xref: /netbsd-src/sys/net/if.c (revision 413d532bcc3f62d122e56d92e13ac64825a40baf)
1 /*	$NetBSD: if.c,v 1.273 2014/04/26 11:16:22 pooka 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 1.273 2014/04/26 11:16:22 pooka Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 
100 #include <sys/param.h>
101 #include <sys/mbuf.h>
102 #include <sys/systm.h>
103 #include <sys/callout.h>
104 #include <sys/proc.h>
105 #include <sys/socket.h>
106 #include <sys/socketvar.h>
107 #include <sys/domain.h>
108 #include <sys/protosw.h>
109 #include <sys/kernel.h>
110 #include <sys/ioctl.h>
111 #include <sys/sysctl.h>
112 #include <sys/syslog.h>
113 #include <sys/kauth.h>
114 #include <sys/kmem.h>
115 
116 #include <net/if.h>
117 #include <net/if_dl.h>
118 #include <net/if_ether.h>
119 #include <net/if_media.h>
120 #include <net80211/ieee80211.h>
121 #include <net80211/ieee80211_ioctl.h>
122 #include <net/if_types.h>
123 #include <net/radix.h>
124 #include <net/route.h>
125 #include <net/netisr.h>
126 #include <sys/module.h>
127 #ifdef NETATALK
128 #include <netatalk/at_extern.h>
129 #include <netatalk/at.h>
130 #endif
131 #include <net/pfil.h>
132 
133 #ifdef INET6
134 #include <netinet/in.h>
135 #include <netinet6/in6_var.h>
136 #include <netinet6/nd6.h>
137 #endif
138 
139 #include "carp.h"
140 #if NCARP > 0
141 #include <netinet/ip_carp.h>
142 #endif
143 
144 #include <compat/sys/sockio.h>
145 #include <compat/sys/socket.h>
146 
147 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
148 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
149 
150 int	ifqmaxlen = IFQ_MAXLEN;
151 callout_t if_slowtimo_ch;
152 
153 int netisr;			/* scheduling bits for network */
154 
155 static int	if_rt_walktree(struct rtentry *, void *);
156 
157 static struct if_clone *if_clone_lookup(const char *, int *);
158 static int	if_clone_list(struct if_clonereq *);
159 
160 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
161 static int if_cloners_count;
162 
163 static uint64_t index_gen;
164 static kmutex_t index_gen_mtx;
165 
166 /* Packet filtering hook for interfaces. */
167 pfil_head_t *	if_pfil;
168 
169 static kauth_listener_t if_listener;
170 
171 static int doifioctl(struct socket *, u_long, void *, struct lwp *);
172 static int ifioctl_attach(struct ifnet *);
173 static void ifioctl_detach(struct ifnet *);
174 static void ifnet_lock_enter(struct ifnet_lock *);
175 static void ifnet_lock_exit(struct ifnet_lock *);
176 static void if_detach_queues(struct ifnet *, struct ifqueue *);
177 static void sysctl_sndq_setup(struct sysctllog **, const char *,
178     struct ifaltq *);
179 
180 #if defined(INET) || defined(INET6)
181 static void sysctl_net_ifq_setup(struct sysctllog **, int, const char *,
182 				 int, const char *, int, struct ifqueue *);
183 #endif
184 
185 static int
186 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
187     void *arg0, void *arg1, void *arg2, void *arg3)
188 {
189 	int result;
190 	enum kauth_network_req req;
191 
192 	result = KAUTH_RESULT_DEFER;
193 	req = (enum kauth_network_req)arg1;
194 
195 	if (action != KAUTH_NETWORK_INTERFACE)
196 		return result;
197 
198 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
199 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
200 		result = KAUTH_RESULT_ALLOW;
201 
202 	return result;
203 }
204 
205 /*
206  * Network interface utility routines.
207  *
208  * Routines with ifa_ifwith* names take sockaddr *'s as
209  * parameters.
210  */
211 void
212 ifinit(void)
213 {
214 #ifdef INET
215 	{extern struct ifqueue ipintrq;
216 	sysctl_net_ifq_setup(NULL, PF_INET, "inet", IPPROTO_IP, "ip",
217 			     IPCTL_IFQ, &ipintrq);}
218 #endif /* INET */
219 #ifdef INET6
220 	{extern struct ifqueue ip6intrq;
221 	sysctl_net_ifq_setup(NULL, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
222 			     IPV6CTL_IFQ, &ip6intrq);}
223 #endif /* INET6 */
224 
225 	callout_init(&if_slowtimo_ch, 0);
226 	if_slowtimo(NULL);
227 
228 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
229 	    if_listener_cb, NULL);
230 
231 	/* interfaces are available, inform socket code */
232 	ifioctl = doifioctl;
233 }
234 
235 /*
236  * XXX Initialization before configure().
237  * XXX hack to get pfil_add_hook working in autoconf.
238  */
239 void
240 ifinit1(void)
241 {
242 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
243 	if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
244 	KASSERT(if_pfil != NULL);
245 }
246 
247 struct ifnet *
248 if_alloc(u_char type)
249 {
250 	return malloc(sizeof(struct ifnet), M_DEVBUF, M_WAITOK|M_ZERO);
251 }
252 
253 void
254 if_free(struct ifnet *ifp)
255 {
256 	free(ifp, M_DEVBUF);
257 }
258 
259 void
260 if_initname(struct ifnet *ifp, const char *name, int unit)
261 {
262 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
263 	    "%s%d", name, unit);
264 }
265 
266 /*
267  * Null routines used while an interface is going away.  These routines
268  * just return an error.
269  */
270 
271 int
272 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
273     const struct sockaddr *so, struct rtentry *rt)
274 {
275 
276 	return ENXIO;
277 }
278 
279 void
280 if_nullinput(struct ifnet *ifp, struct mbuf *m)
281 {
282 
283 	/* Nothing. */
284 }
285 
286 void
287 if_nullstart(struct ifnet *ifp)
288 {
289 
290 	/* Nothing. */
291 }
292 
293 int
294 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
295 {
296 
297 	/* Wake ifioctl_detach(), who may wait for all threads to
298 	 * quit the critical section.
299 	 */
300 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
301 	return ENXIO;
302 }
303 
304 int
305 if_nullinit(struct ifnet *ifp)
306 {
307 
308 	return ENXIO;
309 }
310 
311 void
312 if_nullstop(struct ifnet *ifp, int disable)
313 {
314 
315 	/* Nothing. */
316 }
317 
318 void
319 if_nullwatchdog(struct ifnet *ifp)
320 {
321 
322 	/* Nothing. */
323 }
324 
325 void
326 if_nulldrain(struct ifnet *ifp)
327 {
328 
329 	/* Nothing. */
330 }
331 
332 static u_int if_index = 1;
333 struct ifnet_head ifnet;
334 size_t if_indexlim = 0;
335 struct ifaddr **ifnet_addrs = NULL;
336 struct ifnet **ifindex2ifnet = NULL;
337 struct ifnet *lo0ifp;
338 
339 void
340 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
341 {
342 	struct ifaddr *ifa;
343 	struct sockaddr_dl *sdl;
344 
345 	ifp->if_addrlen = addrlen;
346 	if_alloc_sadl(ifp);
347 	ifa = ifp->if_dl;
348 	sdl = satosdl(ifa->ifa_addr);
349 
350 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
351 	if (factory) {
352 		ifp->if_hwdl = ifp->if_dl;
353 		IFAREF(ifp->if_hwdl);
354 	}
355 	/* TBD routing socket */
356 }
357 
358 struct ifaddr *
359 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
360 {
361 	unsigned socksize, ifasize;
362 	int addrlen, namelen;
363 	struct sockaddr_dl *mask, *sdl;
364 	struct ifaddr *ifa;
365 
366 	namelen = strlen(ifp->if_xname);
367 	addrlen = ifp->if_addrlen;
368 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
369 	ifasize = sizeof(*ifa) + 2 * socksize;
370 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
371 
372 	sdl = (struct sockaddr_dl *)(ifa + 1);
373 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
374 
375 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
376 	    ifp->if_xname, namelen, NULL, addrlen);
377 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
378 	memset(&mask->sdl_data[0], 0xff, namelen);
379 	ifa->ifa_rtrequest = link_rtrequest;
380 	ifa->ifa_addr = (struct sockaddr *)sdl;
381 	ifa->ifa_netmask = (struct sockaddr *)mask;
382 
383 	*sdlp = sdl;
384 
385 	return ifa;
386 }
387 
388 static void
389 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
390 {
391 	const struct sockaddr_dl *sdl;
392 	ifnet_addrs[ifp->if_index] = ifa;
393 	IFAREF(ifa);
394 	ifp->if_dl = ifa;
395 	IFAREF(ifa);
396 	sdl = satosdl(ifa->ifa_addr);
397 	ifp->if_sadl = sdl;
398 }
399 
400 /*
401  * Allocate the link level name for the specified interface.  This
402  * is an attachment helper.  It must be called after ifp->if_addrlen
403  * is initialized, which may not be the case when if_attach() is
404  * called.
