(l->l_cred,
2033 			    KAUTH_NETWORK_INTERFACE,
2034 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
2035 			    (void *)cmd, NULL);
2036 			if (error != 0)
2037 				return error;
2038 		}
2039 	}
2040 
2041 	oif_flags = ifp->if_flags;
2042 
2043 	ifnet_lock_enter(ifp->if_ioctl_lock);
2044 	error = (*ifp->if_ioctl)(ifp, cmd, data);
2045 	if (error != ENOTTY)
2046 		;
2047 	else if (so->so_proto == NULL)
2048 		error = EOPNOTSUPP;
2049 	else {
2050 #ifdef COMPAT_OSOCK
2051 		error = compat_ifioctl(so, ocmd, cmd, data, l);
2052 #else
2053 		error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so,
2054 		    cmd, data, ifp);
2055 #endif
2056 	}
2057 
2058 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
2059 		if ((ifp->if_flags & IFF_UP) != 0) {
2060 			int s = splnet();
2061 			if_up(ifp);
2062 			splx(s);
2063 		}
2064 	}
2065 #ifdef COMPAT_OIFREQ
2066 	if (cmd != ocmd)
2067 		ifreqn2o(oifr, ifr);
2068 #endif
2069 
2070 	ifnet_lock_exit(ifp->if_ioctl_lock);
2071 	return error;
2072 }
2073 
2074 /* This callback adds to the sum in `arg' the number of
2075  * threads on `ci' who have entered or who wait to enter the
2076  * critical section.
2077  */
2078 static void
2079 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
2080 {
2081 	uint64_t *sum = arg, *nenter = p;
2082 
2083 	*sum += *nenter;
2084 }
2085 
2086 /* Return the number of threads who have entered or who wait
2087  * to enter the critical section on all CPUs.
2088  */
2089 static uint64_t
2090 ifnet_lock_entrances(struct ifnet_lock *il)
2091 {
2092 	uint64_t sum = 0;
2093 
2094 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
2095 
2096 	return sum;
2097 }
2098 
2099 static int
2100 ifioctl_attach(struct ifnet *ifp)
2101 {
2102 	struct ifnet_lock *il;
2103 
2104 	/* If the driver has not supplied its own if_ioctl, then
2105 	 * supply the default.
2106 	 */
2107 	if (ifp->if_ioctl == NULL)
2108 		ifp->if_ioctl = ifioctl_common;
2109 
2110 	/* Create an ifnet_lock for synchronizing ifioctls. */
2111 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
2112 		return ENOMEM;
2113 
2114 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
2115 	if (il->il_nenter == NULL) {
2116 		kmem_free(il, sizeof(*il));
2117 		return ENOMEM;
2118 	}
2119 
2120 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
2121 	cv_init(&il->il_emptied, ifp->if_xname);
2122 
2123 	ifp->if_ioctl_lock = il;
2124 
2125 	return 0;
2126 }
2127 
2128 /*
2129  * This must not be called until after `ifp' has been withdrawn from the
2130  * ifnet tables so that ifioctl() cannot get a handle on it by calling
2131  * ifunit().
2132  */
2133 static void
2134 ifioctl_detach(struct ifnet *ifp)
2135 {
2136 	struct ifnet_lock *il;
2137 
2138 	il = ifp->if_ioctl_lock;
2139 	mutex_enter(&il->il_lock);
2140 	/* Install if_nullioctl to make sure that any thread that
2141 	 * subsequently enters the critical section will quit it
2142 	 * immediately and signal the condition variable that we
2143 	 * wait on, below.
2144 	 */
2145 	ifp->if_ioctl = if_nullioctl;
2146 	/* Sleep while threads are still in the critical section or
2147 	 * wait to enter it.
2148 	 */
2149 	while (ifnet_lock_entrances(il) != il->il_nexit)
2150 		cv_wait(&il->il_emptied, &il->il_lock);
2151 	/* At this point, we are the only thread still in the critical
2152 	 * section, and no new thread can get a handle on the ifioctl
2153 	 * lock, so it is safe to free its memory.
