xref: /netbsd-src/sys/net/if.c (revision 6cf6fe02a981b55727c49c3d37b0d8191a98c0ee)
1 /*	$NetBSD: if.c,v 1.291 2014/09/09 20:16:12 rmind 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 1.291 2014/09/09 20:16:12 rmind Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 #include "opt_wlan.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 #include <sys/kmem.h>
116 #include <sys/xcall.h>
117 
118 #include <net/if.h>
119 #include <net/if_dl.h>
120 #include <net/if_ether.h>
121 #include <net/if_media.h>
122 #include <net80211/ieee80211.h>
123 #include <net80211/ieee80211_ioctl.h>
124 #include <net/if_types.h>
125 #include <net/radix.h>
126 #include <net/route.h>
127 #include <net/netisr.h>
128 #include <sys/module.h>
129 #ifdef NETATALK
130 #include <netatalk/at_extern.h>
131 #include <netatalk/at.h>
132 #endif
133 #include <net/pfil.h>
134 #include <netinet/in.h>
135 #include <netinet/in_var.h>
136 
137 #ifdef INET6
138 #include <netinet6/in6_var.h>
139 #include <netinet6/nd6.h>
140 #endif
141 
142 #include "ether.h"
143 #include "fddi.h"
144 #include "token.h"
145 
146 #include "carp.h"
147 #if NCARP > 0
148 #include <netinet/ip_carp.h>
149 #endif
150 
151 #include <compat/sys/sockio.h>
152 #include <compat/sys/socket.h>
153 
154 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
155 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
156 
157 /*
158  * Global list of interfaces.
159  */
160 struct ifnet_head		ifnet_list;
161 static ifnet_t **		ifindex2ifnet = NULL;
162 
163 static u_int			if_index = 1;
164 static size_t			if_indexlim = 0;
165 static uint64_t			index_gen;
166 static kmutex_t			index_gen_mtx;
167 
168 static struct ifaddr **		ifnet_addrs = NULL;
169 
170 static callout_t		if_slowtimo_ch;
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 static int	if_clone_list(struct if_clonereq *);
179 
180 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
181 static int if_cloners_count;
182 
183 /* Packet filtering hook for interfaces. */
184 pfil_head_t *	if_pfil;
185 
186 static kauth_listener_t if_listener;
187 
188 static int doifioctl(struct socket *, u_long, void *, struct lwp *);
189 static int ifioctl_attach(struct ifnet *);
190 static void ifioctl_detach(struct ifnet *);
191 static void ifnet_lock_enter(struct ifnet_lock *);
192 static void ifnet_lock_exit(struct ifnet_lock *);
193 static void if_detach_queues(struct ifnet *, struct ifqueue *);
194 static void sysctl_sndq_setup(struct sysctllog **, const char *,
195     struct ifaltq *);
196 
197 #if defined(INET) || defined(INET6)
198 static void sysctl_net_pktq_setup(struct sysctllog **, int);
199 #endif
200 
201 static int
202 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
203     void *arg0, void *arg1, void *arg2, void *arg3)
204 {
205 	int result;
206 	enum kauth_network_req req;
207 
208 	result = KAUTH_RESULT_DEFER;
209 	req = (enum kauth_network_req)arg1;
210 
211 	if (action != KAUTH_NETWORK_INTERFACE)
212 		return result;
213 
214 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
215 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
216 		result = KAUTH_RESULT_ALLOW;
217 
218 	return result;
219 }
220 
221 /*
222  * Network interface utility routines.
223  *
224  * Routines with ifa_ifwith* names take sockaddr *'s as
225  * parameters.
226  */
227 void
228 ifinit(void)
229 {
230 #if defined(INET)
231 	sysctl_net_pktq_setup(NULL, PF_INET);
232 #endif
233 #ifdef INET6
234 	sysctl_net_pktq_setup(NULL, PF_INET6);
235 #endif
236 
237 	callout_init(&if_slowtimo_ch, 0);
238 	if_slowtimo(NULL);
239 
240 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
241 	    if_listener_cb, NULL);
242 
243 	/* interfaces are available, inform socket code */
244 	ifioctl = doifioctl;
245 }
246 
247 /*
248  * XXX Initialization before configure().
249  * XXX hack to get pfil_add_hook working in autoconf.
250  */
251 void
252 ifinit1(void)
253 {
254 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
255 	TAILQ_INIT(&ifnet_list);
256 	if_indexlim = 8;
257 
258 	if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
259 	KASSERT(if_pfil != NULL);
260 
261 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN)
262 	etherinit();
263 #endif
264 }
265 
266 ifnet_t *
267 if_alloc(u_char type)
268 {
269 	return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP);
270 }
271 
272 void
273 if_free(ifnet_t *ifp)
274 {
275 	kmem_free(ifp, sizeof(ifnet_t));
276 }
277 
278 void
279 if_initname(struct ifnet *ifp, const char *name, int unit)
280 {
281 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
282 	    "%s%d", name, unit);
283 }
284 
285 /*
286  * Null routines used while an interface is going away.  These routines
287  * just return an error.
288  */
289 
290 int
291 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
292     const struct sockaddr *so, struct rtentry *rt)
293 {
294 
295 	return ENXIO;
296 }
297 
298 void
299 if_nullinput(struct ifnet *ifp, struct mbuf *m)
300 {
301 
302 	/* Nothing. */
303 }
304 
305 void
306 if_nullstart(struct ifnet *ifp)
307 {
308 
309 	/* Nothing. */
310 }
311 
312 int
313 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
314 {
315 
316 	/* Wake ifioctl_detach(), who may wait for all threads to
317 	 * quit the critical section.
318 	 */
319 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
320 	return ENXIO;
321 }
322 
323 int
324 if_nullinit(struct ifnet *ifp)
325 {
326 
327 	return ENXIO;
328 }
329 
330 void
331 if_nullstop(struct ifnet *ifp, int disable)
332 {
333 
334 	/* Nothing. */
335 }
336 
337 void
338 if_nullwatchdog(struct ifnet *ifp)
339 {
340 
341 	/* Nothing. */
342 }
343 
344 void
345 if_nulldrain(struct ifnet *ifp)
346 {
347 
348 	/* Nothing. */
349 }
350 
351 void
352 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
353 {
354 	struct ifaddr *ifa;
355 	struct sockaddr_dl *sdl;
356 
357 	ifp->if_addrlen = addrlen;
358 	if_alloc_sadl(ifp);
359 	ifa = ifp->if_dl;
360 	sdl = satosdl(ifa->ifa_addr);
361 
362 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
363 	if (factory) {
364 		ifp->if_hwdl = ifp->if_dl;
365 		ifaref(ifp->if_hwdl);
366 	}
367 	/* TBD routing socket */
368 }
369 
370 struct ifaddr *
371 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
372 {
373 	unsigned socksize, ifasize;
374 	int addrlen, namelen;
375 	struct sockaddr_dl *mask, *sdl;
376 	struct ifaddr *ifa;
377 
378 	namelen = strlen(ifp->if_xname);
379 	addrlen = ifp->if_addrlen;
380 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
381 	ifasize = sizeof(*ifa) + 2 * socksize;
382 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
383 
384 	sdl = (struct sockaddr_dl *)(ifa + 1);
385 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
386 
387 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
388 	    ifp->if_xname, namelen, NULL, addrlen);
389 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
390 	memset(&mask->sdl_data[0], 0xff, namelen);
391 	ifa->ifa_rtrequest = link_rtrequest;
392 	ifa->ifa_addr = (struct sockaddr *)sdl;
393 	ifa->ifa_netmask = (struct sockaddr *)mask;
394 
395 	*sdlp = sdl;
396 
397 	return ifa;
398 }
399 
400 static void
401 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
402 {
403 	const struct sockaddr_dl *sdl;
404 	ifnet_addrs[ifp->if_index] = ifa;
405 	ifaref(ifa);
406 	ifp->if_dl = ifa;
407 	ifaref(ifa);
408 	sdl = satosdl(ifa->ifa_addr);
409 	ifp->if_sadl = sdl;
410 }
411 
412 /*
413  * Allocate the link level name for the specified interface.  This
414  * is an attachment helper.  It must be called after ifp->if_addrlen
415  * is initialized, which may not be the case when if_attach() is
416  * called.
417  */
418 void
419 if_alloc_sadl(struct ifnet *ifp)
420 {
421 	struct ifaddr *ifa;
422 	const struct sockaddr_dl *sdl;
423 
424 	/*
425 	 * If the interface already has a link name, release it
426 	 * now.  This is useful for interfaces that can change
427 	 * link types, and thus switch link names often.
428 	 */
429 	if (ifp->if_sadl != NULL)
430 		if_free_sadl(ifp);
431 
432 	ifa = if_dl_create(ifp, &sdl);
433 
434 	ifa_insert(ifp, ifa);
435 	if_sadl_setrefs(ifp, ifa);
436 }
437 
438 static void
439 if_deactivate_sadl(struct ifnet *ifp)
440 {
441 	struct ifaddr *ifa;
442 
443 	KASSERT(ifp->if_dl != NULL);
444 
445 	ifa = ifp->if_dl;
446 
447 	ifp->if_sadl = NULL;
448 
449 	ifnet_addrs[ifp->if_index] = NULL;
450 	ifafree(ifa);
451 	ifp->if_dl = NULL;
452 	ifafree(ifa);
453 }
454 
455 void
456 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
457     const struct sockaddr_dl *sdl)
458 {
459 	int s;
460 
461 	s = splnet();
462 
463 	if_deactivate_sadl(ifp);
464 
465 	if_sadl_setrefs(ifp, ifa);
466 	IFADDR_FOREACH(ifa, ifp)
467 		rtinit(ifa, RTM_LLINFO_UPD, 0);
468 	splx(s);
469 }
470 
471 /*
472  * Free the link level name for the specified interface.  This is
473  * a detach helper.  This is called from if_detach() or from
474  * link layer type specific detach functions.
