xref: /netbsd-src/sys/net/if.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: if.c,v 1.208 2007/12/06 02:23:42 dyoung Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000, 2001 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by William Studenmund and Jason R. Thorpe.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the NetBSD
21  *	Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 
39 /*
40  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
41  * All rights reserved.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. Neither the name of the project nor the names of its contributors
52  *    may be used to endorse or promote products derived from this software
53  *    without specific prior written permission.
54  *
55  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65  * SUCH DAMAGE.
66  */
67 
68 /*
69  * Copyright (c) 1980, 1986, 1993
70  *	The Regents of the University of California.  All rights reserved.
71  *
72  * Redistribution and use in source and binary forms, with or without
73  * modification, are permitted provided that the following conditions
74  * are met:
75  * 1. Redistributions of source code must retain the above copyright
76  *    notice, this list of conditions and the following disclaimer.
77  * 2. Redistributions in binary form must reproduce the above copyright
78  *    notice, this list of conditions and the following disclaimer in the
79  *    documentation and/or other materials provided with the distribution.
80  * 3. Neither the name of the University nor the names of its contributors
81  *    may be used to endorse or promote products derived from this software
82  *    without specific prior written permission.
83  *
84  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
85  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
86  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
87  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
88  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
89  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
90  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
91  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
92  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
93  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
94  * SUCH DAMAGE.
95  *
96  *	@(#)if.c	8.5 (Berkeley) 1/9/95
97  */
98 
99 #include <sys/cdefs.h>
100 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.208 2007/12/06 02:23:42 dyoung Exp $");
101 
102 #include "opt_inet.h"
103 
104 #include "opt_atalk.h"
105 #include "opt_natm.h"
106 #include "opt_pfil_hooks.h"
107 
108 #include <sys/param.h>
109 #include <sys/mbuf.h>
110 #include <sys/systm.h>
111 #include <sys/callout.h>
112 #include <sys/proc.h>
113 #include <sys/socket.h>
114 #include <sys/socketvar.h>
115 #include <sys/domain.h>
116 #include <sys/protosw.h>
117 #include <sys/kernel.h>
118 #include <sys/ioctl.h>
119 #include <sys/sysctl.h>
120 #include <sys/syslog.h>
121 #include <sys/kauth.h>
122 
123 #include <net/if.h>
124 #include <net/if_dl.h>
125 #include <net/if_ether.h>
126 #include <net/if_media.h>
127 #include <net80211/ieee80211.h>
128 #include <net80211/ieee80211_ioctl.h>
129 #include <net/if_types.h>
130 #include <net/radix.h>
131 #include <net/route.h>
132 #include <net/netisr.h>
133 #ifdef NETATALK
134 #include <netatalk/at_extern.h>
135 #include <netatalk/at.h>
136 #endif
137 #include <net/pfil.h>
138 
139 #ifdef INET6
140 #include <netinet/in.h>
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 #include "carp.h"
146 #if NCARP > 0
147 #include <netinet/ip_carp.h>
148 #endif
149 
150 #include <compat/sys/sockio.h>
151 #include <compat/sys/socket.h>
152 
153 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
154 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
155 
156 int	ifqmaxlen = IFQ_MAXLEN;
157 callout_t if_slowtimo_ch;
158 
159 int netisr;			/* scheduling bits for network */
160 
161 static int	if_rt_walktree(struct rtentry *, void *);
162 
163 static struct if_clone *if_clone_lookup(const char *, int *);
164 static int	if_clone_list(struct if_clonereq *);
165 
166 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
167 static int if_cloners_count;
168 
169 #ifdef PFIL_HOOKS
170 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
171 #endif
172 
173 static void if_detach_queues(struct ifnet *, struct ifqueue *);
174 
175 /*
176  * Network interface utility routines.
177  *
178  * Routines with ifa_ifwith* names take sockaddr *'s as
179  * parameters.
180  */
181 void
182 ifinit(void)
183 {
184 
185 	callout_init(&if_slowtimo_ch, 0);
186 	if_slowtimo(NULL);
187 #ifdef PFIL_HOOKS
188 	if_pfil.ph_type = PFIL_TYPE_IFNET;
189 	if_pfil.ph_ifnet = NULL;
190 	if (pfil_head_register(&if_pfil) != 0)
191 		printf("WARNING: unable to register pfil hook\n");
192 #endif
193 }
194 
195 /*
196  * Null routines used while an interface is going away.  These routines
197  * just return an error.
198  */
199 
200 int
201 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
202     const struct sockaddr *so, struct rtentry *rt)
203 {
204 
205 	return ENXIO;
206 }
207 
208 void
209 if_nullinput(struct ifnet *ifp, struct mbuf *m)
210 {
211 
212 	/* Nothing. */
213 }
214 
215 void
216 if_nullstart(struct ifnet *ifp)
217 {
218 
219 	/* Nothing. */
220 }
221 
222 int
223 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
224 {
225 
226 	return ENXIO;
227 }
228 
229 int
230 if_nullinit(struct ifnet *ifp)
231 {
232 
233 	return ENXIO;
234 }
235 
236 void
237 if_nullstop(struct ifnet *ifp, int disable)
238 {
239 
240 	/* Nothing. */
241 }
242 
243 void
244 if_nullwatchdog(struct ifnet *ifp)
245 {
246 
247 	/* Nothing. */
248 }
249 
250 void
251 if_nulldrain(struct ifnet *ifp)
252 {
253 
254 	/* Nothing. */
255 }
256 
257 static u_int if_index = 1;
258 struct ifnet_head ifnet;
259 size_t if_indexlim = 0;
260 struct ifaddr **ifnet_addrs = NULL;
261 struct ifnet **ifindex2ifnet = NULL;
262 struct ifnet *lo0ifp;
263 
264 /*
265  * Allocate the link level name for the specified interface.  This
266  * is an attachment helper.  It must be called after ifp->if_addrlen
267  * is initialized, which may not be the case when if_attach() is
268  * called.
269  */
270 void
271 if_alloc_sadl(struct ifnet *ifp)
272 {
273 	unsigned socksize, ifasize;
274 	int addrlen, namelen;
275 	struct sockaddr_dl *mask, *sdl;
276 	struct ifaddr *ifa;
277 
278 	/*
279 	 * If the interface already has a link name, release it
280 	 * now.  This is useful for interfaces that can change
281 	 * link types, and thus switch link names often.
282 	 */
283 	if (ifp->if_sadl != NULL)
284 		if_free_sadl(ifp);
285 
286 	namelen = strlen(ifp->if_xname);
287 	addrlen = ifp->if_addrlen;
288 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
289 	ifasize = sizeof(*ifa) + 2 * socksize;
290 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
291 
292 	sdl = (struct sockaddr_dl *)(ifa + 1);
293 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
294 
295 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
296 	    ifp->if_xname, namelen, NULL, addrlen);
297 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
298 	memset(&mask->sdl_data[0], 0xff, namelen);
299 
300 	ifnet_addrs[ifp->if_index] = ifa;
301 	IFAREF(ifa);
302 	ifa_insert(ifp, ifa);
303 	ifa->ifa_rtrequest = link_rtrequest;
304 	ifa->ifa_addr = (struct sockaddr *)sdl;
305 	ifp->if_sadl = sdl;
306 	ifa->ifa_netmask = (struct sockaddr *)mask;
307 }
308 
309 /*
310  * Free the link level name for the specified interface.  This is
311  * a detach helper.  This is called from if_detach() or from
312  * link layer type specific detach functions.
