xref: /netbsd-src/sys/net/if.c (revision 9aa0541bdf64142d9a27c2cf274394d60182818f)
1 /*	$NetBSD: if.c,v 1.251 2011/08/12 22:09:36 dyoung 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 1.251 2011/08/12 22:09:36 dyoung 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 
116 #include <net/if.h>
117 #include <net/if_dl.h>
118 #include <net/if_ether.h>
119 #include <net/if_media.h>
120 #include <net80211/ieee80211.h>
121 #include <net80211/ieee80211_ioctl.h>
122 #include <net/if_types.h>
123 #include <net/radix.h>
124 #include <net/route.h>
125 #include <net/netisr.h>
126 #ifdef NETATALK
127 #include <netatalk/at_extern.h>
128 #include <netatalk/at.h>
129 #endif
130 #include <net/pfil.h>
131 
132 #ifdef INET6
133 #include <netinet/in.h>
134 #include <netinet6/in6_var.h>
135 #include <netinet6/nd6.h>
136 #endif
137 
138 #include "carp.h"
139 #if NCARP > 0
140 #include <netinet/ip_carp.h>
141 #endif
142 
143 #include <compat/sys/sockio.h>
144 #include <compat/sys/socket.h>
145 
146 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
147 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
148 
149 int	ifqmaxlen = IFQ_MAXLEN;
150 callout_t if_slowtimo_ch;
151 
152 int netisr;			/* scheduling bits for network */
153 
154 static int	if_rt_walktree(struct rtentry *, void *);
155 
156 static struct if_clone *if_clone_lookup(const char *, int *);
157 static int	if_clone_list(struct if_clonereq *);
158 
159 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
160 static int if_cloners_count;
161 
162 static uint64_t index_gen;
163 static kmutex_t index_gen_mtx;
164 
165 #ifdef PFIL_HOOKS
166 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
167 #endif
168 
169 static kauth_listener_t if_listener;
170 
171 static void if_detach_queues(struct ifnet *, struct ifqueue *);
172 static void sysctl_sndq_setup(struct sysctllog **, const char *,
173     struct ifaltq *);
174 
175 #if defined(INET) || defined(INET6)
176 static void sysctl_net_ifq_setup(struct sysctllog **, int, const char *,
177 				 int, const char *, int, struct ifqueue *);
178 #endif
179 
180 static int
181 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
182     void *arg0, void *arg1, void *arg2, void *arg3)
183 {
184 	int result;
185 	enum kauth_network_req req;
186 
187 	result = KAUTH_RESULT_DEFER;
188 	req = (enum kauth_network_req)arg1;
189 
190 	if (action != KAUTH_NETWORK_INTERFACE)
191 		return result;
192 
193 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
194 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
195 		result = KAUTH_RESULT_ALLOW;
196 
197 	return result;
198 }
199 
200 /*
201  * Network interface utility routines.
202  *
203  * Routines with ifa_ifwith* names take sockaddr *'s as
204  * parameters.
205  */
206 void
207 ifinit(void)
208 {
209 #ifdef INET
210 	{extern struct ifqueue ipintrq;
211 	sysctl_net_ifq_setup(NULL, PF_INET, "inet", IPPROTO_IP, "ip",
212 			     IPCTL_IFQ, &ipintrq);}
213 #endif /* INET */
214 #ifdef INET6
215 	{extern struct ifqueue ip6intrq;
216 	sysctl_net_ifq_setup(NULL, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
217 			     IPV6CTL_IFQ, &ip6intrq);}
218 #endif /* INET6 */
219 
220 	callout_init(&if_slowtimo_ch, 0);
221 	if_slowtimo(NULL);
222 
223 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
224 	    if_listener_cb, NULL);
225 }
226 
227 /*
228  * XXX Initialization before configure().
229  * XXX hack to get pfil_add_hook working in autoconf.
230  */
231 void
232 ifinit1(void)
233 {
234 
235 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
236 
237 #ifdef PFIL_HOOKS
238 	if_pfil.ph_type = PFIL_TYPE_IFNET;
239 	if_pfil.ph_ifnet = NULL;
240 	if (pfil_head_register(&if_pfil) != 0)
241 		printf("WARNING: unable to register pfil hook\n");
242 #endif
243 }
244 
245 struct ifnet *
246 if_alloc(u_char type)
247 {
248 	return malloc(sizeof(struct ifnet), M_DEVBUF, M_WAITOK|M_ZERO);
249 }
250 
251 void
252 if_free(struct ifnet *ifp)
253 {
254 	free(ifp, M_DEVBUF);
255 }
256 
257 void
258 if_initname(struct ifnet *ifp, const char *name, int unit)
259 {
260 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
261 	    "%s%d", name, unit);
262 }
263 
264 /*
265  * Null routines used while an interface is going away.  These routines
266  * just return an error.
267  */
268 
269 int
270 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
271     const struct sockaddr *so, struct rtentry *rt)
272 {
273 
274 	return ENXIO;
275 }
276 
277 void
278 if_nullinput(struct ifnet *ifp, struct mbuf *m)
279 {
280 
281 	/* Nothing. */
282 }
283 
284 void
285 if_nullstart(struct ifnet *ifp)
286 {
287 
288 	/* Nothing. */
289 }
290 
291 int
292 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
293 {
294 
295 	return ENXIO;
296 }
297 
298 int
299 if_nullinit(struct ifnet *ifp)
300 {
301 
302 	return ENXIO;
303 }
304 
305 void
306 if_nullstop(struct ifnet *ifp, int disable)
307 {
308 
309 	/* Nothing. */
310 }
311 
312 void
313 if_nullwatchdog(struct ifnet *ifp)
314 {
315 
316 	/* Nothing. */
317 }
318 
319 void
320 if_nulldrain(struct ifnet *ifp)
321 {
322 
323 	/* Nothing. */
324 }
325 
326 static u_int if_index = 1;
327 struct ifnet_head ifnet;
328 size_t if_indexlim = 0;
329 struct ifaddr **ifnet_addrs = NULL;
330 struct ifnet **ifindex2ifnet = NULL;
331 struct ifnet *lo0ifp;
332 
333 void
334 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
335 {
336 	struct ifaddr *ifa;
337 	struct sockaddr_dl *sdl;
338 
339 	ifp->if_addrlen = addrlen;
340 	if_alloc_sadl(ifp);
341 	ifa = ifp->if_dl;
342 	sdl = satosdl(ifa->ifa_addr);
343 
344 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
345 	if (factory) {
346 		ifp->if_hwdl = ifp->if_dl;
347 		IFAREF(ifp->if_hwdl);
348 	}
349 	/* TBD routing socket */
350 }
351 
352 struct ifaddr *
353 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
354 {
355 	unsigned socksize, ifasize;
356 	int addrlen, namelen;
357 	struct sockaddr_dl *mask, *sdl;
358 	struct ifaddr *ifa;
359 
360 	namelen = strlen(ifp->if_xname);
361 	addrlen = ifp->if_addrlen;
362 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
363 	ifasize = sizeof(*ifa) + 2 * socksize;
364 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
365 
366 	sdl = (struct sockaddr_dl *)(ifa + 1);
367 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
368 
369 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
370 	    ifp->if_xname, namelen, NULL, addrlen);
371 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
372 	memset(&mask->sdl_data[0], 0xff, namelen);
373 	ifa->ifa_rtrequest = link_rtrequest;
374 	ifa->ifa_addr = (struct sockaddr *)sdl;
375 	ifa->ifa_netmask = (struct sockaddr *)mask;
376 
377 	*sdlp = sdl;
378 
379 	return ifa;
380 }
381 
382 static void
383 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
384 {
385 	const struct sockaddr_dl *sdl;
386 	ifnet_addrs[ifp->if_index] = ifa;
387 	IFAREF(ifa);
388 	ifp->if_dl = ifa;
389 	IFAREF(ifa);
390 	sdl = satosdl(ifa->ifa_addr);
391 	ifp->if_sadl = sdl;
392 }
393 
394 /*
395  * Allocate the link level name for the specified interface.  This
396  * is an attachment helper.  It must be called after ifp->if_addrlen
397  * is initialized, which may not be the case when if_attach() is
398  * called.
