xref: /freebsd/sys/net/if.c (revision cded07a8783dedbb0634ebde78764f935f26467d)
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 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <machine/stdarg.h>
62 #include <vm/uma.h>
63 
64 #include <net/if.h>
65 #include <net/if_arp.h>
66 #include <net/if_clone.h>
67 #include <net/if_dl.h>
68 #include <net/if_types.h>
69 #include <net/if_var.h>
70 #include <net/radix.h>
71 #include <net/route.h>
72 #include <net/vnet.h>
73 
74 #if defined(INET) || defined(INET6)
75 /*XXX*/
76 #include <netinet/in.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip_carp.h>
79 #ifdef INET6
80 #include <netinet6/in6_var.h>
81 #include <netinet6/in6_ifattach.h>
82 #endif
83 #endif
84 #ifdef INET
85 #include <netinet/if_ether.h>
86 #endif
87 
88 #include <security/mac/mac_framework.h>
89 
90 #ifdef COMPAT_FREEBSD32
91 #include <sys/mount.h>
92 #include <compat/freebsd32/freebsd32.h>
93 #endif
94 
95 struct ifindex_entry {
96 	struct  ifnet *ife_ifnet;
97 };
98 
99 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
100 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
101 
102 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
103 SYSCTL_UINT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
104     &ifqmaxlen, 0, "max send queue size");
105 
106 /* Log link state change events */
107 static int log_link_state_change = 1;
108 
109 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
110 	&log_link_state_change, 0,
111 	"log interface link state change events");
112 
113 /* Interface description */
114 static unsigned int ifdescr_maxlen = 1024;
115 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
116 	&ifdescr_maxlen, 0,
117 	"administrative maximum length for interface description");
118 
119 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
120 
121 /* global sx for non-critical path ifdescr */
122 static struct sx ifdescr_sx;
123 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
124 
125 void	(*bstp_linkstate_p)(struct ifnet *ifp, int state);
126 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
127 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
128 /* These are external hooks for CARP. */
129 void	(*carp_linkstate_p)(struct ifnet *ifp);
130 #if defined(INET) || defined(INET6)
131 struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
132 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
133     struct sockaddr *sa, struct rtentry *rt);
134 #endif
135 #ifdef INET
136 int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
137     u_int8_t **);
138 #endif
139 #ifdef INET6
140 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
141 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
142     const struct in6_addr *taddr);
143 #endif
144 
145 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
146 
147 /*
148  * XXX: Style; these should be sorted alphabetically, and unprototyped
149  * static functions should be prototyped. Currently they are sorted by
150  * declaration order.
151  */
152 static void	if_attachdomain(void *);
153 static void	if_attachdomain1(struct ifnet *);
154 static int	ifconf(u_long, caddr_t);
155 static void	if_freemulti(struct ifmultiaddr *);
156 static void	if_init(void *);
157 static void	if_grow(void);
158 static void	if_route(struct ifnet *, int flag, int fam);
159 static int	if_setflag(struct ifnet *, int, int, int *, int);
160 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
161 static void	if_unroute(struct ifnet *, int flag, int fam);
162 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
163 static int	if_rtdel(struct radix_node *, void *);
164 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
165 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
166 static void	do_link_state_change(void *, int);
167 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
168 static int	if_getgroupmembers(struct ifgroupreq *);
169 static void	if_delgroups(struct ifnet *);
170 static void	if_attach_internal(struct ifnet *, int);
171 static void	if_detach_internal(struct ifnet *, int);
172 
173 #ifdef INET6
174 /*
175  * XXX: declare here to avoid to include many inet6 related files..
176  * should be more generalized?
177  */
178 extern void	nd6_setmtu(struct ifnet *);
179 #endif
180 
181 VNET_DEFINE(int, if_index);
182 int	ifqmaxlen = IFQ_MAXLEN;
183 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
184 VNET_DEFINE(struct ifgrouphead, ifg_head);
185 
186 static VNET_DEFINE(int, if_indexlim) = 8;
187 
188 /* Table of ifnet by index. */
189 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
190 
191 #define	V_if_indexlim		VNET(if_indexlim)
192 #define	V_ifindex_table		VNET(ifindex_table)
193 
194 /*
195  * The global network interface list (V_ifnet) and related state (such as
196  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
197  * an rwlock.  Either may be acquired shared to stablize the list, but both
198  * must be acquired writable to modify the list.  This model allows us to
199  * both stablize the interface list during interrupt thread processing, but
200  * also to stablize it over long-running ioctls, without introducing priority
201  * inversions and deadlocks.
202  */
203 struct rwlock ifnet_rwlock;
204 struct sx ifnet_sxlock;
205 
206 /*
207  * The allocation of network interfaces is a rather non-atomic affair; we
208  * need to select an index before we are ready to expose the interface for
209  * use, so will use this pointer value to indicate reservation.
210  */
211 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
212 
213 static	if_com_alloc_t *if_com_alloc[256];
214 static	if_com_free_t *if_com_free[256];
215 
216 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
217 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
218 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
219 
220 struct ifnet *
221 ifnet_byindex_locked(u_short idx)
222 {
223 
224 	if (idx > V_if_index)
225 		return (NULL);
226 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
227 		return (NULL);
228 	return (V_ifindex_table[idx].ife_ifnet);
229 }
230 
231 struct ifnet *
232 ifnet_byindex(u_short idx)
233 {
234 	struct ifnet *ifp;
235 
236 	IFNET_RLOCK_NOSLEEP();
237 	ifp = ifnet_byindex_locked(idx);
238 	IFNET_RUNLOCK_NOSLEEP();
239 	return (ifp);
240 }
241 
242 struct ifnet *
243 ifnet_byindex_ref(u_short idx)
244 {
245 	struct ifnet *ifp;
246 
247 	IFNET_RLOCK_NOSLEEP();
248 	ifp = ifnet_byindex_locked(idx);
249 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
250 		IFNET_RUNLOCK_NOSLEEP();
251 		return (NULL);
252 	}
253 	if_ref(ifp);
254 	IFNET_RUNLOCK_NOSLEEP();
255 	return (ifp);
256 }
257 
258 /*
259  * Allocate an ifindex array entry; return 0 on success or an error on
260  * failure.
261  */
262 static int
263 ifindex_alloc_locked(u_short *idxp)
264 {
265 	u_short idx;
266 
267 	IFNET_WLOCK_ASSERT();
268 
269 	/*
270 	 * Try to find an empty slot below V_if_index.  If we fail, take the
271 	 * next slot.
272 	 */
273 	for (idx = 1; idx <= V_if_index; idx++) {
274 		if (V_ifindex_table[idx].ife_ifnet == NULL)
275 			break;
276 	}
277 
278 	/* Catch if_index overflow. */
279 	if (idx < 1)
280 		return (ENOSPC);
281 	if (idx > V_if_index)
282 		V_if_index = idx;
283 	if (V_if_index >= V_if_indexlim)
284 		if_grow();
285 	*idxp = idx;
286 	return (0);
287 }
288 
289 static void
290 ifindex_free_locked(u_short idx)
291 {
292 
293 	IFNET_WLOCK_ASSERT();
294 
295 	V_ifindex_table[idx].ife_ifnet = NULL;
296 	while (V_if_index > 0 &&
297 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
298 		V_if_index--;
299 }
300 
301 static void
302 ifindex_free(u_short idx)
303 {
304 
305 	IFNET_WLOCK();
306 	ifindex_free_locked(idx);
307 	IFNET_WUNLOCK();
308 }
309 
310 static void
311 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
312 {
313 
314 	IFNET_WLOCK_ASSERT();
315 
316 	V_ifindex_table[idx].ife_ifnet = ifp;
317 }
318 
319 static void
320 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
321 {
322 
323 	IFNET_WLOCK();
324 	ifnet_setbyindex_locked(idx, ifp);
325 	IFNET_WUNLOCK();
326 }
327 
328 struct ifaddr *
329 ifaddr_byindex(u_short idx)
330 {
331 	struct ifaddr *ifa;
332 
333 	IFNET_RLOCK_NOSLEEP();
334 	ifa = ifnet_byindex_locked(idx)->if_addr;
335 	if (ifa != NULL)
336 		ifa_ref(ifa);
337 	IFNET_RUNLOCK_NOSLEEP();
338 	return (ifa);
339 }
340 
341 /*
342  * Network interface utility routines.
343  *
344  * Routines with ifa_ifwith* names take sockaddr *'s as
345  * parameters.
346  */
347 
348 static void
349 vnet_if_init(const void *unused __unused)
350 {
351 
352 	TAILQ_INIT(&V_ifnet);
353 	TAILQ_INIT(&V_ifg_head);
354 	if_grow();				/* create initial table */
355 	vnet_if_clone_init();
356 }
357 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
358     NULL);
359 
360 /* ARGSUSED*/
361 static void
362 if_init(void *dummy __unused)
363 {
364 
365 	IFNET_LOCK_INIT();
366 	if_clone_init();
367 }
368 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
369 
370 
371 #ifdef VIMAGE
372 static void
373 vnet_if_uninit(const void *unused __unused)
374 {
375 
376 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
377 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
378 
379 	free((caddr_t)V_ifindex_table, M_IFNET);
380 }
381 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
382     vnet_if_uninit, NULL);
383 #endif
384 
385 static void
386 if_grow(void)
387 {
388 	u_int n;
389 	struct ifindex_entry *e;
390 
391 	V_if_indexlim <<= 1;
392 	n = V_if_indexlim * sizeof(*e);
393 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
394 	if (V_ifindex_table != NULL) {
395 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
396 		free((caddr_t)V_ifindex_table, M_IFNET);
397 	}
398 	V_ifindex_table = e;
399 }
400 
401 /*
402  * Allocate a struct ifnet and an index for an interface.  A layer 2
403  * common structure will also be allocated if an allocation routine is
404  * registered for the passed type.
