xref: /netbsd-src/sys/net/if.c (revision f7673b99b4d7c781e1b0efb76be193aa60d399d4)
1 /*	$NetBSD: if.c,v 1.292 2014/11/07 12:57:42 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by William Studenmund and Jason R. Thorpe.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the project nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  */
60 
61 /*
62  * Copyright (c) 1980, 1986, 1993
63  *	The Regents of the University of California.  All rights reserved.
64  *
65  * Redistribution and use in source and binary forms, with or without
66  * modification, are permitted provided that the following conditions
67  * are met:
68  * 1. Redistributions of source code must retain the above copyright
69  *    notice, this list of conditions and the following disclaimer.
70  * 2. Redistributions in binary form must reproduce the above copyright
71  *    notice, this list of conditions and the following disclaimer in the
72  *    documentation and/or other materials provided with the distribution.
73  * 3. Neither the name of the University nor the names of its contributors
74  *    may be used to endorse or promote products derived from this software
75  *    without specific prior written permission.
76  *
77  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
78  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
79  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
80  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
81  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
82  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
83  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
84  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
85  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
86  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
87  * SUCH DAMAGE.
88  *
89  *	@(#)if.c	8.5 (Berkeley) 1/9/95
90  */
91 
92 #include <sys/cdefs.h>
93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.292 2014/11/07 12:57:42 christos Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 #include "opt_wlan.h"
100 
101 #include <sys/param.h>
102 #include <sys/mbuf.h>
103 #include <sys/systm.h>
104 #include <sys/callout.h>
105 #include <sys/proc.h>
106 #include <sys/socket.h>
107 #include <sys/socketvar.h>
108 #include <sys/domain.h>
109 #include <sys/protosw.h>
110 #include <sys/kernel.h>
111 #include <sys/ioctl.h>
112 #include <sys/sysctl.h>
113 #include <sys/syslog.h>
114 #include <sys/kauth.h>
115 #include <sys/kmem.h>
116 #include <sys/xcall.h>
117 
118 #include <net/if.h>
119 #include <net/if_dl.h>
120 #include <net/if_ether.h>
121 #include <net/if_media.h>
122 #include <net80211/ieee80211.h>
123 #include <net80211/ieee80211_ioctl.h>
124 #include <net/if_types.h>
125 #include <net/radix.h>
126 #include <net/route.h>
127 #include <net/netisr.h>
128 #include <sys/module.h>
129 #ifdef NETATALK
130 #include <netatalk/at_extern.h>
131 #include <netatalk/at.h>
132 #endif
133 #include <net/pfil.h>
134 #include <netinet/in.h>
135 #include <netinet/in_var.h>
136 
137 #ifdef INET6
138 #include <netinet6/in6_var.h>
139 #include <netinet6/nd6.h>
140 #endif
141 
142 #include "ether.h"
143 #include "fddi.h"
144 #include "token.h"
145 
146 #include "carp.h"
147 #if NCARP > 0
148 #include <netinet/ip_carp.h>
149 #endif
150 
151 #include <compat/sys/sockio.h>
152 #include <compat/sys/socket.h>
153 
154 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
155 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
156 
157 /*
158  * Global list of interfaces.
159  */
160 struct ifnet_head		ifnet_list;
161 static ifnet_t **		ifindex2ifnet = NULL;
162 
163 static u_int			if_index = 1;
164 static size_t			if_indexlim = 0;
165 static uint64_t			index_gen;
166 static kmutex_t			index_gen_mtx;
167 static kmutex_t			if_clone_mtx;
168 
169 static struct ifaddr **		ifnet_addrs = NULL;
170 
171 static callout_t		if_slowtimo_ch;
172 
173 struct ifnet *lo0ifp;
174 int	ifqmaxlen = IFQ_MAXLEN;
175 
176 static int	if_rt_walktree(struct rtentry *, void *);
177 
178 static struct if_clone *if_clone_lookup(const char *, int *);
179 static int	if_clone_list(struct if_clonereq *);
180 
181 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
182 static int if_cloners_count;
183 
184 /* Packet filtering hook for interfaces. */
185 pfil_head_t *	if_pfil;
186 
187 static kauth_listener_t if_listener;
188 
189 static int doifioctl(struct socket *, u_long, void *, struct lwp *);
190 static int ifioctl_attach(struct ifnet *);
191 static void ifioctl_detach(struct ifnet *);
192 static void ifnet_lock_enter(struct ifnet_lock *);
193 static void ifnet_lock_exit(struct ifnet_lock *);
194 static void if_detach_queues(struct ifnet *, struct ifqueue *);
195 static void sysctl_sndq_setup(struct sysctllog **, const char *,
196     struct ifaltq *);
197 
198 #if defined(INET) || defined(INET6)
199 static void sysctl_net_pktq_setup(struct sysctllog **, int);
200 #endif
201 
202 static int
203 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
204     void *arg0, void *arg1, void *arg2, void *arg3)
205 {
206 	int result;
207 	enum kauth_network_req req;
208 
209 	result = KAUTH_RESULT_DEFER;
210 	req = (enum kauth_network_req)arg1;
211 
212 	if (action != KAUTH_NETWORK_INTERFACE)
213 		return result;
214 
215 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
216 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
217 		result = KAUTH_RESULT_ALLOW;
218 
219 	return result;
220 }
221 
222 /*
223  * Network interface utility routines.
224  *
225  * Routines with ifa_ifwith* names take sockaddr *'s as
226  * parameters.
227  */
228 void
229 ifinit(void)
230 {
231 #if defined(INET)
232 	sysctl_net_pktq_setup(NULL, PF_INET);
233 #endif
234 #ifdef INET6
235 	sysctl_net_pktq_setup(NULL, PF_INET6);
236 #endif
237 
238 	callout_init(&if_slowtimo_ch, 0);
239 	if_slowtimo(NULL);
240 
241 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
242 	    if_listener_cb, NULL);
243 
244 	/* interfaces are available, inform socket code */
245 	ifioctl = doifioctl;
246 }
247 
248 /*
249  * XXX Initialization before configure().
250  * XXX hack to get pfil_add_hook working in autoconf.
251  */
252 void
253 ifinit1(void)
254 {
255 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
256 	mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE);
257 	TAILQ_INIT(&ifnet_list);
258 	if_indexlim = 8;
259 
260 	if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
261 	KASSERT(if_pfil != NULL);
262 
263 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN)
264 	etherinit();
265 #endif
266 }
267 
268 ifnet_t *
269 if_alloc(u_char type)
270 {
271 	return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP);
272 }
273 
274 void
275 if_free(ifnet_t *ifp)
276 {
277 	kmem_free(ifp, sizeof(ifnet_t));
278 }
279 
280 void
281 if_initname(struct ifnet *ifp, const char *name, int unit)
282 {
283 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
284 	    "%s%d", name, unit);
285 }
286 
287 /*
288  * Null routines used while an interface is going away.  These routines
289  * just return an error.
290  */
291 
292 int
293 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
294     const struct sockaddr *so, struct rtentry *rt)
295 {
296 
297 	return ENXIO;
298 }
299 
300 void
301 if_nullinput(struct ifnet *ifp, struct mbuf *m)
302 {
303 
304 	/* Nothing. */
305 }
306 
307 void
308 if_nullstart(struct ifnet *ifp)
309 {
310 
311 	/* Nothing. */
312 }
313 
314 int
315 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
316 {
317 
318 	/* Wake ifioctl_detach(), who may wait for all threads to
319 	 * quit the critical section.
320 	 */
321 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
322 	return ENXIO;
323 }
324 
325 int
326 if_nullinit(struct ifnet *ifp)
327 {
328 
329 	return ENXIO;
330 }
331 
332 void
333 if_nullstop(struct ifnet *ifp, int disable)
334 {
335 
336 	/* Nothing. */
337 }
338 
339 void
340 if_nullwatchdog(struct ifnet *ifp)
341 {
342 
343 	/* Nothing. */
344 }
345 
346 void
347 if_nulldrain(struct ifnet *ifp)
348 {
349 
350 	/* Nothing. */
351 }
352 
353 void
354 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
355 {
356 	struct ifaddr *ifa;
357 	struct sockaddr_dl *sdl;
358 
359 	ifp->if_addrlen = addrlen;
360 	if_alloc_sadl(ifp);
361 	ifa = ifp->if_dl;
362 	sdl = satosdl(ifa->ifa_addr);
363 
364 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
365 	if (factory) {
366 		ifp->if_hwdl = ifp->if_dl;
367 		ifaref(ifp->if_hwdl);
368 	}
369 	/* TBD routing socket */
370 }
371 
372 struct ifaddr *
373 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
374 {
375 	unsigned socksize, ifasize;
376 	int addrlen, namelen;
377 	struct sockaddr_dl *mask, *sdl;
378 	struct ifaddr *ifa;
379 
380 	namelen = strlen(ifp->if_xname);
381 	addrlen = ifp->if_addrlen;
382 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
383 	ifasize = sizeof(*ifa) + 2 * socksize;
384 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
385 
386 	sdl = (struct sockaddr_dl *)(ifa + 1);
387 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
388 
389 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
390 	    ifp->if_xname, namelen, NULL, addrlen);
391 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
392 	memset(&mask->sdl_data[0], 0xff, namelen);
393 	ifa->ifa_rtrequest = link_rtrequest;
394 	ifa->ifa_addr = (struct sockaddr *)sdl;
395 	ifa->ifa_netmask = (struct sockaddr *)mask;
396 
397 	*sdlp = sdl;
398 
399 	return ifa;
400 }
401 
402 static void
403 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
404 {
405 	const struct sockaddr_dl *sdl;
406 	ifnet_addrs[ifp->if_index] = ifa;
407 	ifaref(ifa);
408 	ifp->if_dl = ifa;
409 	ifaref(ifa);
410 	sdl = satosdl(ifa->ifa_addr);
411 	ifp->if_sadl = sdl;
412 }
413 
414 /*
415  * Allocate the link level name for the specified interface.  This
416  * is an attachment helper.  It must be called after ifp->if_addrlen
417  * is initialized, which may not be the case when if_attach() is
418  * called.