399  */
400 void
401 if_alloc_sadl(struct ifnet *ifp)
402 {
403 	struct ifaddr *ifa;
404 	const struct sockaddr_dl *sdl;
405 
406 	/*
407 	 * If the interface already has a link name, release it
408 	 * now.  This is useful for interfaces that can change
409 	 * link types, and thus switch link names often.
410 	 */
411 	if (ifp->if_sadl != NULL)
412 		if_free_sadl(ifp);
413 
414 	ifa = if_dl_create(ifp, &sdl);
415 
416 	ifa_insert(ifp, ifa);
417 	if_sadl_setrefs(ifp, ifa);
418 }
419 
420 static void
421 if_deactivate_sadl(struct ifnet *ifp)
422 {
423 	struct ifaddr *ifa;
424 
425 	KASSERT(ifp->if_dl != NULL);
426 
427 	ifa = ifp->if_dl;
428 
429 	ifp->if_sadl = NULL;
430 
431 	ifnet_addrs[ifp->if_index] = NULL;
432 	IFAFREE(ifa);
433 	ifp->if_dl = NULL;
434 	IFAFREE(ifa);
435 }
436 
437 void
438 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
439     const struct sockaddr_dl *sdl)
440 {
441 	int s;
442 
443 	s = splnet();
444 
445 	if_deactivate_sadl(ifp);
446 
447 	if_sadl_setrefs(ifp, ifa);
448 	IFADDR_FOREACH(ifa, ifp)
449 		rtinit(ifa, RTM_LLINFO_UPD, 0);
450 	splx(s);
451 }
452 
453 /*
454  * Free the link level name for the specified interface.  This is
455  * a detach helper.  This is called from if_detach() or from
456  * link layer type specific detach functions.
457  */
458 void
459 if_free_sadl(struct ifnet *ifp)
460 {
461 	struct ifaddr *ifa;
462 	int s;
463 
464 	ifa = ifnet_addrs[ifp->if_index];
465 	if (ifa == NULL) {
466 		KASSERT(ifp->if_sadl == NULL);
467 		KASSERT(ifp->if_dl == NULL);
468 		return;
469 	}
470 
471 	KASSERT(ifp->if_sadl != NULL);
472 	KASSERT(ifp->if_dl != NULL);
473 
474 	s = splnet();
475 	rtinit(ifa, RTM_DELETE, 0);
476 	ifa_remove(ifp, ifa);
477 	if_deactivate_sadl(ifp);
478 	if (ifp->if_hwdl == ifa) {
479 		IFAFREE(ifa);
480 		ifp->if_hwdl = NULL;
481 	}
482 	splx(s);
483 }
484 
485 /*
486  * Attach an interface to the
487  * list of "active" interfaces.
488  */
489 void
490 if_attach(struct ifnet *ifp)
491 {
492 	int indexlim = 0;
493 
494 	if (if_indexlim == 0) {
495 		TAILQ_INIT(&ifnet);
496 		if_indexlim = 8;
497 	}
498 	TAILQ_INIT(&ifp->if_addrlist);
499 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
500 	if (ifp->if_ioctl == NULL)
501 		ifp->if_ioctl = ifioctl_common;
502 
503 	mutex_enter(&index_gen_mtx);
504 	ifp->if_index_gen = index_gen++;
505 	mutex_exit(&index_gen_mtx);
506 
507 	ifp->if_index = if_index;
508 	if (ifindex2ifnet == NULL)
509 		if_index++;
510 	else
511 		while (ifp->if_index < if_indexlim &&
512 		    ifindex2ifnet[ifp->if_index] != NULL) {
513 			++if_index;
514 			if (if_index == 0)
515 				if_index = 1;
516 			/*
517 			 * If we hit USHRT_MAX, we skip back to 0 since
518 			 * there are a number of places where the value
519 			 * of if_index or if_index itself is compared
520 			 * to or stored in an unsigned short.  By
521 			 * jumping back, we won't botch those assignments
522 			 * or comparisons.
523 			 */
524 			else if (if_index == USHRT_MAX) {
525 				/*
526 				 * However, if we have to jump back to
527 				 * zero *twice* without finding an empty
528 				 * slot in ifindex2ifnet[], then there
529 				 * there are too many (>65535) interfaces.
530 				 */
531 				if (indexlim++)
532 					panic("too many interfaces");
533 				else
534 					if_index = 1;
535 			}
536 			ifp->if_index = if_index;
537 		}
538 
539 	/*
540 	 * We have some arrays that should be indexed by if_index.
541 	 * since if_index will grow dynamically, they should grow too.
542 	 *	struct ifadd **ifnet_addrs
543 	 *	struct ifnet **ifindex2ifnet
544 	 */
545 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
546 	    ifp->if_index >= if_indexlim) {
547 		size_t m, n, oldlim;
548 		void *q;
549 
550 		oldlim = if_indexlim;
551 		while (ifp->if_index >= if_indexlim)
552 			if_indexlim <<= 1;
553 
554 		/* grow ifnet_addrs */
555 		m = oldlim * sizeof(struct ifaddr *);
556 		n = if_indexlim * sizeof(struct ifaddr *);
557 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
558 		if (ifnet_addrs != NULL) {
559 			memcpy(q, ifnet_addrs, m);
560 			free(ifnet_addrs, M_IFADDR);
561 		}
562 		ifnet_addrs = (struct ifaddr **)q;
563 
564 		/* grow ifindex2ifnet */
565 		m = oldlim * sizeof(struct ifnet *);
566 		n = if_indexlim * sizeof(struct ifnet *);
567 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
568 		if (ifindex2ifnet != NULL) {
569 			memcpy(q, ifindex2ifnet, m);
570 			free(ifindex2ifnet, M_IFADDR);
571 		}
572 		ifindex2ifnet = (struct ifnet **)q;
573 	}
574 
575 	ifindex2ifnet[ifp->if_index] = ifp;
576 
577 	/*
578 	 * Link level name is allocated later by a separate call to
579 	 * if_alloc_sadl().
580 	 */
581 
582 	if (ifp->if_snd.ifq_maxlen == 0)
583 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
584 
585 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
586 
587 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
588 
589 	ifp->if_link_state = LINK_STATE_UNKNOWN;
590 
591 	ifp->if_capenable = 0;
592 	ifp->if_csum_flags_tx = 0;
593 	ifp->if_csum_flags_rx = 0;
594 
595 #ifdef ALTQ
596 	ifp->if_snd.altq_type = 0;
597 	ifp->if_snd.altq_disc = NULL;
598 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
599 	ifp->if_snd.altq_tbr  = NULL;
600 	ifp->if_snd.altq_ifp  = ifp;
601 #endif
602 
603 #ifdef PFIL_HOOKS
604 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
605 	ifp->if_pfil.ph_ifnet = ifp;
606 	if (pfil_head_register(&ifp->if_pfil) != 0)
607 		printf("%s: WARNING: unable to register pfil hook\n",
608 		    ifp->if_xname);
609 	(void)pfil_run_hooks(&if_pfil,
610 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
611 #endif
612 
613 	if (!STAILQ_EMPTY(&domains))
614 		if_attachdomain1(ifp);
615 
616 	/* Announce the interface. */
617 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
618 }
619 
620 void
621 if_attachdomain(void)
622 {
623 	struct ifnet *ifp;
624 	int s;
625 
626 	s = splnet();
627 	IFNET_FOREACH(ifp)
628 		if_attachdomain1(ifp);
629 	splx(s);
630 }
631 
632 void
633 if_attachdomain1(struct ifnet *ifp)
634 {
635 	struct domain *dp;
636 	int s;
637 
638 	s = splnet();
639 
640 	/* address family dependent data region */
641 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
642 	DOMAIN_FOREACH(dp) {
643 		if (dp->dom_ifattach != NULL)
644 			ifp->if_afdata[dp->dom_family] =
645 			    (*dp->dom_ifattach)(ifp);
646 	}
647 
648 	splx(s);
649 }
650 
651 /*
652  * Deactivate an interface.  This points all of the procedure
653  * handles at error stubs.  May be called from interrupt context.