405  */
406 struct ifnet *
407 if_alloc(u_char type)
408 {
409 	struct ifnet *ifp;
410 	u_short idx;
411 
412 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
413 	IFNET_WLOCK();
414 	if (ifindex_alloc_locked(&idx) != 0) {
415 		IFNET_WUNLOCK();
416 		free(ifp, M_IFNET);
417 		return (NULL);
418 	}
419 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
420 	IFNET_WUNLOCK();
421 	ifp->if_index = idx;
422 	ifp->if_type = type;
423 	ifp->if_alloctype = type;
424 	if (if_com_alloc[type] != NULL) {
425 		ifp->if_l2com = if_com_alloc[type](type, ifp);
426 		if (ifp->if_l2com == NULL) {
427 			free(ifp, M_IFNET);
428 			ifindex_free(idx);
429 			return (NULL);
430 		}
431 	}
432 
433 	IF_ADDR_LOCK_INIT(ifp);
434 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
435 	ifp->if_afdata_initialized = 0;
436 	IF_AFDATA_LOCK_INIT(ifp);
437 	TAILQ_INIT(&ifp->if_addrhead);
438 	TAILQ_INIT(&ifp->if_prefixhead);
439 	TAILQ_INIT(&ifp->if_multiaddrs);
440 	TAILQ_INIT(&ifp->if_groups);
441 #ifdef MAC
442 	mac_ifnet_init(ifp);
443 #endif
444 	ifq_init(&ifp->if_snd, ifp);
445 
446 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
447 	ifnet_setbyindex(ifp->if_index, ifp);
448 	return (ifp);
449 }
450 
451 /*
452  * Do the actual work of freeing a struct ifnet, associated index, and layer
453  * 2 common structure.  This call is made when the last reference to an
454  * interface is released.
455  */
456 static void
457 if_free_internal(struct ifnet *ifp)
458 {
459 
460 	KASSERT((ifp->if_flags & IFF_DYING),
461 	    ("if_free_internal: interface not dying"));
462 
463 	IFNET_WLOCK();
464 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
465 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
466 
467 	ifindex_free_locked(ifp->if_index);
468 	IFNET_WUNLOCK();
469 
470 	if (if_com_free[ifp->if_alloctype] != NULL)
471 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
472 		    ifp->if_alloctype);
473 
474 #ifdef MAC
475 	mac_ifnet_destroy(ifp);
476 #endif /* MAC */
477 	if (ifp->if_description != NULL)
478 		free(ifp->if_description, M_IFDESCR);
479 	IF_AFDATA_DESTROY(ifp);
480 	IF_ADDR_LOCK_DESTROY(ifp);
481 	ifq_delete(&ifp->if_snd);
482 	free(ifp, M_IFNET);
483 }
484 
485 /*
486  * This version should only be called by intefaces that switch their type
487  * after calling if_alloc().  if_free_type() will go away again now that we
488  * have if_alloctype to cache the original allocation type.  For now, assert
489  * that they match, since we require that in practice.
490  */
491 void
492 if_free_type(struct ifnet *ifp, u_char type)
493 {
494 
495 	KASSERT(ifp->if_alloctype == type,
496 	    ("if_free_type: type (%d) != alloctype (%d)", type,
497 	    ifp->if_alloctype));
498 
499 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
500 	if (!refcount_release(&ifp->if_refcount))
501 		return;
502 	if_free_internal(ifp);
503 }
504 
505 /*
506  * This is the normal version of if_free(), used by device drivers to free a
507  * detached network interface.  The contents of if_free_type() will move into
508  * here when if_free_type() goes away.
509  */
510 void
511 if_free(struct ifnet *ifp)
512 {
513 
514 	if_free_type(ifp, ifp->if_alloctype);
515 }
516 
517 /*
518  * Interfaces to keep an ifnet type-stable despite the possibility of the
519  * driver calling if_free().  If there are additional references, we defer
520  * freeing the underlying data structure.
521  */
522 void
523 if_ref(struct ifnet *ifp)
524 {
525 
526 	/* We don't assert the ifnet list lock here, but arguably should. */
527 	refcount_acquire(&ifp->if_refcount);
528 }
529 
530 void
531 if_rele(struct ifnet *ifp)
532 {
533 
534 	if (!refcount_release(&ifp->if_refcount))
535 		return;
536 	if_free_internal(ifp);
537 }
538 
539 void
540 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
541 {
542 
543 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
544 
545 	if (ifq->ifq_maxlen == 0)
546 		ifq->ifq_maxlen = ifqmaxlen;
547 
548 	ifq->altq_type = 0;
549 	ifq->altq_disc = NULL;
550 	ifq->altq_flags &= ALTQF_CANTCHANGE;
551 	ifq->altq_tbr  = NULL;
552 	ifq->altq_ifp  = ifp;
553 }
554 
555 void
556 ifq_delete(struct ifaltq *ifq)
557 {
558 	mtx_destroy(&ifq->ifq_mtx);
559 }
560 
561 /*
562  * Perform generic interface initalization tasks and attach the interface
563  * to the list of "active" interfaces.  If vmove flag is set on entry
564  * to if_attach_internal(), perform only a limited subset of initialization
565  * tasks, given that we are moving from one vnet to another an ifnet which
566  * has already been fully initialized.
567  *
568  * XXX:
569  *  - The decision to return void and thus require this function to
570  *    succeed is questionable.
571  *  - We should probably do more sanity checking.  For instance we don't
572  *    do anything to insure if_xname is unique or non-empty.
573  */
574 void
575 if_attach(struct ifnet *ifp)
576 {
577 
578 	if_attach_internal(ifp, 0);
579 }
580 
581 static void
582 if_attach_internal(struct ifnet *ifp, int vmove)
583 {
584 	unsigned socksize, ifasize;
585 	int namelen, masklen;
586 	struct sockaddr_dl *sdl;
587 	struct ifaddr *ifa;
588 
589 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
590 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
591 		    ifp->if_xname);
592 
593 #ifdef VIMAGE
594 	ifp->if_vnet = curvnet;
595 	if (ifp->if_home_vnet == NULL)
596 		ifp->if_home_vnet = curvnet;
597 #endif
598 
599 	if_addgroup(ifp, IFG_ALL);
600 
601 	getmicrotime(&ifp->if_lastchange);
602 	ifp->if_data.ifi_epoch = time_uptime;
603 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
604 
605 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
606 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
607 	    ("transmit and qflush must both either be set or both be NULL"));
608 	if (ifp->if_transmit == NULL) {
609 		ifp->if_transmit = if_transmit;
610 		ifp->if_qflush = if_qflush;
611 	}
612 
613 	if (!vmove) {
614 #ifdef MAC
615 		mac_ifnet_create(ifp);
616 #endif
617 
618 		/*
619 		 * Create a Link Level name for this device.
620 		 */
621 		namelen = strlen(ifp->if_xname);
622 		/*
623 		 * Always save enough space for any possiable name so we
624 		 * can do a rename in place later.
625 		 */
626 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
627 		socksize = masklen + ifp->if_addrlen;
628 		if (socksize < sizeof(*sdl))
629 			socksize = sizeof(*sdl);
630 		socksize = roundup2(socksize, sizeof(long));
631 		ifasize = sizeof(*ifa) + 2 * socksize;
632 		ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
633 		ifa_init(ifa);
634 		sdl = (struct sockaddr_dl *)(ifa + 1);
635 		sdl->sdl_len = socksize;
636 		sdl->sdl_family = AF_LINK;
637 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
638 		sdl->sdl_nlen = namelen;
639 		sdl->sdl_index = ifp->if_index;
640 		sdl->sdl_type = ifp->if_type;
641 		ifp->if_addr = ifa;
642 		ifa->ifa_ifp = ifp;
643 		ifa->ifa_rtrequest = link_rtrequest;
644 		ifa->ifa_addr = (struct sockaddr *)sdl;
645 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
646 		ifa->ifa_netmask = (struct sockaddr *)sdl;
647 		sdl->sdl_len = masklen;
648 		while (namelen != 0)
649 			sdl->sdl_data[--namelen] = 0xff;
650 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
651 		/* Reliably crash if used uninitialized. */
652 		ifp->if_broadcastaddr = NULL;
653 	}
654 #ifdef VIMAGE
655 	else {
656 		/*
657 		 * Update the interface index in the link layer address
658 		 * of the interface.
659 		 */
660 		for (ifa = ifp->if_addr; ifa != NULL;
661 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
662 			if (ifa->ifa_addr->sa_family == AF_LINK) {
663 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
664 				sdl->sdl_index = ifp->if_index;
665 			}
666 		}
667 	}
668 #endif
669 
670 	IFNET_WLOCK();
671 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
672 #ifdef VIMAGE
673 	curvnet->vnet_ifcnt++;
674 #endif
675 	IFNET_WUNLOCK();
676 
677 	if (domain_init_status >= 2)
678 		if_attachdomain1(ifp);
679 
680 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
681 	if (IS_DEFAULT_VNET(curvnet))
682 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
683 
684 	/* Announce the interface. */
685 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
686 }
687 
688 static void
689 if_attachdomain(void *dummy)
690 {
691 	struct ifnet *ifp;
692 	int s;
693 
694 	s = splnet();
695 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
696 		if_attachdomain1(ifp);
697 	splx(s);
698 }
699 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
700     if_attachdomain, NULL);
701 
702 static void
703 if_attachdomain1(struct ifnet *ifp)
704 {
705 	struct domain *dp;
706 	int s;
707 
708 	s = splnet();
709 
710 	/*
711 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
712 	 * cannot lock ifp->if_afdata initialization, entirely.
713 	 */
714 	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
715 		splx(s);
716 		return;
717 	}
718 	if (ifp->if_afdata_initialized >= domain_init_status) {
719 		IF_AFDATA_UNLOCK(ifp);
720 		splx(s);
721 		printf("if_attachdomain called more than once on %s\n",
722 		    ifp->if_xname);
723 		return;
724 	}
725 	ifp->if_afdata_initialized = domain_init_status;
726 	IF_AFDATA_UNLOCK(ifp);
727 
728 	/* address family dependent data region */
729 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
730 	for (dp = domains; dp; dp = dp->dom_next) {
731 		if (dp->dom_ifattach)
732 			ifp->if_afdata[dp->dom_family] =
733 			    (*dp->dom_ifattach)(ifp);
734 	}
735 
736 	splx(s);
737 }
738 
739 /*
740  * Remove any unicast or broadcast network addresses from an interface.
