xref: /netbsd-src/sys/net/if.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: if.c,v 1.285 2014/07/01 10:16:02 ozaki-r 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 1.285 2014/07/01 10:16:02 ozaki-r Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 
100 #include <sys/param.h>
101 #include <sys/mbuf.h>
102 #include <sys/systm.h>
103 #include <sys/callout.h>
104 #include <sys/proc.h>
105 #include <sys/socket.h>
106 #include <sys/socketvar.h>
107 #include <sys/domain.h>
108 #include <sys/protosw.h>
109 #include <sys/kernel.h>
110 #include <sys/ioctl.h>
111 #include <sys/sysctl.h>
112 #include <sys/syslog.h>
113 #include <sys/kauth.h>
114 #include <sys/kmem.h>
115 #include <sys/xcall.h>
116 
117 #include <net/if.h>
118 #include <net/if_dl.h>
119 #include <net/if_ether.h>
120 #include <net/if_media.h>
121 #include <net80211/ieee80211.h>
122 #include <net80211/ieee80211_ioctl.h>
123 #include <net/if_types.h>
124 #include <net/radix.h>
125 #include <net/route.h>
126 #include <net/netisr.h>
127 #include <sys/module.h>
128 #ifdef NETATALK
129 #include <netatalk/at_extern.h>
130 #include <netatalk/at.h>
131 #endif
132 #include <net/pfil.h>
133 #include <netinet/in.h>
134 #include <netinet/in_var.h>
135 
136 #ifdef INET6
137 #include <netinet6/in6_var.h>
138 #include <netinet6/nd6.h>
139 #endif
140 
141 #include "carp.h"
142 #if NCARP > 0
143 #include <netinet/ip_carp.h>
144 #endif
145 
146 #include <compat/sys/sockio.h>
147 #include <compat/sys/socket.h>
148 
149 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
150 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
151 
152 /*
153  * Global list of interfaces.
154  */
155 struct ifnet_head		ifnet_list;
156 static ifnet_t **		ifindex2ifnet = NULL;
157 
158 static u_int			if_index = 1;
159 static size_t			if_indexlim = 0;
160 static uint64_t			index_gen;
161 static kmutex_t			index_gen_mtx;
162 
163 static struct ifaddr **		ifnet_addrs = NULL;
164 
165 static callout_t		if_slowtimo_ch;
166 
167 struct ifnet *lo0ifp;
168 int	ifqmaxlen = IFQ_MAXLEN;
169 
170 static int	if_rt_walktree(struct rtentry *, void *);
171 
172 static struct if_clone *if_clone_lookup(const char *, int *);
173 static int	if_clone_list(struct if_clonereq *);
174 
175 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
176 static int if_cloners_count;
177 
178 /* Packet filtering hook for interfaces. */
179 pfil_head_t *	if_pfil;
180 
181 static kauth_listener_t if_listener;
182 
183 static int doifioctl(struct socket *, u_long, void *, struct lwp *);
184 static int ifioctl_attach(struct ifnet *);
185 static void ifioctl_detach(struct ifnet *);
186 static void ifnet_lock_enter(struct ifnet_lock *);
187 static void ifnet_lock_exit(struct ifnet_lock *);
188 static void if_detach_queues(struct ifnet *, struct ifqueue *);
189 static void sysctl_sndq_setup(struct sysctllog **, const char *,
190     struct ifaltq *);
191 
192 #if defined(INET) || defined(INET6)
193 static void sysctl_net_pktq_setup(struct sysctllog **, int);
194 #endif
195 
196 static int
197 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
198     void *arg0, void *arg1, void *arg2, void *arg3)
199 {
200 	int result;
201 	enum kauth_network_req req;
202 
203 	result = KAUTH_RESULT_DEFER;
204 	req = (enum kauth_network_req)arg1;
205 
206 	if (action != KAUTH_NETWORK_INTERFACE)
207 		return result;
208 
209 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
210 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
211 		result = KAUTH_RESULT_ALLOW;
212 
213 	return result;
214 }
215 
216 /*
217  * Network interface utility routines.
218  *
219  * Routines with ifa_ifwith* names take sockaddr *'s as
220  * parameters.
221  */
222 void
223 ifinit(void)
224 {
225 #if defined(INET)
226 	sysctl_net_pktq_setup(NULL, PF_INET);
227 #endif
228 #ifdef INET6
229 	sysctl_net_pktq_setup(NULL, PF_INET6);
230 #endif
231 
232 	callout_init(&if_slowtimo_ch, 0);
233 	if_slowtimo(NULL);
234 
235 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
236 	    if_listener_cb, NULL);
237 
238 	/* interfaces are available, inform socket code */
239 	ifioctl = doifioctl;
240 }
241 
242 /*
243  * XXX Initialization before configure().
244  * XXX hack to get pfil_add_hook working in autoconf.
245  */
246 void
247 ifinit1(void)
248 {
249 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
250 	TAILQ_INIT(&ifnet_list);
251 	if_indexlim = 8;
252 
253 	if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
254 	KASSERT(if_pfil != NULL);
255 }
256 
257 ifnet_t *
258 if_alloc(u_char type)
259 {
260 	return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP);
261 }
262 
263 void
264 if_free(ifnet_t *ifp)
265 {
266 	kmem_free(ifp, sizeof(ifnet_t));
267 }
268 
269 void
270 if_initname(struct ifnet *ifp, const char *name, int unit)
271 {
272 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
273 	    "%s%d", name, unit);
274 }
275 
276 /*
277  * Null routines used while an interface is going away.  These routines
278  * just return an error.
279  */
280 
281 int
282 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
283     const struct sockaddr *so, struct rtentry *rt)
284 {
285 
286 	return ENXIO;
287 }
288 
289 void
290 if_nullinput(struct ifnet *ifp, struct mbuf *m)
291 {
292 
293 	/* Nothing. */
294 }
295 
296 void
297 if_nullstart(struct ifnet *ifp)
298 {
299 
300 	/* Nothing. */
301 }
302 
303 int
304 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
305 {
306 
307 	/* Wake ifioctl_detach(), who may wait for all threads to
308 	 * quit the critical section.
309 	 */
310 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
311 	return ENXIO;
312 }
313 
314 int
315 if_nullinit(struct ifnet *ifp)
316 {
317 
318 	return ENXIO;
319 }
320 
321 void
322 if_nullstop(struct ifnet *ifp, int disable)
323 {
324 
325 	/* Nothing. */
326 }
327 
328 void
329 if_nullwatchdog(struct ifnet *ifp)
330 {
331 
332 	/* Nothing. */
333 }
334 
335 void
336 if_nulldrain(struct ifnet *ifp)
337 {
338 
339 	/* Nothing. */
340 }
341 
342 void
343 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
344 {
345 	struct ifaddr *ifa;
346 	struct sockaddr_dl *sdl;
347 
348 	ifp->if_addrlen = addrlen;
349 	if_alloc_sadl(ifp);
350 	ifa = ifp->if_dl;
351 	sdl = satosdl(ifa->ifa_addr);
352 
353 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
354 	if (factory) {
355 		ifp->if_hwdl = ifp->if_dl;
356 		IFAREF(ifp->if_hwdl);
357 	}
358 	/* TBD routing socket */
359 }
360 
361 struct ifaddr *
362 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
363 {
364 	unsigned socksize, ifasize;
365 	int addrlen, namelen;
366 	struct sockaddr_dl *mask, *sdl;
367 	struct ifaddr *ifa;
368 
369 	namelen = strlen(ifp->if_xname);
370 	addrlen = ifp->if_addrlen;
371 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
372 	ifasize = sizeof(*ifa) + 2 * socksize;
373 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
374 
375 	sdl = (struct sockaddr_dl *)(ifa + 1);
376 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
377 
378 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
379 	    ifp->if_xname, namelen, NULL, addrlen);
380 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
381 	memset(&mask->sdl_data[0], 0xff, namelen);
382 	ifa->ifa_rtrequest = link_rtrequest;
383 	ifa->ifa_addr = (struct sockaddr *)sdl;
384 	ifa->ifa_netmask = (struct sockaddr *)mask;
385 
386 	*sdlp = sdl;
387 
388 	return ifa;
389 }
390 
391 static void
392 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
393 {
394 	const struct sockaddr_dl *sdl;
395 	ifnet_addrs[ifp->if_index] = ifa;
396 	IFAREF(ifa);
397 	ifp->if_dl = ifa;
398 	IFAREF(ifa);
399 	sdl = satosdl(ifa->ifa_addr);
400 	ifp->if_sadl = sdl;
401 }
402 
403 /*
404  * Allocate the link level name for the specified interface.  This
405  * is an attachment helper.  It must be called after ifp->if_addrlen
406  * is initialized, which may not be the case when if_attach() is
407  * called.
408  */
409 void
410 if_alloc_sadl(struct ifnet *ifp)
411 {
412 	struct ifaddr *ifa;
413 	const struct sockaddr_dl *sdl;
414 
415 	/*
416 	 * If the interface already has a link name, release it
417 	 * now.  This is useful for interfaces that can change
418 	 * link types, and thus switch link names often.