654  */
655 void
656 if_deactivate(struct ifnet *ifp)
657 {
658 	int s;
659 
660 	s = splnet();
661 
662 	ifp->if_output	 = if_nulloutput;
663 	ifp->if_input	 = if_nullinput;
664 	ifp->if_start	 = if_nullstart;
665 	ifp->if_ioctl	 = if_nullioctl;
666 	ifp->if_init	 = if_nullinit;
667 	ifp->if_stop	 = if_nullstop;
668 	ifp->if_watchdog = if_nullwatchdog;
669 	ifp->if_drain	 = if_nulldrain;
670 
671 	/* No more packets may be enqueued. */
672 	ifp->if_snd.ifq_maxlen = 0;
673 
674 	splx(s);
675 }
676 
677 void
678 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
679 {
680 	struct ifaddr *ifa, *nifa;
681 
682 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
683 		nifa = IFADDR_NEXT(ifa);
684 		if (ifa->ifa_addr->sa_family != family)
685 			continue;
686 		(*purgeaddr)(ifa);
687 	}
688 }
689 
690 /*
691  * Detach an interface from the list of "active" interfaces,
692  * freeing any resources as we go along.
693  *
694  * NOTE: This routine must be called with a valid thread context,
695  * as it may block.
696  */
697 void
698 if_detach(struct ifnet *ifp)
699 {
700 	struct socket so;
701 	struct ifaddr *ifa;
702 #ifdef IFAREF_DEBUG
703 	struct ifaddr *last_ifa = NULL;
704 #endif
705 	struct domain *dp;
706 	const struct protosw *pr;
707 	int s, i, family, purged;
708 
709 	/*
710 	 * XXX It's kind of lame that we have to have the
711 	 * XXX socket structure...
712 	 */
713 	memset(&so, 0, sizeof(so));
714 
715 	s = splnet();
716 
717 	/*
718 	 * Do an if_down() to give protocols a chance to do something.
719 	 */
720 	if_down(ifp);
721 
722 #ifdef ALTQ
723 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
724 		altq_disable(&ifp->if_snd);
725 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
726 		altq_detach(&ifp->if_snd);
727 #endif
728 
729 	sysctl_teardown(&ifp->if_sysctl_log);
730 
731 #if NCARP > 0
732 	/* Remove the interface from any carp group it is a part of.  */
733 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
734 		carp_ifdetach(ifp);
735 #endif
736 
737 	/*
738 	 * Rip all the addresses off the interface.  This should make
739 	 * all of the routes go away.
740 	 *
741 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
742 	 * from the list, including our "cursor", ifa.  For safety,
743 	 * and to honor the TAILQ abstraction, I just restart the
744 	 * loop after each removal.  Note that the loop will exit
745 	 * when all of the remaining ifaddrs belong to the AF_LINK
746 	 * family.  I am counting on the historical fact that at
747 	 * least one pr_usrreq in each address domain removes at
748 	 * least one ifaddr.
749 	 */
750 again:
751 	IFADDR_FOREACH(ifa, ifp) {
752 		family = ifa->ifa_addr->sa_family;
753 #ifdef IFAREF_DEBUG
754 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
755 		    ifa, family, ifa->ifa_refcnt);
756 		if (last_ifa != NULL && ifa == last_ifa)
757 			panic("if_detach: loop detected");
758 		last_ifa = ifa;
759 #endif
760 		if (family == AF_LINK)
761 			continue;
762 		dp = pffinddomain(family);
763 #ifdef DIAGNOSTIC
764 		if (dp == NULL)
765 			panic("if_detach: no domain for AF %d",
766 			    family);
767 #endif
768 		/*
769 		 * XXX These PURGEIF calls are redundant with the
770 		 * purge-all-families calls below, but are left in for
771 		 * now both to make a smaller change, and to avoid
772 		 * unplanned interactions with clearing of
773 		 * ifp->if_addrlist.
774 		 */
775 		purged = 0;
776 		for (pr = dp->dom_protosw;
777 		     pr < dp->dom_protoswNPROTOSW; pr++) {
778 			so.so_proto = pr;
779 			if (pr->pr_usrreq != NULL) {
780 				(void) (*pr->pr_usrreq)(&so,
781 				    PRU_PURGEIF, NULL, NULL,
782 				    (struct mbuf *) ifp, curlwp);
783 				purged = 1;
784 			}
785 		}
786 		if (purged == 0) {
787 			/*
788 			 * XXX What's really the best thing to do
789 			 * XXX here?  --thorpej@NetBSD.org
790 			 */
791 			printf("if_detach: WARNING: AF %d not purged\n",
792 			    family);
793 			ifa_remove(ifp, ifa);
794 		}
795 		goto again;
796 	}
797 
798 	if_free_sadl(ifp);
799 
800 	/* Walk the routing table looking for stragglers. */
801 	for (i = 0; i <= AF_MAX; i++) {
802 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
803 			;
804 	}
805 
806 	DOMAIN_FOREACH(dp) {
807 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
808 			(*dp->dom_ifdetach)(ifp,
809 			    ifp->if_afdata[dp->dom_family]);
810 
811 		/*
812 		 * One would expect multicast memberships (INET and
813 		 * INET6) on UDP sockets to be purged by the PURGEIF
814 		 * calls above, but if all addresses were removed from
815 		 * the interface prior to destruction, the calls will
816 		 * not be made (e.g. ppp, for which pppd(8) generally
817 		 * removes addresses before destroying the interface).
818 		 * Because there is no invariant that multicast
819 		 * memberships only exist for interfaces with IPv4
820 		 * addresses, we must call PURGEIF regardless of
821 		 * addresses.  (Protocols which might store ifnet
822 		 * pointers are marked with PR_PURGEIF.)
823 		 */
824 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
825 			so.so_proto = pr;
826 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
827 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
828 				    NULL, (struct mbuf *)ifp, curlwp);
829 		}
830 	}
831 
832 #ifdef PFIL_HOOKS
833 	(void)pfil_run_hooks(&if_pfil,
834 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
835 	(void)pfil_head_unregister(&ifp->if_pfil);
836 #endif
837 
838 	/* Announce that the interface is gone. */
839 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
840 
841 	ifindex2ifnet[ifp->if_index] = NULL;
842 
843 	TAILQ_REMOVE(&ifnet, ifp, if_list);
844 
845 	/*
846 	 * remove packets that came from ifp, from software interrupt queues.
847 	 */
848 	DOMAIN_FOREACH(dp) {
849 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
850 			if (dp->dom_ifqueues[i] == NULL)
851 				break;
852 			if_detach_queues(ifp, dp->dom_ifqueues[i]);
853 		}
854 	}
855 
856 	splx(s);
857 }
858 
859 static void
860 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
861 {
862 	struct mbuf *m, *prev, *next;
863 
864 	prev = NULL;
865 	for (m = q->ifq_head; m != NULL; m = next) {
866 		next = m->m_nextpkt;
867 #ifdef DIAGNOSTIC
868 		if ((m->m_flags & M_PKTHDR) == 0) {
869 			prev = m;
870 			continue;
871 		}
872 #endif
873 		if (m->m_pkthdr.rcvif != ifp) {
874 			prev = m;
875 			continue;
876 		}
877 
878 		if (prev != NULL)
879 			prev->m_nextpkt = m->m_nextpkt;
880 		else
881 			q->ifq_head = m->m_nextpkt;
882 		if (q->ifq_tail == m)
883 			q->ifq_tail = prev;
884 		q->ifq_len--;
885 
886 		m->m_nextpkt = NULL;
887 		m_freem(m);
888 		IF_DROP(q);
889 	}
890 }
891 
892 /*
893  * Callback for a radix tree walk to delete all references to an
894  * ifnet.