2154 	 */
2155 	mutex_exit(&il->il_lock);
2156 	ifp->if_ioctl_lock = NULL;
2157 	percpu_free(il->il_nenter, sizeof(uint64_t));
2158 	il->il_nenter = NULL;
2159 	cv_destroy(&il->il_emptied);
2160 	mutex_destroy(&il->il_lock);
2161 	kmem_free(il, sizeof(*il));
2162 }
2163 
2164 /*
2165  * Return interface configuration
2166  * of system.  List may be used
2167  * in later ioctl's (above) to get
2168  * other information.
2169  *
2170  * Each record is a struct ifreq.  Before the addition of
2171  * sockaddr_storage, the API rule was that sockaddr flavors that did
2172  * not fit would extend beyond the struct ifreq, with the next struct
2173  * ifreq starting sa_len beyond the struct sockaddr.  Because the
2174  * union in struct ifreq includes struct sockaddr_storage, every kind
2175  * of sockaddr must fit.  Thus, there are no longer any overlength
2176  * records.
2177  *
2178  * Records are added to the user buffer if they fit, and ifc_len is
2179  * adjusted to the length that was written.  Thus, the user is only
2180  * assured of getting the complete list if ifc_len on return is at
2181  * least sizeof(struct ifreq) less than it was on entry.
2182  *
2183  * If the user buffer pointer is NULL, this routine copies no data and
2184  * returns the amount of space that would be needed.
2185  *
2186  * Invariants:
2187  * ifrp points to the next part of the user's buffer to be used.  If
2188  * ifrp != NULL, space holds the number of bytes remaining that we may
2189  * write at ifrp.  Otherwise, space holds the number of bytes that
2190  * would have been written had there been adequate space.
2191  */
2192 /*ARGSUSED*/
2193 static int
2194 ifconf(u_long cmd, void *data)
2195 {
2196 	struct ifconf *ifc = (struct ifconf *)data;
2197 	struct ifnet *ifp;
2198 	struct ifaddr *ifa;
2199 	struct ifreq ifr, *ifrp = NULL;
2200 	int space = 0, error = 0;
2201 	const int sz = (int)sizeof(struct ifreq);
2202 	const bool docopy = ifc->ifc_req != NULL;
2203 
2204 	if (docopy) {
2205 		space = ifc->ifc_len;
2206 		ifrp = ifc->ifc_req;
2207 	}
2208 
2209 	IFNET_FOREACH(ifp) {
2210 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2211 		    sizeof(ifr.ifr_name));
2212 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2213 			return ENAMETOOLONG;
2214 		if (IFADDR_EMPTY(ifp)) {
2215 			/* Interface with no addresses - send zero sockaddr. */
2216 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2217 			if (!docopy) {
2218 				space += sz;
2219 				continue;
2220 			}
2221 			if (space >= sz) {
2222 				error = copyout(&ifr, ifrp, sz);
2223 				if (error != 0)
2224 					return error;
2225 				ifrp++;
2226 				space -= sz;
2227 			}
2228 		}
2229 
2230 		IFADDR_FOREACH(ifa, ifp) {
2231 			struct sockaddr *sa = ifa->ifa_addr;
2232 			/* all sockaddrs must fit in sockaddr_storage */
2233 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2234 
2235 			if (!docopy) {
2236 				space += sz;
2237 				continue;
2238 			}
2239 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2240 			if (space >= sz) {
2241 				error = copyout(&ifr, ifrp, sz);
2242 				if (error != 0)
2243 					return (error);
2244 				ifrp++; space -= sz;
2245 			}
2246 		}
2247 	}
2248 	if (docopy) {
2249 		KASSERT(0 <= space && space <= ifc->ifc_len);
2250 		ifc->ifc_len -= space;
2251 	} else {
2252 		KASSERT(space >= 0);
2253 		ifc->ifc_len = space;
2254 	}
2255 	return (0);
2256 }
2257 
2258 int
2259 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2260 {
2261 	uint8_t len;
2262 #ifdef COMPAT_OIFREQ
2263 	struct ifreq ifrb;
2264 	struct oifreq *oifr = NULL;
2265 	u_long ocmd = cmd;
2266 	cmd = compat_cvtcmd(cmd);
2267 	if (cmd != ocmd) {
2268 		oifr = (struct oifreq *)(void *)ifr;
2269 		ifr = &ifrb;
2270 		ifreqo2n(oifr, ifr);
2271 		len = sizeof(oifr->ifr_addr);
2272 	} else
2273 #endif
2274 		len = sizeof(ifr->ifr_ifru.ifru_space);
2275 
2276 	if (len < sa->sa_len)
2277 		return EFBIG;
2278 
2279 	memset(&ifr->ifr_addr, 0, len);
2280 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2281 
2282 #ifdef COMPAT_OIFREQ
2283 	if (cmd != ocmd)
2284 		ifreqn2o(oifr, ifr);
2285 #endif
2286 	return 0;
2287 }
2288 
2289 /*
2290  * Queue message on interface, and start output if interface
2291  * not yet active.
