xref: /netbsd-src/sys/net/if.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: if.c,v 1.178 2006/11/20 04:09:25 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 1.178 2006/11/20 04:09:25 dyoung Exp $");
101 
102 #include "opt_inet.h"
103 
104 #include "opt_compat_linux.h"
105 #include "opt_compat_svr4.h"
106 #include "opt_compat_ultrix.h"
107 #include "opt_compat_43.h"
108 #include "opt_atalk.h"
109 #include "opt_natm.h"
110 #include "opt_pfil_hooks.h"
111 
112 #include <sys/param.h>
113 #include <sys/mbuf.h>
114 #include <sys/systm.h>
115 #include <sys/callout.h>
116 #include <sys/proc.h>
117 #include <sys/socket.h>
118 #include <sys/socketvar.h>
119 #include <sys/domain.h>
120 #include <sys/protosw.h>
121 #include <sys/kernel.h>
122 #include <sys/ioctl.h>
123 #include <sys/sysctl.h>
124 #include <sys/syslog.h>
125 #include <sys/kauth.h>
126 
127 #include <net/if.h>
128 #include <net/if_dl.h>
129 #include <net/if_ether.h>
130 #include <net/if_media.h>
131 #include <net80211/ieee80211.h>
132 #include <net80211/ieee80211_ioctl.h>
133 #include <net/if_types.h>
134 #include <net/radix.h>
135 #include <net/route.h>
136 #include <net/netisr.h>
137 #ifdef NETATALK
138 #include <netatalk/at_extern.h>
139 #include <netatalk/at.h>
140 #endif
141 #include <net/pfil.h>
142 
143 #ifdef INET6
144 #include <netinet/in.h>
145 #include <netinet6/in6_var.h>
146 #include <netinet6/nd6.h>
147 #endif
148 
149 #include "carp.h"
150 #if NCARP > 0
151 #include <netinet/ip_carp.h>
152 #endif
153 
154 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX) || defined(LKM)
155 #define COMPAT_OSOCK
156 #include <compat/sys/socket.h>
157 #endif
158 
159 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
160 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
161 
162 int	ifqmaxlen = IFQ_MAXLEN;
163 struct	callout if_slowtimo_ch;
164 
165 int netisr;			/* scheduling bits for network */
166 
167 static int	if_rt_walktree(struct radix_node *, void *);
168 
169 static struct if_clone *if_clone_lookup(const char *, int *);
170 static int	if_clone_list(struct if_clonereq *);
171 
172 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
173 static int if_cloners_count;
174 
175 #ifdef PFIL_HOOKS
176 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
177 #endif
178 
179 static void if_detach_queues(struct ifnet *, struct ifqueue *);
180 
181 /*
182  * Network interface utility routines.
183  *
184  * Routines with ifa_ifwith* names take sockaddr *'s as
185  * parameters.
186  */
187 void
188 ifinit(void)
189 {
190 
191 	callout_init(&if_slowtimo_ch);
192 	if_slowtimo(NULL);
193 #ifdef PFIL_HOOKS
194 	if_pfil.ph_type = PFIL_TYPE_IFNET;
195 	if_pfil.ph_ifnet = NULL;
196 	if (pfil_head_register(&if_pfil) != 0)
197 		printf("WARNING: unable to register pfil hook\n");
198 #endif
199 }
200 
201 /*
202  * Null routines used while an interface is going away.  These routines
203  * just return an error.
204  */
205 
206 int
207 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
208     struct sockaddr *so, struct rtentry *rt)
209 {
210 
211 	return (ENXIO);
212 }
213 
214 void
215 if_nullinput(struct ifnet *ifp, struct mbuf *m)
216 {
217 
218 	/* Nothing. */
219 }
220 
221 void
222 if_nullstart(struct ifnet *ifp)
223 {
224 
225 	/* Nothing. */
226 }
227 
228 int
229 if_nullioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
230 {
231 
232 	return (ENXIO);
233 }
234 
235 int
236 if_nullinit(struct ifnet *ifp)
237 {
238 
239 	return (ENXIO);
240 }
241 
242 void
243 if_nullstop(struct ifnet *ifp, int disable)
244 {
245 
246 	/* Nothing. */
247 }
248 
249 void
250 if_nullwatchdog(struct ifnet *ifp)
251 {
252 
253 	/* Nothing. */
254 }
255 
256 void
257 if_nulldrain(struct ifnet *ifp)
258 {
259 
260 	/* Nothing. */
261 }
262 
263 static u_int if_index = 1;
264 struct ifnet_head ifnet;
265 size_t if_indexlim = 0;
266 struct ifaddr **ifnet_addrs = NULL;
267 struct ifnet **ifindex2ifnet = NULL;
268 struct ifnet *lo0ifp;
269 
270 /*
271  * Allocate the link level name for the specified interface.  This
272  * is an attachment helper.  It must be called after ifp->if_addrlen
273  * is initialized, which may not be the case when if_attach() is
274  * called.
275  */
276 void
277 if_alloc_sadl(struct ifnet *ifp)
278 {
279 	unsigned socksize, ifasize;
280 	int namelen, masklen;
281 	struct sockaddr_dl *sdl;
282 	struct ifaddr *ifa;
283 
284 	/*
285 	 * If the interface already has a link name, release it
286 	 * now.  This is useful for interfaces that can change
287 	 * link types, and thus switch link names often.
288 	 */
289 	if (ifp->if_sadl != NULL)
290 		if_free_sadl(ifp);
291 
292 	namelen = strlen(ifp->if_xname);
293 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
294 	socksize = masklen + ifp->if_addrlen;
295 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
296 	if (socksize < sizeof(*sdl))
297 		socksize = sizeof(*sdl);
298 	socksize = ROUNDUP(socksize);
299 	ifasize = sizeof(*ifa) + 2 * socksize;
300 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
301 	memset((caddr_t)ifa, 0, ifasize);
302 	sdl = (struct sockaddr_dl *)(ifa + 1);
303 	sdl->sdl_len = socksize;
304 	sdl->sdl_family = AF_LINK;
305 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
306 	sdl->sdl_nlen = namelen;
307 	sdl->sdl_alen = ifp->if_addrlen;
308 	sdl->sdl_index = ifp->if_index;
309 	sdl->sdl_type = ifp->if_type;
310 	ifnet_addrs[ifp->if_index] = ifa;
311 	IFAREF(ifa);
312 	ifa->ifa_ifp = ifp;
313 	ifa->ifa_rtrequest = link_rtrequest;
314 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
315 	IFAREF(ifa);
316 	ifa->ifa_addr = (struct sockaddr *)sdl;
317 	ifp->if_sadl = sdl;
318 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
319 	ifa->ifa_netmask = (struct sockaddr *)sdl;
320 	sdl->sdl_len = masklen;
321 	while (namelen != 0)
322 		sdl->sdl_data[--namelen] = 0xff;
323 }
324 
325 /*
326  * Free the link level name for the specified interface.  This is
327  * a detach helper.  This is called from if_detach() or from
328  * link layer type specific detach functions.