741  */
742 void
743 if_purgeaddrs(struct ifnet *ifp)
744 {
745 	struct ifaddr *ifa, *next;
746 
747 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
748 		if (ifa->ifa_addr->sa_family == AF_LINK)
749 			continue;
750 #ifdef INET
751 		/* XXX: Ugly!! ad hoc just for INET */
752 		if (ifa->ifa_addr->sa_family == AF_INET) {
753 			struct ifaliasreq ifr;
754 
755 			bzero(&ifr, sizeof(ifr));
756 			ifr.ifra_addr = *ifa->ifa_addr;
757 			if (ifa->ifa_dstaddr)
758 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
759 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
760 			    NULL) == 0)
761 				continue;
762 		}
763 #endif /* INET */
764 #ifdef INET6
765 		if (ifa->ifa_addr->sa_family == AF_INET6) {
766 			in6_purgeaddr(ifa);
767 			/* ifp_addrhead is already updated */
768 			continue;
769 		}
770 #endif /* INET6 */
771 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
772 		ifa_free(ifa);
773 	}
774 }
775 
776 /*
777  * Remove any multicast network addresses from an interface when an ifnet
778  * is going away.
779  */
780 static void
781 if_purgemaddrs(struct ifnet *ifp)
782 {
783 	struct ifmultiaddr *ifma;
784 	struct ifmultiaddr *next;
785 
786 	IF_ADDR_LOCK(ifp);
787 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
788 		if_delmulti_locked(ifp, ifma, 1);
789 	IF_ADDR_UNLOCK(ifp);
790 }
791 
792 /*
793  * Detach an interface, removing it from the list of "active" interfaces.
794  * If vmove flag is set on entry to if_detach_internal(), perform only a
795  * limited subset of cleanup tasks, given that we are moving an ifnet from
796  * one vnet to another, where it must be fully operational.
797  *
798  * XXXRW: There are some significant questions about event ordering, and
799  * how to prevent things from starting to use the interface during detach.
800  */
801 void
802 if_detach(struct ifnet *ifp)
803 {
804 
805 	if_detach_internal(ifp, 0);
806 }
807 
808 static void
809 if_detach_internal(struct ifnet *ifp, int vmove)
810 {
811 	struct ifaddr *ifa;
812 	struct radix_node_head	*rnh;
813 	int i, j;
814 	struct domain *dp;
815  	struct ifnet *iter;
816  	int found = 0;
817 
818 	IFNET_WLOCK();
819 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
820 		if (iter == ifp) {
821 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
822 			found = 1;
823 			break;
824 		}
825 #ifdef VIMAGE
826 	if (found)
827 		curvnet->vnet_ifcnt--;
828 #endif
829 	IFNET_WUNLOCK();
830 	if (!found) {
831 		if (vmove)
832 			panic("%s: ifp=%p not on the ifnet tailq %p",
833 			    __func__, ifp, &V_ifnet);
834 		else
835 			return; /* XXX this should panic as well? */
836 	}
837 
838 	/*
839 	 * Remove/wait for pending events.
840 	 */
841 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
842 
843 	/*
844 	 * Remove routes and flush queues.
845 	 */
846 	if_down(ifp);
847 #ifdef ALTQ
848 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
849 		altq_disable(&ifp->if_snd);
850 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
851 		altq_detach(&ifp->if_snd);
852 #endif
853 
854 	if_purgeaddrs(ifp);
855 
856 #ifdef INET
857 	in_ifdetach(ifp);
858 #endif
859 
860 #ifdef INET6
861 	/*
862 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
863 	 * before removing routing entries below, since IPv6 interface direct
864 	 * routes are expected to be removed by the IPv6-specific kernel API.
865 	 * Otherwise, the kernel will detect some inconsistency and bark it.
866 	 */
867 	in6_ifdetach(ifp);
868 #endif
869 	if_purgemaddrs(ifp);
870 
871 	if (!vmove) {
872 		/*
873 		 * Prevent further calls into the device driver via ifnet.
874 		 */
875 		if_dead(ifp);
876 
877 		/*
878 		 * Remove link ifaddr pointer and maybe decrement if_index.
879 		 * Clean up all addresses.
880 		 */
881 		ifp->if_addr = NULL;
882 
883 		/* We can now free link ifaddr. */
884 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
885 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
886 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
887 			ifa_free(ifa);
888 		}
889 	}
890 
891 	/*
892 	 * Delete all remaining routes using this interface
893 	 * Unfortuneatly the only way to do this is to slog through
894 	 * the entire routing table looking for routes which point
895 	 * to this interface...oh well...
896 	 */
897 	for (i = 1; i <= AF_MAX; i++) {
898 		for (j = 0; j < rt_numfibs; j++) {
899 			rnh = rt_tables_get_rnh(j, i);
900 			if (rnh == NULL)
901 				continue;
902 			RADIX_NODE_HEAD_LOCK(rnh);
903 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
904 			RADIX_NODE_HEAD_UNLOCK(rnh);
905 		}
906 	}
907 
908 	/* Announce that the interface is gone. */
909 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
910 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
911 	if (IS_DEFAULT_VNET(curvnet))
912 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
913 	if_delgroups(ifp);
914 
915 	/*
916 	 * We cannot hold the lock over dom_ifdetach calls as they might
917 	 * sleep, for example trying to drain a callout, thus open up the
918 	 * theoretical race with re-attaching.
919 	 */
920 	IF_AFDATA_LOCK(ifp);
921 	i = ifp->if_afdata_initialized;
922 	ifp->if_afdata_initialized = 0;
923 	IF_AFDATA_UNLOCK(ifp);
924 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
925 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
926 			(*dp->dom_ifdetach)(ifp,
927 			    ifp->if_afdata[dp->dom_family]);
928 	}
929 }
930 
931 #ifdef VIMAGE
932 /*
933  * if_vmove() performs a limited version of if_detach() in current
934  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
935  * An attempt is made to shrink if_index in current vnet, find an
936  * unused if_index in target vnet and calls if_grow() if necessary,
937  * and finally find an unused if_xname for the target vnet.
938  */
939 void
940 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
941 {
942 	u_short idx;
943 
944 	/*
945 	 * Detach from current vnet, but preserve LLADDR info, do not
946 	 * mark as dead etc. so that the ifnet can be reattached later.
947 	 */
948 	if_detach_internal(ifp, 1);
949 
950 	/*
951 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
952 	 * the if_index for that vnet if possible.
953 	 *
954 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
955 	 * or we'd lock on one vnet and unlock on another.
956 	 */
957 	IFNET_WLOCK();
958 	ifindex_free_locked(ifp->if_index);
959 	IFNET_WUNLOCK();
960 
961 	/*
962 	 * Perform interface-specific reassignment tasks, if provided by
963 	 * the driver.
964 	 */
965 	if (ifp->if_reassign != NULL)
966 		ifp->if_reassign(ifp, new_vnet, NULL);
967 
968 	/*
969 	 * Switch to the context of the target vnet.
