xref: /netbsd-src/sys/net/if.c (revision ba65fde2d7fefa7d39838fa5fa855e62bd606b5e)
1 /*	$NetBSD: if.c,v 1.261 2012/11/01 06:36:30 msaitoh 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 1.261 2012/11/01 06:36:30 msaitoh Exp $");
94 
95 #include "opt_inet.h"
96 
97 #include "opt_atalk.h"
98 #include "opt_natm.h"
99 #include "opt_pfil_hooks.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 
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 #ifdef NETATALK
128 #include <netatalk/at_extern.h>
129 #include <netatalk/at.h>
130 #endif
131 #include <net/pfil.h>
132 
133 #ifdef INET6
134 #include <netinet/in.h>
135 #include <netinet6/in6_var.h>
136 #include <netinet6/nd6.h>
137 #endif
138 
139 #include "carp.h"
140 #if NCARP > 0
141 #include <netinet/ip_carp.h>
142 #endif
143 
144 #include <compat/sys/sockio.h>
145 #include <compat/sys/socket.h>
146 
147 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
148 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
149 
150 int	ifqmaxlen = IFQ_MAXLEN;
151 callout_t if_slowtimo_ch;
152 
153 int netisr;			/* scheduling bits for network */
154 
155 static int	if_rt_walktree(struct rtentry *, void *);
156 
157 static struct if_clone *if_clone_lookup(const char *, int *);
158 static int	if_clone_list(struct if_clonereq *);
159 
160 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
161 static int if_cloners_count;
162 
163 static uint64_t index_gen;
164 static kmutex_t index_gen_mtx;
165 
166 #ifdef PFIL_HOOKS
167 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
168 #endif
169 
170 static kauth_listener_t if_listener;
171 
172 static int ifioctl_attach(struct ifnet *);
173 static void ifioctl_detach(struct ifnet *);
174 static void ifnet_lock_enter(struct ifnet_lock *);
175 static void ifnet_lock_exit(struct ifnet_lock *);
176 static void if_detach_queues(struct ifnet *, struct ifqueue *);
177 static void sysctl_sndq_setup(struct sysctllog **, const char *,
178     struct ifaltq *);
179 
180 #if defined(INET) || defined(INET6)
181 static void sysctl_net_ifq_setup(struct sysctllog **, int, const char *,
182 				 int, const char *, int, struct ifqueue *);
183 #endif
184 
185 static int
186 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
187     void *arg0, void *arg1, void *arg2, void *arg3)
188 {
189 	int result;
190 	enum kauth_network_req req;
191 
192 	result = KAUTH_RESULT_DEFER;
193 	req = (enum kauth_network_req)arg1;
194 
195 	if (action != KAUTH_NETWORK_INTERFACE)
196 		return result;
197 
198 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
199 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
200 		result = KAUTH_RESULT_ALLOW;
201 
202 	return result;
203 }
204 
205 /*
206  * Network interface utility routines.
207  *
208  * Routines with ifa_ifwith* names take sockaddr *'s as
209  * parameters.
210  */
211 void
212 ifinit(void)
213 {
214 #ifdef INET
215 	{extern struct ifqueue ipintrq;
216 	sysctl_net_ifq_setup(NULL, PF_INET, "inet", IPPROTO_IP, "ip",
217 			     IPCTL_IFQ, &ipintrq);}
218 #endif /* INET */
219 #ifdef INET6
220 	{extern struct ifqueue ip6intrq;
221 	sysctl_net_ifq_setup(NULL, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
222 			     IPV6CTL_IFQ, &ip6intrq);}
223 #endif /* INET6 */
224 
225 	callout_init(&if_slowtimo_ch, 0);
226 	if_slowtimo(NULL);
227 
228 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
229 	    if_listener_cb, NULL);
230 }
231 
232 /*
233  * XXX Initialization before configure().
234  * XXX hack to get pfil_add_hook working in autoconf.
235  */
236 void
237 ifinit1(void)
238 {
239 
240 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
241 
242 #ifdef PFIL_HOOKS
243 	if_pfil.ph_type = PFIL_TYPE_IFNET;
244 	if_pfil.ph_ifnet = NULL;
245 	if (pfil_head_register(&if_pfil) != 0)
246 		printf("WARNING: unable to register pfil hook\n");
247 #endif
248 }
249 
250 struct ifnet *
251 if_alloc(u_char type)
252 {
253 	return malloc(sizeof(struct ifnet), M_DEVBUF, M_WAITOK|M_ZERO);
254 }
255 
256 void
257 if_free(struct ifnet *ifp)
258 {
259 	free(ifp, M_DEVBUF);
260 }
261 
262 void
263 if_initname(struct ifnet *ifp, const char *name, int unit)
264 {
265 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
266 	    "%s%d", name, unit);
267 }
268 
269 /*
270  * Null routines used while an interface is going away.  These routines
271  * just return an error.
272  */
273 
274 int
275 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
276     const struct sockaddr *so, struct rtentry *rt)
277 {
278 
279 	return ENXIO;
280 }
281 
282 void
283 if_nullinput(struct ifnet *ifp, struct mbuf *m)
284 {
285 
286 	/* Nothing. */
287 }
288 
289 void
290 if_nullstart(struct ifnet *ifp)
291 {
292 
293 	/* Nothing. */
294 }
295 
296 int
297 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
298 {
299 
300 	/* Wake ifioctl_detach(), who may wait for all threads to
301 	 * quit the critical section.
302 	 */
303 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
304 	return ENXIO;
305 }
306 
307 int
308 if_nullinit(struct ifnet *ifp)
309 {
310 
311 	return ENXIO;
312 }
313 
314 void
315 if_nullstop(struct ifnet *ifp, int disable)
316 {
317 
318 	/* Nothing. */
319 }
320 
321 void
322 if_nullwatchdog(struct ifnet *ifp)
323 {
324 
325 	/* Nothing. */
326 }
327 
328 void
329 if_nulldrain(struct ifnet *ifp)
330 {
331 
332 	/* Nothing. */
333 }
334 
335 static u_int if_index = 1;
336 struct ifnet_head ifnet;
337 size_t if_indexlim = 0;
338 struct ifaddr **ifnet_addrs = NULL;
339 struct ifnet **ifindex2ifnet = NULL;
340 struct ifnet *lo0ifp;
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 /*
495  * Attach an interface to the
496  * list of "active" interfaces.
497  */
498 void
499 if_attach(struct ifnet *ifp)
500 {
501 	int indexlim = 0;
502 
503 	if (if_indexlim == 0) {
504 		TAILQ_INIT(&ifnet);
505 		if_indexlim = 8;
506 	}
507 	TAILQ_INIT(&ifp->if_addrlist);
508 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
509 
510 	if (ifioctl_attach(ifp) != 0)
511 		panic("%s: ifioctl_attach() failed", __func__);
512 
513 	mutex_enter(&index_gen_mtx);
514 	ifp->if_index_gen = index_gen++;
515 	mutex_exit(&index_gen_mtx);
516 
517 	ifp->if_index = if_index;
518 	if (ifindex2ifnet == NULL)
519 		if_index++;
520 	else
521 		while (ifp->if_index < if_indexlim &&
522 		    ifindex2ifnet[ifp->if_index] != NULL) {
523 			++if_index;
524 			if (if_index == 0)
525 				if_index = 1;
526 			/*
527 			 * If we hit USHRT_MAX, we skip back to 0 since
528 			 * there are a number of places where the value
529 			 * of if_index or if_index itself is compared
530 			 * to or stored in an unsigned short.  By
531 			 * jumping back, we won't botch those assignments
532 			 * or comparisons.
533 			 */
534 			else if (if_index == USHRT_MAX) {
535 				/*
536 				 * However, if we have to jump back to
537 				 * zero *twice* without finding an empty
538 				 * slot in ifindex2ifnet[], then there
539 				 * there are too many (>65535) interfaces.
540 				 */
541 				if (indexlim++)
542 					panic("too many interfaces");
543 				else
544 					if_index = 1;
545 			}
546 			ifp->if_index = if_index;
547 		}
548 
549 	/*
550 	 * We have some arrays that should be indexed by if_index.