329  */
330 void
331 if_free_sadl(struct ifnet *ifp)
332 {
333 	struct ifaddr *ifa;
334 	int s;
335 
336 	ifa = ifnet_addrs[ifp->if_index];
337 	if (ifa == NULL) {
338 		KASSERT(ifp->if_sadl == NULL);
339 		return;
340 	}
341 
342 	KASSERT(ifp->if_sadl != NULL);
343 
344 	s = splnet();
345 	rtinit(ifa, RTM_DELETE, 0);
346 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
347 	IFAFREE(ifa);
348 
349 	ifp->if_sadl = NULL;
350 
351 	ifnet_addrs[ifp->if_index] = NULL;
352 	IFAFREE(ifa);
353 	splx(s);
354 }
355 
356 /*
357  * Attach an interface to the
358  * list of "active" interfaces.
359  */
360 void
361 if_attach(struct ifnet *ifp)
362 {
363 	int indexlim = 0;
364 
365 	if (if_indexlim == 0) {
366 		TAILQ_INIT(&ifnet);
367 		if_indexlim = 8;
368 	}
369 	TAILQ_INIT(&ifp->if_addrlist);
370 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
371 	ifp->if_index = if_index;
372 	if (ifindex2ifnet == 0)
373 		if_index++;
374 	else
375 		while (ifp->if_index < if_indexlim &&
376 		    ifindex2ifnet[ifp->if_index] != NULL) {
377 			++if_index;
378 			if (if_index == 0)
379 				if_index = 1;
380 			/*
381 			 * If we hit USHRT_MAX, we skip back to 0 since
382 			 * there are a number of places where the value
383 			 * of if_index or if_index itself is compared
384 			 * to or stored in an unsigned short.  By
385 			 * jumping back, we won't botch those assignments
386 			 * or comparisons.
387 			 */
388 			else if (if_index == USHRT_MAX) {
389 				/*
390 				 * However, if we have to jump back to
391 				 * zero *twice* without finding an empty
392 				 * slot in ifindex2ifnet[], then there
393 				 * there are too many (>65535) interfaces.
394 				 */
395 				if (indexlim++)
396 					panic("too many interfaces");
397 				else
398 					if_index = 1;
399 			}
400 			ifp->if_index = if_index;
401 		}
402 
403 	/*
404 	 * We have some arrays that should be indexed by if_index.
405 	 * since if_index will grow dynamically, they should grow too.
406 	 *	struct ifadd **ifnet_addrs
407 	 *	struct ifnet **ifindex2ifnet
408 	 */
409 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
410 	    ifp->if_index >= if_indexlim) {
411 		size_t m, n, oldlim;
412 		caddr_t q;
413 
414 		oldlim = if_indexlim;
415 		while (ifp->if_index >= if_indexlim)
416 			if_indexlim <<= 1;
417 
418 		/* grow ifnet_addrs */
419 		m = oldlim * sizeof(struct ifaddr *);
420 		n = if_indexlim * sizeof(struct ifaddr *);
421 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
422 		memset(q, 0, n);
423 		if (ifnet_addrs) {
424 			bcopy((caddr_t)ifnet_addrs, q, m);
425 			free((caddr_t)ifnet_addrs, M_IFADDR);
426 		}
427 		ifnet_addrs = (struct ifaddr **)q;
428 
429 		/* grow ifindex2ifnet */
430 		m = oldlim * sizeof(struct ifnet *);
431 		n = if_indexlim * sizeof(struct ifnet *);
432 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
433 		memset(q, 0, n);
434 		if (ifindex2ifnet) {
435 			bcopy((caddr_t)ifindex2ifnet, q, m);
436 			free((caddr_t)ifindex2ifnet, M_IFADDR);
437 		}
438 		ifindex2ifnet = (struct ifnet **)q;
439 	}
440 
441 	ifindex2ifnet[ifp->if_index] = ifp;
442 
443 	/*
444 	 * Link level name is allocated later by a separate call to
445 	 * if_alloc_sadl().
446 	 */
447 
448 	if (ifp->if_snd.ifq_maxlen == 0)
449 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
450 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
451 
452 	ifp->if_link_state = LINK_STATE_UNKNOWN;
453 
454 	ifp->if_capenable = 0;
455 	ifp->if_csum_flags_tx = 0;
456 	ifp->if_csum_flags_rx = 0;
457 
458 #ifdef ALTQ
459 	ifp->if_snd.altq_type = 0;
460 	ifp->if_snd.altq_disc = NULL;
461 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
462 	ifp->if_snd.altq_tbr  = NULL;
463 	ifp->if_snd.altq_ifp  = ifp;
464 #endif
465 
466 #ifdef PFIL_HOOKS
467 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
468 	ifp->if_pfil.ph_ifnet = ifp;
469 	if (pfil_head_register(&ifp->if_pfil) != 0)
470 		printf("%s: WARNING: unable to register pfil hook\n",
471 		    ifp->if_xname);
472 	(void)pfil_run_hooks(&if_pfil,
473 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
474 #endif
475 
476 	if (!STAILQ_EMPTY(&domains))
477 		if_attachdomain1(ifp);
478 
479 	/* Announce the interface. */
480 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
481 }
482 
483 void
484 if_attachdomain(void)
485 {
486 	struct ifnet *ifp;
487 	int s;
488 
489 	s = splnet();
490 	TAILQ_FOREACH(ifp, &ifnet, if_list)
491 		if_attachdomain1(ifp);
492 	splx(s);
493 }
494 
495 void
496 if_attachdomain1(struct ifnet *ifp)
497 {
498 	struct domain *dp;
499 	int s;
500 
501 	s = splnet();
502 
503 	/* address family dependent data region */
504 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
505 	DOMAIN_FOREACH(dp) {
506 		if (dp->dom_ifattach)
507 			ifp->if_afdata[dp->dom_family] =
508 			    (*dp->dom_ifattach)(ifp);
509 	}
510 
511 	splx(s);
512 }
513 
514 /*
515  * Deactivate an interface.  This points all of the procedure
516  * handles at error stubs.  May be called from interrupt context.
517  */
518 void
519 if_deactivate(struct ifnet *ifp)
520 {
521 	int s;
522 
523 	s = splnet();
524 
525 	ifp->if_output	 = if_nulloutput;
526 	ifp->if_input	 = if_nullinput;
527 	ifp->if_start	 = if_nullstart;
528 	ifp->if_ioctl	 = if_nullioctl;
529 	ifp->if_init	 = if_nullinit;
530 	ifp->if_stop	 = if_nullstop;
531 	ifp->if_watchdog = if_nullwatchdog;
532 	ifp->if_drain	 = if_nulldrain;
533 
534 	/* No more packets may be enqueued. */
535 	ifp->if_snd.ifq_maxlen = 0;
536 
537 	splx(s);
538 }
539 
540 /*
541  * Detach an interface from the list of "active" interfaces,
542  * freeing any resources as we go along.