895  */
896 static int
897 if_rt_walktree(struct rtentry *rt, void *v)
898 {
899 	struct ifnet *ifp = (struct ifnet *)v;
900 	int error;
901 
902 	if (rt->rt_ifp != ifp)
903 		return 0;
904 
905 	/* Delete the entry. */
906 	++rt->rt_refcnt;
907 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
908 	    rt_mask(rt), rt->rt_flags, NULL);
909 	KASSERT((rt->rt_flags & RTF_UP) == 0);
910 	rt->rt_ifp = NULL;
911 	RTFREE(rt);
912 	if (error != 0)
913 		printf("%s: warning: unable to delete rtentry @ %p, "
914 		    "error = %d\n", ifp->if_xname, rt, error);
915 	return ERESTART;
916 }
917 
918 /*
919  * Create a clone network interface.
920  */
921 int
922 if_clone_create(const char *name)
923 {
924 	struct if_clone *ifc;
925 	int unit;
926 
927 	ifc = if_clone_lookup(name, &unit);
928 	if (ifc == NULL)
929 		return EINVAL;
930 
931 	if (ifunit(name) != NULL)
932 		return EEXIST;
933 
934 	return (*ifc->ifc_create)(ifc, unit);
935 }
936 
937 /*
938  * Destroy a clone network interface.
939  */
940 int
941 if_clone_destroy(const char *name)
942 {
943 	struct if_clone *ifc;
944 	struct ifnet *ifp;
945 
946 	ifc = if_clone_lookup(name, NULL);
947 	if (ifc == NULL)
948 		return EINVAL;
949 
950 	ifp = ifunit(name);
951 	if (ifp == NULL)
952 		return ENXIO;
953 
954 	if (ifc->ifc_destroy == NULL)
955 		return EOPNOTSUPP;
956 
957 	return (*ifc->ifc_destroy)(ifp);
958 }
959 
960 /*
961  * Look up a network interface cloner.
962  */
963 static struct if_clone *
964 if_clone_lookup(const char *name, int *unitp)
965 {
966 	struct if_clone *ifc;
967 	const char *cp;
968 	int unit;
969 
970 	/* separate interface name from unit */
971 	for (cp = name;
972 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
973 	    cp++)
974 		continue;
975 
976 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
977 		return NULL;	/* No name or unit number */
978 
979 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
980 		if (strlen(ifc->ifc_name) == cp - name &&
981 		    strncmp(name, ifc->ifc_name, cp - name) == 0)
982 			break;
983 	}
984 
985 	if (ifc == NULL)
986 		return NULL;
987 
988 	unit = 0;
989 	while (cp - name < IFNAMSIZ && *cp) {
990 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
991 			/* Bogus unit number. */
992 			return NULL;
993 		}
994 		unit = (unit * 10) + (*cp++ - '0');
995 	}
996 
997 	if (unitp != NULL)
998 		*unitp = unit;
999 	return ifc;
1000 }
1001 
1002 /*
1003  * Register a network interface cloner.
1004  */
1005 void
1006 if_clone_attach(struct if_clone *ifc)
1007 {
1008 
1009 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1010 	if_cloners_count++;
1011 }
1012 
1013 /*
1014  * Unregister a network interface cloner.
1015  */
1016 void
1017 if_clone_detach(struct if_clone *ifc)
1018 {
1019 
1020 	LIST_REMOVE(ifc, ifc_list);
1021 	if_cloners_count--;
1022 }
1023 
1024 /*
1025  * Provide list of interface cloners to userspace.
1026  */
1027 static int
1028 if_clone_list(struct if_clonereq *ifcr)
1029 {
1030 	char outbuf[IFNAMSIZ], *dst;
1031 	struct if_clone *ifc;
1032 	int count, error = 0;
1033 
1034 	ifcr->ifcr_total = if_cloners_count;
1035 	if ((dst = ifcr->ifcr_buffer) == NULL) {
1036 		/* Just asking how many there are. */
1037 		return 0;
1038 	}
1039 
1040 	if (ifcr->ifcr_count < 0)
1041 		return EINVAL;
1042 
1043 	count = (if_cloners_count < ifcr->ifcr_count) ?
1044 	    if_cloners_count : ifcr->ifcr_count;
1045 
1046 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
1047 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1048 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1049 		if (outbuf[sizeof(outbuf) - 1] != '\0')
1050 			return ENAMETOOLONG;
1051 		error = copyout(outbuf, dst, sizeof(outbuf));
1052 		if (error != 0)
1053 			break;
1054 	}
1055 
1056 	return error;
1057 }
1058 
1059 void
1060 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
1061 {
1062 	ifa->ifa_ifp = ifp;
1063 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1064 	IFAREF(ifa);
1065 }
1066 
1067 void
1068 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
1069 {
1070 	KASSERT(ifa->ifa_ifp == ifp);
1071 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1072 	IFAFREE(ifa);
1073 }
1074 
1075 static inline int
1076 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
1077 {
1078 	return sockaddr_cmp(sa1, sa2) == 0;
1079 }
1080 
1081 /*
1082  * Locate an interface based on a complete address.
1083  */
1084 /*ARGSUSED*/
1085 struct ifaddr *
1086 ifa_ifwithaddr(const struct sockaddr *addr)
1087 {
1088 	struct ifnet *ifp;
1089 	struct ifaddr *ifa;
1090 
1091 	IFNET_FOREACH(ifp) {
1092 		if (ifp->if_output == if_nulloutput)
1093 			continue;
1094 		IFADDR_FOREACH(ifa, ifp) {
1095 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1096 				continue;
1097 			if (equal(addr, ifa->ifa_addr))
1098 				return ifa;
1099 			if ((ifp->if_flags & IFF_BROADCAST) &&
1100 			    ifa->ifa_broadaddr &&
1101 			    /* IP6 doesn't have broadcast */
1102 			    ifa->ifa_broadaddr->sa_len != 0 &&
1103 			    equal(ifa->ifa_broadaddr, addr))
1104 				return ifa;
1105 		}
1106 	}
1107 	return NULL;
1108 }
1109 
1110 /*
1111  * Locate the point to point interface with a given destination address.
1112  */
1113 /*ARGSUSED*/
1114 struct ifaddr *
1115 ifa_ifwithdstaddr(const struct sockaddr *addr)
1116 {
1117 	struct ifnet *ifp;
1118 	struct ifaddr *ifa;
1119 
1120 	IFNET_FOREACH(ifp) {
1121 		if (ifp->if_output == if_nulloutput)
1122 			continue;
1123 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1124 			continue;
1125 		IFADDR_FOREACH(ifa, ifp) {
1126 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
1127 			    ifa->ifa_dstaddr == NULL)
1128 				continue;
1129 			if (equal(addr, ifa->ifa_dstaddr))
1130 				return ifa;
1131 		}
1132 	}
1133 	return NULL;
1134 }
1135 
1136 /*
1137  * Find an interface on a specific network.  If many, choice
1138  * is most specific found.
1139  */
1140 struct ifaddr *
1141 ifa_ifwithnet(const struct sockaddr *addr)
1142 {
1143 	struct ifnet *ifp;
1144 	struct ifaddr *ifa;
1145 	const struct sockaddr_dl *sdl;
1146 	struct ifaddr *ifa_maybe = 0;
1147 	u_int af = addr->sa_family;
1148 	const char *addr_data = addr->sa_data, *cplim;
1149 
1150 	if (af == AF_LINK) {
1151 		sdl = satocsdl(addr);
1152 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
1153 		    ifindex2ifnet[sdl->sdl_index] &&
1154 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
1155 			return ifnet_addrs[sdl->sdl_index];
1156 	}
1157 #ifdef NETATALK
1158 	if (af == AF_APPLETALK) {
1159 		const struct sockaddr_at *sat, *sat2;
1160 		sat = (const struct sockaddr_at *)addr;
1161 		IFNET_FOREACH(ifp) {
1162 			if (ifp->if_output == if_nulloutput)
1163 				continue;
1164 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1165 			if (ifa == NULL)
1166 				continue;
1167 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
1168 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1169 				return ifa; /* exact match */
1170 			if (ifa_maybe == NULL) {
1171 				/* else keep the if with the right range */
1172 				ifa_maybe = ifa;
1173 			}
1174 		}
1175 		return ifa_maybe;
1176 	}
1177 #endif
1178 	IFNET_FOREACH(ifp) {
1179 		if (ifp->if_output == if_nulloutput)
1180 			continue;
1181 		IFADDR_FOREACH(ifa, ifp) {
1182 			const char *cp, *cp2, *cp3;
1183 
1184 			if (ifa->ifa_addr->sa_family != af ||
1185 			    ifa->ifa_netmask == NULL)
1186  next:				continue;
1187 			cp = addr_data;
1188 			cp2 = ifa->ifa_addr->sa_data;
1189 			cp3 = ifa->ifa_netmask->sa_data;
1190 			cplim = (const char *)ifa->ifa_netmask +
1191 			    ifa->ifa_netmask->sa_len;
1192 			while (cp3 < cplim) {
1193 				if ((*cp++ ^ *cp2++) & *cp3++) {
1194 					/* want to continue for() loop */
1195 					goto next;
1196 				}
1197 			}
1198 			if (ifa_maybe == NULL ||
1199 			    rn_refines((void *)ifa->ifa_netmask,
1200 			    (void *)ifa_maybe->ifa_netmask))
1201 				ifa_maybe = ifa;
1202 		}
1203 	}
1204 	return ifa_maybe;
1205 }
1206 
1207 /*
1208  * Find the interface of the addresss.