419 	 */
420 	if (ifp->if_sadl != NULL)
421 		if_free_sadl(ifp);
422 
423 	ifa = if_dl_create(ifp, &sdl);
424 
425 	ifa_insert(ifp, ifa);
426 	if_sadl_setrefs(ifp, ifa);
427 }
428 
429 static void
430 if_deactivate_sadl(struct ifnet *ifp)
431 {
432 	struct ifaddr *ifa;
433 
434 	KASSERT(ifp->if_dl != NULL);
435 
436 	ifa = ifp->if_dl;
437 
438 	ifp->if_sadl = NULL;
439 
440 	ifnet_addrs[ifp->if_index] = NULL;
441 	IFAFREE(ifa);
442 	ifp->if_dl = NULL;
443 	IFAFREE(ifa);
444 }
445 
446 void
447 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
448     const struct sockaddr_dl *sdl)
449 {
450 	int s;
451 
452 	s = splnet();
453 
454 	if_deactivate_sadl(ifp);
455 
456 	if_sadl_setrefs(ifp, ifa);
457 	IFADDR_FOREACH(ifa, ifp)
458 		rtinit(ifa, RTM_LLINFO_UPD, 0);
459 	splx(s);
460 }
461 
462 /*
463  * Free the link level name for the specified interface.  This is
464  * a detach helper.  This is called from if_detach() or from
465  * link layer type specific detach functions.
466  */
467 void
468 if_free_sadl(struct ifnet *ifp)
469 {
470 	struct ifaddr *ifa;
471 	int s;
472 
473 	ifa = ifnet_addrs[ifp->if_index];
474 	if (ifa == NULL) {
475 		KASSERT(ifp->if_sadl == NULL);
476 		KASSERT(ifp->if_dl == NULL);
477 		return;
478 	}
479 
480 	KASSERT(ifp->if_sadl != NULL);
481 	KASSERT(ifp->if_dl != NULL);
482 
483 	s = splnet();
484 	rtinit(ifa, RTM_DELETE, 0);
485 	ifa_remove(ifp, ifa);
486 	if_deactivate_sadl(ifp);
487 	if (ifp->if_hwdl == ifa) {
488 		IFAFREE(ifa);
489 		ifp->if_hwdl = NULL;
490 	}
491 	splx(s);
492 }
493 
494 static void
495 if_getindex(ifnet_t *ifp)
496 {
497 	bool hitlimit = false;
498 
499 	mutex_enter(&index_gen_mtx);
500 	ifp->if_index_gen = index_gen++;
501 	mutex_exit(&index_gen_mtx);
502 
503 	ifp->if_index = if_index;
504 	if (ifindex2ifnet == NULL) {
505 		if_index++;
506 		goto skip;
507 	}
508 	while (if_byindex(ifp->if_index)) {
509 		/*
510 		 * If we hit USHRT_MAX, we skip back to 0 since
511 		 * there are a number of places where the value
512 		 * of if_index or if_index itself is compared
513 		 * to or stored in an unsigned short.  By
514 		 * jumping back, we won't botch those assignments
515 		 * or comparisons.
516 		 */
517 		if (++if_index == 0) {
518 			if_index = 1;
519 		} else if (if_index == USHRT_MAX) {
520 			/*
521 			 * However, if we have to jump back to
522 			 * zero *twice* without finding an empty
523 			 * slot in ifindex2ifnet[], then there
524 			 * there are too many (>65535) interfaces.
525 			 */
526 			if (hitlimit) {
527 				panic("too many interfaces");
528 			}
529 			hitlimit = true;
530 			if_index = 1;
531 		}
532 		ifp->if_index = if_index;
533 	}
534 skip:
535 	/*
536 	 * We have some arrays that should be indexed by if_index.
537 	 * since if_index will grow dynamically, they should grow too.
538 	 *	struct ifadd **ifnet_addrs
539 	 *	struct ifnet **ifindex2ifnet
540 	 */
541 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
542 	    ifp->if_index >= if_indexlim) {
543 		size_t m, n, oldlim;
544 		void *q;
545 
546 		oldlim = if_indexlim;
547 		while (ifp->if_index >= if_indexlim)
548 			if_indexlim <<= 1;
549 
550 		/* grow ifnet_addrs */
551 		m = oldlim * sizeof(struct ifaddr *);
552 		n = if_indexlim * sizeof(struct ifaddr *);
553 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
554 		if (ifnet_addrs != NULL) {
555 			memcpy(q, ifnet_addrs, m);
556 			free(ifnet_addrs, M_IFADDR);
557 		}
558 		ifnet_addrs = (struct ifaddr **)q;
559 
560 		/* grow ifindex2ifnet */
561 		m = oldlim * sizeof(struct ifnet *);
562 		n = if_indexlim * sizeof(struct ifnet *);
563 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
564 		if (ifindex2ifnet != NULL) {
565 			memcpy(q, ifindex2ifnet, m);
566 			free(ifindex2ifnet, M_IFADDR);
567 		}
568 		ifindex2ifnet = (struct ifnet **)q;
569 	}
570 	ifindex2ifnet[ifp->if_index] = ifp;
571 }
572 
573 /*
574  * Attach an interface to the list of "active" interfaces.
575  */
576 void
577 if_attach(ifnet_t *ifp)
578 {
579 	KASSERT(if_indexlim > 0);
580 	TAILQ_INIT(&ifp->if_addrlist);
581 	TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list);
582 
583 	if (ifioctl_attach(ifp) != 0)
584 		panic("%s: ifioctl_attach() failed", __func__);
585 
586 	if_getindex(ifp);
587 
588 	/*
589 	 * Link level name is allocated later by a separate call to
590 	 * if_alloc_sadl().
591 	 */
592 
593 	if (ifp->if_snd.ifq_maxlen == 0)
594 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
595 
596 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
597 
598 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
599 
600 	ifp->if_link_state = LINK_STATE_UNKNOWN;
601 
602 	ifp->if_capenable = 0;
603 	ifp->if_csum_flags_tx = 0;
604 	ifp->if_csum_flags_rx = 0;
605 
606 #ifdef ALTQ
607 	ifp->if_snd.altq_type = 0;
608 	ifp->if_snd.altq_disc = NULL;
609 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
610 	ifp->if_snd.altq_tbr  = NULL;
611 	ifp->if_snd.altq_ifp  = ifp;
612 #endif
613 
614 #ifdef NET_MPSAFE
615 	ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
616 #else
617 	ifp->if_snd.ifq_lock = NULL;
618 #endif
619 
620 	ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
621 	(void)pfil_run_hooks(if_pfil,
622 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
623 
624 	if (!STAILQ_EMPTY(&domains))
625 		if_attachdomain1(ifp);
626 
627 	/* Announce the interface. */
628 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
629 }
630 
631 void
632 if_attachdomain(void)
633 {
634 	struct ifnet *ifp;
635 	int s;
636 
637 	s = splnet();
638 	IFNET_FOREACH(ifp)
639 		if_attachdomain1(ifp);
640 	splx(s);
641 }
642 
643 void
644 if_attachdomain1(struct ifnet *ifp)
645 {
646 	struct domain *dp;
647 	int s;
648 
649 	s = splnet();
650 
651 	/* address family dependent data region */
652 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
653 	DOMAIN_FOREACH(dp) {
654 		if (dp->dom_ifattach != NULL)
655 			ifp->if_afdata[dp->dom_family] =
656 			    (*dp->dom_ifattach)(ifp);
657 	}
658 
659 	splx(s);
660 }
661 
662 /*
663  * Deactivate an interface.  This points all of the procedure
664  * handles at error stubs.  May be called from interrupt context.
665  */
666 void
667 if_deactivate(struct ifnet *ifp)
668 {
669 	int s;
670 
671 	s = splnet();
672 
673 	ifp->if_output	 = if_nulloutput;
674 	ifp->if_input	 = if_nullinput;
675 	ifp->if_start	 = if_nullstart;
676 	ifp->if_ioctl	 = if_nullioctl;
677 	ifp->if_init	 = if_nullinit;
678 	ifp->if_stop	 = if_nullstop;
679 	ifp->if_watchdog = if_nullwatchdog;
680 	ifp->if_drain	 = if_nulldrain;
681 
682 	/* No more packets may be enqueued. */
683 	ifp->if_snd.ifq_maxlen = 0;
684 
685 	splx(s);
686 }
687 
688 void
689 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
690 {
691 	struct ifaddr *ifa, *nifa;
692 
693 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
694 		nifa = IFADDR_NEXT(ifa);
695 		if (ifa->ifa_addr->sa_family != family)
696 			continue;
697 		(*purgeaddr)(ifa);
698 	}
699 }
700 
701 /*
702  * Detach an interface from the list of "active" interfaces,
703  * freeing any resources as we go along.
704  *
705  * NOTE: This routine must be called with a valid thread context,
706  * as it may block.
707  */
708 void
709 if_detach(struct ifnet *ifp)
710 {
711 	struct socket so;
712 	struct ifaddr *ifa;
713 #ifdef IFAREF_DEBUG
714 	struct ifaddr *last_ifa = NULL;
715 #endif
716 	struct domain *dp;
717 	const struct protosw *pr;
718 	int s, i, family, purged;
719 	uint64_t xc;
720 
721 	/*
722 	 * XXX It's kind of lame that we have to have the
723 	 * XXX socket structure...
724 	 */
725 	memset(&so, 0, sizeof(so));
726 
727 	s = splnet();
728 
729 	/*
730 	 * Do an if_down() to give protocols a chance to do something.