313  */
314 void
315 if_free_sadl(struct ifnet *ifp)
316 {
317 	struct ifaddr *ifa;
318 	int s;
319 
320 	ifa = ifnet_addrs[ifp->if_index];
321 	if (ifa == NULL) {
322 		KASSERT(ifp->if_sadl == NULL);
323 		return;
324 	}
325 
326 	KASSERT(ifp->if_sadl != NULL);
327 
328 	s = splnet();
329 	rtinit(ifa, RTM_DELETE, 0);
330 	ifa_remove(ifp, ifa);
331 
332 	ifp->if_sadl = NULL;
333 
334 	ifnet_addrs[ifp->if_index] = NULL;
335 	IFAFREE(ifa);
336 	splx(s);
337 }
338 
339 /*
340  * Attach an interface to the
341  * list of "active" interfaces.
342  */
343 void
344 if_attach(struct ifnet *ifp)
345 {
346 	int indexlim = 0;
347 
348 	if (if_indexlim == 0) {
349 		TAILQ_INIT(&ifnet);
350 		if_indexlim = 8;
351 	}
352 	TAILQ_INIT(&ifp->if_addrlist);
353 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
354 	ifp->if_index = if_index;
355 	if (ifindex2ifnet == NULL)
356 		if_index++;
357 	else
358 		while (ifp->if_index < if_indexlim &&
359 		    ifindex2ifnet[ifp->if_index] != NULL) {
360 			++if_index;
361 			if (if_index == 0)
362 				if_index = 1;
363 			/*
364 			 * If we hit USHRT_MAX, we skip back to 0 since
365 			 * there are a number of places where the value
366 			 * of if_index or if_index itself is compared
367 			 * to or stored in an unsigned short.  By
368 			 * jumping back, we won't botch those assignments
369 			 * or comparisons.
370 			 */
371 			else if (if_index == USHRT_MAX) {
372 				/*
373 				 * However, if we have to jump back to
374 				 * zero *twice* without finding an empty
375 				 * slot in ifindex2ifnet[], then there
376 				 * there are too many (>65535) interfaces.
377 				 */
378 				if (indexlim++)
379 					panic("too many interfaces");
380 				else
381 					if_index = 1;
382 			}
383 			ifp->if_index = if_index;
384 		}
385 
386 	/*
387 	 * We have some arrays that should be indexed by if_index.
388 	 * since if_index will grow dynamically, they should grow too.
389 	 *	struct ifadd **ifnet_addrs
390 	 *	struct ifnet **ifindex2ifnet
391 	 */
392 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
393 	    ifp->if_index >= if_indexlim) {
394 		size_t m, n, oldlim;
395 		void *q;
396 
397 		oldlim = if_indexlim;
398 		while (ifp->if_index >= if_indexlim)
399 			if_indexlim <<= 1;
400 
401 		/* grow ifnet_addrs */
402 		m = oldlim * sizeof(struct ifaddr *);
403 		n = if_indexlim * sizeof(struct ifaddr *);
404 		q = (void *)malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
405 		if (ifnet_addrs != NULL) {
406 			memcpy(q, ifnet_addrs, m);
407 			free((void *)ifnet_addrs, M_IFADDR);
408 		}
409 		ifnet_addrs = (struct ifaddr **)q;
410 
411 		/* grow ifindex2ifnet */
412 		m = oldlim * sizeof(struct ifnet *);
413 		n = if_indexlim * sizeof(struct ifnet *);
414 		q = (void *)malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
415 		if (ifindex2ifnet != NULL) {
416 			memcpy(q, (void *)ifindex2ifnet, m);
417 			free((void *)ifindex2ifnet, M_IFADDR);
418 		}
419 		ifindex2ifnet = (struct ifnet **)q;
420 	}
421 
422 	ifindex2ifnet[ifp->if_index] = ifp;
423 
424 	/*
425 	 * Link level name is allocated later by a separate call to
426 	 * if_alloc_sadl().
427 	 */
428 
429 	if (ifp->if_snd.ifq_maxlen == 0)
430 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
431 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
432 
433 	ifp->if_link_state = LINK_STATE_UNKNOWN;
434 
435 	ifp->if_capenable = 0;
436 	ifp->if_csum_flags_tx = 0;
437 	ifp->if_csum_flags_rx = 0;
438 
439 #ifdef ALTQ
440 	ifp->if_snd.altq_type = 0;
441 	ifp->if_snd.altq_disc = NULL;
442 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
443 	ifp->if_snd.altq_tbr  = NULL;
444 	ifp->if_snd.altq_ifp  = ifp;
445 #endif
446 
447 #ifdef PFIL_HOOKS
448 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
449 	ifp->if_pfil.ph_ifnet = ifp;
450 	if (pfil_head_register(&ifp->if_pfil) != 0)
451 		printf("%s: WARNING: unable to register pfil hook\n",
452 		    ifp->if_xname);
453 	(void)pfil_run_hooks(&if_pfil,
454 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
455 #endif
456 
457 	if (!STAILQ_EMPTY(&domains))
458 		if_attachdomain1(ifp);
459 
460 	/* Announce the interface. */
461 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
462 }
463 
464 void
465 if_attachdomain(void)
466 {
467 	struct ifnet *ifp;
468 	int s;
469 
470 	s = splnet();
471 	IFNET_FOREACH(ifp)
472 		if_attachdomain1(ifp);
473 	splx(s);
474 }
475 
476 void
477 if_attachdomain1(struct ifnet *ifp)
478 {
479 	struct domain *dp;
480 	int s;
481 
482 	s = splnet();
483 
484 	/* address family dependent data region */
485 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
486 	DOMAIN_FOREACH(dp) {
487 		if (dp->dom_ifattach != NULL)
488 			ifp->if_afdata[dp->dom_family] =
489 			    (*dp->dom_ifattach)(ifp);
490 	}
491 
492 	splx(s);
493 }
494 
495 /*
496  * Deactivate an interface.  This points all of the procedure
497  * handles at error stubs.  May be called from interrupt context.
498  */
499 void
500 if_deactivate(struct ifnet *ifp)
501 {
502 	int s;
503 
504 	s = splnet();
505 
506 	ifp->if_output	 = if_nulloutput;
507 	ifp->if_input	 = if_nullinput;
508 	ifp->if_start	 = if_nullstart;
509 	ifp->if_ioctl	 = if_nullioctl;
510 	ifp->if_init	 = if_nullinit;
511 	ifp->if_stop	 = if_nullstop;
512 	ifp->if_watchdog = if_nullwatchdog;
513 	ifp->if_drain	 = if_nulldrain;
514 
515 	/* No more packets may be enqueued. */
516 	ifp->if_snd.ifq_maxlen = 0;
517 
518 	splx(s);
519 }
520 
521 void
522 if_purgeaddrs(struct ifnet *ifp, int family,
523     void (*purgeaddr)(struct ifaddr *))
524 {
525 	struct ifaddr *ifa, *nifa;
526 
527 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
528 		nifa = IFADDR_NEXT(ifa);
529 		if (ifa->ifa_addr->sa_family != family)
530 			continue;
531 		(*purgeaddr)(ifa);
532 	}
533 }
534 
535 /*
536  * Detach an interface from the list of "active" interfaces,
537  * freeing any resources as we go along.