551 	 * since if_index will grow dynamically, they should grow too.
552 	 *	struct ifadd **ifnet_addrs
553 	 *	struct ifnet **ifindex2ifnet
554 	 */
555 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
556 	    ifp->if_index >= if_indexlim) {
557 		size_t m, n, oldlim;
558 		void *q;
559 
560 		oldlim = if_indexlim;
561 		while (ifp->if_index >= if_indexlim)
562 			if_indexlim <<= 1;
563 
564 		/* grow ifnet_addrs */
565 		m = oldlim * sizeof(struct ifaddr *);
566 		n = if_indexlim * sizeof(struct ifaddr *);
567 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
568 		if (ifnet_addrs != NULL) {
569 			memcpy(q, ifnet_addrs, m);
570 			free(ifnet_addrs, M_IFADDR);
571 		}
572 		ifnet_addrs = (struct ifaddr **)q;
573 
574 		/* grow ifindex2ifnet */
575 		m = oldlim * sizeof(struct ifnet *);
576 		n = if_indexlim * sizeof(struct ifnet *);
577 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
578 		if (ifindex2ifnet != NULL) {
579 			memcpy(q, ifindex2ifnet, m);
580 			free(ifindex2ifnet, M_IFADDR);
581 		}
582 		ifindex2ifnet = (struct ifnet **)q;
583 	}
584 
585 	ifindex2ifnet[ifp->if_index] = ifp;
586 
587 	/*
588 	 * Link level name is allocated later by a separate call to
589 	 * if_alloc_sadl().
590 	 */
591 
592 	if (ifp->if_snd.ifq_maxlen == 0)
593 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
594 
595 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
596 
597 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
598 
599 	ifp->if_link_state = LINK_STATE_UNKNOWN;
600 
601 	ifp->if_capenable = 0;
602 	ifp->if_csum_flags_tx = 0;
603 	ifp->if_csum_flags_rx = 0;
604 
605 #ifdef ALTQ
606 	ifp->if_snd.altq_type = 0;
607 	ifp->if_snd.altq_disc = NULL;
608 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
609 	ifp->if_snd.altq_tbr  = NULL;
610 	ifp->if_snd.altq_ifp  = ifp;
611 #endif
612 
613 #ifdef PFIL_HOOKS
614 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
615 	ifp->if_pfil.ph_ifnet = ifp;
616 	if (pfil_head_register(&ifp->if_pfil) != 0)
617 		printf("%s: WARNING: unable to register pfil hook\n",
618 		    ifp->if_xname);
619 	(void)pfil_run_hooks(&if_pfil,
620 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
621 #endif
622 
623 	if (!STAILQ_EMPTY(&domains))
624 		if_attachdomain1(ifp);
625 
626 	/* Announce the interface. */
627 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
628 }
629 
630 void
631 if_attachdomain(void)
632 {
633 	struct ifnet *ifp;
634 	int s;
635 
636 	s = splnet();
637 	IFNET_FOREACH(ifp)
638 		if_attachdomain1(ifp);
639 	splx(s);
640 }
641 
642 void
643 if_attachdomain1(struct ifnet *ifp)
644 {
645 	struct domain *dp;
646 	int s;
647 
648 	s = splnet();
649 
650 	/* address family dependent data region */
651 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
652 	DOMAIN_FOREACH(dp) {
653 		if (dp->dom_ifattach != NULL)
654 			ifp->if_afdata[dp->dom_family] =
655 			    (*dp->dom_ifattach)(ifp);
656 	}
657 
658 	splx(s);
659 }
660 
661 /*
662  * Deactivate an interface.  This points all of the procedure
663  * handles at error stubs.  May be called from interrupt context.
664  */
665 void
666 if_deactivate(struct ifnet *ifp)
667 {
668 	int s;
669 
670 	s = splnet();
671 
672 	ifp->if_output	 = if_nulloutput;
673 	ifp->if_input	 = if_nullinput;
674 	ifp->if_start	 = if_nullstart;
675 	ifp->if_ioctl	 = if_nullioctl;
676 	ifp->if_init	 = if_nullinit;
677 	ifp->if_stop	 = if_nullstop;
678 	ifp->if_watchdog = if_nullwatchdog;
679 	ifp->if_drain	 = if_nulldrain;
680 
681 	/* No more packets may be enqueued. */
682 	ifp->if_snd.ifq_maxlen = 0;
683 
684 	splx(s);
685 }
686 
687 void
688 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
689 {
690 	struct ifaddr *ifa, *nifa;
691 
692 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
693 		nifa = IFADDR_NEXT(ifa);
694 		if (ifa->ifa_addr->sa_family != family)
695 			continue;
696 		(*purgeaddr)(ifa);
697 	}
698 }
699 
700 /*
701  * Detach an interface from the list of "active" interfaces,
702  * freeing any resources as we go along.
703  *
704  * NOTE: This routine must be called with a valid thread context,
705  * as it may block.
706  */
707 void
708 if_detach(struct ifnet *ifp)
709 {
710 	struct socket so;
711 	struct ifaddr *ifa;
712 #ifdef IFAREF_DEBUG
713 	struct ifaddr *last_ifa = NULL;
714 #endif
715 	struct domain *dp;
716 	const struct protosw *pr;
717 	int s, i, family, purged;
718 
719 	/*
720 	 * XXX It's kind of lame that we have to have the
721 	 * XXX socket structure...
722 	 */
723 	memset(&so, 0, sizeof(so));
724 
725 	s = splnet();
726 
727 	/*
728 	 * Do an if_down() to give protocols a chance to do something.
729 	 */
730 	if_down(ifp);
731 
732 #ifdef ALTQ
733 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
734 		altq_disable(&ifp->if_snd);
735 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
736 		altq_detach(&ifp->if_snd);
737 #endif
738 
739 	sysctl_teardown(&ifp->if_sysctl_log);
740 
741 #if NCARP > 0
742 	/* Remove the interface from any carp group it is a part of.  */
743 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
744 		carp_ifdetach(ifp);
745 #endif
746 
747 	/*
748 	 * Rip all the addresses off the interface.  This should make
749 	 * all of the routes go away.
750 	 *
751 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
752 	 * from the list, including our "cursor", ifa.  For safety,
753 	 * and to honor the TAILQ abstraction, I just restart the
754 	 * loop after each removal.  Note that the loop will exit
755 	 * when all of the remaining ifaddrs belong to the AF_LINK
756 	 * family.  I am counting on the historical fact that at
757 	 * least one pr_usrreq in each address domain removes at
758 	 * least one ifaddr.
759 	 */
760 again:
761 	IFADDR_FOREACH(ifa, ifp) {
762 		family = ifa->ifa_addr->sa_family;
763 #ifdef IFAREF_DEBUG
764 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
765 		    ifa, family, ifa->ifa_refcnt);
766 		if (last_ifa != NULL && ifa == last_ifa)
767 			panic("if_detach: loop detected");
768 		last_ifa = ifa;
769 #endif
770 		if (family == AF_LINK)
771 			continue;
772 		dp = pffinddomain(family);
773 #ifdef DIAGNOSTIC
774 		if (dp == NULL)
775 			panic("if_detach: no domain for AF %d",
776 			    family);
777 #endif
778 		/*
779 		 * XXX These PURGEIF calls are redundant with the
780 		 * purge-all-families calls below, but are left in for
781 		 * now both to make a smaller change, and to avoid
782 		 * unplanned interactions with clearing of
783 		 * ifp->if_addrlist.
