xref: /netbsd-src/sys/net/if.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: if.c,v 1.147 2004/10/07 20:06:58 tron 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 Studnemund 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 1.147 2004/10/07 20:06:58 tron 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_ccitt.h"
110 #include "opt_natm.h"
111 #include "opt_pfil_hooks.h"
112 
113 #include <sys/param.h>
114 #include <sys/mbuf.h>
115 #include <sys/systm.h>
116 #include <sys/callout.h>
117 #include <sys/proc.h>
118 #include <sys/socket.h>
119 #include <sys/socketvar.h>
120 #include <sys/domain.h>
121 #include <sys/protosw.h>
122 #include <sys/kernel.h>
123 #include <sys/ioctl.h>
124 #include <sys/sysctl.h>
125 
126 #include <net/if.h>
127 #include <net/if_dl.h>
128 #include <net/if_ether.h>
129 #include <net/if_media.h>
130 #include <net80211/ieee80211.h>
131 #include <net80211/ieee80211_ioctl.h>
132 #include <net/if_types.h>
133 #include <net/radix.h>
134 #include <net/route.h>
135 #include <net/netisr.h>
136 #ifdef NETATALK
137 #include <netatalk/at_extern.h>
138 #include <netatalk/at.h>
139 #endif
140 #include <net/pfil.h>
141 
142 #ifdef INET6
143 #include <netinet/in.h>
144 #include <netinet6/in6_var.h>
145 #include <netinet6/nd6.h>
146 #endif
147 
148 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
149 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
150 
151 int	ifqmaxlen = IFQ_MAXLEN;
152 struct	callout if_slowtimo_ch;
153 
154 int netisr;			/* scheduling bits for network */
155 
156 int if_rt_walktree __P((struct radix_node *, void *));
157 
158 struct if_clone *if_clone_lookup __P((const char *, int *));
159 int if_clone_list __P((struct if_clonereq *));
160 
161 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
162 int if_cloners_count;
163 
164 #ifdef PFIL_HOOKS
165 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
166 #endif
167 
168 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
169     defined(ISO) || defined(CCITT) || defined(NATM)
170 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
171 #endif
172 
173 /*
174  * Network interface utility routines.
175  *
176  * Routines with ifa_ifwith* names take sockaddr *'s as
177  * parameters.
178  */
179 void
180 ifinit()
181 {
182 
183 	callout_init(&if_slowtimo_ch);
184 	if_slowtimo(NULL);
185 #ifdef PFIL_HOOKS
186 	if_pfil.ph_type = PFIL_TYPE_IFNET;
187 	if_pfil.ph_ifnet = NULL;
188 	if (pfil_head_register(&if_pfil) != 0)
189 		printf("WARNING: unable to register pfil hook\n");
190 #endif
191 }
192 
193 /*
194  * Null routines used while an interface is going away.  These routines
195  * just return an error.
196  */
197 
198 int
199 if_nulloutput(ifp, m, so, rt)
200 	struct ifnet *ifp;
201 	struct mbuf *m;
202 	struct sockaddr *so;
203 	struct rtentry *rt;
204 {
205 
206 	return (ENXIO);
207 }
208 
209 void
210 if_nullinput(ifp, m)
211 	struct ifnet *ifp;
212 	struct mbuf *m;
213 {
214 
215 	/* Nothing. */
216 }
217 
218 void
219 if_nullstart(ifp)
220 	struct ifnet *ifp;
221 {
222 
223 	/* Nothing. */
224 }
225 
226 int
227 if_nullioctl(ifp, cmd, data)
228 	struct ifnet *ifp;
229 	u_long cmd;
230 	caddr_t data;
231 {
232 
233 	return (ENXIO);
234 }
235 
236 int
237 if_nullinit(ifp)
238 	struct ifnet *ifp;
239 {
240 
241 	return (ENXIO);
242 }
243 
244 void
245 if_nullstop(ifp, disable)
246 	struct ifnet *ifp;
247 	int disable;
248 {
249 
250 	/* Nothing. */
251 }
252 
253 void
254 if_nullwatchdog(ifp)
255 	struct ifnet *ifp;
256 {
257 
258 	/* Nothing. */
259 }
260 
261 void
262 if_nulldrain(ifp)
263 	struct ifnet *ifp;
264 {
265 
266 	/* Nothing. */
267 }
268 
269 static u_int if_index = 1;
270 struct ifnet_head ifnet;
271 size_t if_indexlim = 0;
272 struct ifaddr **ifnet_addrs = NULL;
273 struct ifnet **ifindex2ifnet = NULL;
274 
275 /*
276  * Allocate the link level name for the specified interface.  This
277  * is an attachment helper.  It must be called after ifp->if_addrlen
278  * is initialized, which may not be the case when if_attach() is
279  * called.
280  */
281 void
282 if_alloc_sadl(struct ifnet *ifp)
283 {
284 	unsigned socksize, ifasize;
285 	int namelen, masklen;
286 	struct sockaddr_dl *sdl;
287 	struct ifaddr *ifa;
288 
289 	/*
290 	 * If the interface already has a link name, release it
291 	 * now.  This is useful for interfaces that can change
292 	 * link types, and thus switch link names often.
293 	 */
294 	if (ifp->if_sadl != NULL)
295 		if_free_sadl(ifp);
296 
297 	namelen = strlen(ifp->if_xname);
298 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
299 	socksize = masklen + ifp->if_addrlen;
300 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
301 	if (socksize < sizeof(*sdl))
302 		socksize = sizeof(*sdl);
303 	socksize = ROUNDUP(socksize);
304 	ifasize = sizeof(*ifa) + 2 * socksize;
305 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
306 	memset((caddr_t)ifa, 0, ifasize);
307 	sdl = (struct sockaddr_dl *)(ifa + 1);
308 	sdl->sdl_len = socksize;
309 	sdl->sdl_family = AF_LINK;
310 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
311 	sdl->sdl_nlen = namelen;
312 	sdl->sdl_alen = ifp->if_addrlen;
313 	sdl->sdl_index = ifp->if_index;
314 	sdl->sdl_type = ifp->if_type;
315 	ifnet_addrs[ifp->if_index] = ifa;
316 	IFAREF(ifa);
317 	ifa->ifa_ifp = ifp;
318 	ifa->ifa_rtrequest = link_rtrequest;
319 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
320 	IFAREF(ifa);
321 	ifa->ifa_addr = (struct sockaddr *)sdl;
322 	ifp->if_sadl = sdl;
323 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
324 	ifa->ifa_netmask = (struct sockaddr *)sdl;
325 	sdl->sdl_len = masklen;
326 	while (namelen != 0)
327 		sdl->sdl_data[--namelen] = 0xff;
328 }
329 
330 /*
331  * Free the link level name for the specified interface.  This is
332  * a detach helper.  This is called from if_detach() or from
333  * link layer type specific detach functions.