731 	 */
732 	if_down(ifp);
733 
734 #ifdef ALTQ
735 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
736 		altq_disable(&ifp->if_snd);
737 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
738 		altq_detach(&ifp->if_snd);
739 #endif
740 
741 	if (ifp->if_snd.ifq_lock)
742 		mutex_obj_free(ifp->if_snd.ifq_lock);
743 
744 	sysctl_teardown(&ifp->if_sysctl_log);
745 
746 #if NCARP > 0
747 	/* Remove the interface from any carp group it is a part of.  */
748 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
749 		carp_ifdetach(ifp);
750 #endif
751 
752 	/*
753 	 * Rip all the addresses off the interface.  This should make
754 	 * all of the routes go away.
755 	 *
756 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
757 	 * from the list, including our "cursor", ifa.  For safety,
758 	 * and to honor the TAILQ abstraction, I just restart the
759 	 * loop after each removal.  Note that the loop will exit
760 	 * when all of the remaining ifaddrs belong to the AF_LINK
761 	 * family.  I am counting on the historical fact that at
762 	 * least one pr_usrreq in each address domain removes at
763 	 * least one ifaddr.
764 	 */
765 again:
766 	IFADDR_FOREACH(ifa, ifp) {
767 		family = ifa->ifa_addr->sa_family;
768 #ifdef IFAREF_DEBUG
769 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
770 		    ifa, family, ifa->ifa_refcnt);
771 		if (last_ifa != NULL && ifa == last_ifa)
772 			panic("if_detach: loop detected");
773 		last_ifa = ifa;
774 #endif
775 		if (family == AF_LINK)
776 			continue;
777 		dp = pffinddomain(family);
778 #ifdef DIAGNOSTIC
779 		if (dp == NULL)
780 			panic("if_detach: no domain for AF %d",
781 			    family);
782 #endif
783 		/*
784 		 * XXX These PURGEIF calls are redundant with the
785 		 * purge-all-families calls below, but are left in for
786 		 * now both to make a smaller change, and to avoid
787 		 * unplanned interactions with clearing of
788 		 * ifp->if_addrlist.
789 		 */
790 		purged = 0;
791 		for (pr = dp->dom_protosw;
792 		     pr < dp->dom_protoswNPROTOSW; pr++) {
793 			so.so_proto = pr;
794 			if (pr->pr_usrreqs) {
795 				(void) (*pr->pr_usrreqs->pr_generic)(&so,
796 				    PRU_PURGEIF, NULL, NULL,
797 				    (struct mbuf *) ifp, curlwp);
798 				purged = 1;
799 			}
800 		}
801 		if (purged == 0) {
802 			/*
803 			 * XXX What's really the best thing to do
804 			 * XXX here?  --thorpej@NetBSD.org
805 			 */
806 			printf("if_detach: WARNING: AF %d not purged\n",
807 			    family);
808 			ifa_remove(ifp, ifa);
809 		}
810 		goto again;
811 	}
812 
813 	if_free_sadl(ifp);
814 
815 	/* Walk the routing table looking for stragglers. */
816 	for (i = 0; i <= AF_MAX; i++) {
817 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
818 			continue;
819 	}
820 
821 	DOMAIN_FOREACH(dp) {
822 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
823 		{
824 			void *p = ifp->if_afdata[dp->dom_family];
825 			if (p) {
826 				ifp->if_afdata[dp->dom_family] = NULL;
827 				(*dp->dom_ifdetach)(ifp, p);
828 			}
829 		}
830 
831 		/*
832 		 * One would expect multicast memberships (INET and
833 		 * INET6) on UDP sockets to be purged by the PURGEIF
834 		 * calls above, but if all addresses were removed from
835 		 * the interface prior to destruction, the calls will
836 		 * not be made (e.g. ppp, for which pppd(8) generally
837 		 * removes addresses before destroying the interface).
838 		 * Because there is no invariant that multicast
839 		 * memberships only exist for interfaces with IPv4
840 		 * addresses, we must call PURGEIF regardless of
841 		 * addresses.  (Protocols which might store ifnet
842 		 * pointers are marked with PR_PURGEIF.)
843 		 */
844 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
845 			so.so_proto = pr;
846 			if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF)
847 				(void)(*pr->pr_usrreqs->pr_generic)(&so,
848 				    PRU_PURGEIF, NULL, NULL,
849 				    (struct mbuf *)ifp, curlwp);
850 		}
851 	}
852 
853 	(void)pfil_run_hooks(if_pfil,
854 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
855 	(void)pfil_head_destroy(ifp->if_pfil);
856 
857 	/* Announce that the interface is gone. */
858 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
859 
860 	ifindex2ifnet[ifp->if_index] = NULL;
861 
862 	TAILQ_REMOVE(&ifnet_list, ifp, if_list);
863 
864 	ifioctl_detach(ifp);
865 
866 	/*
867 	 * remove packets that came from ifp, from software interrupt queues.
868 	 */
869 	DOMAIN_FOREACH(dp) {
870 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
871 			struct ifqueue *iq = dp->dom_ifqueues[i];
872 			if (iq == NULL)
873 				break;
874 			dp->dom_ifqueues[i] = NULL;
875 			if_detach_queues(ifp, iq);
876 		}
877 	}
878 
879 	/*
880 	 * IP queues have to be processed separately: net-queue barrier
881 	 * ensures that the packets are dequeued while a cross-call will
882 	 * ensure that the interrupts have completed. FIXME: not quite..
883 	 */
884 #ifdef INET
885 	pktq_barrier(ip_pktq);
886 #endif
887 #ifdef INET6
888 	pktq_barrier(ip6_pktq);
889 #endif
890 	xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
891 	xc_wait(xc);
892 
893 	splx(s);
894 }
895 
896 static void
897 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
898 {
899 	struct mbuf *m, *prev, *next;
900 
901 	prev = NULL;
902 	for (m = q->ifq_head; m != NULL; m = next) {
903 		KASSERT((m->m_flags & M_PKTHDR) != 0);
904 
905 		next = m->m_nextpkt;
906 		if (m->m_pkthdr.rcvif != ifp) {
907 			prev = m;
908 			continue;
909 		}
910 
911 		if (prev != NULL)
912 			prev->m_nextpkt = m->m_nextpkt;
913 		else
914 			q->ifq_head = m->m_nextpkt;
915 		if (q->ifq_tail == m)
916 			q->ifq_tail = prev;
917 		q->ifq_len--;
918 
919 		m->m_nextpkt = NULL;
920 		m_freem(m);
921 		IF_DROP(q);
922 	}
923 }
924 
925 /*
926  * Callback for a radix tree walk to delete all references to an
927  * ifnet.
928  */
929 static int
930 if_rt_walktree(struct rtentry *rt, void *v)
931 {
932 	struct ifnet *ifp = (struct ifnet *)v;
933 	int error;
934 
935 	if (rt->rt_ifp != ifp)
936 		return 0;
937 
938 	/* Delete the entry. */
939 	++rt->rt_refcnt;
940 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
941 	    rt_mask(rt), rt->rt_flags, NULL);
942 	KASSERT((rt->rt_flags & RTF_UP) == 0);
943 	rt->rt_ifp = NULL;
944 	rtfree(rt);
945 	if (error != 0)
946 		printf("%s: warning: unable to delete rtentry @ %p, "
947 		    "error = %d\n", ifp->if_xname, rt, error);
948 	return ERESTART;
949 }
950 
951 /*
952  * Create a clone network interface.
953  */
954 int
955 if_clone_create(const char *name)
956 {
957 	struct if_clone *ifc;
958 	int unit;
959 
960 	ifc = if_clone_lookup(name, &unit);
961 	if (ifc == NULL)
962 		return EINVAL;
963 
964 	if (ifunit(name) != NULL)
965 		return EEXIST;
966 
967 	return (*ifc->ifc_create)(ifc, unit);
968 }
969 
970 /*
971  * Destroy a clone network interface.
972  */
973 int
974 if_clone_destroy(const char *name)
975 {
976 	struct if_clone *ifc;
977 	struct ifnet *ifp;
978 
979 	ifc = if_clone_lookup(name, NULL);
980 	if (ifc == NULL)
981 		return EINVAL;
982 
983 	ifp = ifunit(name);
984 	if (ifp == NULL)
985 		return ENXIO;
986 
987 	if (ifc->ifc_destroy == NULL)
988 		return EOPNOTSUPP;
989 
990 	return (*ifc->ifc_destroy)(ifp);
991 }
992 
993 /*
994  * Look up a network interface cloner.
995  */
996 static struct if_clone *
997 if_clone_lookup(const char *name, int *unitp)
998 {
999 	struct if_clone *ifc;
1000 	const char *cp;
1001 	char *dp, ifname[IFNAMSIZ + 3];
1002 	int unit;
1003 
1004 	strcpy(ifname, "if_");
1005 	/* separate interface name from unit */
1006 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
1007 	    *cp && (*cp < '0' || *cp > '9');)
1008 		*dp++ = *cp++;
1009 
1010 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
1011 		return NULL;	/* No name or unit number */
1012 	*dp++ = '\0';
1013 
1014 again:
1015 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
1016 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
1017 			break;
1018 	}
1019 
1020 	if (ifc == NULL) {
1021 		if (*ifname == '\0' ||
1022 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
1023 			return NULL;
1024 		*ifname = '\0';
1025 		goto again;
1026 	}
1027 
1028 	unit = 0;
1029 	while (cp - name < IFNAMSIZ && *cp) {
1030 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1031 			/* Bogus unit number. */
1032 			return NULL;
1033 		}
1034 		unit = (unit * 10) + (*cp++ - '0');
1035 	}
1036 
1037 	if (unitp != NULL)
1038 		*unitp = unit;
1039 	return ifc;
1040 }
1041 
1042 /*
1043  * Register a network interface cloner.