1209  */
1210 struct ifaddr *
1211 ifa_ifwithladdr(const struct sockaddr *addr)
1212 {
1213 	struct ifaddr *ia;
1214 
1215 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
1216 	    (ia = ifa_ifwithnet(addr)))
1217 		return ia;
1218 	return NULL;
1219 }
1220 
1221 /*
1222  * Find an interface using a specific address family
1223  */
1224 struct ifaddr *
1225 ifa_ifwithaf(int af)
1226 {
1227 	struct ifnet *ifp;
1228 	struct ifaddr *ifa;
1229 
1230 	IFNET_FOREACH(ifp) {
1231 		if (ifp->if_output == if_nulloutput)
1232 			continue;
1233 		IFADDR_FOREACH(ifa, ifp) {
1234 			if (ifa->ifa_addr->sa_family == af)
1235 				return ifa;
1236 		}
1237 	}
1238 	return NULL;
1239 }
1240 
1241 /*
1242  * Find an interface address specific to an interface best matching
1243  * a given address.
1244  */
1245 struct ifaddr *
1246 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1247 {
1248 	struct ifaddr *ifa;
1249 	const char *cp, *cp2, *cp3;
1250 	const char *cplim;
1251 	struct ifaddr *ifa_maybe = 0;
1252 	u_int af = addr->sa_family;
1253 
1254 	if (ifp->if_output == if_nulloutput)
1255 		return NULL;
1256 
1257 	if (af >= AF_MAX)
1258 		return NULL;
1259 
1260 	IFADDR_FOREACH(ifa, ifp) {
1261 		if (ifa->ifa_addr->sa_family != af)
1262 			continue;
1263 		ifa_maybe = ifa;
1264 		if (ifa->ifa_netmask == NULL) {
1265 			if (equal(addr, ifa->ifa_addr) ||
1266 			    (ifa->ifa_dstaddr &&
1267 			     equal(addr, ifa->ifa_dstaddr)))
1268 				return ifa;
1269 			continue;
1270 		}
1271 		cp = addr->sa_data;
1272 		cp2 = ifa->ifa_addr->sa_data;
1273 		cp3 = ifa->ifa_netmask->sa_data;
1274 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1275 		for (; cp3 < cplim; cp3++) {
1276 			if ((*cp++ ^ *cp2++) & *cp3)
1277 				break;
1278 		}
1279 		if (cp3 == cplim)
1280 			return ifa;
1281 	}
1282 	return ifa_maybe;
1283 }
1284 
1285 /*
1286  * Default action when installing a route with a Link Level gateway.
1287  * Lookup an appropriate real ifa to point to.
1288  * This should be moved to /sys/net/link.c eventually.
1289  */
1290 void
1291 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1292 {
1293 	struct ifaddr *ifa;
1294 	const struct sockaddr *dst;
1295 	struct ifnet *ifp;
1296 
1297 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
1298 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1299 		return;
1300 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
1301 		rt_replace_ifa(rt, ifa);
1302 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1303 			ifa->ifa_rtrequest(cmd, rt, info);
1304 	}
1305 }
1306 
1307 /*
1308  * Handle a change in the interface link state.
1309  */
1310 void
1311 if_link_state_change(struct ifnet *ifp, int link_state)
1312 {
1313 	if (ifp->if_link_state == link_state)
1314 		return;
1315 	ifp->if_link_state = link_state;
1316 	/* Notify that the link state has changed. */
1317 	rt_ifmsg(ifp);
1318 #if NCARP > 0
1319 	if (ifp->if_carp)
1320 		carp_carpdev_state(ifp);
1321 #endif
1322 }
1323 
1324 /*
1325  * Mark an interface down and notify protocols of
1326  * the transition.
1327  * NOTE: must be called at splsoftnet or equivalent.
1328  */
1329 void
1330 if_down(struct ifnet *ifp)
1331 {
1332 	struct ifaddr *ifa;
1333 
1334 	ifp->if_flags &= ~IFF_UP;
1335 	nanotime(&ifp->if_lastchange);
1336 	IFADDR_FOREACH(ifa, ifp)
1337 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1338 	IFQ_PURGE(&ifp->if_snd);
1339 #if NCARP > 0
1340 	if (ifp->if_carp)
1341 		carp_carpdev_state(ifp);
1342 #endif
1343 	rt_ifmsg(ifp);
1344 }
1345 
1346 /*
1347  * Mark an interface up and notify protocols of
1348  * the transition.
1349  * NOTE: must be called at splsoftnet or equivalent.
1350  */
1351 void
1352 if_up(struct ifnet *ifp)
1353 {
1354 #ifdef notyet
1355 	struct ifaddr *ifa;
1356 #endif
1357 
1358 	ifp->if_flags |= IFF_UP;
1359 	nanotime(&ifp->if_lastchange);
1360 #ifdef notyet
1361 	/* this has no effect on IP, and will kill all ISO connections XXX */
1362 	IFADDR_FOREACH(ifa, ifp)
1363 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
1364 #endif
1365 #if NCARP > 0
1366 	if (ifp->if_carp)
1367 		carp_carpdev_state(ifp);
1368 #endif
1369 	rt_ifmsg(ifp);
1370 #ifdef INET6
1371 	in6_if_up(ifp);
1372 #endif
1373 }
1374 
1375 /*
1376  * Handle interface watchdog timer routines.  Called
1377  * from softclock, we decrement timers (if set) and
1378  * call the appropriate interface routine on expiration.
1379  */
1380 void
1381 if_slowtimo(void *arg)
1382 {
1383 	struct ifnet *ifp;
1384 	int s = splnet();
1385 
1386 	IFNET_FOREACH(ifp) {
1387 		if (ifp->if_timer == 0 || --ifp->if_timer)
1388 			continue;
1389 		if (ifp->if_watchdog != NULL)
1390 			(*ifp->if_watchdog)(ifp);
1391 	}
1392 	splx(s);
1393 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1394 }
1395 
1396 /*
1397  * Set/clear promiscuous mode on interface ifp based on the truth value
1398  * of pswitch.  The calls are reference counted so that only the first
1399  * "on" request actually has an effect, as does the final "off" request.
1400  * Results are undefined if the "off" and "on" requests are not matched.
1401  */
1402 int
1403 ifpromisc(struct ifnet *ifp, int pswitch)
1404 {
1405 	int pcount, ret;
1406 	short flags;
1407 	struct ifreq ifr;
1408 
1409 	pcount = ifp->if_pcount;
1410 	flags = ifp->if_flags;
1411 	if (pswitch) {
1412 		/*
1413 		 * Allow the device to be "placed" into promiscuous
1414 		 * mode even if it is not configured up.  It will
1415 		 * consult IFF_PROMISC when it is brought up.