475  */
476 void
477 if_free_sadl(struct ifnet *ifp)
478 {
479 	struct ifaddr *ifa;
480 	int s;
481 
482 	ifa = ifnet_addrs[ifp->if_index];
483 	if (ifa == NULL) {
484 		KASSERT(ifp->if_sadl == NULL);
485 		KASSERT(ifp->if_dl == NULL);
486 		return;
487 	}
488 
489 	KASSERT(ifp->if_sadl != NULL);
490 	KASSERT(ifp->if_dl != NULL);
491 
492 	s = splnet();
493 	rtinit(ifa, RTM_DELETE, 0);
494 	ifa_remove(ifp, ifa);
495 	if_deactivate_sadl(ifp);
496 	if (ifp->if_hwdl == ifa) {
497 		ifafree(ifa);
498 		ifp->if_hwdl = NULL;
499 	}
500 	splx(s);
501 }
502 
503 static void
504 if_getindex(ifnet_t *ifp)
505 {
506 	bool hitlimit = false;
507 
508 	mutex_enter(&index_gen_mtx);
509 	ifp->if_index_gen = index_gen++;
510 	mutex_exit(&index_gen_mtx);
511 
512 	ifp->if_index = if_index;
513 	if (ifindex2ifnet == NULL) {
514 		if_index++;
515 		goto skip;
516 	}
517 	while (if_byindex(ifp->if_index)) {
518 		/*
519 		 * If we hit USHRT_MAX, we skip back to 0 since
520 		 * there are a number of places where the value
521 		 * of if_index or if_index itself is compared
522 		 * to or stored in an unsigned short.  By
523 		 * jumping back, we won't botch those assignments
524 		 * or comparisons.
525 		 */
526 		if (++if_index == 0) {
527 			if_index = 1;
528 		} else if (if_index == USHRT_MAX) {
529 			/*
530 			 * However, if we have to jump back to
531 			 * zero *twice* without finding an empty
532 			 * slot in ifindex2ifnet[], then there
533 			 * there are too many (>65535) interfaces.
534 			 */
535 			if (hitlimit) {
536 				panic("too many interfaces");
537 			}
538 			hitlimit = true;
539 			if_index = 1;
540 		}
541 		ifp->if_index = if_index;
542 	}
543 skip:
544 	/*
545 	 * We have some arrays that should be indexed by if_index.
546 	 * since if_index will grow dynamically, they should grow too.
547 	 *	struct ifadd **ifnet_addrs
548 	 *	struct ifnet **ifindex2ifnet
549 	 */
550 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
551 	    ifp->if_index >= if_indexlim) {
552 		size_t m, n, oldlim;
553 		void *q;
554 
555 		oldlim = if_indexlim;
556 		while (ifp->if_index >= if_indexlim)
557 			if_indexlim <<= 1;
558 
559 		/* grow ifnet_addrs */
560 		m = oldlim * sizeof(struct ifaddr *);
561 		n = if_indexlim * sizeof(struct ifaddr *);
562 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
563 		if (ifnet_addrs != NULL) {
564 			memcpy(q, ifnet_addrs, m);
565 			free(ifnet_addrs, M_IFADDR);
566 		}
567 		ifnet_addrs = (struct ifaddr **)q;
568 
569 		/* grow ifindex2ifnet */
570 		m = oldlim * sizeof(struct ifnet *);
571 		n = if_indexlim * sizeof(struct ifnet *);
572 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
573 		if (ifindex2ifnet != NULL) {
574 			memcpy(q, ifindex2ifnet, m);
575 			free(ifindex2ifnet, M_IFADDR);
576 		}
577 		ifindex2ifnet = (struct ifnet **)q;
578 	}
579 	ifindex2ifnet[ifp->if_index] = ifp;
580 }
581 
582 /*
583  * Attach an interface to the list of "active" interfaces.
584  */
585 void
586 if_attach(ifnet_t *ifp)
587 {
588 	KASSERT(if_indexlim > 0);
589 	TAILQ_INIT(&ifp->if_addrlist);
590 	TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list);
591 
592 	if (ifioctl_attach(ifp) != 0)
593 		panic("%s: ifioctl_attach() failed", __func__);
594 
595 	if_getindex(ifp);
596 
597 	/*
598 	 * Link level name is allocated later by a separate call to
599 	 * if_alloc_sadl().
600 	 */
601 
602 	if (ifp->if_snd.ifq_maxlen == 0)
603 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
604 
605 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
606 
607 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
608 
609 	ifp->if_link_state = LINK_STATE_UNKNOWN;
610 
611 	ifp->if_capenable = 0;
612 	ifp->if_csum_flags_tx = 0;
613 	ifp->if_csum_flags_rx = 0;
614 
615 #ifdef ALTQ
616 	ifp->if_snd.altq_type = 0;
617 	ifp->if_snd.altq_disc = NULL;
618 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
619 	ifp->if_snd.altq_tbr  = NULL;
620 	ifp->if_snd.altq_ifp  = ifp;
621 #endif
622 
623 #ifdef NET_MPSAFE
624 	ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
625 #else
626 	ifp->if_snd.ifq_lock = NULL;
627 #endif
628 
629 	ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
630 	(void)pfil_run_hooks(if_pfil,
631 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
632 
633 	if (!STAILQ_EMPTY(&domains))
634 		if_attachdomain1(ifp);
635 
636 	/* Announce the interface. */
637 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
638 }
639 
640 void
641 if_attachdomain(void)
642 {
643 	struct ifnet *ifp;
644 	int s;
645 
646 	s = splnet();
647 	IFNET_FOREACH(ifp)
648 		if_attachdomain1(ifp);
649 	splx(s);
650 }
651 
652 void
653 if_attachdomain1(struct ifnet *ifp)
654 {
655 	struct domain *dp;
656 	int s;
657 
658 	s = splnet();
659 
660 	/* address family dependent data region */
661 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
662 	DOMAIN_FOREACH(dp) {
663 		if (dp->dom_ifattach != NULL)
664 			ifp->if_afdata[dp->dom_family] =
665 			    (*dp->dom_ifattach)(ifp);
666 	}
667 
668 	splx(s);
669 }
670 
671 /*
672  * Deactivate an interface.  This points all of the procedure
673  * handles at error stubs.  May be called from interrupt context.
674  */
675 void
676 if_deactivate(struct ifnet *ifp)
677 {
678 	int s;
679 
680 	s = splnet();
681 
682 	ifp->if_output	 = if_nulloutput;
683 	ifp->if_input	 = if_nullinput;
684 	ifp->if_start	 = if_nullstart;
685 	ifp->if_ioctl	 = if_nullioctl;
686 	ifp->if_init	 = if_nullinit;
687 	ifp->if_stop	 = if_nullstop;
688 	ifp->if_watchdog = if_nullwatchdog;
689 	ifp->if_drain	 = if_nulldrain;
690 
691 	/* No more packets may be enqueued. */
692 	ifp->if_snd.ifq_maxlen = 0;
693 
694 	splx(s);
695 }
696 
697 void
698 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
699 {
700 	struct ifaddr *ifa, *nifa;
701 
702 	IFADDR_FOREACH_SAFE(ifa, ifp, nifa) {
703 		if (ifa->ifa_addr->sa_family != family)
704 			continue;
705 		(*purgeaddr)(ifa);
706 	}
707 }
708 
709 /*
710  * Detach an interface from the list of "active" interfaces,
711  * freeing any resources as we go along.
712  *
713  * NOTE: This routine must be called with a valid thread context,
714  * as it may block.
715  */
716 void
717 if_detach(struct ifnet *ifp)
718 {
719 	struct socket so;
720 	struct ifaddr *ifa;
721 #ifdef IFAREF_DEBUG
722 	struct ifaddr *last_ifa = NULL;
723 #endif
724 	struct domain *dp;
725 	const struct protosw *pr;
726 	int s, i, family, purged;
727 	uint64_t xc;
728 
729 	/*
730 	 * XXX It's kind of lame that we have to have the
731 	 * XXX socket structure...
732 	 */
733 	memset(&so, 0, sizeof(so));
734 
735 	s = splnet();
736 
737 	/*
738 	 * Do an if_down() to give protocols a chance to do something.
739 	 */
740 	if_down(ifp);
741 
742 #ifdef ALTQ
743 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
744 		altq_disable(&ifp->if_snd);
745 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
746 		altq_detach(&ifp->if_snd);
747 #endif
748 
749 	if (ifp->if_snd.ifq_lock)
750 		mutex_obj_free(ifp->if_snd.ifq_lock);
751 
752 	sysctl_teardown(&ifp->if_sysctl_log);
753 
754 #if NCARP > 0
755 	/* Remove the interface from any carp group it is a part of.  */
756 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
757 		carp_ifdetach(ifp);
758 #endif
759 
760 	/*
761 	 * Rip all the addresses off the interface.  This should make
762 	 * all of the routes go away.