334  */
335 void
336 if_free_sadl(struct ifnet *ifp)
337 {
338 	struct ifaddr *ifa;
339 	int s;
340 
341 	ifa = ifnet_addrs[ifp->if_index];
342 	if (ifa == NULL) {
343 		KASSERT(ifp->if_sadl == NULL);
344 		return;
345 	}
346 
347 	KASSERT(ifp->if_sadl != NULL);
348 
349 	s = splnet();
350 	rtinit(ifa, RTM_DELETE, 0);
351 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
352 	IFAFREE(ifa);
353 
354 	ifp->if_sadl = NULL;
355 
356 	ifnet_addrs[ifp->if_index] = NULL;
357 	IFAFREE(ifa);
358 	splx(s);
359 }
360 
361 /*
362  * Attach an interface to the
363  * list of "active" interfaces.
364  */
365 void
366 if_attach(ifp)
367 	struct ifnet *ifp;
368 {
369 	int indexlim = 0;
370 
371 	if (if_indexlim == 0) {
372 		TAILQ_INIT(&ifnet);
373 		if_indexlim = 8;
374 	}
375 	TAILQ_INIT(&ifp->if_addrlist);
376 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
377 	ifp->if_index = if_index;
378 	if (ifindex2ifnet == 0)
379 		if_index++;
380 	else
381 		while (ifp->if_index < if_indexlim &&
382 		    ifindex2ifnet[ifp->if_index] != NULL) {
383 			++if_index;
384 			if (if_index == 0)
385 				if_index = 1;
386 			/*
387 			 * If we hit USHRT_MAX, we skip back to 0 since
388 			 * there are a number of places where the value
389 			 * of if_index or if_index itself is compared
390 			 * to or stored in an unsigned short.  By
391 			 * jumping back, we won't botch those assignments
392 			 * or comparisons.
393 			 */
394 			else if (if_index == USHRT_MAX) {
395 				/*
396 				 * However, if we have to jump back to
397 				 * zero *twice* without finding an empty
398 				 * slot in ifindex2ifnet[], then there
399 				 * there are too many (>65535) interfaces.
400 				 */
401 				if (indexlim++)
402 					panic("too many interfaces");
403 				else
404 					if_index = 1;
405 			}
406 			ifp->if_index = if_index;
407 		}
408 
409 	/*
410 	 * We have some arrays that should be indexed by if_index.
411 	 * since if_index will grow dynamically, they should grow too.
412 	 *	struct ifadd **ifnet_addrs
413 	 *	struct ifnet **ifindex2ifnet
414 	 */
415 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
416 	    ifp->if_index >= if_indexlim) {
417 		size_t m, n, oldlim;
418 		caddr_t q;
419 
420 		oldlim = if_indexlim;
421 		while (ifp->if_index >= if_indexlim)
422 			if_indexlim <<= 1;
423 
424 		/* grow ifnet_addrs */
425 		m = oldlim * sizeof(struct ifaddr *);
426 		n = if_indexlim * sizeof(struct ifaddr *);
427 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
428 		memset(q, 0, n);
429 		if (ifnet_addrs) {
430 			bcopy((caddr_t)ifnet_addrs, q, m);
431 			free((caddr_t)ifnet_addrs, M_IFADDR);
432 		}
433 		ifnet_addrs = (struct ifaddr **)q;
434 
435 		/* grow ifindex2ifnet */
436 		m = oldlim * sizeof(struct ifnet *);
437 		n = if_indexlim * sizeof(struct ifnet *);
438 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
439 		memset(q, 0, n);
440 		if (ifindex2ifnet) {
441 			bcopy((caddr_t)ifindex2ifnet, q, m);
442 			free((caddr_t)ifindex2ifnet, M_IFADDR);
443 		}
444 		ifindex2ifnet = (struct ifnet **)q;
445 	}
446 
447 	ifindex2ifnet[ifp->if_index] = ifp;
448 
449 	/*
450 	 * Link level name is allocated later by a separate call to
451 	 * if_alloc_sadl().
452 	 */
453 
454 	if (ifp->if_snd.ifq_maxlen == 0)
455 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
456 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
457 
458 	ifp->if_link_state = LINK_STATE_UNKNOWN;
459 
460 	ifp->if_capenable = 0;
461 	ifp->if_csum_flags_tx = 0;
462 	ifp->if_csum_flags_rx = 0;
463 
464 #ifdef ALTQ
465 	ifp->if_snd.altq_type = 0;
466 	ifp->if_snd.altq_disc = NULL;
467 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
468 	ifp->if_snd.altq_tbr  = NULL;
469 	ifp->if_snd.altq_ifp  = ifp;
470 #endif
471 
472 #ifdef PFIL_HOOKS
473 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
474 	ifp->if_pfil.ph_ifnet = ifp;
475 	if (pfil_head_register(&ifp->if_pfil) != 0)
476 		printf("%s: WARNING: unable to register pfil hook\n",
477 		    ifp->if_xname);
478 	(void)pfil_run_hooks(&if_pfil,
479 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
480 #endif
481 
482 	if (domains)
483 		if_attachdomain1(ifp);
484 
485 	/* Announce the interface. */
486 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
487 }
488 
489 void
490 if_attachdomain()
491 {
492 	struct ifnet *ifp;
493 	int s;
494 
495 	s = splnet();
496 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
497 		if_attachdomain1(ifp);
498 	splx(s);
499 }
500 
501 void
502 if_attachdomain1(ifp)
503 	struct ifnet *ifp;
504 {
505 	struct domain *dp;
506 	int s;
507 
508 	s = splnet();
509 
510 	/* address family dependent data region */
511 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
512 	for (dp = domains; dp; dp = dp->dom_next) {
513 		if (dp->dom_ifattach)
514 			ifp->if_afdata[dp->dom_family] =
515 			    (*dp->dom_ifattach)(ifp);
516 	}
517 
518 	splx(s);
519 }
520 
521 /*
522  * Deactivate an interface.  This points all of the procedure
523  * handles at error stubs.  May be called from interrupt context.