543  *
544  * NOTE: This routine must be called with a valid thread context,
545  * as it may block.
546  */
547 void
548 if_detach(struct ifnet *ifp)
549 {
550 	struct socket so;
551 	struct ifaddr *ifa;
552 #ifdef IFAREF_DEBUG
553 	struct ifaddr *last_ifa = NULL;
554 #endif
555 	struct domain *dp;
556 	const struct protosw *pr;
557 	struct radix_node_head *rnh;
558 	int s, i, family, purged;
559 
560 	/*
561 	 * XXX It's kind of lame that we have to have the
562 	 * XXX socket structure...
563 	 */
564 	memset(&so, 0, sizeof(so));
565 
566 	s = splnet();
567 
568 	/*
569 	 * Do an if_down() to give protocols a chance to do something.
570 	 */
571 	if_down(ifp);
572 
573 #ifdef ALTQ
574 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
575 		altq_disable(&ifp->if_snd);
576 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
577 		altq_detach(&ifp->if_snd);
578 #endif
579 
580 
581 #if NCARP > 0
582 	/* Remove the interface from any carp group it is a part of.  */
583 	if (ifp->if_carp && ifp->if_type != IFT_CARP)
584 		carp_ifdetach(ifp);
585 #endif
586 
587 #ifdef PFIL_HOOKS
588 	(void)pfil_run_hooks(&if_pfil,
589 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
590 	(void)pfil_head_unregister(&ifp->if_pfil);
591 #endif
592 
593 	/*
594 	 * Rip all the addresses off the interface.  This should make
595 	 * all of the routes go away.
596 	 *
597 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
598 	 * from the list, including our "cursor", ifa.  For safety,
599 	 * and to honor the TAILQ abstraction, I just restart the
600 	 * loop after each removal.  Note that the loop will exit
601 	 * when all of the remaining ifaddrs belong to the AF_LINK
602 	 * family.  I am counting on the historical fact that at
603 	 * least one pr_usrreq in each address domain removes at
604 	 * least one ifaddr.
605 	 */
606 again:
607 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
608 		family = ifa->ifa_addr->sa_family;
609 #ifdef IFAREF_DEBUG
610 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
611 		    ifa, family, ifa->ifa_refcnt);
612 		if (last_ifa != NULL && ifa == last_ifa)
613 			panic("if_detach: loop detected");
614 		last_ifa = ifa;
615 #endif
616 		if (family == AF_LINK)
617 			continue;
618 		dp = pffinddomain(family);
619 #ifdef DIAGNOSTIC
620 		if (dp == NULL)
621 			panic("if_detach: no domain for AF %d",
622 			    family);
623 #endif
624 		/*
625 		 * XXX These PURGEIF calls are redundant with the
626 		 * purge-all-families calls below, but are left in for
627 		 * now both to make a smaller change, and to avoid
628 		 * unplanned interactions with clearing of
629 		 * ifp->if_addrlist.
630 		 */
631 		purged = 0;
632 		for (pr = dp->dom_protosw;
633 		     pr < dp->dom_protoswNPROTOSW; pr++) {
634 			so.so_proto = pr;
635 			if (pr->pr_usrreq != NULL) {
636 				(void) (*pr->pr_usrreq)(&so,
637 				    PRU_PURGEIF, NULL, NULL,
638 				    (struct mbuf *) ifp, curlwp);
639 				purged = 1;
640 			}
641 		}
642 		if (purged == 0) {
643 			/*
644 			 * XXX What's really the best thing to do
645 			 * XXX here?  --thorpej@NetBSD.org
646 			 */
647 			printf("if_detach: WARNING: AF %d not purged\n",
648 			    family);
649 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
650 		}
651 		goto again;
652 	}
653 
654 	if_free_sadl(ifp);
655 
656 	/* Walk the routing table looking for straglers. */
657 	for (i = 0; i <= AF_MAX; i++) {
658 		if ((rnh = rt_tables[i]) != NULL)
659 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
660 	}
661 
662 	DOMAIN_FOREACH(dp) {
663 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
664 			(*dp->dom_ifdetach)(ifp,
665 			    ifp->if_afdata[dp->dom_family]);
666 
667 		/*
668 		 * One would expect multicast memberships (INET and
669 		 * INET6) on UDP sockets to be purged by the PURGEIF
670 		 * calls above, but if all addresses were removed from
671 		 * the interface prior to destruction, the calls will
672 		 * not be made (e.g. ppp, for which pppd(8) generally
673 		 * removes addresses before destroying the interface).
674 		 * Because there is no invariant that multicast
675 		 * memberships only exist for interfaces with IPv4
676 		 * addresses, we must call PURGEIF regardless of
677 		 * addresses.  (Protocols which might store ifnet
678 		 * pointers are marked with PR_PURGEIF.)
679 		 */
680 		for (pr = dp->dom_protosw;
681 		     pr < dp->dom_protoswNPROTOSW; pr++) {
682 			so.so_proto = pr;
683 			if (pr->pr_usrreq != NULL &&
684 			    pr->pr_flags & PR_PURGEIF)
685 				(void) (*pr->pr_usrreq)(&so,
686 				    PRU_PURGEIF, NULL, NULL,
687 				    (struct mbuf *) ifp, curlwp);
688 		}
689 	}
690 
691 	/* Announce that the interface is gone. */
692 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
693 
694 	ifindex2ifnet[ifp->if_index] = NULL;
695 
696 	TAILQ_REMOVE(&ifnet, ifp, if_list);
697 
698 	/*
699 	 * remove packets that came from ifp, from software interrupt queues.
700 	 */
701 	DOMAIN_FOREACH(dp) {
702 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
703 			if (dp->dom_ifqueues[i] == NULL)
704 				break;
705 			if_detach_queues(ifp, dp->dom_ifqueues[i]);
706 		}
707 	}
708 
709 	splx(s);
710 }
711 
712 static void
713 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
714 {
715 	struct mbuf *m, *prev, *next;
716 
717 	prev = NULL;
718 	for (m = q->ifq_head; m; m = next) {
719 		next = m->m_nextpkt;
720 #ifdef DIAGNOSTIC
721 		if ((m->m_flags & M_PKTHDR) == 0) {
722 			prev = m;
723 			continue;
724 		}
725 #endif
726 		if (m->m_pkthdr.rcvif != ifp) {
727 			prev = m;
728 			continue;
729 		}
730 
731 		if (prev)
732 			prev->m_nextpkt = m->m_nextpkt;
733 		else
734 			q->ifq_head = m->m_nextpkt;
735 		if (q->ifq_tail == m)
736 			q->ifq_tail = prev;
737 		q->ifq_len--;
738 
739 		m->m_nextpkt = NULL;
740 		m_freem(m);
741 		IF_DROP(q);
742 	}
743 }
744 
745 /*
746  * Callback for a radix tree walk to delete all references to an
747  * ifnet.