763 	 *
764 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
765 	 * from the list, including our "cursor", ifa.  For safety,
766 	 * and to honor the TAILQ abstraction, I just restart the
767 	 * loop after each removal.  Note that the loop will exit
768 	 * when all of the remaining ifaddrs belong to the AF_LINK
769 	 * family.  I am counting on the historical fact that at
770 	 * least one pr_usrreq in each address domain removes at
771 	 * least one ifaddr.
772 	 */
773 again:
774 	IFADDR_FOREACH(ifa, ifp) {
775 		family = ifa->ifa_addr->sa_family;
776 #ifdef IFAREF_DEBUG
777 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
778 		    ifa, family, ifa->ifa_refcnt);
779 		if (last_ifa != NULL && ifa == last_ifa)
780 			panic("if_detach: loop detected");
781 		last_ifa = ifa;
782 #endif
783 		if (family == AF_LINK)
784 			continue;
785 		dp = pffinddomain(family);
786 #ifdef DIAGNOSTIC
787 		if (dp == NULL)
788 			panic("if_detach: no domain for AF %d",
789 			    family);
790 #endif
791 		/*
792 		 * XXX These PURGEIF calls are redundant with the
793 		 * purge-all-families calls below, but are left in for
794 		 * now both to make a smaller change, and to avoid
795 		 * unplanned interactions with clearing of
796 		 * ifp->if_addrlist.
797 		 */
798 		purged = 0;
799 		for (pr = dp->dom_protosw;
800 		     pr < dp->dom_protoswNPROTOSW; pr++) {
801 			so.so_proto = pr;
802 			if (pr->pr_usrreqs) {
803 				(void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
804 				purged = 1;
805 			}
806 		}
807 		if (purged == 0) {
808 			/*
809 			 * XXX What's really the best thing to do
810 			 * XXX here?  --thorpej@NetBSD.org
811 			 */
812 			printf("if_detach: WARNING: AF %d not purged\n",
813 			    family);
814 			ifa_remove(ifp, ifa);
815 		}
816 		goto again;
817 	}
818 
819 	if_free_sadl(ifp);
820 
821 	/* Walk the routing table looking for stragglers. */
822 	for (i = 0; i <= AF_MAX; i++) {
823 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
824 			continue;
825 	}
826 
827 	DOMAIN_FOREACH(dp) {
828 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
829 		{
830 			void *p = ifp->if_afdata[dp->dom_family];
831 			if (p) {
832 				ifp->if_afdata[dp->dom_family] = NULL;
833 				(*dp->dom_ifdetach)(ifp, p);
834 			}
835 		}
836 
837 		/*
838 		 * One would expect multicast memberships (INET and
839 		 * INET6) on UDP sockets to be purged by the PURGEIF
840 		 * calls above, but if all addresses were removed from
841 		 * the interface prior to destruction, the calls will
842 		 * not be made (e.g. ppp, for which pppd(8) generally
843 		 * removes addresses before destroying the interface).
844 		 * Because there is no invariant that multicast
845 		 * memberships only exist for interfaces with IPv4
846 		 * addresses, we must call PURGEIF regardless of
847 		 * addresses.  (Protocols which might store ifnet
848 		 * pointers are marked with PR_PURGEIF.)
849 		 */
850 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
851 			so.so_proto = pr;
852 			if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF)
853 				(void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
854 		}
855 	}
856 
857 	(void)pfil_run_hooks(if_pfil,
858 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
859 	(void)pfil_head_destroy(ifp->if_pfil);
860 
861 	/* Announce that the interface is gone. */
862 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
863 
864 	ifindex2ifnet[ifp->if_index] = NULL;
865 
866 	TAILQ_REMOVE(&ifnet_list, ifp, if_list);
867 
868 	ifioctl_detach(ifp);
869 
870 	/*
871 	 * remove packets that came from ifp, from software interrupt queues.
872 	 */
873 	DOMAIN_FOREACH(dp) {
874 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
875 			struct ifqueue *iq = dp->dom_ifqueues[i];
876 			if (iq == NULL)
877 				break;
878 			dp->dom_ifqueues[i] = NULL;
879 			if_detach_queues(ifp, iq);
880 		}
881 	}
882 
883 	/*
884 	 * IP queues have to be processed separately: net-queue barrier
885 	 * ensures that the packets are dequeued while a cross-call will
886 	 * ensure that the interrupts have completed. FIXME: not quite..
887 	 */
888 #ifdef INET
889 	pktq_barrier(ip_pktq);
890 #endif
891 #ifdef INET6
892 	pktq_barrier(ip6_pktq);
893 #endif
894 	xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
895 	xc_wait(xc);
896 
897 	splx(s);
898 }
899 
900 static void
901 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
902 {
903 	struct mbuf *m, *prev, *next;
904 
905 	prev = NULL;
906 	for (m = q->ifq_head; m != NULL; m = next) {
907 		KASSERT((m->m_flags & M_PKTHDR) != 0);
908 
909 		next = m->m_nextpkt;
910 		if (m->m_pkthdr.rcvif != ifp) {
911 			prev = m;
912 			continue;
913 		}
914 
915 		if (prev != NULL)
916 			prev->m_nextpkt = m->m_nextpkt;
917 		else
918 			q->ifq_head = m->m_nextpkt;
919 		if (q->ifq_tail == m)
920 			q->ifq_tail = prev;
921 		q->ifq_len--;
922 
923 		m->m_nextpkt = NULL;
924 		m_freem(m);
925 		IF_DROP(q);
926 	}
927 }
928 
929 /*
930  * Callback for a radix tree walk to delete all references to an
931  * ifnet.
932  */
933 static int
934 if_rt_walktree(struct rtentry *rt, void *v)
935 {
936 	struct ifnet *ifp = (struct ifnet *)v;
937 	int error;
938 
939 	if (rt->rt_ifp != ifp)
940 		return 0;
941 
942 	/* Delete the entry. */
943 	++rt->rt_refcnt;
944 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
945 	    rt_mask(rt), rt->rt_flags, NULL);
946 	KASSERT((rt->rt_flags & RTF_UP) == 0);
947 	rt->rt_ifp = NULL;
948 	rtfree(rt);
949 	if (error != 0)
950 		printf("%s: warning: unable to delete rtentry @ %p, "
951 		    "error = %d\n", ifp->if_xname, rt, error);
952 	return ERESTART;
953 }
954 
955 /*
956  * Create a clone network interface.
957  */
958 int
959 if_clone_create(const char *name)
960 {
961 	struct if_clone *ifc;
962 	int unit;
963 
964 	ifc = if_clone_lookup(name, &unit);
965 	if (ifc == NULL)
966 		return EINVAL;
967 
968 	if (ifunit(name) != NULL)
969 		return EEXIST;
970 
971 	return (*ifc->ifc_create)(ifc, unit);
972 }
973 
974 /*
975  * Destroy a clone network interface.
976  */
977 int
978 if_clone_destroy(const char *name)
979 {
980 	struct if_clone *ifc;
981 	struct ifnet *ifp;
982 
983 	ifc = if_clone_lookup(name, NULL);
984 	if (ifc == NULL)
985 		return EINVAL;
986 
987 	ifp = ifunit(name);
988 	if (ifp == NULL)
989 		return ENXIO;
990 
991 	if (ifc->ifc_destroy == NULL)
992 		return EOPNOTSUPP;
993 
994 	return (*ifc->ifc_destroy)(ifp);
995 }
996 
997 /*
998  * Look up a network interface cloner.
999  */
1000 static struct if_clone *
1001 if_clone_lookup(const char *name, int *unitp)
1002 {
1003 	struct if_clone *ifc;
1004 	const char *cp;
1005 	char *dp, ifname[IFNAMSIZ + 3];
1006 	int unit;
1007 
1008 	strcpy(ifname, "if_");
1009 	/* separate interface name from unit */
1010 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
1011 	    *cp && (*cp < '0' || *cp > '9');)
1012 		*dp++ = *cp++;
1013 
1014 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
1015 		return NULL;	/* No name or unit number */
1016 	*dp++ = '\0';
1017 
1018 again:
1019 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
1020 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
1021 			break;
1022 	}
1023 
1024 	if (ifc == NULL) {
1025 		if (*ifname == '\0' ||
1026 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
1027 			return NULL;
1028 		*ifname = '\0';
1029 		goto again;
1030 	}
1031 
1032 	unit = 0;
1033 	while (cp - name < IFNAMSIZ && *cp) {
1034 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1035 			/* Bogus unit number. */
1036 			return NULL;
1037 		}
1038 		unit = (unit * 10) + (*cp++ - '0');
1039 	}
1040 
1041 	if (unitp != NULL)
1042 		*unitp = unit;
1043 	return ifc;
1044 }
1045 
1046 /*
1047  * Register a network interface cloner.
1048  */
1049 void
1050 if_clone_attach(struct if_clone *ifc)
1051 {
1052 
1053 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1054 	if_cloners_count++;
1055 }
1056 
1057 /*
1058  * Unregister a network interface cloner.