1416 		 */
1417 		if (ifp->if_pcount++ != 0)
1418 			return 0;
1419 		ifp->if_flags |= IFF_PROMISC;
1420 		if ((ifp->if_flags & IFF_UP) == 0)
1421 			return 0;
1422 	} else {
1423 		if (--ifp->if_pcount > 0)
1424 			return 0;
1425 		ifp->if_flags &= ~IFF_PROMISC;
1426 		/*
1427 		 * If the device is not configured up, we should not need to
1428 		 * turn off promiscuous mode (device should have turned it
1429 		 * off when interface went down; and will look at IFF_PROMISC
1430 		 * again next time interface comes up).
1431 		 */
1432 		if ((ifp->if_flags & IFF_UP) == 0)
1433 			return 0;
1434 	}
1435 	memset(&ifr, 0, sizeof(ifr));
1436 	ifr.ifr_flags = ifp->if_flags;
1437 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
1438 	/* Restore interface state if not successful. */
1439 	if (ret != 0) {
1440 		ifp->if_pcount = pcount;
1441 		ifp->if_flags = flags;
1442 	}
1443 	return ret;
1444 }
1445 
1446 /*
1447  * Map interface name to
1448  * interface structure pointer.
1449  */
1450 struct ifnet *
1451 ifunit(const char *name)
1452 {
1453 	struct ifnet *ifp;
1454 	const char *cp = name;
1455 	u_int unit = 0;
1456 	u_int i;
1457 
1458 	/*
1459 	 * If the entire name is a number, treat it as an ifindex.
1460 	 */
1461 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
1462 		unit = unit * 10 + (*cp - '0');
1463 	}
1464 
1465 	/*
1466 	 * If the number took all of the name, then it's a valid ifindex.
1467 	 */
1468 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
1469 		if (unit >= if_indexlim)
1470 			return NULL;
1471 		ifp = ifindex2ifnet[unit];
1472 		if (ifp == NULL || ifp->if_output == if_nulloutput)
1473 			return NULL;
1474 		return ifp;
1475 	}
1476 
1477 	IFNET_FOREACH(ifp) {
1478 		if (ifp->if_output == if_nulloutput)
1479 			continue;
1480 	 	if (strcmp(ifp->if_xname, name) == 0)
1481 			return ifp;
1482 	}
1483 	return NULL;
1484 }
1485 
1486 ifnet_t *
1487 if_byindex(u_int idx)
1488 {
1489 
1490 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
1491 }
1492 
1493 /* common */
1494 int
1495 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1496 {
1497 	int s;
1498 	struct ifreq *ifr;
1499 	struct ifcapreq *ifcr;
1500 	struct ifdatareq *ifdr;
1501 
1502 	switch (cmd) {
1503 	case SIOCSIFCAP:
1504 		ifcr = data;
1505 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
1506 			return EINVAL;
1507 
1508 		if (ifcr->ifcr_capenable == ifp->if_capenable)
1509 			return 0;
1510 
1511 		ifp->if_capenable = ifcr->ifcr_capenable;
1512 
1513 		/* Pre-compute the checksum flags mask. */
1514 		ifp->if_csum_flags_tx = 0;
1515 		ifp->if_csum_flags_rx = 0;
1516 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
1517 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
1518 		}
1519 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
1520 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
1521 		}
1522 
1523 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
1524 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
1525 		}
1526 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
1527 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
1528 		}
1529 
1530 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
1531 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
1532 		}
1533 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
1534 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
1535 		}
1536 
1537 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
1538 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
1539 		}
1540 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
1541 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
1542 		}
1543 
1544 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
1545 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1546 		}
1547 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
1548 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
1549 		}
1550 		if (ifp->if_flags & IFF_UP)
1551 			return ENETRESET;
1552 		return 0;
1553 	case SIOCSIFFLAGS:
1554 		ifr = data;
1555 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1556 			s = splnet();
1557 			if_down(ifp);
1558 			splx(s);
1559 		}
1560 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1561 			s = splnet();
1562 			if_up(ifp);
1563 			splx(s);
1564 		}
1565 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1566 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1567 		break;
1568 	case SIOCGIFFLAGS:
1569 		ifr = data;
1570 		ifr->ifr_flags = ifp->if_flags;
1571 		break;
1572 
1573 	case SIOCGIFMETRIC:
1574 		ifr = data;
1575 		ifr->ifr_metric = ifp->if_metric;
1576 		break;
1577 
1578 	case SIOCGIFMTU:
1579 		ifr = data;
1580 		ifr->ifr_mtu = ifp->if_mtu;
1581 		break;
1582 
1583 	case SIOCGIFDLT:
1584 		ifr = data;
1585 		ifr->ifr_dlt = ifp->if_dlt;
1586 		break;
1587 
1588 	case SIOCGIFCAP:
1589 		ifcr = data;
1590 		ifcr->ifcr_capabilities = ifp->if_capabilities;
1591 		ifcr->ifcr_capenable = ifp->if_capenable;
1592 		break;
1593 
1594 	case SIOCSIFMETRIC:
1595 		ifr = data;
1596 		ifp->if_metric = ifr->ifr_metric;
1597 		break;
1598 
1599 	case SIOCGIFDATA:
1600 		ifdr = data;
1601 		ifdr->ifdr_data = ifp->if_data;
1602 		break;
1603 
1604 	case SIOCZIFDATA:
1605 		ifdr = data;
1606 		ifdr->ifdr_data = ifp->if_data;
1607 		/*
1608 		 * Assumes that the volatile counters that can be
1609 		 * zero'ed are at the end of if_data.
1610 		 */
1611 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
1612 		    offsetof(struct if_data, ifi_ipackets));
1613 		break;
1614 	case SIOCSIFMTU:
1615 		ifr = data;
1616 		if (ifp->if_mtu == ifr->ifr_mtu)
1617 			break;
1618 		ifp->if_mtu = ifr->ifr_mtu;
1619 		/*
1620 		 * If the link MTU changed, do network layer specific procedure.
1621 		 */
1622 #ifdef INET6
1623 		nd6_setmtu(ifp);
1624 #endif
1625 		return ENETRESET;
1626 	default:
1627 		return ENOTTY;
1628 	}
1629 	return 0;
1630 }
1631 
1632 int
1633 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
1634     lwp_t *l)
1635 {
1636 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
1637 	struct ifaddr *ifa;
1638 	const struct sockaddr *any, *sa;
1639 	union {
1640 		struct sockaddr sa;
1641 		struct sockaddr_storage ss;
1642 	} u, v;
1643 
1644 	switch (cmd) {
1645 	case SIOCSIFADDRPREF:
1646 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
1647 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1648 		    NULL) != 0)
1649 			return EPERM;
1650 	case SIOCGIFADDRPREF:
1651 		break;
1652 	default:
1653 		return EOPNOTSUPP;
1654 	}
1655 
1656 	/* sanity checks */
1657 	if (data == NULL || ifp == NULL) {
1658 		panic("invalid argument to %s", __func__);
1659 		/*NOTREACHED*/
1660 	}
1661 
1662 	/* address must be specified on ADD and DELETE */
1663 	sa = sstocsa(&ifap->ifap_addr);
1664 	if (sa->sa_family != sofamily(so))
1665 		return EINVAL;
1666 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
1667 		return EINVAL;
1668 
1669 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
1670 
1671 	IFADDR_FOREACH(ifa, ifp) {
1672 		if (ifa->ifa_addr->sa_family != sa->sa_family)
1673 			continue;
1674 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1675 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1676 			break;
1677 	}
1678 	if (ifa == NULL)
1679 		return EADDRNOTAVAIL;
1680 
1681 	switch (cmd) {
1682 	case SIOCSIFADDRPREF:
1683 		ifa->ifa_preference = ifap->ifap_preference;
1684 		return 0;
1685 	case SIOCGIFADDRPREF:
1686 		/* fill in the if_laddrreq structure */
1687 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
1688 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
1689 		ifap->ifap_preference = ifa->ifa_preference;
1690 		return 0;
1691 	default:
1692 		return EOPNOTSUPP;
1693 	}
1694 }
1695 
1696 /*
1697  * Interface ioctls.