524  */
525 void
526 if_deactivate(ifp)
527 	struct ifnet *ifp;
528 {
529 	int s;
530 
531 	s = splnet();
532 
533 	ifp->if_output	 = if_nulloutput;
534 	ifp->if_input	 = if_nullinput;
535 	ifp->if_start	 = if_nullstart;
536 	ifp->if_ioctl	 = if_nullioctl;
537 	ifp->if_init	 = if_nullinit;
538 	ifp->if_stop	 = if_nullstop;
539 	ifp->if_watchdog = if_nullwatchdog;
540 	ifp->if_drain	 = if_nulldrain;
541 
542 	/* No more packets may be enqueued. */
543 	ifp->if_snd.ifq_maxlen = 0;
544 
545 	splx(s);
546 }
547 
548 /*
549  * Detach an interface from the list of "active" interfaces,
550  * freeing any resources as we go along.
551  *
552  * NOTE: This routine must be called with a valid thread context,
553  * as it may block.
554  */
555 void
556 if_detach(ifp)
557 	struct ifnet *ifp;
558 {
559 	struct socket so;
560 	struct ifaddr *ifa, **ifap;
561 #ifdef IFAREF_DEBUG
562 	struct ifaddr *last_ifa = NULL;
563 #endif
564 	struct domain *dp;
565 	const struct protosw *pr;
566 	struct radix_node_head *rnh;
567 	int s, i, family, purged;
568 
569 	/*
570 	 * XXX It's kind of lame that we have to have the
571 	 * XXX socket structure...
572 	 */
573 	memset(&so, 0, sizeof(so));
574 
575 	s = splnet();
576 
577 	/*
578 	 * Do an if_down() to give protocols a chance to do something.
579 	 */
580 	if_down(ifp);
581 
582 #ifdef ALTQ
583 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
584 		altq_disable(&ifp->if_snd);
585 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
586 		altq_detach(&ifp->if_snd);
587 #endif
588 
589 #ifdef PFIL_HOOKS
590 	(void)pfil_run_hooks(&if_pfil,
591 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
592 	(void)pfil_head_unregister(&ifp->if_pfil);
593 #endif
594 
595 	/*
596 	 * Rip all the addresses off the interface.  This should make
597 	 * all of the routes go away.
598 	 */
599 	ifap = &TAILQ_FIRST(&ifp->if_addrlist); /* XXX abstraction violation */
600 	while ((ifa = *ifap)) {
601 		family = ifa->ifa_addr->sa_family;
602 #ifdef IFAREF_DEBUG
603 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
604 		    ifa, family, ifa->ifa_refcnt);
605 		if (last_ifa != NULL && ifa == last_ifa)
606 			panic("if_detach: loop detected");
607 		last_ifa = ifa;
608 #endif
609 		if (family == AF_LINK) {
610 			ifap = &TAILQ_NEXT(ifa, ifa_list);
611 			continue;
612 		}
613 		dp = pffinddomain(family);
614 #ifdef DIAGNOSTIC
615 		if (dp == NULL)
616 			panic("if_detach: no domain for AF %d",
617 			    family);
618 #endif
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, curproc);
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 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
638 		}
639 	}
640 
641 	if_free_sadl(ifp);
642 
643 	/* Walk the routing table looking for straglers. */
644 	for (i = 0; i <= AF_MAX; i++) {
645 		if ((rnh = rt_tables[i]) != NULL)
646 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
647 	}
648 
649 	for (dp = domains; dp; dp = dp->dom_next) {
650 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
651 			(*dp->dom_ifdetach)(ifp,
652 			    ifp->if_afdata[dp->dom_family]);
653 	}
654 
655 	/* Announce that the interface is gone. */
656 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
657 
658 	ifindex2ifnet[ifp->if_index] = NULL;
659 
660 	TAILQ_REMOVE(&ifnet, ifp, if_list);
661 
662 	/*
663 	 * remove packets came from ifp, from software interrupt queues.
664 	 * net/netisr_dispatch.h is not usable, as some of them use
665 	 * strange queue names.
666 	 */
667 #define IF_DETACH_QUEUES(x) \
668 do { \
669 	extern struct ifqueue x; \
670 	if_detach_queues(ifp, & x); \
671 } while (/*CONSTCOND*/ 0)
672 #ifdef INET
673 #if NARP > 0
674 	IF_DETACH_QUEUES(arpintrq);
675 #endif
676 	IF_DETACH_QUEUES(ipintrq);
677 #endif
678 #ifdef INET6
679 	IF_DETACH_QUEUES(ip6intrq);
680 #endif
681 #ifdef NETATALK
682 	IF_DETACH_QUEUES(atintrq1);
683 	IF_DETACH_QUEUES(atintrq2);
684 #endif
685 #ifdef NS
686 	IF_DETACH_QUEUES(nsintrq);
687 #endif
688 #ifdef ISO
689 	IF_DETACH_QUEUES(clnlintrq);
690 #endif
691 #ifdef CCITT
692 	IF_DETACH_QUEUES(llcintrq);
693 	IF_DETACH_QUEUES(hdintrq);
694 #endif
695 #ifdef NATM
696 	IF_DETACH_QUEUES(natmintrq);
697 #endif
698 #ifdef DECNET
699 	IF_DETACH_QUEUES(decnetintrq);
700 #endif
701 #undef IF_DETACH_QUEUES
702 
703 	splx(s);
704 }
705 
706 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
707     defined(ISO) || defined(CCITT) || defined(NATM) || defined(DECNET)
708 static void
709 if_detach_queues(ifp, q)
710 	struct ifnet *ifp;
711 	struct ifqueue *q;
712 {
713 	struct mbuf *m, *prev, *next;
714 
715 	prev = NULL;
716 	for (m = q->ifq_head; m; m = next) {
717 		next = m->m_nextpkt;
718 #ifdef DIAGNOSTIC
719 		if ((m->m_flags & M_PKTHDR) == 0) {
720 			prev = m;
721 			continue;
722 		}
723 #endif
724 		if (m->m_pkthdr.rcvif != ifp) {
725 			prev = m;
726 			continue;
727 		}
728 
729 		if (prev)
730 			prev->m_nextpkt = m->m_nextpkt;
731 		else
732 			q->ifq_head = m->m_nextpkt;
733 		if (q->ifq_tail == m)
734 			q->ifq_tail = prev;
735 		q->ifq_len--;
736 
737 		m->m_nextpkt = NULL;
738 		m_freem(m);
739 		IF_DROP(q);
740 	}
741 }
742 #endif /* defined(INET) || ... */
743 
744 /*
745  * Callback for a radix tree walk to delete all references to an
746  * ifnet.