419  */
420 void
421 if_alloc_sadl(struct ifnet *ifp)
422 {
423 	struct ifaddr *ifa;
424 	const struct sockaddr_dl *sdl;
425 
426 	/*
427 	 * If the interface already has a link name, release it
428 	 * now.  This is useful for interfaces that can change
429 	 * link types, and thus switch link names often.
430 	 */
431 	if (ifp->if_sadl != NULL)
432 		if_free_sadl(ifp);
433 
434 	ifa = if_dl_create(ifp, &sdl);
435 
436 	ifa_insert(ifp, ifa);
437 	if_sadl_setrefs(ifp, ifa);
438 }
439 
440 static void
441 if_deactivate_sadl(struct ifnet *ifp)
442 {
443 	struct ifaddr *ifa;
444 
445 	KASSERT(ifp->if_dl != NULL);
446 
447 	ifa = ifp->if_dl;
448 
449 	ifp->if_sadl = NULL;
450 
451 	ifnet_addrs[ifp->if_index] = NULL;
452 	ifafree(ifa);
453 	ifp->if_dl = NULL;
454 	ifafree(ifa);
455 }
456 
457 void
458 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
459     const struct sockaddr_dl *sdl)
460 {
461 	int s;
462 
463 	s = splnet();
464 
465 	if_deactivate_sadl(ifp);
466 
467 	if_sadl_setrefs(ifp, ifa);
468 	IFADDR_FOREACH(ifa, ifp)
469 		rtinit(ifa, RTM_LLINFO_UPD, 0);
470 	splx(s);
471 }
472 
473 /*
474  * Free the link level name for the specified interface.  This is
475  * a detach helper.  This is called from if_detach() or from
476  * link layer type specific detach functions.
477  */
478 void
479 if_free_sadl(struct ifnet *ifp)
480 {
481 	struct ifaddr *ifa;
482 	int s;
483 
484 	ifa = ifnet_addrs[ifp->if_index];
485 	if (ifa == NULL) {
486 		KASSERT(ifp->if_sadl == NULL);
487 		KASSERT(ifp->if_dl == NULL);
488 		return;
489 	}
490 
491 	KASSERT(ifp->if_sadl != NULL);
492 	KASSERT(ifp->if_dl != NULL);
493 
494 	s = splnet();
495 	rtinit(ifa, RTM_DELETE, 0);
496 	ifa_remove(ifp, ifa);
497 	if_deactivate_sadl(ifp);
498 	if (ifp->if_hwdl == ifa) {
499 		ifafree(ifa);
500 		ifp->if_hwdl = NULL;
501 	}
502 	splx(s);
503 }
504 
505 static void
506 if_getindex(ifnet_t *ifp)
507 {
508 	bool hitlimit = false;
509 
510 	mutex_enter(&index_gen_mtx);
511 	ifp->if_index_gen = index_gen++;
512 	mutex_exit(&index_gen_mtx);
513 
514 	ifp->if_index = if_index;
515 	if (ifindex2ifnet == NULL) {
516 		if_index++;
517 		goto skip;
518 	}
519 	while (if_byindex(ifp->if_index)) {
520 		/*
521 		 * If we hit USHRT_MAX, we skip back to 0 since
522 		 * there are a number of places where the value
523 		 * of if_index or if_index itself is compared
524 		 * to or stored in an unsigned short.  By
525 		 * jumping back, we won't botch those assignments
526 		 * or comparisons.
527 		 */
528 		if (++if_index == 0) {
529 			if_index = 1;
530 		} else if (if_index == USHRT_MAX) {
531 			/*
532 			 * However, if we have to jump back to
533 			 * zero *twice* without finding an empty
534 			 * slot in ifindex2ifnet[], then there
535 			 * there are too many (>65535) interfaces.
536 			 */
537 			if (hitlimit) {
538 				panic("too many interfaces");
539 			}
540 			hitlimit = true;
541 			if_index = 1;
542 		}
543 		ifp->if_index = if_index;
544 	}
545 skip:
546 	/*
547 	 * We have some arrays that should be indexed by if_index.
548 	 * since if_index will grow dynamically, they should grow too.
549 	 *	struct ifadd **ifnet_addrs
550 	 *	struct ifnet **ifindex2ifnet
551 	 */
552 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
553 	    ifp->if_index >= if_indexlim) {
554 		size_t m, n, oldlim;
555 		void *q;
556 
557 		oldlim = if_indexlim;
558 		while (ifp->if_index >= if_indexlim)
559 			if_indexlim <<= 1;
560 
561 		/* grow ifnet_addrs */
562 		m = oldlim * sizeof(struct ifaddr *);
563 		n = if_indexlim * sizeof(struct ifaddr *);
564 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
565 		if (ifnet_addrs != NULL) {
566 			memcpy(q, ifnet_addrs, m);
567 			free(ifnet_addrs, M_IFADDR);
568 		}
569 		ifnet_addrs = (struct ifaddr **)q;
570 
571 		/* grow ifindex2ifnet */
572 		m = oldlim * sizeof(struct ifnet *);
573 		n = if_indexlim * sizeof(struct ifnet *);
574 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
575 		if (ifindex2ifnet != NULL) {
576 			memcpy(q, ifindex2ifnet, m);
577 			free(ifindex2ifnet, M_IFADDR);
578 		}
579 		ifindex2ifnet = (struct ifnet **)q;
580 	}
581 	ifindex2ifnet[ifp->if_index] = ifp;
582 }
583 
584 /*
585  * Attach an interface to the list of "active" interfaces.
586  */
587 void
588 if_attach(ifnet_t *ifp)
589 {
590 	KASSERT(if_indexlim > 0);
591 	TAILQ_INIT(&ifp->if_addrlist);
592 	TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list);
593 
594 	if (ifioctl_attach(ifp) != 0)
595 		panic("%s: ifioctl_attach() failed", __func__);
596 
597 	if_getindex(ifp);
598 
599 	/*
600 	 * Link level name is allocated later by a separate call to
601 	 * if_alloc_sadl().
602 	 */
603 
604 	if (ifp->if_snd.ifq_maxlen == 0)
605 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
606 
607 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
608 
609 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
610 
611 	ifp->if_link_state = LINK_STATE_UNKNOWN;
612 
613 	ifp->if_capenable = 0;
614 	ifp->if_csum_flags_tx = 0;
615 	ifp->if_csum_flags_rx = 0;
616 
617 #ifdef ALTQ
618 	ifp->if_snd.altq_type = 0;
619 	ifp->if_snd.altq_disc = NULL;
620 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
621 	ifp->if_snd.altq_tbr  = NULL;
622 	ifp->if_snd.altq_ifp  = ifp;
623 #endif
624 
625 #ifdef NET_MPSAFE
626 	ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
627 #else
628 	ifp->if_snd.ifq_lock = NULL;
629 #endif
630 
631 	ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
632 	(void)pfil_run_hooks(if_pfil,
633 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
634 
635 	if (!STAILQ_EMPTY(&domains))
636 		if_attachdomain1(ifp);
637 
638 	/* Announce the interface. */
639 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
640 }
641 
642 void
643 if_attachdomain(void)
644 {
645 	struct ifnet *ifp;
646 	int s;
647 
648 	s = splnet();
649 	IFNET_FOREACH(ifp)
650 		if_attachdomain1(ifp);
651 	splx(s);
652 }
653 
654 void
655 if_attachdomain1(struct ifnet *ifp)
656 {
657 	struct domain *dp;
658 	int s;
659 
660 	s = splnet();
661 
662 	/* address family dependent data region */
663 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
664 	DOMAIN_FOREACH(dp) {
665 		if (dp->dom_ifattach != NULL)
666 			ifp->if_afdata[dp->dom_family] =
667 			    (*dp->dom_ifattach)(ifp);
668 	}
669 
670 	splx(s);
671 }
672 
673 /*
674  * Deactivate an interface.  This points all of the procedure
675  * handles at error stubs.  May be called from interrupt context.
676  */
677 void
678 if_deactivate(struct ifnet *ifp)
679 {
680 	int s;
681 
682 	s = splnet();
683 
684 	ifp->if_output	 = if_nulloutput;
685 	ifp->if_input	 = if_nullinput;
686 	ifp->if_start	 = if_nullstart;
687 	ifp->if_ioctl	 = if_nullioctl;
688 	ifp->if_init	 = if_nullinit;
689 	ifp->if_stop	 = if_nullstop;
690 	ifp->if_watchdog = if_nullwatchdog;
691 	ifp->if_drain	 = if_nulldrain;
692 
693 	/* No more packets may be enqueued. */
694 	ifp->if_snd.ifq_maxlen = 0;
695 
696 	splx(s);
697 }
698 
699 void
700 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
701 {
702 	struct ifaddr *ifa, *nifa;
703 
704 	IFADDR_FOREACH_SAFE(ifa, ifp, nifa) {
705 		if (ifa->ifa_addr->sa_family != family)
706 			continue;
707 		(*purgeaddr)(ifa);
708 	}
709 }
710 
711 /*
712  * Detach an interface from the list of "active" interfaces,
713  * freeing any resources as we go along.