405  */
406 void
407 if_alloc_sadl(struct ifnet *ifp)
408 {
409 	struct ifaddr *ifa;
410 	const struct sockaddr_dl *sdl;
411 
412 	/*
413 	 * If the interface already has a link name, release it
414 	 * now.  This is useful for interfaces that can change
415 	 * link types, and thus switch link names often.
416 	 */
417 	if (ifp->if_sadl != NULL)
418 		if_free_sadl(ifp);
419 
420 	ifa = if_dl_create(ifp, &sdl);
421 
422 	ifa_insert(ifp, ifa);
423 	if_sadl_setrefs(ifp, ifa);
424 }
425 
426 static void
427 if_deactivate_sadl(struct ifnet *ifp)
428 {
429 	struct ifaddr *ifa;
430 
431 	KASSERT(ifp->if_dl != NULL);
432 
433 	ifa = ifp->if_dl;
434 
435 	ifp->if_sadl = NULL;
436 
437 	ifnet_addrs[ifp->if_index] = NULL;
438 	IFAFREE(ifa);
439 	ifp->if_dl = NULL;
440 	IFAFREE(ifa);
441 }
442 
443 void
444 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
445     const struct sockaddr_dl *sdl)
446 {
447 	int s;
448 
449 	s = splnet();
450 
451 	if_deactivate_sadl(ifp);
452 
453 	if_sadl_setrefs(ifp, ifa);
454 	IFADDR_FOREACH(ifa, ifp)
455 		rtinit(ifa, RTM_LLINFO_UPD, 0);
456 	splx(s);
457 }
458 
459 /*
460  * Free the link level name for the specified interface.  This is
461  * a detach helper.  This is called from if_detach() or from
462  * link layer type specific detach functions.
463  */
464 void
465 if_free_sadl(struct ifnet *ifp)
466 {
467 	struct ifaddr *ifa;
468 	int s;
469 
470 	ifa = ifnet_addrs[ifp->if_index];
471 	if (ifa == NULL) {
472 		KASSERT(ifp->if_sadl == NULL);
473 		KASSERT(ifp->if_dl == NULL);
474 		return;
475 	}
476 
477 	KASSERT(ifp->if_sadl != NULL);
478 	KASSERT(ifp->if_dl != NULL);
479 
480 	s = splnet();
481 	rtinit(ifa, RTM_DELETE, 0);
482 	ifa_remove(ifp, ifa);
483 	if_deactivate_sadl(ifp);
484 	if (ifp->if_hwdl == ifa) {
485 		IFAFREE(ifa);
486 		ifp->if_hwdl = NULL;
487 	}
488 	splx(s);
489 }
490 
491 /*
492  * Attach an interface to the
493  * list of "active" interfaces.
494  */
495 void
496 if_attach(struct ifnet *ifp)
497 {
498 	int indexlim = 0;
499 
500 	if (if_indexlim == 0) {
501 		TAILQ_INIT(&ifnet);
502 		if_indexlim = 8;
503 	}
504 	TAILQ_INIT(&ifp->if_addrlist);
505 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
506 
507 	if (ifioctl_attach(ifp) != 0)
508 		panic("%s: ifioctl_attach() failed", __func__);
509 
510 	mutex_enter(&index_gen_mtx);
511 	ifp->if_index_gen = index_gen++;
512 	mutex_exit(&index_gen_mtx);
513 
514 	ifp->if_index = if_index;
515 	if (ifindex2ifnet == NULL)
516 		if_index++;
517 	else
518 		while (ifp->if_index < if_indexlim &&
519 		    ifindex2ifnet[ifp->if_index] != NULL) {
520 			++if_index;
521 			if (if_index == 0)
522 				if_index = 1;
523 			/*
524 			 * If we hit USHRT_MAX, we skip back to 0 since
525 			 * there are a number of places where the value
526 			 * of if_index or if_index itself is compared
527 			 * to or stored in an unsigned short.  By
528 			 * jumping back, we won't botch those assignments
529 			 * or comparisons.
530 			 */
531 			else if (if_index == USHRT_MAX) {
532 				/*
533 				 * However, if we have to jump back to
534 				 * zero *twice* without finding an empty
535 				 * slot in ifindex2ifnet[], then there
536 				 * there are too many (>65535) interfaces.
537 				 */
538 				if (indexlim++)
539 					panic("too many interfaces");
540 				else
541 					if_index = 1;
542 			}
543 			ifp->if_index = if_index;
544 		}
545 
546 	/*
547 	 * We have some arrays that should be indexed by if_index.
548 	 * since if_index will grow dynamically, they should grow too.
549 	 *	struct ifadd **ifnet_addrs
550 	 *	struct ifnet **ifindex2ifnet
551 	 */
552 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
553 	    ifp->if_index >= if_indexlim) {
554 		size_t m, n, oldlim;
555 		void *q;
556 
557 		oldlim = if_indexlim;
558 		while (ifp->if_index >= if_indexlim)
559 			if_indexlim <<= 1;
560 
561 		/* grow ifnet_addrs */
562 		m = oldlim * sizeof(struct ifaddr *);
563 		n = if_indexlim * sizeof(struct ifaddr *);
564 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
565 		if (ifnet_addrs != NULL) {
566 			memcpy(q, ifnet_addrs, m);
567 			free(ifnet_addrs, M_IFADDR);
568 		}
569 		ifnet_addrs = (struct ifaddr **)q;
570 
571 		/* grow ifindex2ifnet */
572 		m = oldlim * sizeof(struct ifnet *);
573 		n = if_indexlim * sizeof(struct ifnet *);
574 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
575 		if (ifindex2ifnet != NULL) {
576 			memcpy(q, ifindex2ifnet, m);
577 			free(ifindex2ifnet, M_IFADDR);
578 		}
579 		ifindex2ifnet = (struct ifnet **)q;
580 	}
581 
582 	ifindex2ifnet[ifp->if_index] = ifp;
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 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
593 
594 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
595 
596 	ifp->if_link_state = LINK_STATE_UNKNOWN;
597 
598 	ifp->if_capenable = 0;
599 	ifp->if_csum_flags_tx = 0;
600 	ifp->if_csum_flags_rx = 0;
601 
602 #ifdef ALTQ
603 	ifp->if_snd.altq_type = 0;
604 	ifp->if_snd.altq_disc = NULL;
605 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
606 	ifp->if_snd.altq_tbr  = NULL;
607 	ifp->if_snd.altq_ifp  = ifp;
608 #endif
609 
610 	ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
611 	(void)pfil_run_hooks(if_pfil,
612 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
613 
614 	if (!STAILQ_EMPTY(&domains))
615 		if_attachdomain1(ifp);
616 
617 	/* Announce the interface. */
618 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
619 }
620 
621 void
622 if_attachdomain(void)
623 {
624 	struct ifnet *ifp;
625 	int s;
626 
627 	s = splnet();
628 	IFNET_FOREACH(ifp)
629 		if_attachdomain1(ifp);
630 	splx(s);
631 }
632 
633 void
634 if_attachdomain1(struct ifnet *ifp)
635 {
636 	struct domain *dp;
637 	int s;
638 
639 	s = splnet();
640 
641 	/* address family dependent data region */
642 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
643 	DOMAIN_FOREACH(dp) {
644 		if (dp->dom_ifattach != NULL)
645 			ifp->if_afdata[dp->dom_family] =
646 			    (*dp->dom_ifattach)(ifp);
647 	}
648 
649 	splx(s);
650 }
651 
652 /*
653  * Deactivate an interface.  This points all of the procedure
654  * handles at error stubs.  May be called from interrupt context.
655  */
656 void
657 if_deactivate(struct ifnet *ifp)
658 {
659 	int s;
660 
661 	s = splnet();
662 
663 	ifp->if_output	 = if_nulloutput;
664 	ifp->if_input	 = if_nullinput;
665 	ifp->if_start	 = if_nullstart;
666 	ifp->if_ioctl	 = if_nullioctl;
667 	ifp->if_init	 = if_nullinit;
668 	ifp->if_stop	 = if_nullstop;
669 	ifp->if_watchdog = if_nullwatchdog;
670 	ifp->if_drain	 = if_nulldrain;
671 
672 	/* No more packets may be enqueued. */
673 	ifp->if_snd.ifq_maxlen = 0;
674 
675 	splx(s);
676 }
677 
678 void
679 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
680 {
681 	struct ifaddr *ifa, *nifa;
682 
683 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
684 		nifa = IFADDR_NEXT(ifa);
685 		if (ifa->ifa_addr->sa_family != family)
686 			continue;
687 		(*purgeaddr)(ifa);
688 	}
689 }
690 
691 /*
692  * Detach an interface from the list of "active" interfaces,
693  * freeing any resources as we go along.
694  *
695  * NOTE: This routine must be called with a valid thread context,
696  * as it may block.