784 		 */
785 		purged = 0;
786 		for (pr = dp->dom_protosw;
787 		     pr < dp->dom_protoswNPROTOSW; pr++) {
788 			so.so_proto = pr;
789 			if (pr->pr_usrreq != NULL) {
790 				(void) (*pr->pr_usrreq)(&so,
791 				    PRU_PURGEIF, NULL, NULL,
792 				    (struct mbuf *) ifp, curlwp);
793 				purged = 1;
794 			}
795 		}
796 		if (purged == 0) {
797 			/*
798 			 * XXX What's really the best thing to do
799 			 * XXX here?  --thorpej@NetBSD.org
800 			 */
801 			printf("if_detach: WARNING: AF %d not purged\n",
802 			    family);
803 			ifa_remove(ifp, ifa);
804 		}
805 		goto again;
806 	}
807 
808 	if_free_sadl(ifp);
809 
810 	/* Walk the routing table looking for stragglers. */
811 	for (i = 0; i <= AF_MAX; i++) {
812 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
813 			continue;
814 	}
815 
816 	DOMAIN_FOREACH(dp) {
817 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
818 		{
819 			void *p = ifp->if_afdata[dp->dom_family];
820 			if (p) {
821 				ifp->if_afdata[dp->dom_family] = NULL;
822 				(*dp->dom_ifdetach)(ifp, p);
823 			}
824 		}
825 
826 		/*
827 		 * One would expect multicast memberships (INET and
828 		 * INET6) on UDP sockets to be purged by the PURGEIF
829 		 * calls above, but if all addresses were removed from
830 		 * the interface prior to destruction, the calls will
831 		 * not be made (e.g. ppp, for which pppd(8) generally
832 		 * removes addresses before destroying the interface).
833 		 * Because there is no invariant that multicast
834 		 * memberships only exist for interfaces with IPv4
835 		 * addresses, we must call PURGEIF regardless of
836 		 * addresses.  (Protocols which might store ifnet
837 		 * pointers are marked with PR_PURGEIF.)
838 		 */
839 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
840 			so.so_proto = pr;
841 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
842 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
843 				    NULL, (struct mbuf *)ifp, curlwp);
844 		}
845 	}
846 
847 #ifdef PFIL_HOOKS
848 	(void)pfil_run_hooks(&if_pfil,
849 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
850 	(void)pfil_head_unregister(&ifp->if_pfil);
851 #endif
852 
853 	/* Announce that the interface is gone. */
854 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
855 
856 	ifindex2ifnet[ifp->if_index] = NULL;
857 
858 	TAILQ_REMOVE(&ifnet, ifp, if_list);
859 
860 	ifioctl_detach(ifp);
861 
862 	/*
863 	 * remove packets that came from ifp, from software interrupt queues.
864 	 */
865 	DOMAIN_FOREACH(dp) {
866 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
867 			struct ifqueue *iq = dp->dom_ifqueues[i];
868 			if (iq == NULL)
869 				break;
870 			dp->dom_ifqueues[i] = NULL;
871 			if_detach_queues(ifp, iq);
872 		}
873 	}
874 
875 	splx(s);
876 }
877 
878 static void
879 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
880 {
881 	struct mbuf *m, *prev, *next;
882 
883 	prev = NULL;
884 	for (m = q->ifq_head; m != NULL; m = next) {
885 		next = m->m_nextpkt;
886 #ifdef DIAGNOSTIC
887 		if ((m->m_flags & M_PKTHDR) == 0) {
888 			prev = m;
889 			continue;
890 		}
891 #endif
892 		if (m->m_pkthdr.rcvif != ifp) {
893 			prev = m;
894 			continue;
895 		}
896 
897 		if (prev != NULL)
898 			prev->m_nextpkt = m->m_nextpkt;
899 		else
900 			q->ifq_head = m->m_nextpkt;
901 		if (q->ifq_tail == m)
902 			q->ifq_tail = prev;
903 		q->ifq_len--;
904 
905 		m->m_nextpkt = NULL;
906 		m_freem(m);
907 		IF_DROP(q);
908 	}
909 }
910 
911 /*
912  * Callback for a radix tree walk to delete all references to an
913  * ifnet.
914  */
915 static int
916 if_rt_walktree(struct rtentry *rt, void *v)
917 {
918 	struct ifnet *ifp = (struct ifnet *)v;
919 	int error;
920 
921 	if (rt->rt_ifp != ifp)
922 		return 0;
923 
924 	/* Delete the entry. */
925 	++rt->rt_refcnt;
926 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
927 	    rt_mask(rt), rt->rt_flags, NULL);
928 	KASSERT((rt->rt_flags & RTF_UP) == 0);
929 	rt->rt_ifp = NULL;
930 	RTFREE(rt);
931 	if (error != 0)
932 		printf("%s: warning: unable to delete rtentry @ %p, "
933 		    "error = %d\n", ifp->if_xname, rt, error);
934 	return ERESTART;
935 }
936 
937 /*
938  * Create a clone network interface.
939  */
940 int
941 if_clone_create(const char *name)
942 {
943 	struct if_clone *ifc;
944 	int unit;
945 
946 	ifc = if_clone_lookup(name, &unit);
947 	if (ifc == NULL)
948 		return EINVAL;
949 
950 	if (ifunit(name) != NULL)
951 		return EEXIST;
952 
953 	return (*ifc->ifc_create)(ifc, unit);
954 }
955 
956 /*
957  * Destroy a clone network interface.
958  */
959 int
960 if_clone_destroy(const char *name)
961 {
962 	struct if_clone *ifc;
963 	struct ifnet *ifp;
964 
965 	ifc = if_clone_lookup(name, NULL);
966 	if (ifc == NULL)
967 		return EINVAL;
968 
969 	ifp = ifunit(name);
970 	if (ifp == NULL)
971 		return ENXIO;
972 
973 	if (ifc->ifc_destroy == NULL)
974 		return EOPNOTSUPP;
975 
976 	return (*ifc->ifc_destroy)(ifp);
977 }
978 
979 /*
980  * Look up a network interface cloner.
981  */
982 static struct if_clone *
983 if_clone_lookup(const char *name, int *unitp)
984 {
985 	struct if_clone *ifc;
986 	const char *cp;
987 	int unit;
988 
989 	/* separate interface name from unit */
990 	for (cp = name;
991 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
992 	    cp++)
993 		continue;
994 
995 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
996 		return NULL;	/* No name or unit number */
997 
998 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
999 		if (strlen(ifc->ifc_name) == cp - name &&
1000 		    strncmp(name, ifc->ifc_name, cp - name) == 0)
1001 			break;
1002 	}
1003 
1004 	if (ifc == NULL)
1005 		return NULL;
1006 
1007 	unit = 0;
1008 	while (cp - name < IFNAMSIZ && *cp) {
1009 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1010 			/* Bogus unit number. */
1011 			return NULL;
1012 		}
1013 		unit = (unit * 10) + (*cp++ - '0');
1014 	}
1015 
1016 	if (unitp != NULL)
1017 		*unitp = unit;
1018 	return ifc;
1019 }
1020 
1021 /*
1022  * Register a network interface cloner.
1023  */
1024 void
1025 if_clone_attach(struct if_clone *ifc)
1026 {
1027 
1028 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1029 	if_cloners_count++;
1030 }
1031 
1032 /*
1033  * Unregister a network interface cloner.
1034  */
1035 void
1036 if_clone_detach(struct if_clone *ifc)
1037 {
1038 
1039 	LIST_REMOVE(ifc, ifc_list);
1040 	if_cloners_count--;
1041 }
1042 
1043 /*
1044  * Provide list of interface cloners to userspace.
1045  */
1046 static int
1047 if_clone_list(struct if_clonereq *ifcr)
1048 {
1049 	char outbuf[IFNAMSIZ], *dst;
1050 	struct if_clone *ifc;
1051 	int count, error = 0;
1052 
1053 	ifcr->ifcr_total = if_cloners_count;
1054 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1055 		/* Just asking how many there are. */
1056 		return 0;
1057 	}
1058 
1059 	if (ifcr->ifcr_count < 0)
1060 		return EINVAL;
1061 
1062 	count = (if_cloners_count < ifcr->ifcr_count) ?
1063 	    if_cloners_count : ifcr->ifcr_count;
1064 
1065 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1066 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1067 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1068 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1069 			return ENAMETOOLONG;
1070 		error = copyout(outbuf, dst, sizeof(outbuf));
1071 		if (error != 0)
1072 			break;
1073 	}
1074 
1075 	return error;
1076 }
1077 
1078 void
1079 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1080 {
1081 	ifa->ifa_ifp = ifp;
1082 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1083 	IFAREF(ifa);
1084 }
1085 
1086 void
1087 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1088 {
1089 	KASSERT(ifa->ifa_ifp == ifp);
1090 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1091 	IFAFREE(ifa);
1092 }
1093 
1094 static inline int
1095 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1096 {
1097 	return sockaddr_cmp(sa1, sa2) == 0;
1098 }
1099 
1100 /*
1101  * Locate an interface based on a complete address.