1044  */
1045 void
1046 if_clone_attach(struct if_clone *ifc)
1047 {
1048 
1049 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1050 	if_cloners_count++;
1051 }
1052 
1053 /*
1054  * Unregister a network interface cloner.
1055  */
1056 void
1057 if_clone_detach(struct if_clone *ifc)
1058 {
1059 
1060 	LIST_REMOVE(ifc, ifc_list);
1061 	if_cloners_count--;
1062 }
1063 
1064 /*
1065  * Provide list of interface cloners to userspace.
1066  */
1067 static int
1068 if_clone_list(struct if_clonereq *ifcr)
1069 {
1070 	char outbuf[IFNAMSIZ], *dst;
1071 	struct if_clone *ifc;
1072 	int count, error = 0;
1073 
1074 	ifcr->ifcr_total = if_cloners_count;
1075 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1076 		/* Just asking how many there are. */
1077 		return 0;
1078 	}
1079 
1080 	if (ifcr->ifcr_count < 0)
1081 		return EINVAL;
1082 
1083 	count = (if_cloners_count < ifcr->ifcr_count) ?
1084 	    if_cloners_count : ifcr->ifcr_count;
1085 
1086 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1087 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1088 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1089 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1090 			return ENAMETOOLONG;
1091 		error = copyout(outbuf, dst, sizeof(outbuf));
1092 		if (error != 0)
1093 			break;
1094 	}
1095 
1096 	return error;
1097 }
1098 
1099 void
1100 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1101 {
1102 	ifa->ifa_ifp = ifp;
1103 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1104 	IFAREF(ifa);
1105 }
1106 
1107 void
1108 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1109 {
1110 	KASSERT(ifa->ifa_ifp == ifp);
1111 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1112 	IFAFREE(ifa);
1113 }
1114 
1115 static inline int
1116 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1117 {
1118 	return sockaddr_cmp(sa1, sa2) == 0;
1119 }
1120 
1121 /*
1122  * Locate an interface based on a complete address.
1123  */
1124 /*ARGSUSED*/
1125 struct ifaddr *
1126 ifa_ifwithaddr(const struct sockaddr *addr)
1127 {
1128 	struct ifnet *ifp;
1129 	struct ifaddr *ifa;
1130 
1131 	IFNET_FOREACH(ifp) {
1132 		if (ifp->if_output == if_nulloutput)
1133 			continue;
1134 		IFADDR_FOREACH(ifa, ifp) {
1135 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1136 				continue;
1137 			if (equal(addr, ifa->ifa_addr))
1138 				return ifa;
1139 			if ((ifp->if_flags & IFF_BROADCAST) &&
1140 			    ifa->ifa_broadaddr &&
1141 			    /* IP6 doesn't have broadcast */
1142 			    ifa->ifa_broadaddr->sa_len != 0 &&
1143 			    equal(ifa->ifa_broadaddr, addr))
1144 				return ifa;
1145 		}
1146 	}
1147 	return NULL;
1148 }
1149 
1150 /*
1151  * Locate the point to point interface with a given destination address.
1152  */
1153 /*ARGSUSED*/
1154 struct ifaddr *
1155 ifa_ifwithdstaddr(const struct sockaddr *addr)
1156 {
1157 	struct ifnet *ifp;
1158 	struct ifaddr *ifa;
1159 
1160 	IFNET_FOREACH(ifp) {
1161 		if (ifp->if_output == if_nulloutput)
1162 			continue;
1163 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1164 			continue;
1165 		IFADDR_FOREACH(ifa, ifp) {
1166 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1167 			    ifa->ifa_dstaddr == NULL)
1168 				continue;
1169 			if (equal(addr, ifa->ifa_dstaddr))
1170 				return ifa;
1171 		}
1172 	}
1173 	return NULL;
1174 }
1175 
1176 /*
1177  * Find an interface on a specific network.  If many, choice
1178  * is most specific found.
1179  */
1180 struct ifaddr *
1181 ifa_ifwithnet(const struct sockaddr *addr)
1182 {
1183 	struct ifnet *ifp;
1184 	struct ifaddr *ifa;
1185 	const struct sockaddr_dl *sdl;
1186 	struct ifaddr *ifa_maybe = 0;
1187 	u_int af = addr->sa_family;
1188 	const char *addr_data = addr->sa_data, *cplim;
1189 
1190 	if (af == AF_LINK) {
1191 		sdl = satocsdl(addr);
1192 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1193 		    ifindex2ifnet[sdl->sdl_index] &&
1194 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1195 			return ifnet_addrs[sdl->sdl_index];
1196 	}
1197 #ifdef NETATALK
1198 	if (af == AF_APPLETALK) {
1199 		const struct sockaddr_at *sat, *sat2;
1200 		sat = (const struct sockaddr_at *)addr;
1201 		IFNET_FOREACH(ifp) {
1202 			if (ifp->if_output == if_nulloutput)
1203 				continue;
1204 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1205 			if (ifa == NULL)
1206 				continue;
1207 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1208 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1209 				return ifa; /* exact match */
1210 			if (ifa_maybe == NULL) {
1211 				/* else keep the if with the right range */
1212 				ifa_maybe = ifa;
1213 			}
1214 		}
1215 		return ifa_maybe;
1216 	}
1217 #endif
1218 	IFNET_FOREACH(ifp) {
1219 		if (ifp->if_output == if_nulloutput)
1220 			continue;
1221 		IFADDR_FOREACH(ifa, ifp) {
1222 			const char *cp, *cp2, *cp3;
1223 
1224 			if (ifa->ifa_addr->sa_family != af ||
1225 			    ifa->ifa_netmask == NULL)
1226  next:				continue;
1227 			cp = addr_data;
1228 			cp2 = ifa->ifa_addr->sa_data;
1229 			cp3 = ifa->ifa_netmask->sa_data;
1230 			cplim = (const char *)ifa->ifa_netmask +
1231 			    ifa->ifa_netmask->sa_len;
1232 			while (cp3 < cplim) {
1233 				if ((*cp++ ^ *cp2++) & *cp3++) {
1234 					/* want to continue for() loop */
1235 					goto next;
1236 				}
1237 			}
1238 			if (ifa_maybe == NULL ||
1239 			    rn_refines((void *)ifa->ifa_netmask,
1240 			    (void *)ifa_maybe->ifa_netmask))
1241 				ifa_maybe = ifa;
1242 		}
1243 	}
1244 	return ifa_maybe;
1245 }
1246 
1247 /*
1248  * Find the interface of the addresss.
1249  */
1250 struct ifaddr *
1251 ifa_ifwithladdr(const struct sockaddr *addr)
1252 {
1253 	struct ifaddr *ia;
1254 
1255 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1256 	    (ia = ifa_ifwithnet(addr)))
1257 		return ia;
1258 	return NULL;
1259 }
1260 
1261 /*
1262  * Find an interface using a specific address family
1263  */
1264 struct ifaddr *
1265 ifa_ifwithaf(int af)
1266 {
1267 	struct ifnet *ifp;
1268 	struct ifaddr *ifa;
1269 
1270 	IFNET_FOREACH(ifp) {
1271 		if (ifp->if_output == if_nulloutput)
1272 			continue;
1273 		IFADDR_FOREACH(ifa, ifp) {
1274 			if (ifa->ifa_addr->sa_family == af)
1275 				return ifa;
1276 		}
1277 	}
1278 	return NULL;
1279 }
1280 
1281 /*
1282  * Find an interface address specific to an interface best matching
1283  * a given address.
1284  */
1285 struct ifaddr *
1286 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1287 {
1288 	struct ifaddr *ifa;
1289 	const char *cp, *cp2, *cp3;
1290 	const char *cplim;
1291 	struct ifaddr *ifa_maybe = 0;
1292 	u_int af = addr->sa_family;
1293 
1294 	if (ifp->if_output == if_nulloutput)
1295 		return NULL;
1296 
1297 	if (af >= AF_MAX)
1298 		return NULL;
1299 
1300 	IFADDR_FOREACH(ifa, ifp) {
1301 		if (ifa->ifa_addr->sa_family != af)
1302 			continue;
1303 		ifa_maybe = ifa;
1304 		if (ifa->ifa_netmask == NULL) {
1305 			if (equal(addr, ifa->ifa_addr) ||
1306 			    (ifa->ifa_dstaddr &&
1307 			     equal(addr, ifa->ifa_dstaddr)))
1308 				return ifa;
1309 			continue;
1310 		}
1311 		cp = addr->sa_data;
1312 		cp2 = ifa->ifa_addr->sa_data;
1313 		cp3 = ifa->ifa_netmask->sa_data;
1314 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1315 		for (; cp3 < cplim; cp3++) {
1316 			if ((*cp++ ^ *cp2++) & *cp3)
1317 				break;
1318 		}
1319 		if (cp3 == cplim)
1320 			return ifa;
1321 	}
1322 	return ifa_maybe;
1323 }
1324 
1325 /*
1326  * Default action when installing a route with a Link Level gateway.