1698  */
1699 int
1700 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1701 {
1702 	struct ifnet *ifp;
1703 	struct ifreq *ifr;
1704 	int error = 0;
1705 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
1706 	u_long ocmd = cmd;
1707 #endif
1708 	short oif_flags;
1709 #ifdef COMPAT_OIFREQ
1710 	struct ifreq ifrb;
1711 	struct oifreq *oifr = NULL;
1712 #endif
1713 
1714 	switch (cmd) {
1715 #ifdef COMPAT_OIFREQ
1716 	case OSIOCGIFCONF:
1717 	case OOSIOCGIFCONF:
1718 		return compat_ifconf(cmd, data);
1719 #endif
1720 #ifdef COMPAT_OIFDATA
1721 	case OSIOCGIFDATA:
1722 	case OSIOCZIFDATA:
1723 		return compat_ifdatareq(l, cmd, data);
1724 #endif
1725 	case SIOCGIFCONF:
1726 		return ifconf(cmd, data);
1727 	case SIOCINITIFADDR:
1728 		return EPERM;
1729 	}
1730 
1731 #ifdef COMPAT_OIFREQ
1732 	cmd = compat_cvtcmd(cmd);
1733 	if (cmd != ocmd) {
1734 		oifr = data;
1735 		data = ifr = &ifrb;
1736 		ifreqo2n(oifr, ifr);
1737 	} else
1738 #endif
1739 		ifr = data;
1740 
1741 	ifp = ifunit(ifr->ifr_name);
1742 
1743 	switch (cmd) {
1744 	case SIOCIFCREATE:
1745 	case SIOCIFDESTROY:
1746 		if (l != NULL) {
1747 			error = kauth_authorize_network(l->l_cred,
1748 			    KAUTH_NETWORK_INTERFACE,
1749 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1750 			    (void *)cmd, NULL);
1751 			if (error != 0)
1752 				return error;
1753 		}
1754 		return (cmd == SIOCIFCREATE) ?
1755 			if_clone_create(ifr->ifr_name) :
1756 			if_clone_destroy(ifr->ifr_name);
1757 
1758 	case SIOCIFGCLONERS:
1759 		return if_clone_list((struct if_clonereq *)data);
1760 	}
1761 
1762 	if (ifp == NULL)
1763 		return ENXIO;
1764 
1765 	switch (cmd) {
1766 	case SIOCALIFADDR:
1767 	case SIOCDLIFADDR:
1768 	case SIOCSIFADDRPREF:
1769 	case SIOCSIFFLAGS:
1770 	case SIOCSIFCAP:
1771 	case SIOCSIFMETRIC:
1772 	case SIOCZIFDATA:
1773 	case SIOCSIFMTU:
1774 	case SIOCSIFPHYADDR:
1775 	case SIOCDIFPHYADDR:
1776 #ifdef INET6
1777 	case SIOCSIFPHYADDR_IN6:
1778 #endif
1779 	case SIOCSLIFPHYADDR:
1780 	case SIOCADDMULTI:
1781 	case SIOCDELMULTI:
1782 	case SIOCSIFMEDIA:
1783 	case SIOCSDRVSPEC:
1784 	case SIOCG80211:
1785 	case SIOCS80211:
1786 	case SIOCS80211NWID:
1787 	case SIOCS80211NWKEY:
1788 	case SIOCS80211POWER:
1789 	case SIOCS80211BSSID:
1790 	case SIOCS80211CHANNEL:
1791 	case SIOCSLINKSTR:
1792 		if (l != NULL) {
1793 			error = kauth_authorize_network(l->l_cred,
1794 			    KAUTH_NETWORK_INTERFACE,
1795 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
1796 			    (void *)cmd, NULL);
1797 			if (error != 0)
1798 				return error;
1799 		}
1800 	}
1801 
1802 	oif_flags = ifp->if_flags;
1803 
1804 	error = (*ifp->if_ioctl)(ifp, cmd, data);
1805 	if (error != ENOTTY)
1806 		;
1807 	else if (so->so_proto == NULL)
1808 		return EOPNOTSUPP;
1809 	else {
1810 #ifdef COMPAT_OSOCK
1811 		error = compat_ifioctl(so, ocmd, cmd, data, l);
1812 #else
1813 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
1814 		    (struct mbuf *)cmd, (struct mbuf *)data,
1815 		    (struct mbuf *)ifp, l);
1816 #endif
1817 	}
1818 
1819 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1820 #ifdef INET6
1821 		if ((ifp->if_flags & IFF_UP) != 0) {
1822 			int s = splnet();
1823 			in6_if_up(ifp);
1824 			splx(s);
1825 		}
1826 #endif
1827 	}
1828 #ifdef COMPAT_OIFREQ
1829 	if (cmd != ocmd)
1830 		ifreqn2o(oifr, ifr);
1831 #endif
1832 
1833 	return error;
1834 }
1835 
1836 /*
1837  * Return interface configuration
1838  * of system.  List may be used
1839  * in later ioctl's (above) to get
1840  * other information.
1841  *
1842  * Each record is a struct ifreq.  Before the addition of
1843  * sockaddr_storage, the API rule was that sockaddr flavors that did
1844  * not fit would extend beyond the struct ifreq, with the next struct
1845  * ifreq starting sa_len beyond the struct sockaddr.  Because the
1846  * union in struct ifreq includes struct sockaddr_storage, every kind
1847  * of sockaddr must fit.  Thus, there are no longer any overlength
1848  * records.
1849  *
1850  * Records are added to the user buffer if they fit, and ifc_len is
1851  * adjusted to the length that was written.  Thus, the user is only
1852  * assured of getting the complete list if ifc_len on return is at
1853  * least sizeof(struct ifreq) less than it was on entry.
1854  *
1855  * If the user buffer pointer is NULL, this routine copies no data and
1856  * returns the amount of space that would be needed.
1857  *
1858  * Invariants:
1859  * ifrp points to the next part of the user's buffer to be used.  If
1860  * ifrp != NULL, space holds the number of bytes remaining that we may
1861  * write at ifrp.  Otherwise, space holds the number of bytes that
1862  * would have been written had there been adequate space.
1863  */
1864 /*ARGSUSED*/
1865 int
1866 ifconf(u_long cmd, void *data)
1867 {
1868 	struct ifconf *ifc = (struct ifconf *)data;
1869 	struct ifnet *ifp;
1870 	struct ifaddr *ifa;
1871 	struct ifreq ifr, *ifrp;
1872 	int space, error = 0;
1873 	const int sz = (int)sizeof(struct ifreq);
1874 
1875 	if ((ifrp = ifc->ifc_req) == NULL)
1876 		space = 0;
1877 	else
1878 		space = ifc->ifc_len;
1879 	IFNET_FOREACH(ifp) {
1880 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1881 		    sizeof(ifr.ifr_name));
1882 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1883 			return ENAMETOOLONG;
1884 		if (IFADDR_EMPTY(ifp)) {
1885 			/* Interface with no addresses - send zero sockaddr. */
1886 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1887 			if (ifrp == NULL) {
1888 				space += sz;
1889 				continue;
1890 			}
1891 			if (space >= sz) {
1892 				error = copyout(&ifr, ifrp, sz);
1893 				if (error != 0)
1894 					return error;
1895 				ifrp++;
1896 				space -= sz;
1897 			}
1898 		}
1899 
1900 		IFADDR_FOREACH(ifa, ifp) {
1901 			struct sockaddr *sa = ifa->ifa_addr;
1902 			/* all sockaddrs must fit in sockaddr_storage */
1903 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
1904 
1905 			if (ifrp == NULL) {
1906 				space += sz;
1907 				continue;
1908 			}
1909 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
1910 			if (space >= sz) {
1911 				error = copyout(&ifr, ifrp, sz);
1912 				if (error != 0)
1913 					return (error);
1914 				ifrp++; space -= sz;
1915 			}
1916 		}
1917 	}
1918 	if (ifrp != NULL) {
1919 		KASSERT(0 <= space && space <= ifc->ifc_len);
1920 		ifc->ifc_len -= space;
1921 	} else {
1922 		KASSERT(space >= 0);
1923 		ifc->ifc_len = space;
1924 	}
1925 	return (0);
1926 }
1927 
1928 int
1929 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
1930 {
1931 	uint8_t len;
1932 #ifdef COMPAT_OIFREQ
1933 	struct ifreq ifrb;
1934 	struct oifreq *oifr = NULL;
1935 	u_long ocmd = cmd;
1936 	cmd = compat_cvtcmd(cmd);
1937 	if (cmd != ocmd) {
1938 		oifr = (struct oifreq *)(void *)ifr;
1939 		ifr = &ifrb;
1940 		ifreqo2n(oifr, ifr);
1941 		len = sizeof(oifr->ifr_addr);
1942 	} else
1943 #endif
1944 		len = sizeof(ifr->ifr_ifru.ifru_space);
1945 
1946 	if (len < sa->sa_len)
1947 		return EFBIG;
1948 
1949 	memset(&ifr->ifr_addr, 0, len);
1950 	sockaddr_copy(&ifr->ifr_addr, len, sa);
1951 
1952 #ifdef COMPAT_OIFREQ
1953 	if (cmd != ocmd)
1954 		ifreqn2o(oifr, ifr);
1955 #endif
1956 	return 0;
1957 }
1958 
1959 /*
1960  * Queue message on interface, and start output if interface
1961  * not yet active.