748  */
749 static int
750 if_rt_walktree(struct radix_node *rn, void *v)
751 {
752 	struct ifnet *ifp = (struct ifnet *)v;
753 	struct rtentry *rt = (struct rtentry *)rn;
754 	int error;
755 
756 	if (rt->rt_ifp == ifp) {
757 		/* Delete the entry. */
758 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
759 		    rt_mask(rt), rt->rt_flags, NULL);
760 		if (error)
761 			printf("%s: warning: unable to delete rtentry @ %p, "
762 			    "error = %d\n", ifp->if_xname, rt, error);
763 	}
764 	return (0);
765 }
766 
767 /*
768  * Create a clone network interface.
769  */
770 int
771 if_clone_create(const char *name)
772 {
773 	struct if_clone *ifc;
774 	int unit;
775 
776 	ifc = if_clone_lookup(name, &unit);
777 	if (ifc == NULL)
778 		return (EINVAL);
779 
780 	if (ifunit(name) != NULL)
781 		return (EEXIST);
782 
783 	return ((*ifc->ifc_create)(ifc, unit));
784 }
785 
786 /*
787  * Destroy a clone network interface.
788  */
789 int
790 if_clone_destroy(const char *name)
791 {
792 	struct if_clone *ifc;
793 	struct ifnet *ifp;
794 
795 	ifc = if_clone_lookup(name, NULL);
796 	if (ifc == NULL)
797 		return (EINVAL);
798 
799 	ifp = ifunit(name);
800 	if (ifp == NULL)
801 		return (ENXIO);
802 
803 	if (ifc->ifc_destroy == NULL)
804 		return (EOPNOTSUPP);
805 
806 	return ((*ifc->ifc_destroy)(ifp));
807 }
808 
809 /*
810  * Look up a network interface cloner.
811  */
812 static struct if_clone *
813 if_clone_lookup(const char *name, int *unitp)
814 {
815 	struct if_clone *ifc;
816 	const char *cp;
817 	int unit;
818 
819 	/* separate interface name from unit */
820 	for (cp = name;
821 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
822 	    cp++)
823 		continue;
824 
825 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
826 		return (NULL);	/* No name or unit number */
827 
828 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
829 		if (strlen(ifc->ifc_name) == cp - name &&
830 		    !strncmp(name, ifc->ifc_name, cp - name))
831 			break;
832 	}
833 
834 	if (ifc == NULL)
835 		return (NULL);
836 
837 	unit = 0;
838 	while (cp - name < IFNAMSIZ && *cp) {
839 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
840 			/* Bogus unit number. */
841 			return (NULL);
842 		}
843 		unit = (unit * 10) + (*cp++ - '0');
844 	}
845 
846 	if (unitp != NULL)
847 		*unitp = unit;
848 	return (ifc);
849 }
850 
851 /*
852  * Register a network interface cloner.
853  */
854 void
855 if_clone_attach(struct if_clone *ifc)
856 {
857 
858 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
859 	if_cloners_count++;
860 }
861 
862 /*
863  * Unregister a network interface cloner.
864  */
865 void
866 if_clone_detach(struct if_clone *ifc)
867 {
868 
869 	LIST_REMOVE(ifc, ifc_list);
870 	if_cloners_count--;
871 }
872 
873 /*
874  * Provide list of interface cloners to userspace.
875  */
876 static int
877 if_clone_list(struct if_clonereq *ifcr)
878 {
879 	char outbuf[IFNAMSIZ], *dst;
880 	struct if_clone *ifc;
881 	int count, error = 0;
882 
883 	ifcr->ifcr_total = if_cloners_count;
884 	if ((dst = ifcr->ifcr_buffer) == NULL) {
885 		/* Just asking how many there are. */
886 		return (0);
887 	}
888 
889 	if (ifcr->ifcr_count < 0)
890 		return (EINVAL);
891 
892 	count = (if_cloners_count < ifcr->ifcr_count) ?
893 	    if_cloners_count : ifcr->ifcr_count;
894 
895 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
896 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
897 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
898 		if (outbuf[sizeof(outbuf) - 1] != '\0')
899 			return ENAMETOOLONG;
900 		error = copyout(outbuf, dst, sizeof(outbuf));
901 		if (error)
902 			break;
903 	}
904 
905 	return (error);
906 }
907 
908 /*
909  * Locate an interface based on a complete address.
910  */
911 /*ARGSUSED*/
912 struct ifaddr *
913 ifa_ifwithaddr(const struct sockaddr *addr)
914 {
915 	struct ifnet *ifp;
916 	struct ifaddr *ifa;
917 
918 #define	equal(a1, a2) \
919   (bcmp((a1), (a2), ((const struct sockaddr *)(a1))->sa_len) == 0)
920 
921 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
922 		if (ifp->if_output == if_nulloutput)
923 			continue;
924 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
925 			if (ifa->ifa_addr->sa_family != addr->sa_family)
926 				continue;
927 			if (equal(addr, ifa->ifa_addr))
928 				return (ifa);
929 			if ((ifp->if_flags & IFF_BROADCAST) &&
930 			    ifa->ifa_broadaddr &&
931 			    /* IP6 doesn't have broadcast */
932 			    ifa->ifa_broadaddr->sa_len != 0 &&
933 			    equal(ifa->ifa_broadaddr, addr))
934 				return (ifa);
935 		}
936 	}
937 	return (NULL);
938 }
939 
940 /*
941  * Locate the point to point interface with a given destination address.
942  */
943 /*ARGSUSED*/
944 struct ifaddr *
945 ifa_ifwithdstaddr(const struct sockaddr *addr)
946 {
947 	struct ifnet *ifp;
948 	struct ifaddr *ifa;
949 
950 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
951 		if (ifp->if_output == if_nulloutput)
952 			continue;
953 		if (ifp->if_flags & IFF_POINTOPOINT) {
954 			TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
955 				if (ifa->ifa_addr->sa_family !=
956 				      addr->sa_family ||
957 				    ifa->ifa_dstaddr == NULL)
958 					continue;
959 				if (equal(addr, ifa->ifa_dstaddr))
960 					return (ifa);
961 			}
962 		}
963 	}
964 	return (NULL);
965 }
966 
967 /*
968  * Find an interface on a specific network.  If many, choice
969  * is most specific found.
