xref: /netbsd-src/sys/net/if.c (revision 4b896b232495b7a9b8b94a1cf1e21873296d53b8)
1 /*	$NetBSD: if.c,v 1.142 2004/05/25 04:33:59 atatat 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 1.142 2004/05/25 04:33:59 atatat 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 
141 #ifdef INET6
142 #include <netinet/in.h>
143 #include <netinet6/in6_var.h>
144 #include <netinet6/nd6.h>
145 #endif
146 
147 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
148 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
149 
150 int	ifqmaxlen = IFQ_MAXLEN;
151 struct	callout if_slowtimo_ch;
152 
153 int netisr;			/* scheduling bits for network */
154 
155 int if_rt_walktree __P((struct radix_node *, void *));
156 
157 struct if_clone *if_clone_lookup __P((const char *, int *));
158 int if_clone_list __P((struct if_clonereq *));
159 
160 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
161 int if_cloners_count;
162 
163 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
164     defined(ISO) || defined(CCITT) || defined(NATM)
165 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
166 #endif
167 
168 /*
169  * Network interface utility routines.
170  *
171  * Routines with ifa_ifwith* names take sockaddr *'s as
172  * parameters.
173  */
174 void
175 ifinit()
176 {
177 
178 	callout_init(&if_slowtimo_ch);
179 	if_slowtimo(NULL);
180 }
181 
182 /*
183  * Null routines used while an interface is going away.  These routines
184  * just return an error.
185  */
186 
187 int
188 if_nulloutput(ifp, m, so, rt)
189 	struct ifnet *ifp;
190 	struct mbuf *m;
191 	struct sockaddr *so;
192 	struct rtentry *rt;
193 {
194 
195 	return (ENXIO);
196 }
197 
198 void
199 if_nullinput(ifp, m)
200 	struct ifnet *ifp;
201 	struct mbuf *m;
202 {
203 
204 	/* Nothing. */
205 }
206 
207 void
208 if_nullstart(ifp)
209 	struct ifnet *ifp;
210 {
211 
212 	/* Nothing. */
213 }
214 
215 int
216 if_nullioctl(ifp, cmd, data)
217 	struct ifnet *ifp;
218 	u_long cmd;
219 	caddr_t data;
220 {
221 
222 	return (ENXIO);
223 }
224 
225 int
226 if_nullinit(ifp)
227 	struct ifnet *ifp;
228 {
229 
230 	return (ENXIO);
231 }
232 
233 void
234 if_nullstop(ifp, disable)
235 	struct ifnet *ifp;
236 	int disable;
237 {
238 
239 	/* Nothing. */
240 }
241 
242 void
243 if_nullwatchdog(ifp)
244 	struct ifnet *ifp;
245 {
246 
247 	/* Nothing. */
248 }
249 
250 void
251 if_nulldrain(ifp)
252 	struct ifnet *ifp;
253 {
254 
255 	/* Nothing. */
256 }
257 
258 static u_int if_index = 1;
259 struct ifnet_head ifnet;
260 size_t if_indexlim = 0;
261 struct ifaddr **ifnet_addrs = NULL;
262 struct ifnet **ifindex2ifnet = NULL;
263 
264 /*
265  * Allocate the link level name for the specified interface.  This
266  * is an attachment helper.  It must be called after ifp->if_addrlen
267  * is initialized, which may not be the case when if_attach() is
268  * called.
269  */
270 void
271 if_alloc_sadl(struct ifnet *ifp)
272 {
273 	unsigned socksize, ifasize;
274 	int namelen, masklen;
275 	struct sockaddr_dl *sdl;
276 	struct ifaddr *ifa;
277 
278 	/*
279 	 * If the interface already has a link name, release it
280 	 * now.  This is useful for interfaces that can change
281 	 * link types, and thus switch link names often.
282 	 */
283 	if (ifp->if_sadl != NULL)
284 		if_free_sadl(ifp);
285 
286 	namelen = strlen(ifp->if_xname);
287 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
288 	socksize = masklen + ifp->if_addrlen;
289 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
290 	if (socksize < sizeof(*sdl))
291 		socksize = sizeof(*sdl);
292 	socksize = ROUNDUP(socksize);
293 	ifasize = sizeof(*ifa) + 2 * socksize;
294 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
295 	memset((caddr_t)ifa, 0, ifasize);
296 	sdl = (struct sockaddr_dl *)(ifa + 1);
297 	sdl->sdl_len = socksize;
298 	sdl->sdl_family = AF_LINK;
299 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
300 	sdl->sdl_nlen = namelen;
301 	sdl->sdl_alen = ifp->if_addrlen;
302 	sdl->sdl_index = ifp->if_index;
303 	sdl->sdl_type = ifp->if_type;
304 	ifnet_addrs[ifp->if_index] = ifa;
305 	IFAREF(ifa);
306 	ifa->ifa_ifp = ifp;
307 	ifa->ifa_rtrequest = link_rtrequest;
308 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
309 	IFAREF(ifa);
310 	ifa->ifa_addr = (struct sockaddr *)sdl;
311 	ifp->if_sadl = sdl;
312 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
313 	ifa->ifa_netmask = (struct sockaddr *)sdl;
314 	sdl->sdl_len = masklen;
315 	while (namelen != 0)
316 		sdl->sdl_data[--namelen] = 0xff;
317 }
318 
319 /*
320  * Free the link level name for the specified interface.  This is
321  * a detach helper.  This is called from if_detach() or from
322  * link layer type specific detach functions.
323  */
324 void
325 if_free_sadl(struct ifnet *ifp)
326 {
327 	struct ifaddr *ifa;
328 	int s;
329 
330 	ifa = ifnet_addrs[ifp->if_index];
331 	if (ifa == NULL) {
332 		KASSERT(ifp->if_sadl == NULL);
333 		return;
334 	}
335 
336 	KASSERT(ifp->if_sadl != NULL);
337 
338 	s = splnet();
339 	rtinit(ifa, RTM_DELETE, 0);
340 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
341 	IFAFREE(ifa);
342 
343 	ifp->if_sadl = NULL;
344 
345 	ifnet_addrs[ifp->if_index] = NULL;
346 	IFAFREE(ifa);
347 	splx(s);
348 }
349 
350 /*
351  * Attach an interface to the
352  * list of "active" interfaces.