970 	 */
971 	CURVNET_SET_QUIET(new_vnet);
972 
973 	IFNET_WLOCK();
974 	if (ifindex_alloc_locked(&idx) != 0) {
975 		IFNET_WUNLOCK();
976 		panic("if_index overflow");
977 	}
978 	ifp->if_index = idx;
979 	ifnet_setbyindex_locked(ifp->if_index, ifp);
980 	IFNET_WUNLOCK();
981 
982 	if_attach_internal(ifp, 1);
983 
984 	CURVNET_RESTORE();
985 }
986 
987 /*
988  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
989  */
990 static int
991 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
992 {
993 	struct prison *pr;
994 	struct ifnet *difp;
995 
996 	/* Try to find the prison within our visibility. */
997 	sx_slock(&allprison_lock);
998 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
999 	sx_sunlock(&allprison_lock);
1000 	if (pr == NULL)
1001 		return (ENXIO);
1002 	prison_hold_locked(pr);
1003 	mtx_unlock(&pr->pr_mtx);
1004 
1005 	/* Do not try to move the iface from and to the same prison. */
1006 	if (pr->pr_vnet == ifp->if_vnet) {
1007 		prison_free(pr);
1008 		return (EEXIST);
1009 	}
1010 
1011 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1012 	/* XXX Lock interfaces to avoid races. */
1013 	CURVNET_SET_QUIET(pr->pr_vnet);
1014 	difp = ifunit(ifname);
1015 	CURVNET_RESTORE();
1016 	if (difp != NULL) {
1017 		prison_free(pr);
1018 		return (EEXIST);
1019 	}
1020 
1021 	/* Move the interface into the child jail/vnet. */
1022 	if_vmove(ifp, pr->pr_vnet);
1023 
1024 	/* Report the new if_xname back to the userland. */
1025 	sprintf(ifname, "%s", ifp->if_xname);
1026 
1027 	prison_free(pr);
1028 	return (0);
1029 }
1030 
1031 static int
1032 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1033 {
1034 	struct prison *pr;
1035 	struct vnet *vnet_dst;
1036 	struct ifnet *ifp;
1037 
1038 	/* Try to find the prison within our visibility. */
1039 	sx_slock(&allprison_lock);
1040 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1041 	sx_sunlock(&allprison_lock);
1042 	if (pr == NULL)
1043 		return (ENXIO);
1044 	prison_hold_locked(pr);
1045 	mtx_unlock(&pr->pr_mtx);
1046 
1047 	/* Make sure the named iface exists in the source prison/vnet. */
1048 	CURVNET_SET(pr->pr_vnet);
1049 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1050 	if (ifp == NULL) {
1051 		CURVNET_RESTORE();
1052 		prison_free(pr);
1053 		return (ENXIO);
1054 	}
1055 
1056 	/* Do not try to move the iface from and to the same prison. */
1057 	vnet_dst = TD_TO_VNET(td);
1058 	if (vnet_dst == ifp->if_vnet) {
1059 		CURVNET_RESTORE();
1060 		prison_free(pr);
1061 		return (EEXIST);
1062 	}
1063 
1064 	/* Get interface back from child jail/vnet. */
1065 	if_vmove(ifp, vnet_dst);
1066 	CURVNET_RESTORE();
1067 
1068 	/* Report the new if_xname back to the userland. */
1069 	sprintf(ifname, "%s", ifp->if_xname);
1070 
1071 	prison_free(pr);
1072 	return (0);
1073 }
1074 #endif /* VIMAGE */
1075 
1076 /*
1077  * Add a group to an interface
1078  */
1079 int
1080 if_addgroup(struct ifnet *ifp, const char *groupname)
1081 {
1082 	struct ifg_list		*ifgl;
1083 	struct ifg_group	*ifg = NULL;
1084 	struct ifg_member	*ifgm;
1085 
1086 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1087 	    groupname[strlen(groupname) - 1] <= '9')
1088 		return (EINVAL);
1089 
1090 	IFNET_WLOCK();
1091 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1092 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1093 			IFNET_WUNLOCK();
1094 			return (EEXIST);
1095 		}
1096 
1097 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1098 	    M_NOWAIT)) == NULL) {
1099 	    	IFNET_WUNLOCK();
1100 		return (ENOMEM);
1101 	}
1102 
1103 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1104 	    M_TEMP, M_NOWAIT)) == NULL) {
1105 		free(ifgl, M_TEMP);
1106 		IFNET_WUNLOCK();
1107 		return (ENOMEM);
1108 	}
1109 
1110 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1111 		if (!strcmp(ifg->ifg_group, groupname))
1112 			break;
1113 
1114 	if (ifg == NULL) {
1115 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1116 		    M_TEMP, M_NOWAIT)) == NULL) {
1117 			free(ifgl, M_TEMP);
1118 			free(ifgm, M_TEMP);
1119 			IFNET_WUNLOCK();
1120 			return (ENOMEM);
1121 		}
1122 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1123 		ifg->ifg_refcnt = 0;
1124 		TAILQ_INIT(&ifg->ifg_members);
1125 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1126 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1127 	}
1128 
1129 	ifg->ifg_refcnt++;
1130 	ifgl->ifgl_group = ifg;
1131 	ifgm->ifgm_ifp = ifp;
1132 
1133 	IF_ADDR_LOCK(ifp);
1134 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1135 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1136 	IF_ADDR_UNLOCK(ifp);
1137 
1138 	IFNET_WUNLOCK();
1139 
1140 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1141 
1142 	return (0);
1143 }
1144 
1145 /*
1146  * Remove a group from an interface
1147  */
1148 int
1149 if_delgroup(struct ifnet *ifp, const char *groupname)
1150 {
1151 	struct ifg_list		*ifgl;
1152 	struct ifg_member	*ifgm;
1153 
1154 	IFNET_WLOCK();
1155 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1156 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1157 			break;
1158 	if (ifgl == NULL) {
1159 		IFNET_WUNLOCK();
1160 		return (ENOENT);
1161 	}
1162 
1163 	IF_ADDR_LOCK(ifp);
1164 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1165 	IF_ADDR_UNLOCK(ifp);
1166 
1167 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1168 		if (ifgm->ifgm_ifp == ifp)
1169 			break;
1170 
1171 	if (ifgm != NULL) {
1172 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1173 		free(ifgm, M_TEMP);
1174 	}
1175 
1176 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1177 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1178 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1179 		free(ifgl->ifgl_group, M_TEMP);
1180 	}
1181 	IFNET_WUNLOCK();
1182 
1183 	free(ifgl, M_TEMP);
1184 
1185 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1186 
1187 	return (0);
1188 }
1189 
1190 /*
1191  * Remove an interface from all groups
1192  */
1193 static void
1194 if_delgroups(struct ifnet *ifp)
1195 {
1196 	struct ifg_list		*ifgl;
1197 	struct ifg_member	*ifgm;
1198 	char groupname[IFNAMSIZ];
1199 
1200 	IFNET_WLOCK();
1201 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1202 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1203 
1204 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1205 
1206 		IF_ADDR_LOCK(ifp);
1207 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1208 		IF_ADDR_UNLOCK(ifp);
1209 
1210 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1211 			if (ifgm->ifgm_ifp == ifp)
1212 				break;
1213 
1214 		if (ifgm != NULL) {
1215 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1216 			    ifgm_next);
1217 			free(ifgm, M_TEMP);
1218 		}
1219 
1220 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1221 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1222 			EVENTHANDLER_INVOKE(group_detach_event,
1223 			    ifgl->ifgl_group);
1224 			free(ifgl->ifgl_group, M_TEMP);
1225 		}
1226 		IFNET_WUNLOCK();
1227 
1228 		free(ifgl, M_TEMP);
1229 
1230 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1231 
1232 		IFNET_WLOCK();
1233 	}
1234 	IFNET_WUNLOCK();
1235 }
1236 
1237 /*
1238  * Stores all groups from an interface in memory pointed
1239  * to by data
1240  */
1241 static int
1242 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1243 {
1244 	int			 len, error;
1245 	struct ifg_list		*ifgl;
1246 	struct ifg_req		 ifgrq, *ifgp;
1247 	struct ifgroupreq	*ifgr = data;
1248 
1249 	if (ifgr->ifgr_len == 0) {
1250 		IF_ADDR_LOCK(ifp);
1251 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1252 			ifgr->ifgr_len += sizeof(struct ifg_req);
1253 		IF_ADDR_UNLOCK(ifp);
1254 		return (0);
1255 	}
1256 
1257 	len = ifgr->ifgr_len;
1258 	ifgp = ifgr->ifgr_groups;
1259 	/* XXX: wire */
1260 	IF_ADDR_LOCK(ifp);
1261 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1262 		if (len < sizeof(ifgrq)) {
1263 			IF_ADDR_UNLOCK(ifp);
1264 			return (EINVAL);
1265 		}
1266 		bzero(&ifgrq, sizeof ifgrq);
1267 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1268 		    sizeof(ifgrq.ifgrq_group));
1269 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1270 		    	IF_ADDR_UNLOCK(ifp);
1271 			return (error);
1272 		}
1273 		len -= sizeof(ifgrq);
1274 		ifgp++;
1275 	}
1276 	IF_ADDR_UNLOCK(ifp);
1277 
1278 	return (0);
1279 }
1280 
1281 /*
1282  * Stores all members of a group in memory pointed to by data
1283  */
1284 static int
1285 if_getgroupmembers(struct ifgroupreq *data)
1286 {
1287 	struct ifgroupreq	*ifgr = data;
1288 	struct ifg_group	*ifg;
1289 	struct ifg_member	*ifgm;
1290 	struct ifg_req		 ifgrq, *ifgp;
1291 	int			 len, error;
1292 
1293 	IFNET_RLOCK();
1294 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1295 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1296 			break;
1297 	if (ifg == NULL) {
1298 		IFNET_RUNLOCK();
1299 		return (ENOENT);
1300 	}
1301 
1302 	if (ifgr->ifgr_len == 0) {
1303 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1304 			ifgr->ifgr_len += sizeof(ifgrq);
1305 		IFNET_RUNLOCK();
1306 		return (0);
1307 	}
1308 
1309 	len = ifgr->ifgr_len;
1310 	ifgp = ifgr->ifgr_groups;
1311 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1312 		if (len < sizeof(ifgrq)) {
1313 			IFNET_RUNLOCK();
1314 			return (EINVAL);
1315 		}
1316 		bzero(&ifgrq, sizeof ifgrq);
1317 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1318 		    sizeof(ifgrq.ifgrq_member));
1319 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1320 			IFNET_RUNLOCK();
1321 			return (error);
1322 		}
1323 		len -= sizeof(ifgrq);
1324 		ifgp++;
1325 	}
1326 	IFNET_RUNLOCK();
1327 
1328 	return (0);
1329 }
1330 
1331 /*
1332  * Delete Routes for a Network Interface
1333  *
1334  * Called for each routing entry via the rnh->rnh_walktree() call above
1335  * to delete all route entries referencing a detaching network interface.
1336  *
1337  * Arguments:
1338  *	rn	pointer to node in the routing table
1339  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1340  *
1341  * Returns:
1342  *	0	successful
1343  *	errno	failed - reason indicated
1344  *
1345  */
1346 static int
1347 if_rtdel(struct radix_node *rn, void *arg)
1348 {
1349 	struct rtentry	*rt = (struct rtentry *)rn;
1350 	struct ifnet	*ifp = arg;
1351 	int		err;
1352 
1353 	if (rt->rt_ifp == ifp) {
1354 
1355 		/*
1356 		 * Protect (sorta) against walktree recursion problems
1357 		 * with cloned routes
1358 		 */
1359 		if ((rt->rt_flags & RTF_UP) == 0)
1360 			return (0);
1361 
1362 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1363 				rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
1364 				(struct rtentry **) NULL, rt->rt_fibnum);
1365 		if (err) {
1366 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1367 		}
1368 	}
1369 
1370 	return (0);
1371 }
1372 
1373 /*
1374  * Wrapper functions for struct ifnet address list locking macros.  These are
1375  * used by kernel modules to avoid encoding programming interface or binary
1376  * interface assumptions that may be violated when kernel-internal locking
1377  * approaches change.