1102  */
1103 /*ARGSUSED*/
1104 struct ifaddr *
1105 ifa_ifwithaddr(const struct sockaddr *addr)
1106 {
1107 	struct ifnet *ifp;
1108 	struct ifaddr *ifa;
1109 
1110 	IFNET_FOREACH(ifp) {
1111 		if (ifp->if_output == if_nulloutput)
1112 			continue;
1113 		IFADDR_FOREACH(ifa, ifp) {
1114 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1115 				continue;
1116 			if (equal(addr, ifa->ifa_addr))
1117 				return ifa;
1118 			if ((ifp->if_flags & IFF_BROADCAST) &&
1119 			    ifa->ifa_broadaddr &&
1120 			    /* IP6 doesn't have broadcast */
1121 			    ifa->ifa_broadaddr->sa_len != 0 &&
1122 			    equal(ifa->ifa_broadaddr, addr))
1123 				return ifa;
1124 		}
1125 	}
1126 	return NULL;
1127 }
1128 
1129 /*
1130  * Locate the point to point interface with a given destination address.
1131  */
1132 /*ARGSUSED*/
1133 struct ifaddr *
1134 ifa_ifwithdstaddr(const struct sockaddr *addr)
1135 {
1136 	struct ifnet *ifp;
1137 	struct ifaddr *ifa;
1138 
1139 	IFNET_FOREACH(ifp) {
1140 		if (ifp->if_output == if_nulloutput)
1141 			continue;
1142 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1143 			continue;
1144 		IFADDR_FOREACH(ifa, ifp) {
1145 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1146 			    ifa->ifa_dstaddr == NULL)
1147 				continue;
1148 			if (equal(addr, ifa->ifa_dstaddr))
1149 				return ifa;
1150 		}
1151 	}
1152 	return NULL;
1153 }
1154 
1155 /*
1156  * Find an interface on a specific network.  If many, choice
1157  * is most specific found.
1158  */
1159 struct ifaddr *
1160 ifa_ifwithnet(const struct sockaddr *addr)
1161 {
1162 	struct ifnet *ifp;
1163 	struct ifaddr *ifa;
1164 	const struct sockaddr_dl *sdl;
1165 	struct ifaddr *ifa_maybe = 0;
1166 	u_int af = addr->sa_family;
1167 	const char *addr_data = addr->sa_data, *cplim;
1168 
1169 	if (af == AF_LINK) {
1170 		sdl = satocsdl(addr);
1171 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1172 		    ifindex2ifnet[sdl->sdl_index] &&
1173 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1174 			return ifnet_addrs[sdl->sdl_index];
1175 	}
1176 #ifdef NETATALK
1177 	if (af == AF_APPLETALK) {
1178 		const struct sockaddr_at *sat, *sat2;
1179 		sat = (const struct sockaddr_at *)addr;
1180 		IFNET_FOREACH(ifp) {
1181 			if (ifp->if_output == if_nulloutput)
1182 				continue;
1183 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1184 			if (ifa == NULL)
1185 				continue;
1186 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1187 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1188 				return ifa; /* exact match */
1189 			if (ifa_maybe == NULL) {
1190 				/* else keep the if with the right range */
1191 				ifa_maybe = ifa;
1192 			}
1193 		}
1194 		return ifa_maybe;
1195 	}
1196 #endif
1197 	IFNET_FOREACH(ifp) {
1198 		if (ifp->if_output == if_nulloutput)
1199 			continue;
1200 		IFADDR_FOREACH(ifa, ifp) {
1201 			const char *cp, *cp2, *cp3;
1202 
1203 			if (ifa->ifa_addr->sa_family != af ||
1204 			    ifa->ifa_netmask == NULL)
1205  next:				continue;
1206 			cp = addr_data;
1207 			cp2 = ifa->ifa_addr->sa_data;
1208 			cp3 = ifa->ifa_netmask->sa_data;
1209 			cplim = (const char *)ifa->ifa_netmask +
1210 			    ifa->ifa_netmask->sa_len;
1211 			while (cp3 < cplim) {
1212 				if ((*cp++ ^ *cp2++) & *cp3++) {
1213 					/* want to continue for() loop */
1214 					goto next;
1215 				}
1216 			}
1217 			if (ifa_maybe == NULL ||
1218 			    rn_refines((void *)ifa->ifa_netmask,
1219 			    (void *)ifa_maybe->ifa_netmask))
1220 				ifa_maybe = ifa;
1221 		}
1222 	}
1223 	return ifa_maybe;
1224 }
1225 
1226 /*
1227  * Find the interface of the addresss.
1228  */
1229 struct ifaddr *
1230 ifa_ifwithladdr(const struct sockaddr *addr)
1231 {
1232 	struct ifaddr *ia;
1233 
1234 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1235 	    (ia = ifa_ifwithnet(addr)))
1236 		return ia;
1237 	return NULL;
1238 }
1239 
1240 /*
1241  * Find an interface using a specific address family
1242  */
1243 struct ifaddr *
1244 ifa_ifwithaf(int af)
1245 {
1246 	struct ifnet *ifp;
1247 	struct ifaddr *ifa;
1248 
1249 	IFNET_FOREACH(ifp) {
1250 		if (ifp->if_output == if_nulloutput)
1251 			continue;
1252 		IFADDR_FOREACH(ifa, ifp) {
1253 			if (ifa->ifa_addr->sa_family == af)
1254 				return ifa;
1255 		}
1256 	}
1257 	return NULL;
1258 }
1259 
1260 /*
1261  * Find an interface address specific to an interface best matching
1262  * a given address.
1263  */
1264 struct ifaddr *
1265 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1266 {
1267 	struct ifaddr *ifa;
1268 	const char *cp, *cp2, *cp3;
1269 	const char *cplim;
1270 	struct ifaddr *ifa_maybe = 0;
1271 	u_int af = addr->sa_family;
1272 
1273 	if (ifp->if_output == if_nulloutput)
1274 		return NULL;
1275 
1276 	if (af >= AF_MAX)
1277 		return NULL;
1278 
1279 	IFADDR_FOREACH(ifa, ifp) {
1280 		if (ifa->ifa_addr->sa_family != af)
1281 			continue;
1282 		ifa_maybe = ifa;
1283 		if (ifa->ifa_netmask == NULL) {
1284 			if (equal(addr, ifa->ifa_addr) ||
1285 			    (ifa->ifa_dstaddr &&
1286 			     equal(addr, ifa->ifa_dstaddr)))
1287 				return ifa;
1288 			continue;
1289 		}
1290 		cp = addr->sa_data;
1291 		cp2 = ifa->ifa_addr->sa_data;
1292 		cp3 = ifa->ifa_netmask->sa_data;
1293 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1294 		for (; cp3 < cplim; cp3++) {
1295 			if ((*cp++ ^ *cp2++) & *cp3)
1296 				break;
1297 		}
1298 		if (cp3 == cplim)
1299 			return ifa;
1300 	}
1301 	return ifa_maybe;
1302 }
1303 
1304 /*
1305  * Default action when installing a route with a Link Level gateway.
1306  * Lookup an appropriate real ifa to point to.
1307  * This should be moved to /sys/net/link.c eventually.
1308  */
1309 void
1310 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1311 {
1312 	struct ifaddr *ifa;
1313 	const struct sockaddr *dst;
1314 	struct ifnet *ifp;
1315 
1316 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1317 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1318 		return;
1319 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1320 		rt_replace_ifa(rt, ifa);
1321 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1322 			ifa->ifa_rtrequest(cmd, rt, info);
1323 	}
1324 }
1325 
1326 /*
1327  * Handle a change in the interface link state.