538  *
539  * NOTE: This routine must be called with a valid thread context,
540  * as it may block.
541  */
542 void
543 if_detach(struct ifnet *ifp)
544 {
545 	struct socket so;
546 	struct ifaddr *ifa;
547 #ifdef IFAREF_DEBUG
548 	struct ifaddr *last_ifa = NULL;
549 #endif
550 	struct domain *dp;
551 	const struct protosw *pr;
552 	int s, i, family, purged;
553 
554 	/*
555 	 * XXX It's kind of lame that we have to have the
556 	 * XXX socket structure...
557 	 */
558 	memset(&so, 0, sizeof(so));
559 
560 	s = splnet();
561 
562 	/*
563 	 * Do an if_down() to give protocols a chance to do something.
564 	 */
565 	if_down(ifp);
566 
567 #ifdef ALTQ
568 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
569 		altq_disable(&ifp->if_snd);
570 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
571 		altq_detach(&ifp->if_snd);
572 #endif
573 
574 
575 #if NCARP > 0
576 	/* Remove the interface from any carp group it is a part of.  */
577 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
578 		carp_ifdetach(ifp);
579 #endif
580 
581 	/*
582 	 * Rip all the addresses off the interface.  This should make
583 	 * all of the routes go away.
584 	 *
585 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
586 	 * from the list, including our "cursor", ifa.  For safety,
587 	 * and to honor the TAILQ abstraction, I just restart the
588 	 * loop after each removal.  Note that the loop will exit
589 	 * when all of the remaining ifaddrs belong to the AF_LINK
590 	 * family.  I am counting on the historical fact that at
591 	 * least one pr_usrreq in each address domain removes at
592 	 * least one ifaddr.
593 	 */
594 again:
595 	IFADDR_FOREACH(ifa, ifp) {
596 		family = ifa->ifa_addr->sa_family;
597 #ifdef IFAREF_DEBUG
598 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
599 		    ifa, family, ifa->ifa_refcnt);
600 		if (last_ifa != NULL && ifa == last_ifa)
601 			panic("if_detach: loop detected");
602 		last_ifa = ifa;
603 #endif
604 		if (family == AF_LINK)
605 			continue;
606 		dp = pffinddomain(family);
607 #ifdef DIAGNOSTIC
608 		if (dp == NULL)
609 			panic("if_detach: no domain for AF %d",
610 			    family);
611 #endif
612 		/*
613 		 * XXX These PURGEIF calls are redundant with the
614 		 * purge-all-families calls below, but are left in for
615 		 * now both to make a smaller change, and to avoid
616 		 * unplanned interactions with clearing of
617 		 * ifp->if_addrlist.
618 		 */
619 		purged = 0;
620 		for (pr = dp->dom_protosw;
621 		     pr < dp->dom_protoswNPROTOSW; pr++) {
622 			so.so_proto = pr;
623 			if (pr->pr_usrreq != NULL) {
624 				(void) (*pr->pr_usrreq)(&so,
625 				    PRU_PURGEIF, NULL, NULL,
626 				    (struct mbuf *) ifp, curlwp);
627 				purged = 1;
628 			}
629 		}
630 		if (purged == 0) {
631 			/*
632 			 * XXX What's really the best thing to do
633 			 * XXX here?  --thorpej@NetBSD.org
634 			 */
635 			printf("if_detach: WARNING: AF %d not purged\n",
636 			    family);
637 			ifa_remove(ifp, ifa);
638 		}
639 		goto again;
640 	}
641 
642 	if_free_sadl(ifp);
643 
644 	/* Walk the routing table looking for stragglers. */
645 	for (i = 0; i <= AF_MAX; i++)
646 		(void)rt_walktree(i, if_rt_walktree, ifp);
647 
648 	DOMAIN_FOREACH(dp) {
649 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
650 			(*dp->dom_ifdetach)(ifp,
651 			    ifp->if_afdata[dp->dom_family]);
652 
653 		/*
654 		 * One would expect multicast memberships (INET and
655 		 * INET6) on UDP sockets to be purged by the PURGEIF
656 		 * calls above, but if all addresses were removed from
657 		 * the interface prior to destruction, the calls will
658 		 * not be made (e.g. ppp, for which pppd(8) generally
659 		 * removes addresses before destroying the interface).
660 		 * Because there is no invariant that multicast
661 		 * memberships only exist for interfaces with IPv4
662 		 * addresses, we must call PURGEIF regardless of
663 		 * addresses.  (Protocols which might store ifnet
664 		 * pointers are marked with PR_PURGEIF.)
665 		 */
666 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
667 			so.so_proto = pr;
668 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
669 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
670 				    NULL, (struct mbuf *)ifp, curlwp);
671 		}
672 	}
673 
674 #ifdef PFIL_HOOKS
675 	(void)pfil_run_hooks(&if_pfil,
676 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
677 	(void)pfil_head_unregister(&ifp->if_pfil);
678 #endif
679 
680 	/* Announce that the interface is gone. */
681 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
682 
683 	ifindex2ifnet[ifp->if_index] = NULL;
684 
685 	TAILQ_REMOVE(&ifnet, ifp, if_list);
686 
687 	/*
688 	 * remove packets that came from ifp, from software interrupt queues.
689 	 */
690 	DOMAIN_FOREACH(dp) {
691 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
692 			if (dp->dom_ifqueues[i] == NULL)
693 				break;
694 			if_detach_queues(ifp, dp->dom_ifqueues[i]);
695 		}
696 	}
697 
698 	splx(s);
699 }
700 
701 static void
702 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
703 {
704 	struct mbuf *m, *prev, *next;
705 
706 	prev = NULL;
707 	for (m = q->ifq_head; m != NULL; m = next) {
708 		next = m->m_nextpkt;
709 #ifdef DIAGNOSTIC
710 		if ((m->m_flags & M_PKTHDR) == 0) {
711 			prev = m;
712 			continue;
713 		}
714 #endif
715 		if (m->m_pkthdr.rcvif != ifp) {
716 			prev = m;
717 			continue;
718 		}
719 
720 		if (prev != NULL)
721 			prev->m_nextpkt = m->m_nextpkt;
722 		else
723 			q->ifq_head = m->m_nextpkt;
724 		if (q->ifq_tail == m)
725 			q->ifq_tail = prev;
726 		q->ifq_len--;
727 
728 		m->m_nextpkt = NULL;
729 		m_freem(m);
730 		IF_DROP(q);
731 	}
732 }
733 
734 /*
735  * Callback for a radix tree walk to delete all references to an
736  * ifnet.
737  */
738 static int
739 if_rt_walktree(struct rtentry *rt, void *v)
740 {
741 	struct ifnet *ifp = (struct ifnet *)v;
742 	int error;
743 
744 	if (rt->rt_ifp != ifp)
745 		return 0;
746 
747 	/* Delete the entry. */
748 	++rt->rt_refcnt;
749 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
750 	    rt_mask(rt), rt->rt_flags, NULL);
751 	KASSERT((rt->rt_flags & RTF_UP) == 0);
752 	rt->rt_ifp = NULL;
753 	RTFREE(rt);
754 	if (error != 0)
755 		printf("%s: warning: unable to delete rtentry @ %p, "
756 		    "error = %d\n", ifp->if_xname, rt, error);
757 	return 0;
758 }
759 
760 /*
761  * Create a clone network interface.