1378  */
1379 void
1380 if_addr_rlock(struct ifnet *ifp)
1381 {
1382 
1383 	IF_ADDR_LOCK(ifp);
1384 }
1385 
1386 void
1387 if_addr_runlock(struct ifnet *ifp)
1388 {
1389 
1390 	IF_ADDR_UNLOCK(ifp);
1391 }
1392 
1393 void
1394 if_maddr_rlock(struct ifnet *ifp)
1395 {
1396 
1397 	IF_ADDR_LOCK(ifp);
1398 }
1399 
1400 void
1401 if_maddr_runlock(struct ifnet *ifp)
1402 {
1403 
1404 	IF_ADDR_UNLOCK(ifp);
1405 }
1406 
1407 /*
1408  * Reference count functions for ifaddrs.
1409  */
1410 void
1411 ifa_init(struct ifaddr *ifa)
1412 {
1413 
1414 	mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1415 	refcount_init(&ifa->ifa_refcnt, 1);
1416 }
1417 
1418 void
1419 ifa_ref(struct ifaddr *ifa)
1420 {
1421 
1422 	refcount_acquire(&ifa->ifa_refcnt);
1423 }
1424 
1425 void
1426 ifa_free(struct ifaddr *ifa)
1427 {
1428 
1429 	if (refcount_release(&ifa->ifa_refcnt)) {
1430 		mtx_destroy(&ifa->ifa_mtx);
1431 		free(ifa, M_IFADDR);
1432 	}
1433 }
1434 
1435 int
1436 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1437 {
1438 	int error = 0;
1439 	struct rtentry *rt = NULL;
1440 	struct rt_addrinfo info;
1441 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1442 
1443 	bzero(&info, sizeof(info));
1444 	info.rti_ifp = V_loif;
1445 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1446 	info.rti_info[RTAX_DST] = ia;
1447 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1448 	error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1449 
1450 	if (error == 0 && rt != NULL) {
1451 		RT_LOCK(rt);
1452 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1453 			ifa->ifa_ifp->if_type;
1454 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1455 			ifa->ifa_ifp->if_index;
1456 		RT_REMREF(rt);
1457 		RT_UNLOCK(rt);
1458 	} else if (error != 0)
1459 		log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1460 
1461 	return (error);
1462 }
1463 
1464 int
1465 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1466 {
1467 	int error = 0;
1468 	struct rt_addrinfo info;
1469 	struct sockaddr_dl null_sdl;
1470 
1471 	bzero(&null_sdl, sizeof(null_sdl));
1472 	null_sdl.sdl_len = sizeof(null_sdl);
1473 	null_sdl.sdl_family = AF_LINK;
1474 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1475 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1476 	bzero(&info, sizeof(info));
1477 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1478 	info.rti_info[RTAX_DST] = ia;
1479 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1480 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1481 
1482 	if (error != 0)
1483 		log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1484 
1485 	return (error);
1486 }
1487 
1488 /*
1489  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1490  * structs used to represent other address families, it is necessary
1491  * to perform a different comparison.
1492  */
1493 
1494 #define	sa_equal(a1, a2)	\
1495 	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1496 
1497 #define	sa_dl_equal(a1, a2)	\
1498 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1499 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1500 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1501 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1502 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1503 
1504 /*
1505  * Locate an interface based on a complete address.
1506  */
1507 /*ARGSUSED*/
1508 static struct ifaddr *
1509 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1510 {
1511 	struct ifnet *ifp;
1512 	struct ifaddr *ifa;
1513 
1514 	IFNET_RLOCK_NOSLEEP();
1515 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1516 		IF_ADDR_LOCK(ifp);
1517 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1518 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1519 				continue;
1520 			if (sa_equal(addr, ifa->ifa_addr)) {
1521 				if (getref)
1522 					ifa_ref(ifa);
1523 				IF_ADDR_UNLOCK(ifp);
1524 				goto done;
1525 			}
1526 			/* IP6 doesn't have broadcast */
1527 			if ((ifp->if_flags & IFF_BROADCAST) &&
1528 			    ifa->ifa_broadaddr &&
1529 			    ifa->ifa_broadaddr->sa_len != 0 &&
1530 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1531 				if (getref)
1532 					ifa_ref(ifa);
1533 				IF_ADDR_UNLOCK(ifp);
1534 				goto done;
1535 			}
1536 		}
1537 		IF_ADDR_UNLOCK(ifp);
1538 	}
1539 	ifa = NULL;
1540 done:
1541 	IFNET_RUNLOCK_NOSLEEP();
1542 	return (ifa);
1543 }
1544 
1545 struct ifaddr *
1546 ifa_ifwithaddr(struct sockaddr *addr)
1547 {
1548 
1549 	return (ifa_ifwithaddr_internal(addr, 1));
1550 }
1551 
1552 int
1553 ifa_ifwithaddr_check(struct sockaddr *addr)
1554 {
1555 
1556 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1557 }
1558 
1559 /*
1560  * Locate an interface based on the broadcast address.
1561  */
1562 /* ARGSUSED */
1563 struct ifaddr *
1564 ifa_ifwithbroadaddr(struct sockaddr *addr)
1565 {
1566 	struct ifnet *ifp;
1567 	struct ifaddr *ifa;
1568 
1569 	IFNET_RLOCK_NOSLEEP();
1570 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1571 		IF_ADDR_LOCK(ifp);
1572 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1573 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1574 				continue;
1575 			if ((ifp->if_flags & IFF_BROADCAST) &&
1576 			    ifa->ifa_broadaddr &&
1577 			    ifa->ifa_broadaddr->sa_len != 0 &&
1578 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1579 				ifa_ref(ifa);
1580 				IF_ADDR_UNLOCK(ifp);
1581 				goto done;
1582 			}
1583 		}
1584 		IF_ADDR_UNLOCK(ifp);
1585 	}
1586 	ifa = NULL;
1587 done:
1588 	IFNET_RUNLOCK_NOSLEEP();
1589 	return (ifa);
1590 }
1591 
1592 /*
1593  * Locate the point to point interface with a given destination address.
1594  */
1595 /*ARGSUSED*/
1596 struct ifaddr *
1597 ifa_ifwithdstaddr(struct sockaddr *addr)
1598 {
1599 	struct ifnet *ifp;
1600 	struct ifaddr *ifa;
1601 
1602 	IFNET_RLOCK_NOSLEEP();
1603 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1604 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1605 			continue;
1606 		IF_ADDR_LOCK(ifp);
1607 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1608 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1609 				continue;
1610 			if (ifa->ifa_dstaddr != NULL &&
1611 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1612 				ifa_ref(ifa);
1613 				IF_ADDR_UNLOCK(ifp);
1614 				goto done;
1615 			}
1616 		}
1617 		IF_ADDR_UNLOCK(ifp);
1618 	}
1619 	ifa = NULL;
1620 done:
1621 	IFNET_RUNLOCK_NOSLEEP();
1622 	return (ifa);
1623 }
1624 
1625 /*
1626  * Find an interface on a specific network.  If many, choice
1627  * is most specific found.
1628  */
1629 struct ifaddr *
1630 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1631 {
1632 	struct ifnet *ifp;
1633 	struct ifaddr *ifa;
1634 	struct ifaddr *ifa_maybe = NULL;
1635 	u_int af = addr->sa_family;
1636 	char *addr_data = addr->sa_data, *cplim;
1637 
1638 	/*
1639 	 * AF_LINK addresses can be looked up directly by their index number,
1640 	 * so do that if we can.
1641 	 */
1642 	if (af == AF_LINK) {
1643 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1644 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1645 		return (ifaddr_byindex(sdl->sdl_index));
1646 	}
1647 
1648 	/*
1649 	 * Scan though each interface, looking for ones that have addresses
1650 	 * in this address family.  Maintain a reference on ifa_maybe once
1651 	 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1652 	 * when we move onto the next interface.
1653 	 */
1654 	IFNET_RLOCK_NOSLEEP();
1655 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1656 		IF_ADDR_LOCK(ifp);
1657 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1658 			char *cp, *cp2, *cp3;
1659 
1660 			if (ifa->ifa_addr->sa_family != af)
1661 next:				continue;
1662 			if (af == AF_INET &&
1663 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1664 				/*
1665 				 * This is a bit broken as it doesn't
1666 				 * take into account that the remote end may
1667 				 * be a single node in the network we are
1668 				 * looking for.
1669 				 * The trouble is that we don't know the
1670 				 * netmask for the remote end.
1671 				 */
1672 				if (ifa->ifa_dstaddr != NULL &&
1673 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1674 					ifa_ref(ifa);
1675 					IF_ADDR_UNLOCK(ifp);
1676 					goto done;
1677 				}
1678 			} else {
1679 				/*
1680 				 * if we have a special address handler,
1681 				 * then use it instead of the generic one.
1682 				 */
1683 				if (ifa->ifa_claim_addr) {
1684 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1685 						ifa_ref(ifa);
1686 						IF_ADDR_UNLOCK(ifp);
1687 						goto done;
1688 					}
1689 					continue;
1690 				}
1691 
1692 				/*
1693 				 * Scan all the bits in the ifa's address.
1694 				 * If a bit dissagrees with what we are
1695 				 * looking for, mask it with the netmask
1696 				 * to see if it really matters.
1697 				 * (A byte at a time)
1698 				 */
1699 				if (ifa->ifa_netmask == 0)
1700 					continue;
1701 				cp = addr_data;
1702 				cp2 = ifa->ifa_addr->sa_data;
1703 				cp3 = ifa->ifa_netmask->sa_data;
1704 				cplim = ifa->ifa_netmask->sa_len
1705 					+ (char *)ifa->ifa_netmask;
1706 				while (cp3 < cplim)
1707 					if ((*cp++ ^ *cp2++) & *cp3++)
1708 						goto next; /* next address! */
1709 				/*
1710 				 * If the netmask of what we just found
1711 				 * is more specific than what we had before
1712 				 * (if we had one) then remember the new one
1713 				 * before continuing to search
1714 				 * for an even better one.