1328  */
1329 void
1330 if_link_state_change(struct ifnet *ifp, int link_state)
1331 {
1332 	if (ifp->if_link_state == link_state)
1333 		return;
1334 	ifp->if_link_state = link_state;
1335 	/* Notify that the link state has changed. */
1336 	rt_ifmsg(ifp);
1337 #if NCARP > 0
1338 	if (ifp->if_carp)
1339 		carp_carpdev_state(ifp);
1340 #endif
1341 }
1342 
1343 /*
1344  * Mark an interface down and notify protocols of
1345  * the transition.
1346  * NOTE: must be called at splsoftnet or equivalent.
1347  */
1348 void
1349 if_down(struct ifnet *ifp)
1350 {
1351 	struct ifaddr *ifa;
1352 
1353 	ifp->if_flags &= ~IFF_UP;
1354 	nanotime(&ifp->if_lastchange);
1355 	IFADDR_FOREACH(ifa, ifp)
1356 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1357 	IFQ_PURGE(&ifp->if_snd);
1358 #if NCARP > 0
1359 	if (ifp->if_carp)
1360 		carp_carpdev_state(ifp);
1361 #endif
1362 	rt_ifmsg(ifp);
1363 }
1364 
1365 /*
1366  * Mark an interface up and notify protocols of
1367  * the transition.
1368  * NOTE: must be called at splsoftnet or equivalent.
1369  */
1370 void
1371 if_up(struct ifnet *ifp)
1372 {
1373 #ifdef notyet
1374 	struct ifaddr *ifa;
1375 #endif
1376 
1377 	ifp->if_flags |= IFF_UP;
1378 	nanotime(&ifp->if_lastchange);
1379 #ifdef notyet
1380 	/* this has no effect on IP, and will kill all ISO connections XXX */
1381 	IFADDR_FOREACH(ifa, ifp)
1382 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1383 #endif
1384 #if NCARP > 0
1385 	if (ifp->if_carp)
1386 		carp_carpdev_state(ifp);
1387 #endif
1388 	rt_ifmsg(ifp);
1389 #ifdef INET6
1390 	in6_if_up(ifp);
1391 #endif
1392 }
1393 
1394 /*
1395  * Handle interface watchdog timer routines.  Called
1396  * from softclock, we decrement timers (if set) and
1397  * call the appropriate interface routine on expiration.
1398  */
1399 void
1400 if_slowtimo(void *arg)
1401 {
1402 	struct ifnet *ifp;
1403 	int s = splnet();
1404 
1405 	IFNET_FOREACH(ifp) {
1406 		if (ifp->if_timer == 0 || --ifp->if_timer)
1407 			continue;
1408 		if (ifp->if_watchdog != NULL)
1409 			(*ifp->if_watchdog)(ifp);
1410 	}
1411 	splx(s);
1412 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1413 }
1414 
1415 /*
1416  * Set/clear promiscuous mode on interface ifp based on the truth value
1417  * of pswitch.  The calls are reference counted so that only the first
1418  * "on" request actually has an effect, as does the final "off" request.
1419  * Results are undefined if the "off" and "on" requests are not matched.
1420  */
1421 int
1422 ifpromisc(struct ifnet *ifp, int pswitch)
1423 {
1424 	int pcount, ret;
1425 	short nflags;
1426 
1427 	pcount = ifp->if_pcount;
1428 	if (pswitch) {
1429 		/*
1430 		 * Allow the device to be "placed" into promiscuous
1431 		 * mode even if it is not configured up.  It will
1432 		 * consult IFF_PROMISC when it is brought up.
1433 		 */
1434 		if (ifp->if_pcount++ != 0)
1435 			return 0;
1436 		nflags = ifp->if_flags | IFF_PROMISC;
1437 	} else {
1438 		if (--ifp->if_pcount > 0)
1439 			return 0;
1440 		nflags = ifp->if_flags & ~IFF_PROMISC;
1441 	}
1442 	ret = if_flags_set(ifp, nflags);
1443 	/* Restore interface state if not successful. */
1444 	if (ret != 0) {
1445 		ifp->if_pcount = pcount;
1446 	}
1447 	return ret;
1448 }
1449 
1450 /*
1451  * Map interface name to
1452  * interface structure pointer.
1453  */
1454 struct ifnet *
1455 ifunit(const char *name)
1456 {
1457 	struct ifnet *ifp;
1458 	const char *cp = name;
1459 	u_int unit = 0;
1460 	u_int i;
1461 
1462 	/*
1463 	 * If the entire name is a number, treat it as an ifindex.
1464 	 */
1465 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1466 		unit = unit * 10 + (*cp - '0');
1467 	}
1468 
1469 	/*
1470 	 * If the number took all of the name, then it's a valid ifindex.
1471 	 */
1472 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1473 		if (unit >= if_indexlim)
1474 			return NULL;
1475 		ifp = ifindex2ifnet[unit];
1476 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1477 			return NULL;
1478 		return ifp;
1479 	}
1480 
1481 	IFNET_FOREACH(ifp) {
1482 		if (ifp->if_output == if_nulloutput)
1483 			continue;
1484 	 	if (strcmp(ifp->if_xname, name) == 0)
1485 			return ifp;
1486 	}
1487 	return NULL;
1488 }
1489 
1490 ifnet_t *
1491 if_byindex(u_int idx)
1492 {
1493 
1494 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1495 }
1496 
1497 /* common */
1498 int
1499 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1500 {
1501 	int s;
1502 	struct ifreq *ifr;
1503 	struct ifcapreq *ifcr;
1504 	struct ifdatareq *ifdr;
1505 
1506 	switch (cmd) {
1507 	case SIOCSIFCAP:
1508 		ifcr = data;
1509 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1510 			return EINVAL;
1511 
1512 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1513 			return 0;
1514 
1515 		ifp->if_capenable = ifcr->ifcr_capenable;
1516 
1517 		/* Pre-compute the checksum flags mask. */
1518 		ifp->if_csum_flags_tx = 0;
1519 		ifp->if_csum_flags_rx = 0;
1520 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1521 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1522 		}
1523 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1524 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1525 		}
1526 
1527 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1528 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1529 		}
1530 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1531 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1532 		}
1533 
1534 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1535 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1536 		}
1537 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1538 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1539 		}
1540 
1541 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1542 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1543 		}
1544 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1545 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1546 		}
1547 
1548 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1549 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1550 		}
1551 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1552 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1553 		}
1554 		if (ifp->if_flags & IFF_UP)
1555 			return ENETRESET;
1556 		return 0;
1557 	case SIOCSIFFLAGS:
1558 		ifr = data;
1559 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1560 			s = splnet();
1561 			if_down(ifp);
1562 			splx(s);
1563 		}
1564 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1565 			s = splnet();
1566 			if_up(ifp);
1567 			splx(s);
1568 		}
1569 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1570 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1571 		break;
1572 	case SIOCGIFFLAGS:
1573 		ifr = data;
1574 		ifr->ifr_flags = ifp->if_flags;
1575 		break;
1576 
1577 	case SIOCGIFMETRIC:
1578 		ifr = data;
1579 		ifr->ifr_metric = ifp->if_metric;
1580 		break;
1581 
1582 	case SIOCGIFMTU:
1583 		ifr = data;
1584 		ifr->ifr_mtu = ifp->if_mtu;
1585 		break;
1586 
1587 	case SIOCGIFDLT:
1588 		ifr = data;
1589 		ifr->ifr_dlt = ifp->if_dlt;
1590 		break;
1591 
1592 	case SIOCGIFCAP:
1593 		ifcr = data;
1594 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1595 		ifcr->ifcr_capenable = ifp->if_capenable;
1596 		break;
1597 
1598 	case SIOCSIFMETRIC:
1599 		ifr = data;
1600 		ifp->if_metric = ifr->ifr_metric;
1601 		break;
1602 
1603 	case SIOCGIFDATA:
1604 		ifdr = data;
1605 		ifdr->ifdr_data = ifp->if_data;
1606 		break;
1607 
1608 	case SIOCZIFDATA:
1609 		ifdr = data;
1610 		ifdr->ifdr_data = ifp->if_data;
1611 		/*
1612 		 * Assumes that the volatile counters that can be
1613 		 * zero'ed are at the end of if_data.
1614 		 */
1615 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1616 		    offsetof(struct if_data, ifi_ipackets));
1617 		/*
1618 		 * The memset() clears to the bottm of if_data. In the area,
1619 		 * if_lastchange is included. Please be careful if new entry
1620 		 * will be added into if_data or rewite this.