1059  */
1060 void
1061 if_clone_detach(struct if_clone *ifc)
1062 {
1063 
1064 	LIST_REMOVE(ifc, ifc_list);
1065 	if_cloners_count--;
1066 }
1067 
1068 /*
1069  * Provide list of interface cloners to userspace.
1070  */
1071 static int
1072 if_clone_list(struct if_clonereq *ifcr)
1073 {
1074 	char outbuf[IFNAMSIZ], *dst;
1075 	struct if_clone *ifc;
1076 	int count, error = 0;
1077 
1078 	ifcr->ifcr_total = if_cloners_count;
1079 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1080 		/* Just asking how many there are. */
1081 		return 0;
1082 	}
1083 
1084 	if (ifcr->ifcr_count < 0)
1085 		return EINVAL;
1086 
1087 	count = (if_cloners_count < ifcr->ifcr_count) ?
1088 	    if_cloners_count : ifcr->ifcr_count;
1089 
1090 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1091 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1092 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1093 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1094 			return ENAMETOOLONG;
1095 		error = copyout(outbuf, dst, sizeof(outbuf));
1096 		if (error != 0)
1097 			break;
1098 	}
1099 
1100 	return error;
1101 }
1102 
1103 void
1104 ifaref(struct ifaddr *ifa)
1105 {
1106 	ifa->ifa_refcnt++;
1107 }
1108 
1109 void
1110 ifafree(struct ifaddr *ifa)
1111 {
1112 	KASSERT(ifa != NULL);
1113 	KASSERT(ifa->ifa_refcnt > 0);
1114 
1115 	if (--ifa->ifa_refcnt == 0) {
1116 		free(ifa, M_IFADDR);
1117 	}
1118 }
1119 
1120 void
1121 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1122 {
1123 	ifa->ifa_ifp = ifp;
1124 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1125 	ifaref(ifa);
1126 }
1127 
1128 void
1129 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1130 {
1131 	KASSERT(ifa->ifa_ifp == ifp);
1132 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1133 	ifafree(ifa);
1134 }
1135 
1136 static inline int
1137 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1138 {
1139 	return sockaddr_cmp(sa1, sa2) == 0;
1140 }
1141 
1142 /*
1143  * Locate an interface based on a complete address.
1144  */
1145 /*ARGSUSED*/
1146 struct ifaddr *
1147 ifa_ifwithaddr(const struct sockaddr *addr)
1148 {
1149 	struct ifnet *ifp;
1150 	struct ifaddr *ifa;
1151 
1152 	IFNET_FOREACH(ifp) {
1153 		if (ifp->if_output == if_nulloutput)
1154 			continue;
1155 		IFADDR_FOREACH(ifa, ifp) {
1156 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1157 				continue;
1158 			if (equal(addr, ifa->ifa_addr))
1159 				return ifa;
1160 			if ((ifp->if_flags & IFF_BROADCAST) &&
1161 			    ifa->ifa_broadaddr &&
1162 			    /* IP6 doesn't have broadcast */
1163 			    ifa->ifa_broadaddr->sa_len != 0 &&
1164 			    equal(ifa->ifa_broadaddr, addr))
1165 				return ifa;
1166 		}
1167 	}
1168 	return NULL;
1169 }
1170 
1171 /*
1172  * Locate the point to point interface with a given destination address.
1173  */
1174 /*ARGSUSED*/
1175 struct ifaddr *
1176 ifa_ifwithdstaddr(const struct sockaddr *addr)
1177 {
1178 	struct ifnet *ifp;
1179 	struct ifaddr *ifa;
1180 
1181 	IFNET_FOREACH(ifp) {
1182 		if (ifp->if_output == if_nulloutput)
1183 			continue;
1184 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1185 			continue;
1186 		IFADDR_FOREACH(ifa, ifp) {
1187 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1188 			    ifa->ifa_dstaddr == NULL)
1189 				continue;
1190 			if (equal(addr, ifa->ifa_dstaddr))
1191 				return ifa;
1192 		}
1193 	}
1194 	return NULL;
1195 }
1196 
1197 /*
1198  * Find an interface on a specific network.  If many, choice
1199  * is most specific found.
1200  */
1201 struct ifaddr *
1202 ifa_ifwithnet(const struct sockaddr *addr)
1203 {
1204 	struct ifnet *ifp;
1205 	struct ifaddr *ifa;
1206 	const struct sockaddr_dl *sdl;
1207 	struct ifaddr *ifa_maybe = 0;
1208 	u_int af = addr->sa_family;
1209 	const char *addr_data = addr->sa_data, *cplim;
1210 
1211 	if (af == AF_LINK) {
1212 		sdl = satocsdl(addr);
1213 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1214 		    ifindex2ifnet[sdl->sdl_index] &&
1215 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1216 			return ifnet_addrs[sdl->sdl_index];
1217 	}
1218 #ifdef NETATALK
1219 	if (af == AF_APPLETALK) {
1220 		const struct sockaddr_at *sat, *sat2;
1221 		sat = (const struct sockaddr_at *)addr;
1222 		IFNET_FOREACH(ifp) {
1223 			if (ifp->if_output == if_nulloutput)
1224 				continue;
1225 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1226 			if (ifa == NULL)
1227 				continue;
1228 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1229 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1230 				return ifa; /* exact match */
1231 			if (ifa_maybe == NULL) {
1232 				/* else keep the if with the right range */
1233 				ifa_maybe = ifa;
1234 			}
1235 		}
1236 		return ifa_maybe;
1237 	}
1238 #endif
1239 	IFNET_FOREACH(ifp) {
1240 		if (ifp->if_output == if_nulloutput)
1241 			continue;
1242 		IFADDR_FOREACH(ifa, ifp) {
1243 			const char *cp, *cp2, *cp3;
1244 
1245 			if (ifa->ifa_addr->sa_family != af ||
1246 			    ifa->ifa_netmask == NULL)
1247  next:				continue;
1248 			cp = addr_data;
1249 			cp2 = ifa->ifa_addr->sa_data;
1250 			cp3 = ifa->ifa_netmask->sa_data;
1251 			cplim = (const char *)ifa->ifa_netmask +
1252 			    ifa->ifa_netmask->sa_len;
1253 			while (cp3 < cplim) {
1254 				if ((*cp++ ^ *cp2++) & *cp3++) {
1255 					/* want to continue for() loop */
1256 					goto next;
1257 				}
1258 			}
1259 			if (ifa_maybe == NULL ||
1260 			    rn_refines((void *)ifa->ifa_netmask,
1261 			    (void *)ifa_maybe->ifa_netmask))
1262 				ifa_maybe = ifa;
1263 		}
1264 	}
1265 	return ifa_maybe;
1266 }
1267 
1268 /*
1269  * Find the interface of the addresss.
1270  */
1271 struct ifaddr *
1272 ifa_ifwithladdr(const struct sockaddr *addr)
1273 {
1274 	struct ifaddr *ia;
1275 
1276 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1277 	    (ia = ifa_ifwithnet(addr)))
1278 		return ia;
1279 	return NULL;
1280 }
1281 
1282 /*
1283  * Find an interface using a specific address family
1284  */
1285 struct ifaddr *
1286 ifa_ifwithaf(int af)
1287 {
1288 	struct ifnet *ifp;
1289 	struct ifaddr *ifa;
1290 
1291 	IFNET_FOREACH(ifp) {
1292 		if (ifp->if_output == if_nulloutput)
1293 			continue;
1294 		IFADDR_FOREACH(ifa, ifp) {
1295 			if (ifa->ifa_addr->sa_family == af)
1296 				return ifa;
1297 		}
1298 	}
1299 	return NULL;
1300 }
1301 
1302 /*
1303  * Find an interface address specific to an interface best matching
1304  * a given address.
1305  */
1306 struct ifaddr *
1307 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1308 {
1309 	struct ifaddr *ifa;
1310 	const char *cp, *cp2, *cp3;
1311 	const char *cplim;
1312 	struct ifaddr *ifa_maybe = 0;
1313 	u_int af = addr->sa_family;
1314 
1315 	if (ifp->if_output == if_nulloutput)
1316 		return NULL;
1317 
1318 	if (af >= AF_MAX)
1319 		return NULL;
1320 
1321 	IFADDR_FOREACH(ifa, ifp) {
1322 		if (ifa->ifa_addr->sa_family != af)
1323 			continue;
1324 		ifa_maybe = ifa;
1325 		if (ifa->ifa_netmask == NULL) {
1326 			if (equal(addr, ifa->ifa_addr) ||
1327 			    (ifa->ifa_dstaddr &&
1328 			     equal(addr, ifa->ifa_dstaddr)))
1329 				return ifa;
1330 			continue;
1331 		}
1332 		cp = addr->sa_data;
1333 		cp2 = ifa->ifa_addr->sa_data;
1334 		cp3 = ifa->ifa_netmask->sa_data;
1335 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1336 		for (; cp3 < cplim; cp3++) {
1337 			if ((*cp++ ^ *cp2++) & *cp3)
1338 				break;
1339 		}
1340 		if (cp3 == cplim)
1341 			return ifa;
1342 	}
1343 	return ifa_maybe;
1344 }
1345 
1346 /*
1347  * Default action when installing a route with a Link Level gateway.
1348  * Lookup an appropriate real ifa to point to.