353  */
354 void
355 if_attach(ifp)
356 	struct ifnet *ifp;
357 {
358 	int indexlim = 0;
359 
360 	if (if_indexlim == 0) {
361 		TAILQ_INIT(&ifnet);
362 		if_indexlim = 8;
363 	}
364 	TAILQ_INIT(&ifp->if_addrlist);
365 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
366 	ifp->if_index = if_index;
367 	if (ifindex2ifnet == 0)
368 		if_index++;
369 	else
370 		while (ifp->if_index < if_indexlim &&
371 		    ifindex2ifnet[ifp->if_index] != NULL) {
372 			++if_index;
373 			if (if_index == 0)
374 				if_index = 1;
375 			/*
376 			 * If we hit USHRT_MAX, we skip back to 0 since
377 			 * there are a number of places where the value
378 			 * of if_index or if_index itself is compared
379 			 * to or stored in an unsigned short.  By
380 			 * jumping back, we won't botch those assignments
381 			 * or comparisons.
382 			 */
383 			else if (if_index == USHRT_MAX) {
384 				/*
385 				 * However, if we have to jump back to
386 				 * zero *twice* without finding an empty
387 				 * slot in ifindex2ifnet[], then there
388 				 * there are too many (>65535) interfaces.
389 				 */
390 				if (indexlim++)
391 					panic("too many interfaces");
392 				else
393 					if_index = 1;
394 			}
395 			ifp->if_index = if_index;
396 		}
397 
398 	/*
399 	 * We have some arrays that should be indexed by if_index.
400 	 * since if_index will grow dynamically, they should grow too.
401 	 *	struct ifadd **ifnet_addrs
402 	 *	struct ifnet **ifindex2ifnet
403 	 */
404 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
405 	    ifp->if_index >= if_indexlim) {
406 		size_t m, n, oldlim;
407 		caddr_t q;
408 
409 		oldlim = if_indexlim;
410 		while (ifp->if_index >= if_indexlim)
411 			if_indexlim <<= 1;
412 
413 		/* grow ifnet_addrs */
414 		m = oldlim * sizeof(struct ifaddr *);
415 		n = if_indexlim * sizeof(struct ifaddr *);
416 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
417 		memset(q, 0, n);
418 		if (ifnet_addrs) {
419 			bcopy((caddr_t)ifnet_addrs, q, m);
420 			free((caddr_t)ifnet_addrs, M_IFADDR);
421 		}
422 		ifnet_addrs = (struct ifaddr **)q;
423 
424 		/* grow ifindex2ifnet */
425 		m = oldlim * sizeof(struct ifnet *);
426 		n = if_indexlim * sizeof(struct ifnet *);
427 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
428 		memset(q, 0, n);
429 		if (ifindex2ifnet) {
430 			bcopy((caddr_t)ifindex2ifnet, q, m);
431 			free((caddr_t)ifindex2ifnet, M_IFADDR);
432 		}
433 		ifindex2ifnet = (struct ifnet **)q;
434 	}
435 
436 	ifindex2ifnet[ifp->if_index] = ifp;
437 
438 	/*
439 	 * Link level name is allocated later by a separate call to
440 	 * if_alloc_sadl().
441 	 */
442 
443 	if (ifp->if_snd.ifq_maxlen == 0)
444 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
445 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
446 
447 	ifp->if_link_state = LINK_STATE_UNKNOWN;
448 
449 	ifp->if_capenable = 0;
450 	ifp->if_csum_flags_tx = 0;
451 	ifp->if_csum_flags_rx = 0;
452 
453 #ifdef ALTQ
454 	ifp->if_snd.altq_type = 0;
455 	ifp->if_snd.altq_disc = NULL;
456 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
457 	ifp->if_snd.altq_tbr  = NULL;
458 	ifp->if_snd.altq_ifp  = ifp;
459 #endif
460 
461 #ifdef PFIL_HOOKS
462 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
463 	ifp->if_pfil.ph_ifnet = ifp;
464 	if (pfil_head_register(&ifp->if_pfil) != 0)
465 		printf("%s: WARNING: unable to register pfil hook\n",
466 		    ifp->if_xname);
467 #endif
468 
469 	if (domains)
470 		if_attachdomain1(ifp);
471 
472 	/* Announce the interface. */
473 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
474 }
475 
476 void
477 if_attachdomain()
478 {
479 	struct ifnet *ifp;
480 	int s;
481 
482 	s = splnet();
483 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
484 		if_attachdomain1(ifp);
485 	splx(s);
486 }
487 
488 void
489 if_attachdomain1(ifp)
490 	struct ifnet *ifp;
491 {
492 	struct domain *dp;
493 	int s;
494 
495 	s = splnet();
496 
497 	/* address family dependent data region */
498 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
499 	for (dp = domains; dp; dp = dp->dom_next) {
500 		if (dp->dom_ifattach)
501 			ifp->if_afdata[dp->dom_family] =
502 			    (*dp->dom_ifattach)(ifp);
503 	}
504 
505 	splx(s);
506 }
507 
508 /*
509  * Deactivate an interface.  This points all of the procedure
510  * handles at error stubs.  May be called from interrupt context.
511  */
512 void
513 if_deactivate(ifp)
514 	struct ifnet *ifp;
515 {
516 	int s;
517 
518 	s = splnet();
519 
520 	ifp->if_output	 = if_nulloutput;
521 	ifp->if_input	 = if_nullinput;
522 	ifp->if_start	 = if_nullstart;
523 	ifp->if_ioctl	 = if_nullioctl;
524 	ifp->if_init	 = if_nullinit;
525 	ifp->if_stop	 = if_nullstop;
526 	ifp->if_watchdog = if_nullwatchdog;
527 	ifp->if_drain	 = if_nulldrain;
528 
529 	/* No more packets may be enqueued. */
530 	ifp->if_snd.ifq_maxlen = 0;
531 
532 	splx(s);
533 }
534 
535 /*
536  * Detach an interface from the list of "active" interfaces,
537  * freeing any resources as we go along.
538  *
539  * NOTE: This routine must be called with a valid thread context,
540  * as it may block.
541  */
542 void
543 if_detach(ifp)
544 	struct ifnet *ifp;
545 {
546 	struct socket so;
547 	struct ifaddr *ifa, **ifap;
548 #ifdef IFAREF_DEBUG
549 	struct ifaddr *last_ifa = NULL;
550 #endif
551 	struct domain *dp;
552 	const struct protosw *pr;
553 	struct radix_node_head *rnh;
554 	int s, i, family, purged;
555 
556 	/*
557 	 * XXX It's kind of lame that we have to have the
558 	 * XXX socket structure...