970  */
971 struct ifaddr *
972 ifa_ifwithnet(const struct sockaddr *addr)
973 {
974 	struct ifnet *ifp;
975 	struct ifaddr *ifa;
976 	const struct sockaddr_dl *sdl;
977 	struct ifaddr *ifa_maybe = 0;
978 	u_int af = addr->sa_family;
979 	const char *addr_data = addr->sa_data, *cplim;
980 
981 	if (af == AF_LINK) {
982 		sdl = (const struct sockaddr_dl *)addr;
983 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
984 		    ifindex2ifnet[sdl->sdl_index] &&
985 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
986 			return (ifnet_addrs[sdl->sdl_index]);
987 	}
988 #ifdef NETATALK
989 	if (af == AF_APPLETALK) {
990 		const struct sockaddr_at *sat, *sat2;
991 		sat = (const struct sockaddr_at *)addr;
992 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
993 			if (ifp->if_output == if_nulloutput)
994 				continue;
995 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
996 			if (ifa == NULL)
997 				continue;
998 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
999 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1000 				return (ifa); /* exact match */
1001 			if (ifa_maybe == NULL) {
1002 				/* else keep the if with the right range */
1003 				ifa_maybe = ifa;
1004 			}
1005 		}
1006 		return (ifa_maybe);
1007 	}
1008 #endif
1009 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1010 		if (ifp->if_output == if_nulloutput)
1011 			continue;
1012 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1013 			const char *cp, *cp2, *cp3;
1014 
1015 			if (ifa->ifa_addr->sa_family != af ||
1016 			    ifa->ifa_netmask == 0)
1017  next:				continue;
1018 			cp = addr_data;
1019 			cp2 = ifa->ifa_addr->sa_data;
1020 			cp3 = ifa->ifa_netmask->sa_data;
1021 			cplim = (const char *)ifa->ifa_netmask +
1022 			    ifa->ifa_netmask->sa_len;
1023 			while (cp3 < cplim) {
1024 				if ((*cp++ ^ *cp2++) & *cp3++) {
1025 					/* want to continue for() loop */
1026 					goto next;
1027 				}
1028 			}
1029 			if (ifa_maybe == 0 ||
1030 			    rn_refines((caddr_t)ifa->ifa_netmask,
1031 			    (caddr_t)ifa_maybe->ifa_netmask))
1032 				ifa_maybe = ifa;
1033 		}
1034 	}
1035 	return (ifa_maybe);
1036 }
1037 
1038 /*
1039  * Find the interface of the addresss.
1040  */
1041 struct ifaddr *
1042 ifa_ifwithladdr(const struct sockaddr *addr)
1043 {
1044 	struct ifaddr *ia;
1045 
1046 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1047 	    (ia = ifa_ifwithnet(addr)))
1048 		return (ia);
1049 	return (NULL);
1050 }
1051 
1052 /*
1053  * Find an interface using a specific address family
1054  */
1055 struct ifaddr *
1056 ifa_ifwithaf(int af)
1057 {
1058 	struct ifnet *ifp;
1059 	struct ifaddr *ifa;
1060 
1061 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1062 		if (ifp->if_output == if_nulloutput)
1063 			continue;
1064 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1065 			if (ifa->ifa_addr->sa_family == af)
1066 				return ifa;
1067 		}
1068 	}
1069 	return NULL;
1070 }
1071 
1072 /*
1073  * Find an interface address specific to an interface best matching
1074  * a given address.
1075  */
1076 struct ifaddr *
1077 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1078 {
1079 	struct ifaddr *ifa;
1080 	const char *cp, *cp2, *cp3;
1081 	const char *cplim;
1082 	struct ifaddr *ifa_maybe = 0;
1083 	u_int af = addr->sa_family;
1084 
1085 	if (ifp->if_output == if_nulloutput)
1086 		return (NULL);
1087 
1088 	if (af >= AF_MAX)
1089 		return (NULL);
1090 
1091 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1092 		if (ifa->ifa_addr->sa_family != af)
1093 			continue;
1094 		ifa_maybe = ifa;
1095 		if (ifa->ifa_netmask == 0) {
1096 			if (equal(addr, ifa->ifa_addr) ||
1097 			    (ifa->ifa_dstaddr &&
1098 			     equal(addr, ifa->ifa_dstaddr)))
1099 				return (ifa);
1100 			continue;
1101 		}
1102 		cp = addr->sa_data;
1103 		cp2 = ifa->ifa_addr->sa_data;
1104 		cp3 = ifa->ifa_netmask->sa_data;
1105 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1106 		for (; cp3 < cplim; cp3++) {
1107 			if ((*cp++ ^ *cp2++) & *cp3)
1108 				break;
1109 		}
1110 		if (cp3 == cplim)
1111 			return (ifa);
1112 	}
1113 	return (ifa_maybe);
1114 }
1115 
1116 /*
1117  * Default action when installing a route with a Link Level gateway.
1118  * Lookup an appropriate real ifa to point to.
1119  * This should be moved to /sys/net/link.c eventually.
1120  */
1121 void
1122 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1123 {
1124 	struct ifaddr *ifa;
1125 	struct sockaddr *dst;
1126 	struct ifnet *ifp;
1127 
1128 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1129 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1130 		return;
1131 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1132 		rt_replace_ifa(rt, ifa);
1133 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1134 			ifa->ifa_rtrequest(cmd, rt, info);
1135 	}
1136 }
1137 
1138 /*
1139  * Handle a change in the interface link state.
1140  */
1141 void
1142 if_link_state_change(struct ifnet *ifp, int link_state)
1143 {
1144 	/* Notify that the link state has changed. */
1145 	if (ifp->if_link_state != link_state) {
1146 		ifp->if_link_state = link_state;
1147 		rt_ifmsg(ifp);
1148 #if NCARP > 0
1149 		if (ifp->if_carp)
1150 			carp_carpdev_state(ifp);
1151 #endif
1152 	}
1153 }
1154 
1155 /*
1156  * Mark an interface down and notify protocols of
1157  * the transition.
1158  * NOTE: must be called at splsoftnet or equivalent.
1159  */
1160 void
1161 if_down(struct ifnet *ifp)
1162 {
1163 	struct ifaddr *ifa;
1164 
1165 	ifp->if_flags &= ~IFF_UP;
1166 	microtime(&ifp->if_lastchange);
1167 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1168 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1169 	IFQ_PURGE(&ifp->if_snd);
1170 #if NCARP > 0
1171 	if (ifp->if_carp)
1172 		carp_carpdev_state(ifp);
1173 #endif
1174 	rt_ifmsg(ifp);
1175 }
1176 
1177 /*
1178  * Mark an interface up and notify protocols of
1179  * the transition.
1180  * NOTE: must be called at splsoftnet or equivalent.