714  *
715  * NOTE: This routine must be called with a valid thread context,
716  * as it may block.
717  */
718 void
719 if_detach(struct ifnet *ifp)
720 {
721 	struct socket so;
722 	struct ifaddr *ifa;
723 #ifdef IFAREF_DEBUG
724 	struct ifaddr *last_ifa = NULL;
725 #endif
726 	struct domain *dp;
727 	const struct protosw *pr;
728 	int s, i, family, purged;
729 	uint64_t xc;
730 
731 	/*
732 	 * XXX It's kind of lame that we have to have the
733 	 * XXX socket structure...
734 	 */
735 	memset(&so, 0, sizeof(so));
736 
737 	s = splnet();
738 
739 	/*
740 	 * Do an if_down() to give protocols a chance to do something.
741 	 */
742 	if_down(ifp);
743 
744 #ifdef ALTQ
745 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
746 		altq_disable(&ifp->if_snd);
747 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
748 		altq_detach(&ifp->if_snd);
749 #endif
750 
751 	if (ifp->if_snd.ifq_lock)
752 		mutex_obj_free(ifp->if_snd.ifq_lock);
753 
754 	sysctl_teardown(&ifp->if_sysctl_log);
755 
756 #if NCARP > 0
757 	/* Remove the interface from any carp group it is a part of.  */
758 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
759 		carp_ifdetach(ifp);
760 #endif
761 
762 	/*
763 	 * Rip all the addresses off the interface.  This should make
764 	 * all of the routes go away.
765 	 *
766 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
767 	 * from the list, including our "cursor", ifa.  For safety,
768 	 * and to honor the TAILQ abstraction, I just restart the
769 	 * loop after each removal.  Note that the loop will exit
770 	 * when all of the remaining ifaddrs belong to the AF_LINK
771 	 * family.  I am counting on the historical fact that at
772 	 * least one pr_usrreq in each address domain removes at
773 	 * least one ifaddr.
774 	 */
775 again:
776 	IFADDR_FOREACH(ifa, ifp) {
777 		family = ifa->ifa_addr->sa_family;
778 #ifdef IFAREF_DEBUG
779 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
780 		    ifa, family, ifa->ifa_refcnt);
781 		if (last_ifa != NULL && ifa == last_ifa)
782 			panic("if_detach: loop detected");
783 		last_ifa = ifa;
784 #endif
785 		if (family == AF_LINK)
786 			continue;
787 		dp = pffinddomain(family);
788 #ifdef DIAGNOSTIC
789 		if (dp == NULL)
790 			panic("if_detach: no domain for AF %d",
791 			    family);
792 #endif
793 		/*
794 		 * XXX These PURGEIF calls are redundant with the
795 		 * purge-all-families calls below, but are left in for
796 		 * now both to make a smaller change, and to avoid
797 		 * unplanned interactions with clearing of
798 		 * ifp->if_addrlist.
799 		 */
800 		purged = 0;
801 		for (pr = dp->dom_protosw;
802 		     pr < dp->dom_protoswNPROTOSW; pr++) {
803 			so.so_proto = pr;
804 			if (pr->pr_usrreqs) {
805 				(void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
806 				purged = 1;
807 			}
808 		}
809 		if (purged == 0) {
810 			/*
811 			 * XXX What's really the best thing to do
812 			 * XXX here?  --thorpej@NetBSD.org
813 			 */
814 			printf("if_detach: WARNING: AF %d not purged\n",
815 			    family);
816 			ifa_remove(ifp, ifa);
817 		}
818 		goto again;
819 	}
820 
821 	if_free_sadl(ifp);
822 
823 	/* Walk the routing table looking for stragglers. */
824 	for (i = 0; i <= AF_MAX; i++) {
825 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
826 			continue;
827 	}
828 
829 	DOMAIN_FOREACH(dp) {
830 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
831 		{
832 			void *p = ifp->if_afdata[dp->dom_family];
833 			if (p) {
834 				ifp->if_afdata[dp->dom_family] = NULL;
835 				(*dp->dom_ifdetach)(ifp, p);
836 			}
837 		}
838 
839 		/*
840 		 * One would expect multicast memberships (INET and
841 		 * INET6) on UDP sockets to be purged by the PURGEIF
842 		 * calls above, but if all addresses were removed from
843 		 * the interface prior to destruction, the calls will
844 		 * not be made (e.g. ppp, for which pppd(8) generally
845 		 * removes addresses before destroying the interface).
846 		 * Because there is no invariant that multicast
847 		 * memberships only exist for interfaces with IPv4
848 		 * addresses, we must call PURGEIF regardless of
849 		 * addresses.  (Protocols which might store ifnet
850 		 * pointers are marked with PR_PURGEIF.)
851 		 */
852 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
853 			so.so_proto = pr;
854 			if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF)
855 				(void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
856 		}
857 	}
858 
859 	(void)pfil_run_hooks(if_pfil,
860 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
861 	(void)pfil_head_destroy(ifp->if_pfil);
862 
863 	/* Announce that the interface is gone. */
864 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
865 
866 	ifindex2ifnet[ifp->if_index] = NULL;
867 
868 	TAILQ_REMOVE(&ifnet_list, ifp, if_list);
869 
870 	ifioctl_detach(ifp);
871 
872 	/*
873 	 * remove packets that came from ifp, from software interrupt queues.
874 	 */
875 	DOMAIN_FOREACH(dp) {
876 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
877 			struct ifqueue *iq = dp->dom_ifqueues[i];
878 			if (iq == NULL)
879 				break;
880 			dp->dom_ifqueues[i] = NULL;
881 			if_detach_queues(ifp, iq);
882 		}
883 	}
884 
885 	/*
886 	 * IP queues have to be processed separately: net-queue barrier
887 	 * ensures that the packets are dequeued while a cross-call will
888 	 * ensure that the interrupts have completed. FIXME: not quite..
889 	 */
890 #ifdef INET
891 	pktq_barrier(ip_pktq);
892 #endif
893 #ifdef INET6
894 	pktq_barrier(ip6_pktq);
895 #endif
896 	xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
897 	xc_wait(xc);
898 
899 	splx(s);
900 }
901 
902 static void
903 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
904 {
905 	struct mbuf *m, *prev, *next;
906 
907 	prev = NULL;
908 	for (m = q->ifq_head; m != NULL; m = next) {
909 		KASSERT((m->m_flags & M_PKTHDR) != 0);
910 
911 		next = m->m_nextpkt;
912 		if (m->m_pkthdr.rcvif != ifp) {
913 			prev = m;
914 			continue;
915 		}
916 
917 		if (prev != NULL)
918 			prev->m_nextpkt = m->m_nextpkt;
919 		else
920 			q->ifq_head = m->m_nextpkt;
921 		if (q->ifq_tail == m)
922 			q->ifq_tail = prev;
923 		q->ifq_len--;
924 
925 		m->m_nextpkt = NULL;
926 		m_freem(m);
927 		IF_DROP(q);
928 	}
929 }
930 
931 /*
932  * Callback for a radix tree walk to delete all references to an
933  * ifnet.
934  */
935 static int
936 if_rt_walktree(struct rtentry *rt, void *v)
937 {
938 	struct ifnet *ifp = (struct ifnet *)v;
939 	int error;
940 
941 	if (rt->rt_ifp != ifp)
942 		return 0;
943 
944 	/* Delete the entry. */
945 	++rt->rt_refcnt;
946 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
947 	    rt_mask(rt), rt->rt_flags, NULL);
948 	KASSERT((rt->rt_flags & RTF_UP) == 0);
949 	rt->rt_ifp = NULL;
950 	rtfree(rt);
951 	if (error != 0)
952 		printf("%s: warning: unable to delete rtentry @ %p, "
953 		    "error = %d\n", ifp->if_xname, rt, error);
954 	return ERESTART;
955 }
956 
957 /*
958  * Create a clone network interface.
959  */
960 int
961 if_clone_create(const char *name)
962 {
963 	struct if_clone *ifc;
964 	int unit;
965 
966 	ifc = if_clone_lookup(name, &unit);
967 	if (ifc == NULL)
968 		return EINVAL;
969 
970 	if (ifunit(name) != NULL)
971 		return EEXIST;
972 
973 	return (*ifc->ifc_create)(ifc, unit);
974 }
975 
976 /*
977  * Destroy a clone network interface.
978  */
979 int
980 if_clone_destroy(const char *name)
981 {
982 	struct if_clone *ifc;
983 	struct ifnet *ifp;
984 
985 	ifc = if_clone_lookup(name, NULL);
986 	if (ifc == NULL)
987 		return EINVAL;
988 
989 	ifp = ifunit(name);
990 	if (ifp == NULL)
991 		return ENXIO;
992 
993 	if (ifc->ifc_destroy == NULL)
994 		return EOPNOTSUPP;
995 
996 	return (*ifc->ifc_destroy)(ifp);
997 }
998 
999 /*
1000  * Look up a network interface cloner.
