xref: /freebsd/sys/net/if.c (revision 7aa383846770374466b1dcb2cefd71bde9acf463)
1 /*-
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_carp.h"
37 #include "opt_ddb.h"
38 
39 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/conf.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #ifdef DDB
67 #include <ddb/ddb.h>
68 #endif
69 
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 /*XXX*/
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #ifdef INET6
85 #include <netinet6/in6_var.h>
86 #include <netinet6/in6_ifattach.h>
87 #endif
88 #endif
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #endif
92 #if defined(INET) || defined(INET6)
93 #ifdef DEV_CARP
94 #include <netinet/ip_carp.h>
95 #endif
96 #endif
97 
98 #include <security/mac/mac_framework.h>
99 
100 struct ifindex_entry {
101 	struct  ifnet *ife_ifnet;
102 };
103 
104 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
105 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
106 
107 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
108 SYSCTL_UINT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
109     &ifqmaxlen, 0, "max send queue size");
110 
111 /* Log link state change events */
112 static int log_link_state_change = 1;
113 
114 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
115 	&log_link_state_change, 0,
116 	"log interface link state change events");
117 
118 /* Interface description */
119 static unsigned int ifdescr_maxlen = 1024;
120 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
121 	&ifdescr_maxlen, 0,
122 	"administrative maximum length for interface description");
123 
124 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
125 
126 /* global sx for non-critical path ifdescr */
127 static struct sx ifdescr_sx;
128 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
129 
130 void	(*bstp_linkstate_p)(struct ifnet *ifp, int state);
131 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
132 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
133 
134 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
135 
136 /*
137  * XXX: Style; these should be sorted alphabetically, and unprototyped
138  * static functions should be prototyped. Currently they are sorted by
139  * declaration order.
140  */
141 static void	if_attachdomain(void *);
142 static void	if_attachdomain1(struct ifnet *);
143 static int	ifconf(u_long, caddr_t);
144 static void	if_freemulti(struct ifmultiaddr *);
145 static void	if_init(void *);
146 static void	if_grow(void);
147 static void	if_route(struct ifnet *, int flag, int fam);
148 static int	if_setflag(struct ifnet *, int, int, int *, int);
149 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
150 static void	if_unroute(struct ifnet *, int flag, int fam);
151 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
152 static int	if_rtdel(struct radix_node *, void *);
153 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
154 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
155 static void	do_link_state_change(void *, int);
156 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
157 static int	if_getgroupmembers(struct ifgroupreq *);
158 static void	if_delgroups(struct ifnet *);
159 static void	if_attach_internal(struct ifnet *, int);
160 static void	if_detach_internal(struct ifnet *, int);
161 
162 #ifdef INET6
163 /*
164  * XXX: declare here to avoid to include many inet6 related files..
165  * should be more generalized?
166  */
167 extern void	nd6_setmtu(struct ifnet *);
168 #endif
169 
170 VNET_DEFINE(int, if_index);
171 int	ifqmaxlen = IFQ_MAXLEN;
172 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
173 VNET_DEFINE(struct ifgrouphead, ifg_head);
174 
175 static VNET_DEFINE(int, if_indexlim) = 8;
176 
177 /* Table of ifnet by index. */
178 static VNET_DEFINE(struct ifindex_entry *, ifindex_table);
179 
180 #define	V_if_indexlim		VNET(if_indexlim)
181 #define	V_ifindex_table		VNET(ifindex_table)
182 
183 /*
184  * The global network interface list (V_ifnet) and related state (such as
185  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
186  * an rwlock.  Either may be acquired shared to stablize the list, but both
187  * must be acquired writable to modify the list.  This model allows us to
188  * both stablize the interface list during interrupt thread processing, but
189  * also to stablize it over long-running ioctls, without introducing priority
190  * inversions and deadlocks.
191  */
192 struct rwlock ifnet_rwlock;
193 struct sx ifnet_sxlock;
194 
195 /*
196  * The allocation of network interfaces is a rather non-atomic affair; we
197  * need to select an index before we are ready to expose the interface for
198  * use, so will use this pointer value to indicate reservation.
199  */
200 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
201 
202 static	if_com_alloc_t *if_com_alloc[256];
203 static	if_com_free_t *if_com_free[256];
204 
205 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
206 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
207 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
208 
209 struct ifnet *
210 ifnet_byindex_locked(u_short idx)
211 {
212 
213 	if (idx > V_if_index)
214 		return (NULL);
215 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
216 		return (NULL);
217 	return (V_ifindex_table[idx].ife_ifnet);
218 }
219 
220 struct ifnet *
221 ifnet_byindex(u_short idx)
222 {
223 	struct ifnet *ifp;
224 
225 	IFNET_RLOCK_NOSLEEP();
226 	ifp = ifnet_byindex_locked(idx);
227 	IFNET_RUNLOCK_NOSLEEP();
228 	return (ifp);
229 }
230 
231 struct ifnet *
232 ifnet_byindex_ref(u_short idx)
233 {
234 	struct ifnet *ifp;
235 
236 	IFNET_RLOCK_NOSLEEP();
237 	ifp = ifnet_byindex_locked(idx);
238 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
239 		IFNET_RUNLOCK_NOSLEEP();
240 		return (NULL);
241 	}
242 	if_ref(ifp);
243 	IFNET_RUNLOCK_NOSLEEP();
244 	return (ifp);
245 }
246 
247 /*
248  * Allocate an ifindex array entry; return 0 on success or an error on
249  * failure.
250  */
251 static int
252 ifindex_alloc_locked(u_short *idxp)
253 {
254 	u_short idx;
255 
256 	IFNET_WLOCK_ASSERT();
257 
258 	/*
259 	 * Try to find an empty slot below V_if_index.  If we fail, take the
260 	 * next slot.
261 	 */
262 	for (idx = 1; idx <= V_if_index; idx++) {
263 		if (V_ifindex_table[idx].ife_ifnet == NULL)
264 			break;
265 	}
266 
267 	/* Catch if_index overflow. */
268 	if (idx < 1)
269 		return (ENOSPC);
270 	if (idx > V_if_index)
271 		V_if_index = idx;
272 	if (V_if_index >= V_if_indexlim)
273 		if_grow();
274 	*idxp = idx;
275 	return (0);
276 }
277 
278 static void
279 ifindex_free_locked(u_short idx)
280 {
281 
282 	IFNET_WLOCK_ASSERT();
283 
284 	V_ifindex_table[idx].ife_ifnet = NULL;
285 	while (V_if_index > 0 &&
286 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
287 		V_if_index--;
288 }
289 
290 static void
291 ifindex_free(u_short idx)
292 {
293 
294 	IFNET_WLOCK();
295 	ifindex_free_locked(idx);
296 	IFNET_WUNLOCK();
297 }
298 
299 static void
300 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
301 {
302 
303 	IFNET_WLOCK_ASSERT();
304 
305 	V_ifindex_table[idx].ife_ifnet = ifp;
306 }
307 
308 static void
309 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
310 {
311 
312 	IFNET_WLOCK();
313 	ifnet_setbyindex_locked(idx, ifp);
314 	IFNET_WUNLOCK();
315 }
316 
317 struct ifaddr *
318 ifaddr_byindex(u_short idx)
319 {
320 	struct ifaddr *ifa;
321 
322 	IFNET_RLOCK_NOSLEEP();
323 	ifa = ifnet_byindex_locked(idx)->if_addr;
324 	if (ifa != NULL)
325 		ifa_ref(ifa);
326 	IFNET_RUNLOCK_NOSLEEP();
327 	return (ifa);
328 }
329 
330 /*
331  * Network interface utility routines.
332  *
333  * Routines with ifa_ifwith* names take sockaddr *'s as
334  * parameters.
335  */
336 
337 static void
338 vnet_if_init(const void *unused __unused)
339 {
340 
341 	TAILQ_INIT(&V_ifnet);
342 	TAILQ_INIT(&V_ifg_head);
343 	if_grow();				/* create initial table */
344 	vnet_if_clone_init();
345 }
346 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
347     NULL);
348 
349 /* ARGSUSED*/
350 static void
351 if_init(void *dummy __unused)
352 {
353 
354 	IFNET_LOCK_INIT();
355 	if_clone_init();
356 }
357 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
358 
359 
360 #ifdef VIMAGE
361 static void
362 vnet_if_uninit(const void *unused __unused)
363 {
364 
365 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
366 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
367 
368 	free((caddr_t)V_ifindex_table, M_IFNET);
369 }
370 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
371     vnet_if_uninit, NULL);
372 #endif
373 
374 static void
375 if_grow(void)
376 {
377 	u_int n;
378 	struct ifindex_entry *e;
379 
380 	V_if_indexlim <<= 1;
381 	n = V_if_indexlim * sizeof(*e);
382 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
383 	if (V_ifindex_table != NULL) {
384 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
385 		free((caddr_t)V_ifindex_table, M_IFNET);
386 	}
387 	V_ifindex_table = e;
388 }
389 
390 /*
391  * Allocate a struct ifnet and an index for an interface.  A layer 2
392  * common structure will also be allocated if an allocation routine is
393  * registered for the passed type.
394  */
395 struct ifnet *
396 if_alloc(u_char type)
397 {
398 	struct ifnet *ifp;
399 	u_short idx;
400 
401 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
402 	IFNET_WLOCK();
403 	if (ifindex_alloc_locked(&idx) != 0) {
404 		IFNET_WUNLOCK();
405 		free(ifp, M_IFNET);
406 		return (NULL);
407 	}
408 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
409 	IFNET_WUNLOCK();
410 	ifp->if_index = idx;
411 	ifp->if_type = type;
412 	ifp->if_alloctype = type;
413 	if (if_com_alloc[type] != NULL) {
414 		ifp->if_l2com = if_com_alloc[type](type, ifp);
415 		if (ifp->if_l2com == NULL) {
416 			free(ifp, M_IFNET);
417 			ifindex_free(idx);
418 			return (NULL);
419 		}
420 	}
421 
422 	IF_ADDR_LOCK_INIT(ifp);
423 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
424 	ifp->if_afdata_initialized = 0;
425 	IF_AFDATA_LOCK_INIT(ifp);
426 	TAILQ_INIT(&ifp->if_addrhead);
427 	TAILQ_INIT(&ifp->if_prefixhead);
428 	TAILQ_INIT(&ifp->if_multiaddrs);
429 	TAILQ_INIT(&ifp->if_groups);
430 #ifdef MAC
431 	mac_ifnet_init(ifp);
432 #endif
433 	ifq_init(&ifp->if_snd, ifp);
434 
435 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
436 	ifnet_setbyindex(ifp->if_index, ifp);
437 	return (ifp);
438 }
439 
440 /*
441  * Do the actual work of freeing a struct ifnet, associated index, and layer
442  * 2 common structure.  This call is made when the last reference to an
443  * interface is released.
444  */
445 static void
446 if_free_internal(struct ifnet *ifp)
447 {
448 
449 	KASSERT((ifp->if_flags & IFF_DYING),
450 	    ("if_free_internal: interface not dying"));
451 
452 	IFNET_WLOCK();
453 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
454 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
455 
456 	ifindex_free_locked(ifp->if_index);
457 	IFNET_WUNLOCK();
458 
459 	if (if_com_free[ifp->if_alloctype] != NULL)
460 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
461 		    ifp->if_alloctype);
462 
463 #ifdef MAC
464 	mac_ifnet_destroy(ifp);
465 #endif /* MAC */
466 	if (ifp->if_description != NULL)
467 		free(ifp->if_description, M_IFDESCR);
468 	IF_AFDATA_DESTROY(ifp);
469 	IF_ADDR_LOCK_DESTROY(ifp);
470 	ifq_delete(&ifp->if_snd);
471 	free(ifp, M_IFNET);
472 }
473 
474 /*
475  * This version should only be called by intefaces that switch their type
476  * after calling if_alloc().  if_free_type() will go away again now that we
477  * have if_alloctype to cache the original allocation type.  For now, assert
478  * that they match, since we require that in practice.
479  */
480 void
481 if_free_type(struct ifnet *ifp, u_char type)
482 {
483 
484 	KASSERT(ifp->if_alloctype == type,
485 	    ("if_free_type: type (%d) != alloctype (%d)", type,
486 	    ifp->if_alloctype));
487 
488 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
489 	if (!refcount_release(&ifp->if_refcount))
490 		return;
491 	if_free_internal(ifp);
492 }
493 
494 /*
495  * This is the normal version of if_free(), used by device drivers to free a
496  * detached network interface.  The contents of if_free_type() will move into
497  * here when if_free_type() goes away.
498  */
499 void
500 if_free(struct ifnet *ifp)
501 {
502 
503 	if_free_type(ifp, ifp->if_alloctype);
504 }
505 
506 /*
507  * Interfaces to keep an ifnet type-stable despite the possibility of the
508  * driver calling if_free().  If there are additional references, we defer
509  * freeing the underlying data structure.