559 	 */
560 	memset(&so, 0, sizeof(so));
561 
562 	s = splnet();
563 
564 	/*
565 	 * Do an if_down() to give protocols a chance to do something.
566 	 */
567 	if_down(ifp);
568 
569 #ifdef ALTQ
570 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
571 		altq_disable(&ifp->if_snd);
572 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
573 		altq_detach(&ifp->if_snd);
574 #endif
575 
576 #ifdef PFIL_HOOKS
577 	(void) pfil_head_unregister(&ifp->if_pfil);
578 #endif
579 
580 	/*
581 	 * Rip all the addresses off the interface.  This should make
582 	 * all of the routes go away.
583 	 */
584 	ifap = &TAILQ_FIRST(&ifp->if_addrlist); /* XXX abstraction violation */
585 	while ((ifa = *ifap)) {
586 		family = ifa->ifa_addr->sa_family;
587 #ifdef IFAREF_DEBUG
588 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
589 		    ifa, family, ifa->ifa_refcnt);
590 		if (last_ifa != NULL && ifa == last_ifa)
591 			panic("if_detach: loop detected");
592 		last_ifa = ifa;
593 #endif
594 		if (family == AF_LINK) {
595 			ifap = &TAILQ_NEXT(ifa, ifa_list);
596 			continue;
597 		}
598 		dp = pffinddomain(family);
599 #ifdef DIAGNOSTIC
600 		if (dp == NULL)
601 			panic("if_detach: no domain for AF %d",
602 			    family);
603 #endif
604 		purged = 0;
605 		for (pr = dp->dom_protosw;
606 		     pr < dp->dom_protoswNPROTOSW; pr++) {
607 			so.so_proto = pr;
608 			if (pr->pr_usrreq != NULL) {
609 				(void) (*pr->pr_usrreq)(&so,
610 				    PRU_PURGEIF, NULL, NULL,
611 				    (struct mbuf *) ifp, curproc);
612 				purged = 1;
613 			}
614 		}
615 		if (purged == 0) {
616 			/*
617 			 * XXX What's really the best thing to do
618 			 * XXX here?  --thorpej@NetBSD.org
619 			 */
620 			printf("if_detach: WARNING: AF %d not purged\n",
621 			    family);
622 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
623 		}
624 	}
625 
626 	if_free_sadl(ifp);
627 
628 	/* Walk the routing table looking for straglers. */
629 	for (i = 0; i <= AF_MAX; i++) {
630 		if ((rnh = rt_tables[i]) != NULL)
631 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
632 	}
633 
634 	for (dp = domains; dp; dp = dp->dom_next) {
635 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
636 			(*dp->dom_ifdetach)(ifp,
637 			    ifp->if_afdata[dp->dom_family]);
638 	}
639 
640 	/* Announce that the interface is gone. */
641 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
642 
643 	ifindex2ifnet[ifp->if_index] = NULL;
644 
645 	TAILQ_REMOVE(&ifnet, ifp, if_list);
646 
647 	/*
648 	 * remove packets came from ifp, from software interrupt queues.
649 	 * net/netisr_dispatch.h is not usable, as some of them use
650 	 * strange queue names.
651 	 */
652 #define IF_DETACH_QUEUES(x) \
653 do { \
654 	extern struct ifqueue x; \
655 	if_detach_queues(ifp, & x); \
656 } while (/*CONSTCOND*/ 0)
657 #ifdef INET
658 #if NARP > 0
659 	IF_DETACH_QUEUES(arpintrq);
660 #endif
661 	IF_DETACH_QUEUES(ipintrq);
662 #endif
663 #ifdef INET6
664 	IF_DETACH_QUEUES(ip6intrq);
665 #endif
666 #ifdef NETATALK
667 	IF_DETACH_QUEUES(atintrq1);
668 	IF_DETACH_QUEUES(atintrq2);
669 #endif
670 #ifdef NS
671 	IF_DETACH_QUEUES(nsintrq);
672 #endif
673 #ifdef ISO
674 	IF_DETACH_QUEUES(clnlintrq);
675 #endif
676 #ifdef CCITT
677 	IF_DETACH_QUEUES(llcintrq);
678 	IF_DETACH_QUEUES(hdintrq);
679 #endif
680 #ifdef NATM
681 	IF_DETACH_QUEUES(natmintrq);
682 #endif
683 #ifdef DECNET
684 	IF_DETACH_QUEUES(decnetintrq);
685 #endif
686 #undef IF_DETACH_QUEUES
687 
688 	splx(s);
689 }
690 
691 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
692     defined(ISO) || defined(CCITT) || defined(NATM) || defined(DECNET)
693 static void
694 if_detach_queues(ifp, q)
695 	struct ifnet *ifp;
696 	struct ifqueue *q;
697 {
698 	struct mbuf *m, *prev, *next;
699 
700 	prev = NULL;
701 	for (m = q->ifq_head; m; m = next) {
702 		next = m->m_nextpkt;
703 #ifdef DIAGNOSTIC
704 		if ((m->m_flags & M_PKTHDR) == 0) {
705 			prev = m;
706 			continue;
707 		}
708 #endif
709 		if (m->m_pkthdr.rcvif != ifp) {
710 			prev = m;
711 			continue;
712 		}
713 
714 		if (prev)
715 			prev->m_nextpkt = m->m_nextpkt;
716 		else
717 			q->ifq_head = m->m_nextpkt;
718 		if (q->ifq_tail == m)
719 			q->ifq_tail = prev;
720 		q->ifq_len--;
721 
722 		m->m_nextpkt = NULL;
723 		m_freem(m);
724 		IF_DROP(q);
725 	}
726 }
727 #endif /* defined(INET) || ... */
728 
729 /*
730  * Callback for a radix tree walk to delete all references to an
731  * ifnet.
732  */
733 int
734 if_rt_walktree(rn, v)
735 	struct radix_node *rn;
736 	void *v;
737 {
738 	struct ifnet *ifp = (struct ifnet *)v;
739 	struct rtentry *rt = (struct rtentry *)rn;
740 	int error;
741 
742 	if (rt->rt_ifp == ifp) {
743 		/* Delete the entry. */
744 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
745 		    rt_mask(rt), rt->rt_flags, NULL);
746 		if (error)
747 			printf("%s: warning: unable to delete rtentry @ %p, "
748 			    "error = %d\n", ifp->if_xname, rt, error);
749 	}
750 	return (0);
751 }
752 
753 /*
754  * Create a clone network interface.