697  */
698 void
699 if_detach(struct ifnet *ifp)
700 {
701 	struct socket so;
702 	struct ifaddr *ifa;
703 #ifdef IFAREF_DEBUG
704 	struct ifaddr *last_ifa = NULL;
705 #endif
706 	struct domain *dp;
707 	const struct protosw *pr;
708 	int s, i, family, purged;
709 
710 	/*
711 	 * XXX It's kind of lame that we have to have the
712 	 * XXX socket structure...
713 	 */
714 	memset(&so, 0, sizeof(so));
715 
716 	s = splnet();
717 
718 	/*
719 	 * Do an if_down() to give protocols a chance to do something.
720 	 */
721 	if_down(ifp);
722 
723 #ifdef ALTQ
724 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
725 		altq_disable(&ifp->if_snd);
726 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
727 		altq_detach(&ifp->if_snd);
728 #endif
729 
730 	sysctl_teardown(&ifp->if_sysctl_log);
731 
732 #if NCARP > 0
733 	/* Remove the interface from any carp group it is a part of.  */
734 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
735 		carp_ifdetach(ifp);
736 #endif
737 
738 	/*
739 	 * Rip all the addresses off the interface.  This should make
740 	 * all of the routes go away.
741 	 *
742 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
743 	 * from the list, including our "cursor", ifa.  For safety,
744 	 * and to honor the TAILQ abstraction, I just restart the
745 	 * loop after each removal.  Note that the loop will exit
746 	 * when all of the remaining ifaddrs belong to the AF_LINK
747 	 * family.  I am counting on the historical fact that at
748 	 * least one pr_usrreq in each address domain removes at
749 	 * least one ifaddr.
750 	 */
751 again:
752 	IFADDR_FOREACH(ifa, ifp) {
753 		family = ifa->ifa_addr->sa_family;
754 #ifdef IFAREF_DEBUG
755 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
756 		    ifa, family, ifa->ifa_refcnt);
757 		if (last_ifa != NULL && ifa == last_ifa)
758 			panic("if_detach: loop detected");
759 		last_ifa = ifa;
760 #endif
761 		if (family == AF_LINK)
762 			continue;
763 		dp = pffinddomain(family);
764 #ifdef DIAGNOSTIC
765 		if (dp == NULL)
766 			panic("if_detach: no domain for AF %d",
767 			    family);
768 #endif
769 		/*
770 		 * XXX These PURGEIF calls are redundant with the
771 		 * purge-all-families calls below, but are left in for
772 		 * now both to make a smaller change, and to avoid
773 		 * unplanned interactions with clearing of
774 		 * ifp->if_addrlist.
775 		 */
776 		purged = 0;
777 		for (pr = dp->dom_protosw;
778 		     pr < dp->dom_protoswNPROTOSW; pr++) {
779 			so.so_proto = pr;
780 			if (pr->pr_usrreq != NULL) {
781 				(void) (*pr->pr_usrreq)(&so,
782 				    PRU_PURGEIF, NULL, NULL,
783 				    (struct mbuf *) ifp, curlwp);
784 				purged = 1;
785 			}
786 		}
787 		if (purged == 0) {
788 			/*
789 			 * XXX What's really the best thing to do
790 			 * XXX here?  --thorpej@NetBSD.org
791 			 */
792 			printf("if_detach: WARNING: AF %d not purged\n",
793 			    family);
794 			ifa_remove(ifp, ifa);
795 		}
796 		goto again;
797 	}
798 
799 	if_free_sadl(ifp);
800 
801 	/* Walk the routing table looking for stragglers. */
802 	for (i = 0; i <= AF_MAX; i++) {
803 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
804 			continue;
805 	}
806 
807 	DOMAIN_FOREACH(dp) {
808 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
809 		{
810 			void *p = ifp->if_afdata[dp->dom_family];
811 			if (p) {
812 				ifp->if_afdata[dp->dom_family] = NULL;
813 				(*dp->dom_ifdetach)(ifp, p);
814 			}
815 		}
816 
817 		/*
818 		 * One would expect multicast memberships (INET and
819 		 * INET6) on UDP sockets to be purged by the PURGEIF
820 		 * calls above, but if all addresses were removed from
821 		 * the interface prior to destruction, the calls will
822 		 * not be made (e.g. ppp, for which pppd(8) generally
823 		 * removes addresses before destroying the interface).
824 		 * Because there is no invariant that multicast
825 		 * memberships only exist for interfaces with IPv4
826 		 * addresses, we must call PURGEIF regardless of
827 		 * addresses.  (Protocols which might store ifnet
828 		 * pointers are marked with PR_PURGEIF.)
829 		 */
830 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
831 			so.so_proto = pr;
832 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
833 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
834 				    NULL, (struct mbuf *)ifp, curlwp);
835 		}
836 	}
837 
838 	(void)pfil_run_hooks(if_pfil,
839 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
840 	(void)pfil_head_destroy(ifp->if_pfil);
841 
842 	/* Announce that the interface is gone. */
843 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
844 
845 	ifindex2ifnet[ifp->if_index] = NULL;
846 
847 	TAILQ_REMOVE(&ifnet, ifp, if_list);
848 
849 	ifioctl_detach(ifp);
850 
851 	/*
852 	 * remove packets that came from ifp, from software interrupt queues.
853 	 */
854 	DOMAIN_FOREACH(dp) {
855 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
856 			struct ifqueue *iq = dp->dom_ifqueues[i];
857 			if (iq == NULL)
858 				break;
859 			dp->dom_ifqueues[i] = NULL;
860 			if_detach_queues(ifp, iq);
861 		}
862 	}
863 
864 	splx(s);
865 }
866 
867 static void
868 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
869 {
870 	struct mbuf *m, *prev, *next;
871 
872 	prev = NULL;
873 	for (m = q->ifq_head; m != NULL; m = next) {
874 		next = m->m_nextpkt;
875 #ifdef DIAGNOSTIC
876 		if ((m->m_flags & M_PKTHDR) == 0) {
877 			prev = m;
878 			continue;
879 		}
880 #endif
881 		if (m->m_pkthdr.rcvif != ifp) {
882 			prev = m;
883 			continue;
884 		}
885 
886 		if (prev != NULL)
887 			prev->m_nextpkt = m->m_nextpkt;
888 		else
889 			q->ifq_head = m->m_nextpkt;
890 		if (q->ifq_tail == m)
891 			q->ifq_tail = prev;
892 		q->ifq_len--;
893 
894 		m->m_nextpkt = NULL;
895 		m_freem(m);
896 		IF_DROP(q);
897 	}
898 }
899 
900 /*
901  * Callback for a radix tree walk to delete all references to an
902  * ifnet.
903  */
904 static int
905 if_rt_walktree(struct rtentry *rt, void *v)
906 {
907 	struct ifnet *ifp = (struct ifnet *)v;
908 	int error;
909 
910 	if (rt->rt_ifp != ifp)
911 		return 0;
912 
913 	/* Delete the entry. */
914 	++rt->rt_refcnt;
915 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
916 	    rt_mask(rt), rt->rt_flags, NULL);
917 	KASSERT((rt->rt_flags & RTF_UP) == 0);
918 	rt->rt_ifp = NULL;
919 	RTFREE(rt);
920 	if (error != 0)
921 		printf("%s: warning: unable to delete rtentry @ %p, "
922 		    "error = %d\n", ifp->if_xname, rt, error);
923 	return ERESTART;
924 }
925 
926 /*
927  * Create a clone network interface.
928  */
929 int
930 if_clone_create(const char *name)
931 {
932 	struct if_clone *ifc;
933 	int unit;
934 
935 	ifc = if_clone_lookup(name, &unit);
936 	if (ifc == NULL)
937 		return EINVAL;
938 
939 	if (ifunit(name) != NULL)
940 		return EEXIST;
941 
942 	return (*ifc->ifc_create)(ifc, unit);
943 }
944 
945 /*
946  * Destroy a clone network interface.
947  */
948 int
949 if_clone_destroy(const char *name)
950 {
951 	struct if_clone *ifc;
952 	struct ifnet *ifp;
953 
954 	ifc = if_clone_lookup(name, NULL);
955 	if (ifc == NULL)
956 		return EINVAL;
957 
958 	ifp = ifunit(name);
959 	if (ifp == NULL)
960 		return ENXIO;
961 
962 	if (ifc->ifc_destroy == NULL)
963 		return EOPNOTSUPP;
964 
965 	return (*ifc->ifc_destroy)(ifp);
966 }
967 
968 /*
969  * Look up a network interface cloner.