510  */
511 void
512 if_ref(struct ifnet *ifp)
513 {
514 
515 	/* We don't assert the ifnet list lock here, but arguably should. */
516 	refcount_acquire(&ifp->if_refcount);
517 }
518 
519 void
520 if_rele(struct ifnet *ifp)
521 {
522 
523 	if (!refcount_release(&ifp->if_refcount))
524 		return;
525 	if_free_internal(ifp);
526 }
527 
528 void
529 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
530 {
531 
532 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
533 
534 	if (ifq->ifq_maxlen == 0)
535 		ifq->ifq_maxlen = ifqmaxlen;
536 
537 	ifq->altq_type = 0;
538 	ifq->altq_disc = NULL;
539 	ifq->altq_flags &= ALTQF_CANTCHANGE;
540 	ifq->altq_tbr  = NULL;
541 	ifq->altq_ifp  = ifp;
542 }
543 
544 void
545 ifq_delete(struct ifaltq *ifq)
546 {
547 	mtx_destroy(&ifq->ifq_mtx);
548 }
549 
550 /*
551  * Perform generic interface initalization tasks and attach the interface
552  * to the list of "active" interfaces.  If vmove flag is set on entry
553  * to if_attach_internal(), perform only a limited subset of initialization
554  * tasks, given that we are moving from one vnet to another an ifnet which
555  * has already been fully initialized.
556  *
557  * XXX:
558  *  - The decision to return void and thus require this function to
559  *    succeed is questionable.
560  *  - We should probably do more sanity checking.  For instance we don't
561  *    do anything to insure if_xname is unique or non-empty.
562  */
563 void
564 if_attach(struct ifnet *ifp)
565 {
566 
567 	if_attach_internal(ifp, 0);
568 }
569 
570 static void
571 if_attach_internal(struct ifnet *ifp, int vmove)
572 {
573 	unsigned socksize, ifasize;
574 	int namelen, masklen;
575 	struct sockaddr_dl *sdl;
576 	struct ifaddr *ifa;
577 
578 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
579 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
580 		    ifp->if_xname);
581 
582 #ifdef VIMAGE
583 	ifp->if_vnet = curvnet;
584 	if (ifp->if_home_vnet == NULL)
585 		ifp->if_home_vnet = curvnet;
586 #endif
587 
588 	if_addgroup(ifp, IFG_ALL);
589 
590 	getmicrotime(&ifp->if_lastchange);
591 	ifp->if_data.ifi_epoch = time_uptime;
592 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
593 
594 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
595 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
596 	    ("transmit and qflush must both either be set or both be NULL"));
597 	if (ifp->if_transmit == NULL) {
598 		ifp->if_transmit = if_transmit;
599 		ifp->if_qflush = if_qflush;
600 	}
601 
602 	if (!vmove) {
603 #ifdef MAC
604 		mac_ifnet_create(ifp);
605 #endif
606 
607 		/*
608 		 * Create a Link Level name for this device.
609 		 */
610 		namelen = strlen(ifp->if_xname);
611 		/*
612 		 * Always save enough space for any possiable name so we
613 		 * can do a rename in place later.
614 		 */
615 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
616 		socksize = masklen + ifp->if_addrlen;
617 		if (socksize < sizeof(*sdl))
618 			socksize = sizeof(*sdl);
619 		socksize = roundup2(socksize, sizeof(long));
620 		ifasize = sizeof(*ifa) + 2 * socksize;
621 		ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
622 		ifa_init(ifa);
623 		sdl = (struct sockaddr_dl *)(ifa + 1);
624 		sdl->sdl_len = socksize;
625 		sdl->sdl_family = AF_LINK;
626 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
627 		sdl->sdl_nlen = namelen;
628 		sdl->sdl_index = ifp->if_index;
629 		sdl->sdl_type = ifp->if_type;
630 		ifp->if_addr = ifa;
631 		ifa->ifa_ifp = ifp;
632 		ifa->ifa_rtrequest = link_rtrequest;
633 		ifa->ifa_addr = (struct sockaddr *)sdl;
634 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
635 		ifa->ifa_netmask = (struct sockaddr *)sdl;
636 		sdl->sdl_len = masklen;
637 		while (namelen != 0)
638 			sdl->sdl_data[--namelen] = 0xff;
639 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
640 		/* Reliably crash if used uninitialized. */
641 		ifp->if_broadcastaddr = NULL;
642 	}
643 #ifdef VIMAGE
644 	else {
645 		/*
646 		 * Update the interface index in the link layer address
647 		 * of the interface.
648 		 */
649 		for (ifa = ifp->if_addr; ifa != NULL;
650 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
651 			if (ifa->ifa_addr->sa_family == AF_LINK) {
652 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
653 				sdl->sdl_index = ifp->if_index;
654 			}
655 		}
656 	}
657 #endif
658 
659 	IFNET_WLOCK();
660 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
661 #ifdef VIMAGE
662 	curvnet->vnet_ifcnt++;
663 #endif
664 	IFNET_WUNLOCK();
665 
666 	if (domain_init_status >= 2)
667 		if_attachdomain1(ifp);
668 
669 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
670 	if (IS_DEFAULT_VNET(curvnet))
671 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
672 
673 	/* Announce the interface. */
674 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
675 }
676 
677 static void
678 if_attachdomain(void *dummy)
679 {
680 	struct ifnet *ifp;
681 	int s;
682 
683 	s = splnet();
684 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
685 		if_attachdomain1(ifp);
686 	splx(s);
687 }
688 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
689     if_attachdomain, NULL);
690 
691 static void
692 if_attachdomain1(struct ifnet *ifp)
693 {
694 	struct domain *dp;
695 	int s;
696 
697 	s = splnet();
698 
699 	/*
700 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
701 	 * cannot lock ifp->if_afdata initialization, entirely.
702 	 */
703 	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
704 		splx(s);
705 		return;
706 	}
707 	if (ifp->if_afdata_initialized >= domain_init_status) {
708 		IF_AFDATA_UNLOCK(ifp);
709 		splx(s);
710 		printf("if_attachdomain called more than once on %s\n",
711 		    ifp->if_xname);
712 		return;
713 	}
714 	ifp->if_afdata_initialized = domain_init_status;
715 	IF_AFDATA_UNLOCK(ifp);
716 
717 	/* address family dependent data region */
718 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
719 	for (dp = domains; dp; dp = dp->dom_next) {
720 		if (dp->dom_ifattach)
721 			ifp->if_afdata[dp->dom_family] =
722 			    (*dp->dom_ifattach)(ifp);
723 	}
724 
725 	splx(s);
726 }
727 
728 /*
729  * Remove any unicast or broadcast network addresses from an interface.
730  */
731 void
732 if_purgeaddrs(struct ifnet *ifp)
733 {
734 	struct ifaddr *ifa, *next;
735 
736 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
737 		if (ifa->ifa_addr->sa_family == AF_LINK)
738 			continue;
739 #ifdef INET
740 		/* XXX: Ugly!! ad hoc just for INET */
741 		if (ifa->ifa_addr->sa_family == AF_INET) {
742 			struct ifaliasreq ifr;
743 
744 			bzero(&ifr, sizeof(ifr));
745 			ifr.ifra_addr = *ifa->ifa_addr;
746 			if (ifa->ifa_dstaddr)
747 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
748 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
749 			    NULL) == 0)
750 				continue;
751 		}
752 #endif /* INET */
753 #ifdef INET6
754 		if (ifa->ifa_addr->sa_family == AF_INET6) {
755 			in6_purgeaddr(ifa);
756 			/* ifp_addrhead is already updated */
757 			continue;
758 		}
759 #endif /* INET6 */
760 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
761 		ifa_free(ifa);
762 	}
763 }
764 
765 /*
766  * Remove any multicast network addresses from an interface when an ifnet
767  * is going away.
768  */
769 static void
770 if_purgemaddrs(struct ifnet *ifp)
771 {
772 	struct ifmultiaddr *ifma;
773 	struct ifmultiaddr *next;
774 
775 	IF_ADDR_LOCK(ifp);
776 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
777 		if_delmulti_locked(ifp, ifma, 1);
778 	IF_ADDR_UNLOCK(ifp);
779 }
780 
781 /*
782  * Detach an interface, removing it from the list of "active" interfaces.
783  * If vmove flag is set on entry to if_detach_internal(), perform only a
784  * limited subset of cleanup tasks, given that we are moving an ifnet from
785  * one vnet to another, where it must be fully operational.
786  *
787  * XXXRW: There are some significant questions about event ordering, and
788  * how to prevent things from starting to use the interface during detach.
789  */
790 void
791 if_detach(struct ifnet *ifp)
792 {
793 
794 	if_detach_internal(ifp, 0);
795 }
796 
797 static void
798 if_detach_internal(struct ifnet *ifp, int vmove)
799 {
800 	struct ifaddr *ifa;
801 	struct radix_node_head	*rnh;
802 	int i, j;
803 	struct domain *dp;
804  	struct ifnet *iter;
805  	int found = 0;
806 
807 	IFNET_WLOCK();
808 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
809 		if (iter == ifp) {
810 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
811 			found = 1;
812 			break;
813 		}
814 #ifdef VIMAGE
815 	if (found)
816 		curvnet->vnet_ifcnt--;
817 #endif
818 	IFNET_WUNLOCK();
819 	if (!found) {
820 		if (vmove)
821 			panic("%s: ifp=%p not on the ifnet tailq %p",
822 			    __func__, ifp, &V_ifnet);
823 		else
824 			return; /* XXX this should panic as well? */
825 	}
826 
827 	/*
828 	 * Remove/wait for pending events.
829 	 */
830 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
831 
832 	/*
833 	 * Remove routes and flush queues.
834 	 */
835 	if_down(ifp);
836 #ifdef ALTQ
837 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
838 		altq_disable(&ifp->if_snd);
839 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
840 		altq_detach(&ifp->if_snd);
841 #endif
842 
843 	if_purgeaddrs(ifp);
844 
845 #ifdef INET
846 	in_ifdetach(ifp);
847 #endif
848 
849 #ifdef INET6
850 	/*
851 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
852 	 * before removing routing entries below, since IPv6 interface direct
853 	 * routes are expected to be removed by the IPv6-specific kernel API.
854 	 * Otherwise, the kernel will detect some inconsistency and bark it.
855 	 */
856 	in6_ifdetach(ifp);
857 #endif
858 	if_purgemaddrs(ifp);
859 
860 	if (!vmove) {
861 		/*
862 		 * Prevent further calls into the device driver via ifnet.
863 		 */
864 		if_dead(ifp);
865 
866 		/*
867 		 * Remove link ifaddr pointer and maybe decrement if_index.
868 		 * Clean up all addresses.
869 		 */
870 		ifp->if_addr = NULL;
871 
872 		/* We can now free link ifaddr. */
873 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
874 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
875 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
876 			ifa_free(ifa);
877 		}
878 	}
879 
880 	/*
881 	 * Delete all remaining routes using this interface
882 	 * Unfortuneatly the only way to do this is to slog through
883 	 * the entire routing table looking for routes which point
884 	 * to this interface...oh well...
885 	 */
886 	for (i = 1; i <= AF_MAX; i++) {
887 		for (j = 0; j < rt_numfibs; j++) {
888 			rnh = rt_tables_get_rnh(j, i);
889 			if (rnh == NULL)
890 				continue;
891 			RADIX_NODE_HEAD_LOCK(rnh);
892 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
893 			RADIX_NODE_HEAD_UNLOCK(rnh);
894 		}
895 	}
896 
897 	/* Announce that the interface is gone. */
898 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
899 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
900 	if (IS_DEFAULT_VNET(curvnet))
901 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
902 	if_delgroups(ifp);
903 
904 	/*
905 	 * We cannot hold the lock over dom_ifdetach calls as they might
906 	 * sleep, for example trying to drain a callout, thus open up the
907 	 * theoretical race with re-attaching.
908 	 */
909 	IF_AFDATA_LOCK(ifp);
910 	i = ifp->if_afdata_initialized;
911 	ifp->if_afdata_initialized = 0;
912 	IF_AFDATA_UNLOCK(ifp);
913 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
914 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
915 			(*dp->dom_ifdetach)(ifp,
916 			    ifp->if_afdata[dp->dom_family]);
917 	}
918 }
919 
920 #ifdef VIMAGE
921 /*
922  * if_vmove() performs a limited version of if_detach() in current
923  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
924  * An attempt is made to shrink if_index in current vnet, find an
925  * unused if_index in target vnet and calls if_grow() if necessary,
926  * and finally find an unused if_xname for the target vnet.
927  */
928 void
929 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
930 {
931 	u_short idx;
932 
933 	/*
934 	 * Detach from current vnet, but preserve LLADDR info, do not
935 	 * mark as dead etc. so that the ifnet can be reattached later.
936 	 */
937 	if_detach_internal(ifp, 1);
938 
939 	/*
940 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
941 	 * the if_index for that vnet if possible.
942 	 *
943 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
944 	 * or we'd lock on one vnet and unlock on another.
945 	 */
946 	IFNET_WLOCK();
947 	ifindex_free_locked(ifp->if_index);
948 
949 	/*
950 	 * Switch to the context of the target vnet.