1621 		 *
1622 		 * And also, update if_lastchnage.
1623 		 */
1624 		getnanotime(&ifp->if_lastchange);
1625 		break;
1626 	case SIOCSIFMTU:
1627 		ifr = data;
1628 		if (ifp->if_mtu == ifr->ifr_mtu)
1629 			break;
1630 		ifp->if_mtu = ifr->ifr_mtu;
1631 		/*
1632 		 * If the link MTU changed, do network layer specific procedure.
1633 		 */
1634 #ifdef INET6
1635 		nd6_setmtu(ifp);
1636 #endif
1637 		return ENETRESET;
1638 	default:
1639 		return ENOTTY;
1640 	}
1641 	return 0;
1642 }
1643 
1644 int
1645 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
1646     lwp_t *l)
1647 {
1648 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1649 	struct ifaddr *ifa;
1650 	const struct sockaddr *any, *sa;
1651 	union {
1652 		struct sockaddr sa;
1653 		struct sockaddr_storage ss;
1654 	} u, v;
1655 
1656 	switch (cmd) {
1657 	case SIOCSIFADDRPREF:
1658 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
1659 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1660 		    NULL) != 0)
1661 			return EPERM;
1662 	case SIOCGIFADDRPREF:
1663 		break;
1664 	default:
1665 		return EOPNOTSUPP;
1666 	}
1667 
1668 	/* sanity checks */
1669 	if (data == NULL || ifp == NULL) {
1670 		panic("invalid argument to %s", __func__);
1671 		/*NOTREACHED*/
1672 	}
1673 
1674 	/* address must be specified on ADD and DELETE */
1675 	sa = sstocsa(&ifap->ifap_addr);
1676 	if (sa->sa_family != sofamily(so))
1677 		return EINVAL;
1678 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1679 		return EINVAL;
1680 
1681 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1682 
1683 	IFADDR_FOREACH(ifa, ifp) {
1684 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1685 			continue;
1686 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1687 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1688 			break;
1689 	}
1690 	if (ifa == NULL)
1691 		return EADDRNOTAVAIL;
1692 
1693 	switch (cmd) {
1694 	case SIOCSIFADDRPREF:
1695 		ifa->ifa_preference = ifap->ifap_preference;
1696 		return 0;
1697 	case SIOCGIFADDRPREF:
1698 		/* fill in the if_laddrreq structure */
1699 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1700 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1701 		ifap->ifap_preference = ifa->ifa_preference;
1702 		return 0;
1703 	default:
1704 		return EOPNOTSUPP;
1705 	}
1706 }
1707 
1708 static void
1709 ifnet_lock_enter(struct ifnet_lock *il)
1710 {
1711 	uint64_t *nenter;
1712 
1713 	/* Before trying to acquire the mutex, increase the count of threads
1714 	 * who have entered or who wait to enter the critical section.
1715 	 * Avoid one costly locked memory transaction by keeping a count for
1716 	 * each CPU.
1717 	 */
1718 	nenter = percpu_getref(il->il_nenter);
1719 	(*nenter)++;
1720 	percpu_putref(il->il_nenter);
1721 	mutex_enter(&il->il_lock);
1722 }
1723 
1724 static void
1725 ifnet_lock_exit(struct ifnet_lock *il)
1726 {
1727 	/* Increase the count of threads who have exited the critical
1728 	 * section.  Increase while we still hold the lock.
1729 	 */
1730 	il->il_nexit++;
1731 	mutex_exit(&il->il_lock);
1732 }
1733 
1734 /*
1735  * Interface ioctls.
1736  */
1737 int
1738 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1739 {
1740 	struct ifnet *ifp;
1741 	struct ifreq *ifr;
1742 	int error = 0;
1743 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1744 	u_long ocmd = cmd;
1745 #endif
1746 	short oif_flags;
1747 #ifdef COMPAT_OIFREQ
1748 	struct ifreq ifrb;
1749 	struct oifreq *oifr = NULL;
1750 #endif
1751 
1752 	switch (cmd) {
1753 #ifdef COMPAT_OIFREQ
1754 	case OSIOCGIFCONF:
1755 	case OOSIOCGIFCONF:
1756 		return compat_ifconf(cmd, data);
1757 #endif
1758 #ifdef COMPAT_OIFDATA
1759 	case OSIOCGIFDATA:
1760 	case OSIOCZIFDATA:
1761 		return compat_ifdatareq(l, cmd, data);
1762 #endif
1763 	case SIOCGIFCONF:
1764 		return ifconf(cmd, data);
1765 	case SIOCINITIFADDR:
1766 		return EPERM;
1767 	}
1768 
1769 #ifdef COMPAT_OIFREQ
1770 	cmd = compat_cvtcmd(cmd);
1771 	if (cmd != ocmd) {
1772 		oifr = data;
1773 		data = ifr = &ifrb;
1774 		ifreqo2n(oifr, ifr);
1775 	} else
1776 #endif
1777 		ifr = data;
1778 
1779 	ifp = ifunit(ifr->ifr_name);
1780 
1781 	switch (cmd) {
1782 	case SIOCIFCREATE:
1783 	case SIOCIFDESTROY:
1784 		if (l != NULL) {
1785 			error = kauth_authorize_network(l->l_cred,
1786 			    KAUTH_NETWORK_INTERFACE,
1787 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1788 			    (void *)cmd, NULL);
1789 			if (error != 0)
1790 				return error;
1791 		}
1792 		return (cmd == SIOCIFCREATE) ?
1793 			if_clone_create(ifr->ifr_name) :
1794 			if_clone_destroy(ifr->ifr_name);
1795 
1796 	case SIOCIFGCLONERS:
1797 		return if_clone_list((struct if_clonereq *)data);
1798 	}
1799 
1800 	if (ifp == NULL)
1801 		return ENXIO;
1802 
1803 	switch (cmd) {
1804 	case SIOCALIFADDR:
1805 	case SIOCDLIFADDR:
1806 	case SIOCSIFADDRPREF:
1807 	case SIOCSIFFLAGS:
1808 	case SIOCSIFCAP:
1809 	case SIOCSIFMETRIC:
1810 	case SIOCZIFDATA:
1811 	case SIOCSIFMTU:
1812 	case SIOCSIFPHYADDR:
1813 	case SIOCDIFPHYADDR:
1814 #ifdef INET6
1815 	case SIOCSIFPHYADDR_IN6:
1816 #endif
1817 	case SIOCSLIFPHYADDR:
1818 	case SIOCADDMULTI:
1819 	case SIOCDELMULTI:
1820 	case SIOCSIFMEDIA:
1821 	case SIOCSDRVSPEC:
1822 	case SIOCG80211:
1823 	case SIOCS80211:
1824 	case SIOCS80211NWID:
1825 	case SIOCS80211NWKEY:
1826 	case SIOCS80211POWER:
1827 	case SIOCS80211BSSID:
1828 	case SIOCS80211CHANNEL:
1829 	case SIOCSLINKSTR:
1830 		if (l != NULL) {
1831 			error = kauth_authorize_network(l->l_cred,
1832 			    KAUTH_NETWORK_INTERFACE,
1833 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1834 			    (void *)cmd, NULL);
1835 			if (error != 0)
1836 				return error;
1837 		}
1838 	}
1839 
1840 	oif_flags = ifp->if_flags;
1841 
1842 	ifnet_lock_enter(ifp->if_ioctl_lock);
1843 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1844 	if (error != ENOTTY)
1845 		;
1846 	else if (so->so_proto == NULL)
1847 		error = EOPNOTSUPP;
1848 	else {
1849 #ifdef COMPAT_OSOCK
1850 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1851 #else
1852 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1853 		    (struct mbuf *)cmd, (struct mbuf *)data,
1854 		    (struct mbuf *)ifp, l);
1855 #endif
1856 	}
1857 
1858 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1859 #ifdef INET6
1860 		if ((ifp->if_flags & IFF_UP) != 0) {
1861 			int s = splnet();
1862 			in6_if_up(ifp);
1863 			splx(s);
1864 		}
1865 #endif
1866 	}
1867 #ifdef COMPAT_OIFREQ
1868 	if (cmd != ocmd)
1869 		ifreqn2o(oifr, ifr);
1870 #endif
1871 
1872 	ifnet_lock_exit(ifp->if_ioctl_lock);
1873 	return error;
1874 }
1875 
1876 /* This callback adds to the sum in `arg' the number of
1877  * threads on `ci' who have entered or who wait to enter the
1878  * critical section.