755  */
756 int
757 if_clone_create(name)
758 	const char *name;
759 {
760 	struct if_clone *ifc;
761 	int unit;
762 
763 	ifc = if_clone_lookup(name, &unit);
764 	if (ifc == NULL)
765 		return (EINVAL);
766 
767 	if (ifunit(name) != NULL)
768 		return (EEXIST);
769 
770 	return ((*ifc->ifc_create)(ifc, unit));
771 }
772 
773 /*
774  * Destroy a clone network interface.
775  */
776 int
777 if_clone_destroy(name)
778 	const char *name;
779 {
780 	struct if_clone *ifc;
781 	struct ifnet *ifp;
782 
783 	ifc = if_clone_lookup(name, NULL);
784 	if (ifc == NULL)
785 		return (EINVAL);
786 
787 	ifp = ifunit(name);
788 	if (ifp == NULL)
789 		return (ENXIO);
790 
791 	if (ifc->ifc_destroy == NULL)
792 		return (EOPNOTSUPP);
793 
794 	(*ifc->ifc_destroy)(ifp);
795 	return (0);
796 }
797 
798 /*
799  * Look up a network interface cloner.
800  */
801 struct if_clone *
802 if_clone_lookup(name, unitp)
803 	const char *name;
804 	int *unitp;
805 {
806 	struct if_clone *ifc;
807 	const char *cp;
808 	int unit;
809 
810 	/* separate interface name from unit */
811 	for (cp = name;
812 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
813 	    cp++)
814 		continue;
815 
816 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
817 		return (NULL);	/* No name or unit number */
818 
819 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
820 		if (strlen(ifc->ifc_name) == cp - name &&
821 		    !strncmp(name, ifc->ifc_name, cp - name))
822 			break;
823 	}
824 
825 	if (ifc == NULL)
826 		return (NULL);
827 
828 	unit = 0;
829 	while (cp - name < IFNAMSIZ && *cp) {
830 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
831 			/* Bogus unit number. */
832 			return (NULL);
833 		}
834 		unit = (unit * 10) + (*cp++ - '0');
835 	}
836 
837 	if (unitp != NULL)
838 		*unitp = unit;
839 	return (ifc);
840 }
841 
842 /*
843  * Register a network interface cloner.
844  */
845 void
846 if_clone_attach(ifc)
847 	struct if_clone *ifc;
848 {
849 
850 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
851 	if_cloners_count++;
852 }
853 
854 /*
855  * Unregister a network interface cloner.
856  */
857 void
858 if_clone_detach(ifc)
859 	struct if_clone *ifc;
860 {
861 
862 	LIST_REMOVE(ifc, ifc_list);
863 	if_cloners_count--;
864 }
865 
866 /*
867  * Provide list of interface cloners to userspace.
868  */
869 int
870 if_clone_list(ifcr)
871 	struct if_clonereq *ifcr;
872 {
873 	char outbuf[IFNAMSIZ], *dst;
874 	struct if_clone *ifc;
875 	int count, error = 0;
876 
877 	ifcr->ifcr_total = if_cloners_count;
878 	if ((dst = ifcr->ifcr_buffer) == NULL) {
879 		/* Just asking how many there are. */
880 		return (0);
881 	}
882 
883 	if (ifcr->ifcr_count < 0)
884 		return (EINVAL);
885 
886 	count = (if_cloners_count < ifcr->ifcr_count) ?
887 	    if_cloners_count : ifcr->ifcr_count;
888 
889 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
890 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
891 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
892 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
893 		error = copyout(outbuf, dst, IFNAMSIZ);
894 		if (error)
895 			break;
896 	}
897 
898 	return (error);
899 }
900 
901 /*
902  * Locate an interface based on a complete address.
903  */
904 /*ARGSUSED*/
905 struct ifaddr *
906 ifa_ifwithaddr(addr)
907 	const struct sockaddr *addr;
908 {
909 	struct ifnet *ifp;
910 	struct ifaddr *ifa;
911 
912 #define	equal(a1, a2) \
913   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
914 
915 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
916 	     ifp = TAILQ_NEXT(ifp, if_list)) {
917 		if (ifp->if_output == if_nulloutput)
918 			continue;
919 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
920 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
921 			if (ifa->ifa_addr->sa_family != addr->sa_family)
922 				continue;
923 			if (equal(addr, ifa->ifa_addr))
924 				return (ifa);
925 			if ((ifp->if_flags & IFF_BROADCAST) &&
926 			    ifa->ifa_broadaddr &&
927 			    /* IP6 doesn't have broadcast */
928 			    ifa->ifa_broadaddr->sa_len != 0 &&
929 			    equal(ifa->ifa_broadaddr, addr))
930 				return (ifa);
931 		}
932 	}
933 	return (NULL);
934 }
935 
936 /*
937  * Locate the point to point interface with a given destination address.
938  */
939 /*ARGSUSED*/
940 struct ifaddr *
941 ifa_ifwithdstaddr(addr)
942 	const struct sockaddr *addr;
943 {
944 	struct ifnet *ifp;
945 	struct ifaddr *ifa;
946 
947 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
948 	     ifp = TAILQ_NEXT(ifp, if_list)) {
949 		if (ifp->if_output == if_nulloutput)
950 			continue;
951 		if (ifp->if_flags & IFF_POINTOPOINT) {
952 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
953 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
954 				if (ifa->ifa_addr->sa_family !=
955 				      addr->sa_family ||
956 				    ifa->ifa_dstaddr == NULL)
957 					continue;
958 				if (equal(addr, ifa->ifa_dstaddr))
959 					return (ifa);
960 			}
961 		}
962 	}
963 	return (NULL);
964 }
965 
966 /*
967  * Find an interface on a specific network.  If many, choice
968  * is most specific found.