970  */
971 static struct if_clone *
972 if_clone_lookup(const char *name, int *unitp)
973 {
974 	struct if_clone *ifc;
975 	const char *cp;
976 	char *dp, ifname[IFNAMSIZ + 3];
977 	int unit;
978 
979 	strcpy(ifname, "if_");
980 	/* separate interface name from unit */
981 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
982 	    *cp && (*cp < '0' || *cp > '9');)
983 		*dp++ = *cp++;
984 
985 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
986 		return NULL;	/* No name or unit number */
987 	*dp++ = '\0';
988 
989 again:
990 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
991 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
992 			break;
993 	}
994 
995 	if (ifc == NULL) {
996 		if (*ifname == '\0' ||
997 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
998 			return NULL;
999 		*ifname = '\0';
1000 		goto again;
1001 	}
1002 
1003 	unit = 0;
1004 	while (cp - name < IFNAMSIZ && *cp) {
1005 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1006 			/* Bogus unit number. */
1007 			return NULL;
1008 		}
1009 		unit = (unit * 10) + (*cp++ - '0');
1010 	}
1011 
1012 	if (unitp != NULL)
1013 		*unitp = unit;
1014 	return ifc;
1015 }
1016 
1017 /*
1018  * Register a network interface cloner.
1019  */
1020 void
1021 if_clone_attach(struct if_clone *ifc)
1022 {
1023 
1024 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1025 	if_cloners_count++;
1026 }
1027 
1028 /*
1029  * Unregister a network interface cloner.
1030  */
1031 void
1032 if_clone_detach(struct if_clone *ifc)
1033 {
1034 
1035 	LIST_REMOVE(ifc, ifc_list);
1036 	if_cloners_count--;
1037 }
1038 
1039 /*
1040  * Provide list of interface cloners to userspace.
1041  */
1042 static int
1043 if_clone_list(struct if_clonereq *ifcr)
1044 {
1045 	char outbuf[IFNAMSIZ], *dst;
1046 	struct if_clone *ifc;
1047 	int count, error = 0;
1048 
1049 	ifcr->ifcr_total = if_cloners_count;
1050 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1051 		/* Just asking how many there are. */
1052 		return 0;
1053 	}
1054 
1055 	if (ifcr->ifcr_count < 0)
1056 		return EINVAL;
1057 
1058 	count = (if_cloners_count < ifcr->ifcr_count) ?
1059 	    if_cloners_count : ifcr->ifcr_count;
1060 
1061 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1062 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1063 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1064 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1065 			return ENAMETOOLONG;
1066 		error = copyout(outbuf, dst, sizeof(outbuf));
1067 		if (error != 0)
1068 			break;
1069 	}
1070 
1071 	return error;
1072 }
1073 
1074 void
1075 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1076 {
1077 	ifa->ifa_ifp = ifp;
1078 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1079 	IFAREF(ifa);
1080 }
1081 
1082 void
1083 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1084 {
1085 	KASSERT(ifa->ifa_ifp == ifp);
1086 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1087 	IFAFREE(ifa);
1088 }
1089 
1090 static inline int
1091 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1092 {
1093 	return sockaddr_cmp(sa1, sa2) == 0;
1094 }
1095 
1096 /*
1097  * Locate an interface based on a complete address.
1098  */
1099 /*ARGSUSED*/
1100 struct ifaddr *
1101 ifa_ifwithaddr(const struct sockaddr *addr)
1102 {
1103 	struct ifnet *ifp;
1104 	struct ifaddr *ifa;
1105 
1106 	IFNET_FOREACH(ifp) {
1107 		if (ifp->if_output == if_nulloutput)
1108 			continue;
1109 		IFADDR_FOREACH(ifa, ifp) {
1110 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1111 				continue;
1112 			if (equal(addr, ifa->ifa_addr))
1113 				return ifa;
1114 			if ((ifp->if_flags & IFF_BROADCAST) &&
1115 			    ifa->ifa_broadaddr &&
1116 			    /* IP6 doesn't have broadcast */
1117 			    ifa->ifa_broadaddr->sa_len != 0 &&
1118 			    equal(ifa->ifa_broadaddr, addr))
1119 				return ifa;
1120 		}
1121 	}
1122 	return NULL;
1123 }
1124 
1125 /*
1126  * Locate the point to point interface with a given destination address.
1127  */
1128 /*ARGSUSED*/
1129 struct ifaddr *
1130 ifa_ifwithdstaddr(const struct sockaddr *addr)
1131 {
1132 	struct ifnet *ifp;
1133 	struct ifaddr *ifa;
1134 
1135 	IFNET_FOREACH(ifp) {
1136 		if (ifp->if_output == if_nulloutput)
1137 			continue;
1138 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1139 			continue;
1140 		IFADDR_FOREACH(ifa, ifp) {
1141 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1142 			    ifa->ifa_dstaddr == NULL)
1143 				continue;
1144 			if (equal(addr, ifa->ifa_dstaddr))
1145 				return ifa;
1146 		}
1147 	}
1148 	return NULL;
1149 }
1150 
1151 /*
1152  * Find an interface on a specific network.  If many, choice
1153  * is most specific found.
1154  */
1155 struct ifaddr *
1156 ifa_ifwithnet(const struct sockaddr *addr)
1157 {
1158 	struct ifnet *ifp;
1159 	struct ifaddr *ifa;
1160 	const struct sockaddr_dl *sdl;
1161 	struct ifaddr *ifa_maybe = 0;
1162 	u_int af = addr->sa_family;
1163 	const char *addr_data = addr->sa_data, *cplim;
1164 
1165 	if (af == AF_LINK) {
1166 		sdl = satocsdl(addr);
1167 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1168 		    ifindex2ifnet[sdl->sdl_index] &&
1169 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1170 			return ifnet_addrs[sdl->sdl_index];
1171 	}
1172 #ifdef NETATALK
1173 	if (af == AF_APPLETALK) {
1174 		const struct sockaddr_at *sat, *sat2;
1175 		sat = (const struct sockaddr_at *)addr;
1176 		IFNET_FOREACH(ifp) {
1177 			if (ifp->if_output == if_nulloutput)
1178 				continue;
1179 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1180 			if (ifa == NULL)
1181 				continue;
1182 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1183 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1184 				return ifa; /* exact match */
1185 			if (ifa_maybe == NULL) {
1186 				/* else keep the if with the right range */
1187 				ifa_maybe = ifa;
1188 			}
1189 		}
1190 		return ifa_maybe;
1191 	}
1192 #endif
1193 	IFNET_FOREACH(ifp) {
1194 		if (ifp->if_output == if_nulloutput)
1195 			continue;
1196 		IFADDR_FOREACH(ifa, ifp) {
1197 			const char *cp, *cp2, *cp3;
1198 
1199 			if (ifa->ifa_addr->sa_family != af ||
1200 			    ifa->ifa_netmask == NULL)
1201  next:				continue;
1202 			cp = addr_data;
1203 			cp2 = ifa->ifa_addr->sa_data;
1204 			cp3 = ifa->ifa_netmask->sa_data;
1205 			cplim = (const char *)ifa->ifa_netmask +
1206 			    ifa->ifa_netmask->sa_len;
1207 			while (cp3 < cplim) {
1208 				if ((*cp++ ^ *cp2++) & *cp3++) {
1209 					/* want to continue for() loop */
1210 					goto next;
1211 				}
1212 			}
1213 			if (ifa_maybe == NULL ||
1214 			    rn_refines((void *)ifa->ifa_netmask,
1215 			    (void *)ifa_maybe->ifa_netmask))
1216 				ifa_maybe = ifa;
1217 		}
1218 	}
1219 	return ifa_maybe;
1220 }
1221 
1222 /*
1223  * Find the interface of the addresss.
1224  */
1225 struct ifaddr *
1226 ifa_ifwithladdr(const struct sockaddr *addr)
1227 {
1228 	struct ifaddr *ia;
1229 
1230 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1231 	    (ia = ifa_ifwithnet(addr)))
1232 		return ia;
1233 	return NULL;
1234 }
1235 
1236 /*
1237  * Find an interface using a specific address family
1238  */
1239 struct ifaddr *
1240 ifa_ifwithaf(int af)
1241 {
1242 	struct ifnet *ifp;
1243 	struct ifaddr *ifa;
1244 
1245 	IFNET_FOREACH(ifp) {
1246 		if (ifp->if_output == if_nulloutput)
1247 			continue;
1248 		IFADDR_FOREACH(ifa, ifp) {
1249 			if (ifa->ifa_addr->sa_family == af)
1250 				return ifa;
1251 		}
1252 	}
1253 	return NULL;
1254 }
1255 
1256 /*
1257  * Find an interface address specific to an interface best matching
1258  * a given address.