1001  */
1002 static struct if_clone *
1003 if_clone_lookup(const char *name, int *unitp)
1004 {
1005 	struct if_clone *ifc;
1006 	const char *cp;
1007 	char *dp, ifname[IFNAMSIZ + 3];
1008 	int unit;
1009 
1010 	strcpy(ifname, "if_");
1011 	/* separate interface name from unit */
1012 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
1013 	    *cp && (*cp < '0' || *cp > '9');)
1014 		*dp++ = *cp++;
1015 
1016 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
1017 		return NULL;	/* No name or unit number */
1018 	*dp++ = '\0';
1019 
1020 again:
1021 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
1022 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
1023 			break;
1024 	}
1025 
1026 	if (ifc == NULL) {
1027 		if (*ifname == '\0' ||
1028 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
1029 			return NULL;
1030 		*ifname = '\0';
1031 		goto again;
1032 	}
1033 
1034 	unit = 0;
1035 	while (cp - name < IFNAMSIZ && *cp) {
1036 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1037 			/* Bogus unit number. */
1038 			return NULL;
1039 		}
1040 		unit = (unit * 10) + (*cp++ - '0');
1041 	}
1042 
1043 	if (unitp != NULL)
1044 		*unitp = unit;
1045 	return ifc;
1046 }
1047 
1048 /*
1049  * Register a network interface cloner.
1050  */
1051 void
1052 if_clone_attach(struct if_clone *ifc)
1053 {
1054 
1055 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1056 	if_cloners_count++;
1057 }
1058 
1059 /*
1060  * Unregister a network interface cloner.
1061  */
1062 void
1063 if_clone_detach(struct if_clone *ifc)
1064 {
1065 
1066 	LIST_REMOVE(ifc, ifc_list);
1067 	if_cloners_count--;
1068 }
1069 
1070 /*
1071  * Provide list of interface cloners to userspace.
1072  */
1073 static int
1074 if_clone_list(struct if_clonereq *ifcr)
1075 {
1076 	char outbuf[IFNAMSIZ], *dst;
1077 	struct if_clone *ifc;
1078 	int count, error = 0;
1079 
1080 	ifcr->ifcr_total = if_cloners_count;
1081 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1082 		/* Just asking how many there are. */
1083 		return 0;
1084 	}
1085 
1086 	if (ifcr->ifcr_count < 0)
1087 		return EINVAL;
1088 
1089 	count = (if_cloners_count < ifcr->ifcr_count) ?
1090 	    if_cloners_count : ifcr->ifcr_count;
1091 
1092 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1093 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1094 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1095 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1096 			return ENAMETOOLONG;
1097 		error = copyout(outbuf, dst, sizeof(outbuf));
1098 		if (error != 0)
1099 			break;
1100 	}
1101 
1102 	return error;
1103 }
1104 
1105 void
1106 ifaref(struct ifaddr *ifa)
1107 {
1108 	ifa->ifa_refcnt++;
1109 }
1110 
1111 void
1112 ifafree(struct ifaddr *ifa)
1113 {
1114 	KASSERT(ifa != NULL);
1115 	KASSERT(ifa->ifa_refcnt > 0);
1116 
1117 	if (--ifa->ifa_refcnt == 0) {
1118 		free(ifa, M_IFADDR);
1119 	}
1120 }
1121 
1122 void
1123 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1124 {
1125 	ifa->ifa_ifp = ifp;
1126 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1127 	ifaref(ifa);
1128 }
1129 
1130 void
1131 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1132 {
1133 	KASSERT(ifa->ifa_ifp == ifp);
1134 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1135 	ifafree(ifa);
1136 }
1137 
1138 static inline int
1139 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1140 {
1141 	return sockaddr_cmp(sa1, sa2) == 0;
1142 }
1143 
1144 /*
1145  * Locate an interface based on a complete address.
1146  */
1147 /*ARGSUSED*/
1148 struct ifaddr *
1149 ifa_ifwithaddr(const struct sockaddr *addr)
1150 {
1151 	struct ifnet *ifp;
1152 	struct ifaddr *ifa;
1153 
1154 	IFNET_FOREACH(ifp) {
1155 		if (ifp->if_output == if_nulloutput)
1156 			continue;
1157 		IFADDR_FOREACH(ifa, ifp) {
1158 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1159 				continue;
1160 			if (equal(addr, ifa->ifa_addr))
1161 				return ifa;
1162 			if ((ifp->if_flags & IFF_BROADCAST) &&
1163 			    ifa->ifa_broadaddr &&
1164 			    /* IP6 doesn't have broadcast */
1165 			    ifa->ifa_broadaddr->sa_len != 0 &&
1166 			    equal(ifa->ifa_broadaddr, addr))
1167 				return ifa;
1168 		}
1169 	}
1170 	return NULL;
1171 }
1172 
1173 /*
1174  * Locate the point to point interface with a given destination address.
1175  */
1176 /*ARGSUSED*/
1177 struct ifaddr *
1178 ifa_ifwithdstaddr(const struct sockaddr *addr)
1179 {
1180 	struct ifnet *ifp;
1181 	struct ifaddr *ifa;
1182 
1183 	IFNET_FOREACH(ifp) {
1184 		if (ifp->if_output == if_nulloutput)
1185 			continue;
1186 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1187 			continue;
1188 		IFADDR_FOREACH(ifa, ifp) {
1189 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1190 			    ifa->ifa_dstaddr == NULL)
1191 				continue;
1192 			if (equal(addr, ifa->ifa_dstaddr))
1193 				return ifa;
1194 		}
1195 	}
1196 	return NULL;
1197 }
1198 
1199 /*
1200  * Find an interface on a specific network.  If many, choice
1201  * is most specific found.
1202  */
1203 struct ifaddr *
1204 ifa_ifwithnet(const struct sockaddr *addr)
1205 {
1206 	struct ifnet *ifp;
1207 	struct ifaddr *ifa;
1208 	const struct sockaddr_dl *sdl;
1209 	struct ifaddr *ifa_maybe = 0;
1210 	u_int af = addr->sa_family;
1211 	const char *addr_data = addr->sa_data, *cplim;
1212 
1213 	if (af == AF_LINK) {
1214 		sdl = satocsdl(addr);
1215 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1216 		    ifindex2ifnet[sdl->sdl_index] &&
1217 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1218 			return ifnet_addrs[sdl->sdl_index];
1219 	}
1220 #ifdef NETATALK
1221 	if (af == AF_APPLETALK) {
1222 		const struct sockaddr_at *sat, *sat2;
1223 		sat = (const struct sockaddr_at *)addr;
1224 		IFNET_FOREACH(ifp) {
1225 			if (ifp->if_output == if_nulloutput)
1226 				continue;
1227 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1228 			if (ifa == NULL)
1229 				continue;
1230 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1231 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1232 				return ifa; /* exact match */
1233 			if (ifa_maybe == NULL) {
1234 				/* else keep the if with the right range */
1235 				ifa_maybe = ifa;
1236 			}
1237 		}
1238 		return ifa_maybe;
1239 	}
1240 #endif
1241 	IFNET_FOREACH(ifp) {
1242 		if (ifp->if_output == if_nulloutput)
1243 			continue;
1244 		IFADDR_FOREACH(ifa, ifp) {
1245 			const char *cp, *cp2, *cp3;
1246 
1247 			if (ifa->ifa_addr->sa_family != af ||
1248 			    ifa->ifa_netmask == NULL)
1249  next:				continue;
1250 			cp = addr_data;
1251 			cp2 = ifa->ifa_addr->sa_data;
1252 			cp3 = ifa->ifa_netmask->sa_data;
1253 			cplim = (const char *)ifa->ifa_netmask +
1254 			    ifa->ifa_netmask->sa_len;
1255 			while (cp3 < cplim) {
1256 				if ((*cp++ ^ *cp2++) & *cp3++) {
1257 					/* want to continue for() loop */
1258 					goto next;
1259 				}
1260 			}
1261 			if (ifa_maybe == NULL ||
1262 			    rn_refines((void *)ifa->ifa_netmask,
1263 			    (void *)ifa_maybe->ifa_netmask))
1264 				ifa_maybe = ifa;
1265 		}
1266 	}
1267 	return ifa_maybe;
1268 }
1269 
1270 /*
1271  * Find the interface of the addresss.
1272  */
1273 struct ifaddr *
1274 ifa_ifwithladdr(const struct sockaddr *addr)
1275 {
1276 	struct ifaddr *ia;
1277 
1278 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1279 	    (ia = ifa_ifwithnet(addr)))
1280 		return ia;
1281 	return NULL;
1282 }
1283 
1284 /*
1285  * Find an interface using a specific address family
1286  */
1287 struct ifaddr *
1288 ifa_ifwithaf(int af)
1289 {
1290 	struct ifnet *ifp;
1291 	struct ifaddr *ifa;
1292 
1293 	IFNET_FOREACH(ifp) {
1294 		if (ifp->if_output == if_nulloutput)
1295 			continue;
1296 		IFADDR_FOREACH(ifa, ifp) {
1297 			if (ifa->ifa_addr->sa_family == af)
1298 				return ifa;
1299 		}
1300 	}
1301 	return NULL;
1302 }
1303 
1304 /*
1305  * Find an interface address specific to an interface best matching
1306  * a given address.