1181  */
1182 void
1183 if_up(struct ifnet *ifp)
1184 {
1185 #ifdef notyet
1186 	struct ifaddr *ifa;
1187 #endif
1188 
1189 	ifp->if_flags |= IFF_UP;
1190 	microtime(&ifp->if_lastchange);
1191 #ifdef notyet
1192 	/* this has no effect on IP, and will kill all ISO connections XXX */
1193 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1194 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1195 #endif
1196 #if NCARP > 0
1197 	if (ifp->if_carp)
1198 		carp_carpdev_state(ifp);
1199 #endif
1200 	rt_ifmsg(ifp);
1201 #ifdef INET6
1202 	in6_if_up(ifp);
1203 #endif
1204 }
1205 
1206 /*
1207  * Handle interface watchdog timer routines.  Called
1208  * from softclock, we decrement timers (if set) and
1209  * call the appropriate interface routine on expiration.
1210  */
1211 void
1212 if_slowtimo(void *arg)
1213 {
1214 	struct ifnet *ifp;
1215 	int s = splnet();
1216 
1217 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1218 		if (ifp->if_timer == 0 || --ifp->if_timer)
1219 			continue;
1220 		if (ifp->if_watchdog)
1221 			(*ifp->if_watchdog)(ifp);
1222 	}
1223 	splx(s);
1224 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1225 	    if_slowtimo, NULL);
1226 }
1227 
1228 /*
1229  * Set/clear promiscuous mode on interface ifp based on the truth value
1230  * of pswitch.  The calls are reference counted so that only the first
1231  * "on" request actually has an effect, as does the final "off" request.
1232  * Results are undefined if the "off" and "on" requests are not matched.
1233  */
1234 int
1235 ifpromisc(struct ifnet *ifp, int pswitch)
1236 {
1237 	int pcount, ret;
1238 	short flags;
1239 	struct ifreq ifr;
1240 
1241 	pcount = ifp->if_pcount;
1242 	flags = ifp->if_flags;
1243 	if (pswitch) {
1244 		/*
1245 		 * Allow the device to be "placed" into promiscuous
1246 		 * mode even if it is not configured up.  It will
1247 		 * consult IFF_PROMISC when it is is brought up.
1248 		 */
1249 		if (ifp->if_pcount++ != 0)
1250 			return (0);
1251 		ifp->if_flags |= IFF_PROMISC;
1252 		if ((ifp->if_flags & IFF_UP) == 0)
1253 			return (0);
1254 	} else {
1255 		if (--ifp->if_pcount > 0)
1256 			return (0);
1257 		ifp->if_flags &= ~IFF_PROMISC;
1258 		/*
1259 		 * If the device is not configured up, we should not need to
1260 		 * turn off promiscuous mode (device should have turned it
1261 		 * off when interface went down; and will look at IFF_PROMISC
1262 		 * again next time interface comes up).
1263 		 */
1264 		if ((ifp->if_flags & IFF_UP) == 0)
1265 			return (0);
1266 	}
1267 	memset(&ifr, 0, sizeof(ifr));
1268 	ifr.ifr_flags = ifp->if_flags;
1269 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1270 	/* Restore interface state if not successful. */
1271 	if (ret != 0) {
1272 		ifp->if_pcount = pcount;
1273 		ifp->if_flags = flags;
1274 	}
1275 	return (ret);
1276 }
1277 
1278 /*
1279  * Map interface name to
1280  * interface structure pointer.
1281  */
1282 struct ifnet *
1283 ifunit(const char *name)
1284 {
1285 	struct ifnet *ifp;
1286 	const char *cp = name;
1287 	u_int unit = 0;
1288 	u_int i;
1289 
1290 	/*
1291 	 * If the entire name is a number, treat it as an ifindex.
1292 	 */
1293 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1294 		unit = unit * 10 + (*cp - '0');
1295 	}
1296 
1297 	/*
1298 	 * If the number took all of the name, then it's a valid ifindex.
1299 	 */
1300 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1301 		if (unit >= if_indexlim)
1302 			return (NULL);
1303 		ifp = ifindex2ifnet[unit];
1304 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1305 			return (NULL);
1306 		return (ifp);
1307 	}
1308 
1309 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1310 		if (ifp->if_output == if_nulloutput)
1311 			continue;
1312 	 	if (strcmp(ifp->if_xname, name) == 0)
1313 			return (ifp);
1314 	}
1315 	return (NULL);
1316 }
1317 
1318 /*
1319  * Interface ioctls.
1320  */
1321 int
1322 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct lwp *l)
1323 {
1324 	struct ifnet *ifp;
1325 	struct ifreq *ifr;
1326 	struct ifcapreq *ifcr;
1327 	struct ifdatareq *ifdr;
1328 	int s, error = 0;
1329 	short oif_flags;
1330 
1331 	switch (cmd) {
1332 
1333 	case SIOCGIFCONF:
1334 	case OSIOCGIFCONF:
1335 		return (ifconf(cmd, data));
1336 	}
1337 	ifr = (struct ifreq *)data;
1338 	ifcr = (struct ifcapreq *)data;
1339 	ifdr = (struct ifdatareq *)data;
1340 
1341 	ifp = ifunit(ifr->ifr_name);
1342 
1343 	switch (cmd) {
1344 	case SIOCIFCREATE:
1345 	case SIOCIFDESTROY:
1346 		if (l) {
1347 			error = kauth_authorize_network(l->l_cred,
1348 			    KAUTH_NETWORK_INTERFACE,
1349 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1350 			    (void *)cmd, NULL);
1351 			if (error)
1352 				return error;
1353 		}
1354 		return ((cmd == SIOCIFCREATE) ?