951 	 */
952 	CURVNET_SET_QUIET(new_vnet);
953 
954 	if (ifindex_alloc_locked(&idx) != 0) {
955 		IFNET_WUNLOCK();
956 		panic("if_index overflow");
957 	}
958 	ifp->if_index = idx;
959 	ifnet_setbyindex_locked(ifp->if_index, ifp);
960 	IFNET_WUNLOCK();
961 
962 	if_attach_internal(ifp, 1);
963 
964 	CURVNET_RESTORE();
965 }
966 
967 /*
968  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
969  */
970 static int
971 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
972 {
973 	struct prison *pr;
974 	struct ifnet *difp;
975 
976 	/* Try to find the prison within our visibility. */
977 	sx_slock(&allprison_lock);
978 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
979 	sx_sunlock(&allprison_lock);
980 	if (pr == NULL)
981 		return (ENXIO);
982 	prison_hold_locked(pr);
983 	mtx_unlock(&pr->pr_mtx);
984 
985 	/* Do not try to move the iface from and to the same prison. */
986 	if (pr->pr_vnet == ifp->if_vnet) {
987 		prison_free(pr);
988 		return (EEXIST);
989 	}
990 
991 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
992 	/* XXX Lock interfaces to avoid races. */
993 	CURVNET_SET_QUIET(pr->pr_vnet);
994 	difp = ifunit(ifname);
995 	CURVNET_RESTORE();
996 	if (difp != NULL) {
997 		prison_free(pr);
998 		return (EEXIST);
999 	}
1000 
1001 	/* Move the interface into the child jail/vnet. */
1002 	if_vmove(ifp, pr->pr_vnet);
1003 
1004 	/* Report the new if_xname back to the userland. */
1005 	sprintf(ifname, "%s", ifp->if_xname);
1006 
1007 	prison_free(pr);
1008 	return (0);
1009 }
1010 
1011 static int
1012 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1013 {
1014 	struct prison *pr;
1015 	struct vnet *vnet_dst;
1016 	struct ifnet *ifp;
1017 
1018 	/* Try to find the prison within our visibility. */
1019 	sx_slock(&allprison_lock);
1020 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1021 	sx_sunlock(&allprison_lock);
1022 	if (pr == NULL)
1023 		return (ENXIO);
1024 	prison_hold_locked(pr);
1025 	mtx_unlock(&pr->pr_mtx);
1026 
1027 	/* Make sure the named iface exists in the source prison/vnet. */
1028 	CURVNET_SET(pr->pr_vnet);
1029 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1030 	if (ifp == NULL) {
1031 		CURVNET_RESTORE();
1032 		prison_free(pr);
1033 		return (ENXIO);
1034 	}
1035 
1036 	/* Do not try to move the iface from and to the same prison. */
1037 	vnet_dst = TD_TO_VNET(td);
1038 	if (vnet_dst == ifp->if_vnet) {
1039 		CURVNET_RESTORE();
1040 		prison_free(pr);
1041 		return (EEXIST);
1042 	}
1043 
1044 	/* Get interface back from child jail/vnet. */
1045 	if_vmove(ifp, vnet_dst);
1046 	CURVNET_RESTORE();
1047 
1048 	/* Report the new if_xname back to the userland. */
1049 	sprintf(ifname, "%s", ifp->if_xname);
1050 
1051 	prison_free(pr);
1052 	return (0);
1053 }
1054 #endif /* VIMAGE */
1055 
1056 /*
1057  * Add a group to an interface
1058  */
1059 int
1060 if_addgroup(struct ifnet *ifp, const char *groupname)
1061 {
1062 	struct ifg_list		*ifgl;
1063 	struct ifg_group	*ifg = NULL;
1064 	struct ifg_member	*ifgm;
1065 
1066 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1067 	    groupname[strlen(groupname) - 1] <= '9')
1068 		return (EINVAL);
1069 
1070 	IFNET_WLOCK();
1071 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1072 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1073 			IFNET_WUNLOCK();
1074 			return (EEXIST);
1075 		}
1076 
1077 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1078 	    M_NOWAIT)) == NULL) {
1079 	    	IFNET_WUNLOCK();
1080 		return (ENOMEM);
1081 	}
1082 
1083 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1084 	    M_TEMP, M_NOWAIT)) == NULL) {
1085 		free(ifgl, M_TEMP);
1086 		IFNET_WUNLOCK();
1087 		return (ENOMEM);
1088 	}
1089 
1090 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1091 		if (!strcmp(ifg->ifg_group, groupname))
1092 			break;
1093 
1094 	if (ifg == NULL) {
1095 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1096 		    M_TEMP, M_NOWAIT)) == NULL) {
1097 			free(ifgl, M_TEMP);
1098 			free(ifgm, M_TEMP);
1099 			IFNET_WUNLOCK();
1100 			return (ENOMEM);
1101 		}
1102 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1103 		ifg->ifg_refcnt = 0;
1104 		TAILQ_INIT(&ifg->ifg_members);
1105 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1106 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1107 	}
1108 
1109 	ifg->ifg_refcnt++;
1110 	ifgl->ifgl_group = ifg;
1111 	ifgm->ifgm_ifp = ifp;
1112 
1113 	IF_ADDR_LOCK(ifp);
1114 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1115 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1116 	IF_ADDR_UNLOCK(ifp);
1117 
1118 	IFNET_WUNLOCK();
1119 
1120 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1121 
1122 	return (0);
1123 }
1124 
1125 /*
1126  * Remove a group from an interface
1127  */
1128 int
1129 if_delgroup(struct ifnet *ifp, const char *groupname)
1130 {
1131 	struct ifg_list		*ifgl;
1132 	struct ifg_member	*ifgm;
1133 
1134 	IFNET_WLOCK();
1135 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1136 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1137 			break;
1138 	if (ifgl == NULL) {
1139 		IFNET_WUNLOCK();
1140 		return (ENOENT);
1141 	}
1142 
1143 	IF_ADDR_LOCK(ifp);
1144 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1145 	IF_ADDR_UNLOCK(ifp);
1146 
1147 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1148 		if (ifgm->ifgm_ifp == ifp)
1149 			break;
1150 
1151 	if (ifgm != NULL) {
1152 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1153 		free(ifgm, M_TEMP);
1154 	}
1155 
1156 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1157 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1158 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1159 		free(ifgl->ifgl_group, M_TEMP);
1160 	}
1161 	IFNET_WUNLOCK();
1162 
1163 	free(ifgl, M_TEMP);
1164 
1165 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1166 
1167 	return (0);
1168 }
1169 
1170 /*
1171  * Remove an interface from all groups
1172  */
1173 static void
1174 if_delgroups(struct ifnet *ifp)
1175 {
1176 	struct ifg_list		*ifgl;
1177 	struct ifg_member	*ifgm;
1178 	char groupname[IFNAMSIZ];
1179 
1180 	IFNET_WLOCK();
1181 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1182 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1183 
1184 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1185 
1186 		IF_ADDR_LOCK(ifp);
1187 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1188 		IF_ADDR_UNLOCK(ifp);
1189 
1190 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1191 			if (ifgm->ifgm_ifp == ifp)
1192 				break;
1193 
1194 		if (ifgm != NULL) {
1195 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1196 			    ifgm_next);
1197 			free(ifgm, M_TEMP);
1198 		}
1199 
1200 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1201 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1202 			EVENTHANDLER_INVOKE(group_detach_event,
1203 			    ifgl->ifgl_group);
1204 			free(ifgl->ifgl_group, M_TEMP);
1205 		}
1206 		IFNET_WUNLOCK();
1207 
1208 		free(ifgl, M_TEMP);
1209 
1210 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1211 
1212 		IFNET_WLOCK();
1213 	}
1214 	IFNET_WUNLOCK();
1215 }
1216 
1217 /*
1218  * Stores all groups from an interface in memory pointed
1219  * to by data
1220  */
1221 static int
1222 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1223 {
1224 	int			 len, error;
1225 	struct ifg_list		*ifgl;
1226 	struct ifg_req		 ifgrq, *ifgp;
1227 	struct ifgroupreq	*ifgr = data;
1228 
1229 	if (ifgr->ifgr_len == 0) {
1230 		IF_ADDR_LOCK(ifp);
1231 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1232 			ifgr->ifgr_len += sizeof(struct ifg_req);
1233 		IF_ADDR_UNLOCK(ifp);
1234 		return (0);
1235 	}
1236 
1237 	len = ifgr->ifgr_len;
1238 	ifgp = ifgr->ifgr_groups;
1239 	/* XXX: wire */
1240 	IF_ADDR_LOCK(ifp);
1241 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1242 		if (len < sizeof(ifgrq)) {
1243 			IF_ADDR_UNLOCK(ifp);
1244 			return (EINVAL);
1245 		}
1246 		bzero(&ifgrq, sizeof ifgrq);
1247 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1248 		    sizeof(ifgrq.ifgrq_group));
1249 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1250 		    	IF_ADDR_UNLOCK(ifp);
1251 			return (error);
1252 		}
1253 		len -= sizeof(ifgrq);
1254 		ifgp++;
1255 	}
1256 	IF_ADDR_UNLOCK(ifp);
1257 
1258 	return (0);
1259 }
1260 
1261 /*
1262  * Stores all members of a group in memory pointed to by data
1263  */
1264 static int
1265 if_getgroupmembers(struct ifgroupreq *data)
1266 {
1267 	struct ifgroupreq	*ifgr = data;
1268 	struct ifg_group	*ifg;
1269 	struct ifg_member	*ifgm;
1270 	struct ifg_req		 ifgrq, *ifgp;
1271 	int			 len, error;
1272 
1273 	IFNET_RLOCK();
1274 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1275 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1276 			break;
1277 	if (ifg == NULL) {
1278 		IFNET_RUNLOCK();
1279 		return (ENOENT);
1280 	}
1281 
1282 	if (ifgr->ifgr_len == 0) {
1283 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1284 			ifgr->ifgr_len += sizeof(ifgrq);
1285 		IFNET_RUNLOCK();
1286 		return (0);
1287 	}
1288 
1289 	len = ifgr->ifgr_len;
1290 	ifgp = ifgr->ifgr_groups;
1291 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1292 		if (len < sizeof(ifgrq)) {
1293 			IFNET_RUNLOCK();
1294 			return (EINVAL);
1295 		}
1296 		bzero(&ifgrq, sizeof ifgrq);
1297 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1298 		    sizeof(ifgrq.ifgrq_member));
1299 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1300 			IFNET_RUNLOCK();
1301 			return (error);
1302 		}
1303 		len -= sizeof(ifgrq);
1304 		ifgp++;
1305 	}
1306 	IFNET_RUNLOCK();
1307 
1308 	return (0);
1309 }
1310 
1311 /*
1312  * Delete Routes for a Network Interface
1313  *
1314  * Called for each routing entry via the rnh->rnh_walktree() call above
1315  * to delete all route entries referencing a detaching network interface.
1316  *
1317  * Arguments:
1318  *	rn	pointer to node in the routing table
1319  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1320  *
1321  * Returns:
1322  *	0	successful
1323  *	errno	failed - reason indicated
1324  *
1325  */
1326 static int
1327 if_rtdel(struct radix_node *rn, void *arg)
1328 {
1329 	struct rtentry	*rt = (struct rtentry *)rn;
1330 	struct ifnet	*ifp = arg;
1331 	int		err;
1332 
1333 	if (rt->rt_ifp == ifp) {
1334 
1335 		/*
1336 		 * Protect (sorta) against walktree recursion problems
1337 		 * with cloned routes
1338 		 */
1339 		if ((rt->rt_flags & RTF_UP) == 0)
1340 			return (0);
1341 
1342 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1343 				rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
1344 				(struct rtentry **) NULL, rt->rt_fibnum);
1345 		if (err) {
1346 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1347 		}
1348 	}
1349 
1350 	return (0);
1351 }
1352 
1353 /*
1354  * Wrapper functions for struct ifnet address list locking macros.  These are
1355  * used by kernel modules to avoid encoding programming interface or binary
1356  * interface assumptions that may be violated when kernel-internal locking
1357  * approaches change.
1358  */
1359 void
1360 if_addr_rlock(struct ifnet *ifp)
1361 {
1362 
1363 	IF_ADDR_LOCK(ifp);
1364 }
1365 
1366 void
1367 if_addr_runlock(struct ifnet *ifp)
1368 {
1369 
1370 	IF_ADDR_UNLOCK(ifp);
1371 }
1372 
1373 void
1374 if_maddr_rlock(struct ifnet *ifp)
1375 {
1376 
1377 	IF_ADDR_LOCK(ifp);
1378 }
1379 
1380 void
1381 if_maddr_runlock(struct ifnet *ifp)
1382 {
1383 
1384 	IF_ADDR_UNLOCK(ifp);
1385 }
1386 
1387 /*
1388  * Reference count functions for ifaddrs.