1879  */
1880 static void
1881 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
1882 {
1883 	uint64_t *sum = arg, *nenter = p;
1884 
1885 	*sum += *nenter;
1886 }
1887 
1888 /* Return the number of threads who have entered or who wait
1889  * to enter the critical section on all CPUs.
1890  */
1891 static uint64_t
1892 ifnet_lock_entrances(struct ifnet_lock *il)
1893 {
1894 	uint64_t sum = 0;
1895 
1896 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
1897 
1898 	return sum;
1899 }
1900 
1901 static int
1902 ifioctl_attach(struct ifnet *ifp)
1903 {
1904 	struct ifnet_lock *il;
1905 
1906 	/* If the driver has not supplied its own if_ioctl, then
1907 	 * supply the default.
1908 	 */
1909 	if (ifp->if_ioctl == NULL)
1910 		ifp->if_ioctl = ifioctl_common;
1911 
1912 	/* Create an ifnet_lock for synchronizing ifioctls. */
1913 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
1914 		return ENOMEM;
1915 
1916 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
1917 	if (il->il_nenter == NULL) {
1918 		kmem_free(il, sizeof(*il));
1919 		return ENOMEM;
1920 	}
1921 
1922 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
1923 	cv_init(&il->il_emptied, ifp->if_xname);
1924 
1925 	ifp->if_ioctl_lock = il;
1926 
1927 	return 0;
1928 }
1929 
1930 /*
1931  * This must not be called until after `ifp' has been withdrawn from the
1932  * ifnet tables so that ifioctl() cannot get a handle on it by calling
1933  * ifunit().
1934  */
1935 static void
1936 ifioctl_detach(struct ifnet *ifp)
1937 {
1938 	struct ifnet_lock *il;
1939 
1940 	il = ifp->if_ioctl_lock;
1941 	mutex_enter(&il->il_lock);
1942 	/* Install if_nullioctl to make sure that any thread that
1943 	 * subsequently enters the critical section will quit it
1944 	 * immediately and signal the condition variable that we
1945 	 * wait on, below.
1946 	 */
1947 	ifp->if_ioctl = if_nullioctl;
1948 	/* Sleep while threads are still in the critical section or
1949 	 * wait to enter it.
1950 	 */
1951 	while (ifnet_lock_entrances(il) != il->il_nexit)
1952 		cv_wait(&il->il_emptied, &il->il_lock);
1953 	/* At this point, we are the only thread still in the critical
1954 	 * section, and no new thread can get a handle on the ifioctl
1955 	 * lock, so it is safe to free its memory.
1956 	 */
1957 	mutex_exit(&il->il_lock);
1958 	ifp->if_ioctl_lock = NULL;
1959 	percpu_free(il->il_nenter, sizeof(uint64_t));
1960 	il->il_nenter = NULL;
1961 	cv_destroy(&il->il_emptied);
1962 	mutex_destroy(&il->il_lock);
1963 	kmem_free(il, sizeof(*il));
1964 }
1965 
1966 /*
1967  * Return interface configuration
1968  * of system.  List may be used
1969  * in later ioctl's (above) to get
1970  * other information.
1971  *
1972  * Each record is a struct ifreq.  Before the addition of
1973  * sockaddr_storage, the API rule was that sockaddr flavors that did
1974  * not fit would extend beyond the struct ifreq, with the next struct
1975  * ifreq starting sa_len beyond the struct sockaddr.  Because the
1976  * union in struct ifreq includes struct sockaddr_storage, every kind
1977  * of sockaddr must fit.  Thus, there are no longer any overlength
1978  * records.
1979  *
1980  * Records are added to the user buffer if they fit, and ifc_len is
1981  * adjusted to the length that was written.  Thus, the user is only
1982  * assured of getting the complete list if ifc_len on return is at
1983  * least sizeof(struct ifreq) less than it was on entry.
1984  *
1985  * If the user buffer pointer is NULL, this routine copies no data and
1986  * returns the amount of space that would be needed.
1987  *
1988  * Invariants:
1989  * ifrp points to the next part of the user's buffer to be used.  If
1990  * ifrp != NULL, space holds the number of bytes remaining that we may
1991  * write at ifrp.  Otherwise, space holds the number of bytes that
1992  * would have been written had there been adequate space.
1993  */
1994 /*ARGSUSED*/
1995 int
1996 ifconf(u_long cmd, void *data)
1997 {
1998 	struct ifconf *ifc = (struct ifconf *)data;
1999 	struct ifnet *ifp;
2000 	struct ifaddr *ifa;
2001 	struct ifreq ifr, *ifrp;
2002 	int space, error = 0;
2003 	const int sz = (int)sizeof(struct ifreq);
2004 
2005 	if ((ifrp = ifc->ifc_req) == NULL)
2006 		space = 0;
2007 	else
2008 		space = ifc->ifc_len;
2009 	IFNET_FOREACH(ifp) {
2010 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
2011 		    sizeof(ifr.ifr_name));
2012 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
2013 			return ENAMETOOLONG;
2014 		if (IFADDR_EMPTY(ifp)) {
2015 			/* Interface with no addresses - send zero sockaddr. */
2016 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
2017 			if (ifrp == NULL) {
2018 				space += sz;
2019 				continue;
2020 			}
2021 			if (space >= sz) {
2022 				error = copyout(&ifr, ifrp, sz);
2023 				if (error != 0)
2024 					return error;
2025 				ifrp++;
2026 				space -= sz;
2027 			}
2028 		}
2029 
2030 		IFADDR_FOREACH(ifa, ifp) {
2031 			struct sockaddr *sa = ifa->ifa_addr;
2032 			/* all sockaddrs must fit in sockaddr_storage */
2033 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
2034 
2035 			if (ifrp == NULL) {
2036 				space += sz;
2037 				continue;
2038 			}
2039 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
2040 			if (space >= sz) {
2041 				error = copyout(&ifr, ifrp, sz);
2042 				if (error != 0)
2043 					return (error);
2044 				ifrp++; space -= sz;
2045 			}
2046 		}
2047 	}
2048 	if (ifrp != NULL) {
2049 		KASSERT(0 <= space && space <= ifc->ifc_len);
2050 		ifc->ifc_len -= space;
2051 	} else {
2052 		KASSERT(space >= 0);
2053 		ifc->ifc_len = space;
2054 	}
2055 	return (0);
2056 }
2057 
2058 int
2059 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
2060 {
2061 	uint8_t len;
2062 #ifdef COMPAT_OIFREQ
2063 	struct ifreq ifrb;
2064 	struct oifreq *oifr = NULL;
2065 	u_long ocmd = cmd;
2066 	cmd = compat_cvtcmd(cmd);
2067 	if (cmd != ocmd) {
2068 		oifr = (struct oifreq *)(void *)ifr;
2069 		ifr = &ifrb;
2070 		ifreqo2n(oifr, ifr);
2071 		len = sizeof(oifr->ifr_addr);
2072 	} else
2073 #endif
2074 		len = sizeof(ifr->ifr_ifru.ifru_space);
2075 
2076 	if (len < sa->sa_len)
2077 		return EFBIG;
2078 
2079 	memset(&ifr->ifr_addr, 0, len);
2080 	sockaddr_copy(&ifr->ifr_addr, len, sa);
2081 
2082 #ifdef COMPAT_OIFREQ
2083 	if (cmd != ocmd)
2084 		ifreqn2o(oifr, ifr);
2085 #endif
2086 	return 0;
2087 }
2088 
2089 /*
2090  * Queue message on interface, and start output if interface
2091  * not yet active.