1715 				 */
1716 				if (ifa_maybe == NULL ||
1717 				    rn_refines((caddr_t)ifa->ifa_netmask,
1718 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1719 					if (ifa_maybe != NULL)
1720 						ifa_free(ifa_maybe);
1721 					ifa_maybe = ifa;
1722 					ifa_ref(ifa_maybe);
1723 				}
1724 			}
1725 		}
1726 		IF_ADDR_UNLOCK(ifp);
1727 	}
1728 	ifa = ifa_maybe;
1729 	ifa_maybe = NULL;
1730 done:
1731 	IFNET_RUNLOCK_NOSLEEP();
1732 	if (ifa_maybe != NULL)
1733 		ifa_free(ifa_maybe);
1734 	return (ifa);
1735 }
1736 
1737 /*
1738  * Find an interface address specific to an interface best matching
1739  * a given address.
1740  */
1741 struct ifaddr *
1742 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1743 {
1744 	struct ifaddr *ifa;
1745 	char *cp, *cp2, *cp3;
1746 	char *cplim;
1747 	struct ifaddr *ifa_maybe = NULL;
1748 	u_int af = addr->sa_family;
1749 
1750 	if (af >= AF_MAX)
1751 		return (NULL);
1752 	IF_ADDR_LOCK(ifp);
1753 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1754 		if (ifa->ifa_addr->sa_family != af)
1755 			continue;
1756 		if (ifa_maybe == NULL)
1757 			ifa_maybe = ifa;
1758 		if (ifa->ifa_netmask == 0) {
1759 			if (sa_equal(addr, ifa->ifa_addr) ||
1760 			    (ifa->ifa_dstaddr &&
1761 			    sa_equal(addr, ifa->ifa_dstaddr)))
1762 				goto done;
1763 			continue;
1764 		}
1765 		if (ifp->if_flags & IFF_POINTOPOINT) {
1766 			if (sa_equal(addr, ifa->ifa_dstaddr))
1767 				goto done;
1768 		} else {
1769 			cp = addr->sa_data;
1770 			cp2 = ifa->ifa_addr->sa_data;
1771 			cp3 = ifa->ifa_netmask->sa_data;
1772 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1773 			for (; cp3 < cplim; cp3++)
1774 				if ((*cp++ ^ *cp2++) & *cp3)
1775 					break;
1776 			if (cp3 == cplim)
1777 				goto done;
1778 		}
1779 	}
1780 	ifa = ifa_maybe;
1781 done:
1782 	if (ifa != NULL)
1783 		ifa_ref(ifa);
1784 	IF_ADDR_UNLOCK(ifp);
1785 	return (ifa);
1786 }
1787 
1788 #include <net/if_llatbl.h>
1789 
1790 /*
1791  * Default action when installing a route with a Link Level gateway.
1792  * Lookup an appropriate real ifa to point to.
1793  * This should be moved to /sys/net/link.c eventually.
1794  */
1795 static void
1796 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1797 {
1798 	struct ifaddr *ifa, *oifa;
1799 	struct sockaddr *dst;
1800 	struct ifnet *ifp;
1801 
1802 	RT_LOCK_ASSERT(rt);
1803 
1804 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1805 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1806 		return;
1807 	ifa = ifaof_ifpforaddr(dst, ifp);
1808 	if (ifa) {
1809 		oifa = rt->rt_ifa;
1810 		rt->rt_ifa = ifa;
1811 		ifa_free(oifa);
1812 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1813 			ifa->ifa_rtrequest(cmd, rt, info);
1814 	}
1815 }
1816 
1817 /*
1818  * Mark an interface down and notify protocols of
1819  * the transition.
1820  * NOTE: must be called at splnet or eqivalent.
1821  */
1822 static void
1823 if_unroute(struct ifnet *ifp, int flag, int fam)
1824 {
1825 	struct ifaddr *ifa;
1826 
1827 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1828 
1829 	ifp->if_flags &= ~flag;
1830 	getmicrotime(&ifp->if_lastchange);
1831 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1832 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1833 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1834 	ifp->if_qflush(ifp);
1835 
1836 	if (ifp->if_carp)
1837 		(*carp_linkstate_p)(ifp);
1838 	rt_ifmsg(ifp);
1839 }
1840 
1841 /*
1842  * Mark an interface up and notify protocols of
1843  * the transition.
1844  * NOTE: must be called at splnet or eqivalent.
1845  */
1846 static void
1847 if_route(struct ifnet *ifp, int flag, int fam)
1848 {
1849 	struct ifaddr *ifa;
1850 
1851 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1852 
1853 	ifp->if_flags |= flag;
1854 	getmicrotime(&ifp->if_lastchange);
1855 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1856 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1857 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1858 	if (ifp->if_carp)
1859 		(*carp_linkstate_p)(ifp);
1860 	rt_ifmsg(ifp);
1861 #ifdef INET6
1862 	in6_if_up(ifp);
1863 #endif
1864 }
1865 
1866 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1867 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1868 
1869 /*
1870  * Handle a change in the interface link state. To avoid LORs
1871  * between driver lock and upper layer locks, as well as possible
1872  * recursions, we post event to taskqueue, and all job
1873  * is done in static do_link_state_change().
1874  */
1875 void
1876 if_link_state_change(struct ifnet *ifp, int link_state)
1877 {
1878 	/* Return if state hasn't changed. */
1879 	if (ifp->if_link_state == link_state)
1880 		return;
1881 
1882 	ifp->if_link_state = link_state;
1883 
1884 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1885 }
1886 
1887 static void
1888 do_link_state_change(void *arg, int pending)
1889 {
1890 	struct ifnet *ifp = (struct ifnet *)arg;
1891 	int link_state = ifp->if_link_state;
1892 	CURVNET_SET(ifp->if_vnet);
1893 
1894 	/* Notify that the link state has changed. */
1895 	rt_ifmsg(ifp);
1896 	if (ifp->if_vlantrunk != NULL)
1897 		(*vlan_link_state_p)(ifp);
1898 
1899 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1900 	    IFP2AC(ifp)->ac_netgraph != NULL)
1901 		(*ng_ether_link_state_p)(ifp, link_state);
1902 	if (ifp->if_carp)
1903 		(*carp_linkstate_p)(ifp);
1904 	if (ifp->if_bridge) {
1905 		KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1906 		(*bstp_linkstate_p)(ifp, link_state);
1907 	}
1908 	if (ifp->if_lagg) {
1909 		KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1910 		(*lagg_linkstate_p)(ifp, link_state);
1911 	}
1912 
1913 	if (IS_DEFAULT_VNET(curvnet))
1914 		devctl_notify("IFNET", ifp->if_xname,
1915 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1916 		    NULL);
1917 	if (pending > 1)
1918 		if_printf(ifp, "%d link states coalesced\n", pending);
1919 	if (log_link_state_change)
1920 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1921 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1922 	CURVNET_RESTORE();
1923 }
1924 
1925 /*
1926  * Mark an interface down and notify protocols of
1927  * the transition.
1928  * NOTE: must be called at splnet or eqivalent.
1929  */
1930 void
1931 if_down(struct ifnet *ifp)
1932 {
1933 
1934 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1935 }
1936 
1937 /*
1938  * Mark an interface up and notify protocols of
1939  * the transition.
1940  * NOTE: must be called at splnet or eqivalent.
1941  */
1942 void
1943 if_up(struct ifnet *ifp)
1944 {
1945 
1946 	if_route(ifp, IFF_UP, AF_UNSPEC);
1947 }
1948 
1949 /*
1950  * Flush an interface queue.
1951  */
1952 void
1953 if_qflush(struct ifnet *ifp)
1954 {
1955 	struct mbuf *m, *n;
1956 	struct ifaltq *ifq;
1957 
1958 	ifq = &ifp->if_snd;
1959 	IFQ_LOCK(ifq);
1960 #ifdef ALTQ
1961 	if (ALTQ_IS_ENABLED(ifq))
1962 		ALTQ_PURGE(ifq);
1963 #endif
1964 	n = ifq->ifq_head;
1965 	while ((m = n) != 0) {
1966 		n = m->m_act;
1967 		m_freem(m);
1968 	}
1969 	ifq->ifq_head = 0;
1970 	ifq->ifq_tail = 0;
1971 	ifq->ifq_len = 0;
1972 	IFQ_UNLOCK(ifq);
1973 }
1974 
1975 /*
1976  * Map interface name to interface structure pointer, with or without
1977  * returning a reference.
1978  */
1979 struct ifnet *
1980 ifunit_ref(const char *name)
1981 {
1982 	struct ifnet *ifp;
1983 
1984 	IFNET_RLOCK_NOSLEEP();
1985 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1986 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
1987 		    !(ifp->if_flags & IFF_DYING))
1988 			break;
1989 	}
1990 	if (ifp != NULL)
1991 		if_ref(ifp);
1992 	IFNET_RUNLOCK_NOSLEEP();
1993 	return (ifp);
1994 }
1995 
1996 struct ifnet *
1997 ifunit(const char *name)
1998 {
1999 	struct ifnet *ifp;
2000 
2001 	IFNET_RLOCK_NOSLEEP();
2002 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2003 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2004 			break;
2005 	}
2006 	IFNET_RUNLOCK_NOSLEEP();
2007 	return (ifp);
2008 }
2009 
2010 /*
2011  * Hardware specific interface ioctls.