747  */
748 int
749 if_rt_walktree(rn, v)
750 	struct radix_node *rn;
751 	void *v;
752 {
753 	struct ifnet *ifp = (struct ifnet *)v;
754 	struct rtentry *rt = (struct rtentry *)rn;
755 	int error;
756 
757 	if (rt->rt_ifp == ifp) {
758 		/* Delete the entry. */
759 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
760 		    rt_mask(rt), rt->rt_flags, NULL);
761 		if (error)
762 			printf("%s: warning: unable to delete rtentry @ %p, "
763 			    "error = %d\n", ifp->if_xname, rt, error);
764 	}
765 	return (0);
766 }
767 
768 /*
769  * Create a clone network interface.
770  */
771 int
772 if_clone_create(name)
773 	const char *name;
774 {
775 	struct if_clone *ifc;
776 	int unit;
777 
778 	ifc = if_clone_lookup(name, &unit);
779 	if (ifc == NULL)
780 		return (EINVAL);
781 
782 	if (ifunit(name) != NULL)
783 		return (EEXIST);
784 
785 	return ((*ifc->ifc_create)(ifc, unit));
786 }
787 
788 /*
789  * Destroy a clone network interface.
790  */
791 int
792 if_clone_destroy(name)
793 	const char *name;
794 {
795 	struct if_clone *ifc;
796 	struct ifnet *ifp;
797 
798 	ifc = if_clone_lookup(name, NULL);
799 	if (ifc == NULL)
800 		return (EINVAL);
801 
802 	ifp = ifunit(name);
803 	if (ifp == NULL)
804 		return (ENXIO);
805 
806 	if (ifc->ifc_destroy == NULL)
807 		return (EOPNOTSUPP);
808 
809 	(*ifc->ifc_destroy)(ifp);
810 	return (0);
811 }
812 
813 /*
814  * Look up a network interface cloner.
815  */
816 struct if_clone *
817 if_clone_lookup(name, unitp)
818 	const char *name;
819 	int *unitp;
820 {
821 	struct if_clone *ifc;
822 	const char *cp;
823 	int unit;
824 
825 	/* separate interface name from unit */
826 	for (cp = name;
827 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
828 	    cp++)
829 		continue;
830 
831 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
832 		return (NULL);	/* No name or unit number */
833 
834 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
835 		if (strlen(ifc->ifc_name) == cp - name &&
836 		    !strncmp(name, ifc->ifc_name, cp - name))
837 			break;
838 	}
839 
840 	if (ifc == NULL)
841 		return (NULL);
842 
843 	unit = 0;
844 	while (cp - name < IFNAMSIZ && *cp) {
845 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
846 			/* Bogus unit number. */
847 			return (NULL);
848 		}
849 		unit = (unit * 10) + (*cp++ - '0');
850 	}
851 
852 	if (unitp != NULL)
853 		*unitp = unit;
854 	return (ifc);
855 }
856 
857 /*
858  * Register a network interface cloner.
859  */
860 void
861 if_clone_attach(ifc)
862 	struct if_clone *ifc;
863 {
864 
865 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
866 	if_cloners_count++;
867 }
868 
869 /*
870  * Unregister a network interface cloner.
871  */
872 void
873 if_clone_detach(ifc)
874 	struct if_clone *ifc;
875 {
876 
877 	LIST_REMOVE(ifc, ifc_list);
878 	if_cloners_count--;
879 }
880 
881 /*
882  * Provide list of interface cloners to userspace.
883  */
884 int
885 if_clone_list(ifcr)
886 	struct if_clonereq *ifcr;
887 {
888 	char outbuf[IFNAMSIZ], *dst;
889 	struct if_clone *ifc;
890 	int count, error = 0;
891 
892 	ifcr->ifcr_total = if_cloners_count;
893 	if ((dst = ifcr->ifcr_buffer) == NULL) {
894 		/* Just asking how many there are. */
895 		return (0);
896 	}
897 
898 	if (ifcr->ifcr_count < 0)
899 		return (EINVAL);
900 
901 	count = (if_cloners_count < ifcr->ifcr_count) ?
902 	    if_cloners_count : ifcr->ifcr_count;
903 
904 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
905 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
906 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
907 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
908 		error = copyout(outbuf, dst, IFNAMSIZ);
909 		if (error)
910 			break;
911 	}
912 
913 	return (error);
914 }
915 
916 /*
917  * Locate an interface based on a complete address.
918  */
919 /*ARGSUSED*/
920 struct ifaddr *
921 ifa_ifwithaddr(addr)
922 	const struct sockaddr *addr;
923 {
924 	struct ifnet *ifp;
925 	struct ifaddr *ifa;
926 
927 #define	equal(a1, a2) \
928   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
929 
930 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
931 	     ifp = TAILQ_NEXT(ifp, if_list)) {
932 		if (ifp->if_output == if_nulloutput)
933 			continue;
934 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
935 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
936 			if (ifa->ifa_addr->sa_family != addr->sa_family)
937 				continue;
938 			if (equal(addr, ifa->ifa_addr))
939 				return (ifa);
940 			if ((ifp->if_flags & IFF_BROADCAST) &&
941 			    ifa->ifa_broadaddr &&
942 			    /* IP6 doesn't have broadcast */
943 			    ifa->ifa_broadaddr->sa_len != 0 &&
944 			    equal(ifa->ifa_broadaddr, addr))
945 				return (ifa);
946 		}
947 	}
948 	return (NULL);
949 }
950 
951 /*
952  * Locate the point to point interface with a given destination address.
953  */
954 /*ARGSUSED*/
955 struct ifaddr *
956 ifa_ifwithdstaddr(addr)
957 	const struct sockaddr *addr;
958 {
959 	struct ifnet *ifp;
960 	struct ifaddr *ifa;
961 
962 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
963 	     ifp = TAILQ_NEXT(ifp, if_list)) {
964 		if (ifp->if_output == if_nulloutput)
965 			continue;
966 		if (ifp->if_flags & IFF_POINTOPOINT) {
967 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
968 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
969 				if (ifa->ifa_addr->sa_family !=
970 				      addr->sa_family ||
971 				    ifa->ifa_dstaddr == NULL)
972 					continue;
973 				if (equal(addr, ifa->ifa_dstaddr))
974 					return (ifa);
975 			}
976 		}
977 	}
978 	return (NULL);
979 }
980 
981 /*
982  * Find an interface on a specific network.  If many, choice
983  * is most specific found.