1327  * Lookup an appropriate real ifa to point to.
1328  * This should be moved to /sys/net/link.c eventually.
1329  */
1330 void
1331 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1332 {
1333 	struct ifaddr *ifa;
1334 	const struct sockaddr *dst;
1335 	struct ifnet *ifp;
1336 
1337 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1338 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1339 		return;
1340 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1341 		rt_replace_ifa(rt, ifa);
1342 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1343 			ifa->ifa_rtrequest(cmd, rt, info);
1344 	}
1345 }
1346 
1347 /*
1348  * Handle a change in the interface link state.
1349  * XXX: We should listen to the routing socket in-kernel rather
1350  * than calling in6_if_link_* functions directly from here.
1351  */
1352 void
1353 if_link_state_change(struct ifnet *ifp, int link_state)
1354 {
1355 	int s;
1356 #if defined(DEBUG) || defined(INET6)
1357 	int old_link_state;
1358 #endif
1359 
1360 	s = splnet();
1361 	if (ifp->if_link_state == link_state) {
1362 		splx(s);
1363 		return;
1364 	}
1365 
1366 #if defined(DEBUG) || defined(INET6)
1367 	old_link_state = ifp->if_link_state;
1368 #endif
1369 	ifp->if_link_state = link_state;
1370 #ifdef DEBUG
1371 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
1372 		link_state == LINK_STATE_UP ? "UP" :
1373 		link_state == LINK_STATE_DOWN ? "DOWN" :
1374 		"UNKNOWN",
1375 		 old_link_state == LINK_STATE_UP ? "UP" :
1376 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
1377 		"UNKNOWN");
1378 #endif
1379 
1380 #ifdef INET6
1381 	/*
1382 	 * When going from UNKNOWN to UP, we need to mark existing
1383 	 * IPv6 addresses as tentative and restart DAD as we may have
1384 	 * erroneously not found a duplicate.
1385 	 *
1386 	 * This needs to happen before rt_ifmsg to avoid a race where
1387 	 * listeners would have an address and expect it to work right
1388 	 * away.
1389 	 */
1390 	if (in6_present && link_state == LINK_STATE_UP &&
1391 	    old_link_state == LINK_STATE_UNKNOWN)
1392 		in6_if_link_down(ifp);
1393 #endif
1394 
1395 	/* Notify that the link state has changed. */
1396 	rt_ifmsg(ifp);
1397 
1398 #if NCARP > 0
1399 	if (ifp->if_carp)
1400 		carp_carpdev_state(ifp);
1401 #endif
1402 
1403 #ifdef INET6
1404 	if (in6_present) {
1405 		if (link_state == LINK_STATE_DOWN)
1406 			in6_if_link_down(ifp);
1407 		else if (link_state == LINK_STATE_UP)
1408 			in6_if_link_up(ifp);
1409 	}
1410 #endif
1411 
1412 	splx(s);
1413 }
1414 
1415 /*
1416  * Mark an interface down and notify protocols of
1417  * the transition.
1418  * NOTE: must be called at splsoftnet or equivalent.
1419  */
1420 void
1421 if_down(struct ifnet *ifp)
1422 {
1423 	struct ifaddr *ifa;
1424 
1425 	ifp->if_flags &= ~IFF_UP;
1426 	nanotime(&ifp->if_lastchange);
1427 	IFADDR_FOREACH(ifa, ifp)
1428 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1429 	IFQ_PURGE(&ifp->if_snd);
1430 #if NCARP > 0
1431 	if (ifp->if_carp)
1432 		carp_carpdev_state(ifp);
1433 #endif
1434 	rt_ifmsg(ifp);
1435 #ifdef INET6
1436 	if (in6_present)
1437 		in6_if_down(ifp);
1438 #endif
1439 }
1440 
1441 /*
1442  * Mark an interface up and notify protocols of
1443  * the transition.
1444  * NOTE: must be called at splsoftnet or equivalent.
1445  */
1446 void
1447 if_up(struct ifnet *ifp)
1448 {
1449 #ifdef notyet
1450 	struct ifaddr *ifa;
1451 #endif
1452 
1453 	ifp->if_flags |= IFF_UP;
1454 	nanotime(&ifp->if_lastchange);
1455 #ifdef notyet
1456 	/* this has no effect on IP, and will kill all ISO connections XXX */
1457 	IFADDR_FOREACH(ifa, ifp)
1458 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1459 #endif
1460 #if NCARP > 0
1461 	if (ifp->if_carp)
1462 		carp_carpdev_state(ifp);
1463 #endif
1464 	rt_ifmsg(ifp);
1465 #ifdef INET6
1466 	if (in6_present)
1467 		in6_if_up(ifp);
1468 #endif
1469 }
1470 
1471 /*
1472  * Handle interface watchdog timer routines.  Called
1473  * from softclock, we decrement timers (if set) and
1474  * call the appropriate interface routine on expiration.
1475  */
1476 void
1477 if_slowtimo(void *arg)
1478 {
1479 	struct ifnet *ifp;
1480 	int s = splnet();
1481 
1482 	IFNET_FOREACH(ifp) {
1483 		if (ifp->if_timer == 0 || --ifp->if_timer)
1484 			continue;
1485 		if (ifp->if_watchdog != NULL)
1486 			(*ifp->if_watchdog)(ifp);
1487 	}
1488 	splx(s);
1489 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1490 }
1491 
1492 /*
1493  * Set/clear promiscuous mode on interface ifp based on the truth value
1494  * of pswitch.  The calls are reference counted so that only the first
1495  * "on" request actually has an effect, as does the final "off" request.
1496  * Results are undefined if the "off" and "on" requests are not matched.
1497  */
1498 int
1499 ifpromisc(struct ifnet *ifp, int pswitch)
1500 {
1501 	int pcount, ret;
1502 	short nflags;
1503 
1504 	pcount = ifp->if_pcount;
1505 	if (pswitch) {
1506 		/*
1507 		 * Allow the device to be "placed" into promiscuous
1508 		 * mode even if it is not configured up.  It will
1509 		 * consult IFF_PROMISC when it is brought up.
1510 		 */
1511 		if (ifp->if_pcount++ != 0)
1512 			return 0;
1513 		nflags = ifp->if_flags | IFF_PROMISC;
1514 	} else {
1515 		if (--ifp->if_pcount > 0)
1516 			return 0;
1517 		nflags = ifp->if_flags & ~IFF_PROMISC;
1518 	}
1519 	ret = if_flags_set(ifp, nflags);
1520 	/* Restore interface state if not successful. */
1521 	if (ret != 0) {
1522 		ifp->if_pcount = pcount;
1523 	}
1524 	return ret;
1525 }
1526 
1527 /*
1528  * Map interface name to
1529  * interface structure pointer.
1530  */
1531 struct ifnet *
1532 ifunit(const char *name)
1533 {
1534 	struct ifnet *ifp;
1535 	const char *cp = name;
1536 	u_int unit = 0;
1537 	u_int i;
1538 
1539 	/*
1540 	 * If the entire name is a number, treat it as an ifindex.
1541 	 */
1542 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1543 		unit = unit * 10 + (*cp - '0');
1544 	}
1545 
1546 	/*
1547 	 * If the number took all of the name, then it's a valid ifindex.
1548 	 */
1549 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1550 		if (unit >= if_indexlim)
1551 			return NULL;
1552 		ifp = ifindex2ifnet[unit];
1553 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1554 			return NULL;
1555 		return ifp;
1556 	}
1557 
1558 	IFNET_FOREACH(ifp) {
1559 		if (ifp->if_output == if_nulloutput)
1560 			continue;
1561 	 	if (strcmp(ifp->if_xname, name) == 0)
1562 			return ifp;
1563 	}
1564 	return NULL;
1565 }
1566 
1567 ifnet_t *
1568 if_byindex(u_int idx)
1569 {
1570 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1571 }
1572 
1573 /* common */
1574 int
1575 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1576 {
1577 	int s;
1578 	struct ifreq *ifr;
1579 	struct ifcapreq *ifcr;
1580 	struct ifdatareq *ifdr;
1581 
1582 	switch (cmd) {
1583 	case SIOCSIFCAP:
1584 		ifcr = data;
1585 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1586 			return EINVAL;
1587 
1588 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1589 			return 0;
1590 
1591 		ifp->if_capenable = ifcr->ifcr_capenable;
1592 
1593 		/* Pre-compute the checksum flags mask. */
1594 		ifp->if_csum_flags_tx = 0;
1595 		ifp->if_csum_flags_rx = 0;
1596 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1597 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1598 		}
1599 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1600 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1601 		}
1602 
1603 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1604 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1605 		}
1606 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1607 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1608 		}
1609 
1610 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1611 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1612 		}
1613 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1614 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1615 		}
1616 
1617 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1618 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1619 		}
1620 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1621 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1622 		}
1623 
1624 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1625 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1626 		}
1627 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1628 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1629 		}
1630 		if (ifp->if_flags & IFF_UP)
1631 			return ENETRESET;
1632 		return 0;
1633 	case SIOCSIFFLAGS:
1634 		ifr = data;
1635 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1636 			s = splnet();
1637 			if_down(ifp);
1638 			splx(s);
1639 		}
1640 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1641 			s = splnet();
1642 			if_up(ifp);
1643 			splx(s);
1644 		}
1645 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1646 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1647 		break;
1648 	case SIOCGIFFLAGS:
1649 		ifr = data;
1650 		ifr->ifr_flags = ifp->if_flags;
1651 		break;
1652 
1653 	case SIOCGIFMETRIC:
1654 		ifr = data;
1655 		ifr->ifr_metric = ifp->if_metric;
1656 		break;
1657 
1658 	case SIOCGIFMTU:
1659 		ifr = data;
1660 		ifr->ifr_mtu = ifp->if_mtu;
1661 		break;
1662 
1663 	case SIOCGIFDLT:
1664 		ifr = data;
1665 		ifr->ifr_dlt = ifp->if_dlt;
1666 		break;
1667 
1668 	case SIOCGIFCAP:
1669 		ifcr = data;
1670 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1671 		ifcr->ifcr_capenable = ifp->if_capenable;
1672 		break;
1673 
1674 	case SIOCSIFMETRIC:
1675 		ifr = data;
1676 		ifp->if_metric = ifr->ifr_metric;
1677 		break;
1678 
1679 	case SIOCGIFDATA:
1680 		ifdr = data;
1681 		ifdr->ifdr_data = ifp->if_data;
1682 		break;
1683 
1684 	case SIOCGIFINDEX:
1685 		ifr = data;
1686 		ifr->ifr_index = ifp->if_index;
1687 		break;
1688 
1689 	case SIOCZIFDATA:
1690 		ifdr = data;
1691 		ifdr->ifdr_data = ifp->if_data;
1692 		/*
1693 		 * Assumes that the volatile counters that can be
1694 		 * zero'ed are at the end of if_data.