1389  */
1390 void
1391 ifa_init(struct ifaddr *ifa)
1392 {
1393 
1394 	mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1395 	refcount_init(&ifa->ifa_refcnt, 1);
1396 }
1397 
1398 void
1399 ifa_ref(struct ifaddr *ifa)
1400 {
1401 
1402 	refcount_acquire(&ifa->ifa_refcnt);
1403 }
1404 
1405 void
1406 ifa_free(struct ifaddr *ifa)
1407 {
1408 
1409 	if (refcount_release(&ifa->ifa_refcnt)) {
1410 		mtx_destroy(&ifa->ifa_mtx);
1411 		free(ifa, M_IFADDR);
1412 	}
1413 }
1414 
1415 int
1416 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1417 {
1418 	int error = 0;
1419 	struct rtentry *rt = NULL;
1420 	struct rt_addrinfo info;
1421 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1422 
1423 	bzero(&info, sizeof(info));
1424 	info.rti_ifp = V_loif;
1425 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1426 	info.rti_info[RTAX_DST] = ia;
1427 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1428 	error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1429 
1430 	if (error == 0 && rt != NULL) {
1431 		RT_LOCK(rt);
1432 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1433 			ifa->ifa_ifp->if_type;
1434 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1435 			ifa->ifa_ifp->if_index;
1436 		RT_REMREF(rt);
1437 		RT_UNLOCK(rt);
1438 	} else if (error != 0)
1439 		log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1440 
1441 	return (error);
1442 }
1443 
1444 int
1445 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1446 {
1447 	int error = 0;
1448 	struct rt_addrinfo info;
1449 	struct sockaddr_dl null_sdl;
1450 
1451 	bzero(&null_sdl, sizeof(null_sdl));
1452 	null_sdl.sdl_len = sizeof(null_sdl);
1453 	null_sdl.sdl_family = AF_LINK;
1454 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1455 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1456 	bzero(&info, sizeof(info));
1457 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1458 	info.rti_info[RTAX_DST] = ia;
1459 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1460 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1461 
1462 	if (error != 0)
1463 		log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1464 
1465 	return (error);
1466 }
1467 
1468 /*
1469  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1470  * structs used to represent other address families, it is necessary
1471  * to perform a different comparison.
1472  */
1473 
1474 #define	sa_equal(a1, a2)	\
1475 	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1476 
1477 #define	sa_dl_equal(a1, a2)	\
1478 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1479 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1480 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1481 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1482 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1483 
1484 /*
1485  * Locate an interface based on a complete address.
1486  */
1487 /*ARGSUSED*/
1488 static struct ifaddr *
1489 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1490 {
1491 	struct ifnet *ifp;
1492 	struct ifaddr *ifa;
1493 
1494 	IFNET_RLOCK_NOSLEEP();
1495 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1496 		IF_ADDR_LOCK(ifp);
1497 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1498 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1499 				continue;
1500 			if (sa_equal(addr, ifa->ifa_addr)) {
1501 				if (getref)
1502 					ifa_ref(ifa);
1503 				IF_ADDR_UNLOCK(ifp);
1504 				goto done;
1505 			}
1506 			/* IP6 doesn't have broadcast */
1507 			if ((ifp->if_flags & IFF_BROADCAST) &&
1508 			    ifa->ifa_broadaddr &&
1509 			    ifa->ifa_broadaddr->sa_len != 0 &&
1510 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1511 				if (getref)
1512 					ifa_ref(ifa);
1513 				IF_ADDR_UNLOCK(ifp);
1514 				goto done;
1515 			}
1516 		}
1517 		IF_ADDR_UNLOCK(ifp);
1518 	}
1519 	ifa = NULL;
1520 done:
1521 	IFNET_RUNLOCK_NOSLEEP();
1522 	return (ifa);
1523 }
1524 
1525 struct ifaddr *
1526 ifa_ifwithaddr(struct sockaddr *addr)
1527 {
1528 
1529 	return (ifa_ifwithaddr_internal(addr, 1));
1530 }
1531 
1532 int
1533 ifa_ifwithaddr_check(struct sockaddr *addr)
1534 {
1535 
1536 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1537 }
1538 
1539 /*
1540  * Locate an interface based on the broadcast address.
1541  */
1542 /* ARGSUSED */
1543 struct ifaddr *
1544 ifa_ifwithbroadaddr(struct sockaddr *addr)
1545 {
1546 	struct ifnet *ifp;
1547 	struct ifaddr *ifa;
1548 
1549 	IFNET_RLOCK_NOSLEEP();
1550 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1551 		IF_ADDR_LOCK(ifp);
1552 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1553 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1554 				continue;
1555 			if ((ifp->if_flags & IFF_BROADCAST) &&
1556 			    ifa->ifa_broadaddr &&
1557 			    ifa->ifa_broadaddr->sa_len != 0 &&
1558 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1559 				ifa_ref(ifa);
1560 				IF_ADDR_UNLOCK(ifp);
1561 				goto done;
1562 			}
1563 		}
1564 		IF_ADDR_UNLOCK(ifp);
1565 	}
1566 	ifa = NULL;
1567 done:
1568 	IFNET_RUNLOCK_NOSLEEP();
1569 	return (ifa);
1570 }
1571 
1572 /*
1573  * Locate the point to point interface with a given destination address.
1574  */
1575 /*ARGSUSED*/
1576 struct ifaddr *
1577 ifa_ifwithdstaddr(struct sockaddr *addr)
1578 {
1579 	struct ifnet *ifp;
1580 	struct ifaddr *ifa;
1581 
1582 	IFNET_RLOCK_NOSLEEP();
1583 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1584 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1585 			continue;
1586 		IF_ADDR_LOCK(ifp);
1587 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1588 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1589 				continue;
1590 			if (ifa->ifa_dstaddr != NULL &&
1591 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1592 				ifa_ref(ifa);
1593 				IF_ADDR_UNLOCK(ifp);
1594 				goto done;
1595 			}
1596 		}
1597 		IF_ADDR_UNLOCK(ifp);
1598 	}
1599 	ifa = NULL;
1600 done:
1601 	IFNET_RUNLOCK_NOSLEEP();
1602 	return (ifa);
1603 }
1604 
1605 /*
1606  * Find an interface on a specific network.  If many, choice
1607  * is most specific found.
1608  */
1609 struct ifaddr *
1610 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1611 {
1612 	struct ifnet *ifp;
1613 	struct ifaddr *ifa;
1614 	struct ifaddr *ifa_maybe = NULL;
1615 	u_int af = addr->sa_family;
1616 	char *addr_data = addr->sa_data, *cplim;
1617 
1618 	/*
1619 	 * AF_LINK addresses can be looked up directly by their index number,
1620 	 * so do that if we can.
1621 	 */
1622 	if (af == AF_LINK) {
1623 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1624 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1625 		return (ifaddr_byindex(sdl->sdl_index));
1626 	}
1627 
1628 	/*
1629 	 * Scan though each interface, looking for ones that have addresses
1630 	 * in this address family.  Maintain a reference on ifa_maybe once
1631 	 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1632 	 * when we move onto the next interface.
1633 	 */
1634 	IFNET_RLOCK_NOSLEEP();
1635 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1636 		IF_ADDR_LOCK(ifp);
1637 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1638 			char *cp, *cp2, *cp3;
1639 
1640 			if (ifa->ifa_addr->sa_family != af)
1641 next:				continue;
1642 			if (af == AF_INET &&
1643 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1644 				/*
1645 				 * This is a bit broken as it doesn't
1646 				 * take into account that the remote end may
1647 				 * be a single node in the network we are
1648 				 * looking for.
1649 				 * The trouble is that we don't know the
1650 				 * netmask for the remote end.
1651 				 */
1652 				if (ifa->ifa_dstaddr != NULL &&
1653 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1654 					ifa_ref(ifa);
1655 					IF_ADDR_UNLOCK(ifp);
1656 					goto done;
1657 				}
1658 			} else {
1659 				/*
1660 				 * if we have a special address handler,
1661 				 * then use it instead of the generic one.
1662 				 */
1663 				if (ifa->ifa_claim_addr) {
1664 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1665 						ifa_ref(ifa);
1666 						IF_ADDR_UNLOCK(ifp);
1667 						goto done;
1668 					}
1669 					continue;
1670 				}
1671 
1672 				/*
1673 				 * Scan all the bits in the ifa's address.
1674 				 * If a bit dissagrees with what we are
1675 				 * looking for, mask it with the netmask
1676 				 * to see if it really matters.
1677 				 * (A byte at a time)
1678 				 */
1679 				if (ifa->ifa_netmask == 0)
1680 					continue;
1681 				cp = addr_data;
1682 				cp2 = ifa->ifa_addr->sa_data;
1683 				cp3 = ifa->ifa_netmask->sa_data;
1684 				cplim = ifa->ifa_netmask->sa_len
1685 					+ (char *)ifa->ifa_netmask;
1686 				while (cp3 < cplim)
1687 					if ((*cp++ ^ *cp2++) & *cp3++)
1688 						goto next; /* next address! */
1689 				/*
1690 				 * If the netmask of what we just found
1691 				 * is more specific than what we had before
1692 				 * (if we had one) then remember the new one
1693 				 * before continuing to search
1694 				 * for an even better one.
1695 				 */
1696 				if (ifa_maybe == NULL ||
1697 				    rn_refines((caddr_t)ifa->ifa_netmask,
1698 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1699 					if (ifa_maybe != NULL)
1700 						ifa_free(ifa_maybe);
1701 					ifa_maybe = ifa;
1702 					ifa_ref(ifa_maybe);
1703 				}
1704 			}
1705 		}
1706 		IF_ADDR_UNLOCK(ifp);
1707 	}
1708 	ifa = ifa_maybe;
1709 	ifa_maybe = NULL;
1710 done:
1711 	IFNET_RUNLOCK_NOSLEEP();
1712 	if (ifa_maybe != NULL)
1713 		ifa_free(ifa_maybe);
1714 	return (ifa);
1715 }
1716 
1717 /*
1718  * Find an interface address specific to an interface best matching
1719  * a given address.
1720  */
1721 struct ifaddr *
1722 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1723 {
1724 	struct ifaddr *ifa;
1725 	char *cp, *cp2, *cp3;
1726 	char *cplim;
1727 	struct ifaddr *ifa_maybe = NULL;
1728 	u_int af = addr->sa_family;
1729 
1730 	if (af >= AF_MAX)
1731 		return (0);
1732 	IF_ADDR_LOCK(ifp);
1733 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1734 		if (ifa->ifa_addr->sa_family != af)
1735 			continue;
1736 		if (ifa_maybe == NULL)
1737 			ifa_maybe = ifa;
1738 		if (ifa->ifa_netmask == 0) {
1739 			if (sa_equal(addr, ifa->ifa_addr) ||
1740 			    (ifa->ifa_dstaddr &&
1741 			    sa_equal(addr, ifa->ifa_dstaddr)))
1742 				goto done;
1743 			continue;
1744 		}
1745 		if (ifp->if_flags & IFF_POINTOPOINT) {
1746 			if (sa_equal(addr, ifa->ifa_dstaddr))
1747 				goto done;
1748 		} else {
1749 			cp = addr->sa_data;
1750 			cp2 = ifa->ifa_addr->sa_data;
1751 			cp3 = ifa->ifa_netmask->sa_data;
1752 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1753 			for (; cp3 < cplim; cp3++)
1754 				if ((*cp++ ^ *cp2++) & *cp3)
1755 					break;
1756 			if (cp3 == cplim)
1757 				goto done;
1758 		}
1759 	}
1760 	ifa = ifa_maybe;
1761 done:
1762 	if (ifa != NULL)
1763 		ifa_ref(ifa);
1764 	IF_ADDR_UNLOCK(ifp);
1765 	return (ifa);
1766 }
1767 
1768 #include <net/if_llatbl.h>
1769 
1770 /*
1771  * Default action when installing a route with a Link Level gateway.
1772  * Lookup an appropriate real ifa to point to.
1773  * This should be moved to /sys/net/link.c eventually.
1774  */
1775 static void
1776 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1777 {
1778 	struct ifaddr *ifa, *oifa;
1779 	struct sockaddr *dst;
1780 	struct ifnet *ifp;
1781 
1782 	RT_LOCK_ASSERT(rt);
1783 
1784 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1785 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1786 		return;
1787 	ifa = ifaof_ifpforaddr(dst, ifp);
1788 	if (ifa) {
1789 		oifa = rt->rt_ifa;
1790 		rt->rt_ifa = ifa;
1791 		ifa_free(oifa);
1792 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1793 			ifa->ifa_rtrequest(cmd, rt, info);
1794 	}
1795 }
1796 
1797 /*
1798  * Mark an interface down and notify protocols of
1799  * the transition.
1800  * NOTE: must be called at splnet or eqivalent.