2292  */
2293 int
2294 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2295     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2296 {
2297 	int len = m->m_pkthdr.len;
2298 	int mflags = m->m_flags;
2299 	int s = splnet();
2300 	int error;
2301 
2302 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2303 	if (error != 0)
2304 		goto out;
2305 	ifp->if_obytes += len;
2306 	if (mflags & M_MCAST)
2307 		ifp->if_omcasts++;
2308 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2309 		(*ifp->if_start)(ifp);
2310 out:
2311 	splx(s);
2312 	return error;
2313 }
2314 
2315 /*
2316  * Queue message on interface, possibly using a second fast queue
2317  */
2318 int
2319 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2320     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2321 {
2322 	int error = 0;
2323 
2324 	if (ifq != NULL
2325 #ifdef ALTQ
2326 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2327 #endif
2328 	    ) {
2329 		if (IF_QFULL(ifq)) {
2330 			IF_DROP(&ifp->if_snd);
2331 			m_freem(m);
2332 			if (error == 0)
2333 				error = ENOBUFS;
2334 		} else
2335 			IF_ENQUEUE(ifq, m);
2336 	} else
2337 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2338 	if (error != 0) {
2339 		++ifp->if_oerrors;
2340 		return error;
2341 	}
2342 	return 0;
2343 }
2344 
2345 int
2346 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2347 {
2348 	int rc;
2349 
2350 	if (ifp->if_initaddr != NULL)
2351 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2352 	else if (src ||
2353 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2354 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2355 
2356 	return rc;
2357 }
2358 
2359 int
2360 if_do_dad(struct ifnet *ifp)
2361 {
2362 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2363 		return 0;
2364 
2365 	switch (ifp->if_type) {
2366 	case IFT_FAITH:
2367 		/*
2368 		 * These interfaces do not have the IFF_LOOPBACK flag,
2369 		 * but loop packets back.  We do not have to do DAD on such
2370 		 * interfaces.  We should even omit it, because loop-backed
2371 		 * responses would confuse the DAD procedure.
2372 		 */
2373 		return 0;
2374 	default:
2375 		/*
2376 		 * Our DAD routine requires the interface up and running.
2377 		 * However, some interfaces can be up before the RUNNING
2378 		 * status.  Additionaly, users may try to assign addresses
2379 		 * before the interface becomes up (or running).
2380 		 * We simply skip DAD in such a case as a work around.
2381 		 * XXX: we should rather mark "tentative" on such addresses,
2382 		 * and do DAD after the interface becomes ready.
2383 		 */
2384 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2385 		    (IFF_UP|IFF_RUNNING))
2386 			return 0;
2387 
2388 		return 1;
2389 	}
2390 }
2391 
2392 int
2393 if_flags_set(ifnet_t *ifp, const short flags)
2394 {
2395 	int rc;
2396 
2397 	if (ifp->if_setflags != NULL)
2398 		rc = (*ifp->if_setflags)(ifp, flags);
2399 	else {
2400 		short cantflags, chgdflags;
2401 		struct ifreq ifr;
2402 
2403 		chgdflags = ifp->if_flags ^ flags;
2404 		cantflags = chgdflags & IFF_CANTCHANGE;
2405 
2406 		if (cantflags != 0)
2407 			ifp->if_flags ^= cantflags;
2408 
2409                 /* Traditionally, we do not call if_ioctl after
2410                  * setting/clearing only IFF_PROMISC if the interface
2411                  * isn't IFF_UP.  Uphold that tradition.