1695 		 */
1696 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1697 		    offsetof(struct if_data, ifi_ipackets));
1698 		/*
1699 		 * The memset() clears to the bottm of if_data. In the area,
1700 		 * if_lastchange is included. Please be careful if new entry
1701 		 * will be added into if_data or rewite this.
1702 		 *
1703 		 * And also, update if_lastchnage.
1704 		 */
1705 		getnanotime(&ifp->if_lastchange);
1706 		break;
1707 	case SIOCSIFMTU:
1708 		ifr = data;
1709 		if (ifp->if_mtu == ifr->ifr_mtu)
1710 			break;
1711 		ifp->if_mtu = ifr->ifr_mtu;
1712 		/*
1713 		 * If the link MTU changed, do network layer specific procedure.
1714 		 */
1715 #ifdef INET6
1716 		if (in6_present)
1717 			nd6_setmtu(ifp);
1718 #endif
1719 		return ENETRESET;
1720 	default:
1721 		return ENOTTY;
1722 	}
1723 	return 0;
1724 }
1725 
1726 int
1727 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
1728 {
1729 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1730 	struct ifaddr *ifa;
1731 	const struct sockaddr *any, *sa;
1732 	union {
1733 		struct sockaddr sa;
1734 		struct sockaddr_storage ss;
1735 	} u, v;
1736 
1737 	switch (cmd) {
1738 	case SIOCSIFADDRPREF:
1739 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
1740 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1741 		    NULL) != 0)
1742 			return EPERM;
1743 	case SIOCGIFADDRPREF:
1744 		break;
1745 	default:
1746 		return EOPNOTSUPP;
1747 	}
1748 
1749 	/* sanity checks */
1750 	if (data == NULL || ifp == NULL) {
1751 		panic("invalid argument to %s", __func__);
1752 		/*NOTREACHED*/
1753 	}
1754 
1755 	/* address must be specified on ADD and DELETE */
1756 	sa = sstocsa(&ifap->ifap_addr);
1757 	if (sa->sa_family != sofamily(so))
1758 		return EINVAL;
1759 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1760 		return EINVAL;
1761 
1762 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1763 
1764 	IFADDR_FOREACH(ifa, ifp) {
1765 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1766 			continue;
1767 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1768 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1769 			break;
1770 	}
1771 	if (ifa == NULL)
1772 		return EADDRNOTAVAIL;
1773 
1774 	switch (cmd) {
1775 	case SIOCSIFADDRPREF:
1776 		ifa->ifa_preference = ifap->ifap_preference;
1777 		return 0;
1778 	case SIOCGIFADDRPREF:
1779 		/* fill in the if_laddrreq structure */
1780 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1781 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1782 		ifap->ifap_preference = ifa->ifa_preference;
1783 		return 0;
1784 	default:
1785 		return EOPNOTSUPP;
1786 	}
1787 }
1788 
1789 static void
1790 ifnet_lock_enter(struct ifnet_lock *il)
1791 {
1792 	uint64_t *nenter;
1793 
1794 	/* Before trying to acquire the mutex, increase the count of threads
1795 	 * who have entered or who wait to enter the critical section.
1796 	 * Avoid one costly locked memory transaction by keeping a count for
1797 	 * each CPU.
1798 	 */
1799 	nenter = percpu_getref(il->il_nenter);
1800 	(*nenter)++;
1801 	percpu_putref(il->il_nenter);
1802 	mutex_enter(&il->il_lock);
1803 }
1804 
1805 static void
1806 ifnet_lock_exit(struct ifnet_lock *il)
1807 {
1808 	/* Increase the count of threads who have exited the critical
1809 	 * section.  Increase while we still hold the lock.
1810 	 */
1811 	il->il_nexit++;
1812 	mutex_exit(&il->il_lock);
1813 }
1814 
1815 /*
1816  * Interface ioctls.
1817  */
1818 static int
1819 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1820 {
1821 	struct ifnet *ifp;
1822 	struct ifreq *ifr;
1823 	int error = 0;
1824 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1825 	u_long ocmd = cmd;
1826 #endif
1827 	short oif_flags;
1828 #ifdef COMPAT_OIFREQ
1829 	struct ifreq ifrb;
1830 	struct oifreq *oifr = NULL;
1831 #endif
1832 
1833 	switch (cmd) {
1834 #ifdef COMPAT_OIFREQ
1835 	case OSIOCGIFCONF:
1836 	case OOSIOCGIFCONF:
1837 		return compat_ifconf(cmd, data);
1838 #endif
1839 #ifdef COMPAT_OIFDATA
1840 	case OSIOCGIFDATA:
1841 	case OSIOCZIFDATA:
1842 		return compat_ifdatareq(l, cmd, data);
1843 #endif
1844 	case SIOCGIFCONF:
1845 		return ifconf(cmd, data);
1846 	case SIOCINITIFADDR:
1847 		return EPERM;
1848 	}
1849 
1850 #ifdef COMPAT_OIFREQ
1851 	cmd = compat_cvtcmd(cmd);
1852 	if (cmd != ocmd) {
1853 		oifr = data;
1854 		data = ifr = &ifrb;
1855 		ifreqo2n(oifr, ifr);
1856 	} else
1857 #endif
1858 		ifr = data;
1859 
1860 	ifp = ifunit(ifr->ifr_name);
1861 
1862 	switch (cmd) {
1863 	case SIOCIFCREATE:
1864 	case SIOCIFDESTROY:
1865 		if (l != NULL) {
1866 			error = kauth_authorize_network(l->l_cred,
1867 			    KAUTH_NETWORK_INTERFACE,
1868 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1869 			    (void *)cmd, NULL);
1870 			if (error != 0)
1871 				return error;
1872 		}
1873 		return (cmd == SIOCIFCREATE) ?
1874 			if_clone_create(ifr->ifr_name) :
1875 			if_clone_destroy(ifr->ifr_name);
1876 
1877 	case SIOCIFGCLONERS:
1878 		return if_clone_list((struct if_clonereq *)data);
1879 	}
1880 
1881 	if (ifp == NULL)
1882 		return ENXIO;
1883 
1884 	switch (cmd) {
1885 	case SIOCALIFADDR:
1886 	case SIOCDLIFADDR:
1887 	case SIOCSIFADDRPREF:
1888 	case SIOCSIFFLAGS:
1889 	case SIOCSIFCAP:
1890 	case SIOCSIFMETRIC:
1891 	case SIOCZIFDATA:
1892 	case SIOCSIFMTU:
1893 	case SIOCSIFPHYADDR:
1894 	case SIOCDIFPHYADDR:
1895 #ifdef INET6
1896 	case SIOCSIFPHYADDR_IN6:
1897 #endif
1898 	case SIOCSLIFPHYADDR:
1899 	case SIOCADDMULTI:
1900 	case SIOCDELMULTI:
1901 	case SIOCSIFMEDIA:
1902 	case SIOCSDRVSPEC:
1903 	case SIOCG80211:
1904 	case SIOCS80211:
1905 	case SIOCS80211NWID:
1906 	case SIOCS80211NWKEY:
1907 	case SIOCS80211POWER:
1908 	case SIOCS80211BSSID:
1909 	case SIOCS80211CHANNEL:
1910 	case SIOCSLINKSTR:
1911 		if (l != NULL) {
1912 			error = kauth_authorize_network(l->l_cred,
1913 			    KAUTH_NETWORK_INTERFACE,
1914 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1915 			    (void *)cmd, NULL);
1916 			if (error != 0)
1917 				return error;
1918 		}
1919 	}
1920 
1921 	oif_flags = ifp->if_flags;
1922 
1923 	ifnet_lock_enter(ifp->if_ioctl_lock);
1924 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1925 	if (error != ENOTTY)
1926 		;
1927 	else if (so->so_proto == NULL)
1928 		error = EOPNOTSUPP;
1929 	else {
1930 #ifdef COMPAT_OSOCK
1931 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1932 #else
1933 		error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so,
1934 		    cmd, data, ifp);
1935 #endif
1936 	}
1937 
1938 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1939 #ifdef INET6
1940 		if (in6_present && (ifp->if_flags & IFF_UP) != 0) {
1941 			int s = splnet();
1942 			in6_if_up(ifp);
1943 			splx(s);
1944 		}
1945 #endif
1946 	}
1947 #ifdef COMPAT_OIFREQ
1948 	if (cmd != ocmd)
1949 		ifreqn2o(oifr, ifr);
1950 #endif
1951 
1952 	ifnet_lock_exit(ifp->if_ioctl_lock);
1953 	return error;
1954 }
1955 
1956 /* This callback adds to the sum in `arg' the number of
1957  * threads on `ci' who have entered or who wait to enter the
1958  * critical section.