1259  */
1260 struct ifaddr *
1261 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1262 {
1263 	struct ifaddr *ifa;
1264 	const char *cp, *cp2, *cp3;
1265 	const char *cplim;
1266 	struct ifaddr *ifa_maybe = 0;
1267 	u_int af = addr->sa_family;
1268 
1269 	if (ifp->if_output == if_nulloutput)
1270 		return NULL;
1271 
1272 	if (af >= AF_MAX)
1273 		return NULL;
1274 
1275 	IFADDR_FOREACH(ifa, ifp) {
1276 		if (ifa->ifa_addr->sa_family != af)
1277 			continue;
1278 		ifa_maybe = ifa;
1279 		if (ifa->ifa_netmask == NULL) {
1280 			if (equal(addr, ifa->ifa_addr) ||
1281 			    (ifa->ifa_dstaddr &&
1282 			     equal(addr, ifa->ifa_dstaddr)))
1283 				return ifa;
1284 			continue;
1285 		}
1286 		cp = addr->sa_data;
1287 		cp2 = ifa->ifa_addr->sa_data;
1288 		cp3 = ifa->ifa_netmask->sa_data;
1289 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1290 		for (; cp3 < cplim; cp3++) {
1291 			if ((*cp++ ^ *cp2++) & *cp3)
1292 				break;
1293 		}
1294 		if (cp3 == cplim)
1295 			return ifa;
1296 	}
1297 	return ifa_maybe;
1298 }
1299 
1300 /*
1301  * Default action when installing a route with a Link Level gateway.
1302  * Lookup an appropriate real ifa to point to.
1303  * This should be moved to /sys/net/link.c eventually.
1304  */
1305 void
1306 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1307 {
1308 	struct ifaddr *ifa;
1309 	const struct sockaddr *dst;
1310 	struct ifnet *ifp;
1311 
1312 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1313 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1314 		return;
1315 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1316 		rt_replace_ifa(rt, ifa);
1317 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1318 			ifa->ifa_rtrequest(cmd, rt, info);
1319 	}
1320 }
1321 
1322 /*
1323  * Handle a change in the interface link state.
1324  * XXX: We should listen to the routing socket in-kernel rather
1325  * than calling in6_if_link_* functions directly from here.
1326  */
1327 void
1328 if_link_state_change(struct ifnet *ifp, int link_state)
1329 {
1330 	int s;
1331 #if defined(DEBUG) || defined(INET6)
1332 	int old_link_state;
1333 #endif
1334 
1335 	s = splnet();
1336 	if (ifp->if_link_state == link_state) {
1337 		splx(s);
1338 		return;
1339 	}
1340 
1341 #if defined(DEBUG) || defined(INET6)
1342 	old_link_state = ifp->if_link_state;
1343 #endif
1344 	ifp->if_link_state = link_state;
1345 #ifdef DEBUG
1346 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
1347 		link_state == LINK_STATE_UP ? "UP" :
1348 		link_state == LINK_STATE_DOWN ? "DOWN" :
1349 		"UNKNOWN",
1350 		 old_link_state == LINK_STATE_UP ? "UP" :
1351 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
1352 		"UNKNOWN");
1353 #endif
1354 
1355 #ifdef INET6
1356 	/*
1357 	 * When going from UNKNOWN to UP, we need to mark existing
1358 	 * IPv6 addresses as tentative and restart DAD as we may have
1359 	 * erroneously not found a duplicate.
1360 	 *
1361 	 * This needs to happen before rt_ifmsg to avoid a race where
1362 	 * listeners would have an address and expect it to work right
1363 	 * away.
1364 	 */
1365 	if (in6_present && link_state == LINK_STATE_UP &&
1366 	    old_link_state == LINK_STATE_UNKNOWN)
1367 		in6_if_link_down(ifp);
1368 #endif
1369 
1370 	/* Notify that the link state has changed. */
1371 	rt_ifmsg(ifp);
1372 
1373 #if NCARP > 0
1374 	if (ifp->if_carp)
1375 		carp_carpdev_state(ifp);
1376 #endif
1377 
1378 #ifdef INET6
1379 	if (in6_present) {
1380 		if (link_state == LINK_STATE_DOWN)
1381 			in6_if_link_down(ifp);
1382 		else if (link_state == LINK_STATE_UP)
1383 			in6_if_link_up(ifp);
1384 	}
1385 #endif
1386 
1387 	splx(s);
1388 }
1389 
1390 /*
1391  * Mark an interface down and notify protocols of
1392  * the transition.
1393  * NOTE: must be called at splsoftnet or equivalent.
1394  */
1395 void
1396 if_down(struct ifnet *ifp)
1397 {
1398 	struct ifaddr *ifa;
1399 
1400 	ifp->if_flags &= ~IFF_UP;
1401 	nanotime(&ifp->if_lastchange);
1402 	IFADDR_FOREACH(ifa, ifp)
1403 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1404 	IFQ_PURGE(&ifp->if_snd);
1405 #if NCARP > 0
1406 	if (ifp->if_carp)
1407 		carp_carpdev_state(ifp);
1408 #endif
1409 	rt_ifmsg(ifp);
1410 #ifdef INET6
1411 	if (in6_present)
1412 		in6_if_down(ifp);
1413 #endif
1414 }
1415 
1416 /*
1417  * Mark an interface up and notify protocols of
1418  * the transition.
1419  * NOTE: must be called at splsoftnet or equivalent.
1420  */
1421 void
1422 if_up(struct ifnet *ifp)
1423 {
1424 #ifdef notyet
1425 	struct ifaddr *ifa;
1426 #endif
1427 
1428 	ifp->if_flags |= IFF_UP;
1429 	nanotime(&ifp->if_lastchange);
1430 #ifdef notyet
1431 	/* this has no effect on IP, and will kill all ISO connections XXX */
1432 	IFADDR_FOREACH(ifa, ifp)
1433 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1434 #endif
1435 #if NCARP > 0
1436 	if (ifp->if_carp)
1437 		carp_carpdev_state(ifp);
1438 #endif
1439 	rt_ifmsg(ifp);
1440 #ifdef INET6
1441 	if (in6_present)
1442 		in6_if_up(ifp);
1443 #endif
1444 }
1445 
1446 /*
1447  * Handle interface watchdog timer routines.  Called
1448  * from softclock, we decrement timers (if set) and
1449  * call the appropriate interface routine on expiration.
1450  */
1451 void
1452 if_slowtimo(void *arg)
1453 {
1454 	struct ifnet *ifp;
1455 	int s = splnet();
1456 
1457 	IFNET_FOREACH(ifp) {
1458 		if (ifp->if_timer == 0 || --ifp->if_timer)
1459 			continue;
1460 		if (ifp->if_watchdog != NULL)
1461 			(*ifp->if_watchdog)(ifp);
1462 	}
1463 	splx(s);
1464 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1465 }
1466 
1467 /*
1468  * Set/clear promiscuous mode on interface ifp based on the truth value
1469  * of pswitch.  The calls are reference counted so that only the first
1470  * "on" request actually has an effect, as does the final "off" request.
1471  * Results are undefined if the "off" and "on" requests are not matched.
1472  */
1473 int
1474 ifpromisc(struct ifnet *ifp, int pswitch)
1475 {
1476 	int pcount, ret;
1477 	short nflags;
1478 
1479 	pcount = ifp->if_pcount;
1480 	if (pswitch) {
1481 		/*
1482 		 * Allow the device to be "placed" into promiscuous
1483 		 * mode even if it is not configured up.  It will
1484 		 * consult IFF_PROMISC when it is brought up.
1485 		 */
1486 		if (ifp->if_pcount++ != 0)
1487 			return 0;
1488 		nflags = ifp->if_flags | IFF_PROMISC;
1489 	} else {
1490 		if (--ifp->if_pcount > 0)
1491 			return 0;
1492 		nflags = ifp->if_flags & ~IFF_PROMISC;
1493 	}
1494 	ret = if_flags_set(ifp, nflags);
1495 	/* Restore interface state if not successful. */
1496 	if (ret != 0) {
1497 		ifp->if_pcount = pcount;
1498 	}
1499 	return ret;
1500 }
1501 
1502 /*
1503  * Map interface name to
1504  * interface structure pointer.
1505  */
1506 struct ifnet *
1507 ifunit(const char *name)
1508 {
1509 	struct ifnet *ifp;
1510 	const char *cp = name;
1511 	u_int unit = 0;
1512 	u_int i;
1513 
1514 	/*
1515 	 * If the entire name is a number, treat it as an ifindex.
1516 	 */
1517 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1518 		unit = unit * 10 + (*cp - '0');
1519 	}
1520 
1521 	/*
1522 	 * If the number took all of the name, then it's a valid ifindex.