1801  */
1802 static void
1803 if_unroute(struct ifnet *ifp, int flag, int fam)
1804 {
1805 	struct ifaddr *ifa;
1806 
1807 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1808 
1809 	ifp->if_flags &= ~flag;
1810 	getmicrotime(&ifp->if_lastchange);
1811 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1812 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1813 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1814 	ifp->if_qflush(ifp);
1815 
1816 #if defined(INET) || defined(INET6)
1817 #ifdef DEV_CARP
1818 	if (ifp->if_carp)
1819 		carp_carpdev_state(ifp->if_carp);
1820 #endif
1821 #endif
1822 	rt_ifmsg(ifp);
1823 }
1824 
1825 /*
1826  * Mark an interface up and notify protocols of
1827  * the transition.
1828  * NOTE: must be called at splnet or eqivalent.
1829  */
1830 static void
1831 if_route(struct ifnet *ifp, int flag, int fam)
1832 {
1833 	struct ifaddr *ifa;
1834 
1835 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1836 
1837 	ifp->if_flags |= flag;
1838 	getmicrotime(&ifp->if_lastchange);
1839 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1840 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1841 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1842 #if defined(INET) || defined(INET6)
1843 #ifdef DEV_CARP
1844 	if (ifp->if_carp)
1845 		carp_carpdev_state(ifp->if_carp);
1846 #endif
1847 #endif
1848 	rt_ifmsg(ifp);
1849 #ifdef INET6
1850 	in6_if_up(ifp);
1851 #endif
1852 }
1853 
1854 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1855 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1856 
1857 /*
1858  * Handle a change in the interface link state. To avoid LORs
1859  * between driver lock and upper layer locks, as well as possible
1860  * recursions, we post event to taskqueue, and all job
1861  * is done in static do_link_state_change().
1862  */
1863 void
1864 if_link_state_change(struct ifnet *ifp, int link_state)
1865 {
1866 	/* Return if state hasn't changed. */
1867 	if (ifp->if_link_state == link_state)
1868 		return;
1869 
1870 	ifp->if_link_state = link_state;
1871 
1872 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1873 }
1874 
1875 static void
1876 do_link_state_change(void *arg, int pending)
1877 {
1878 	struct ifnet *ifp = (struct ifnet *)arg;
1879 	int link_state = ifp->if_link_state;
1880 	CURVNET_SET(ifp->if_vnet);
1881 
1882 	/* Notify that the link state has changed. */
1883 	rt_ifmsg(ifp);
1884 	if (ifp->if_vlantrunk != NULL)
1885 		(*vlan_link_state_p)(ifp);
1886 
1887 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1888 	    IFP2AC(ifp)->ac_netgraph != NULL)
1889 		(*ng_ether_link_state_p)(ifp, link_state);
1890 #if defined(INET) || defined(INET6)
1891 #ifdef DEV_CARP
1892 	if (ifp->if_carp)
1893 		carp_carpdev_state(ifp->if_carp);
1894 #endif
1895 #endif
1896 	if (ifp->if_bridge) {
1897 		KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1898 		(*bstp_linkstate_p)(ifp, link_state);
1899 	}
1900 	if (ifp->if_lagg) {
1901 		KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1902 		(*lagg_linkstate_p)(ifp, link_state);
1903 	}
1904 
1905 	if (IS_DEFAULT_VNET(curvnet))
1906 		devctl_notify("IFNET", ifp->if_xname,
1907 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1908 		    NULL);
1909 	if (pending > 1)
1910 		if_printf(ifp, "%d link states coalesced\n", pending);
1911 	if (log_link_state_change)
1912 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1913 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1914 	CURVNET_RESTORE();
1915 }
1916 
1917 /*
1918  * Mark an interface down and notify protocols of
1919  * the transition.
1920  * NOTE: must be called at splnet or eqivalent.
1921  */
1922 void
1923 if_down(struct ifnet *ifp)
1924 {
1925 
1926 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1927 }
1928 
1929 /*
1930  * Mark an interface up and notify protocols of
1931  * the transition.
1932  * NOTE: must be called at splnet or eqivalent.
1933  */
1934 void
1935 if_up(struct ifnet *ifp)
1936 {
1937 
1938 	if_route(ifp, IFF_UP, AF_UNSPEC);
1939 }
1940 
1941 /*
1942  * Flush an interface queue.
1943  */
1944 void
1945 if_qflush(struct ifnet *ifp)
1946 {
1947 	struct mbuf *m, *n;
1948 	struct ifaltq *ifq;
1949 
1950 	ifq = &ifp->if_snd;
1951 	IFQ_LOCK(ifq);
1952 #ifdef ALTQ
1953 	if (ALTQ_IS_ENABLED(ifq))
1954 		ALTQ_PURGE(ifq);
1955 #endif
1956 	n = ifq->ifq_head;
1957 	while ((m = n) != 0) {
1958 		n = m->m_act;
1959 		m_freem(m);
1960 	}
1961 	ifq->ifq_head = 0;
1962 	ifq->ifq_tail = 0;
1963 	ifq->ifq_len = 0;
1964 	IFQ_UNLOCK(ifq);
1965 }
1966 
1967 /*
1968  * Map interface name to interface structure pointer, with or without
1969  * returning a reference.
1970  */
1971 struct ifnet *
1972 ifunit_ref(const char *name)
1973 {
1974 	struct ifnet *ifp;
1975 
1976 	IFNET_RLOCK_NOSLEEP();
1977 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1978 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
1979 		    !(ifp->if_flags & IFF_DYING))
1980 			break;
1981 	}
1982 	if (ifp != NULL)
1983 		if_ref(ifp);
1984 	IFNET_RUNLOCK_NOSLEEP();
1985 	return (ifp);
1986 }
1987 
1988 struct ifnet *
1989 ifunit(const char *name)
1990 {
1991 	struct ifnet *ifp;
1992 
1993 	IFNET_RLOCK_NOSLEEP();
1994 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1995 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1996 			break;
1997 	}
1998 	IFNET_RUNLOCK_NOSLEEP();
1999 	return (ifp);
2000 }
2001 
2002 /*
2003  * Hardware specific interface ioctls.
2004  */
2005 static int
2006 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2007 {
2008 	struct ifreq *ifr;
2009 	struct ifstat *ifs;
2010 	int error = 0;
2011 	int new_flags, temp_flags;
2012 	size_t namelen, onamelen;
2013 	size_t descrlen;
2014 	char *descrbuf, *odescrbuf;
2015 	char new_name[IFNAMSIZ];
2016 	struct ifaddr *ifa;
2017 	struct sockaddr_dl *sdl;
2018 
2019 	ifr = (struct ifreq *)data;
2020 	switch (cmd) {
2021 	case SIOCGIFINDEX:
2022 		ifr->ifr_index = ifp->if_index;
2023 		break;
2024 
2025 	case SIOCGIFFLAGS:
2026 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2027 		ifr->ifr_flags = temp_flags & 0xffff;
2028 		ifr->ifr_flagshigh = temp_flags >> 16;
2029 		break;
2030 
2031 	case SIOCGIFCAP:
2032 		ifr->ifr_reqcap = ifp->if_capabilities;
2033 		ifr->ifr_curcap = ifp->if_capenable;
2034 		break;
2035 
2036 #ifdef MAC
2037 	case SIOCGIFMAC:
2038 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2039 		break;
2040 #endif
2041 
2042 	case SIOCGIFMETRIC:
2043 		ifr->ifr_metric = ifp->if_metric;
2044 		break;
2045 
2046 	case SIOCGIFMTU:
2047 		ifr->ifr_mtu = ifp->if_mtu;
2048 		break;
2049 
2050 	case SIOCGIFPHYS:
2051 		ifr->ifr_phys = ifp->if_physical;
2052 		break;
2053 
2054 	case SIOCGIFDESCR:
2055 		error = 0;
2056 		sx_slock(&ifdescr_sx);
2057 		if (ifp->if_description == NULL)
2058 			error = ENOMSG;
2059 		else {
2060 			/* space for terminating nul */
2061 			descrlen = strlen(ifp->if_description) + 1;
2062 			if (ifr->ifr_buffer.length < descrlen)
2063 				ifr->ifr_buffer.buffer = NULL;
2064 			else
2065 				error = copyout(ifp->if_description,
2066 				    ifr->ifr_buffer.buffer, descrlen);
2067 			ifr->ifr_buffer.length = descrlen;
2068 		}
2069 		sx_sunlock(&ifdescr_sx);
2070 		break;
2071 
2072 	case SIOCSIFDESCR:
2073 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2074 		if (error)
2075 			return (error);
2076 
2077 		/*
2078 		 * Copy only (length-1) bytes to make sure that
2079 		 * if_description is always nul terminated.  The
2080 		 * length parameter is supposed to count the
2081 		 * terminating nul in.
2082 		 */
2083 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2084 			return (ENAMETOOLONG);
2085 		else if (ifr->ifr_buffer.length == 0)
2086 			descrbuf = NULL;
2087 		else {
2088 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2089 			    M_WAITOK | M_ZERO);
2090 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2091 			    ifr->ifr_buffer.length - 1);
2092 			if (error) {
2093 				free(descrbuf, M_IFDESCR);
2094 				break;
2095 			}
2096 		}
2097 
2098 		sx_xlock(&ifdescr_sx);
2099 		odescrbuf = ifp->if_description;
2100 		ifp->if_description = descrbuf;
2101 		sx_xunlock(&ifdescr_sx);
2102 
2103 		getmicrotime(&ifp->if_lastchange);
2104 		free(odescrbuf, M_IFDESCR);
2105 		break;
2106 
2107 	case SIOCSIFFLAGS:
2108 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2109 		if (error)
2110 			return (error);
2111 		/*
2112 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2113 		 * check, so we don't need special handling here yet.
2114 		 */
2115 		new_flags = (ifr->ifr_flags & 0xffff) |
2116 		    (ifr->ifr_flagshigh << 16);
2117 		if (ifp->if_flags & IFF_SMART) {
2118 			/* Smart drivers twiddle their own routes */
2119 		} else if (ifp->if_flags & IFF_UP &&
2120 		    (new_flags & IFF_UP) == 0) {
2121 			int s = splimp();
2122 			if_down(ifp);
2123 			splx(s);
2124 		} else if (new_flags & IFF_UP &&
2125 		    (ifp->if_flags & IFF_UP) == 0) {
2126 			int s = splimp();
2127 			if_up(ifp);
2128 			splx(s);
2129 		}
2130 		/* See if permanently promiscuous mode bit is about to flip */
2131 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2132 			if (new_flags & IFF_PPROMISC)
2133 				ifp->if_flags |= IFF_PROMISC;
2134 			else if (ifp->if_pcount == 0)
2135 				ifp->if_flags &= ~IFF_PROMISC;
2136 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2137 			    ifp->if_xname,
2138 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2139 		}
2140 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2141 			(new_flags &~ IFF_CANTCHANGE);
2142 		if (ifp->if_ioctl) {
2143 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2144 		}
2145 		getmicrotime(&ifp->if_lastchange);
2146 		break;
2147 
2148 	case SIOCSIFCAP:
2149 		error = priv_check(td, PRIV_NET_SETIFCAP);
2150 		if (error)
2151 			return (error);
2152 		if (ifp->if_ioctl == NULL)
2153 			return (EOPNOTSUPP);
2154 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2155 			return (EINVAL);
2156 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2157 		if (error == 0)
2158 			getmicrotime(&ifp->if_lastchange);
2159 		break;
2160 
2161 #ifdef MAC
2162 	case SIOCSIFMAC:
2163 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2164 		break;
2165 #endif
2166 
2167 	case SIOCSIFNAME:
2168 		error = priv_check(td, PRIV_NET_SETIFNAME);
2169 		if (error)
2170 			return (error);
2171 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2172 		if (error != 0)
2173 			return (error);
2174 		if (new_name[0] == '\0')
2175 			return (EINVAL);
2176 		if (ifunit(new_name) != NULL)
2177 			return (EEXIST);
2178 
2179 		/*
2180 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2181 		 * and there are numerous other races with the
2182 		 * ifunit() checks not being atomic with namespace
2183 		 * changes (renames, vmoves, if_attach, etc).
2184 		 */
2185 		ifp->if_flags |= IFF_RENAMING;
2186 
2187 		/* Announce the departure of the interface. */
2188 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2189 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2190 
2191 		log(LOG_INFO, "%s: changing name to '%s'\n",
2192 		    ifp->if_xname, new_name);
2193 
2194 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2195 		ifa = ifp->if_addr;
2196 		IFA_LOCK(ifa);
2197 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2198 		namelen = strlen(new_name);
2199 		onamelen = sdl->sdl_nlen;
2200 		/*
2201 		 * Move the address if needed.  This is safe because we
2202 		 * allocate space for a name of length IFNAMSIZ when we
2203 		 * create this in if_attach().