1959  */
1960 static void
1961 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
1962 {
1963 	uint64_t *sum = arg, *nenter = p;
1964 
1965 	*sum += *nenter;
1966 }
1967 
1968 /* Return the number of threads who have entered or who wait
1969  * to enter the critical section on all CPUs.
1970  */
1971 static uint64_t
1972 ifnet_lock_entrances(struct ifnet_lock *il)
1973 {
1974 	uint64_t sum = 0;
1975 
1976 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
1977 
1978 	return sum;
1979 }
1980 
1981 static int
1982 ifioctl_attach(struct ifnet *ifp)
1983 {
1984 	struct ifnet_lock *il;
1985 
1986 	/* If the driver has not supplied its own if_ioctl, then
1987 	 * supply the default.
1988 	 */
1989 	if (ifp->if_ioctl == NULL)
1990 		ifp->if_ioctl = ifioctl_common;
1991 
1992 	/* Create an ifnet_lock for synchronizing ifioctls. */
1993 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
1994 		return ENOMEM;
1995 
1996 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
1997 	if (il->il_nenter == NULL) {
1998 		kmem_free(il, sizeof(*il));
1999 		return ENOMEM;
2000 	}
2001 
2002 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
2003 	cv_init(&il->il_emptied, ifp->if_xname);
2004 
2005 	ifp->if_ioctl_lock = il;
2006 
2007 	return 0;
2008 }
2009 
2010 /*
2011  * This must not be called until after `ifp' has been withdrawn from the
2012  * ifnet tables so that ifioctl() cannot get a handle on it by calling
2013  * ifunit().
2014  */
2015 static void
2016 ifioctl_detach(struct ifnet *ifp)
2017 {
2018 	struct ifnet_lock *il;
2019 
2020 	il = ifp->if_ioctl_lock;
2021 	mutex_enter(&il->il_lock);
2022 	/* Install if_nullioctl to make sure that any thread that
2023 	 * subsequently enters the critical section will quit it
2024 	 * immediately and signal the condition variable that we
2025 	 * wait on, below.
2026 	 */
2027 	ifp->if_ioctl = if_nullioctl;
2028 	/* Sleep while threads are still in the critical section or
2029 	 * wait to enter it.
2030 	 */
2031 	while (ifnet_lock_entrances(il) != il->il_nexit)
2032 		cv_wait(&il->il_emptied, &il->il_lock);
2033 	/* At this point, we are the only thread still in the critical
2034 	 * section, and no new thread can get a handle on the ifioctl
2035 	 * lock, so it is safe to free its memory.
2036 	 */
2037 	mutex_exit(&il->il_lock);
2038 	ifp->if_ioctl_lock = NULL;
2039 	percpu_free(il->il_nenter, sizeof(uint64_t));
2040 	il->il_nenter = NULL;
2041 	cv_destroy(&il->il_emptied);
2042 	mutex_destroy(&il->il_lock);
2043 	kmem_free(il, sizeof(*il));
2044 }
2045 
2046 /*
2047  * Return interface configuration
2048  * of system.  List may be used
2049  * in later ioctl's (above) to get
2050  * other information.
2051  *
2052  * Each record is a struct ifreq.  Before the addition of
2053  * sockaddr_storage, the API rule was that sockaddr flavors that did
2054  * not fit would extend beyond the struct ifreq, with the next struct
2055  * ifreq starting sa_len beyond the struct sockaddr.  Because the
2056  * union in struct ifreq includes struct sockaddr_storage, every kind
2057  * of sockaddr must fit.  Thus, there are no longer any overlength
2058  * records.
2059  *
2060  * Records are added to the user buffer if they fit, and ifc_len is
2061  * adjusted to the length that was written.  Thus, the user is only
2062  * assured of getting the complete list if ifc_len on return is at
2063  * least sizeof(struct ifreq) less than it was on entry.
2064  *
2065  * If the user buffer pointer is NULL, this routine copies no data and
2066  * returns the amount of space that would be needed.
2067  *
2068  * Invariants:
2069  * ifrp points to the next part of the user's buffer to be used.  If
2070  * ifrp != NULL, space holds the number of bytes remaining that we may
2071  * write at ifrp.  Otherwise, space holds the number of bytes that
2072  * would have been written had there been adequate space.
2073  */
2074 /*ARGSUSED*/
2075 int
2076 ifconf(u_long cmd, void *data)
2077 {
2078 	struct ifconf *ifc = (struct ifconf *)data;
2079 	struct ifnet *ifp;
2080 	struct ifaddr *ifa;
2081 	struct ifreq ifr, *ifrp;
2082 	int space, error = 0;
2083 	const int sz = (int)sizeof(struct ifreq);
2084 
2085 	if ((ifrp = ifc->ifc_req) == NULL)
2086 		space = 0;
2087 	else
2088 		space = ifc->ifc_len;
2089 	IFNET_FOREACH(ifp) {
2090 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2091 		    sizeof(ifr.ifr_name));
2092 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2093 			return ENAMETOOLONG;
2094 		if (IFADDR_EMPTY(ifp)) {
2095 			/* Interface with no addresses - send zero sockaddr. */
2096 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2097 			if (ifrp == NULL) {
2098 				space += sz;
2099 				continue;
2100 			}
2101 			if (space >= sz) {
2102 				error = copyout(&ifr, ifrp, sz);
2103 				if (error != 0)
2104 					return error;
2105 				ifrp++;
2106 				space -= sz;
2107 			}
2108 		}
2109 
2110 		IFADDR_FOREACH(ifa, ifp) {
2111 			struct sockaddr *sa = ifa->ifa_addr;
2112 			/* all sockaddrs must fit in sockaddr_storage */
2113 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2114 
2115 			if (ifrp == NULL) {
2116 				space += sz;
2117 				continue;
2118 			}
2119 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2120 			if (space >= sz) {
2121 				error = copyout(&ifr, ifrp, sz);
2122 				if (error != 0)
2123 					return (error);
2124 				ifrp++; space -= sz;
2125 			}
2126 		}
2127 	}
2128 	if (ifrp != NULL) {
2129 		KASSERT(0 <= space && space <= ifc->ifc_len);
2130 		ifc->ifc_len -= space;
2131 	} else {
2132 		KASSERT(space >= 0);
2133 		ifc->ifc_len = space;
2134 	}
2135 	return (0);
2136 }
2137 
2138 int
2139 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2140 {
2141 	uint8_t len;
2142 #ifdef COMPAT_OIFREQ
2143 	struct ifreq ifrb;
2144 	struct oifreq *oifr = NULL;
2145 	u_long ocmd = cmd;
2146 	cmd = compat_cvtcmd(cmd);
2147 	if (cmd != ocmd) {
2148 		oifr = (struct oifreq *)(void *)ifr;
2149 		ifr = &ifrb;
2150 		ifreqo2n(oifr, ifr);
2151 		len = sizeof(oifr->ifr_addr);
2152 	} else
2153 #endif
2154 		len = sizeof(ifr->ifr_ifru.ifru_space);
2155 
2156 	if (len < sa->sa_len)
2157 		return EFBIG;
2158 
2159 	memset(&ifr->ifr_addr, 0, len);
2160 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2161 
2162 #ifdef COMPAT_OIFREQ
2163 	if (cmd != ocmd)
2164 		ifreqn2o(oifr, ifr);
2165 #endif
2166 	return 0;
2167 }
2168 
2169 /*
2170  * Queue message on interface, and start output if interface
2171  * not yet active.