1349  * This should be moved to /sys/net/link.c eventually.
1350  */
1351 void
1352 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1353 {
1354 	struct ifaddr *ifa;
1355 	const struct sockaddr *dst;
1356 	struct ifnet *ifp;
1357 
1358 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1359 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1360 		return;
1361 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1362 		rt_replace_ifa(rt, ifa);
1363 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1364 			ifa->ifa_rtrequest(cmd, rt, info);
1365 	}
1366 }
1367 
1368 /*
1369  * Handle a change in the interface link state.
1370  * XXX: We should listen to the routing socket in-kernel rather
1371  * than calling in6_if_link_* functions directly from here.
1372  */
1373 void
1374 if_link_state_change(struct ifnet *ifp, int link_state)
1375 {
1376 	int s;
1377 #if defined(DEBUG) || defined(INET6)
1378 	int old_link_state;
1379 #endif
1380 
1381 	s = splnet();
1382 	if (ifp->if_link_state == link_state) {
1383 		splx(s);
1384 		return;
1385 	}
1386 
1387 #if defined(DEBUG) || defined(INET6)
1388 	old_link_state = ifp->if_link_state;
1389 #endif
1390 	ifp->if_link_state = link_state;
1391 #ifdef DEBUG
1392 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
1393 		link_state == LINK_STATE_UP ? "UP" :
1394 		link_state == LINK_STATE_DOWN ? "DOWN" :
1395 		"UNKNOWN",
1396 		 old_link_state == LINK_STATE_UP ? "UP" :
1397 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
1398 		"UNKNOWN");
1399 #endif
1400 
1401 #ifdef INET6
1402 	/*
1403 	 * When going from UNKNOWN to UP, we need to mark existing
1404 	 * IPv6 addresses as tentative and restart DAD as we may have
1405 	 * erroneously not found a duplicate.
1406 	 *
1407 	 * This needs to happen before rt_ifmsg to avoid a race where
1408 	 * listeners would have an address and expect it to work right
1409 	 * away.
1410 	 */
1411 	if (in6_present && link_state == LINK_STATE_UP &&
1412 	    old_link_state == LINK_STATE_UNKNOWN)
1413 		in6_if_link_down(ifp);
1414 #endif
1415 
1416 	/* Notify that the link state has changed. */
1417 	rt_ifmsg(ifp);
1418 
1419 #if NCARP > 0
1420 	if (ifp->if_carp)
1421 		carp_carpdev_state(ifp);
1422 #endif
1423 
1424 #ifdef INET6
1425 	if (in6_present) {
1426 		if (link_state == LINK_STATE_DOWN)
1427 			in6_if_link_down(ifp);
1428 		else if (link_state == LINK_STATE_UP)
1429 			in6_if_link_up(ifp);
1430 	}
1431 #endif
1432 
1433 	splx(s);
1434 }
1435 
1436 /*
1437  * Mark an interface down and notify protocols of
1438  * the transition.
1439  * NOTE: must be called at splsoftnet or equivalent.
1440  */
1441 void
1442 if_down(struct ifnet *ifp)
1443 {
1444 	struct ifaddr *ifa;
1445 
1446 	ifp->if_flags &= ~IFF_UP;
1447 	nanotime(&ifp->if_lastchange);
1448 	IFADDR_FOREACH(ifa, ifp)
1449 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1450 	IFQ_PURGE(&ifp->if_snd);
1451 #if NCARP > 0
1452 	if (ifp->if_carp)
1453 		carp_carpdev_state(ifp);
1454 #endif
1455 	rt_ifmsg(ifp);
1456 #ifdef INET6
1457 	if (in6_present)
1458 		in6_if_down(ifp);
1459 #endif
1460 }
1461 
1462 /*
1463  * Mark an interface up and notify protocols of
1464  * the transition.
1465  * NOTE: must be called at splsoftnet or equivalent.
1466  */
1467 void
1468 if_up(struct ifnet *ifp)
1469 {
1470 #ifdef notyet
1471 	struct ifaddr *ifa;
1472 #endif
1473 
1474 	ifp->if_flags |= IFF_UP;
1475 	nanotime(&ifp->if_lastchange);
1476 #ifdef notyet
1477 	/* this has no effect on IP, and will kill all ISO connections XXX */
1478 	IFADDR_FOREACH(ifa, ifp)
1479 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1480 #endif
1481 #if NCARP > 0
1482 	if (ifp->if_carp)
1483 		carp_carpdev_state(ifp);
1484 #endif
1485 	rt_ifmsg(ifp);
1486 #ifdef INET6
1487 	if (in6_present)
1488 		in6_if_up(ifp);
1489 #endif
1490 }
1491 
1492 /*
1493  * Handle interface watchdog timer routines.  Called
1494  * from softclock, we decrement timers (if set) and
1495  * call the appropriate interface routine on expiration.
1496  */
1497 void
1498 if_slowtimo(void *arg)
1499 {
1500 	struct ifnet *ifp;
1501 	int s = splnet();
1502 
1503 	IFNET_FOREACH(ifp) {
1504 		if (ifp->if_timer == 0 || --ifp->if_timer)
1505 			continue;
1506 		if (ifp->if_watchdog != NULL)
1507 			(*ifp->if_watchdog)(ifp);
1508 	}
1509 	splx(s);
1510 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1511 }
1512 
1513 /*
1514  * Set/clear promiscuous mode on interface ifp based on the truth value
1515  * of pswitch.  The calls are reference counted so that only the first
1516  * "on" request actually has an effect, as does the final "off" request.
1517  * Results are undefined if the "off" and "on" requests are not matched.
1518  */
1519 int
1520 ifpromisc(struct ifnet *ifp, int pswitch)
1521 {
1522 	int pcount, ret;
1523 	short nflags;
1524 
1525 	pcount = ifp->if_pcount;
1526 	if (pswitch) {
1527 		/*
1528 		 * Allow the device to be "placed" into promiscuous
1529 		 * mode even if it is not configured up.  It will
1530 		 * consult IFF_PROMISC when it is brought up.
1531 		 */
1532 		if (ifp->if_pcount++ != 0)
1533 			return 0;
1534 		nflags = ifp->if_flags | IFF_PROMISC;
1535 	} else {
1536 		if (--ifp->if_pcount > 0)
1537 			return 0;
1538 		nflags = ifp->if_flags & ~IFF_PROMISC;
1539 	}
1540 	ret = if_flags_set(ifp, nflags);
1541 	/* Restore interface state if not successful. */
1542 	if (ret != 0) {
1543 		ifp->if_pcount = pcount;
1544 	}
1545 	return ret;
1546 }
1547 
1548 /*
1549  * Map interface name to
1550  * interface structure pointer.
1551  */
1552 struct ifnet *
1553 ifunit(const char *name)
1554 {
1555 	struct ifnet *ifp;
1556 	const char *cp = name;
1557 	u_int unit = 0;
1558 	u_int i;
1559 
1560 	/*
1561 	 * If the entire name is a number, treat it as an ifindex.
1562 	 */
1563 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1564 		unit = unit * 10 + (*cp - '0');
1565 	}
1566 
1567 	/*
1568 	 * If the number took all of the name, then it's a valid ifindex.
1569 	 */
1570 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1571 		if (unit >= if_indexlim)
1572 			return NULL;
1573 		ifp = ifindex2ifnet[unit];
1574 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1575 			return NULL;
1576 		return ifp;
1577 	}
1578 
1579 	IFNET_FOREACH(ifp) {
1580 		if (ifp->if_output == if_nulloutput)
1581 			continue;
1582 	 	if (strcmp(ifp->if_xname, name) == 0)
1583 			return ifp;
1584 	}
1585 	return NULL;
1586 }
1587 
1588 ifnet_t *
1589 if_byindex(u_int idx)
1590 {
1591 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1592 }
1593 
1594 /* common */
1595 int
1596 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1597 {
1598 	int s;
1599 	struct ifreq *ifr;
1600 	struct ifcapreq *ifcr;
1601 	struct ifdatareq *ifdr;
1602 
1603 	switch (cmd) {
1604 	case SIOCSIFCAP:
1605 		ifcr = data;
1606 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1607 			return EINVAL;
1608 
1609 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1610 			return 0;
1611 
1612 		ifp->if_capenable = ifcr->ifcr_capenable;
1613 
1614 		/* Pre-compute the checksum flags mask. */
1615 		ifp->if_csum_flags_tx = 0;
1616 		ifp->if_csum_flags_rx = 0;
1617 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1618 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1619 		}
1620 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1621 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1622 		}
1623 
1624 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1625 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1626 		}
1627 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1628 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1629 		}
1630 
1631 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1632 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1633 		}
1634 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1635 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1636 		}
1637 
1638 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1639 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1640 		}
1641 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1642 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1643 		}
1644 
1645 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1646 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1647 		}
1648 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1649 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1650 		}
1651 		if (ifp->if_flags & IFF_UP)
1652 			return ENETRESET;
1653 		return 0;
1654 	case SIOCSIFFLAGS:
1655 		ifr = data;
1656 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1657 			s = splnet();
1658 			if_down(ifp);
1659 			splx(s);
1660 		}
1661 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1662 			s = splnet();
1663 			if_up(ifp);
1664 			splx(s);
1665 		}
1666 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1667 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1668 		break;
1669 	case SIOCGIFFLAGS:
1670 		ifr = data;
1671 		ifr->ifr_flags = ifp->if_flags;
1672 		break;
1673 
1674 	case SIOCGIFMETRIC:
1675 		ifr = data;
1676 		ifr->ifr_metric = ifp->if_metric;
1677 		break;
1678 
1679 	case SIOCGIFMTU:
1680 		ifr = data;
1681 		ifr->ifr_mtu = ifp->if_mtu;
1682 		break;
1683 
1684 	case SIOCGIFDLT:
1685 		ifr = data;
1686 		ifr->ifr_dlt = ifp->if_dlt;
1687 		break;
1688 
1689 	case SIOCGIFCAP:
1690 		ifcr = data;
1691 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1692 		ifcr->ifcr_capenable = ifp->if_capenable;
1693 		break;
1694 
1695 	case SIOCSIFMETRIC:
1696 		ifr = data;
1697 		ifp->if_metric = ifr->ifr_metric;
1698 		break;
1699 
1700 	case SIOCGIFDATA:
1701 		ifdr = data;
1702 		ifdr->ifdr_data = ifp->if_data;
1703 		break;
1704 
1705 	case SIOCGIFINDEX:
1706 		ifr = data;
1707 		ifr->ifr_index = ifp->if_index;
1708 		break;
1709 
1710 	case SIOCZIFDATA:
1711 		ifdr = data;
1712 		ifdr->ifdr_data = ifp->if_data;
1713 		/*
1714 		 * Assumes that the volatile counters that can be
1715 		 * zero'ed are at the end of if_data.