762  */
763 int
764 if_clone_create(const char *name)
765 {
766 	struct if_clone *ifc;
767 	int unit;
768 
769 	ifc = if_clone_lookup(name, &unit);
770 	if (ifc == NULL)
771 		return EINVAL;
772 
773 	if (ifunit(name) != NULL)
774 		return EEXIST;
775 
776 	return (*ifc->ifc_create)(ifc, unit);
777 }
778 
779 /*
780  * Destroy a clone network interface.
781  */
782 int
783 if_clone_destroy(const char *name)
784 {
785 	struct if_clone *ifc;
786 	struct ifnet *ifp;
787 
788 	ifc = if_clone_lookup(name, NULL);
789 	if (ifc == NULL)
790 		return EINVAL;
791 
792 	ifp = ifunit(name);
793 	if (ifp == NULL)
794 		return ENXIO;
795 
796 	if (ifc->ifc_destroy == NULL)
797 		return EOPNOTSUPP;
798 
799 	return (*ifc->ifc_destroy)(ifp);
800 }
801 
802 /*
803  * Look up a network interface cloner.
804  */
805 static struct if_clone *
806 if_clone_lookup(const char *name, int *unitp)
807 {
808 	struct if_clone *ifc;
809 	const char *cp;
810 	int unit;
811 
812 	/* separate interface name from unit */
813 	for (cp = name;
814 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
815 	    cp++)
816 		continue;
817 
818 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
819 		return NULL;	/* No name or unit number */
820 
821 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
822 		if (strlen(ifc->ifc_name) == cp - name &&
823 		    strncmp(name, ifc->ifc_name, cp - name) == 0)
824 			break;
825 	}
826 
827 	if (ifc == NULL)
828 		return NULL;
829 
830 	unit = 0;
831 	while (cp - name < IFNAMSIZ && *cp) {
832 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
833 			/* Bogus unit number. */
834 			return NULL;
835 		}
836 		unit = (unit * 10) + (*cp++ - '0');
837 	}
838 
839 	if (unitp != NULL)
840 		*unitp = unit;
841 	return ifc;
842 }
843 
844 /*
845  * Register a network interface cloner.
846  */
847 void
848 if_clone_attach(struct if_clone *ifc)
849 {
850 
851 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
852 	if_cloners_count++;
853 }
854 
855 /*
856  * Unregister a network interface cloner.
857  */
858 void
859 if_clone_detach(struct if_clone *ifc)
860 {
861 
862 	LIST_REMOVE(ifc, ifc_list);
863 	if_cloners_count--;
864 }
865 
866 /*
867  * Provide list of interface cloners to userspace.
868  */
869 static int
870 if_clone_list(struct if_clonereq *ifcr)
871 {
872 	char outbuf[IFNAMSIZ], *dst;
873 	struct if_clone *ifc;
874 	int count, error = 0;
875 
876 	ifcr->ifcr_total = if_cloners_count;
877 	if ((dst = ifcr->ifcr_buffer) == NULL) {
878 		/* Just asking how many there are. */
879 		return 0;
880 	}
881 
882 	if (ifcr->ifcr_count < 0)
883 		return EINVAL;
884 
885 	count = (if_cloners_count < ifcr->ifcr_count) ?
886 	    if_cloners_count : ifcr->ifcr_count;
887 
888 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
889 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
890 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
891 		if (outbuf[sizeof(outbuf) - 1] != '\0')
892 			return ENAMETOOLONG;
893 		error = copyout(outbuf, dst, sizeof(outbuf));
894 		if (error != 0)
895 			break;
896 	}
897 
898 	return error;
899 }
900 
901 void
902 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
903 {
904 	ifa->ifa_ifp = ifp;
905 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
906 	IFAREF(ifa);
907 }
908 
909 void
910 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
911 {
912 	KASSERT(ifa->ifa_ifp == ifp);
913 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
914 	IFAFREE(ifa);
915 }
916 
917 static inline int
918 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
919 {
920 	return sockaddr_cmp(sa1, sa2) == 0;
921 }
922 
923 /*
924  * Locate an interface based on a complete address.
925  */
926 /*ARGSUSED*/
927 struct ifaddr *
928 ifa_ifwithaddr(const struct sockaddr *addr)
929 {
930 	struct ifnet *ifp;
931 	struct ifaddr *ifa;
932 
933 	IFNET_FOREACH(ifp) {
934 		if (ifp->if_output == if_nulloutput)
935 			continue;
936 		IFADDR_FOREACH(ifa, ifp) {
937 			if (ifa->ifa_addr->sa_family != addr->sa_family)
938 				continue;
939 			if (equal(addr, ifa->ifa_addr))
940 				return ifa;
941 			if ((ifp->if_flags & IFF_BROADCAST) &&
942 			    ifa->ifa_broadaddr &&
943 			    /* IP6 doesn't have broadcast */
944 			    ifa->ifa_broadaddr->sa_len != 0 &&
945 			    equal(ifa->ifa_broadaddr, addr))
946 				return ifa;
947 		}
948 	}
949 	return NULL;
950 }
951 
952 /*
953  * Locate the point to point interface with a given destination address.
954  */
955 /*ARGSUSED*/
956 struct ifaddr *
957 ifa_ifwithdstaddr(const struct sockaddr *addr)
958 {
959 	struct ifnet *ifp;
960 	struct ifaddr *ifa;
961 
962 	IFNET_FOREACH(ifp) {
963 		if (ifp->if_output == if_nulloutput)
964 			continue;
965 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
966 			continue;
967 		IFADDR_FOREACH(ifa, ifp) {
968 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
969 			    ifa->ifa_dstaddr == NULL)
970 				continue;
971 			if (equal(addr, ifa->ifa_dstaddr))
972 				return ifa;
973 		}
974 	}
975 	return NULL;
976 }
977 
978 /*
979  * Find an interface on a specific network.  If many, choice
980  * is most specific found.