1307  */
1308 struct ifaddr *
1309 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1310 {
1311 	struct ifaddr *ifa;
1312 	const char *cp, *cp2, *cp3;
1313 	const char *cplim;
1314 	struct ifaddr *ifa_maybe = 0;
1315 	u_int af = addr->sa_family;
1316 
1317 	if (ifp->if_output == if_nulloutput)
1318 		return NULL;
1319 
1320 	if (af >= AF_MAX)
1321 		return NULL;
1322 
1323 	IFADDR_FOREACH(ifa, ifp) {
1324 		if (ifa->ifa_addr->sa_family != af)
1325 			continue;
1326 		ifa_maybe = ifa;
1327 		if (ifa->ifa_netmask == NULL) {
1328 			if (equal(addr, ifa->ifa_addr) ||
1329 			    (ifa->ifa_dstaddr &&
1330 			     equal(addr, ifa->ifa_dstaddr)))
1331 				return ifa;
1332 			continue;
1333 		}
1334 		cp = addr->sa_data;
1335 		cp2 = ifa->ifa_addr->sa_data;
1336 		cp3 = ifa->ifa_netmask->sa_data;
1337 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1338 		for (; cp3 < cplim; cp3++) {
1339 			if ((*cp++ ^ *cp2++) & *cp3)
1340 				break;
1341 		}
1342 		if (cp3 == cplim)
1343 			return ifa;
1344 	}
1345 	return ifa_maybe;
1346 }
1347 
1348 /*
1349  * Default action when installing a route with a Link Level gateway.
1350  * Lookup an appropriate real ifa to point to.
1351  * This should be moved to /sys/net/link.c eventually.
1352  */
1353 void
1354 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1355 {
1356 	struct ifaddr *ifa;
1357 	const struct sockaddr *dst;
1358 	struct ifnet *ifp;
1359 
1360 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1361 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1362 		return;
1363 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1364 		rt_replace_ifa(rt, ifa);
1365 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1366 			ifa->ifa_rtrequest(cmd, rt, info);
1367 	}
1368 }
1369 
1370 /*
1371  * Handle a change in the interface link state.
1372  * XXX: We should listen to the routing socket in-kernel rather
1373  * than calling in6_if_link_* functions directly from here.
1374  */
1375 void
1376 if_link_state_change(struct ifnet *ifp, int link_state)
1377 {
1378 	int s;
1379 #if defined(DEBUG) || defined(INET6)
1380 	int old_link_state;
1381 #endif
1382 
1383 	s = splnet();
1384 	if (ifp->if_link_state == link_state) {
1385 		splx(s);
1386 		return;
1387 	}
1388 
1389 #if defined(DEBUG) || defined(INET6)
1390 	old_link_state = ifp->if_link_state;
1391 #endif
1392 	ifp->if_link_state = link_state;
1393 #ifdef DEBUG
1394 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
1395 		link_state == LINK_STATE_UP ? "UP" :
1396 		link_state == LINK_STATE_DOWN ? "DOWN" :
1397 		"UNKNOWN",
1398 		 old_link_state == LINK_STATE_UP ? "UP" :
1399 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
1400 		"UNKNOWN");
1401 #endif
1402 
1403 #ifdef INET6
1404 	/*
1405 	 * When going from UNKNOWN to UP, we need to mark existing
1406 	 * IPv6 addresses as tentative and restart DAD as we may have
1407 	 * erroneously not found a duplicate.
1408 	 *
1409 	 * This needs to happen before rt_ifmsg to avoid a race where
1410 	 * listeners would have an address and expect it to work right
1411 	 * away.
1412 	 */
1413 	if (in6_present && link_state == LINK_STATE_UP &&
1414 	    old_link_state == LINK_STATE_UNKNOWN)
1415 		in6_if_link_down(ifp);
1416 #endif
1417 
1418 	/* Notify that the link state has changed. */
1419 	rt_ifmsg(ifp);
1420 
1421 #if NCARP > 0
1422 	if (ifp->if_carp)
1423 		carp_carpdev_state(ifp);
1424 #endif
1425 
1426 #ifdef INET6
1427 	if (in6_present) {
1428 		if (link_state == LINK_STATE_DOWN)
1429 			in6_if_link_down(ifp);
1430 		else if (link_state == LINK_STATE_UP)
1431 			in6_if_link_up(ifp);
1432 	}
1433 #endif
1434 
1435 	splx(s);
1436 }
1437 
1438 /*
1439  * Mark an interface down and notify protocols of
1440  * the transition.
1441  * NOTE: must be called at splsoftnet or equivalent.
1442  */
1443 void
1444 if_down(struct ifnet *ifp)
1445 {
1446 	struct ifaddr *ifa;
1447 
1448 	ifp->if_flags &= ~IFF_UP;
1449 	nanotime(&ifp->if_lastchange);
1450 	IFADDR_FOREACH(ifa, ifp)
1451 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1452 	IFQ_PURGE(&ifp->if_snd);
1453 #if NCARP > 0
1454 	if (ifp->if_carp)
1455 		carp_carpdev_state(ifp);
1456 #endif
1457 	rt_ifmsg(ifp);
1458 #ifdef INET6
1459 	if (in6_present)
1460 		in6_if_down(ifp);
1461 #endif
1462 }
1463 
1464 /*
1465  * Mark an interface up and notify protocols of
1466  * the transition.
1467  * NOTE: must be called at splsoftnet or equivalent.
1468  */
1469 void
1470 if_up(struct ifnet *ifp)
1471 {
1472 #ifdef notyet
1473 	struct ifaddr *ifa;
1474 #endif
1475 
1476 	ifp->if_flags |= IFF_UP;
1477 	nanotime(&ifp->if_lastchange);
1478 #ifdef notyet
1479 	/* this has no effect on IP, and will kill all ISO connections XXX */
1480 	IFADDR_FOREACH(ifa, ifp)
1481 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1482 #endif
1483 #if NCARP > 0
1484 	if (ifp->if_carp)
1485 		carp_carpdev_state(ifp);
1486 #endif
1487 	rt_ifmsg(ifp);
1488 #ifdef INET6
1489 	if (in6_present)
1490 		in6_if_up(ifp);
1491 #endif
1492 }
1493 
1494 /*
1495  * Handle interface watchdog timer routines.  Called
1496  * from softclock, we decrement timers (if set) and
1497  * call the appropriate interface routine on expiration.
1498  */
1499 void
1500 if_slowtimo(void *arg)
1501 {
1502 	struct ifnet *ifp;
1503 	int s = splnet();
1504 
1505 	IFNET_FOREACH(ifp) {
1506 		if (ifp->if_timer == 0 || --ifp->if_timer)
1507 			continue;
1508 		if (ifp->if_watchdog != NULL)
1509 			(*ifp->if_watchdog)(ifp);
1510 	}
1511 	splx(s);
1512 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1513 }
1514 
1515 /*
1516  * Set/clear promiscuous mode on interface ifp based on the truth value
1517  * of pswitch.  The calls are reference counted so that only the first
1518  * "on" request actually has an effect, as does the final "off" request.
1519  * Results are undefined if the "off" and "on" requests are not matched.
1520  */
1521 int
1522 ifpromisc(struct ifnet *ifp, int pswitch)
1523 {
1524 	int pcount, ret;
1525 	short nflags;
1526 
1527 	pcount = ifp->if_pcount;
1528 	if (pswitch) {
1529 		/*
1530 		 * Allow the device to be "placed" into promiscuous
1531 		 * mode even if it is not configured up.  It will
1532 		 * consult IFF_PROMISC when it is brought up.
1533 		 */
1534 		if (ifp->if_pcount++ != 0)
1535 			return 0;
1536 		nflags = ifp->if_flags | IFF_PROMISC;
1537 	} else {
1538 		if (--ifp->if_pcount > 0)
1539 			return 0;
1540 		nflags = ifp->if_flags & ~IFF_PROMISC;
1541 	}
1542 	ret = if_flags_set(ifp, nflags);
1543 	/* Restore interface state if not successful. */
1544 	if (ret != 0) {
1545 		ifp->if_pcount = pcount;
1546 	}
1547 	return ret;
1548 }
1549 
1550 /*
1551  * Map interface name to
1552  * interface structure pointer.
1553  */
1554 struct ifnet *
1555 ifunit(const char *name)
1556 {
1557 	struct ifnet *ifp;
1558 	const char *cp = name;
1559 	u_int unit = 0;
1560 	u_int i;
1561 
1562 	/*
1563 	 * If the entire name is a number, treat it as an ifindex.
1564 	 */
1565 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1566 		unit = unit * 10 + (*cp - '0');
1567 	}
1568 
1569 	/*
1570 	 * If the number took all of the name, then it's a valid ifindex.