2204 		 */
2205 		if (namelen != onamelen) {
2206 			bcopy(sdl->sdl_data + onamelen,
2207 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2208 		}
2209 		bcopy(new_name, sdl->sdl_data, namelen);
2210 		sdl->sdl_nlen = namelen;
2211 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2212 		bzero(sdl->sdl_data, onamelen);
2213 		while (namelen != 0)
2214 			sdl->sdl_data[--namelen] = 0xff;
2215 		IFA_UNLOCK(ifa);
2216 
2217 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2218 		/* Announce the return of the interface. */
2219 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2220 
2221 		ifp->if_flags &= ~IFF_RENAMING;
2222 		break;
2223 
2224 #ifdef VIMAGE
2225 	case SIOCSIFVNET:
2226 		error = priv_check(td, PRIV_NET_SETIFVNET);
2227 		if (error)
2228 			return (error);
2229 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2230 		break;
2231 #endif
2232 
2233 	case SIOCSIFMETRIC:
2234 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2235 		if (error)
2236 			return (error);
2237 		ifp->if_metric = ifr->ifr_metric;
2238 		getmicrotime(&ifp->if_lastchange);
2239 		break;
2240 
2241 	case SIOCSIFPHYS:
2242 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2243 		if (error)
2244 			return (error);
2245 		if (ifp->if_ioctl == NULL)
2246 			return (EOPNOTSUPP);
2247 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2248 		if (error == 0)
2249 			getmicrotime(&ifp->if_lastchange);
2250 		break;
2251 
2252 	case SIOCSIFMTU:
2253 	{
2254 		u_long oldmtu = ifp->if_mtu;
2255 
2256 		error = priv_check(td, PRIV_NET_SETIFMTU);
2257 		if (error)
2258 			return (error);
2259 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2260 			return (EINVAL);
2261 		if (ifp->if_ioctl == NULL)
2262 			return (EOPNOTSUPP);
2263 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2264 		if (error == 0) {
2265 			getmicrotime(&ifp->if_lastchange);
2266 			rt_ifmsg(ifp);
2267 		}
2268 		/*
2269 		 * If the link MTU changed, do network layer specific procedure.
2270 		 */
2271 		if (ifp->if_mtu != oldmtu) {
2272 #ifdef INET6
2273 			nd6_setmtu(ifp);
2274 #endif
2275 		}
2276 		break;
2277 	}
2278 
2279 	case SIOCADDMULTI:
2280 	case SIOCDELMULTI:
2281 		if (cmd == SIOCADDMULTI)
2282 			error = priv_check(td, PRIV_NET_ADDMULTI);
2283 		else
2284 			error = priv_check(td, PRIV_NET_DELMULTI);
2285 		if (error)
2286 			return (error);
2287 
2288 		/* Don't allow group membership on non-multicast interfaces. */
2289 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2290 			return (EOPNOTSUPP);
2291 
2292 		/* Don't let users screw up protocols' entries. */
2293 		if (ifr->ifr_addr.sa_family != AF_LINK)
2294 			return (EINVAL);
2295 
2296 		if (cmd == SIOCADDMULTI) {
2297 			struct ifmultiaddr *ifma;
2298 
2299 			/*
2300 			 * Userland is only permitted to join groups once
2301 			 * via the if_addmulti() KPI, because it cannot hold
2302 			 * struct ifmultiaddr * between calls. It may also
2303 			 * lose a race while we check if the membership
2304 			 * already exists.
2305 			 */
2306 			IF_ADDR_LOCK(ifp);
2307 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2308 			IF_ADDR_UNLOCK(ifp);
2309 			if (ifma != NULL)
2310 				error = EADDRINUSE;
2311 			else
2312 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2313 		} else {
2314 			error = if_delmulti(ifp, &ifr->ifr_addr);
2315 		}
2316 		if (error == 0)
2317 			getmicrotime(&ifp->if_lastchange);
2318 		break;
2319 
2320 	case SIOCSIFPHYADDR:
2321 	case SIOCDIFPHYADDR:
2322 #ifdef INET6
2323 	case SIOCSIFPHYADDR_IN6:
2324 #endif
2325 	case SIOCSLIFPHYADDR:
2326 	case SIOCSIFMEDIA:
2327 	case SIOCSIFGENERIC:
2328 		error = priv_check(td, PRIV_NET_HWIOCTL);
2329 		if (error)
2330 			return (error);
2331 		if (ifp->if_ioctl == NULL)
2332 			return (EOPNOTSUPP);
2333 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2334 		if (error == 0)
2335 			getmicrotime(&ifp->if_lastchange);
2336 		break;
2337 
2338 	case SIOCGIFSTATUS:
2339 		ifs = (struct ifstat *)data;
2340 		ifs->ascii[0] = '\0';
2341 
2342 	case SIOCGIFPSRCADDR:
2343 	case SIOCGIFPDSTADDR:
2344 	case SIOCGLIFPHYADDR:
2345 	case SIOCGIFMEDIA:
2346 	case SIOCGIFGENERIC:
2347 		if (ifp->if_ioctl == NULL)
2348 			return (EOPNOTSUPP);
2349 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2350 		break;
2351 
2352 	case SIOCSIFLLADDR:
2353 		error = priv_check(td, PRIV_NET_SETLLADDR);
2354 		if (error)
2355 			return (error);
2356 		error = if_setlladdr(ifp,
2357 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2358 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2359 		break;
2360 
2361 	case SIOCAIFGROUP:
2362 	{
2363 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2364 
2365 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2366 		if (error)
2367 			return (error);
2368 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2369 			return (error);
2370 		break;
2371 	}
2372 
2373 	case SIOCGIFGROUP:
2374 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2375 			return (error);
2376 		break;
2377 
2378 	case SIOCDIFGROUP:
2379 	{
2380 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2381 
2382 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2383 		if (error)
2384 			return (error);
2385 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2386 			return (error);
2387 		break;
2388 	}
2389 
2390 	default:
2391 		error = ENOIOCTL;
2392 		break;
2393 	}
2394 	return (error);
2395 }
2396 
2397 /*
2398  * Interface ioctls.
2399  */
2400 int
2401 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2402 {
2403 	struct ifnet *ifp;
2404 	struct ifreq *ifr;
2405 	int error;
2406 	int oif_flags;
2407 
2408 	switch (cmd) {
2409 	case SIOCGIFCONF:
2410 	case OSIOCGIFCONF:
2411 #ifdef __amd64__
2412 	case SIOCGIFCONF32:
2413 #endif
2414 		return (ifconf(cmd, data));
2415 	}
2416 	ifr = (struct ifreq *)data;
2417 
2418 	switch (cmd) {
2419 #ifdef VIMAGE
2420 	case SIOCSIFRVNET:
2421 		error = priv_check(td, PRIV_NET_SETIFVNET);
2422 		if (error)
2423 			return (error);
2424 		return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
2425 #endif
2426 	case SIOCIFCREATE:
2427 	case SIOCIFCREATE2:
2428 		error = priv_check(td, PRIV_NET_IFCREATE);
2429 		if (error)
2430 			return (error);
2431 		return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
2432 			cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
2433 	case SIOCIFDESTROY:
2434 		error = priv_check(td, PRIV_NET_IFDESTROY);
2435 		if (error)
2436 			return (error);
2437 		return if_clone_destroy(ifr->ifr_name);
2438 
2439 	case SIOCIFGCLONERS:
2440 		return (if_clone_list((struct if_clonereq *)data));
2441 	case SIOCGIFGMEMB:
2442 		return (if_getgroupmembers((struct ifgroupreq *)data));
2443 	}
2444 
2445 	ifp = ifunit_ref(ifr->ifr_name);
2446 	if (ifp == NULL)
2447 		return (ENXIO);
2448 
2449 	error = ifhwioctl(cmd, ifp, data, td);
2450 	if (error != ENOIOCTL) {
2451 		if_rele(ifp);
2452 		return (error);
2453 	}
2454 
2455 	oif_flags = ifp->if_flags;
2456 	if (so->so_proto == NULL) {
2457 		if_rele(ifp);
2458 		return (EOPNOTSUPP);
2459 	}
2460 #ifndef COMPAT_43
2461 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2462 								 data,
2463 								 ifp, td));
2464 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2465 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2466 #else
2467 	{
2468 		u_long ocmd = cmd;
2469 
2470 		switch (cmd) {
2471 
2472 		case SIOCSIFDSTADDR:
2473 		case SIOCSIFADDR:
2474 		case SIOCSIFBRDADDR:
2475 		case SIOCSIFNETMASK:
2476 #if BYTE_ORDER != BIG_ENDIAN
2477 			if (ifr->ifr_addr.sa_family == 0 &&
2478 			    ifr->ifr_addr.sa_len < 16) {
2479 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2480 				ifr->ifr_addr.sa_len = 16;
2481 			}
2482 #else
2483 			if (ifr->ifr_addr.sa_len == 0)
2484 				ifr->ifr_addr.sa_len = 16;
2485 #endif
2486 			break;
2487 
2488 		case OSIOCGIFADDR:
2489 			cmd = SIOCGIFADDR;
2490 			break;
2491 
2492 		case OSIOCGIFDSTADDR:
2493 			cmd = SIOCGIFDSTADDR;
2494 			break;
2495 
2496 		case OSIOCGIFBRDADDR:
2497 			cmd = SIOCGIFBRDADDR;
2498 			break;
2499 
2500 		case OSIOCGIFNETMASK:
2501 			cmd = SIOCGIFNETMASK;
2502 		}
2503 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
2504 								   cmd,
2505 								   data,
2506 								   ifp, td));
2507 		if (error == EOPNOTSUPP && ifp != NULL &&
2508 		    ifp->if_ioctl != NULL)
2509 			error = (*ifp->if_ioctl)(ifp, cmd, data);
2510 		switch (ocmd) {
2511 
2512 		case OSIOCGIFADDR:
2513 		case OSIOCGIFDSTADDR:
2514 		case OSIOCGIFBRDADDR:
2515 		case OSIOCGIFNETMASK:
2516 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2517 
2518 		}
2519 	}
2520 #endif /* COMPAT_43 */
2521 
2522 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2523 #ifdef INET6
2524 		if (ifp->if_flags & IFF_UP) {
2525 			int s = splimp();
2526 			in6_if_up(ifp);
2527 			splx(s);
2528 		}
2529 #endif
2530 	}
2531 	if_rele(ifp);
2532 	return (error);
2533 }
2534 
2535 /*
2536  * The code common to handling reference counted flags,
2537  * e.g., in ifpromisc() and if_allmulti().
2538  * The "pflag" argument can specify a permanent mode flag to check,
2539  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2540  *
2541  * Only to be used on stack-owned flags, not driver-owned flags.
2542  */
2543 static int
2544 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2545 {
2546 	struct ifreq ifr;
2547 	int error;
2548 	int oldflags, oldcount;
2549 
2550 	/* Sanity checks to catch programming errors */
2551 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2552 	    ("%s: setting driver-owned flag %d", __func__, flag));
2553 
2554 	if (onswitch)
2555 		KASSERT(*refcount >= 0,
2556 		    ("%s: increment negative refcount %d for flag %d",
2557 		    __func__, *refcount, flag));
2558 	else
2559 		KASSERT(*refcount > 0,
2560 		    ("%s: decrement non-positive refcount %d for flag %d",
2561 		    __func__, *refcount, flag));
2562 
2563 	/* In case this mode is permanent, just touch refcount */
2564 	if (ifp->if_flags & pflag) {
2565 		*refcount += onswitch ? 1 : -1;
2566 		return (0);
2567 	}
2568 
2569 	/* Save ifnet parameters for if_ioctl() may fail */
2570 	oldcount = *refcount;
2571 	oldflags = ifp->if_flags;
2572 
2573 	/*
2574 	 * See if we aren't the only and touching refcount is enough.
2575 	 * Actually toggle interface flag if we are the first or last.
2576 	 */
2577 	if (onswitch) {
2578 		if ((*refcount)++)
2579 			return (0);
2580 		ifp->if_flags |= flag;
2581 	} else {
2582 		if (--(*refcount))
2583 			return (0);
2584 		ifp->if_flags &= ~flag;
2585 	}
2586 
2587 	/* Call down the driver since we've changed interface flags */
2588 	if (ifp->if_ioctl == NULL) {
2589 		error = EOPNOTSUPP;
2590 		goto recover;
2591 	}
2592 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2593 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2594 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2595 	if (error)
2596 		goto recover;
2597 	/* Notify userland that interface flags have changed */
2598 	rt_ifmsg(ifp);
2599 	return (0);
2600 
2601 recover:
2602 	/* Recover after driver error */
2603 	*refcount = oldcount;
2604 	ifp->if_flags = oldflags;
2605 	return (error);
2606 }
2607 
2608 /*
2609  * Set/clear promiscuous mode on interface ifp based on the truth value
2610  * of pswitch.  The calls are reference counted so that only the first
2611  * "on" request actually has an effect, as does the final "off" request.
2612  * Results are undefined if the "off" and "on" requests are not matched.
2613  */
2614 int
2615 ifpromisc(struct ifnet *ifp, int pswitch)
2616 {
2617 	int error;
2618 	int oldflags = ifp->if_flags;
2619 
2620 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2621 			   &ifp->if_pcount, pswitch);
2622 	/* If promiscuous mode status has changed, log a message */
2623 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2624 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2625 		    ifp->if_xname,
2626 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2627 	return (error);
2628 }
2629 
2630 /*
2631  * Return interface configuration
2632  * of system.  List may be used
2633  * in later ioctl's (above) to get
2634  * other information.