969  */
970 struct ifaddr *
971 ifa_ifwithnet(addr)
972 	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 	char *addr_data = addr->sa_data, *cplim;
980 
981 	if (af == AF_LINK) {
982 		sdl = (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 = (struct sockaddr_at *)addr;
992 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
993 		     ifp = TAILQ_NEXT(ifp, if_list)) {
994 			if (ifp->if_output == if_nulloutput)
995 				continue;
996 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
997 			if (ifa == NULL)
998 				continue;
999 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1000 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1001 				return (ifa); /* exact match */
1002 			if (ifa_maybe == NULL) {
1003 				/* else keep the if with the right range */
1004 				ifa_maybe = ifa;
1005 			}
1006 		}
1007 		return (ifa_maybe);
1008 	}
1009 #endif
1010 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1011 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1012 		if (ifp->if_output == if_nulloutput)
1013 			continue;
1014 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1015 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1016 			char *cp, *cp2, *cp3;
1017 
1018 			if (ifa->ifa_addr->sa_family != af ||
1019 			    ifa->ifa_netmask == 0)
1020  next:				continue;
1021 			cp = addr_data;
1022 			cp2 = ifa->ifa_addr->sa_data;
1023 			cp3 = ifa->ifa_netmask->sa_data;
1024 			cplim = (char *)ifa->ifa_netmask +
1025 			    ifa->ifa_netmask->sa_len;
1026 			while (cp3 < cplim) {
1027 				if ((*cp++ ^ *cp2++) & *cp3++) {
1028 					/* want to continue for() loop */
1029 					goto next;
1030 				}
1031 			}
1032 			if (ifa_maybe == 0 ||
1033 			    rn_refines((caddr_t)ifa->ifa_netmask,
1034 			    (caddr_t)ifa_maybe->ifa_netmask))
1035 				ifa_maybe = ifa;
1036 		}
1037 	}
1038 	return (ifa_maybe);
1039 }
1040 
1041 /*
1042  * Find the interface of the addresss.
1043  */
1044 struct ifaddr *
1045 ifa_ifwithladdr(addr)
1046 	const struct sockaddr *addr;
1047 {
1048 	struct ifaddr *ia;
1049 
1050 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1051 	    (ia = ifa_ifwithnet(addr)))
1052 		return (ia);
1053 	return (NULL);
1054 }
1055 
1056 /*
1057  * Find an interface using a specific address family
1058  */
1059 struct ifaddr *
1060 ifa_ifwithaf(af)
1061 	int af;
1062 {
1063 	struct ifnet *ifp;
1064 	struct ifaddr *ifa;
1065 
1066 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1067 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1068 		if (ifp->if_output == if_nulloutput)
1069 			continue;
1070 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1071 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1072 			if (ifa->ifa_addr->sa_family == af)
1073 				return (ifa);
1074 		}
1075 	}
1076 	return (NULL);
1077 }
1078 
1079 /*
1080  * Find an interface address specific to an interface best matching
1081  * a given address.
1082  */
1083 struct ifaddr *
1084 ifaof_ifpforaddr(addr, ifp)
1085 	const struct sockaddr *addr;
1086 	struct ifnet *ifp;
1087 {
1088 	struct ifaddr *ifa;
1089 	const char *cp, *cp2, *cp3;
1090 	const char *cplim;
1091 	struct ifaddr *ifa_maybe = 0;
1092 	u_int af = addr->sa_family;
1093 
1094 	if (ifp->if_output == if_nulloutput)
1095 		return (NULL);
1096 
1097 	if (af >= AF_MAX)
1098 		return (NULL);
1099 
1100 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1101 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
1102 		if (ifa->ifa_addr->sa_family != af)
1103 			continue;
1104 		ifa_maybe = ifa;
1105 		if (ifa->ifa_netmask == 0) {
1106 			if (equal(addr, ifa->ifa_addr) ||
1107 			    (ifa->ifa_dstaddr &&
1108 			     equal(addr, ifa->ifa_dstaddr)))
1109 				return (ifa);
1110 			continue;
1111 		}
1112 		cp = addr->sa_data;
1113 		cp2 = ifa->ifa_addr->sa_data;
1114 		cp3 = ifa->ifa_netmask->sa_data;
1115 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1116 		for (; cp3 < cplim; cp3++) {
1117 			if ((*cp++ ^ *cp2++) & *cp3)
1118 				break;
1119 		}
1120 		if (cp3 == cplim)
1121 			return (ifa);
1122 	}
1123 	return (ifa_maybe);
1124 }
1125 
1126 /*
1127  * Default action when installing a route with a Link Level gateway.
1128  * Lookup an appropriate real ifa to point to.
1129  * This should be moved to /sys/net/link.c eventually.
1130  */
1131 void
1132 link_rtrequest(cmd, rt, info)
1133 	int cmd;
1134 	struct rtentry *rt;
1135 	struct rt_addrinfo *info;
1136 {
1137 	struct ifaddr *ifa;
1138 	struct sockaddr *dst;
1139 	struct ifnet *ifp;
1140 
1141 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1142 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1143 		return;
1144 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1145 		IFAFREE(rt->rt_ifa);
1146 		rt->rt_ifa = ifa;
1147 		IFAREF(ifa);
1148 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1149 			ifa->ifa_rtrequest(cmd, rt, info);
1150 	}
1151 }
1152 
1153 /*
1154  * Mark an interface down and notify protocols of
1155  * the transition.
1156  * NOTE: must be called at splsoftnet or equivalent.
1157  */
1158 void
1159 if_down(ifp)
1160 	struct ifnet *ifp;
1161 {
1162 	struct ifaddr *ifa;
1163 
1164 	ifp->if_flags &= ~IFF_UP;
1165 	microtime(&ifp->if_lastchange);
1166 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1167 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1168 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1169 	IFQ_PURGE(&ifp->if_snd);
1170 	rt_ifmsg(ifp);
1171 }
1172 
1173 /*
1174  * Mark an interface up and notify protocols of
1175  * the transition.
1176  * NOTE: must be called at splsoftnet or equivalent.
1177  */
1178 void
1179 if_up(ifp)
1180 	struct ifnet *ifp;
1181 {
1182 #ifdef notyet
1183 	struct ifaddr *ifa;
1184 #endif
1185 
1186 	ifp->if_flags |= IFF_UP;
1187 	microtime(&ifp->if_lastchange);
1188 #ifdef notyet
1189 	/* this has no effect on IP, and will kill all ISO connections XXX */
1190 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
1191 	     ifa = TAILQ_NEXT(ifa, ifa_list))
1192 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1193 #endif
1194 	rt_ifmsg(ifp);
1195 #ifdef INET6
1196 	in6_if_up(ifp);
1197 #endif
1198 }
1199 
1200 /*
1201  * Handle interface watchdog timer routines.  Called
1202  * from softclock, we decrement timers (if set) and
1203  * call the appropriate interface routine on expiration.