1962  */
1963 int
1964 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
1965     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1966 {
1967 	int len = m->m_pkthdr.len;
1968 	int mflags = m->m_flags;
1969 	int s = splnet();
1970 	int error;
1971 
1972 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
1973 	if (error != 0)
1974 		goto out;
1975 	ifp->if_obytes += len;
1976 	if (mflags & M_MCAST)
1977 		ifp->if_omcasts++;
1978 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
1979 		(*ifp->if_start)(ifp);
1980 out:
1981 	splx(s);
1982 	return error;
1983 }
1984 
1985 /*
1986  * Queue message on interface, possibly using a second fast queue
1987  */
1988 int
1989 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
1990     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
1991 {
1992 	int error = 0;
1993 
1994 	if (ifq != NULL
1995 #ifdef ALTQ
1996 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
1997 #endif
1998 	    ) {
1999 		if (IF_QFULL(ifq)) {
2000 			IF_DROP(&ifp->if_snd);
2001 			m_freem(m);
2002 			if (error == 0)
2003 				error = ENOBUFS;
2004 		} else
2005 			IF_ENQUEUE(ifq, m);
2006 	} else
2007 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
2008 	if (error != 0) {
2009 		++ifp->if_oerrors;
2010 		return error;
2011 	}
2012 	return 0;
2013 }
2014 
2015 
2016 static void
2017 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
2018     struct ifaltq *ifq)
2019 {
2020 	const struct sysctlnode *cnode, *rnode;
2021 
2022 	if (sysctl_createv(clog, 0, NULL, &rnode,
2023 		       CTLFLAG_PERMANENT,
2024 		       CTLTYPE_NODE, "net", NULL,
2025 		       NULL, 0, NULL, 0,
2026 		       CTL_NET, CTL_EOL) != 0)
2027 		goto bad;
2028 
2029 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2030 		       CTLFLAG_PERMANENT,
2031 		       CTLTYPE_NODE, "interfaces",
2032 		       SYSCTL_DESCR("Per-interface controls"),
2033 		       NULL, 0, NULL, 0,
2034 		       CTL_CREATE, CTL_EOL) != 0)
2035 		goto bad;
2036 
2037 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2038 		       CTLFLAG_PERMANENT,
2039 		       CTLTYPE_NODE, ifname,
2040 		       SYSCTL_DESCR("Interface controls"),
2041 		       NULL, 0, NULL, 0,
2042 		       CTL_CREATE, CTL_EOL) != 0)
2043 		goto bad;
2044 
2045 	if (sysctl_createv(clog, 0, &rnode, &rnode,
2046 		       CTLFLAG_PERMANENT,
2047 		       CTLTYPE_NODE, "sndq",
2048 		       SYSCTL_DESCR("Interface output queue controls"),
2049 		       NULL, 0, NULL, 0,
2050 		       CTL_CREATE, CTL_EOL) != 0)
2051 		goto bad;
2052 
2053 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2054 		       CTLFLAG_PERMANENT,
2055 		       CTLTYPE_INT, "len",
2056 		       SYSCTL_DESCR("Current output queue length"),
2057 		       NULL, 0, &ifq->ifq_len, 0,
2058 		       CTL_CREATE, CTL_EOL) != 0)
2059 		goto bad;
2060 
2061 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2062 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2063 		       CTLTYPE_INT, "maxlen",
2064 		       SYSCTL_DESCR("Maximum allowed output queue length"),
2065 		       NULL, 0, &ifq->ifq_maxlen, 0,
2066 		       CTL_CREATE, CTL_EOL) != 0)
2067 		goto bad;
2068 
2069 	if (sysctl_createv(clog, 0, &rnode, &cnode,
2070 		       CTLFLAG_PERMANENT,
2071 		       CTLTYPE_INT, "drops",
2072 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
2073 		       NULL, 0, &ifq->ifq_drops, 0,
2074 		       CTL_CREATE, CTL_EOL) != 0)
2075 		goto bad;
2076 
2077 	return;
2078 bad:
2079 	printf("%s: could not attach sysctl nodes\n", ifname);
2080 	return;
2081 }
2082 
2083 #if defined(INET) || defined(INET6)
2084 static void
2085 sysctl_net_ifq_setup(struct sysctllog **clog,
2086 		     int pf, const char *pfname,
2087 		     int ipn, const char *ipname,
2088 		     int qid, struct ifqueue *ifq)
2089 {
2090 
2091 	sysctl_createv(clog, 0, NULL, NULL,
2092 		       CTLFLAG_PERMANENT,
2093 		       CTLTYPE_NODE, "net", NULL,
2094 		       NULL, 0, NULL, 0,
2095 		       CTL_NET, CTL_EOL);
2096 	sysctl_createv(clog, 0, NULL, NULL,
2097 		       CTLFLAG_PERMANENT,
2098 		       CTLTYPE_NODE, pfname, NULL,
2099 		       NULL, 0, NULL, 0,
2100 		       CTL_NET, pf, CTL_EOL);
2101 	sysctl_createv(clog, 0, NULL, NULL,
2102 		       CTLFLAG_PERMANENT,
2103 		       CTLTYPE_NODE, ipname, NULL,
2104 		       NULL, 0, NULL, 0,
2105 		       CTL_NET, pf, ipn, CTL_EOL);
2106 	sysctl_createv(clog, 0, NULL, NULL,
2107 		       CTLFLAG_PERMANENT,
2108 		       CTLTYPE_NODE, "ifq",
2109 		       SYSCTL_DESCR("Protocol input queue controls"),
2110 		       NULL, 0, NULL, 0,
2111 		       CTL_NET, pf, ipn, qid, CTL_EOL);
2112 
2113 	sysctl_createv(clog, 0, NULL, NULL,
2114 		       CTLFLAG_PERMANENT,
2115 		       CTLTYPE_INT, "len",
2116 		       SYSCTL_DESCR("Current input queue length"),
2117 		       NULL, 0, &ifq->ifq_len, 0,
2118 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
2119 	sysctl_createv(clog, 0, NULL, NULL,
2120 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2121 		       CTLTYPE_INT, "maxlen",
2122 		       SYSCTL_DESCR("Maximum allowed input queue length"),
2123 		       NULL, 0, &ifq->ifq_maxlen, 0,
2124 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
2125 #ifdef notyet
2126 	sysctl_createv(clog, 0, NULL, NULL,
2127 		       CTLFLAG_PERMANENT,
2128 		       CTLTYPE_INT, "peak",
2129 		       SYSCTL_DESCR("Highest input queue length"),
2130 		       NULL, 0, &ifq->ifq_peak, 0,
2131 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
2132 #endif
2133 	sysctl_createv(clog, 0, NULL, NULL,
2134 		       CTLFLAG_PERMANENT,
2135 		       CTLTYPE_INT, "drops",
2136 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
2137 		       NULL, 0, &ifq->ifq_drops, 0,
2138 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
2139 }
2140 #endif /* INET || INET6 */
2141