1523 	 */
1524 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1525 		if (unit >= if_indexlim)
1526 			return NULL;
1527 		ifp = ifindex2ifnet[unit];
1528 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1529 			return NULL;
1530 		return ifp;
1531 	}
1532 
1533 	IFNET_FOREACH(ifp) {
1534 		if (ifp->if_output == if_nulloutput)
1535 			continue;
1536 	 	if (strcmp(ifp->if_xname, name) == 0)
1537 			return ifp;
1538 	}
1539 	return NULL;
1540 }
1541 
1542 ifnet_t *
1543 if_byindex(u_int idx)
1544 {
1545 
1546 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1547 }
1548 
1549 /* common */
1550 int
1551 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1552 {
1553 	int s;
1554 	struct ifreq *ifr;
1555 	struct ifcapreq *ifcr;
1556 	struct ifdatareq *ifdr;
1557 
1558 	switch (cmd) {
1559 	case SIOCSIFCAP:
1560 		ifcr = data;
1561 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1562 			return EINVAL;
1563 
1564 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1565 			return 0;
1566 
1567 		ifp->if_capenable = ifcr->ifcr_capenable;
1568 
1569 		/* Pre-compute the checksum flags mask. */
1570 		ifp->if_csum_flags_tx = 0;
1571 		ifp->if_csum_flags_rx = 0;
1572 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1573 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1574 		}
1575 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1576 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1577 		}
1578 
1579 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1580 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1581 		}
1582 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1583 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1584 		}
1585 
1586 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1587 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1588 		}
1589 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1590 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1591 		}
1592 
1593 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1594 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1595 		}
1596 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1597 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1598 		}
1599 
1600 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1601 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1602 		}
1603 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1604 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1605 		}
1606 		if (ifp->if_flags & IFF_UP)
1607 			return ENETRESET;
1608 		return 0;
1609 	case SIOCSIFFLAGS:
1610 		ifr = data;
1611 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1612 			s = splnet();
1613 			if_down(ifp);
1614 			splx(s);
1615 		}
1616 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1617 			s = splnet();
1618 			if_up(ifp);
1619 			splx(s);
1620 		}
1621 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1622 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1623 		break;
1624 	case SIOCGIFFLAGS:
1625 		ifr = data;
1626 		ifr->ifr_flags = ifp->if_flags;
1627 		break;
1628 
1629 	case SIOCGIFMETRIC:
1630 		ifr = data;
1631 		ifr->ifr_metric = ifp->if_metric;
1632 		break;
1633 
1634 	case SIOCGIFMTU:
1635 		ifr = data;
1636 		ifr->ifr_mtu = ifp->if_mtu;
1637 		break;
1638 
1639 	case SIOCGIFDLT:
1640 		ifr = data;
1641 		ifr->ifr_dlt = ifp->if_dlt;
1642 		break;
1643 
1644 	case SIOCGIFCAP:
1645 		ifcr = data;
1646 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1647 		ifcr->ifcr_capenable = ifp->if_capenable;
1648 		break;
1649 
1650 	case SIOCSIFMETRIC:
1651 		ifr = data;
1652 		ifp->if_metric = ifr->ifr_metric;
1653 		break;
1654 
1655 	case SIOCGIFDATA:
1656 		ifdr = data;
1657 		ifdr->ifdr_data = ifp->if_data;
1658 		break;
1659 
1660 	case SIOCGIFINDEX:
1661 		ifr = data;
1662 		ifr->ifr_index = ifp->if_index;
1663 		break;
1664 
1665 	case SIOCZIFDATA:
1666 		ifdr = data;
1667 		ifdr->ifdr_data = ifp->if_data;
1668 		/*
1669 		 * Assumes that the volatile counters that can be
1670 		 * zero'ed are at the end of if_data.
1671 		 */
1672 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1673 		    offsetof(struct if_data, ifi_ipackets));
1674 		/*
1675 		 * The memset() clears to the bottm of if_data. In the area,
1676 		 * if_lastchange is included. Please be careful if new entry
1677 		 * will be added into if_data or rewite this.
1678 		 *
1679 		 * And also, update if_lastchnage.
1680 		 */
1681 		getnanotime(&ifp->if_lastchange);
1682 		break;
1683 	case SIOCSIFMTU:
1684 		ifr = data;
1685 		if (ifp->if_mtu == ifr->ifr_mtu)
1686 			break;
1687 		ifp->if_mtu = ifr->ifr_mtu;
1688 		/*
1689 		 * If the link MTU changed, do network layer specific procedure.
1690 		 */
1691 #ifdef INET6
1692 		if (in6_present)
1693 			nd6_setmtu(ifp);
1694 #endif
1695 		return ENETRESET;
1696 	default:
1697 		return ENOTTY;
1698 	}
1699 	return 0;
1700 }
1701 
1702 int
1703 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
1704     lwp_t *l)
1705 {
1706 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1707 	struct ifaddr *ifa;
1708 	const struct sockaddr *any, *sa;
1709 	union {
1710 		struct sockaddr sa;
1711 		struct sockaddr_storage ss;
1712 	} u, v;
1713 
1714 	switch (cmd) {
1715 	case SIOCSIFADDRPREF:
1716 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
1717 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1718 		    NULL) != 0)
1719 			return EPERM;
1720 	case SIOCGIFADDRPREF:
1721 		break;
1722 	default:
1723 		return EOPNOTSUPP;
1724 	}
1725 
1726 	/* sanity checks */
1727 	if (data == NULL || ifp == NULL) {
1728 		panic("invalid argument to %s", __func__);
1729 		/*NOTREACHED*/
1730 	}
1731 
1732 	/* address must be specified on ADD and DELETE */
1733 	sa = sstocsa(&ifap->ifap_addr);
1734 	if (sa->sa_family != sofamily(so))
1735 		return EINVAL;
1736 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1737 		return EINVAL;
1738 
1739 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1740 
1741 	IFADDR_FOREACH(ifa, ifp) {
1742 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1743 			continue;
1744 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1745 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1746 			break;
1747 	}
1748 	if (ifa == NULL)
1749 		return EADDRNOTAVAIL;
1750 
1751 	switch (cmd) {
1752 	case SIOCSIFADDRPREF:
1753 		ifa->ifa_preference = ifap->ifap_preference;
1754 		return 0;
1755 	case SIOCGIFADDRPREF:
1756 		/* fill in the if_laddrreq structure */
1757 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1758 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1759 		ifap->ifap_preference = ifa->ifa_preference;
1760 		return 0;
1761 	default:
1762 		return EOPNOTSUPP;
1763 	}
1764 }
1765 
1766 static void
1767 ifnet_lock_enter(struct ifnet_lock *il)
1768 {
1769 	uint64_t *nenter;
1770 
1771 	/* Before trying to acquire the mutex, increase the count of threads
1772 	 * who have entered or who wait to enter the critical section.
1773 	 * Avoid one costly locked memory transaction by keeping a count for
1774 	 * each CPU.
1775 	 */
1776 	nenter = percpu_getref(il->il_nenter);
1777 	(*nenter)++;
1778 	percpu_putref(il->il_nenter);
1779 	mutex_enter(&il->il_lock);
1780 }
1781 
1782 static void
1783 ifnet_lock_exit(struct ifnet_lock *il)
1784 {
1785 	/* Increase the count of threads who have exited the critical
1786 	 * section.  Increase while we still hold the lock.
1787 	 */
1788 	il->il_nexit++;
1789 	mutex_exit(&il->il_lock);
1790 }
1791 
1792 /*
1793  * Interface ioctls.
1794  */
1795 static int
1796 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1797 {
1798 	struct ifnet *ifp;
1799 	struct ifreq *ifr;
1800 	int error = 0;
1801 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1802 	u_long ocmd = cmd;
1803 #endif
1804 	short oif_flags;
1805 #ifdef COMPAT_OIFREQ
1806 	struct ifreq ifrb;
1807 	struct oifreq *oifr = NULL;
1808 #endif
1809 
1810 	switch (cmd) {
1811 #ifdef COMPAT_OIFREQ
1812 	case OSIOCGIFCONF:
1813 	case OOSIOCGIFCONF:
1814 		return compat_ifconf(cmd, data);
1815 #endif
1816 #ifdef COMPAT_OIFDATA
1817 	case OSIOCGIFDATA:
1818 	case OSIOCZIFDATA:
1819 		return compat_ifdatareq(l, cmd, data);
1820 #endif
1821 	case SIOCGIFCONF:
1822 		return ifconf(cmd, data);
1823 	case SIOCINITIFADDR:
1824 		return EPERM;
1825 	}
1826 
1827 #ifdef COMPAT_OIFREQ
1828 	cmd = compat_cvtcmd(cmd);
1829 	if (cmd != ocmd) {
1830 		oifr = data;
1831 		data = ifr = &ifrb;
1832 		ifreqo2n(oifr, ifr);
1833 	} else
1834 #endif
1835 		ifr = data;
1836 
1837 	ifp = ifunit(ifr->ifr_name);
1838 
1839 	switch (cmd) {
1840 	case SIOCIFCREATE:
1841 	case SIOCIFDESTROY:
1842 		if (l != NULL) {
1843 			error = kauth_authorize_network(l->l_cred,
1844 			    KAUTH_NETWORK_INTERFACE,
1845 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1846 			    (void *)cmd, NULL);
1847 			if (error != 0)
1848 				return error;
1849 		}
1850 		return (cmd == SIOCIFCREATE) ?