1571 	 */
1572 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1573 		if (unit >= if_indexlim)
1574 			return NULL;
1575 		ifp = ifindex2ifnet[unit];
1576 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1577 			return NULL;
1578 		return ifp;
1579 	}
1580 
1581 	IFNET_FOREACH(ifp) {
1582 		if (ifp->if_output == if_nulloutput)
1583 			continue;
1584 	 	if (strcmp(ifp->if_xname, name) == 0)
1585 			return ifp;
1586 	}
1587 	return NULL;
1588 }
1589 
1590 ifnet_t *
1591 if_byindex(u_int idx)
1592 {
1593 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1594 }
1595 
1596 /* common */
1597 int
1598 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1599 {
1600 	int s;
1601 	struct ifreq *ifr;
1602 	struct ifcapreq *ifcr;
1603 	struct ifdatareq *ifdr;
1604 
1605 	switch (cmd) {
1606 	case SIOCSIFCAP:
1607 		ifcr = data;
1608 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1609 			return EINVAL;
1610 
1611 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1612 			return 0;
1613 
1614 		ifp->if_capenable = ifcr->ifcr_capenable;
1615 
1616 		/* Pre-compute the checksum flags mask. */
1617 		ifp->if_csum_flags_tx = 0;
1618 		ifp->if_csum_flags_rx = 0;
1619 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1620 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1621 		}
1622 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1623 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1624 		}
1625 
1626 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1627 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1628 		}
1629 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1630 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1631 		}
1632 
1633 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1634 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1635 		}
1636 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1637 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1638 		}
1639 
1640 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1641 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1642 		}
1643 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1644 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1645 		}
1646 
1647 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1648 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1649 		}
1650 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1651 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1652 		}
1653 		if (ifp->if_flags & IFF_UP)
1654 			return ENETRESET;
1655 		return 0;
1656 	case SIOCSIFFLAGS:
1657 		ifr = data;
1658 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1659 			s = splnet();
1660 			if_down(ifp);
1661 			splx(s);
1662 		}
1663 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1664 			s = splnet();
1665 			if_up(ifp);
1666 			splx(s);
1667 		}
1668 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1669 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1670 		break;
1671 	case SIOCGIFFLAGS:
1672 		ifr = data;
1673 		ifr->ifr_flags = ifp->if_flags;
1674 		break;
1675 
1676 	case SIOCGIFMETRIC:
1677 		ifr = data;
1678 		ifr->ifr_metric = ifp->if_metric;
1679 		break;
1680 
1681 	case SIOCGIFMTU:
1682 		ifr = data;
1683 		ifr->ifr_mtu = ifp->if_mtu;
1684 		break;
1685 
1686 	case SIOCGIFDLT:
1687 		ifr = data;
1688 		ifr->ifr_dlt = ifp->if_dlt;
1689 		break;
1690 
1691 	case SIOCGIFCAP:
1692 		ifcr = data;
1693 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1694 		ifcr->ifcr_capenable = ifp->if_capenable;
1695 		break;
1696 
1697 	case SIOCSIFMETRIC:
1698 		ifr = data;
1699 		ifp->if_metric = ifr->ifr_metric;
1700 		break;
1701 
1702 	case SIOCGIFDATA:
1703 		ifdr = data;
1704 		ifdr->ifdr_data = ifp->if_data;
1705 		break;
1706 
1707 	case SIOCGIFINDEX:
1708 		ifr = data;
1709 		ifr->ifr_index = ifp->if_index;
1710 		break;
1711 
1712 	case SIOCZIFDATA:
1713 		ifdr = data;
1714 		ifdr->ifdr_data = ifp->if_data;
1715 		/*
1716 		 * Assumes that the volatile counters that can be
1717 		 * zero'ed are at the end of if_data.
1718 		 */
1719 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1720 		    offsetof(struct if_data, ifi_ipackets));
1721 		/*
1722 		 * The memset() clears to the bottm of if_data. In the area,
1723 		 * if_lastchange is included. Please be careful if new entry
1724 		 * will be added into if_data or rewite this.
1725 		 *
1726 		 * And also, update if_lastchnage.
1727 		 */
1728 		getnanotime(&ifp->if_lastchange);
1729 		break;
1730 	case SIOCSIFMTU:
1731 		ifr = data;
1732 		if (ifp->if_mtu == ifr->ifr_mtu)
1733 			break;
1734 		ifp->if_mtu = ifr->ifr_mtu;
1735 		/*
1736 		 * If the link MTU changed, do network layer specific procedure.
1737 		 */
1738 #ifdef INET6
1739 		if (in6_present)
1740 			nd6_setmtu(ifp);
1741 #endif
1742 		return ENETRESET;
1743 	default:
1744 		return ENOTTY;
1745 	}
1746 	return 0;
1747 }
1748 
1749 int
1750 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
1751 {
1752 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1753 	struct ifaddr *ifa;
1754 	const struct sockaddr *any, *sa;
1755 	union {
1756 		struct sockaddr sa;
1757 		struct sockaddr_storage ss;
1758 	} u, v;
1759 
1760 	switch (cmd) {
1761 	case SIOCSIFADDRPREF:
1762 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
1763 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1764 		    NULL) != 0)
1765 			return EPERM;
1766 	case SIOCGIFADDRPREF:
1767 		break;
1768 	default:
1769 		return EOPNOTSUPP;
1770 	}
1771 
1772 	/* sanity checks */
1773 	if (data == NULL || ifp == NULL) {
1774 		panic("invalid argument to %s", __func__);
1775 		/*NOTREACHED*/
1776 	}
1777 
1778 	/* address must be specified on ADD and DELETE */
1779 	sa = sstocsa(&ifap->ifap_addr);
1780 	if (sa->sa_family != sofamily(so))
1781 		return EINVAL;
1782 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1783 		return EINVAL;
1784 
1785 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1786 
1787 	IFADDR_FOREACH(ifa, ifp) {
1788 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1789 			continue;
1790 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1791 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1792 			break;
1793 	}
1794 	if (ifa == NULL)
1795 		return EADDRNOTAVAIL;
1796 
1797 	switch (cmd) {
1798 	case SIOCSIFADDRPREF:
1799 		ifa->ifa_preference = ifap->ifap_preference;
1800 		return 0;
1801 	case SIOCGIFADDRPREF:
1802 		/* fill in the if_laddrreq structure */
1803 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1804 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1805 		ifap->ifap_preference = ifa->ifa_preference;
1806 		return 0;
1807 	default:
1808 		return EOPNOTSUPP;
1809 	}
1810 }
1811 
1812 static void
1813 ifnet_lock_enter(struct ifnet_lock *il)
1814 {
1815 	uint64_t *nenter;
1816 
1817 	/* Before trying to acquire the mutex, increase the count of threads
1818 	 * who have entered or who wait to enter the critical section.
1819 	 * Avoid one costly locked memory transaction by keeping a count for
1820 	 * each CPU.
1821 	 */
1822 	nenter = percpu_getref(il->il_nenter);
1823 	(*nenter)++;
1824 	percpu_putref(il->il_nenter);
1825 	mutex_enter(&il->il_lock);
1826 }
1827 
1828 static void
1829 ifnet_lock_exit(struct ifnet_lock *il)
1830 {
1831 	/* Increase the count of threads who have exited the critical
1832 	 * section.  Increase while we still hold the lock.
1833 	 */
1834 	il->il_nexit++;
1835 	mutex_exit(&il->il_lock);
1836 }
1837 
1838 /*
1839  * Interface ioctls.
1840  */
1841 static int
1842 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1843 {
1844 	struct ifnet *ifp;
1845 	struct ifreq *ifr;
1846 	int error = 0;
1847 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1848 	u_long ocmd = cmd;
1849 #endif
1850 	short oif_flags;
1851 #ifdef COMPAT_OIFREQ
1852 	struct ifreq ifrb;
1853 	struct oifreq *oifr = NULL;
1854 #endif
1855 	int r;
1856 
1857 	switch (cmd) {
1858 #ifdef COMPAT_OIFREQ
1859 	case OSIOCGIFCONF:
1860 	case OOSIOCGIFCONF:
1861 		return compat_ifconf(cmd, data);
1862 #endif
1863 #ifdef COMPAT_OIFDATA
1864 	case OSIOCGIFDATA:
1865 	case OSIOCZIFDATA:
1866 		return compat_ifdatareq(l, cmd, data);
1867 #endif
1868 	case SIOCGIFCONF:
1869 		return ifconf(cmd, data);
1870 	case SIOCINITIFADDR:
1871 		return EPERM;
1872 	}
1873 
1874 #ifdef COMPAT_OIFREQ
1875 	cmd = compat_cvtcmd(cmd);
1876 	if (cmd != ocmd) {
1877 		oifr = data;
1878 		data = ifr = &ifrb;
1879 		ifreqo2n(oifr, ifr);
1880 	} else
1881 #endif
1882 		ifr = data;
1883 
1884 	ifp = ifunit(ifr->ifr_name);
1885 
1886 	switch (cmd) {
1887 	case SIOCIFCREATE:
1888 	case SIOCIFDESTROY:
1889 		if (l != NULL) {
1890 			error = kauth_authorize_network(l->l_cred,
1891 			    KAUTH_NETWORK_INTERFACE,
1892 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1893 			    (void *)cmd, NULL);
1894 			if (error != 0)
1895 				return error;
1896 		}
1897 		mutex_enter(&if_clone_mtx);
1898 		r = (cmd == SIOCIFCREATE) ?
1899 			if_clone_create(ifr->ifr_name) :
1900 			if_clone_destroy(ifr->ifr_name);
1901 		mutex_exit(&if_clone_mtx);
1902 		return r;
1903 
1904 	case SIOCIFGCLONERS:
1905 		return if_clone_list((struct if_clonereq *)data);
1906 	}
1907 
1908 	if (ifp == NULL)
1909 		return ENXIO;
1910 
1911 	switch (cmd) {
1912 	case SIOCALIFADDR:
1913 	case SIOCDLIFADDR:
1914 	case SIOCSIFADDRPREF:
1915 	case SIOCSIFFLAGS:
1916 	case SIOCSIFCAP:
1917 	case SIOCSIFMETRIC:
1918 	case SIOCZIFDATA:
1919 	case SIOCSIFMTU:
1920 	case SIOCSIFPHYADDR:
1921 	case SIOCDIFPHYADDR:
1922 #ifdef INET6
1923 	case SIOCSIFPHYADDR_IN6:
1924 #endif
1925 	case SIOCSLIFPHYADDR:
1926 	case SIOCADDMULTI:
1927 	case SIOCDELMULTI:
1928 	case SIOCSIFMEDIA:
1929 	case SIOCSDRVSPEC:
1930 	case SIOCG80211:
1931 	case SIOCS80211:
1932 	case SIOCS80211NWID:
1933 	case SIOCS80211NWKEY:
1934 	case SIOCS80211POWER:
1935 	case SIOCS80211BSSID:
1936 	case SIOCS80211CHANNEL:
1937 	case SIOCSLINKSTR:
1938 		if (l != NULL) {
1939 			error = kauth_authorize_network(l->l_cred,
1940 			    KAUTH_NETWORK_INTERFACE,
1941 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1942 			    (void *)cmd, NULL);
1943 			if (error != 0)
1944 				return error;
1945 		}
1946 	}
1947 
1948 	oif_flags = ifp->if_flags;
1949 
1950 	ifnet_lock_enter(ifp->if_ioctl_lock);
1951 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1952 	if (error != ENOTTY)
1953 		;
1954 	else if (so->so_proto == NULL)
1955 		error = EOPNOTSUPP;
1956 	else {
1957 #ifdef COMPAT_OSOCK
1958 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1959 #else
1960 		error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so,
1961 		    cmd, data, ifp);
1962 #endif
1963 	}
1964 
1965 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1966 #ifdef INET6
1967 		if (in6_present && (ifp->if_flags & IFF_UP) != 0) {
1968 			int s = splnet();
1969 			in6_if_up(ifp);
1970 			splx(s);
1971 		}
1972 #endif
1973 	}
1974 #ifdef COMPAT_OIFREQ
1975 	if (cmd != ocmd)
1976 		ifreqn2o(oifr, ifr);
1977 #endif
1978 
1979 	ifnet_lock_exit(ifp->if_ioctl_lock);
1980 	return error;
1981 }
1982 
1983 /* This callback adds to the sum in `arg' the number of
1984  * threads on `ci' who have entered or who wait to enter the
1985  * critical section.