984  */
985 struct ifaddr *
986 ifa_ifwithnet(addr)
987 	const struct sockaddr *addr;
988 {
989 	struct ifnet *ifp;
990 	struct ifaddr *ifa;
991 	const struct sockaddr_dl *sdl;
992 	struct ifaddr *ifa_maybe = 0;
993 	u_int af = addr->sa_family;
994 	char *addr_data = addr->sa_data, *cplim;
995 
996 	if (af == AF_LINK) {
997 		sdl = (struct sockaddr_dl *)addr;
998 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
999 		    ifindex2ifnet[sdl->sdl_index] &&
1000 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1001 			return (ifnet_addrs[sdl->sdl_index]);
1002 	}
1003 #ifdef NETATALK
1004 	if (af == AF_APPLETALK) {
1005 		const struct sockaddr_at *sat, *sat2;
1006 		sat = (struct sockaddr_at *)addr;
1007 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1008 		     ifp = TAILQ_NEXT(ifp, if_list)) {
1009 			if (ifp->if_output == if_nulloutput)
1010 				continue;
1011 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
1012 			if (ifa == NULL)
1013 				continue;
1014 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1015 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1016 				return (ifa); /* exact match */
1017 			if (ifa_maybe == NULL) {
1018 				/* else keep the if with the right range */
1019 				ifa_maybe = ifa;
1020 			}
1021 		}
1022 		return (ifa_maybe);
1023 	}
1024 #endif
1025 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1026 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1027 		if (ifp->if_output == if_nulloutput)
1028 			continue;
1029 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1030 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1031 			char *cp, *cp2, *cp3;
1032 
1033 			if (ifa->ifa_addr->sa_family != af ||
1034 			    ifa->ifa_netmask == 0)
1035  next:				continue;
1036 			cp = addr_data;
1037 			cp2 = ifa->ifa_addr->sa_data;
1038 			cp3 = ifa->ifa_netmask->sa_data;
1039 			cplim = (char *)ifa->ifa_netmask +
1040 			    ifa->ifa_netmask->sa_len;
1041 			while (cp3 < cplim) {
1042 				if ((*cp++ ^ *cp2++) & *cp3++) {
1043 					/* want to continue for() loop */
1044 					goto next;
1045 				}
1046 			}
1047 			if (ifa_maybe == 0 ||
1048 			    rn_refines((caddr_t)ifa->ifa_netmask,
1049 			    (caddr_t)ifa_maybe->ifa_netmask))
1050 				ifa_maybe = ifa;
1051 		}
1052 	}
1053 	return (ifa_maybe);
1054 }
1055 
1056 /*
1057  * Find the interface of the addresss.
1058  */
1059 struct ifaddr *
1060 ifa_ifwithladdr(addr)
1061 	const struct sockaddr *addr;
1062 {
1063 	struct ifaddr *ia;
1064 
1065 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1066 	    (ia = ifa_ifwithnet(addr)))
1067 		return (ia);
1068 	return (NULL);
1069 }
1070 
1071 /*
1072  * Find an interface using a specific address family
1073  */
1074 struct ifaddr *
1075 ifa_ifwithaf(af)
1076 	int af;
1077 {
1078 	struct ifnet *ifp;
1079 	struct ifaddr *ifa;
1080 
1081 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1082 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1083 		if (ifp->if_output == if_nulloutput)
1084 			continue;
1085 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1086 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1087 			if (ifa->ifa_addr->sa_family == af)
1088 				return (ifa);
1089 		}
1090 	}
1091 	return (NULL);
1092 }
1093 
1094 /*
1095  * Find an interface address specific to an interface best matching
1096  * a given address.
1097  */
1098 struct ifaddr *
1099 ifaof_ifpforaddr(addr, ifp)
1100 	const struct sockaddr *addr;
1101 	struct ifnet *ifp;
1102 {
1103 	struct ifaddr *ifa;
1104 	const char *cp, *cp2, *cp3;
1105 	const char *cplim;
1106 	struct ifaddr *ifa_maybe = 0;
1107 	u_int af = addr->sa_family;
1108 
1109 	if (ifp->if_output == if_nulloutput)
1110 		return (NULL);
1111 
1112 	if (af >= AF_MAX)
1113 		return (NULL);
1114 
1115 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1116 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1117 		if (ifa->ifa_addr->sa_family != af)
1118 			continue;
1119 		ifa_maybe = ifa;
1120 		if (ifa->ifa_netmask == 0) {
1121 			if (equal(addr, ifa->ifa_addr) ||
1122 			    (ifa->ifa_dstaddr &&
1123 			     equal(addr, ifa->ifa_dstaddr)))
1124 				return (ifa);
1125 			continue;
1126 		}
1127 		cp = addr->sa_data;
1128 		cp2 = ifa->ifa_addr->sa_data;
1129 		cp3 = ifa->ifa_netmask->sa_data;
1130 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1131 		for (; cp3 < cplim; cp3++) {
1132 			if ((*cp++ ^ *cp2++) & *cp3)
1133 				break;
1134 		}
1135 		if (cp3 == cplim)
1136 			return (ifa);
1137 	}
1138 	return (ifa_maybe);
1139 }
1140 
1141 /*
1142  * Default action when installing a route with a Link Level gateway.
1143  * Lookup an appropriate real ifa to point to.
1144  * This should be moved to /sys/net/link.c eventually.
1145  */
1146 void
1147 link_rtrequest(cmd, rt, info)
1148 	int cmd;
1149 	struct rtentry *rt;
1150 	struct rt_addrinfo *info;
1151 {
1152 	struct ifaddr *ifa;
1153 	struct sockaddr *dst;
1154 	struct ifnet *ifp;
1155 
1156 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1157 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1158 		return;
1159 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1160 		IFAFREE(rt->rt_ifa);
1161 		rt->rt_ifa = ifa;
1162 		IFAREF(ifa);
1163 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1164 			ifa->ifa_rtrequest(cmd, rt, info);
1165 	}
1166 }
1167 
1168 /*
1169  * Mark an interface down and notify protocols of
1170  * the transition.
1171  * NOTE: must be called at splsoftnet or equivalent.
1172  */
1173 void
1174 if_down(ifp)
1175 	struct ifnet *ifp;
1176 {
1177 	struct ifaddr *ifa;
1178 
1179 	ifp->if_flags &= ~IFF_UP;
1180 	microtime(&ifp->if_lastchange);
1181 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1182 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1183 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1184 	IFQ_PURGE(&ifp->if_snd);
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(ifp)
1195 	struct ifnet *ifp;
1196 {
1197 #ifdef notyet
1198 	struct ifaddr *ifa;
1199 #endif
1200 
1201 	ifp->if_flags |= IFF_UP;
1202 	microtime(&ifp->if_lastchange);
1203 #ifdef notyet
1204 	/* this has no effect on IP, and will kill all ISO connections XXX */
1205 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1206 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1207 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1208 #endif
1209 	rt_ifmsg(ifp);
1210 #ifdef INET6
1211 	in6_if_up(ifp);
1212 #endif
1213 }
1214 
1215 /*
1216  * Handle interface watchdog timer routines.  Called
1217  * from softclock, we decrement timers (if set) and
1218  * call the appropriate interface routine on expiration.