2092  */
2093 int
2094 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
2095     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2096 {
2097 	int len = m->m_pkthdr.len;
2098 	int mflags = m->m_flags;
2099 	int s = splnet();
2100 	int error;
2101 
2102 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2103 	if (error != 0)
2104 		goto out;
2105 	ifp->if_obytes += len;
2106 	if (mflags & M_MCAST)
2107 		ifp->if_omcasts++;
2108 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
2109 		(*ifp->if_start)(ifp);
2110 out:
2111 	splx(s);
2112 	return error;
2113 }
2114 
2115 /*
2116  * Queue message on interface, possibly using a second fast queue
2117  */
2118 int
2119 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
2120     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
2121 {
2122 	int error = 0;
2123 
2124 	if (ifq != NULL
2125 #ifdef ALTQ
2126 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
2127 #endif
2128 	    ) {
2129 		if (IF_QFULL(ifq)) {
2130 			IF_DROP(&ifp->if_snd);
2131 			m_freem(m);
2132 			if (error == 0)
2133 				error = ENOBUFS;
2134 		} else
2135 			IF_ENQUEUE(ifq, m);
2136 	} else
2137 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2138 	if (error != 0) {
2139 		++ifp->if_oerrors;
2140 		return error;
2141 	}
2142 	return 0;
2143 }
2144 
2145 int
2146 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
2147 {
2148 	int rc;
2149 
2150 	if (ifp->if_initaddr != NULL)
2151 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
2152 	else if (src ||
2153 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
2154 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
2155 
2156 	return rc;
2157 }
2158 
2159 int
2160 if_flags_set(ifnet_t *ifp, const short flags)
2161 {
2162 	int rc;
2163 
2164 	if (ifp->if_setflags != NULL)
2165 		rc = (*ifp->if_setflags)(ifp, flags);
2166 	else {
2167 		short cantflags, chgdflags;
2168 		struct ifreq ifr;
2169 
2170 		chgdflags = ifp->if_flags ^ flags;
2171 		cantflags = chgdflags & IFF_CANTCHANGE;
2172 
2173 		if (cantflags != 0)
2174 			ifp->if_flags ^= cantflags;
2175 
2176                 /* Traditionally, we do not call if_ioctl after
2177                  * setting/clearing only IFF_PROMISC if the interface
2178                  * isn't IFF_UP.  Uphold that tradition.
2179 		 */
2180 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
2181 			return 0;
2182 
2183 		memset(&ifr, 0, sizeof(ifr));
2184 
2185 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
2186 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
2187 
2188 		if (rc != 0 && cantflags != 0)
2189 			ifp->if_flags ^= cantflags;
2190 	}
2191 
2192 	return rc;
2193 }
2194 
2195 int
2196 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
2197 {
2198 	int rc;
2199 	struct ifreq ifr;
2200 
2201 	if (ifp->if_mcastop != NULL)
2202 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
2203 	else {
2204 		ifreq_setaddr(cmd, &ifr, sa);
2205 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
2206 	}
2207 
2208 	return rc;
2209 }
2210 
2211 static void
2212 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2213     struct ifaltq *ifq)
2214 {
2215 	const struct sysctlnode *cnode, *rnode;
2216 
2217 	if (sysctl_createv(clog, 0, NULL, &rnode,
2218 		       CTLFLAG_PERMANENT,
2219 		       CTLTYPE_NODE, "net", NULL,
2220 		       NULL, 0, NULL, 0,
2221 		       CTL_NET, CTL_EOL) != 0)
2222 		goto bad;
2223 
2224 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2225 		       CTLFLAG_PERMANENT,
2226 		       CTLTYPE_NODE, "interfaces",
2227 		       SYSCTL_DESCR("Per-interface controls"),
2228 		       NULL, 0, NULL, 0,
2229 		       CTL_CREATE, CTL_EOL) != 0)
2230 		goto bad;
2231 
2232 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2233 		       CTLFLAG_PERMANENT,
2234 		       CTLTYPE_NODE, ifname,
2235 		       SYSCTL_DESCR("Interface controls"),
2236 		       NULL, 0, NULL, 0,
2237 		       CTL_CREATE, CTL_EOL) != 0)
2238 		goto bad;
2239 
2240 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2241 		       CTLFLAG_PERMANENT,
2242 		       CTLTYPE_NODE, "sndq",
2243 		       SYSCTL_DESCR("Interface output queue controls"),
2244 		       NULL, 0, NULL, 0,
2245 		       CTL_CREATE, CTL_EOL) != 0)
2246 		goto bad;
2247 
2248 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2249 		       CTLFLAG_PERMANENT,
2250 		       CTLTYPE_INT, "len",
2251 		       SYSCTL_DESCR("Current output queue length"),
2252 		       NULL, 0, &ifq->ifq_len, 0,
2253 		       CTL_CREATE, CTL_EOL) != 0)
2254 		goto bad;
2255 
2256 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2257 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2258 		       CTLTYPE_INT, "maxlen",
2259 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2260 		       NULL, 0, &ifq->ifq_maxlen, 0,
2261 		       CTL_CREATE, CTL_EOL) != 0)
2262 		goto bad;
2263 
2264 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2265 		       CTLFLAG_PERMANENT,
2266 		       CTLTYPE_INT, "drops",
2267 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2268 		       NULL, 0, &ifq->ifq_drops, 0,
2269 		       CTL_CREATE, CTL_EOL) != 0)
2270 		goto bad;
2271 
2272 	return;
2273 bad:
2274 	printf("%s: could not attach sysctl nodes\n", ifname);
2275 	return;
2276 }
2277 
2278 #if defined(INET) || defined(INET6)
2279 static void
2280 sysctl_net_ifq_setup(struct sysctllog **clog,
2281 		     int pf, const char *pfname,
2282 		     int ipn, const char *ipname,
2283 		     int qid, struct ifqueue *ifq)
2284 {
2285 
2286 	sysctl_createv(clog, 0, NULL, NULL,
2287 		       CTLFLAG_PERMANENT,
2288 		       CTLTYPE_NODE, "net", NULL,
2289 		       NULL, 0, NULL, 0,
2290 		       CTL_NET, CTL_EOL);
2291 	sysctl_createv(clog, 0, NULL, NULL,
2292 		       CTLFLAG_PERMANENT,
2293 		       CTLTYPE_NODE, pfname, NULL,
2294 		       NULL, 0, NULL, 0,
2295 		       CTL_NET, pf, CTL_EOL);
2296 	sysctl_createv(clog, 0, NULL, NULL,
2297 		       CTLFLAG_PERMANENT,
2298 		       CTLTYPE_NODE, ipname, NULL,
2299 		       NULL, 0, NULL, 0,
2300 		       CTL_NET, pf, ipn, CTL_EOL);
2301 	sysctl_createv(clog, 0, NULL, NULL,
2302 		       CTLFLAG_PERMANENT,
2303 		       CTLTYPE_NODE, "ifq",
2304 		       SYSCTL_DESCR("Protocol input queue controls"),
2305 		       NULL, 0, NULL, 0,
2306 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2307 
2308 	sysctl_createv(clog, 0, NULL, NULL,
2309 		       CTLFLAG_PERMANENT,
2310 		       CTLTYPE_INT, "len",
2311 		       SYSCTL_DESCR("Current input queue length"),
2312 		       NULL, 0, &ifq->ifq_len, 0,
2313 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2314 	sysctl_createv(clog, 0, NULL, NULL,
2315 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2316 		       CTLTYPE_INT, "maxlen",
2317 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2318 		       NULL, 0, &ifq->ifq_maxlen, 0,
2319 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2320 #ifdef notyet
2321 	sysctl_createv(clog, 0, NULL, NULL,
2322 		       CTLFLAG_PERMANENT,
2323 		       CTLTYPE_INT, "peak",
2324 		       SYSCTL_DESCR("Highest input queue length"),
2325 		       NULL, 0, &ifq->ifq_peak, 0,
2326 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
2327 #endif
2328 	sysctl_createv(clog, 0, NULL, NULL,
2329 		       CTLFLAG_PERMANENT,
2330 		       CTLTYPE_INT, "drops",
2331 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2332 		       NULL, 0, &ifq->ifq_drops, 0,
2333 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2334 }
2335 #endif /* INET || INET6 */
2336