2635  */
2636 /*ARGSUSED*/
2637 static int
2638 ifconf(u_long cmd, caddr_t data)
2639 {
2640 	struct ifconf *ifc = (struct ifconf *)data;
2641 #ifdef __amd64__
2642 	struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2643 	struct ifconf ifc_swab;
2644 #endif
2645 	struct ifnet *ifp;
2646 	struct ifaddr *ifa;
2647 	struct ifreq ifr;
2648 	struct sbuf *sb;
2649 	int error, full = 0, valid_len, max_len;
2650 
2651 #ifdef __amd64__
2652 	if (cmd == SIOCGIFCONF32) {
2653 		ifc_swab.ifc_len = ifc32->ifc_len;
2654 		ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2655 		ifc = &ifc_swab;
2656 	}
2657 #endif
2658 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2659 	max_len = MAXPHYS - 1;
2660 
2661 	/* Prevent hostile input from being able to crash the system */
2662 	if (ifc->ifc_len <= 0)
2663 		return (EINVAL);
2664 
2665 again:
2666 	if (ifc->ifc_len <= max_len) {
2667 		max_len = ifc->ifc_len;
2668 		full = 1;
2669 	}
2670 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2671 	max_len = 0;
2672 	valid_len = 0;
2673 
2674 	IFNET_RLOCK();
2675 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2676 		int addrs;
2677 
2678 		/*
2679 		 * Zero the ifr_name buffer to make sure we don't
2680 		 * disclose the contents of the stack.
2681 		 */
2682 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2683 
2684 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2685 		    >= sizeof(ifr.ifr_name)) {
2686 			sbuf_delete(sb);
2687 			IFNET_RUNLOCK();
2688 			return (ENAMETOOLONG);
2689 		}
2690 
2691 		addrs = 0;
2692 		IF_ADDR_LOCK(ifp);
2693 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2694 			struct sockaddr *sa = ifa->ifa_addr;
2695 
2696 			if (prison_if(curthread->td_ucred, sa) != 0)
2697 				continue;
2698 			addrs++;
2699 #ifdef COMPAT_43
2700 			if (cmd == OSIOCGIFCONF) {
2701 				struct osockaddr *osa =
2702 					 (struct osockaddr *)&ifr.ifr_addr;
2703 				ifr.ifr_addr = *sa;
2704 				osa->sa_family = sa->sa_family;
2705 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2706 				max_len += sizeof(ifr);
2707 			} else
2708 #endif
2709 			if (sa->sa_len <= sizeof(*sa)) {
2710 				ifr.ifr_addr = *sa;
2711 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2712 				max_len += sizeof(ifr);
2713 			} else {
2714 				sbuf_bcat(sb, &ifr,
2715 				    offsetof(struct ifreq, ifr_addr));
2716 				max_len += offsetof(struct ifreq, ifr_addr);
2717 				sbuf_bcat(sb, sa, sa->sa_len);
2718 				max_len += sa->sa_len;
2719 			}
2720 
2721 			if (!sbuf_overflowed(sb))
2722 				valid_len = sbuf_len(sb);
2723 		}
2724 		IF_ADDR_UNLOCK(ifp);
2725 		if (addrs == 0) {
2726 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2727 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2728 			max_len += sizeof(ifr);
2729 
2730 			if (!sbuf_overflowed(sb))
2731 				valid_len = sbuf_len(sb);
2732 		}
2733 	}
2734 	IFNET_RUNLOCK();
2735 
2736 	/*
2737 	 * If we didn't allocate enough space (uncommon), try again.  If
2738 	 * we have already allocated as much space as we are allowed,
2739 	 * return what we've got.
2740 	 */
2741 	if (valid_len != max_len && !full) {
2742 		sbuf_delete(sb);
2743 		goto again;
2744 	}
2745 
2746 	ifc->ifc_len = valid_len;
2747 #ifdef __amd64__
2748 	if (cmd == SIOCGIFCONF32)
2749 		ifc32->ifc_len = valid_len;
2750 #endif
2751 	sbuf_finish(sb);
2752 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2753 	sbuf_delete(sb);
2754 	return (error);
2755 }
2756 
2757 /*
2758  * Just like ifpromisc(), but for all-multicast-reception mode.
2759  */
2760 int
2761 if_allmulti(struct ifnet *ifp, int onswitch)
2762 {
2763 
2764 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2765 }
2766 
2767 struct ifmultiaddr *
2768 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2769 {
2770 	struct ifmultiaddr *ifma;
2771 
2772 	IF_ADDR_LOCK_ASSERT(ifp);
2773 
2774 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2775 		if (sa->sa_family == AF_LINK) {
2776 			if (sa_dl_equal(ifma->ifma_addr, sa))
2777 				break;
2778 		} else {
2779 			if (sa_equal(ifma->ifma_addr, sa))
2780 				break;
2781 		}
2782 	}
2783 
2784 	return ifma;
2785 }
2786 
2787 /*
2788  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2789  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2790  * the ifnet multicast address list here, so the caller must do that and
2791  * other setup work (such as notifying the device driver).  The reference
2792  * count is initialized to 1.
2793  */
2794 static struct ifmultiaddr *
2795 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2796     int mflags)
2797 {
2798 	struct ifmultiaddr *ifma;
2799 	struct sockaddr *dupsa;
2800 
2801 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2802 	    M_ZERO);
2803 	if (ifma == NULL)
2804 		return (NULL);
2805 
2806 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2807 	if (dupsa == NULL) {
2808 		free(ifma, M_IFMADDR);
2809 		return (NULL);
2810 	}
2811 	bcopy(sa, dupsa, sa->sa_len);
2812 	ifma->ifma_addr = dupsa;
2813 
2814 	ifma->ifma_ifp = ifp;
2815 	ifma->ifma_refcount = 1;
2816 	ifma->ifma_protospec = NULL;
2817 
2818 	if (llsa == NULL) {
2819 		ifma->ifma_lladdr = NULL;
2820 		return (ifma);
2821 	}
2822 
2823 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2824 	if (dupsa == NULL) {
2825 		free(ifma->ifma_addr, M_IFMADDR);
2826 		free(ifma, M_IFMADDR);
2827 		return (NULL);
2828 	}
2829 	bcopy(llsa, dupsa, llsa->sa_len);
2830 	ifma->ifma_lladdr = dupsa;
2831 
2832 	return (ifma);
2833 }
2834 
2835 /*
2836  * if_freemulti: free ifmultiaddr structure and possibly attached related
2837  * addresses.  The caller is responsible for implementing reference
2838  * counting, notifying the driver, handling routing messages, and releasing
2839  * any dependent link layer state.
2840  */
2841 static void
2842 if_freemulti(struct ifmultiaddr *ifma)
2843 {
2844 
2845 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2846 	    ifma->ifma_refcount));
2847 	KASSERT(ifma->ifma_protospec == NULL,
2848 	    ("if_freemulti: protospec not NULL"));
2849 
2850 	if (ifma->ifma_lladdr != NULL)
2851 		free(ifma->ifma_lladdr, M_IFMADDR);
2852 	free(ifma->ifma_addr, M_IFMADDR);
2853 	free(ifma, M_IFMADDR);
2854 }
2855 
2856 /*
2857  * Register an additional multicast address with a network interface.
2858  *
2859  * - If the address is already present, bump the reference count on the
2860  *   address and return.
2861  * - If the address is not link-layer, look up a link layer address.
2862  * - Allocate address structures for one or both addresses, and attach to the
2863  *   multicast address list on the interface.  If automatically adding a link
2864  *   layer address, the protocol address will own a reference to the link
2865  *   layer address, to be freed when it is freed.
2866  * - Notify the network device driver of an addition to the multicast address
2867  *   list.
2868  *
2869  * 'sa' points to caller-owned memory with the desired multicast address.
2870  *
2871  * 'retifma' will be used to return a pointer to the resulting multicast
2872  * address reference, if desired.
2873  */
2874 int
2875 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2876     struct ifmultiaddr **retifma)
2877 {
2878 	struct ifmultiaddr *ifma, *ll_ifma;
2879 	struct sockaddr *llsa;
2880 	int error;
2881 
2882 	/*
2883 	 * If the address is already present, return a new reference to it;
2884 	 * otherwise, allocate storage and set up a new address.
2885 	 */
2886 	IF_ADDR_LOCK(ifp);
2887 	ifma = if_findmulti(ifp, sa);
2888 	if (ifma != NULL) {
2889 		ifma->ifma_refcount++;
2890 		if (retifma != NULL)
2891 			*retifma = ifma;
2892 		IF_ADDR_UNLOCK(ifp);
2893 		return (0);
2894 	}
2895 
2896 	/*
2897 	 * The address isn't already present; resolve the protocol address
2898 	 * into a link layer address, and then look that up, bump its
2899 	 * refcount or allocate an ifma for that also.  If 'llsa' was
2900 	 * returned, we will need to free it later.
2901 	 */
2902 	llsa = NULL;
2903 	ll_ifma = NULL;
2904 	if (ifp->if_resolvemulti != NULL) {
2905 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2906 		if (error)
2907 			goto unlock_out;
2908 	}
2909 
2910 	/*
2911 	 * Allocate the new address.  Don't hook it up yet, as we may also
2912 	 * need to allocate a link layer multicast address.
2913 	 */
2914 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2915 	if (ifma == NULL) {
2916 		error = ENOMEM;
2917 		goto free_llsa_out;
2918 	}
2919 
2920 	/*
2921 	 * If a link layer address is found, we'll need to see if it's
2922 	 * already present in the address list, or allocate is as well.
2923 	 * When this block finishes, the link layer address will be on the
2924 	 * list.
2925 	 */
2926 	if (llsa != NULL) {
2927 		ll_ifma = if_findmulti(ifp, llsa);
2928 		if (ll_ifma == NULL) {
2929 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2930 			if (ll_ifma == NULL) {
2931 				--ifma->ifma_refcount;
2932 				if_freemulti(ifma);
2933 				error = ENOMEM;
2934 				goto free_llsa_out;
2935 			}
2936 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2937 			    ifma_link);
2938 		} else
2939 			ll_ifma->ifma_refcount++;
2940 		ifma->ifma_llifma = ll_ifma;
2941 	}
2942 
2943 	/*
2944 	 * We now have a new multicast address, ifma, and possibly a new or
2945 	 * referenced link layer address.  Add the primary address to the
2946 	 * ifnet address list.
2947 	 */
2948 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2949 
2950 	if (retifma != NULL)
2951 		*retifma = ifma;
2952 
2953 	/*
2954 	 * Must generate the message while holding the lock so that 'ifma'
2955 	 * pointer is still valid.
2956 	 */
2957 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2958 	IF_ADDR_UNLOCK(ifp);
2959 
2960 	/*
2961 	 * We are certain we have added something, so call down to the
2962 	 * interface to let them know about it.
2963 	 */
2964 	if (ifp->if_ioctl != NULL) {
2965 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2966 	}
2967 
2968 	if (llsa != NULL)
2969 		free(llsa, M_IFMADDR);
2970 
2971 	return (0);
2972 
2973 free_llsa_out:
2974 	if (llsa != NULL)
2975 		free(llsa, M_IFMADDR);
2976 
2977 unlock_out:
2978 	IF_ADDR_UNLOCK(ifp);
2979 	return (error);
2980 }
2981 
2982 /*
2983  * Delete a multicast group membership by network-layer group address.
2984  *
2985  * Returns ENOENT if the entry could not be found. If ifp no longer
2986  * exists, results are undefined. This entry point should only be used
2987  * from subsystems which do appropriate locking to hold ifp for the
2988  * duration of the call.
2989  * Network-layer protocol domains must use if_delmulti_ifma().
2990  */
2991 int
2992 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2993 {
2994 	struct ifmultiaddr *ifma;
2995 	int lastref;
2996 #ifdef INVARIANTS
2997 	struct ifnet *oifp;
2998 
2999 	IFNET_RLOCK_NOSLEEP();
3000 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3001 		if (ifp == oifp)
3002 			break;
3003 	if (ifp != oifp)
3004 		ifp = NULL;
3005 	IFNET_RUNLOCK_NOSLEEP();
3006 
3007 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3008 #endif
3009 	if (ifp == NULL)
3010 		return (ENOENT);
3011 
3012 	IF_ADDR_LOCK(ifp);
3013 	lastref = 0;
3014 	ifma = if_findmulti(ifp, sa);
3015 	if (ifma != NULL)
3016 		lastref = if_delmulti_locked(ifp, ifma, 0);
3017 	IF_ADDR_UNLOCK(ifp);
3018 
3019 	if (ifma == NULL)
3020 		return (ENOENT);
3021 
3022 	if (lastref && ifp->if_ioctl != NULL) {
3023 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3024 	}
3025 
3026 	return (0);
3027 }
3028 
3029 /*
3030  * Delete all multicast group membership for an interface.
3031  * Should be used to quickly flush all multicast filters.