1716 		 */
1717 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1718 		    offsetof(struct if_data, ifi_ipackets));
1719 		/*
1720 		 * The memset() clears to the bottm of if_data. In the area,
1721 		 * if_lastchange is included. Please be careful if new entry
1722 		 * will be added into if_data or rewite this.
1723 		 *
1724 		 * And also, update if_lastchnage.
1725 		 */
1726 		getnanotime(&ifp->if_lastchange);
1727 		break;
1728 	case SIOCSIFMTU:
1729 		ifr = data;
1730 		if (ifp->if_mtu == ifr->ifr_mtu)
1731 			break;
1732 		ifp->if_mtu = ifr->ifr_mtu;
1733 		/*
1734 		 * If the link MTU changed, do network layer specific procedure.
1735 		 */
1736 #ifdef INET6
1737 		if (in6_present)
1738 			nd6_setmtu(ifp);
1739 #endif
1740 		return ENETRESET;
1741 	default:
1742 		return ENOTTY;
1743 	}
1744 	return 0;
1745 }
1746 
1747 int
1748 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
1749 {
1750 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1751 	struct ifaddr *ifa;
1752 	const struct sockaddr *any, *sa;
1753 	union {
1754 		struct sockaddr sa;
1755 		struct sockaddr_storage ss;
1756 	} u, v;
1757 
1758 	switch (cmd) {
1759 	case SIOCSIFADDRPREF:
1760 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
1761 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1762 		    NULL) != 0)
1763 			return EPERM;
1764 	case SIOCGIFADDRPREF:
1765 		break;
1766 	default:
1767 		return EOPNOTSUPP;
1768 	}
1769 
1770 	/* sanity checks */
1771 	if (data == NULL || ifp == NULL) {
1772 		panic("invalid argument to %s", __func__);
1773 		/*NOTREACHED*/
1774 	}
1775 
1776 	/* address must be specified on ADD and DELETE */
1777 	sa = sstocsa(&ifap->ifap_addr);
1778 	if (sa->sa_family != sofamily(so))
1779 		return EINVAL;
1780 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1781 		return EINVAL;
1782 
1783 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1784 
1785 	IFADDR_FOREACH(ifa, ifp) {
1786 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1787 			continue;
1788 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1789 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1790 			break;
1791 	}
1792 	if (ifa == NULL)
1793 		return EADDRNOTAVAIL;
1794 
1795 	switch (cmd) {
1796 	case SIOCSIFADDRPREF:
1797 		ifa->ifa_preference = ifap->ifap_preference;
1798 		return 0;
1799 	case SIOCGIFADDRPREF:
1800 		/* fill in the if_laddrreq structure */
1801 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1802 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1803 		ifap->ifap_preference = ifa->ifa_preference;
1804 		return 0;
1805 	default:
1806 		return EOPNOTSUPP;
1807 	}
1808 }
1809 
1810 static void
1811 ifnet_lock_enter(struct ifnet_lock *il)
1812 {
1813 	uint64_t *nenter;
1814 
1815 	/* Before trying to acquire the mutex, increase the count of threads
1816 	 * who have entered or who wait to enter the critical section.
1817 	 * Avoid one costly locked memory transaction by keeping a count for
1818 	 * each CPU.
1819 	 */
1820 	nenter = percpu_getref(il->il_nenter);
1821 	(*nenter)++;
1822 	percpu_putref(il->il_nenter);
1823 	mutex_enter(&il->il_lock);
1824 }
1825 
1826 static void
1827 ifnet_lock_exit(struct ifnet_lock *il)
1828 {
1829 	/* Increase the count of threads who have exited the critical
1830 	 * section.  Increase while we still hold the lock.
1831 	 */
1832 	il->il_nexit++;
1833 	mutex_exit(&il->il_lock);
1834 }
1835 
1836 /*
1837  * Interface ioctls.
1838  */
1839 static int
1840 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1841 {
1842 	struct ifnet *ifp;
1843 	struct ifreq *ifr;
1844 	int error = 0;
1845 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1846 	u_long ocmd = cmd;
1847 #endif
1848 	short oif_flags;
1849 #ifdef COMPAT_OIFREQ
1850 	struct ifreq ifrb;
1851 	struct oifreq *oifr = NULL;
1852 #endif
1853 
1854 	switch (cmd) {
1855 #ifdef COMPAT_OIFREQ
1856 	case OSIOCGIFCONF:
1857 	case OOSIOCGIFCONF:
1858 		return compat_ifconf(cmd, data);
1859 #endif
1860 #ifdef COMPAT_OIFDATA
1861 	case OSIOCGIFDATA:
1862 	case OSIOCZIFDATA:
1863 		return compat_ifdatareq(l, cmd, data);
1864 #endif
1865 	case SIOCGIFCONF:
1866 		return ifconf(cmd, data);
1867 	case SIOCINITIFADDR:
1868 		return EPERM;
1869 	}
1870 
1871 #ifdef COMPAT_OIFREQ
1872 	cmd = compat_cvtcmd(cmd);
1873 	if (cmd != ocmd) {
1874 		oifr = data;
1875 		data = ifr = &ifrb;
1876 		ifreqo2n(oifr, ifr);
1877 	} else
1878 #endif
1879 		ifr = data;
1880 
1881 	ifp = ifunit(ifr->ifr_name);
1882 
1883 	switch (cmd) {
1884 	case SIOCIFCREATE:
1885 	case SIOCIFDESTROY:
1886 		if (l != NULL) {
1887 			error = kauth_authorize_network(l->l_cred,
1888 			    KAUTH_NETWORK_INTERFACE,
1889 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1890 			    (void *)cmd, NULL);
1891 			if (error != 0)
1892 				return error;
1893 		}
1894 		return (cmd == SIOCIFCREATE) ?
1895 			if_clone_create(ifr->ifr_name) :
1896 			if_clone_destroy(ifr->ifr_name);
1897 
1898 	case SIOCIFGCLONERS:
1899 		return if_clone_list((struct if_clonereq *)data);
1900 	}
1901 
1902 	if (ifp == NULL)
1903 		return ENXIO;
1904 
1905 	switch (cmd) {
1906 	case SIOCALIFADDR:
1907 	case SIOCDLIFADDR:
1908 	case SIOCSIFADDRPREF:
1909 	case SIOCSIFFLAGS:
1910 	case SIOCSIFCAP:
1911 	case SIOCSIFMETRIC:
1912 	case SIOCZIFDATA:
1913 	case SIOCSIFMTU:
1914 	case SIOCSIFPHYADDR:
1915 	case SIOCDIFPHYADDR:
1916 #ifdef INET6
1917 	case SIOCSIFPHYADDR_IN6:
1918 #endif
1919 	case SIOCSLIFPHYADDR:
1920 	case SIOCADDMULTI:
1921 	case SIOCDELMULTI:
1922 	case SIOCSIFMEDIA:
1923 	case SIOCSDRVSPEC:
1924 	case SIOCG80211:
1925 	case SIOCS80211:
1926 	case SIOCS80211NWID:
1927 	case SIOCS80211NWKEY:
1928 	case SIOCS80211POWER:
1929 	case SIOCS80211BSSID:
1930 	case SIOCS80211CHANNEL:
1931 	case SIOCSLINKSTR:
1932 		if (l != NULL) {
1933 			error = kauth_authorize_network(l->l_cred,
1934 			    KAUTH_NETWORK_INTERFACE,
1935 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1936 			    (void *)cmd, NULL);
1937 			if (error != 0)
1938 				return error;
1939 		}
1940 	}
1941 
1942 	oif_flags = ifp->if_flags;
1943 
1944 	ifnet_lock_enter(ifp->if_ioctl_lock);
1945 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1946 	if (error != ENOTTY)
1947 		;
1948 	else if (so->so_proto == NULL)
1949 		error = EOPNOTSUPP;
1950 	else {
1951 #ifdef COMPAT_OSOCK
1952 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1953 #else
1954 		error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so,
1955 		    cmd, data, ifp);
1956 #endif
1957 	}
1958 
1959 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1960 #ifdef INET6
1961 		if (in6_present && (ifp->if_flags & IFF_UP) != 0) {
1962 			int s = splnet();
1963 			in6_if_up(ifp);
1964 			splx(s);
1965 		}
1966 #endif
1967 	}
1968 #ifdef COMPAT_OIFREQ
1969 	if (cmd != ocmd)
1970 		ifreqn2o(oifr, ifr);
1971 #endif
1972 
1973 	ifnet_lock_exit(ifp->if_ioctl_lock);
1974 	return error;
1975 }
1976 
1977 /* This callback adds to the sum in `arg' the number of
1978  * threads on `ci' who have entered or who wait to enter the
1979  * critical section.