2012  */
2013 static int
2014 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2015 {
2016 	struct ifreq *ifr;
2017 	struct ifstat *ifs;
2018 	int error = 0;
2019 	int new_flags, temp_flags;
2020 	size_t namelen, onamelen;
2021 	size_t descrlen;
2022 	char *descrbuf, *odescrbuf;
2023 	char new_name[IFNAMSIZ];
2024 	struct ifaddr *ifa;
2025 	struct sockaddr_dl *sdl;
2026 
2027 	ifr = (struct ifreq *)data;
2028 	switch (cmd) {
2029 	case SIOCGIFINDEX:
2030 		ifr->ifr_index = ifp->if_index;
2031 		break;
2032 
2033 	case SIOCGIFFLAGS:
2034 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2035 		ifr->ifr_flags = temp_flags & 0xffff;
2036 		ifr->ifr_flagshigh = temp_flags >> 16;
2037 		break;
2038 
2039 	case SIOCGIFCAP:
2040 		ifr->ifr_reqcap = ifp->if_capabilities;
2041 		ifr->ifr_curcap = ifp->if_capenable;
2042 		break;
2043 
2044 #ifdef MAC
2045 	case SIOCGIFMAC:
2046 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2047 		break;
2048 #endif
2049 
2050 	case SIOCGIFMETRIC:
2051 		ifr->ifr_metric = ifp->if_metric;
2052 		break;
2053 
2054 	case SIOCGIFMTU:
2055 		ifr->ifr_mtu = ifp->if_mtu;
2056 		break;
2057 
2058 	case SIOCGIFPHYS:
2059 		ifr->ifr_phys = ifp->if_physical;
2060 		break;
2061 
2062 	case SIOCGIFDESCR:
2063 		error = 0;
2064 		sx_slock(&ifdescr_sx);
2065 		if (ifp->if_description == NULL)
2066 			error = ENOMSG;
2067 		else {
2068 			/* space for terminating nul */
2069 			descrlen = strlen(ifp->if_description) + 1;
2070 			if (ifr->ifr_buffer.length < descrlen)
2071 				ifr->ifr_buffer.buffer = NULL;
2072 			else
2073 				error = copyout(ifp->if_description,
2074 				    ifr->ifr_buffer.buffer, descrlen);
2075 			ifr->ifr_buffer.length = descrlen;
2076 		}
2077 		sx_sunlock(&ifdescr_sx);
2078 		break;
2079 
2080 	case SIOCSIFDESCR:
2081 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2082 		if (error)
2083 			return (error);
2084 
2085 		/*
2086 		 * Copy only (length-1) bytes to make sure that
2087 		 * if_description is always nul terminated.  The
2088 		 * length parameter is supposed to count the
2089 		 * terminating nul in.
2090 		 */
2091 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2092 			return (ENAMETOOLONG);
2093 		else if (ifr->ifr_buffer.length == 0)
2094 			descrbuf = NULL;
2095 		else {
2096 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2097 			    M_WAITOK | M_ZERO);
2098 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2099 			    ifr->ifr_buffer.length - 1);
2100 			if (error) {
2101 				free(descrbuf, M_IFDESCR);
2102 				break;
2103 			}
2104 		}
2105 
2106 		sx_xlock(&ifdescr_sx);
2107 		odescrbuf = ifp->if_description;
2108 		ifp->if_description = descrbuf;
2109 		sx_xunlock(&ifdescr_sx);
2110 
2111 		getmicrotime(&ifp->if_lastchange);
2112 		free(odescrbuf, M_IFDESCR);
2113 		break;
2114 
2115 	case SIOCSIFFLAGS:
2116 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2117 		if (error)
2118 			return (error);
2119 		/*
2120 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2121 		 * check, so we don't need special handling here yet.
2122 		 */
2123 		new_flags = (ifr->ifr_flags & 0xffff) |
2124 		    (ifr->ifr_flagshigh << 16);
2125 		if (ifp->if_flags & IFF_SMART) {
2126 			/* Smart drivers twiddle their own routes */
2127 		} else if (ifp->if_flags & IFF_UP &&
2128 		    (new_flags & IFF_UP) == 0) {
2129 			int s = splimp();
2130 			if_down(ifp);
2131 			splx(s);
2132 		} else if (new_flags & IFF_UP &&
2133 		    (ifp->if_flags & IFF_UP) == 0) {
2134 			int s = splimp();
2135 			if_up(ifp);
2136 			splx(s);
2137 		}
2138 		/* See if permanently promiscuous mode bit is about to flip */
2139 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2140 			if (new_flags & IFF_PPROMISC)
2141 				ifp->if_flags |= IFF_PROMISC;
2142 			else if (ifp->if_pcount == 0)
2143 				ifp->if_flags &= ~IFF_PROMISC;
2144 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2145 			    ifp->if_xname,
2146 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2147 		}
2148 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2149 			(new_flags &~ IFF_CANTCHANGE);
2150 		if (ifp->if_ioctl) {
2151 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2152 		}
2153 		getmicrotime(&ifp->if_lastchange);
2154 		break;
2155 
2156 	case SIOCSIFCAP:
2157 		error = priv_check(td, PRIV_NET_SETIFCAP);
2158 		if (error)
2159 			return (error);
2160 		if (ifp->if_ioctl == NULL)
2161 			return (EOPNOTSUPP);
2162 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2163 			return (EINVAL);
2164 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2165 		if (error == 0)
2166 			getmicrotime(&ifp->if_lastchange);
2167 		break;
2168 
2169 #ifdef MAC
2170 	case SIOCSIFMAC:
2171 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2172 		break;
2173 #endif
2174 
2175 	case SIOCSIFNAME:
2176 		error = priv_check(td, PRIV_NET_SETIFNAME);
2177 		if (error)
2178 			return (error);
2179 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2180 		if (error != 0)
2181 			return (error);
2182 		if (new_name[0] == '\0')
2183 			return (EINVAL);
2184 		if (ifunit(new_name) != NULL)
2185 			return (EEXIST);
2186 
2187 		/*
2188 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2189 		 * and there are numerous other races with the
2190 		 * ifunit() checks not being atomic with namespace
2191 		 * changes (renames, vmoves, if_attach, etc).
2192 		 */
2193 		ifp->if_flags |= IFF_RENAMING;
2194 
2195 		/* Announce the departure of the interface. */
2196 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2197 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2198 
2199 		log(LOG_INFO, "%s: changing name to '%s'\n",
2200 		    ifp->if_xname, new_name);
2201 
2202 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2203 		ifa = ifp->if_addr;
2204 		IFA_LOCK(ifa);
2205 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2206 		namelen = strlen(new_name);
2207 		onamelen = sdl->sdl_nlen;
2208 		/*
2209 		 * Move the address if needed.  This is safe because we
2210 		 * allocate space for a name of length IFNAMSIZ when we
2211 		 * create this in if_attach().
2212 		 */
2213 		if (namelen != onamelen) {
2214 			bcopy(sdl->sdl_data + onamelen,
2215 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2216 		}
2217 		bcopy(new_name, sdl->sdl_data, namelen);
2218 		sdl->sdl_nlen = namelen;
2219 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2220 		bzero(sdl->sdl_data, onamelen);
2221 		while (namelen != 0)
2222 			sdl->sdl_data[--namelen] = 0xff;
2223 		IFA_UNLOCK(ifa);
2224 
2225 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2226 		/* Announce the return of the interface. */
2227 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2228 
2229 		ifp->if_flags &= ~IFF_RENAMING;
2230 		break;
2231 
2232 #ifdef VIMAGE
2233 	case SIOCSIFVNET:
2234 		error = priv_check(td, PRIV_NET_SETIFVNET);
2235 		if (error)
2236 			return (error);
2237 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2238 		break;
2239 #endif
2240 
2241 	case SIOCSIFMETRIC:
2242 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2243 		if (error)
2244 			return (error);
2245 		ifp->if_metric = ifr->ifr_metric;
2246 		getmicrotime(&ifp->if_lastchange);
2247 		break;
2248 
2249 	case SIOCSIFPHYS:
2250 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2251 		if (error)
2252 			return (error);
2253 		if (ifp->if_ioctl == NULL)
2254 			return (EOPNOTSUPP);
2255 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2256 		if (error == 0)
2257 			getmicrotime(&ifp->if_lastchange);
2258 		break;
2259 
2260 	case SIOCSIFMTU:
2261 	{
2262 		u_long oldmtu = ifp->if_mtu;
2263 
2264 		error = priv_check(td, PRIV_NET_SETIFMTU);
2265 		if (error)
2266 			return (error);
2267 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2268 			return (EINVAL);
2269 		if (ifp->if_ioctl == NULL)
2270 			return (EOPNOTSUPP);
2271 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2272 		if (error == 0) {
2273 			getmicrotime(&ifp->if_lastchange);
2274 			rt_ifmsg(ifp);
2275 		}
2276 		/*
2277 		 * If the link MTU changed, do network layer specific procedure.
2278 		 */
2279 		if (ifp->if_mtu != oldmtu) {
2280 #ifdef INET6
2281 			nd6_setmtu(ifp);
2282 #endif
2283 		}
2284 		break;
2285 	}
2286 
2287 	case SIOCADDMULTI:
2288 	case SIOCDELMULTI:
2289 		if (cmd == SIOCADDMULTI)
2290 			error = priv_check(td, PRIV_NET_ADDMULTI);
2291 		else
2292 			error = priv_check(td, PRIV_NET_DELMULTI);
2293 		if (error)
2294 			return (error);
2295 
2296 		/* Don't allow group membership on non-multicast interfaces. */
2297 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2298 			return (EOPNOTSUPP);
2299 
2300 		/* Don't let users screw up protocols' entries. */
2301 		if (ifr->ifr_addr.sa_family != AF_LINK)
2302 			return (EINVAL);
2303 
2304 		if (cmd == SIOCADDMULTI) {
2305 			struct ifmultiaddr *ifma;
2306 
2307 			/*
2308 			 * Userland is only permitted to join groups once
2309 			 * via the if_addmulti() KPI, because it cannot hold
2310 			 * struct ifmultiaddr * between calls. It may also
2311 			 * lose a race while we check if the membership
2312 			 * already exists.