1219  */
1220 void
1221 if_slowtimo(arg)
1222 	void *arg;
1223 {
1224 	struct ifnet *ifp;
1225 	int s = splnet();
1226 
1227 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1228 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1229 		if (ifp->if_timer == 0 || --ifp->if_timer)
1230 			continue;
1231 		if (ifp->if_watchdog)
1232 			(*ifp->if_watchdog)(ifp);
1233 	}
1234 	splx(s);
1235 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1236 	    if_slowtimo, NULL);
1237 }
1238 
1239 /*
1240  * Set/clear promiscuous mode on interface ifp based on the truth value
1241  * of pswitch.  The calls are reference counted so that only the first
1242  * "on" request actually has an effect, as does the final "off" request.
1243  * Results are undefined if the "off" and "on" requests are not matched.
1244  */
1245 int
1246 ifpromisc(ifp, pswitch)
1247 	struct ifnet *ifp;
1248 	int pswitch;
1249 {
1250 	int pcount, ret;
1251 	short flags;
1252 	struct ifreq ifr;
1253 
1254 	pcount = ifp->if_pcount;
1255 	flags = ifp->if_flags;
1256 	if (pswitch) {
1257 		/*
1258 		 * Allow the device to be "placed" into promiscuous
1259 		 * mode even if it is not configured up.  It will
1260 		 * consult IFF_PROMISC when it is is brought up.
1261 		 */
1262 		if (ifp->if_pcount++ != 0)
1263 			return (0);
1264 		ifp->if_flags |= IFF_PROMISC;
1265 		if ((ifp->if_flags & IFF_UP) == 0)
1266 			return (0);
1267 	} else {
1268 		if (--ifp->if_pcount > 0)
1269 			return (0);
1270 		ifp->if_flags &= ~IFF_PROMISC;
1271 		/*
1272 		 * If the device is not configured up, we should not need to
1273 		 * turn off promiscuous mode (device should have turned it
1274 		 * off when interface went down; and will look at IFF_PROMISC
1275 		 * again next time interface comes up).
1276 		 */
1277 		if ((ifp->if_flags & IFF_UP) == 0)
1278 			return (0);
1279 	}
1280 	memset(&ifr, 0, sizeof(ifr));
1281 	ifr.ifr_flags = ifp->if_flags;
1282 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1283 	/* Restore interface state if not successful. */
1284 	if (ret != 0) {
1285 		ifp->if_pcount = pcount;
1286 		ifp->if_flags = flags;
1287 	}
1288 	return (ret);
1289 }
1290 
1291 /*
1292  * Map interface name to
1293  * interface structure pointer.
1294  */
1295 struct ifnet *
1296 ifunit(name)
1297 	const char *name;
1298 {
1299 	struct ifnet *ifp;
1300 	const char *cp = name;
1301 	u_int unit = 0;
1302 	u_int i;
1303 
1304 	/*
1305 	 * If the entire name is a number, treat it as an ifindex.
1306 	 */
1307 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1308 		unit = unit * 10 + (*cp - '0');
1309 	}
1310 
1311 	/*
1312 	 * If the number took all of the name, then it's a valid ifindex.
1313 	 */
1314 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1315 		if (unit >= if_indexlim)
1316 			return (NULL);
1317 		ifp = ifindex2ifnet[unit];
1318 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1319 			return (NULL);
1320 		return (ifp);
1321 	}
1322 
1323 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1324 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1325 		if (ifp->if_output == if_nulloutput)
1326 			continue;
1327 	 	if (strcmp(ifp->if_xname, name) == 0)
1328 			return (ifp);
1329 	}
1330 	return (NULL);
1331 }
1332 
1333 /*
1334  * Interface ioctls.
1335  */
1336 int
1337 ifioctl(so, cmd, data, p)
1338 	struct socket *so;
1339 	u_long cmd;
1340 	caddr_t data;
1341 	struct proc *p;
1342 {
1343 	struct ifnet *ifp;
1344 	struct ifreq *ifr;
1345 	struct ifcapreq *ifcr;
1346 	struct ifdatareq *ifdr;
1347 	int s, error = 0;
1348 	short oif_flags;
1349 	int prived_error;
1350 
1351 	if (p)
1352 		prived_error = suser(p->p_ucred, &p->p_acflag);
1353 	else
1354 		prived_error = 0;
1355 
1356 	switch (cmd) {
1357 
1358 	case SIOCGIFCONF:
1359 	case OSIOCGIFCONF:
1360 		return (ifconf(cmd, data));
1361 	}
1362 	ifr = (struct ifreq *)data;
1363 	ifcr = (struct ifcapreq *)data;
1364 	ifdr = (struct ifdatareq *)data;
1365 
1366 	switch (cmd) {
1367 	case SIOCIFCREATE:
1368 	case SIOCIFDESTROY:
1369 		if (prived_error)
1370 			return (prived_error);
1371 		return ((cmd == SIOCIFCREATE) ?