1980  */
1981 static void
1982 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
1983 {
1984 	uint64_t *sum = arg, *nenter = p;
1985 
1986 	*sum += *nenter;
1987 }
1988 
1989 /* Return the number of threads who have entered or who wait
1990  * to enter the critical section on all CPUs.
1991  */
1992 static uint64_t
1993 ifnet_lock_entrances(struct ifnet_lock *il)
1994 {
1995 	uint64_t sum = 0;
1996 
1997 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
1998 
1999 	return sum;
2000 }
2001 
2002 static int
2003 ifioctl_attach(struct ifnet *ifp)
2004 {
2005 	struct ifnet_lock *il;
2006 
2007 	/* If the driver has not supplied its own if_ioctl, then
2008 	 * supply the default.
2009 	 */
2010 	if (ifp->if_ioctl == NULL)
2011 		ifp->if_ioctl = ifioctl_common;
2012 
2013 	/* Create an ifnet_lock for synchronizing ifioctls. */
2014 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
2015 		return ENOMEM;
2016 
2017 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
2018 	if (il->il_nenter == NULL) {
2019 		kmem_free(il, sizeof(*il));
2020 		return ENOMEM;
2021 	}
2022 
2023 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
2024 	cv_init(&il->il_emptied, ifp->if_xname);
2025 
2026 	ifp->if_ioctl_lock = il;
2027 
2028 	return 0;
2029 }
2030 
2031 /*
2032  * This must not be called until after `ifp' has been withdrawn from the
2033  * ifnet tables so that ifioctl() cannot get a handle on it by calling
2034  * ifunit().
2035  */
2036 static void
2037 ifioctl_detach(struct ifnet *ifp)
2038 {
2039 	struct ifnet_lock *il;
2040 
2041 	il = ifp->if_ioctl_lock;
2042 	mutex_enter(&il->il_lock);
2043 	/* Install if_nullioctl to make sure that any thread that
2044 	 * subsequently enters the critical section will quit it
2045 	 * immediately and signal the condition variable that we
2046 	 * wait on, below.
2047 	 */
2048 	ifp->if_ioctl = if_nullioctl;
2049 	/* Sleep while threads are still in the critical section or
2050 	 * wait to enter it.
2051 	 */
2052 	while (ifnet_lock_entrances(il) != il->il_nexit)
2053 		cv_wait(&il->il_emptied, &il->il_lock);
2054 	/* At this point, we are the only thread still in the critical
2055 	 * section, and no new thread can get a handle on the ifioctl
2056 	 * lock, so it is safe to free its memory.
2057 	 */
2058 	mutex_exit(&il->il_lock);
2059 	ifp->if_ioctl_lock = NULL;
2060 	percpu_free(il->il_nenter, sizeof(uint64_t));
2061 	il->il_nenter = NULL;
2062 	cv_destroy(&il->il_emptied);
2063 	mutex_destroy(&il->il_lock);
2064 	kmem_free(il, sizeof(*il));
2065 }
2066 
2067 /*
2068  * Return interface configuration
2069  * of system.  List may be used
2070  * in later ioctl's (above) to get
2071  * other information.
2072  *
2073  * Each record is a struct ifreq.  Before the addition of
2074  * sockaddr_storage, the API rule was that sockaddr flavors that did
2075  * not fit would extend beyond the struct ifreq, with the next struct
2076  * ifreq starting sa_len beyond the struct sockaddr.  Because the
2077  * union in struct ifreq includes struct sockaddr_storage, every kind
2078  * of sockaddr must fit.  Thus, there are no longer any overlength
2079  * records.
2080  *
2081  * Records are added to the user buffer if they fit, and ifc_len is
2082  * adjusted to the length that was written.  Thus, the user is only
2083  * assured of getting the complete list if ifc_len on return is at
2084  * least sizeof(struct ifreq) less than it was on entry.
2085  *
2086  * If the user buffer pointer is NULL, this routine copies no data and
2087  * returns the amount of space that would be needed.
2088  *
2089  * Invariants:
2090  * ifrp points to the next part of the user's buffer to be used.  If
2091  * ifrp != NULL, space holds the number of bytes remaining that we may
2092  * write at ifrp.  Otherwise, space holds the number of bytes that
2093  * would have been written had there been adequate space.
2094  */
2095 /*ARGSUSED*/
2096 int
2097 ifconf(u_long cmd, void *data)
2098 {
2099 	struct ifconf *ifc = (struct ifconf *)data;
2100 	struct ifnet *ifp;
2101 	struct ifaddr *ifa;
2102 	struct ifreq ifr, *ifrp;
2103 	int space, error = 0;
2104 	const int sz = (int)sizeof(struct ifreq);
2105 
2106 	if ((ifrp = ifc->ifc_req) == NULL)
2107 		space = 0;
2108 	else
2109 		space = ifc->ifc_len;
2110 	IFNET_FOREACH(ifp) {
2111 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2112 		    sizeof(ifr.ifr_name));
2113 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2114 			return ENAMETOOLONG;
2115 		if (IFADDR_EMPTY(ifp)) {
2116 			/* Interface with no addresses - send zero sockaddr. */
2117 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2118 			if (ifrp == NULL) {
2119 				space += sz;
2120 				continue;
2121 			}
2122 			if (space >= sz) {
2123 				error = copyout(&ifr, ifrp, sz);
2124 				if (error != 0)
2125 					return error;
2126 				ifrp++;
2127 				space -= sz;
2128 			}
2129 		}
2130 
2131 		IFADDR_FOREACH(ifa, ifp) {
2132 			struct sockaddr *sa = ifa->ifa_addr;
2133 			/* all sockaddrs must fit in sockaddr_storage */
2134 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2135 
2136 			if (ifrp == NULL) {
2137 				space += sz;
2138 				continue;
2139 			}
2140 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2141 			if (space >= sz) {
2142 				error = copyout(&ifr, ifrp, sz);
2143 				if (error != 0)
2144 					return (error);
2145 				ifrp++; space -= sz;
2146 			}
2147 		}
2148 	}
2149 	if (ifrp != NULL) {
2150 		KASSERT(0 <= space && space <= ifc->ifc_len);
2151 		ifc->ifc_len -= space;
2152 	} else {
2153 		KASSERT(space >= 0);
2154 		ifc->ifc_len = space;
2155 	}
2156 	return (0);
2157 }
2158 
2159 int
2160 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2161 {
2162 	uint8_t len;
2163 #ifdef COMPAT_OIFREQ
2164 	struct ifreq ifrb;
2165 	struct oifreq *oifr = NULL;
2166 	u_long ocmd = cmd;
2167 	cmd = compat_cvtcmd(cmd);
2168 	if (cmd != ocmd) {
2169 		oifr = (struct oifreq *)(void *)ifr;
2170 		ifr = &ifrb;
2171 		ifreqo2n(oifr, ifr);
2172 		len = sizeof(oifr->ifr_addr);
2173 	} else
2174 #endif
2175 		len = sizeof(ifr->ifr_ifru.ifru_space);
2176 
2177 	if (len < sa->sa_len)
2178 		return EFBIG;
2179 
2180 	memset(&ifr->ifr_addr, 0, len);
2181 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2182 
2183 #ifdef COMPAT_OIFREQ
2184 	if (cmd != ocmd)
2185 		ifreqn2o(oifr, ifr);
2186 #endif
2187 	return 0;
2188 }
2189 
2190 /*
2191  * Queue message on interface, and start output if interface
2192  * not yet active.