1355 			if_clone_create(ifr->ifr_name) :
1356 			if_clone_destroy(ifr->ifr_name));
1357 
1358 	case SIOCIFGCLONERS:
1359 		return (if_clone_list((struct if_clonereq *)data));
1360 	}
1361 
1362 	if (ifp == 0)
1363 		return (ENXIO);
1364 
1365 	switch (cmd) {
1366 	case SIOCSIFFLAGS:
1367 	case SIOCSIFCAP:
1368 	case SIOCSIFMETRIC:
1369 	case SIOCZIFDATA:
1370 	case SIOCSIFMTU:
1371 	case SIOCSIFPHYADDR:
1372 	case SIOCDIFPHYADDR:
1373 #ifdef INET6
1374 	case SIOCSIFPHYADDR_IN6:
1375 #endif
1376 	case SIOCSLIFPHYADDR:
1377 	case SIOCADDMULTI:
1378 	case SIOCDELMULTI:
1379 	case SIOCSIFMEDIA:
1380 	case SIOCSDRVSPEC:
1381 	case SIOCS80211NWID:
1382 	case SIOCS80211NWKEY:
1383 	case SIOCS80211POWER:
1384 	case SIOCS80211BSSID:
1385 	case SIOCS80211CHANNEL:
1386 		if (l) {
1387 			error = kauth_authorize_network(l->l_cred,
1388 			    KAUTH_NETWORK_INTERFACE,
1389 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1390 			    (void *)cmd, NULL);
1391 			if (error)
1392 				return error;
1393 		}
1394 	}
1395 
1396 	oif_flags = ifp->if_flags;
1397 	switch (cmd) {
1398 
1399 	case SIOCGIFFLAGS:
1400 		ifr->ifr_flags = ifp->if_flags;
1401 		break;
1402 
1403 	case SIOCGIFMETRIC:
1404 		ifr->ifr_metric = ifp->if_metric;
1405 		break;
1406 
1407 	case SIOCGIFMTU:
1408 		ifr->ifr_mtu = ifp->if_mtu;
1409 		break;
1410 
1411 	case SIOCGIFDLT:
1412 		ifr->ifr_dlt = ifp->if_dlt;
1413 		break;
1414 
1415 	case SIOCSIFFLAGS:
1416 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1417 			s = splnet();
1418 			if_down(ifp);
1419 			splx(s);
1420 		}
1421 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1422 			s = splnet();
1423 			if_up(ifp);
1424 			splx(s);
1425 		}
1426 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1427 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1428 		if (ifp->if_ioctl)
1429 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1430 		break;
1431 
1432 	case SIOCGIFCAP:
1433 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1434 		ifcr->ifcr_capenable = ifp->if_capenable;
1435 		break;
1436 
1437 	case SIOCSIFCAP:
1438 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1439 			return (EINVAL);
1440 		if (ifp->if_ioctl == NULL)
1441 			return (EOPNOTSUPP);
1442 
1443 		/* Must prevent race with packet reception here. */
1444 		s = splnet();
1445 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1446 			struct ifreq ifrq;
1447 
1448 			ifrq.ifr_flags = ifp->if_flags;
1449 			ifp->if_capenable = ifcr->ifcr_capenable;
1450 
1451 			/* Pre-compute the checksum flags mask. */
1452 			ifp->if_csum_flags_tx = 0;
1453 			ifp->if_csum_flags_rx = 0;
1454 			if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1455 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1456 			}
1457 			if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1458 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1459 			}
1460 
1461 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1462 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1463 			}
1464 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1465 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1466 			}
1467 
1468 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1469 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1470 			}
1471 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1472 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1473 			}
1474 
1475 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1476 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1477 			}
1478 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1479 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1480 			}
1481 
1482 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1483 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1484 			}
1485 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1486 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1487 			}
1488 
1489 			/*
1490 			 * Only kick the interface if it's up.  If it's
1491 			 * not up now, it will notice the cap enables
1492 			 * when it is brought up later.
1493 			 */
1494 			if (ifp->if_flags & IFF_UP)
1495 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1496 				    (caddr_t) &ifrq);
1497 		}
1498 		splx(s);
1499 		break;
1500 
1501 	case SIOCSIFMETRIC:
1502 		ifp->if_metric = ifr->ifr_metric;
1503 		break;
1504 
1505 	case SIOCGIFDATA:
1506 		ifdr->ifdr_data = ifp->if_data;
1507 		break;
1508 
1509 	case SIOCZIFDATA:
1510 		ifdr->ifdr_data = ifp->if_data;
1511 		/*
1512 		 * Assumes that the volatile counters that can be
1513 		 * zero'ed are at the end of if_data.
1514 		 */
1515 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1516 		    offsetof(struct if_data, ifi_ipackets));
1517 		break;
1518 
1519 	case SIOCSIFMTU:
1520 	{
1521 		u_long oldmtu = ifp->if_mtu;
1522 
1523 		if (ifp->if_ioctl == NULL)
1524 			return (EOPNOTSUPP);
1525 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1526 
1527 		/*
1528 		 * If the link MTU changed, do network layer specific procedure.
1529 		 */
1530 		if (ifp->if_mtu != oldmtu) {
1531 #ifdef INET6
1532 			nd6_setmtu(ifp);
1533 #endif
1534 		}
1535 		break;
1536 	}
1537 	case SIOCSIFPHYADDR:
1538 	case SIOCDIFPHYADDR:
1539 #ifdef INET6
1540 	case SIOCSIFPHYADDR_IN6:
1541 #endif
1542 	case SIOCSLIFPHYADDR:
1543 	case SIOCADDMULTI:
1544 	case SIOCDELMULTI:
1545 	case SIOCSIFMEDIA:
1546 	case SIOCGIFPSRCADDR:
1547 	case SIOCGIFPDSTADDR:
1548 	case SIOCGLIFPHYADDR:
1549 	case SIOCGIFMEDIA:
1550 		if (ifp->if_ioctl == 0)
1551 			return (EOPNOTSUPP);
1552 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1553 		break;
1554 
1555 	case SIOCSDRVSPEC:
1556 	case SIOCS80211NWID:
1557 	case SIOCS80211NWKEY:
1558 	case SIOCS80211POWER:
1559 	case SIOCS80211BSSID:
1560 	case SIOCS80211CHANNEL:
1561 	default:
1562 		if (so->so_proto == 0)
1563 			return (EOPNOTSUPP);
1564 #ifdef COMPAT_OSOCK
1565 		error = compat_ifioctl(so, cmd, data, l);
1566 #else
1567 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1568 		    (struct mbuf *)cmd, (struct mbuf *)data,
1569 		    (struct mbuf *)ifp, l));
1570 #endif
1571 		break;
1572 	}
1573 
1574 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1575 #ifdef INET6
1576 		if ((ifp->if_flags & IFF_UP) != 0) {
1577 			s = splnet();
1578 			in6_if_up(ifp);
1579 			splx(s);
1580 		}
1581 #endif
1582 	}
1583 
1584 	return (error);
1585 }
1586 
1587 /*
1588  * Return interface configuration
1589  * of system.  List may be used
1590  * in later ioctl's (above) to get
1591  * other information.