1204  */
1205 void
1206 if_slowtimo(arg)
1207 	void *arg;
1208 {
1209 	struct ifnet *ifp;
1210 	int s = splnet();
1211 
1212 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1213 	     ifp = TAILQ_NEXT(ifp, if_list)) {
1214 		if (ifp->if_timer == 0 || --ifp->if_timer)
1215 			continue;
1216 		if (ifp->if_watchdog)
1217 			(*ifp->if_watchdog)(ifp);
1218 	}
1219 	splx(s);
1220 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
1221 	    if_slowtimo, NULL);
1222 }
1223 
1224 /*
1225  * Set/clear promiscuous mode on interface ifp based on the truth value
1226  * of pswitch.  The calls are reference counted so that only the first
1227  * "on" request actually has an effect, as does the final "off" request.
1228  * Results are undefined if the "off" and "on" requests are not matched.
1229  */
1230 int
1231 ifpromisc(ifp, pswitch)
1232 	struct ifnet *ifp;
1233 	int pswitch;
1234 {
1235 	int pcount, ret;
1236 	short flags;
1237 	struct ifreq ifr;
1238 
1239 	pcount = ifp->if_pcount;
1240 	flags = ifp->if_flags;
1241 	if (pswitch) {
1242 		/*
1243 		 * Allow the device to be "placed" into promiscuous
1244 		 * mode even if it is not configured up.  It will
1245 		 * consult IFF_PROMISC when it is is brought up.
1246 		 */
1247 		if (ifp->if_pcount++ != 0)
1248 			return (0);
1249 		ifp->if_flags |= IFF_PROMISC;
1250 		if ((ifp->if_flags & IFF_UP) == 0)
1251 			return (0);
1252 	} else {
1253 		if (--ifp->if_pcount > 0)
1254 			return (0);
1255 		ifp->if_flags &= ~IFF_PROMISC;
1256 		/*
1257 		 * If the device is not configured up, we should not need to
1258 		 * turn off promiscuous mode (device should have turned it
1259 		 * off when interface went down; and will look at IFF_PROMISC
1260 		 * again next time interface comes up).
1261 		 */
1262 		if ((ifp->if_flags & IFF_UP) == 0)
1263 			return (0);
1264 	}
1265 	memset(&ifr, 0, sizeof(ifr));
1266 	ifr.ifr_flags = ifp->if_flags;
1267 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
1268 	/* Restore interface state if not successful. */
1269 	if (ret != 0) {
1270 		ifp->if_pcount = pcount;
1271 		ifp->if_flags = flags;
1272 	}
1273 	return (ret);
1274 }
1275 
1276 /*
1277  * Map interface name to
1278  * interface structure pointer.
1279  */
1280 struct ifnet *
1281 ifunit(name)
1282 	const char *name;
1283 {
1284 	struct ifnet *ifp;
1285 	const char *cp = name;
1286 	u_int unit = 0;
1287 	u_int i;
1288 
1289 	/*
1290 	 * If the entire name is a number, treat it as an ifindex.
1291 	 */
1292 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1293 		unit = unit * 10 + (*cp - '0');
1294 	}
1295 
1296 	/*
1297 	 * If the number took all of the name, then it's a valid ifindex.
1298 	 */
1299 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1300 		if (unit >= if_indexlim)
1301 			return (NULL);
1302 		ifp = ifindex2ifnet[unit];
1303 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1304 			return (NULL);
1305 		return (ifp);
1306 	}
1307 
1308 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
1309 	     ifp = TAILQ_NEXT(ifp, 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(so, cmd, data, p)
1323 	struct socket *so;
1324 	u_long cmd;
1325 	caddr_t data;
1326 	struct proc *p;
1327 {
1328 	struct ifnet *ifp;
1329 	struct ifreq *ifr;
1330 	struct ifcapreq *ifcr;
1331 	struct ifdatareq *ifdr;
1332 	int s, error = 0;
1333 	short oif_flags;
1334 	int prived_error;
1335 
1336 	if (p)
1337 		prived_error = suser(p->p_ucred, &p->p_acflag);
1338 	else
1339 		prived_error = 0;
1340 
1341 	switch (cmd) {
1342 
1343 	case SIOCGIFCONF:
1344 	case OSIOCGIFCONF:
1345 		return (ifconf(cmd, data));
1346 	}
1347 	ifr = (struct ifreq *)data;
1348 	ifcr = (struct ifcapreq *)data;
1349 	ifdr = (struct ifdatareq *)data;
1350 
1351 	switch (cmd) {
1352 	case SIOCIFCREATE:
1353 	case SIOCIFDESTROY:
1354 		if (prived_error)
1355 			return (prived_error);
1356 		return ((cmd == SIOCIFCREATE) ?
1357 			if_clone_create(ifr->ifr_name) :
1358 			if_clone_destroy(ifr->ifr_name));
1359 
1360 	case SIOCIFGCLONERS:
1361 		return (if_clone_list((struct if_clonereq *)data));
1362 	}
1363 
1364 	ifp = ifunit(ifr->ifr_name);
1365 	if (ifp == 0)
1366 		return (ENXIO);
1367 	oif_flags = ifp->if_flags;
1368 	switch (cmd) {
1369 
1370 	case SIOCGIFFLAGS:
1371 		ifr->ifr_flags = ifp->if_flags;
1372 		break;
1373 
1374 	case SIOCGIFMETRIC:
1375 		ifr->ifr_metric = ifp->if_metric;
1376 		break;
1377 
1378 	case SIOCGIFMTU:
1379 		ifr->ifr_mtu = ifp->if_mtu;
1380 		break;
1381 
1382 	case SIOCGIFDLT:
1383 		ifr->ifr_dlt = ifp->if_dlt;
1384 		break;
1385 
1386 	case SIOCSIFFLAGS:
1387 		if (prived_error != 0)
1388 			return (prived_error);
1389 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1390 			s = splnet();
1391 			if_down(ifp);
1392 			splx(s);
1393 		}
1394 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1395 			s = splnet();
1396 			if_up(ifp);
1397 			splx(s);
1398 		}
1399 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1400 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1401 		if (ifp->if_ioctl)
1402 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1403 		break;
1404 
1405 	case SIOCGIFCAP:
1406 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1407 		ifcr->ifcr_capenable = ifp->if_capenable;
1408 		break;
1409 
1410 	case SIOCSIFCAP:
1411 		if (prived_error != 0)
1412 			return (prived_error);
1413 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1414 			return (EINVAL);
1415 		if (ifp->if_ioctl == NULL)
1416 			return (EOPNOTSUPP);
1417 
1418 		/* Must prevent race with packet reception here. */
1419 		s = splnet();
1420 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
1421 			struct ifreq ifrq;
1422 
1423 			ifrq.ifr_flags = ifp->if_flags;
1424 			ifp->if_capenable = ifcr->ifcr_capenable;
1425 
1426 			/* Pre-compute the checksum flags mask. */
1427 			ifp->if_csum_flags_tx = 0;
1428 			ifp->if_csum_flags_rx = 0;
1429 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
1430 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1431 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1432 			}
1433 
1434 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
1435 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1436 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1437 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
1438 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1439 
1440 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
1441 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1442 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1443 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
1444 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1445 
1446 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
1447 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1448 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1449 			}
1450 
1451 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
1452 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1453 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1454 			}
1455 
1456 			/*
1457 			 * Only kick the interface if it's up.  If it's
1458 			 * not up now, it will notice the cap enables
1459 			 * when it is brought up later.