2172  */
2173 int
2174 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2175     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2176 {
2177 	int len = m->m_pkthdr.len;
2178 	int mflags = m->m_flags;
2179 	int s = splnet();
2180 	int error;
2181 
2182 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2183 	if (error != 0)
2184 		goto out;
2185 	ifp->if_obytes += len;
2186 	if (mflags & M_MCAST)
2187 		ifp->if_omcasts++;
2188 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2189 		(*ifp->if_start)(ifp);
2190 out:
2191 	splx(s);
2192 	return error;
2193 }
2194 
2195 /*
2196  * Queue message on interface, possibly using a second fast queue
2197  */
2198 int
2199 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2200     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2201 {
2202 	int error = 0;
2203 
2204 	if (ifq != NULL
2205 #ifdef ALTQ
2206 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2207 #endif
2208 	    ) {
2209 		if (IF_QFULL(ifq)) {
2210 			IF_DROP(&ifp->if_snd);
2211 			m_freem(m);
2212 			if (error == 0)
2213 				error = ENOBUFS;
2214 		} else
2215 			IF_ENQUEUE(ifq, m);
2216 	} else
2217 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2218 	if (error != 0) {
2219 		++ifp->if_oerrors;
2220 		return error;
2221 	}
2222 	return 0;
2223 }
2224 
2225 int
2226 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2227 {
2228 	int rc;
2229 
2230 	if (ifp->if_initaddr != NULL)
2231 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2232 	else if (src ||
2233 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2234 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2235 
2236 	return rc;
2237 }
2238 
2239 int
2240 if_flags_set(ifnet_t *ifp, const short flags)
2241 {
2242 	int rc;
2243 
2244 	if (ifp->if_setflags != NULL)
2245 		rc = (*ifp->if_setflags)(ifp, flags);
2246 	else {
2247 		short cantflags, chgdflags;
2248 		struct ifreq ifr;
2249 
2250 		chgdflags = ifp->if_flags ^ flags;
2251 		cantflags = chgdflags & IFF_CANTCHANGE;
2252 
2253 		if (cantflags != 0)
2254 			ifp->if_flags ^= cantflags;
2255 
2256                 /* Traditionally, we do not call if_ioctl after
2257                  * setting/clearing only IFF_PROMISC if the interface
2258                  * isn't IFF_UP.  Uphold that tradition.
2259 		 */
2260 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2261 			return 0;
2262 
2263 		memset(&ifr, 0, sizeof(ifr));
2264 
2265 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2266 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2267 
2268 		if (rc != 0 && cantflags != 0)
2269 			ifp->if_flags ^= cantflags;
2270 	}
2271 
2272 	return rc;
2273 }
2274 
2275 int
2276 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2277 {
2278 	int rc;
2279 	struct ifreq ifr;
2280 
2281 	if (ifp->if_mcastop != NULL)
2282 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2283 	else {
2284 		ifreq_setaddr(cmd, &ifr, sa);
2285 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2286 	}
2287 
2288 	return rc;
2289 }
2290 
2291 static void
2292 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2293     struct ifaltq *ifq)
2294 {
2295 	const struct sysctlnode *cnode, *rnode;
2296 
2297 	if (sysctl_createv(clog, 0, NULL, &rnode,
2298 		       CTLFLAG_PERMANENT,
2299 		       CTLTYPE_NODE, "interfaces",
2300 		       SYSCTL_DESCR("Per-interface controls"),
2301 		       NULL, 0, NULL, 0,
2302 		       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
2303 		goto bad;
2304 
2305 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2306 		       CTLFLAG_PERMANENT,
2307 		       CTLTYPE_NODE, ifname,
2308 		       SYSCTL_DESCR("Interface controls"),
2309 		       NULL, 0, NULL, 0,
2310 		       CTL_CREATE, CTL_EOL) != 0)
2311 		goto bad;
2312 
2313 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2314 		       CTLFLAG_PERMANENT,
2315 		       CTLTYPE_NODE, "sndq",
2316 		       SYSCTL_DESCR("Interface output queue controls"),
2317 		       NULL, 0, NULL, 0,
2318 		       CTL_CREATE, CTL_EOL) != 0)
2319 		goto bad;
2320 
2321 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2322 		       CTLFLAG_PERMANENT,
2323 		       CTLTYPE_INT, "len",
2324 		       SYSCTL_DESCR("Current output queue length"),
2325 		       NULL, 0, &ifq->ifq_len, 0,
2326 		       CTL_CREATE, CTL_EOL) != 0)
2327 		goto bad;
2328 
2329 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2330 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2331 		       CTLTYPE_INT, "maxlen",
2332 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2333 		       NULL, 0, &ifq->ifq_maxlen, 0,
2334 		       CTL_CREATE, CTL_EOL) != 0)
2335 		goto bad;
2336 
2337 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2338 		       CTLFLAG_PERMANENT,
2339 		       CTLTYPE_INT, "drops",
2340 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2341 		       NULL, 0, &ifq->ifq_drops, 0,
2342 		       CTL_CREATE, CTL_EOL) != 0)
2343 		goto bad;
2344 
2345 	return;
2346 bad:
2347 	printf("%s: could not attach sysctl nodes\n", ifname);
2348 	return;
2349 }
2350 
2351 #if defined(INET) || defined(INET6)
2352 
2353 #define	SYSCTL_NET_PKTQ(q, cn, c)					\
2354 	static int							\
2355 	sysctl_net_##q##_##cn(SYSCTLFN_ARGS)				\
2356 	{								\
2357 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c);	\
2358 	}
2359 
2360 #if defined(INET)
2361 static int
2362 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS)
2363 {
2364 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq);
2365 }
2366 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS)
2367 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS)
2368 #endif
2369 
2370 #if defined(INET6)
2371 static int
2372 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS)
2373 {
2374 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq);
2375 }
2376 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS)
2377 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS)
2378 #endif
2379 
2380 static void
2381 sysctl_net_pktq_setup(struct sysctllog **clog, int pf)
2382 {
2383 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
2384 	const char *pfname = NULL, *ipname = NULL;
2385 	int ipn = 0, qid = 0;
2386 
2387 	switch (pf) {
2388 #if defined(INET)
2389 	case PF_INET:
2390 		len_func = sysctl_net_ip_pktq_items;
2391 		maxlen_func = sysctl_net_ip_pktq_maxlen;
2392 		drops_func = sysctl_net_ip_pktq_drops;
2393 		pfname = "inet", ipn = IPPROTO_IP;
2394 		ipname = "ip", qid = IPCTL_IFQ;
2395 		break;
2396 #endif
2397 #if defined(INET6)
2398 	case PF_INET6:
2399 		len_func = sysctl_net_ip6_pktq_items;
2400 		maxlen_func = sysctl_net_ip6_pktq_maxlen;
2401 		drops_func = sysctl_net_ip6_pktq_drops;
2402 		pfname = "inet6", ipn = IPPROTO_IPV6;
2403 		ipname = "ip6", qid = IPV6CTL_IFQ;
2404 		break;
2405 #endif
2406 	default:
2407 		KASSERT(false);
2408 	}
2409 
2410 	sysctl_createv(clog, 0, NULL, NULL,
2411 		       CTLFLAG_PERMANENT,
2412 		       CTLTYPE_NODE, pfname, NULL,
2413 		       NULL, 0, NULL, 0,
2414 		       CTL_NET, pf, CTL_EOL);
2415 	sysctl_createv(clog, 0, NULL, NULL,
2416 		       CTLFLAG_PERMANENT,
2417 		       CTLTYPE_NODE, ipname, NULL,
2418 		       NULL, 0, NULL, 0,
2419 		       CTL_NET, pf, ipn, CTL_EOL);
2420 	sysctl_createv(clog, 0, NULL, NULL,
2421 		       CTLFLAG_PERMANENT,
2422 		       CTLTYPE_NODE, "ifq",
2423 		       SYSCTL_DESCR("Protocol input queue controls"),
2424 		       NULL, 0, NULL, 0,
2425 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2426 
2427 	sysctl_createv(clog, 0, NULL, NULL,
2428 		       CTLFLAG_PERMANENT,
2429 		       CTLTYPE_INT, "len",
2430 		       SYSCTL_DESCR("Current input queue length"),
2431 		       len_func, 0, NULL, 0,
2432 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2433 	sysctl_createv(clog, 0, NULL, NULL,
2434 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2435 		       CTLTYPE_INT, "maxlen",
2436 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2437 		       maxlen_func, 0, NULL, 0,
2438 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2439 	sysctl_createv(clog, 0, NULL, NULL,
2440 		       CTLFLAG_PERMANENT,
2441 		       CTLTYPE_INT, "drops",
2442 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2443 		       drops_func, 0, NULL, 0,
2444 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2445 }
2446 #endif /* INET || INET6 */
2447 
2448 static int
2449 if_sdl_sysctl(SYSCTLFN_ARGS)
2450 {
2451 	struct ifnet *ifp;
2452 	const struct sockaddr_dl *sdl;
2453 
2454 	if (namelen != 1)
2455 		return EINVAL;
2456 
2457 	ifp = if_byindex(name[0]);
2458 	if (ifp == NULL)
2459 		return ENODEV;
2460 
2461 	sdl = ifp->if_sadl;
2462 	if (sdl == NULL) {
2463 		*oldlenp = 0;
2464 		return 0;
2465 	}
2466 
2467 	if (oldp == NULL) {
2468 		*oldlenp = sdl->sdl_alen;
2469 		return 0;
2470 	}
2471 
2472 	if (*oldlenp >= sdl->sdl_alen)
2473 		*oldlenp = sdl->sdl_alen;
2474 	return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp);
2475 }
2476 
2477 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup")
2478 {
2479 	const struct sysctlnode *rnode = NULL;
2480 
2481 	sysctl_createv(clog, 0, NULL, &rnode,
2482 		       CTLFLAG_PERMANENT,
2483 		       CTLTYPE_NODE, "sdl",
2484 		       SYSCTL_DESCR("Get active link-layer address"),
2485 		       if_sdl_sysctl, 0, NULL, 0,
2486 		       CTL_NET, CTL_CREATE, CTL_EOL);
2487 }
2488