1592  */
1593 /*ARGSUSED*/
1594 int
1595 ifconf(u_long cmd, caddr_t data)
1596 {
1597 	struct ifconf *ifc = (struct ifconf *)data;
1598 	struct ifnet *ifp;
1599 	struct ifaddr *ifa;
1600 	struct ifreq ifr, *ifrp;
1601 	int space = ifc->ifc_len, error = 0;
1602 	const int sz = (int)sizeof(ifr);
1603 	int sign;
1604 
1605 	if ((ifrp = ifc->ifc_req) == NULL) {
1606 		space = 0;
1607 		sign = -1;
1608 	} else {
1609 		sign = 1;
1610 	}
1611 	IFNET_FOREACH(ifp) {
1612 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1613 		    sizeof(ifr.ifr_name));
1614 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1615 			return ENAMETOOLONG;
1616 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == NULL) {
1617 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1618 			if (ifrp != NULL && space >= sz) {
1619 				error = copyout(&ifr, ifrp, sz);
1620 				if (error)
1621 					break;
1622 				ifrp++;
1623 			}
1624 			space -= sizeof(ifr) * sign;
1625 			continue;
1626 		}
1627 
1628 		for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1629 			struct sockaddr *sa = ifa->ifa_addr;
1630 #ifdef COMPAT_OSOCK
1631 			if (cmd == OSIOCGIFCONF) {
1632 				struct osockaddr *osa =
1633 					 (struct osockaddr *)&ifr.ifr_addr;
1634 				/*
1635 				 * If it does not fit, we don't bother with it
1636 				 */
1637 				if (sa->sa_len > sizeof(*osa))
1638 					continue;
1639 				ifr.ifr_addr = *sa;
1640 				osa->sa_family = sa->sa_family;
1641 				if (ifrp != NULL && space >= sz) {
1642 					error = copyout(&ifr, ifrp, sz);
1643 					ifrp++;
1644 				}
1645 			} else
1646 #endif
1647 			if (sa->sa_len <= sizeof(*sa)) {
1648 				ifr.ifr_addr = *sa;
1649 				if (ifrp != NULL && space >= sz) {
1650 					error = copyout(&ifr, ifrp, sz);
1651 					ifrp++;
1652 				}
1653 			} else {
1654 				space -= (sa->sa_len - sizeof(*sa)) * sign;
1655 				if (ifrp != NULL && space >= sz) {
1656 					error = copyout(&ifr, ifrp,
1657 					    sizeof(ifr.ifr_name));
1658 					if (error == 0) {
1659 						error = copyout(sa,
1660 						    &ifrp->ifr_addr,
1661 						    sa->sa_len);
1662 					}
1663 					ifrp = (struct ifreq *)
1664 						(sa->sa_len +
1665 						 (caddr_t)&ifrp->ifr_addr);
1666 				}
1667 			}
1668 			if (error)
1669 				break;
1670 			space -= sz * sign;
1671 		}
1672 	}
1673 	if (ifrp != NULL)
1674 		ifc->ifc_len -= space;
1675 	else
1676 		ifc->ifc_len = space;
1677 	return (error);
1678 }
1679 
1680 /*
1681  * Queue message on interface, and start output if interface
1682  * not yet active.
1683  */
1684 int
1685 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
1686     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1687 {
1688 	int len = m->m_pkthdr.len;
1689 	int mflags = m->m_flags;
1690 	int s = splnet();
1691 	int error;
1692 
1693 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1694 	if (error) {
1695 		splx(s);
1696 		return error;
1697 	}
1698 	ifp->if_obytes += len;
1699 	if (mflags & M_MCAST)
1700 		ifp->if_omcasts++;
1701 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
1702 		(*ifp->if_start)(ifp);
1703 	splx(s);
1704 	return error;
1705 }
1706 
1707 /*
1708  * Queue message on interface, possibly using a second fast queue
1709  */
1710 int
1711 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
1712     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1713 {
1714 	int error = 0;
1715 
1716 	if (ifq != NULL
1717 #ifdef ALTQ
1718 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
1719 #endif
1720 	    ) {
1721 		if (IF_QFULL(ifq)) {
1722 			IF_DROP(&ifp->if_snd);
1723 			m_freem(m);
1724 			if (error == 0)
1725 				error = ENOBUFS;
1726 		}
1727 		else
1728 			IF_ENQUEUE(ifq, m);
1729 	} else
1730 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1731 	if (error != 0) {
1732 		++ifp->if_oerrors;
1733 		return error;
1734 	}
1735 
1736 	return 0;
1737 }
1738 
1739 
1740 #if defined(INET) || defined(INET6)
1741 static void
1742 sysctl_net_ifq_setup(struct sysctllog **clog,
1743 		     int pf, const char *pfname,
1744 		     int ipn, const char *ipname,
1745 		     int qid, struct ifqueue *ifq)
1746 {
1747 
1748 	sysctl_createv(clog, 0, NULL, NULL,
1749 		       CTLFLAG_PERMANENT,
1750 		       CTLTYPE_NODE, "net", NULL,
1751 		       NULL, 0, NULL, 0,
1752 		       CTL_NET, CTL_EOL);
1753 	sysctl_createv(clog, 0, NULL, NULL,
1754 		       CTLFLAG_PERMANENT,
1755 		       CTLTYPE_NODE, pfname, NULL,
1756 		       NULL, 0, NULL, 0,
1757 		       CTL_NET, pf, CTL_EOL);
1758 	sysctl_createv(clog, 0, NULL, NULL,
1759 		       CTLFLAG_PERMANENT,
1760 		       CTLTYPE_NODE, ipname, NULL,
1761 		       NULL, 0, NULL, 0,
1762 		       CTL_NET, pf, ipn, CTL_EOL);
1763 	sysctl_createv(clog, 0, NULL, NULL,
1764 		       CTLFLAG_PERMANENT,
1765 		       CTLTYPE_NODE, "ifq",
1766 		       SYSCTL_DESCR("Protocol input queue controls"),
1767 		       NULL, 0, NULL, 0,
1768 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1769 
1770 	sysctl_createv(clog, 0, NULL, NULL,
1771 		       CTLFLAG_PERMANENT,
1772 		       CTLTYPE_INT, "len",
1773 		       SYSCTL_DESCR("Current input queue length"),
1774 		       NULL, 0, &ifq->ifq_len, 0,
1775 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1776 	sysctl_createv(clog, 0, NULL, NULL,
1777 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1778 		       CTLTYPE_INT, "maxlen",
1779 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1780 		       NULL, 0, &ifq->ifq_maxlen, 0,
1781 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1782 #ifdef notyet
1783 	sysctl_createv(clog, 0, NULL, NULL,
1784 		       CTLFLAG_PERMANENT,
1785 		       CTLTYPE_INT, "peak",
1786 		       SYSCTL_DESCR("Highest input queue length"),
1787 		       NULL, 0, &ifq->ifq_peak, 0,
1788 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1789 #endif
1790 	sysctl_createv(clog, 0, NULL, NULL,
1791 		       CTLFLAG_PERMANENT,
1792 		       CTLTYPE_INT, "drops",
1793 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1794 		       NULL, 0, &ifq->ifq_drops, 0,
1795 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1796 }
1797 
1798 #ifdef INET
1799 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1800 	     "sysctl net.inet.ip.ifq subtree setup")
1801 {
1802 	extern struct ifqueue ipintrq;
1803 
1804 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1805 			     IPCTL_IFQ, &ipintrq);
1806 }
1807 #endif /* INET */
1808 
1809 #ifdef INET6
1810 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1811 	     "sysctl net.inet6.ip6.ifq subtree setup")
1812 {
1813 	extern struct ifqueue ip6intrq;
1814 
1815 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1816 			     IPV6CTL_IFQ, &ip6intrq);
1817 }
1818 #endif /* INET6 */
1819 #endif /* INET || INET6 */
1820