1460 			 */
1461 			if (ifp->if_flags & IFF_UP)
1462 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
1463 				    (caddr_t) &ifrq);
1464 		}
1465 		splx(s);
1466 		break;
1467 
1468 	case SIOCSIFMETRIC:
1469 		if (prived_error != 0)
1470 			return (prived_error);
1471 		ifp->if_metric = ifr->ifr_metric;
1472 		break;
1473 
1474 	case SIOCGIFDATA:
1475 		ifdr->ifdr_data = ifp->if_data;
1476 		break;
1477 
1478 	case SIOCZIFDATA:
1479 		if (prived_error != 0)
1480 			return (prived_error);
1481 		ifdr->ifdr_data = ifp->if_data;
1482 		/*
1483 		 * Assumes that the volatile counters that can be
1484 		 * zero'ed are at the end of if_data.
1485 		 */
1486 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1487 		    offsetof(struct if_data, ifi_ipackets));
1488 		break;
1489 
1490 	case SIOCSIFMTU:
1491 	{
1492 		u_long oldmtu = ifp->if_mtu;
1493 
1494 		if (prived_error)
1495 			return (prived_error);
1496 		if (ifp->if_ioctl == NULL)
1497 			return (EOPNOTSUPP);
1498 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1499 
1500 		/*
1501 		 * If the link MTU changed, do network layer specific procedure.
1502 		 */
1503 		if (ifp->if_mtu != oldmtu) {
1504 #ifdef INET6
1505 			nd6_setmtu(ifp);
1506 #endif
1507 		}
1508 		break;
1509 	}
1510 	case SIOCSIFPHYADDR:
1511 	case SIOCDIFPHYADDR:
1512 #ifdef INET6
1513 	case SIOCSIFPHYADDR_IN6:
1514 #endif
1515 	case SIOCSLIFPHYADDR:
1516 	case SIOCADDMULTI:
1517 	case SIOCDELMULTI:
1518 	case SIOCSIFMEDIA:
1519 		if (prived_error != 0)
1520 			return (prived_error);
1521 		/* FALLTHROUGH */
1522 	case SIOCGIFPSRCADDR:
1523 	case SIOCGIFPDSTADDR:
1524 	case SIOCGLIFPHYADDR:
1525 	case SIOCGIFMEDIA:
1526 		if (ifp->if_ioctl == 0)
1527 			return (EOPNOTSUPP);
1528 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1529 		break;
1530 
1531 	case SIOCSDRVSPEC:
1532 	case SIOCS80211NWID:
1533 	case SIOCS80211NWKEY:
1534 	case SIOCS80211POWER:
1535 	case SIOCS80211BSSID:
1536 	case SIOCS80211CHANNEL:
1537 		/* XXX:  need to pass proc pointer through to driver... */
1538 		if (prived_error != 0)
1539 			return (prived_error);
1540 	/* FALLTHROUGH */
1541 	default:
1542 		if (so->so_proto == 0)
1543 			return (EOPNOTSUPP);
1544 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(COMPAT_ULTRIX) && !defined(LKM)
1545 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1546 		    (struct mbuf *)cmd, (struct mbuf *)data,
1547 		    (struct mbuf *)ifp, p));
1548 #else
1549 	    {
1550 		int ocmd = cmd;
1551 
1552 		switch (cmd) {
1553 
1554 		case SIOCSIFADDR:
1555 		case SIOCSIFDSTADDR:
1556 		case SIOCSIFBRDADDR:
1557 		case SIOCSIFNETMASK:
1558 #if BYTE_ORDER != BIG_ENDIAN
1559 			if (ifr->ifr_addr.sa_family == 0 &&
1560 			    ifr->ifr_addr.sa_len < 16) {
1561 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1562 				ifr->ifr_addr.sa_len = 16;
1563 			}
1564 #else
1565 			if (ifr->ifr_addr.sa_len == 0)
1566 				ifr->ifr_addr.sa_len = 16;
1567 #endif
1568 			break;
1569 
1570 		case OSIOCGIFADDR:
1571 			cmd = SIOCGIFADDR;
1572 			break;
1573 
1574 		case OSIOCGIFDSTADDR:
1575 			cmd = SIOCGIFDSTADDR;
1576 			break;
1577 
1578 		case OSIOCGIFBRDADDR:
1579 			cmd = SIOCGIFBRDADDR;
1580 			break;
1581 
1582 		case OSIOCGIFNETMASK:
1583 			cmd = SIOCGIFNETMASK;
1584 		}
1585 
1586 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1587 		    (struct mbuf *)cmd, (struct mbuf *)data,
1588 		    (struct mbuf *)ifp, p));
1589 
1590 		switch (ocmd) {
1591 		case OSIOCGIFADDR:
1592 		case OSIOCGIFDSTADDR:
1593 		case OSIOCGIFBRDADDR:
1594 		case OSIOCGIFNETMASK:
1595 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1596 		}
1597 	    }
1598 #endif /* COMPAT_43 */
1599 		break;
1600 	}
1601 
1602 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1603 #ifdef INET6
1604 		if ((ifp->if_flags & IFF_UP) != 0) {
1605 			s = splnet();
1606 			in6_if_up(ifp);
1607 			splx(s);
1608 		}
1609 #endif
1610 	}
1611 
1612 	return (error);
1613 }
1614 
1615 /*
1616  * Return interface configuration
1617  * of system.  List may be used
1618  * in later ioctl's (above) to get
1619  * other information.