1851 			if_clone_create(ifr->ifr_name) :
1852 			if_clone_destroy(ifr->ifr_name);
1853 
1854 	case SIOCIFGCLONERS:
1855 		return if_clone_list((struct if_clonereq *)data);
1856 	}
1857 
1858 	if (ifp == NULL)
1859 		return ENXIO;
1860 
1861 	switch (cmd) {
1862 	case SIOCALIFADDR:
1863 	case SIOCDLIFADDR:
1864 	case SIOCSIFADDRPREF:
1865 	case SIOCSIFFLAGS:
1866 	case SIOCSIFCAP:
1867 	case SIOCSIFMETRIC:
1868 	case SIOCZIFDATA:
1869 	case SIOCSIFMTU:
1870 	case SIOCSIFPHYADDR:
1871 	case SIOCDIFPHYADDR:
1872 #ifdef INET6
1873 	case SIOCSIFPHYADDR_IN6:
1874 #endif
1875 	case SIOCSLIFPHYADDR:
1876 	case SIOCADDMULTI:
1877 	case SIOCDELMULTI:
1878 	case SIOCSIFMEDIA:
1879 	case SIOCSDRVSPEC:
1880 	case SIOCG80211:
1881 	case SIOCS80211:
1882 	case SIOCS80211NWID:
1883 	case SIOCS80211NWKEY:
1884 	case SIOCS80211POWER:
1885 	case SIOCS80211BSSID:
1886 	case SIOCS80211CHANNEL:
1887 	case SIOCSLINKSTR:
1888 		if (l != NULL) {
1889 			error = kauth_authorize_network(l->l_cred,
1890 			    KAUTH_NETWORK_INTERFACE,
1891 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1892 			    (void *)cmd, NULL);
1893 			if (error != 0)
1894 				return error;
1895 		}
1896 	}
1897 
1898 	oif_flags = ifp->if_flags;
1899 
1900 	ifnet_lock_enter(ifp->if_ioctl_lock);
1901 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1902 	if (error != ENOTTY)
1903 		;
1904 	else if (so->so_proto == NULL)
1905 		error = EOPNOTSUPP;
1906 	else {
1907 #ifdef COMPAT_OSOCK
1908 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1909 #else
1910 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1911 		    (struct mbuf *)cmd, (struct mbuf *)data,
1912 		    (struct mbuf *)ifp, l);
1913 #endif
1914 	}
1915 
1916 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1917 #ifdef INET6
1918 		if (in6_present && (ifp->if_flags & IFF_UP) != 0) {
1919 			int s = splnet();
1920 			in6_if_up(ifp);
1921 			splx(s);
1922 		}
1923 #endif
1924 	}
1925 #ifdef COMPAT_OIFREQ
1926 	if (cmd != ocmd)
1927 		ifreqn2o(oifr, ifr);
1928 #endif
1929 
1930 	ifnet_lock_exit(ifp->if_ioctl_lock);
1931 	return error;
1932 }
1933 
1934 /* This callback adds to the sum in `arg' the number of
1935  * threads on `ci' who have entered or who wait to enter the
1936  * critical section.
1937  */
1938 static void
1939 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
1940 {
1941 	uint64_t *sum = arg, *nenter = p;
1942 
1943 	*sum += *nenter;
1944 }
1945 
1946 /* Return the number of threads who have entered or who wait
1947  * to enter the critical section on all CPUs.
1948  */
1949 static uint64_t
1950 ifnet_lock_entrances(struct ifnet_lock *il)
1951 {
1952 	uint64_t sum = 0;
1953 
1954 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
1955 
1956 	return sum;
1957 }
1958 
1959 static int
1960 ifioctl_attach(struct ifnet *ifp)
1961 {
1962 	struct ifnet_lock *il;
1963 
1964 	/* If the driver has not supplied its own if_ioctl, then
1965 	 * supply the default.
1966 	 */
1967 	if (ifp->if_ioctl == NULL)
1968 		ifp->if_ioctl = ifioctl_common;
1969 
1970 	/* Create an ifnet_lock for synchronizing ifioctls. */
1971 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
1972 		return ENOMEM;
1973 
1974 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
1975 	if (il->il_nenter == NULL) {
1976 		kmem_free(il, sizeof(*il));
1977 		return ENOMEM;
1978 	}
1979 
1980 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
1981 	cv_init(&il->il_emptied, ifp->if_xname);
1982 
1983 	ifp->if_ioctl_lock = il;
1984 
1985 	return 0;
1986 }
1987 
1988 /*
1989  * This must not be called until after `ifp' has been withdrawn from the
1990  * ifnet tables so that ifioctl() cannot get a handle on it by calling
1991  * ifunit().
1992  */
1993 static void
1994 ifioctl_detach(struct ifnet *ifp)
1995 {
1996 	struct ifnet_lock *il;
1997 
1998 	il = ifp->if_ioctl_lock;
1999 	mutex_enter(&il->il_lock);
2000 	/* Install if_nullioctl to make sure that any thread that
2001 	 * subsequently enters the critical section will quit it
2002 	 * immediately and signal the condition variable that we
2003 	 * wait on, below.
2004 	 */
2005 	ifp->if_ioctl = if_nullioctl;
2006 	/* Sleep while threads are still in the critical section or
2007 	 * wait to enter it.
2008 	 */
2009 	while (ifnet_lock_entrances(il) != il->il_nexit)
2010 		cv_wait(&il->il_emptied, &il->il_lock);
2011 	/* At this point, we are the only thread still in the critical
2012 	 * section, and no new thread can get a handle on the ifioctl
2013 	 * lock, so it is safe to free its memory.
2014 	 */
2015 	mutex_exit(&il->il_lock);
2016 	ifp->if_ioctl_lock = NULL;
2017 	percpu_free(il->il_nenter, sizeof(uint64_t));
2018 	il->il_nenter = NULL;
2019 	cv_destroy(&il->il_emptied);
2020 	mutex_destroy(&il->il_lock);
2021 	kmem_free(il, sizeof(*il));
2022 }
2023 
2024 /*
2025  * Return interface configuration
2026  * of system.  List may be used
2027  * in later ioctl's (above) to get
2028  * other information.
2029  *
2030  * Each record is a struct ifreq.  Before the addition of
2031  * sockaddr_storage, the API rule was that sockaddr flavors that did
2032  * not fit would extend beyond the struct ifreq, with the next struct
2033  * ifreq starting sa_len beyond the struct sockaddr.  Because the
2034  * union in struct ifreq includes struct sockaddr_storage, every kind
2035  * of sockaddr must fit.  Thus, there are no longer any overlength
2036  * records.
2037  *
2038  * Records are added to the user buffer if they fit, and ifc_len is
2039  * adjusted to the length that was written.  Thus, the user is only
2040  * assured of getting the complete list if ifc_len on return is at
2041  * least sizeof(struct ifreq) less than it was on entry.
2042  *
2043  * If the user buffer pointer is NULL, this routine copies no data and
2044  * returns the amount of space that would be needed.
2045  *
2046  * Invariants:
2047  * ifrp points to the next part of the user's buffer to be used.  If
2048  * ifrp != NULL, space holds the number of bytes remaining that we may
2049  * write at ifrp.  Otherwise, space holds the number of bytes that
2050  * would have been written had there been adequate space.
2051  */
2052 /*ARGSUSED*/
2053 int
2054 ifconf(u_long cmd, void *data)
2055 {
2056 	struct ifconf *ifc = (struct ifconf *)data;
2057 	struct ifnet *ifp;
2058 	struct ifaddr *ifa;
2059 	struct ifreq ifr, *ifrp;
2060 	int space, error = 0;
2061 	const int sz = (int)sizeof(struct ifreq);
2062 
2063 	if ((ifrp = ifc->ifc_req) == NULL)
2064 		space = 0;
2065 	else
2066 		space = ifc->ifc_len;
2067 	IFNET_FOREACH(ifp) {
2068 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2069 		    sizeof(ifr.ifr_name));
2070 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2071 			return ENAMETOOLONG;
2072 		if (IFADDR_EMPTY(ifp)) {
2073 			/* Interface with no addresses - send zero sockaddr. */
2074 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2075 			if (ifrp == NULL) {
2076 				space += sz;
2077 				continue;
2078 			}
2079 			if (space >= sz) {
2080 				error = copyout(&ifr, ifrp, sz);
2081 				if (error != 0)
2082 					return error;
2083 				ifrp++;
2084 				space -= sz;
2085 			}
2086 		}
2087 
2088 		IFADDR_FOREACH(ifa, ifp) {
2089 			struct sockaddr *sa = ifa->ifa_addr;
2090 			/* all sockaddrs must fit in sockaddr_storage */
2091 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2092 
2093 			if (ifrp == NULL) {
2094 				space += sz;
2095 				continue;
2096 			}
2097 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2098 			if (space >= sz) {
2099 				error = copyout(&ifr, ifrp, sz);
2100 				if (error != 0)
2101 					return (error);
2102 				ifrp++; space -= sz;
2103 			}
2104 		}
2105 	}
2106 	if (ifrp != NULL) {
2107 		KASSERT(0 <= space && space <= ifc->ifc_len);
2108 		ifc->ifc_len -= space;
2109 	} else {
2110 		KASSERT(space >= 0);
2111 		ifc->ifc_len = space;
2112 	}
2113 	return (0);
2114 }
2115 
2116 int
2117 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2118 {
2119 	uint8_t len;
2120 #ifdef COMPAT_OIFREQ
2121 	struct ifreq ifrb;
2122 	struct oifreq *oifr = NULL;
2123 	u_long ocmd = cmd;
2124 	cmd = compat_cvtcmd(cmd);
2125 	if (cmd != ocmd) {
2126 		oifr = (struct oifreq *)(void *)ifr;
2127 		ifr = &ifrb;
2128 		ifreqo2n(oifr, ifr);
2129 		len = sizeof(oifr->ifr_addr);
2130 	} else
2131 #endif
2132 		len = sizeof(ifr->ifr_ifru.ifru_space);
2133 
2134 	if (len < sa->sa_len)
2135 		return EFBIG;
2136 
2137 	memset(&ifr->ifr_addr, 0, len);
2138 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2139 
2140 #ifdef COMPAT_OIFREQ
2141 	if (cmd != ocmd)
2142 		ifreqn2o(oifr, ifr);
2143 #endif
2144 	return 0;
2145 }
2146 
2147 /*
2148  * Queue message on interface, and start output if interface
2149  * not yet active.