1372 			if_clone_create(ifr->ifr_name) :
1373 			if_clone_destroy(ifr->ifr_name));
1374 
1375 	case SIOCIFGCLONERS:
1376 		return (if_clone_list((struct if_clonereq *)data));
1377 	}
1378 
1379 	ifp = ifunit(ifr->ifr_name);
1380 	if (ifp == 0)
1381 		return (ENXIO);
1382 	oif_flags = ifp->if_flags;
1383 	switch (cmd) {
1384 
1385 	case SIOCGIFFLAGS:
1386 		ifr->ifr_flags = ifp->if_flags;
1387 		break;
1388 
1389 	case SIOCGIFMETRIC:
1390 		ifr->ifr_metric = ifp->if_metric;
1391 		break;
1392 
1393 	case SIOCGIFMTU:
1394 		ifr->ifr_mtu = ifp->if_mtu;
1395 		break;
1396 
1397 	case SIOCGIFDLT:
1398 		ifr->ifr_dlt = ifp->if_dlt;
1399 		break;
1400 
1401 	case SIOCSIFFLAGS:
1402 		if (prived_error != 0)
1403 			return (prived_error);
1404 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1405 			s = splnet();
1406 			if_down(ifp);
1407 			splx(s);
1408 		}
1409 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1410 			s = splnet();
1411 			if_up(ifp);
1412 			splx(s);
1413 		}
1414 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1415 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1416 		if (ifp->if_ioctl)
1417 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1418 		break;
1419 
1420 	case SIOCGIFCAP:
1421 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1422 		ifcr->ifcr_capenable = ifp->if_capenable;
1423 		break;
1424 
1425 	case SIOCSIFCAP:
1426 		if (prived_error != 0)
1427 			return (prived_error);
1428 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1429 			return (EINVAL);
1430 		if (ifp->if_ioctl == NULL)
1431 			return (EOPNOTSUPP);
1432 
1433 		/* Must prevent race with packet reception here. */
1434 		s = splnet();
1435 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1436 			struct ifreq ifrq;
1437 
1438 			ifrq.ifr_flags = ifp->if_flags;
1439 			ifp->if_capenable = ifcr->ifcr_capenable;
1440 
1441 			/* Pre-compute the checksum flags mask. */
1442 			ifp->if_csum_flags_tx = 0;
1443 			ifp->if_csum_flags_rx = 0;
1444 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1445 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1446 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1447 			}
1448 
1449 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1450 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1451 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1452 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1453 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1454 
1455 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1456 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1457 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1458 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1459 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1460 
1461 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1462 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1463 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1464 			}
1465 
1466 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1467 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1468 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1469 			}
1470 
1471 			/*
1472 			 * Only kick the interface if it's up.  If it's
1473 			 * not up now, it will notice the cap enables
1474 			 * when it is brought up later.
1475 			 */
1476 			if (ifp->if_flags & IFF_UP)
1477 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1478 				    (caddr_t) &ifrq);
1479 		}
1480 		splx(s);
1481 		break;
1482 
1483 	case SIOCSIFMETRIC:
1484 		if (prived_error != 0)
1485 			return (prived_error);
1486 		ifp->if_metric = ifr->ifr_metric;
1487 		break;
1488 
1489 	case SIOCGIFDATA:
1490 		ifdr->ifdr_data = ifp->if_data;
1491 		break;
1492 
1493 	case SIOCZIFDATA:
1494 		if (prived_error != 0)
1495 			return (prived_error);
1496 		ifdr->ifdr_data = ifp->if_data;
1497 		/*
1498 		 * Assumes that the volatile counters that can be
1499 		 * zero'ed are at the end of if_data.
1500 		 */
1501 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1502 		    offsetof(struct if_data, ifi_ipackets));
1503 		break;
1504 
1505 	case SIOCSIFMTU:
1506 	{
1507 		u_long oldmtu = ifp->if_mtu;
1508 
1509 		if (prived_error)
1510 			return (prived_error);
1511 		if (ifp->if_ioctl == NULL)
1512 			return (EOPNOTSUPP);
1513 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1514 
1515 		/*
1516 		 * If the link MTU changed, do network layer specific procedure.
1517 		 */
1518 		if (ifp->if_mtu != oldmtu) {
1519 #ifdef INET6
1520 			nd6_setmtu(ifp);
1521 #endif
1522 		}
1523 		break;
1524 	}
1525 	case SIOCSIFPHYADDR:
1526 	case SIOCDIFPHYADDR:
1527 #ifdef INET6
1528 	case SIOCSIFPHYADDR_IN6:
1529 #endif
1530 	case SIOCSLIFPHYADDR:
1531 	case SIOCADDMULTI:
1532 	case SIOCDELMULTI:
1533 	case SIOCSIFMEDIA:
1534 		if (prived_error != 0)
1535 			return (prived_error);
1536 		/* FALLTHROUGH */
1537 	case SIOCGIFPSRCADDR:
1538 	case SIOCGIFPDSTADDR:
1539 	case SIOCGLIFPHYADDR:
1540 	case SIOCGIFMEDIA:
1541 		if (ifp->if_ioctl == 0)
1542 			return (EOPNOTSUPP);
1543 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1544 		break;
1545 
1546 	case SIOCSDRVSPEC:
1547 	case SIOCS80211NWID:
1548 	case SIOCS80211NWKEY:
1549 	case SIOCS80211POWER:
1550 	case SIOCS80211BSSID:
1551 	case SIOCS80211CHANNEL:
1552 		/* XXX:  need to pass proc pointer through to driver... */
1553 		if (prived_error != 0)
1554 			return (prived_error);
1555 	/* FALLTHROUGH */
1556 	default:
1557 		if (so->so_proto == 0)
1558 			return (EOPNOTSUPP);
1559 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(COMPAT_ULTRIX) && !defined(LKM)
1560 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1561 		    (struct mbuf *)cmd, (struct mbuf *)data,
1562 		    (struct mbuf *)ifp, p));
1563 #else
1564 	    {
1565 		int ocmd = cmd;
1566 
1567 		switch (cmd) {
1568 
1569 		case SIOCSIFADDR:
1570 		case SIOCSIFDSTADDR:
1571 		case SIOCSIFBRDADDR:
1572 		case SIOCSIFNETMASK:
1573 #if BYTE_ORDER != BIG_ENDIAN
1574 			if (ifr->ifr_addr.sa_family == 0 &&
1575 			    ifr->ifr_addr.sa_len < 16) {
1576 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1577 				ifr->ifr_addr.sa_len = 16;
1578 			}
1579 #else
1580 			if (ifr->ifr_addr.sa_len == 0)
1581 				ifr->ifr_addr.sa_len = 16;
1582 #endif
1583 			break;
1584 
1585 		case OSIOCGIFADDR:
1586 			cmd = SIOCGIFADDR;
1587 			break;
1588 
1589 		case OSIOCGIFDSTADDR:
1590 			cmd = SIOCGIFDSTADDR;
1591 			break;
1592 
1593 		case OSIOCGIFBRDADDR:
1594 			cmd = SIOCGIFBRDADDR;
1595 			break;
1596 
1597 		case OSIOCGIFNETMASK:
1598 			cmd = SIOCGIFNETMASK;
1599 		}
1600 
1601 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1602 		    (struct mbuf *)cmd, (struct mbuf *)data,
1603 		    (struct mbuf *)ifp, p));
1604 
1605 		switch (ocmd) {
1606 		case OSIOCGIFADDR:
1607 		case OSIOCGIFDSTADDR:
1608 		case OSIOCGIFBRDADDR:
1609 		case OSIOCGIFNETMASK:
1610 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1611 		}
1612 	    }
1613 #endif /* COMPAT_43 */
1614 		break;
1615 	}
1616 
1617 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1618 #ifdef INET6
1619 		if ((ifp->if_flags & IFF_UP) != 0) {
1620 			s = splnet();
1621 			in6_if_up(ifp);
1622 			splx(s);
1623 		}
1624 #endif
1625 	}
1626 
1627 	return (error);
1628 }
1629 
1630 /*
1631  * Return interface configuration
1632  * of system.  List may be used
1633  * in later ioctl's (above) to get
1634  * other information.