2412 		 */
2413 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2414 			return 0;
2415 
2416 		memset(&ifr, 0, sizeof(ifr));
2417 
2418 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2419 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2420 
2421 		if (rc != 0 && cantflags != 0)
2422 			ifp->if_flags ^= cantflags;
2423 	}
2424 
2425 	return rc;
2426 }
2427 
2428 int
2429 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2430 {
2431 	int rc;
2432 	struct ifreq ifr;
2433 
2434 	if (ifp->if_mcastop != NULL)
2435 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2436 	else {
2437 		ifreq_setaddr(cmd, &ifr, sa);
2438 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2439 	}
2440 
2441 	return rc;
2442 }
2443 
2444 static void
2445 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2446     struct ifaltq *ifq)
2447 {
2448 	const struct sysctlnode *cnode, *rnode;
2449 
2450 	if (sysctl_createv(clog, 0, NULL, &rnode,
2451 		       CTLFLAG_PERMANENT,
2452 		       CTLTYPE_NODE, "interfaces",
2453 		       SYSCTL_DESCR("Per-interface controls"),
2454 		       NULL, 0, NULL, 0,
2455 		       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
2456 		goto bad;
2457 
2458 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2459 		       CTLFLAG_PERMANENT,
2460 		       CTLTYPE_NODE, ifname,
2461 		       SYSCTL_DESCR("Interface controls"),
2462 		       NULL, 0, NULL, 0,
2463 		       CTL_CREATE, CTL_EOL) != 0)
2464 		goto bad;
2465 
2466 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2467 		       CTLFLAG_PERMANENT,
2468 		       CTLTYPE_NODE, "sndq",
2469 		       SYSCTL_DESCR("Interface output queue controls"),
2470 		       NULL, 0, NULL, 0,
2471 		       CTL_CREATE, CTL_EOL) != 0)
2472 		goto bad;
2473 
2474 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2475 		       CTLFLAG_PERMANENT,
2476 		       CTLTYPE_INT, "len",
2477 		       SYSCTL_DESCR("Current output queue length"),
2478 		       NULL, 0, &ifq->ifq_len, 0,
2479 		       CTL_CREATE, CTL_EOL) != 0)
2480 		goto bad;
2481 
2482 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2483 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2484 		       CTLTYPE_INT, "maxlen",
2485 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2486 		       NULL, 0, &ifq->ifq_maxlen, 0,
2487 		       CTL_CREATE, CTL_EOL) != 0)
2488 		goto bad;
2489 
2490 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2491 		       CTLFLAG_PERMANENT,
2492 		       CTLTYPE_INT, "drops",
2493 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2494 		       NULL, 0, &ifq->ifq_drops, 0,
2495 		       CTL_CREATE, CTL_EOL) != 0)
2496 		goto bad;
2497 
2498 	return;
2499 bad:
2500 	printf("%s: could not attach sysctl nodes\n", ifname);
2501 	return;
2502 }
2503 
2504 #if defined(INET) || defined(INET6)
2505 
2506 #define	SYSCTL_NET_PKTQ(q, cn, c)					\
2507 	static int							\
2508 	sysctl_net_##q##_##cn(SYSCTLFN_ARGS)				\
2509 	{								\
2510 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c);	\
2511 	}
2512 
2513 #if defined(INET)
2514 static int
2515 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS)
2516 {
2517 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq);
2518 }
2519 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS)
2520 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS)
2521 #endif
2522 
2523 #if defined(INET6)
2524 static int
2525 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS)
2526 {
2527 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq);
2528 }