1986  */
1987 static void
1988 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
1989 {
1990 	uint64_t *sum = arg, *nenter = p;
1991 
1992 	*sum += *nenter;
1993 }
1994 
1995 /* Return the number of threads who have entered or who wait
1996  * to enter the critical section on all CPUs.
1997  */
1998 static uint64_t
1999 ifnet_lock_entrances(struct ifnet_lock *il)
2000 {
2001 	uint64_t sum = 0;
2002 
2003 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
2004 
2005 	return sum;
2006 }
2007 
2008 static int
2009 ifioctl_attach(struct ifnet *ifp)
2010 {
2011 	struct ifnet_lock *il;
2012 
2013 	/* If the driver has not supplied its own if_ioctl, then
2014 	 * supply the default.
2015 	 */
2016 	if (ifp->if_ioctl == NULL)
2017 		ifp->if_ioctl = ifioctl_common;
2018 
2019 	/* Create an ifnet_lock for synchronizing ifioctls. */
2020 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
2021 		return ENOMEM;
2022 
2023 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
2024 	if (il->il_nenter == NULL) {
2025 		kmem_free(il, sizeof(*il));
2026 		return ENOMEM;
2027 	}
2028 
2029 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
2030 	cv_init(&il->il_emptied, ifp->if_xname);
2031 
2032 	ifp->if_ioctl_lock = il;
2033 
2034 	return 0;
2035 }
2036 
2037 /*
2038  * This must not be called until after `ifp' has been withdrawn from the
2039  * ifnet tables so that ifioctl() cannot get a handle on it by calling
2040  * ifunit().
2041  */
2042 static void
2043 ifioctl_detach(struct ifnet *ifp)
2044 {
2045 	struct ifnet_lock *il;
2046 
2047 	il = ifp->if_ioctl_lock;
2048 	mutex_enter(&il->il_lock);
2049 	/* Install if_nullioctl to make sure that any thread that
2050 	 * subsequently enters the critical section will quit it
2051 	 * immediately and signal the condition variable that we
2052 	 * wait on, below.
2053 	 */
2054 	ifp->if_ioctl = if_nullioctl;
2055 	/* Sleep while threads are still in the critical section or
2056 	 * wait to enter it.
2057 	 */
2058 	while (ifnet_lock_entrances(il) != il->il_nexit)
2059 		cv_wait(&il->il_emptied, &il->il_lock);
2060 	/* At this point, we are the only thread still in the critical
2061 	 * section, and no new thread can get a handle on the ifioctl
2062 	 * lock, so it is safe to free its memory.
2063 	 */
2064 	mutex_exit(&il->il_lock);
2065 	ifp->if_ioctl_lock = NULL;
2066 	percpu_free(il->il_nenter, sizeof(uint64_t));
2067 	il->il_nenter = NULL;
2068 	cv_destroy(&il->il_emptied);
2069 	mutex_destroy(&il->il_lock);
2070 	kmem_free(il, sizeof(*il));
2071 }
2072 
2073 /*
2074  * Return interface configuration
2075  * of system.  List may be used
2076  * in later ioctl's (above) to get
2077  * other information.
2078  *
2079  * Each record is a struct ifreq.  Before the addition of
2080  * sockaddr_storage, the API rule was that sockaddr flavors that did
2081  * not fit would extend beyond the struct ifreq, with the next struct
2082  * ifreq starting sa_len beyond the struct sockaddr.  Because the
2083  * union in struct ifreq includes struct sockaddr_storage, every kind
2084  * of sockaddr must fit.  Thus, there are no longer any overlength
2085  * records.
2086  *
2087  * Records are added to the user buffer if they fit, and ifc_len is
2088  * adjusted to the length that was written.  Thus, the user is only
2089  * assured of getting the complete list if ifc_len on return is at
2090  * least sizeof(struct ifreq) less than it was on entry.
2091  *
2092  * If the user buffer pointer is NULL, this routine copies no data and
2093  * returns the amount of space that would be needed.
2094  *
2095  * Invariants:
2096  * ifrp points to the next part of the user's buffer to be used.  If
2097  * ifrp != NULL, space holds the number of bytes remaining that we may
2098  * write at ifrp.  Otherwise, space holds the number of bytes that
2099  * would have been written had there been adequate space.
2100  */
2101 /*ARGSUSED*/
2102 int
2103 ifconf(u_long cmd, void *data)
2104 {
2105 	struct ifconf *ifc = (struct ifconf *)data;
2106 	struct ifnet *ifp;
2107 	struct ifaddr *ifa;
2108 	struct ifreq ifr, *ifrp;
2109 	int space, error = 0;
2110 	const int sz = (int)sizeof(struct ifreq);
2111 
2112 	if ((ifrp = ifc->ifc_req) == NULL)
2113 		space = 0;
2114 	else
2115 		space = ifc->ifc_len;
2116 	IFNET_FOREACH(ifp) {
2117 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2118 		    sizeof(ifr.ifr_name));
2119 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2120 			return ENAMETOOLONG;
2121 		if (IFADDR_EMPTY(ifp)) {
2122 			/* Interface with no addresses - send zero sockaddr. */
2123 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2124 			if (ifrp == NULL) {
2125 				space += sz;
2126 				continue;
2127 			}
2128 			if (space >= sz) {
2129 				error = copyout(&ifr, ifrp, sz);
2130 				if (error != 0)
2131 					return error;
2132 				ifrp++;
2133 				space -= sz;
2134 			}
2135 		}
2136 
2137 		IFADDR_FOREACH(ifa, ifp) {
2138 			struct sockaddr *sa = ifa->ifa_addr;
2139 			/* all sockaddrs must fit in sockaddr_storage */
2140 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2141 
2142 			if (ifrp == NULL) {
2143 				space += sz;
2144 				continue;
2145 			}
2146 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2147 			if (space >= sz) {
2148 				error = copyout(&ifr, ifrp, sz);
2149 				if (error != 0)
2150 					return (error);
2151 				ifrp++; space -= sz;
2152 			}
2153 		}
2154 	}
2155 	if (ifrp != NULL) {
2156 		KASSERT(0 <= space && space <= ifc->ifc_len);
2157 		ifc->ifc_len -= space;
2158 	} else {
2159 		KASSERT(space >= 0);
2160 		ifc->ifc_len = space;
2161 	}
2162 	return (0);
2163 }
2164 
2165 int
2166 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2167 {
2168 	uint8_t len;
2169 #ifdef COMPAT_OIFREQ
2170 	struct ifreq ifrb;
2171 	struct oifreq *oifr = NULL;
2172 	u_long ocmd = cmd;
2173 	cmd = compat_cvtcmd(cmd);
2174 	if (cmd != ocmd) {
2175 		oifr = (struct oifreq *)(void *)ifr;
2176 		ifr = &ifrb;
2177 		ifreqo2n(oifr, ifr);
2178 		len = sizeof(oifr->ifr_addr);
2179 	} else
2180 #endif
2181 		len = sizeof(ifr->ifr_ifru.ifru_space);
2182 
2183 	if (len < sa->sa_len)
2184 		return EFBIG;
2185 
2186 	memset(&ifr->ifr_addr, 0, len);
2187 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2188 
2189 #ifdef COMPAT_OIFREQ
2190 	if (cmd != ocmd)
2191 		ifreqn2o(oifr, ifr);
2192 #endif
2193 	return 0;
2194 }
2195 
2196 /*
2197  * Queue message on interface, and start output if interface
2198  * not yet active.