2150  */
2151 int
2152 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2153     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2154 {
2155 	int len = m->m_pkthdr.len;
2156 	int mflags = m->m_flags;
2157 	int s = splnet();
2158 	int error;
2159 
2160 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2161 	if (error != 0)
2162 		goto out;
2163 	ifp->if_obytes += len;
2164 	if (mflags & M_MCAST)
2165 		ifp->if_omcasts++;
2166 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2167 		(*ifp->if_start)(ifp);
2168 out:
2169 	splx(s);
2170 	return error;
2171 }
2172 
2173 /*
2174  * Queue message on interface, possibly using a second fast queue
2175  */
2176 int
2177 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2178     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2179 {
2180 	int error = 0;
2181 
2182 	if (ifq != NULL
2183 #ifdef ALTQ
2184 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2185 #endif
2186 	    ) {
2187 		if (IF_QFULL(ifq)) {
2188 			IF_DROP(&ifp->if_snd);
2189 			m_freem(m);
2190 			if (error == 0)
2191 				error = ENOBUFS;
2192 		} else
2193 			IF_ENQUEUE(ifq, m);
2194 	} else
2195 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2196 	if (error != 0) {
2197 		++ifp->if_oerrors;
2198 		return error;
2199 	}
2200 	return 0;
2201 }
2202 
2203 int
2204 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2205 {
2206 	int rc;
2207 
2208 	if (ifp->if_initaddr != NULL)
2209 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2210 	else if (src ||
2211 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2212 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2213 
2214 	return rc;
2215 }
2216 
2217 int
2218 if_flags_set(ifnet_t *ifp, const short flags)
2219 {
2220 	int rc;
2221 
2222 	if (ifp->if_setflags != NULL)
2223 		rc = (*ifp->if_setflags)(ifp, flags);
2224 	else {
2225 		short cantflags, chgdflags;
2226 		struct ifreq ifr;
2227 
2228 		chgdflags = ifp->if_flags ^ flags;
2229 		cantflags = chgdflags & IFF_CANTCHANGE;
2230 
2231 		if (cantflags != 0)
2232 			ifp->if_flags ^= cantflags;
2233 
2234                 /* Traditionally, we do not call if_ioctl after
2235                  * setting/clearing only IFF_PROMISC if the interface
2236                  * isn't IFF_UP.  Uphold that tradition.
2237 		 */
2238 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2239 			return 0;
2240 
2241 		memset(&ifr, 0, sizeof(ifr));
2242 
2243 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2244 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2245 
2246 		if (rc != 0 && cantflags != 0)
2247 			ifp->if_flags ^= cantflags;
2248 	}
2249 
2250 	return rc;
2251 }
2252 
2253 int
2254 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2255 {
2256 	int rc;
2257 	struct ifreq ifr;
2258 
2259 	if (ifp->if_mcastop != NULL)
2260 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2261 	else {
2262 		ifreq_setaddr(cmd, &ifr, sa);
2263 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2264 	}
2265 
2266 	return rc;
2267 }
2268 
2269 static void
2270 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2271     struct ifaltq *ifq)
2272 {
2273 	const struct sysctlnode *cnode, *rnode;
2274 
2275 	if (sysctl_createv(clog, 0, NULL, &rnode,
2276 		       CTLFLAG_PERMANENT,
2277 		       CTLTYPE_NODE, "interfaces",
2278 		       SYSCTL_DESCR("Per-interface controls"),
2279 		       NULL, 0, NULL, 0,
2280 		       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
2281 		goto bad;
2282 
2283 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2284 		       CTLFLAG_PERMANENT,
2285 		       CTLTYPE_NODE, ifname,
2286 		       SYSCTL_DESCR("Interface controls"),
2287 		       NULL, 0, NULL, 0,
2288 		       CTL_CREATE, CTL_EOL) != 0)
2289 		goto bad;
2290 
2291 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2292 		       CTLFLAG_PERMANENT,
2293 		       CTLTYPE_NODE, "sndq",
2294 		       SYSCTL_DESCR("Interface output queue controls"),
2295 		       NULL, 0, NULL, 0,
2296 		       CTL_CREATE, CTL_EOL) != 0)
2297 		goto bad;
2298 
2299 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2300 		       CTLFLAG_PERMANENT,
2301 		       CTLTYPE_INT, "len",
2302 		       SYSCTL_DESCR("Current output queue length"),
2303 		       NULL, 0, &ifq->ifq_len, 0,
2304 		       CTL_CREATE, CTL_EOL) != 0)
2305 		goto bad;
2306 
2307 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2308 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2309 		       CTLTYPE_INT, "maxlen",
2310 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2311 		       NULL, 0, &ifq->ifq_maxlen, 0,
2312 		       CTL_CREATE, CTL_EOL) != 0)
2313 		goto bad;
2314 
2315 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2316 		       CTLFLAG_PERMANENT,
2317 		       CTLTYPE_INT, "drops",
2318 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2319 		       NULL, 0, &ifq->ifq_drops, 0,
2320 		       CTL_CREATE, CTL_EOL) != 0)
2321 		goto bad;
2322 
2323 	return;
2324 bad:
2325 	printf("%s: could not attach sysctl nodes\n", ifname);
2326 	return;
2327 }
2328 
2329 #if defined(INET) || defined(INET6)
2330 static void
2331 sysctl_net_ifq_setup(struct sysctllog **clog,
2332 		     int pf, const char *pfname,
2333 		     int ipn, const char *ipname,
2334 		     int qid, struct ifqueue *ifq)
2335 {
2336 
2337 	sysctl_createv(clog, 0, NULL, NULL,
2338 		       CTLFLAG_PERMANENT,
2339 		       CTLTYPE_NODE, pfname, NULL,
2340 		       NULL, 0, NULL, 0,
2341 		       CTL_NET, pf, CTL_EOL);
2342 	sysctl_createv(clog, 0, NULL, NULL,
2343 		       CTLFLAG_PERMANENT,
2344 		       CTLTYPE_NODE, ipname, NULL,
2345 		       NULL, 0, NULL, 0,
2346 		       CTL_NET, pf, ipn, CTL_EOL);
2347 	sysctl_createv(clog, 0, NULL, NULL,
2348 		       CTLFLAG_PERMANENT,
2349 		       CTLTYPE_NODE, "ifq",
2350 		       SYSCTL_DESCR("Protocol input queue controls"),
2351 		       NULL, 0, NULL, 0,
2352 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2353 
2354 	sysctl_createv(clog, 0, NULL, NULL,
2355 		       CTLFLAG_PERMANENT,
2356 		       CTLTYPE_INT, "len",
2357 		       SYSCTL_DESCR("Current input queue length"),
2358 		       NULL, 0, &ifq->ifq_len, 0,
2359 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2360 	sysctl_createv(clog, 0, NULL, NULL,
2361 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2362 		       CTLTYPE_INT, "maxlen",
2363 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2364 		       NULL, 0, &ifq->ifq_maxlen, 0,
2365 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2366 #ifdef notyet
2367 	sysctl_createv(clog, 0, NULL, NULL,
2368 		       CTLFLAG_PERMANENT,
2369 		       CTLTYPE_INT, "peak",
2370 		       SYSCTL_DESCR("Highest input queue length"),
2371 		       NULL, 0, &ifq->ifq_peak, 0,
2372 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
2373 #endif
2374 	sysctl_createv(clog, 0, NULL, NULL,
2375 		       CTLFLAG_PERMANENT,
2376 		       CTLTYPE_INT, "drops",
2377 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2378 		       NULL, 0, &ifq->ifq_drops, 0,
2379 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2380 }
2381 #endif /* INET || INET6 */
2382