1620  */
1621 /*ARGSUSED*/
1622 int
1623 ifconf(cmd, data)
1624 	u_long cmd;
1625 	caddr_t data;
1626 {
1627 	struct ifconf *ifc = (struct ifconf *)data;
1628 	struct ifnet *ifp;
1629 	struct ifaddr *ifa;
1630 	struct ifreq ifr, *ifrp;
1631 	int space = ifc->ifc_len, error = 0;
1632 	const int sz = (int)sizeof(ifr);
1633 	int sign;
1634 
1635 	if ((ifrp = ifc->ifc_req) == NULL) {
1636 		space = 0;
1637 		sign = -1;
1638 	} else {
1639 		sign = 1;
1640 	}
1641 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1642 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
1643 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
1644 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1645 			if (ifrp != NULL && space >= sz) {
1646 				error = copyout(&ifr, ifrp, sz);
1647 				if (error)
1648 					break;
1649 				ifrp++;
1650 			}
1651 			space -= sizeof(ifr) * sign;
1652 			continue;
1653 		}
1654 
1655 		for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
1656 			struct sockaddr *sa = ifa->ifa_addr;
1657 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4) || defined(COMPAT_ULTRIX)
1658 			if (cmd == OSIOCGIFCONF) {
1659 				struct osockaddr *osa =
1660 					 (struct osockaddr *)&ifr.ifr_addr;
1661 				/*
1662 				 * If it does not fit, we don't bother with it
1663 				 */
1664 				if (sa->sa_len > sizeof(*osa))
1665 					continue;
1666 				ifr.ifr_addr = *sa;
1667 				osa->sa_family = sa->sa_family;
1668 				if (ifrp != NULL && space >= sz) {
1669 					error = copyout(&ifr, ifrp, sz);
1670 					ifrp++;
1671 				}
1672 			} else
1673 #endif
1674 			if (sa->sa_len <= sizeof(*sa)) {
1675 				ifr.ifr_addr = *sa;
1676 				if (ifrp != NULL && space >= sz) {
1677 					error = copyout(&ifr, ifrp, sz);
1678 					ifrp++;
1679 				}
1680 			} else {
1681 				space -= (sa->sa_len - sizeof(*sa)) * sign;
1682 				if (ifrp != NULL && space >= sz) {
1683 					error = copyout(&ifr, ifrp,
1684 					    sizeof(ifr.ifr_name));
1685 					if (error == 0) {
1686 						error = copyout(sa,
1687 						    &ifrp->ifr_addr,
1688 						    sa->sa_len);
1689 					}
1690 					ifrp = (struct ifreq *)
1691 						(sa->sa_len +
1692 						 (caddr_t)&ifrp->ifr_addr);
1693 				}
1694 			}
1695 			if (error)
1696 				break;
1697 			space -= sz * sign;
1698 		}
1699 	}
1700 	if (ifrp != NULL)
1701 		ifc->ifc_len -= space;
1702 	else
1703 		ifc->ifc_len = space;
1704 	return (error);
1705 }
1706 
1707 #if defined(INET) || defined(INET6)
1708 static void
1709 sysctl_net_ifq_setup(struct sysctllog **clog,
1710 		     int pf, const char *pfname,
1711 		     int ipn, const char *ipname,
1712 		     int qid, struct ifqueue *ifq)
1713 {
1714 
1715 	sysctl_createv(clog, 0, NULL, NULL,
1716 		       CTLFLAG_PERMANENT,
1717 		       CTLTYPE_NODE, "net", NULL,
1718 		       NULL, 0, NULL, 0,
1719 		       CTL_NET, CTL_EOL);
1720 	sysctl_createv(clog, 0, NULL, NULL,
1721 		       CTLFLAG_PERMANENT,
1722 		       CTLTYPE_NODE, pfname, NULL,
1723 		       NULL, 0, NULL, 0,
1724 		       CTL_NET, pf, CTL_EOL);
1725 	sysctl_createv(clog, 0, NULL, NULL,
1726 		       CTLFLAG_PERMANENT,
1727 		       CTLTYPE_NODE, ipname, NULL,
1728 		       NULL, 0, NULL, 0,
1729 		       CTL_NET, pf, ipn, CTL_EOL);
1730 	sysctl_createv(clog, 0, NULL, NULL,
1731 		       CTLFLAG_PERMANENT,
1732 		       CTLTYPE_NODE, "ifq",
1733 		       SYSCTL_DESCR("Protocol input queue controls"),
1734 		       NULL, 0, NULL, 0,
1735 		       CTL_NET, pf, ipn, qid, CTL_EOL);
1736 
1737 	sysctl_createv(clog, 0, NULL, NULL,
1738 		       CTLFLAG_PERMANENT,
1739 		       CTLTYPE_INT, "len",
1740 		       SYSCTL_DESCR("Current input queue length"),
1741 		       NULL, 0, &ifq->ifq_len, 0,
1742 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1743 	sysctl_createv(clog, 0, NULL, NULL,
1744 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1745 		       CTLTYPE_INT, "maxlen",
1746 		       SYSCTL_DESCR("Maximum allowed input queue length"),
1747 		       NULL, 0, &ifq->ifq_maxlen, 0,
1748 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1749 #ifdef notyet
1750 	sysctl_createv(clog, 0, NULL, NULL,
1751 		       CTLFLAG_PERMANENT,
1752 		       CTLTYPE_INT, "peak",
1753 		       SYSCTL_DESCR("Highest input queue length"),
1754 		       NULL, 0, &ifq->ifq_peak, 0,
1755 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
1756 #endif
1757 	sysctl_createv(clog, 0, NULL, NULL,
1758 		       CTLFLAG_PERMANENT,
1759 		       CTLTYPE_INT, "drops",
1760 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
1761 		       NULL, 0, &ifq->ifq_drops, 0,
1762 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
1763 }
1764 
1765 #ifdef INET
1766 SYSCTL_SETUP(sysctl_net_inet_ip_ifq_setup,
1767 	     "sysctl net.inet.ip.ifq subtree setup")
1768 {
1769 	extern struct ifqueue ipintrq;
1770 
1771 	sysctl_net_ifq_setup(clog, PF_INET, "inet", IPPROTO_IP, "ip",
1772 			     IPCTL_IFQ, &ipintrq);
1773 }
1774 #endif /* INET */
1775 
1776 #ifdef INET6
1777 SYSCTL_SETUP(sysctl_net_inet6_ip6_ifq_setup,
1778 	     "sysctl net.inet6.ip6.ifq subtree setup")
1779 {
1780 	extern struct ifqueue ip6intrq;
1781 
1782 	sysctl_net_ifq_setup(clog, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
1783 			     IPV6CTL_IFQ, &ip6intrq);
1784 }
1785 #endif /* INET6 */
1786 #endif /* INET || INET6 */
1787