981  */
982 struct ifaddr *
983 ifa_ifwithnet(const struct sockaddr *addr)
984 {
985 	struct ifnet *ifp;
986 	struct ifaddr *ifa;
987 	const struct sockaddr_dl *sdl;
988 	struct ifaddr *ifa_maybe = 0;
989 	u_int af = addr->sa_family;
990 	const char *addr_data = addr->sa_data, *cplim;
991 
992 	if (af == AF_LINK) {
993 		sdl = satocsdl(addr);
994 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
995 		    ifindex2ifnet[sdl->sdl_index] &&
996 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
997 			return ifnet_addrs[sdl->sdl_index];
998 	}
999 #ifdef NETATALK
1000 	if (af == AF_APPLETALK) {
1001 		const struct sockaddr_at *sat, *sat2;
1002 		sat = (const struct sockaddr_at *)addr;
1003 		IFNET_FOREACH(ifp) {
1004 			if (ifp->if_output == if_nulloutput)
1005 				continue;
1006 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1007 			if (ifa == NULL)
1008 				continue;
1009 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1010 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1011 				return ifa; /* exact match */
1012 			if (ifa_maybe == NULL) {
1013 				/* else keep the if with the right range */
1014 				ifa_maybe = ifa;
1015 			}
1016 		}
1017 		return ifa_maybe;
1018 	}
1019 #endif
1020 	IFNET_FOREACH(ifp) {
1021 		if (ifp->if_output == if_nulloutput)
1022 			continue;
1023 		IFADDR_FOREACH(ifa, ifp) {
1024 			const char *cp, *cp2, *cp3;
1025 
1026 			if (ifa->ifa_addr->sa_family != af ||
1027 			    ifa->ifa_netmask == NULL)
1028  next:				continue;
1029 			cp = addr_data;
1030 			cp2 = ifa->ifa_addr->sa_data;
1031 			cp3 = ifa->ifa_netmask->sa_data;
1032 			cplim = (const char *)ifa->ifa_netmask +
1033 			    ifa->ifa_netmask->sa_len;
1034 			while (cp3 < cplim) {
1035 				if ((*cp++ ^ *cp2++) & *cp3++) {
1036 					/* want to continue for() loop */
1037 					goto next;
1038 				}
1039 			}
1040 			if (ifa_maybe == NULL ||
1041 			    rn_refines((void *)ifa->ifa_netmask,
1042 			    (void *)ifa_maybe->ifa_netmask))
1043 				ifa_maybe = ifa;
1044 		}
1045 	}
1046 	return ifa_maybe;
1047 }
1048 
1049 /*
1050  * Find the interface of the addresss.
1051  */
1052 struct ifaddr *
1053 ifa_ifwithladdr(const struct sockaddr *addr)
1054 {
1055 	struct ifaddr *ia;
1056 
1057 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1058 	    (ia = ifa_ifwithnet(addr)))
1059 		return ia;
1060 	return NULL;
1061 }
1062 
1063 /*
1064  * Find an interface using a specific address family
1065  */
1066 struct ifaddr *
1067 ifa_ifwithaf(int af)
1068 {
1069 	struct ifnet *ifp;
1070 	struct ifaddr *ifa;
1071 
1072 	IFNET_FOREACH(ifp) {
1073 		if (ifp->if_output == if_nulloutput)
1074 			continue;
1075 		IFADDR_FOREACH(ifa, ifp) {
1076 			if (ifa->ifa_addr->sa_family == af)
1077 				return ifa;
1078 		}
1079 	}
1080 	return NULL;
1081 }
1082 
1083 /*
1084  * Find an interface address specific to an interface best matching
1085  * a given address.
1086  */
1087 struct ifaddr *
1088 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1089 {
1090 	struct ifaddr *ifa;
1091 	const char *cp, *cp2, *cp3;
1092 	const char *cplim;
1093 	struct ifaddr *ifa_maybe = 0;
1094 	u_int af = addr->sa_family;
1095 
1096 	if (ifp->if_output == if_nulloutput)
1097 		return NULL;
1098 
1099 	if (af >= AF_MAX)
1100 		return NULL;
1101 
1102 	IFADDR_FOREACH(ifa, ifp) {
1103 		if (ifa->ifa_addr->sa_family != af)
1104 			continue;
1105 		ifa_maybe = ifa;
1106 		if (ifa->ifa_netmask == NULL) {
1107 			if (equal(addr, ifa->ifa_addr) ||
1108 			    (ifa->ifa_dstaddr &&
1109 			     equal(addr, ifa->ifa_dstaddr)))
1110 				return ifa;
1111 			continue;
1112 		}
1113 		cp = addr->sa_data;
1114 		cp2 = ifa->ifa_addr->sa_data;
1115 		cp3 = ifa->ifa_netmask->sa_data;
1116 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1117 		for (; cp3 < cplim; cp3++) {
1118 			if ((*cp++ ^ *cp2++) & *cp3)
1119 				break;
1120 		}
1121 		if (cp3 == cplim)
1122 			return ifa;
1123 	}
1124 	return ifa_maybe;
1125 }
1126 
1127 /*
1128  * Default action when installing a route with a Link Level gateway.
1129  * Lookup an appropriate real ifa to point to.
1130  * This should be moved to /sys/net/link.c eventually.
1131  */
1132 void
1133 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1134 {
1135 	struct ifaddr *ifa;
1136 	const struct sockaddr *dst;
1137 	struct ifnet *ifp;
1138 
1139 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
1140 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_getkey(rt)) == NULL))
1141 		return;
1142 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1143 		rt_replace_ifa(rt, ifa);
1144 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1145 			ifa->ifa_rtrequest(cmd, rt, info);
1146 	}
1147 }
1148 
1149 /*
1150  * Handle a change in the interface link state.
1151  */
1152 void
1153 if_link_state_change(struct ifnet *ifp, int link_state)
1154 {
1155 	if (ifp->if_link_state == link_state)
1156 		return;
1157 	ifp->if_link_state = link_state;
1158 	/* Notify that the link state has changed. */
1159 	rt_ifmsg(ifp);
1160 #if NCARP > 0
1161 	if (ifp->if_carp)
1162 		carp_carpdev_state(ifp);
1163 #endif
1164 }
1165 
1166 /*
1167  * Mark an interface down and notify protocols of
1168  * the transition.
1169  * NOTE: must be called at splsoftnet or equivalent.
1170  */
1171 void
1172 if_down(struct ifnet *ifp)
1173 {
1174 	struct ifaddr *ifa;
1175 
1176 	ifp->if_flags &= ~IFF_UP;
1177 	microtime(&ifp->if_lastchange);
1178 	IFADDR_FOREACH(ifa, ifp)
1179 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1180 	IFQ_PURGE(&ifp->if_snd);
1181 #if NCARP > 0
1182 	if (ifp->if_carp)
1183 		carp_carpdev_state(ifp);
1184 #endif
1185 	rt_ifmsg(ifp);
1186 }
1187 
1188 /*
1189  * Mark an interface up and notify protocols of
1190  * the transition.
1191  * NOTE: must be called at splsoftnet or equivalent.
1192  */
1193 void
1194 if_up(struct ifnet *ifp)
1195 {
1196 #ifdef notyet
1197 	struct ifaddr *ifa;
1198 #endif
1199 
1200 	ifp->if_flags |= IFF_UP;
1201 	microtime(&ifp->if_lastchange);
1202 #ifdef notyet
1203 	/* this has no effect on IP, and will kill all ISO connections XXX */
1204 	IFADDR_FOREACH(ifa, ifp)
1205 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1206 #endif
1207 #if NCARP > 0
1208 	if (ifp->if_carp)
1209 		carp_carpdev_state(ifp);
1210 #endif
1211 	rt_ifmsg(ifp);
1212 #ifdef INET6
1213 	in6_if_up(ifp);
1214 #endif
1215 }
1216 
1217 /*
1218  * Handle interface watchdog timer routines.  Called
1219  * from softclock, we decrement timers (if set) and
1220  * call the appropriate interface routine on expiration.