2529 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS)
2530 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS)
2531 #endif
2532 
2533 static void
2534 sysctl_net_pktq_setup(struct sysctllog **clog, int pf)
2535 {
2536 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
2537 	const char *pfname = NULL, *ipname = NULL;
2538 	int ipn = 0, qid = 0;
2539 
2540 	switch (pf) {
2541 #if defined(INET)
2542 	case PF_INET:
2543 		len_func = sysctl_net_ip_pktq_items;
2544 		maxlen_func = sysctl_net_ip_pktq_maxlen;
2545 		drops_func = sysctl_net_ip_pktq_drops;
2546 		pfname = "inet", ipn = IPPROTO_IP;
2547 		ipname = "ip", qid = IPCTL_IFQ;
2548 		break;
2549 #endif
2550 #if defined(INET6)
2551 	case PF_INET6:
2552 		len_func = sysctl_net_ip6_pktq_items;
2553 		maxlen_func = sysctl_net_ip6_pktq_maxlen;
2554 		drops_func = sysctl_net_ip6_pktq_drops;
2555 		pfname = "inet6", ipn = IPPROTO_IPV6;
2556 		ipname = "ip6", qid = IPV6CTL_IFQ;
2557 		break;
2558 #endif
2559 	default:
2560 		KASSERT(false);
2561 	}
2562 
2563 	sysctl_createv(clog, 0, NULL, NULL,
2564 		       CTLFLAG_PERMANENT,
2565 		       CTLTYPE_NODE, pfname, NULL,
2566 		       NULL, 0, NULL, 0,
2567 		       CTL_NET, pf, CTL_EOL);
2568 	sysctl_createv(clog, 0, NULL, NULL,
2569 		       CTLFLAG_PERMANENT,
2570 		       CTLTYPE_NODE, ipname, NULL,
2571 		       NULL, 0, NULL, 0,
2572 		       CTL_NET, pf, ipn, CTL_EOL);
2573 	sysctl_createv(clog, 0, NULL, NULL,
2574 		       CTLFLAG_PERMANENT,
2575 		       CTLTYPE_NODE, "ifq",
2576 		       SYSCTL_DESCR("Protocol input queue controls"),
2577 		       NULL, 0, NULL, 0,
2578 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2579 
2580 	sysctl_createv(clog, 0, NULL, NULL,
2581 		       CTLFLAG_PERMANENT,
2582 		       CTLTYPE_INT, "len",
2583 		       SYSCTL_DESCR("Current input queue length"),
2584 		       len_func, 0, NULL, 0,
2585 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2586 	sysctl_createv(clog, 0, NULL, NULL,
2587 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2588 		       CTLTYPE_INT, "maxlen",
2589 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2590 		       maxlen_func, 0, NULL, 0,
2591 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2592 	sysctl_createv(clog, 0, NULL, NULL,
2593 		       CTLFLAG_PERMANENT,
2594 		       CTLTYPE_INT, "drops",
2595 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2596 		       drops_func, 0, NULL, 0,
2597 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2598 }
2599 #endif /* INET || INET6 */
2600 
2601 static int
2602 if_sdl_sysctl(SYSCTLFN_ARGS)
2603 {
2604 	struct ifnet *ifp;
2605 	const struct sockaddr_dl *sdl;
2606 
2607 	if (namelen != 1)
2608 		return EINVAL;
2609 
2610 	ifp = if_byindex(name[0]);
2611 	if (ifp == NULL)
2612 		return ENODEV;
2613 
2614 	sdl = ifp->if_sadl;
2615 	if (sdl == NULL) {
2616 		*oldlenp = 0;
2617 		return 0;
2618 	}
2619 
2620 	if (oldp == NULL) {
2621 		*oldlenp = sdl->sdl_alen;
2622 		return 0;
2623 	}
2624 
2625 	if (*oldlenp >= sdl->sdl_alen)
2626 		*oldlenp = sdl->sdl_alen;
2627 	return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp);
2628 }
2629 
2630 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup")
2631 {
2632 	const struct sysctlnode *rnode = NULL;
2633 
2634 	sysctl_createv(clog, 0, NULL, &rnode,
2635 		       CTLFLAG_PERMANENT,
2636 		       CTLTYPE_NODE, "sdl",
2637 		       SYSCTL_DESCR("Get active link-layer address"),
2638 		       if_sdl_sysctl, 0, NULL, 0,
2639 		       CTL_NET, CTL_CREATE, CTL_EOL);
2640 }
2641