1635  */
1636 /*ARGSUSED*/
1637 int
1638 ifconf(cmd, data)
1639 	u_long cmd;
1640 	caddr_t data;
1641 {
1642 	struct ifconf *ifc = (struct ifconf *)data;
1643 	struct ifnet *ifp;
1644 	struct ifaddr *ifa;
1645 	struct ifreq ifr, *ifrp;
1646 	int space = ifc->ifc_len, error = 0;
1647 	const int sz = (int)sizeof(ifr);
1648 	int sign;
1649 
1650 	if ((ifrp = ifc->ifc_req) == NULL) {
1651 		space = 0;
1652 		sign = -1;
1653 	} else {
1654 		sign = 1;
1655 	}
1656 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1657 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1658 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1659 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1660 			if (ifrp != NULL && space >= sz) {
1661 				error = copyout(&ifr, ifrp, sz);
1662 				if (error)
1663 					break;
1664 				ifrp++;
1665 			}
1666 			space -= sizeof(ifr) * sign;
1667 			continue;
1668 		}
1669 
1670 		for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1671 			struct sockaddr *sa = ifa->ifa_addr;
1672 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX)
1673 			if (cmd == OSIOCGIFCONF) {
1674 				struct osockaddr *osa =
1675 					 (struct osockaddr *)&ifr.ifr_addr;
1676 				/*
1677 				 * If it does not fit, we don't bother with it
1678 				 */
1679 				if (sa->sa_len > sizeof(*osa))
1680 					continue;
1681 				ifr.ifr_addr = *sa;
1682 				osa->sa_family = sa->sa_family;
1683 				if (ifrp != NULL && space >= sz) {
1684 					error = copyout(&ifr, ifrp, sz);
1685 					ifrp++;
1686 				}
1687 			} else
1688 #endif
1689 			if (sa->sa_len <= sizeof(*sa)) {
1690 				ifr.ifr_addr = *sa;
1691 				if (ifrp != NULL && space >= sz) {
1692 					error = copyout(&ifr, ifrp, sz);
1693 					ifrp++;
1694 				}
1695 			} else {
1696 				space -= (sa->sa_len - sizeof(*sa)) * sign;
1697 				if (ifrp != NULL && space >= sz) {
1698 					error = copyout(&ifr, ifrp,
1699 					    sizeof(ifr.ifr_name));
1700 					if (error == 0) {
1701 						error = copyout(sa,
1702 						    &ifrp->ifr_addr,
1703 						    sa->sa_len);
1704 					}
1705 					ifrp = (struct ifreq *)
1706 						(sa->sa_len +
1707 						 (caddr_t)&ifrp->ifr_addr);
1708 				}
1709 			}
1710 			if (error)
1711 				break;
1712 			space -= sz * sign;
1713 		}
1714 	}
1715 	if (ifrp != NULL)
1716 		ifc->ifc_len -= space;
1717 	else
1718 		ifc->ifc_len = space;
1719 	return (error);
1720 }
1721 
1722 #if defined(INET) || defined(INET6)
1723 static void
1724 sysctl_net_ifq_setup(struct sysctllog **clog,
1725 		     int pf, const char *pfname,
1726 		     int ipn, const char *ipname,
1727 		     int qid, struct ifqueue *ifq)
1728 {
1729 
1730 	sysctl_createv(clog, 0, NULL, NULL,
1731 		       CTLFLAG_PERMANENT,
1732 		       CTLTYPE_NODE, "net", NULL,
1733 		       NULL, 0, NULL, 0,
1734 		       CTL_NET, CTL_EOL);
1735 	sysctl_createv(clog, 0, NULL, NULL,
1736 		       CTLFLAG_PERMANENT,
1737 		       CTLTYPE_NODE, pfname, NULL,
1738 		       NULL, 0, NULL, 0,
1739 		       CTL_NET, pf, CTL_EOL);
1740 	sysctl_createv(clog, 0, NULL, NULL,
1741 		       CTLFLAG_PERMANENT,
1742 		       CTLTYPE_NODE, ipname, NULL,
1743 		       NULL, 0, NULL, 0,
1744 		       CTL_NET, pf, ipn, CTL_EOL);
1745 	sysctl_createv(clog, 0, NULL, NULL,
1746 		       CTLFLAG_PERMANENT,
1747 		       CTLTYPE_NODE, "ifq",
1748 		       SYSCTL_DESCR("Protocol input queue controls"),
1749 		       NULL, 0, NULL, 0,
1750 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1751 
1752 	sysctl_createv(clog, 0, NULL, NULL,
1753 		       CTLFLAG_PERMANENT,
1754 		       CTLTYPE_INT, "len",
1755 		       SYSCTL_DESCR("Current input queue length"),
1756 		       NULL, 0, &ifq->ifq_len, 0,
1757 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1758 	sysctl_createv(clog, 0, NULL, NULL,
1759 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1760 		       CTLTYPE_INT, "maxlen",
1761 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1762 		       NULL, 0, &ifq->ifq_maxlen, 0,
1763 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1764 #ifdef notyet
1765 	sysctl_createv(clog, 0, NULL, NULL,
1766 		       CTLFLAG_PERMANENT,
1767 		       CTLTYPE_INT, "peak",
1768 		       SYSCTL_DESCR("Highest input queue length"),
1769 		       NULL, 0, &ifq->ifq_peak, 0,
1770 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1771 #endif
1772 	sysctl_createv(clog, 0, NULL, NULL,
1773 		       CTLFLAG_PERMANENT,
1774 		       CTLTYPE_INT, "drops",
1775 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1776 		       NULL, 0, &ifq->ifq_drops, 0,
1777 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1778 }
1779 
1780 #ifdef INET
1781 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1782 	     "sysctl net.inet.ip.ifq subtree setup")
1783 {
1784 	extern struct ifqueue ipintrq;
1785 
1786 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1787 			     IPCTL_IFQ, &ipintrq);
1788 }
1789 #endif /* INET */
1790 
1791 #ifdef INET6
1792 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1793 	     "sysctl net.inet6.ip6.ifq subtree setup")
1794 {
1795 	extern struct ifqueue ip6intrq;
1796 
1797 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1798 			     IPV6CTL_IFQ, &ip6intrq);
1799 }
1800 #endif /* INET6 */
1801 #endif /* INET || INET6 */
1802