1221  */
1222 void
1223 if_slowtimo(void *arg)
1224 {
1225 	struct ifnet *ifp;
1226 	int s = splnet();
1227 
1228 	IFNET_FOREACH(ifp) {
1229 		if (ifp->if_timer == 0 || --ifp->if_timer)
1230 			continue;
1231 		if (ifp->if_watchdog != NULL)
1232 			(*ifp->if_watchdog)(ifp);
1233 	}
1234 	splx(s);
1235 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1236 }
1237 
1238 /*
1239  * Set/clear promiscuous mode on interface ifp based on the truth value
1240  * of pswitch.  The calls are reference counted so that only the first
1241  * "on" request actually has an effect, as does the final "off" request.
1242  * Results are undefined if the "off" and "on" requests are not matched.
1243  */
1244 int
1245 ifpromisc(struct ifnet *ifp, int pswitch)
1246 {
1247 	int pcount, ret;
1248 	short flags;
1249 	struct ifreq ifr;
1250 
1251 	pcount = ifp->if_pcount;
1252 	flags = ifp->if_flags;
1253 	if (pswitch) {
1254 		/*
1255 		 * Allow the device to be "placed" into promiscuous
1256 		 * mode even if it is not configured up.  It will
1257 		 * consult IFF_PROMISC when it is is brought up.
1258 		 */
1259 		if (ifp->if_pcount++ != 0)
1260 			return 0;
1261 		ifp->if_flags |= IFF_PROMISC;
1262 		if ((ifp->if_flags & IFF_UP) == 0)
1263 			return 0;
1264 	} else {
1265 		if (--ifp->if_pcount > 0)
1266 			return 0;
1267 		ifp->if_flags &= ~IFF_PROMISC;
1268 		/*
1269 		 * If the device is not configured up, we should not need to
1270 		 * turn off promiscuous mode (device should have turned it
1271 		 * off when interface went down; and will look at IFF_PROMISC
1272 		 * again next time interface comes up).
1273 		 */
1274 		if ((ifp->if_flags & IFF_UP) == 0)
1275 			return 0;
1276 	}
1277 	memset(&ifr, 0, sizeof(ifr));
1278 	ifr.ifr_flags = ifp->if_flags;
1279 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (void *) &ifr);
1280 	/* Restore interface state if not successful. */
1281 	if (ret != 0) {
1282 		ifp->if_pcount = pcount;
1283 		ifp->if_flags = flags;
1284 	}
1285 	return ret;
1286 }
1287 
1288 /*
1289  * Map interface name to
1290  * interface structure pointer.
1291  */
1292 struct ifnet *
1293 ifunit(const char *name)
1294 {
1295 	struct ifnet *ifp;
1296 	const char *cp = name;
1297 	u_int unit = 0;
1298 	u_int i;
1299 
1300 	/*
1301 	 * If the entire name is a number, treat it as an ifindex.
1302 	 */
1303 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1304 		unit = unit * 10 + (*cp - '0');
1305 	}
1306 
1307 	/*
1308 	 * If the number took all of the name, then it's a valid ifindex.
1309 	 */
1310 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1311 		if (unit >= if_indexlim)
1312 			return NULL;
1313 		ifp = ifindex2ifnet[unit];
1314 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1315 			return NULL;
1316 		return ifp;
1317 	}
1318 
1319 	IFNET_FOREACH(ifp) {
1320 		if (ifp->if_output == if_nulloutput)
1321 			continue;
1322 	 	if (strcmp(ifp->if_xname, name) == 0)
1323 			return ifp;
1324 	}
1325 	return NULL;
1326 }
1327 
1328 /*
1329  * Interface ioctls.
1330  */
1331 int
1332 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1333 {
1334 	struct ifnet *ifp;
1335 	struct ifreq *ifr;
1336 	struct ifcapreq *ifcr;
1337 	struct ifdatareq *ifdr;
1338 	int s, error = 0;
1339 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1340 	u_long ocmd = cmd;
1341 #endif
1342 	short oif_flags;
1343 #ifdef COMPAT_OIFREQ
1344 	struct ifreq ifrb;
1345 	struct oifreq *oifr = NULL;
1346 #endif
1347 
1348 	switch (cmd) {
1349 #ifdef COMPAT_OIFREQ
1350 	case OSIOCGIFCONF:
1351 	case OOSIOCGIFCONF:
1352 		return compat_ifconf(cmd, data);
1353 #endif
1354 	case SIOCGIFCONF:
1355 		return ifconf(cmd, data);
1356 	}
1357 
1358 #ifdef COMPAT_OIFREQ
1359 	cmd = compat_cvtcmd(cmd);
1360 	if (cmd != ocmd) {
1361 		oifr = data;
1362 		data = ifr = &ifrb;
1363 		ifreqo2n(oifr, ifr);
1364 	} else
1365 #endif
1366 		ifr = data;
1367 	ifcr = data;
1368 	ifdr = data;
1369 
1370 	ifp = ifunit(ifr->ifr_name);
1371 
1372 	switch (cmd) {
1373 	case SIOCIFCREATE:
1374 	case SIOCIFDESTROY:
1375 		if (l != NULL) {
1376 			error = kauth_authorize_network(l->l_cred,
1377 			    KAUTH_NETWORK_INTERFACE,
1378 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1379 			    (void *)cmd, NULL);
1380 			if (error != 0)
1381 				return error;
1382 		}
1383 		return (cmd == SIOCIFCREATE) ?
1384 			if_clone_create(ifr->ifr_name) :
1385 			if_clone_destroy(ifr->ifr_name);
1386 
1387 	case SIOCIFGCLONERS:
1388 		return if_clone_list((struct if_clonereq *)data);
1389 	}
1390 
1391 	if (ifp == NULL)
1392 		return ENXIO;
1393 
1394 	switch (cmd) {
1395 	case SIOCSIFFLAGS:
1396 	case SIOCSIFCAP:
1397 	case SIOCSIFMETRIC:
1398 	case SIOCZIFDATA:
1399 	case SIOCSIFMTU:
1400 	case SIOCSIFPHYADDR:
1401 	case SIOCDIFPHYADDR:
1402 #ifdef INET6
1403 	case SIOCSIFPHYADDR_IN6:
1404 #endif
1405 	case SIOCSLIFPHYADDR:
1406 	case SIOCADDMULTI:
1407 	case SIOCDELMULTI:
1408 	case SIOCSIFMEDIA:
1409 	case SIOCSDRVSPEC:
1410 	case SIOCG80211:
1411 	case SIOCS80211:
1412 	case SIOCS80211NWID:
1413 	case SIOCS80211NWKEY:
1414 	case SIOCS80211POWER:
1415 	case SIOCS80211BSSID:
1416 	case SIOCS80211CHANNEL:
1417 		if (l != NULL) {
1418 			error = kauth_authorize_network(l->l_cred,
1419 			    KAUTH_NETWORK_INTERFACE,
1420 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1421 			    (void *)cmd, NULL);
1422 			if (error != 0)
1423 				return error;
1424 		}
1425 	}
1426 
1427 	oif_flags = ifp->if_flags;
1428 	switch (cmd) {
1429 
1430 	case SIOCGIFFLAGS:
1431 		ifr->ifr_flags = ifp->if_flags;
1432 		break;
1433 
1434 	case SIOCGIFMETRIC:
1435 		ifr->ifr_metric = ifp->if_metric;
1436 		break;
1437 
1438 	case SIOCGIFMTU:
1439 		ifr->ifr_mtu = ifp->if_mtu;
1440 		break;
1441 
1442 	case SIOCGIFDLT:
1443 		ifr->ifr_dlt = ifp->if_dlt;
1444 		break;
1445 
1446 	case SIOCSIFFLAGS:
1447 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1448 			s = splnet();
1449 			if_down(ifp);
1450 			splx(s);
1451 		}
1452 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1453 			s = splnet();
1454 			if_up(ifp);
1455 			splx(s);
1456 		}
1457 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1458 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1459 		if (ifp->if_ioctl)
1460 			(void)(*ifp->if_ioctl)(ifp, cmd, data);
1461 		break;
1462 
1463 	case SIOCGIFCAP:
1464 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1465 		ifcr->ifcr_capenable = ifp->if_capenable;
1466 		break;
1467 
1468 	case SIOCSIFCAP:
1469 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1470 			return EINVAL;
1471 		if (ifp->if_ioctl == NULL)
1472 			return EOPNOTSUPP;
1473 
1474 		/* Must prevent race with packet reception here. */
1475 		s = splnet();
1476 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1477 			struct ifreq ifrq;
1478 
1479 			ifrq.ifr_flags = ifp->if_flags;
1480 			ifp->if_capenable = ifcr->ifcr_capenable;
1481 
1482 			/* Pre-compute the checksum flags mask. */
1483 			ifp->if_csum_flags_tx = 0;
1484 			ifp->if_csum_flags_rx = 0;
1485 			if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1486 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1487 			}
1488 			if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1489 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1490 			}
1491 
1492 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1493 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1494 			}
1495 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1496 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1497 			}
1498 
1499 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1500 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1501 			}
1502 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1503 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1504 			}
1505 
1506 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1507 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1508 			}
1509 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1510 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1511 			}
1512 
1513 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1514 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1515 			}
1516 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1517 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1518 			}
1519 
1520 			/*
1521 			 * Only kick the interface if it's up.  If it's
1522 			 * not up now, it will notice the cap enables
1523 			 * when it is brought up later.