3032  */
3033 void
3034 if_delallmulti(struct ifnet *ifp)
3035 {
3036 	struct ifmultiaddr *ifma;
3037 	struct ifmultiaddr *next;
3038 
3039 	IF_ADDR_LOCK(ifp);
3040 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3041 		if_delmulti_locked(ifp, ifma, 0);
3042 	IF_ADDR_UNLOCK(ifp);
3043 }
3044 
3045 /*
3046  * Delete a multicast group membership by group membership pointer.
3047  * Network-layer protocol domains must use this routine.
3048  *
3049  * It is safe to call this routine if the ifp disappeared.
3050  */
3051 void
3052 if_delmulti_ifma(struct ifmultiaddr *ifma)
3053 {
3054 	struct ifnet *ifp;
3055 	int lastref;
3056 
3057 	ifp = ifma->ifma_ifp;
3058 #ifdef DIAGNOSTIC
3059 	if (ifp == NULL) {
3060 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3061 	} else {
3062 		struct ifnet *oifp;
3063 
3064 		IFNET_RLOCK_NOSLEEP();
3065 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3066 			if (ifp == oifp)
3067 				break;
3068 		if (ifp != oifp) {
3069 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3070 			ifp = NULL;
3071 		}
3072 		IFNET_RUNLOCK_NOSLEEP();
3073 	}
3074 #endif
3075 	/*
3076 	 * If and only if the ifnet instance exists: Acquire the address lock.
3077 	 */
3078 	if (ifp != NULL)
3079 		IF_ADDR_LOCK(ifp);
3080 
3081 	lastref = if_delmulti_locked(ifp, ifma, 0);
3082 
3083 	if (ifp != NULL) {
3084 		/*
3085 		 * If and only if the ifnet instance exists:
3086 		 *  Release the address lock.
3087 		 *  If the group was left: update the hardware hash filter.
3088 		 */
3089 		IF_ADDR_UNLOCK(ifp);
3090 		if (lastref && ifp->if_ioctl != NULL) {
3091 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3092 		}
3093 	}
3094 }
3095 
3096 /*
3097  * Perform deletion of network-layer and/or link-layer multicast address.
3098  *
3099  * Return 0 if the reference count was decremented.
3100  * Return 1 if the final reference was released, indicating that the
3101  * hardware hash filter should be reprogrammed.
3102  */
3103 static int
3104 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3105 {
3106 	struct ifmultiaddr *ll_ifma;
3107 
3108 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3109 		KASSERT(ifma->ifma_ifp == ifp,
3110 		    ("%s: inconsistent ifp %p", __func__, ifp));
3111 		IF_ADDR_LOCK_ASSERT(ifp);
3112 	}
3113 
3114 	ifp = ifma->ifma_ifp;
3115 
3116 	/*
3117 	 * If the ifnet is detaching, null out references to ifnet,
3118 	 * so that upper protocol layers will notice, and not attempt
3119 	 * to obtain locks for an ifnet which no longer exists. The
3120 	 * routing socket announcement must happen before the ifnet
3121 	 * instance is detached from the system.
3122 	 */
3123 	if (detaching) {
3124 #ifdef DIAGNOSTIC
3125 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3126 #endif
3127 		/*
3128 		 * ifp may already be nulled out if we are being reentered
3129 		 * to delete the ll_ifma.
3130 		 */
3131 		if (ifp != NULL) {
3132 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3133 			ifma->ifma_ifp = NULL;
3134 		}
3135 	}
3136 
3137 	if (--ifma->ifma_refcount > 0)
3138 		return 0;
3139 
3140 	/*
3141 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3142 	 * have been associated with it. Release it first if so.
3143 	 */
3144 	ll_ifma = ifma->ifma_llifma;
3145 	if (ll_ifma != NULL) {
3146 		KASSERT(ifma->ifma_lladdr != NULL,
3147 		    ("%s: llifma w/o lladdr", __func__));
3148 		if (detaching)
3149 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3150 		if (--ll_ifma->ifma_refcount == 0) {
3151 			if (ifp != NULL) {
3152 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3153 				    ifma_link);
3154 			}
3155 			if_freemulti(ll_ifma);
3156 		}
3157 	}
3158 
3159 	if (ifp != NULL)
3160 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3161 
3162 	if_freemulti(ifma);
3163 
3164 	/*
3165 	 * The last reference to this instance of struct ifmultiaddr
3166 	 * was released; the hardware should be notified of this change.
3167 	 */
3168 	return 1;
3169 }
3170 
3171 /*
3172  * Set the link layer address on an interface.
3173  *
3174  * At this time we only support certain types of interfaces,
3175  * and we don't allow the length of the address to change.
3176  */
3177 int
3178 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3179 {
3180 	struct sockaddr_dl *sdl;
3181 	struct ifaddr *ifa;
3182 	struct ifreq ifr;
3183 
3184 	IF_ADDR_LOCK(ifp);
3185 	ifa = ifp->if_addr;
3186 	if (ifa == NULL) {
3187 		IF_ADDR_UNLOCK(ifp);
3188 		return (EINVAL);
3189 	}
3190 	ifa_ref(ifa);
3191 	IF_ADDR_UNLOCK(ifp);
3192 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3193 	if (sdl == NULL) {
3194 		ifa_free(ifa);
3195 		return (EINVAL);
3196 	}
3197 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3198 		ifa_free(ifa);
3199 		return (EINVAL);
3200 	}
3201 	switch (ifp->if_type) {
3202 	case IFT_ETHER:
3203 	case IFT_FDDI:
3204 	case IFT_XETHER:
3205 	case IFT_ISO88025:
3206 	case IFT_L2VLAN:
3207 	case IFT_BRIDGE:
3208 	case IFT_ARCNET:
3209 	case IFT_IEEE8023ADLAG:
3210 	case IFT_IEEE80211:
3211 		bcopy(lladdr, LLADDR(sdl), len);
3212 		ifa_free(ifa);
3213 		break;
3214 	default:
3215 		ifa_free(ifa);
3216 		return (ENODEV);
3217 	}
3218 
3219 	/*
3220 	 * If the interface is already up, we need
3221 	 * to re-init it in order to reprogram its
3222 	 * address filter.
3223 	 */
3224 	if ((ifp->if_flags & IFF_UP) != 0) {
3225 		if (ifp->if_ioctl) {
3226 			ifp->if_flags &= ~IFF_UP;
3227 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3228 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3229 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3230 			ifp->if_flags |= IFF_UP;
3231 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3232 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3233 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3234 		}
3235 #ifdef INET
3236 		/*
3237 		 * Also send gratuitous ARPs to notify other nodes about
3238 		 * the address change.
3239 		 */
3240 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3241 			if (ifa->ifa_addr->sa_family == AF_INET)
3242 				arp_ifinit(ifp, ifa);
3243 		}
3244 #endif
3245 	}
3246 	return (0);
3247 }
3248 
3249 /*
3250  * The name argument must be a pointer to storage which will last as
3251  * long as the interface does.  For physical devices, the result of
3252  * device_get_name(dev) is a good choice and for pseudo-devices a
3253  * static string works well.
3254  */
3255 void
3256 if_initname(struct ifnet *ifp, const char *name, int unit)
3257 {
3258 	ifp->if_dname = name;
3259 	ifp->if_dunit = unit;
3260 	if (unit != IF_DUNIT_NONE)
3261 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3262 	else
3263 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3264 }
3265 
3266 int
3267 if_printf(struct ifnet *ifp, const char * fmt, ...)
3268 {
3269 	va_list ap;
3270 	int retval;
3271 
3272 	retval = printf("%s: ", ifp->if_xname);
3273 	va_start(ap, fmt);
3274 	retval += vprintf(fmt, ap);
3275 	va_end(ap);
3276 	return (retval);
3277 }
3278 
3279 void
3280 if_start(struct ifnet *ifp)
3281 {
3282 
3283 	(*(ifp)->if_start)(ifp);
3284 }
3285 
3286 /*
3287  * Backwards compatibility interface for drivers
3288  * that have not implemented it
3289  */
3290 static int
3291 if_transmit(struct ifnet *ifp, struct mbuf *m)
3292 {
3293 	int error;
3294 
3295 	IFQ_HANDOFF(ifp, m, error);
3296 	return (error);
3297 }
3298 
3299 int
3300 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3301 {
3302 	int active = 0;
3303 
3304 	IF_LOCK(ifq);
3305 	if (_IF_QFULL(ifq)) {
3306 		_IF_DROP(ifq);
3307 		IF_UNLOCK(ifq);
3308 		m_freem(m);
3309 		return (0);
3310 	}
3311 	if (ifp != NULL) {
3312 		ifp->if_obytes += m->m_pkthdr.len + adjust;
3313 		if (m->m_flags & (M_BCAST|M_MCAST))
3314 			ifp->if_omcasts++;
3315 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3316 	}
3317 	_IF_ENQUEUE(ifq, m);
3318 	IF_UNLOCK(ifq);
3319 	if (ifp != NULL && !active)
3320 		(*(ifp)->if_start)(ifp);
3321 	return (1);
3322 }
3323 
3324 void
3325 if_register_com_alloc(u_char type,
3326     if_com_alloc_t *a, if_com_free_t *f)
3327 {
3328 
3329 	KASSERT(if_com_alloc[type] == NULL,
3330 	    ("if_register_com_alloc: %d already registered", type));
3331 	KASSERT(if_com_free[type] == NULL,
3332 	    ("if_register_com_alloc: %d free already registered", type));
3333 
3334 	if_com_alloc[type] = a;
3335 	if_com_free[type] = f;
3336 }
3337 
3338 void
3339 if_deregister_com_alloc(u_char type)
3340 {
3341 
3342 	KASSERT(if_com_alloc[type] != NULL,
3343 	    ("if_deregister_com_alloc: %d not registered", type));
3344 	KASSERT(if_com_free[type] != NULL,
3345 	    ("if_deregister_com_alloc: %d free not registered", type));
3346 	if_com_alloc[type] = NULL;
3347 	if_com_free[type] = NULL;
3348 }
3349 
3350 #ifdef DDB
3351 static void
3352 if_show_ifnet(struct ifnet *ifp)
3353 {
3354 
3355 	if (ifp == NULL)
3356 		return;
3357 	db_printf("%s:\n", ifp->if_xname);
3358 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
3359 	IF_DB_PRINTF("%s", if_dname);
3360 	IF_DB_PRINTF("%d", if_dunit);
3361 	IF_DB_PRINTF("%s", if_description);
3362 	IF_DB_PRINTF("%u", if_index);
3363 	IF_DB_PRINTF("%u", if_refcount);
3364 	IF_DB_PRINTF("%d", if_index_reserved);
3365 	IF_DB_PRINTF("%p", if_softc);
3366 	IF_DB_PRINTF("%p", if_l2com);
3367 	IF_DB_PRINTF("%p", if_vnet);
3368 	IF_DB_PRINTF("%p", if_home_vnet);
3369 	IF_DB_PRINTF("%p", if_addr);
3370 	IF_DB_PRINTF("%p", if_llsoftc);
3371 	IF_DB_PRINTF("%p", if_label);
3372 	IF_DB_PRINTF("%u", if_pcount);
3373 	IF_DB_PRINTF("0x%08x", if_flags);
3374 	IF_DB_PRINTF("0x%08x", if_drv_flags);
3375 	IF_DB_PRINTF("0x%08x", if_capabilities);
3376 	IF_DB_PRINTF("0x%08x", if_capenable);
3377 	IF_DB_PRINTF("%p", if_snd.ifq_head);
3378 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
3379 	IF_DB_PRINTF("%d", if_snd.ifq_len);
3380 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
3381 	IF_DB_PRINTF("%d", if_snd.ifq_drops);
3382 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
3383 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
3384 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
3385 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
3386 	IF_DB_PRINTF("%d", if_snd.altq_type);
3387 	IF_DB_PRINTF("%x", if_snd.altq_flags);
3388 #undef IF_DB_PRINTF
3389 }
3390 
3391 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
3392 {
3393 
3394 	if (!have_addr) {
3395 		db_printf("usage: show ifnet <struct ifnet *>\n");
3396 		return;
3397 	}
3398 
3399 	if_show_ifnet((struct ifnet *)addr);
3400 }
3401 
3402 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
3403 {
3404 	VNET_ITERATOR_DECL(vnet_iter);
3405 	struct ifnet *ifp;
3406 	u_short idx;
3407 
3408 	VNET_FOREACH(vnet_iter) {
3409 		CURVNET_SET_QUIET(vnet_iter);
3410 #ifdef VIMAGE
3411 		db_printf("vnet=%p\n", curvnet);
3412 #endif
3413 		for (idx = 1; idx <= V_if_index; idx++) {
3414 			ifp = V_ifindex_table[idx].ife_ifnet;
3415 			if (ifp == NULL)
3416 				continue;
3417 			db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
3418 			if (db_pager_quit)
3419 				break;
3420 		}
3421 		CURVNET_RESTORE();
3422 	}
3423 }
3424 #endif
3425