2199  */
2200 int
2201 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2202     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2203 {
2204 	int len = m->m_pkthdr.len;
2205 	int mflags = m->m_flags;
2206 	int s = splnet();
2207 	int error;
2208 
2209 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2210 	if (error != 0)
2211 		goto out;
2212 	ifp->if_obytes += len;
2213 	if (mflags & M_MCAST)
2214 		ifp->if_omcasts++;
2215 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2216 		(*ifp->if_start)(ifp);
2217 out:
2218 	splx(s);
2219 	return error;
2220 }
2221 
2222 /*
2223  * Queue message on interface, possibly using a second fast queue
2224  */
2225 int
2226 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2227     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2228 {
2229 	int error = 0;
2230 
2231 	if (ifq != NULL
2232 #ifdef ALTQ
2233 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2234 #endif
2235 	    ) {
2236 		if (IF_QFULL(ifq)) {
2237 			IF_DROP(&ifp->if_snd);
2238 			m_freem(m);
2239 			if (error == 0)
2240 				error = ENOBUFS;
2241 		} else
2242 			IF_ENQUEUE(ifq, m);
2243 	} else
2244 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2245 	if (error != 0) {
2246 		++ifp->if_oerrors;
2247 		return error;
2248 	}
2249 	return 0;
2250 }
2251 
2252 int
2253 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2254 {
2255 	int rc;
2256 
2257 	if (ifp->if_initaddr != NULL)
2258 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2259 	else if (src ||
2260 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2261 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2262 
2263 	return rc;
2264 }
2265 
2266 int
2267 if_flags_set(ifnet_t *ifp, const short flags)
2268 {
2269 	int rc;
2270 
2271 	if (ifp->if_setflags != NULL)
2272 		rc = (*ifp->if_setflags)(ifp, flags);
2273 	else {
2274 		short cantflags, chgdflags;
2275 		struct ifreq ifr;
2276 
2277 		chgdflags = ifp->if_flags ^ flags;
2278 		cantflags = chgdflags & IFF_CANTCHANGE;
2279 
2280 		if (cantflags != 0)
2281 			ifp->if_flags ^= cantflags;
2282 
2283                 /* Traditionally, we do not call if_ioctl after
2284                  * setting/clearing only IFF_PROMISC if the interface
2285                  * isn't IFF_UP.  Uphold that tradition.
2286 		 */
2287 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2288 			return 0;
2289 
2290 		memset(&ifr, 0, sizeof(ifr));
2291 
2292 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2293 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2294 
2295 		if (rc != 0 && cantflags != 0)
2296 			ifp->if_flags ^= cantflags;
2297 	}
2298 
2299 	return rc;
2300 }
2301 
2302 int
2303 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2304 {
2305 	int rc;
2306 	struct ifreq ifr;
2307 
2308 	if (ifp->if_mcastop != NULL)
2309 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2310 	else {
2311 		ifreq_setaddr(cmd, &ifr, sa);
2312 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2313 	}
2314 
2315 	return rc;
2316 }
2317 
2318 static void
2319 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2320     struct ifaltq *ifq)
2321 {
2322 	const struct sysctlnode *cnode, *rnode;
2323 
2324 	if (sysctl_createv(clog, 0, NULL, &rnode,
2325 		       CTLFLAG_PERMANENT,
2326 		       CTLTYPE_NODE, "interfaces",
2327 		       SYSCTL_DESCR("Per-interface controls"),
2328 		       NULL, 0, NULL, 0,
2329 		       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
2330 		goto bad;
2331 
2332 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2333 		       CTLFLAG_PERMANENT,
2334 		       CTLTYPE_NODE, ifname,
2335 		       SYSCTL_DESCR("Interface controls"),
2336 		       NULL, 0, NULL, 0,
2337 		       CTL_CREATE, CTL_EOL) != 0)
2338 		goto bad;
2339 
2340 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2341 		       CTLFLAG_PERMANENT,
2342 		       CTLTYPE_NODE, "sndq",
2343 		       SYSCTL_DESCR("Interface output queue controls"),
2344 		       NULL, 0, NULL, 0,
2345 		       CTL_CREATE, CTL_EOL) != 0)
2346 		goto bad;
2347 
2348 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2349 		       CTLFLAG_PERMANENT,
2350 		       CTLTYPE_INT, "len",
2351 		       SYSCTL_DESCR("Current output queue length"),
2352 		       NULL, 0, &ifq->ifq_len, 0,
2353 		       CTL_CREATE, CTL_EOL) != 0)
2354 		goto bad;
2355 
2356 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2357 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2358 		       CTLTYPE_INT, "maxlen",
2359 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2360 		       NULL, 0, &ifq->ifq_maxlen, 0,
2361 		       CTL_CREATE, CTL_EOL) != 0)
2362 		goto bad;
2363 
2364 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2365 		       CTLFLAG_PERMANENT,
2366 		       CTLTYPE_INT, "drops",
2367 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2368 		       NULL, 0, &ifq->ifq_drops, 0,
2369 		       CTL_CREATE, CTL_EOL) != 0)
2370 		goto bad;
2371 
2372 	return;
2373 bad:
2374 	printf("%s: could not attach sysctl nodes\n", ifname);
2375 	return;
2376 }
2377 
2378 #if defined(INET) || defined(INET6)
2379 
2380 #define	SYSCTL_NET_PKTQ(q, cn, c)					\
2381 	static int							\
2382 	sysctl_net_##q##_##cn(SYSCTLFN_ARGS)				\
2383 	{								\
2384 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c);	\
2385 	}
2386 
2387 #if defined(INET)
2388 static int
2389 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS)
2390 {
2391 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq);
2392 }
2393 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS)
2394 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS)
2395 #endif
2396 
2397 #if defined(INET6)
2398 static int
2399 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS)
2400 {
2401 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq);
2402 }
2403 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS)
2404 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS)
2405 #endif
2406 
2407 static void
2408 sysctl_net_pktq_setup(struct sysctllog **clog, int pf)
2409 {
2410 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
2411 	const char *pfname = NULL, *ipname = NULL;
2412 	int ipn = 0, qid = 0;
2413 
2414 	switch (pf) {
2415 #if defined(INET)
2416 	case PF_INET:
2417 		len_func = sysctl_net_ip_pktq_items;
2418 		maxlen_func = sysctl_net_ip_pktq_maxlen;
2419 		drops_func = sysctl_net_ip_pktq_drops;
2420 		pfname = "inet", ipn = IPPROTO_IP;
2421 		ipname = "ip", qid = IPCTL_IFQ;
2422 		break;
2423 #endif
2424 #if defined(INET6)
2425 	case PF_INET6:
2426 		len_func = sysctl_net_ip6_pktq_items;
2427 		maxlen_func = sysctl_net_ip6_pktq_maxlen;
2428 		drops_func = sysctl_net_ip6_pktq_drops;
2429 		pfname = "inet6", ipn = IPPROTO_IPV6;
2430 		ipname = "ip6", qid = IPV6CTL_IFQ;
2431 		break;
2432 #endif
2433 	default:
2434 		KASSERT(false);
2435 	}
2436 
2437 	sysctl_createv(clog, 0, NULL, NULL,
2438 		       CTLFLAG_PERMANENT,
2439 		       CTLTYPE_NODE, pfname, NULL,
2440 		       NULL, 0, NULL, 0,
2441 		       CTL_NET, pf, CTL_EOL);
2442 	sysctl_createv(clog, 0, NULL, NULL,
2443 		       CTLFLAG_PERMANENT,
2444 		       CTLTYPE_NODE, ipname, NULL,
2445 		       NULL, 0, NULL, 0,
2446 		       CTL_NET, pf, ipn, CTL_EOL);
2447 	sysctl_createv(clog, 0, NULL, NULL,
2448 		       CTLFLAG_PERMANENT,
2449 		       CTLTYPE_NODE, "ifq",
2450 		       SYSCTL_DESCR("Protocol input queue controls"),
2451 		       NULL, 0, NULL, 0,
2452 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2453 
2454 	sysctl_createv(clog, 0, NULL, NULL,
2455 		       CTLFLAG_PERMANENT,
2456 		       CTLTYPE_INT, "len",
2457 		       SYSCTL_DESCR("Current input queue length"),
2458 		       len_func, 0, NULL, 0,
2459 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2460 	sysctl_createv(clog, 0, NULL, NULL,
2461 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2462 		       CTLTYPE_INT, "maxlen",
2463 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2464 		       maxlen_func, 0, NULL, 0,
2465 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2466 	sysctl_createv(clog, 0, NULL, NULL,
2467 		       CTLFLAG_PERMANENT,
2468 		       CTLTYPE_INT, "drops",
2469 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2470 		       drops_func, 0, NULL, 0,
2471 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2472 }
2473 #endif /* INET || INET6 */
2474 
2475 static int
2476 if_sdl_sysctl(SYSCTLFN_ARGS)
2477 {
2478 	struct ifnet *ifp;
2479 	const struct sockaddr_dl *sdl;
2480 
2481 	if (namelen != 1)
2482 		return EINVAL;
2483 
2484 	ifp = if_byindex(name[0]);
2485 	if (ifp == NULL)
2486 		return ENODEV;
2487 
2488 	sdl = ifp->if_sadl;
2489 	if (sdl == NULL) {
2490 		*oldlenp = 0;
2491 		return 0;
2492 	}
2493 
2494 	if (oldp == NULL) {
2495 		*oldlenp = sdl->sdl_alen;
2496 		return 0;
2497 	}
2498 
2499 	if (*oldlenp >= sdl->sdl_alen)
2500 		*oldlenp = sdl->sdl_alen;
2501 	return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp);
2502 }
2503 
2504 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup")
2505 {
2506 	const struct sysctlnode *rnode = NULL;
2507 
2508 	sysctl_createv(clog, 0, NULL, &rnode,
2509 		       CTLFLAG_PERMANENT,
2510 		       CTLTYPE_NODE, "sdl",
2511 		       SYSCTL_DESCR("Get active link-layer address"),
2512 		       if_sdl_sysctl, 0, NULL, 0,
2513 		       CTL_NET, CTL_CREATE, CTL_EOL);
2514 }
2515