1524 			 */
1525 			if (ifp->if_flags & IFF_UP)
1526 				(void)(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1527 				    (void *)&ifrq);
1528 		}
1529 		splx(s);
1530 		break;
1531 
1532 	case SIOCSIFMETRIC:
1533 		ifp->if_metric = ifr->ifr_metric;
1534 		break;
1535 
1536 	case SIOCGIFDATA:
1537 		ifdr->ifdr_data = ifp->if_data;
1538 		break;
1539 
1540 	case SIOCZIFDATA:
1541 		ifdr->ifdr_data = ifp->if_data;
1542 		/*
1543 		 * Assumes that the volatile counters that can be
1544 		 * zero'ed are at the end of if_data.
1545 		 */
1546 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1547 		    offsetof(struct if_data, ifi_ipackets));
1548 		break;
1549 
1550 	case SIOCSIFMTU:
1551 	{
1552 		u_long oldmtu = ifp->if_mtu;
1553 
1554 		if (ifp->if_ioctl == NULL)
1555 			return EOPNOTSUPP;
1556 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1557 
1558 		/*
1559 		 * If the link MTU changed, do network layer specific procedure.
1560 		 */
1561 		if (ifp->if_mtu != oldmtu) {
1562 #ifdef INET6
1563 			nd6_setmtu(ifp);
1564 #endif
1565 		}
1566 		break;
1567 	}
1568 	case SIOCSIFPHYADDR:
1569 	case SIOCDIFPHYADDR:
1570 #ifdef INET6
1571 	case SIOCSIFPHYADDR_IN6:
1572 #endif
1573 	case SIOCSLIFPHYADDR:
1574 	case SIOCADDMULTI:
1575 	case SIOCDELMULTI:
1576 	case SIOCSIFMEDIA:
1577 	case SIOCGIFPSRCADDR:
1578 	case SIOCGIFPDSTADDR:
1579 	case SIOCGLIFPHYADDR:
1580 	case SIOCGIFMEDIA:
1581 	case SIOCG80211:
1582 	case SIOCS80211:
1583 	case SIOCS80211NWID:
1584 	case SIOCS80211NWKEY:
1585 	case SIOCS80211POWER:
1586 	case SIOCS80211BSSID:
1587 	case SIOCS80211CHANNEL:
1588 		if (ifp->if_ioctl == NULL)
1589 			return EOPNOTSUPP;
1590 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1591 		break;
1592 
1593 	case SIOCSDRVSPEC:
1594 	default:
1595 		if (so->so_proto == NULL)
1596 			return EOPNOTSUPP;
1597 #ifdef COMPAT_OSOCK
1598 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1599 #else
1600 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1601 		    (struct mbuf *)cmd, (struct mbuf *)data,
1602 		    (struct mbuf *)ifp, l));
1603 #endif
1604 		break;
1605 	}
1606 
1607 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1608 #ifdef INET6
1609 		if ((ifp->if_flags & IFF_UP) != 0) {
1610 			s = splnet();
1611 			in6_if_up(ifp);
1612 			splx(s);
1613 		}
1614 #endif
1615 	}
1616 #ifdef COMPAT_OIFREQ
1617 	if (cmd != ocmd)
1618 		ifreqn2o(oifr, ifr);
1619 #endif
1620 
1621 	return error;
1622 }
1623 
1624 /*
1625  * Return interface configuration
1626  * of system.  List may be used
1627  * in later ioctl's (above) to get
1628  * other information.
1629  *
1630  * Each record is a struct ifreq.  Before the addition of
1631  * sockaddr_storage, the API rule was that sockaddr flavors that did
1632  * not fit would extend beyond the struct ifreq, with the next struct
1633  * ifreq starting sa_len beyond the struct sockaddr.  Because the
1634  * union in struct ifreq includes struct sockaddr_storage, every kind
1635  * of sockaddr must fit.  Thus, there are no longer any overlength
1636  * records.
1637  *
1638  * Records are added to the user buffer if they fit, and ifc_len is
1639  * adjusted to the length that was written.  Thus, the user is only
1640  * assured of getting the complete list if ifc_len on return is at
1641  * least sizeof(struct ifreq) less than it was on entry.
1642  *
1643  * If the user buffer pointer is NULL, this routine copies no data and
1644  * returns the amount of space that would be needed.
1645  *
1646  * Invariants:
1647  * ifrp points to the next part of the user's buffer to be used.  If
1648  * ifrp != NULL, space holds the number of bytes remaining that we may
1649  * write at ifrp.  Otherwise, space holds the number of bytes that
1650  * would have been written had there been adequate space.