2193  */
2194 int
2195 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2196     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2197 {
2198 	int len = m->m_pkthdr.len;
2199 	int mflags = m->m_flags;
2200 	int s = splnet();
2201 	int error;
2202 
2203 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2204 	if (error != 0)
2205 		goto out;
2206 	ifp->if_obytes += len;
2207 	if (mflags & M_MCAST)
2208 		ifp->if_omcasts++;
2209 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2210 		(*ifp->if_start)(ifp);
2211 out:
2212 	splx(s);
2213 	return error;
2214 }
2215 
2216 /*
2217  * Queue message on interface, possibly using a second fast queue
2218  */
2219 int
2220 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2221     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2222 {
2223 	int error = 0;
2224 
2225 	if (ifq != NULL
2226 #ifdef ALTQ
2227 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2228 #endif
2229 	    ) {
2230 		if (IF_QFULL(ifq)) {
2231 			IF_DROP(&ifp->if_snd);
2232 			m_freem(m);
2233 			if (error == 0)
2234 				error = ENOBUFS;
2235 		} else
2236 			IF_ENQUEUE(ifq, m);
2237 	} else
2238 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2239 	if (error != 0) {
2240 		++ifp->if_oerrors;
2241 		return error;
2242 	}
2243 	return 0;
2244 }
2245 
2246 int
2247 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2248 {
2249 	int rc;
2250 
2251 	if (ifp->if_initaddr != NULL)
2252 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2253 	else if (src ||
2254 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2255 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2256 
2257 	return rc;
2258 }
2259 
2260 int
2261 if_flags_set(ifnet_t *ifp, const short flags)
2262 {
2263 	int rc;
2264 
2265 	if (ifp->if_setflags != NULL)
2266 		rc = (*ifp->if_setflags)(ifp, flags);
2267 	else {
2268 		short cantflags, chgdflags;
2269 		struct ifreq ifr;
2270 
2271 		chgdflags = ifp->if_flags ^ flags;
2272 		cantflags = chgdflags & IFF_CANTCHANGE;
2273 
2274 		if (cantflags != 0)
2275 			ifp->if_flags ^= cantflags;
2276 
2277                 /* Traditionally, we do not call if_ioctl after
2278                  * setting/clearing only IFF_PROMISC if the interface
2279                  * isn't IFF_UP.  Uphold that tradition.
2280 		 */
2281 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2282 			return 0;
2283 
2284 		memset(&ifr, 0, sizeof(ifr));
2285 
2286 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2287 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2288 
2289 		if (rc != 0 && cantflags != 0)
2290 			ifp->if_flags ^= cantflags;
2291 	}
2292 
2293 	return rc;
2294 }
2295 
2296 int
2297 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2298 {
2299 	int rc;
2300 	struct ifreq ifr;
2301 
2302 	if (ifp->if_mcastop != NULL)
2303 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2304 	else {
2305 		ifreq_setaddr(cmd, &ifr, sa);
2306 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2307 	}
2308 
2309 	return rc;
2310 }
2311 
2312 static void
2313 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2314     struct ifaltq *ifq)
2315 {
2316 	const struct sysctlnode *cnode, *rnode;
2317 
2318 	if (sysctl_createv(clog, 0, NULL, &rnode,
2319 		       CTLFLAG_PERMANENT,
2320 		       CTLTYPE_NODE, "interfaces",
2321 		       SYSCTL_DESCR("Per-interface controls"),
2322 		       NULL, 0, NULL, 0,
2323 		       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
2324 		goto bad;
2325 
2326 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2327 		       CTLFLAG_PERMANENT,
2328 		       CTLTYPE_NODE, ifname,
2329 		       SYSCTL_DESCR("Interface controls"),
2330 		       NULL, 0, NULL, 0,
2331 		       CTL_CREATE, CTL_EOL) != 0)
2332 		goto bad;
2333 
2334 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2335 		       CTLFLAG_PERMANENT,
2336 		       CTLTYPE_NODE, "sndq",
2337 		       SYSCTL_DESCR("Interface output queue controls"),
2338 		       NULL, 0, NULL, 0,
2339 		       CTL_CREATE, CTL_EOL) != 0)
2340 		goto bad;
2341 
2342 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2343 		       CTLFLAG_PERMANENT,
2344 		       CTLTYPE_INT, "len",
2345 		       SYSCTL_DESCR("Current output queue length"),
2346 		       NULL, 0, &ifq->ifq_len, 0,
2347 		       CTL_CREATE, CTL_EOL) != 0)
2348 		goto bad;
2349 
2350 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2351 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2352 		       CTLTYPE_INT, "maxlen",
2353 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2354 		       NULL, 0, &ifq->ifq_maxlen, 0,
2355 		       CTL_CREATE, CTL_EOL) != 0)
2356 		goto bad;
2357 
2358 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2359 		       CTLFLAG_PERMANENT,
2360 		       CTLTYPE_INT, "drops",
2361 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2362 		       NULL, 0, &ifq->ifq_drops, 0,
2363 		       CTL_CREATE, CTL_EOL) != 0)
2364 		goto bad;
2365 
2366 	return;
2367 bad:
2368 	printf("%s: could not attach sysctl nodes\n", ifname);
2369 	return;
2370 }
2371 
2372 #if defined(INET) || defined(INET6)
2373 
2374 #define	SYSCTL_NET_PKTQ(q, cn, c)					\
2375 	static int							\
2376 	sysctl_net_##q##_##cn(SYSCTLFN_ARGS)				\
2377 	{								\
2378 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c);	\
2379 	}
2380 
2381 #if defined(INET)
2382 static int
2383 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS)
2384 {
2385 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq);
2386 }
2387 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS)
2388 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS)
2389 #endif
2390 
2391 #if defined(INET6)
2392 static int
2393 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS)
2394 {
2395 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq);
2396 }
2397 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS)
2398 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS)
2399 #endif
2400 
2401 static void
2402 sysctl_net_pktq_setup(struct sysctllog **clog, int pf)
2403 {
2404 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
2405 	const char *pfname = NULL, *ipname = NULL;
2406 	int ipn = 0, qid = 0;
2407 
2408 	switch (pf) {
2409 #if defined(INET)
2410 	case PF_INET:
2411 		len_func = sysctl_net_ip_pktq_items;
2412 		maxlen_func = sysctl_net_ip_pktq_maxlen;
2413 		drops_func = sysctl_net_ip_pktq_drops;
2414 		pfname = "inet", ipn = IPPROTO_IP;
2415 		ipname = "ip", qid = IPCTL_IFQ;
2416 		break;
2417 #endif
2418 #if defined(INET6)
2419 	case PF_INET6:
2420 		len_func = sysctl_net_ip6_pktq_items;
2421 		maxlen_func = sysctl_net_ip6_pktq_maxlen;
2422 		drops_func = sysctl_net_ip6_pktq_drops;
2423 		pfname = "inet6", ipn = IPPROTO_IPV6;
2424 		ipname = "ip6", qid = IPV6CTL_IFQ;
2425 		break;
2426 #endif
2427 	default:
2428 		KASSERT(false);
2429 	}
2430 
2431 	sysctl_createv(clog, 0, NULL, NULL,
2432 		       CTLFLAG_PERMANENT,
2433 		       CTLTYPE_NODE, pfname, NULL,
2434 		       NULL, 0, NULL, 0,
2435 		       CTL_NET, pf, CTL_EOL);
2436 	sysctl_createv(clog, 0, NULL, NULL,
2437 		       CTLFLAG_PERMANENT,
2438 		       CTLTYPE_NODE, ipname, NULL,
2439 		       NULL, 0, NULL, 0,
2440 		       CTL_NET, pf, ipn, CTL_EOL);
2441 	sysctl_createv(clog, 0, NULL, NULL,
2442 		       CTLFLAG_PERMANENT,
2443 		       CTLTYPE_NODE, "ifq",
2444 		       SYSCTL_DESCR("Protocol input queue controls"),
2445 		       NULL, 0, NULL, 0,
2446 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2447 
2448 	sysctl_createv(clog, 0, NULL, NULL,
2449 		       CTLFLAG_PERMANENT,
2450 		       CTLTYPE_INT, "len",
2451 		       SYSCTL_DESCR("Current input queue length"),
2452 		       len_func, 0, NULL, 0,
2453 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2454 	sysctl_createv(clog, 0, NULL, NULL,
2455 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2456 		       CTLTYPE_INT, "maxlen",
2457 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2458 		       maxlen_func, 0, NULL, 0,
2459 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2460 	sysctl_createv(clog, 0, NULL, NULL,
2461 		       CTLFLAG_PERMANENT,
2462 		       CTLTYPE_INT, "drops",
2463 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2464 		       drops_func, 0, NULL, 0,
2465 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2466 }
2467 #endif /* INET || INET6 */
2468 
2469 static int
2470 if_sdl_sysctl(SYSCTLFN_ARGS)
2471 {
2472 	struct ifnet *ifp;
2473 	const struct sockaddr_dl *sdl;
2474 
2475 	if (namelen != 1)
2476 		return EINVAL;
2477 
2478 	ifp = if_byindex(name[0]);
2479 	if (ifp == NULL)
2480 		return ENODEV;
2481 
2482 	sdl = ifp->if_sadl;
2483 	if (sdl == NULL) {
2484 		*oldlenp = 0;
2485 		return 0;
2486 	}
2487 
2488 	if (oldp == NULL) {
2489 		*oldlenp = sdl->sdl_alen;
2490 		return 0;
2491 	}
2492 
2493 	if (*oldlenp >= sdl->sdl_alen)
2494 		*oldlenp = sdl->sdl_alen;
2495 	return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp);
2496 }
2497 
2498 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup")
2499 {
2500 	const struct sysctlnode *rnode = NULL;
2501 
2502 	sysctl_createv(clog, 0, NULL, &rnode,
2503 		       CTLFLAG_PERMANENT,
2504 		       CTLTYPE_NODE, "sdl",
2505 		       SYSCTL_DESCR("Get active link-layer address"),
2506 		       if_sdl_sysctl, 0, NULL, 0,
2507 		       CTL_NET, CTL_CREATE, CTL_EOL);
2508 }
2509