1651  */
1652 /*ARGSUSED*/
1653 int
1654 ifconf(u_long cmd, void *data)
1655 {
1656 	struct ifconf *ifc = (struct ifconf *)data;
1657 	struct ifnet *ifp;
1658 	struct ifaddr *ifa;
1659 	struct ifreq ifr, *ifrp;
1660 	int space, error = 0;
1661 	const int sz = (int)sizeof(struct ifreq);
1662 
1663 	if ((ifrp = ifc->ifc_req) == NULL)
1664 		space = 0;
1665 	else
1666 		space = ifc->ifc_len;
1667 	IFNET_FOREACH(ifp) {
1668 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1669 		    sizeof(ifr.ifr_name));
1670 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1671 			return ENAMETOOLONG;
1672 		if (IFADDR_EMPTY(ifp)) {
1673 			/* Interface with no addresses - send zero sockaddr. */
1674 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1675 			if (ifrp != NULL)
1676 			{
1677 				if (space >= sz) {
1678 					error = copyout(&ifr, ifrp, sz);
1679 					if (error != 0)
1680 						return (error);
1681 					ifrp++; space -= sz;
1682 				}
1683 			}
1684 			else
1685 				space += sz;
1686 			continue;
1687 		}
1688 
1689 		IFADDR_FOREACH(ifa, ifp) {
1690 			struct sockaddr *sa = ifa->ifa_addr;
1691 			/* all sockaddrs must fit in sockaddr_storage */
1692 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
1693 
1694 			if (ifrp != NULL)
1695 			{
1696 				memcpy(&ifr.ifr_space, sa, sa->sa_len);
1697 				if (space >= sz) {
1698 					error = copyout(&ifr, ifrp, sz);
1699 					if (error != 0)
1700 						return (error);
1701 					ifrp++; space -= sz;
1702 				}
1703 			}
1704 			else
1705 				space += sz;
1706 		}
1707 	}
1708 	if (ifrp != NULL)
1709 	{
1710 		KASSERT(0 <= space && space <= ifc->ifc_len);
1711 		ifc->ifc_len -= space;
1712 	}
1713 	else
1714 	{
1715 		KASSERT(space >= 0);
1716 		ifc->ifc_len = space;
1717 	}
1718 	return (0);
1719 }
1720 
1721 int
1722 ifreq_setaddr(const u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
1723 {
1724 	uint8_t len;
1725 	u_long ncmd;
1726 	const uint8_t osockspace = sizeof(ifr->ifr_addr);
1727 	const uint8_t sockspace = sizeof(ifr->ifr_ifru.ifru_space);
1728 
1729 #ifdef INET6
1730 	if (cmd == SIOCGIFPSRCADDR_IN6 || cmd == SIOCGIFPDSTADDR_IN6)
1731 		len = MIN(sizeof(struct sockaddr_in6), sa->sa_len);
1732 	else
1733 #endif /* INET6 */
1734 	if ((ncmd = compat_cvtcmd(cmd)) != cmd)
1735 		len = MIN(osockspace, sa->sa_len);
1736 	else
1737 		len = MIN(sockspace, sa->sa_len);
1738 	if (len < sa->sa_len)
1739 		return EFBIG;
1740 	sockaddr_copy(&ifr->ifr_addr, len, sa);
1741 	return 0;
1742 }
1743 
1744 /*
1745  * Queue message on interface, and start output if interface
1746  * not yet active.
1747  */
1748 int
1749 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
1750     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1751 {
1752 	int len = m->m_pkthdr.len;
1753 	int mflags = m->m_flags;
1754 	int s = splnet();
1755 	int error;
1756 
1757 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1758 	if (error != 0)
1759 		goto out;
1760 	ifp->if_obytes += len;
1761 	if (mflags & M_MCAST)
1762 		ifp->if_omcasts++;
1763 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
1764 		(*ifp->if_start)(ifp);
1765 out:
1766 	splx(s);
1767 	return error;
1768 }
1769 
1770 /*
1771  * Queue message on interface, possibly using a second fast queue
1772  */
1773 int
1774 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
1775     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1776 {
1777 	int error = 0;
1778 
1779 	if (ifq != NULL
1780 #ifdef ALTQ
1781 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
1782 #endif
1783 	    ) {
1784 		if (IF_QFULL(ifq)) {
1785 			IF_DROP(&ifp->if_snd);
1786 			m_freem(m);
1787 			if (error == 0)
1788 				error = ENOBUFS;
1789 		} else
1790 			IF_ENQUEUE(ifq, m);
1791 	} else
1792 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1793 	if (error != 0) {
1794 		++ifp->if_oerrors;
1795 		return error;
1796 	}
1797 	return 0;
1798 }
1799 
1800 
1801 #if defined(INET) || defined(INET6)
1802 static void
1803 sysctl_net_ifq_setup(struct sysctllog **clog,
1804 		     int pf, const char *pfname,
1805 		     int ipn, const char *ipname,
1806 		     int qid, struct ifqueue *ifq)
1807 {
1808 
1809 	sysctl_createv(clog, 0, NULL, NULL,
1810 		       CTLFLAG_PERMANENT,
1811 		       CTLTYPE_NODE, "net", NULL,
1812 		       NULL, 0, NULL, 0,
1813 		       CTL_NET, CTL_EOL);
1814 	sysctl_createv(clog, 0, NULL, NULL,
1815 		       CTLFLAG_PERMANENT,
1816 		       CTLTYPE_NODE, pfname, NULL,
1817 		       NULL, 0, NULL, 0,
1818 		       CTL_NET, pf, CTL_EOL);
1819 	sysctl_createv(clog, 0, NULL, NULL,
1820 		       CTLFLAG_PERMANENT,
1821 		       CTLTYPE_NODE, ipname, NULL,
1822 		       NULL, 0, NULL, 0,
1823 		       CTL_NET, pf, ipn, CTL_EOL);
1824 	sysctl_createv(clog, 0, NULL, NULL,
1825 		       CTLFLAG_PERMANENT,
1826 		       CTLTYPE_NODE, "ifq",
1827 		       SYSCTL_DESCR("Protocol input queue controls"),
1828 		       NULL, 0, NULL, 0,
1829 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1830 
1831 	sysctl_createv(clog, 0, NULL, NULL,
1832 		       CTLFLAG_PERMANENT,
1833 		       CTLTYPE_INT, "len",
1834 		       SYSCTL_DESCR("Current input queue length"),
1835 		       NULL, 0, &ifq->ifq_len, 0,
1836 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1837 	sysctl_createv(clog, 0, NULL, NULL,
1838 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1839 		       CTLTYPE_INT, "maxlen",
1840 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1841 		       NULL, 0, &ifq->ifq_maxlen, 0,
1842 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1843 #ifdef notyet
1844 	sysctl_createv(clog, 0, NULL, NULL,
1845 		       CTLFLAG_PERMANENT,
1846 		       CTLTYPE_INT, "peak",
1847 		       SYSCTL_DESCR("Highest input queue length"),
1848 		       NULL, 0, &ifq->ifq_peak, 0,
1849 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1850 #endif
1851 	sysctl_createv(clog, 0, NULL, NULL,
1852 		       CTLFLAG_PERMANENT,
1853 		       CTLTYPE_INT, "drops",
1854 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1855 		       NULL, 0, &ifq->ifq_drops, 0,
1856 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1857 }
1858 
1859 #ifdef INET
1860 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1861 	     "sysctl net.inet.ip.ifq subtree setup")
1862 {
1863 	extern struct ifqueue ipintrq;
1864 
1865 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1866 			     IPCTL_IFQ, &ipintrq);
1867 }
1868 #endif /* INET */
1869 
1870 #ifdef INET6
1871 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1872 	     "sysctl net.inet6.ip6.ifq subtree setup")
1873 {
1874 	extern struct ifqueue ip6intrq;
1875 
1876 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1877 			     IPV6CTL_IFQ, &ip6intrq);
1878 }
1879 #endif /* INET6 */
1880 #endif /* INET || INET6 */
1881