2313 			 */
2314 			IF_ADDR_LOCK(ifp);
2315 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2316 			IF_ADDR_UNLOCK(ifp);
2317 			if (ifma != NULL)
2318 				error = EADDRINUSE;
2319 			else
2320 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2321 		} else {
2322 			error = if_delmulti(ifp, &ifr->ifr_addr);
2323 		}
2324 		if (error == 0)
2325 			getmicrotime(&ifp->if_lastchange);
2326 		break;
2327 
2328 	case SIOCSIFPHYADDR:
2329 	case SIOCDIFPHYADDR:
2330 #ifdef INET6
2331 	case SIOCSIFPHYADDR_IN6:
2332 #endif
2333 	case SIOCSLIFPHYADDR:
2334 	case SIOCSIFMEDIA:
2335 	case SIOCSIFGENERIC:
2336 		error = priv_check(td, PRIV_NET_HWIOCTL);
2337 		if (error)
2338 			return (error);
2339 		if (ifp->if_ioctl == NULL)
2340 			return (EOPNOTSUPP);
2341 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2342 		if (error == 0)
2343 			getmicrotime(&ifp->if_lastchange);
2344 		break;
2345 
2346 	case SIOCGIFSTATUS:
2347 		ifs = (struct ifstat *)data;
2348 		ifs->ascii[0] = '\0';
2349 
2350 	case SIOCGIFPSRCADDR:
2351 	case SIOCGIFPDSTADDR:
2352 	case SIOCGLIFPHYADDR:
2353 	case SIOCGIFMEDIA:
2354 	case SIOCGIFGENERIC:
2355 		if (ifp->if_ioctl == NULL)
2356 			return (EOPNOTSUPP);
2357 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2358 		break;
2359 
2360 	case SIOCSIFLLADDR:
2361 		error = priv_check(td, PRIV_NET_SETLLADDR);
2362 		if (error)
2363 			return (error);
2364 		error = if_setlladdr(ifp,
2365 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2366 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2367 		break;
2368 
2369 	case SIOCAIFGROUP:
2370 	{
2371 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2372 
2373 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2374 		if (error)
2375 			return (error);
2376 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2377 			return (error);
2378 		break;
2379 	}
2380 
2381 	case SIOCGIFGROUP:
2382 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2383 			return (error);
2384 		break;
2385 
2386 	case SIOCDIFGROUP:
2387 	{
2388 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2389 
2390 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2391 		if (error)
2392 			return (error);
2393 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2394 			return (error);
2395 		break;
2396 	}
2397 
2398 	default:
2399 		error = ENOIOCTL;
2400 		break;
2401 	}
2402 	return (error);
2403 }
2404 
2405 #ifdef COMPAT_FREEBSD32
2406 struct ifconf32 {
2407 	int32_t	ifc_len;
2408 	union {
2409 		uint32_t	ifcu_buf;
2410 		uint32_t	ifcu_req;
2411 	} ifc_ifcu;
2412 };
2413 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2414 #endif
2415 
2416 /*
2417  * Interface ioctls.
2418  */
2419 int
2420 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2421 {
2422 	struct ifnet *ifp;
2423 	struct ifreq *ifr;
2424 	int error;
2425 	int oif_flags;
2426 
2427 	switch (cmd) {
2428 	case SIOCGIFCONF:
2429 	case OSIOCGIFCONF:
2430 		return (ifconf(cmd, data));
2431 
2432 #ifdef COMPAT_FREEBSD32
2433 	case SIOCGIFCONF32:
2434 		{
2435 			struct ifconf32 *ifc32;
2436 			struct ifconf ifc;
2437 
2438 			ifc32 = (struct ifconf32 *)data;
2439 			ifc.ifc_len = ifc32->ifc_len;
2440 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2441 
2442 			return (ifconf(SIOCGIFCONF, (void *)&ifc));
2443 		}
2444 #endif
2445 	}
2446 	ifr = (struct ifreq *)data;
2447 
2448 	switch (cmd) {
2449 #ifdef VIMAGE
2450 	case SIOCSIFRVNET:
2451 		error = priv_check(td, PRIV_NET_SETIFVNET);
2452 		if (error)
2453 			return (error);
2454 		return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
2455 #endif
2456 	case SIOCIFCREATE:
2457 	case SIOCIFCREATE2:
2458 		error = priv_check(td, PRIV_NET_IFCREATE);
2459 		if (error)
2460 			return (error);
2461 		return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
2462 			cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
2463 	case SIOCIFDESTROY:
2464 		error = priv_check(td, PRIV_NET_IFDESTROY);
2465 		if (error)
2466 			return (error);
2467 		return if_clone_destroy(ifr->ifr_name);
2468 
2469 	case SIOCIFGCLONERS:
2470 		return (if_clone_list((struct if_clonereq *)data));
2471 	case SIOCGIFGMEMB:
2472 		return (if_getgroupmembers((struct ifgroupreq *)data));
2473 	}
2474 
2475 	ifp = ifunit_ref(ifr->ifr_name);
2476 	if (ifp == NULL)
2477 		return (ENXIO);
2478 
2479 	error = ifhwioctl(cmd, ifp, data, td);
2480 	if (error != ENOIOCTL) {
2481 		if_rele(ifp);
2482 		return (error);
2483 	}
2484 
2485 	oif_flags = ifp->if_flags;
2486 	if (so->so_proto == NULL) {
2487 		if_rele(ifp);
2488 		return (EOPNOTSUPP);
2489 	}
2490 #ifndef COMPAT_43
2491 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2492 								 data,
2493 								 ifp, td));
2494 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2495 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2496 #else
2497 	{
2498 		u_long ocmd = cmd;
2499 
2500 		switch (cmd) {
2501 
2502 		case SIOCSIFDSTADDR:
2503 		case SIOCSIFADDR:
2504 		case SIOCSIFBRDADDR:
2505 		case SIOCSIFNETMASK:
2506 #if BYTE_ORDER != BIG_ENDIAN
2507 			if (ifr->ifr_addr.sa_family == 0 &&
2508 			    ifr->ifr_addr.sa_len < 16) {
2509 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2510 				ifr->ifr_addr.sa_len = 16;
2511 			}
2512 #else
2513 			if (ifr->ifr_addr.sa_len == 0)
2514 				ifr->ifr_addr.sa_len = 16;
2515 #endif
2516 			break;
2517 
2518 		case OSIOCGIFADDR:
2519 			cmd = SIOCGIFADDR;
2520 			break;
2521 
2522 		case OSIOCGIFDSTADDR:
2523 			cmd = SIOCGIFDSTADDR;
2524 			break;
2525 
2526 		case OSIOCGIFBRDADDR:
2527 			cmd = SIOCGIFBRDADDR;
2528 			break;
2529 
2530 		case OSIOCGIFNETMASK:
2531 			cmd = SIOCGIFNETMASK;
2532 		}
2533 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
2534 								   cmd,
2535 								   data,
2536 								   ifp, td));
2537 		if (error == EOPNOTSUPP && ifp != NULL &&
2538 		    ifp->if_ioctl != NULL)
2539 			error = (*ifp->if_ioctl)(ifp, cmd, data);
2540 		switch (ocmd) {
2541 
2542 		case OSIOCGIFADDR:
2543 		case OSIOCGIFDSTADDR:
2544 		case OSIOCGIFBRDADDR:
2545 		case OSIOCGIFNETMASK:
2546 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2547 
2548 		}
2549 	}
2550 #endif /* COMPAT_43 */
2551 
2552 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2553 #ifdef INET6
2554 		if (ifp->if_flags & IFF_UP) {
2555 			int s = splimp();
2556 			in6_if_up(ifp);
2557 			splx(s);
2558 		}
2559 #endif
2560 	}
2561 	if_rele(ifp);
2562 	return (error);
2563 }
2564 
2565 /*
2566  * The code common to handling reference counted flags,
2567  * e.g., in ifpromisc() and if_allmulti().
2568  * The "pflag" argument can specify a permanent mode flag to check,
2569  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2570  *
2571  * Only to be used on stack-owned flags, not driver-owned flags.
2572  */
2573 static int
2574 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2575 {
2576 	struct ifreq ifr;
2577 	int error;
2578 	int oldflags, oldcount;
2579 
2580 	/* Sanity checks to catch programming errors */
2581 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2582 	    ("%s: setting driver-owned flag %d", __func__, flag));
2583 
2584 	if (onswitch)
2585 		KASSERT(*refcount >= 0,
2586 		    ("%s: increment negative refcount %d for flag %d",
2587 		    __func__, *refcount, flag));
2588 	else
2589 		KASSERT(*refcount > 0,
2590 		    ("%s: decrement non-positive refcount %d for flag %d",
2591 		    __func__, *refcount, flag));
2592 
2593 	/* In case this mode is permanent, just touch refcount */
2594 	if (ifp->if_flags & pflag) {
2595 		*refcount += onswitch ? 1 : -1;
2596 		return (0);
2597 	}
2598 
2599 	/* Save ifnet parameters for if_ioctl() may fail */
2600 	oldcount = *refcount;
2601 	oldflags = ifp->if_flags;
2602 
2603 	/*
2604 	 * See if we aren't the only and touching refcount is enough.
2605 	 * Actually toggle interface flag if we are the first or last.
2606 	 */
2607 	if (onswitch) {
2608 		if ((*refcount)++)
2609 			return (0);
2610 		ifp->if_flags |= flag;
2611 	} else {
2612 		if (--(*refcount))
2613 			return (0);
2614 		ifp->if_flags &= ~flag;
2615 	}
2616 
2617 	/* Call down the driver since we've changed interface flags */
2618 	if (ifp->if_ioctl == NULL) {
2619 		error = EOPNOTSUPP;
2620 		goto recover;
2621 	}
2622 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2623 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2624 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2625 	if (error)
2626 		goto recover;
2627 	/* Notify userland that interface flags have changed */
2628 	rt_ifmsg(ifp);
2629 	return (0);
2630 
2631 recover:
2632 	/* Recover after driver error */
2633 	*refcount = oldcount;
2634 	ifp->if_flags = oldflags;
2635 	return (error);
2636 }
2637 
2638 /*
2639  * Set/clear promiscuous mode on interface ifp based on the truth value
2640  * of pswitch.  The calls are reference counted so that only the first
2641  * "on" request actually has an effect, as does the final "off" request.
2642  * Results are undefined if the "off" and "on" requests are not matched.
2643  */
2644 int
2645 ifpromisc(struct ifnet *ifp, int pswitch)
2646 {
2647 	int error;
2648 	int oldflags = ifp->if_flags;
2649 
2650 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2651 			   &ifp->if_pcount, pswitch);
2652 	/* If promiscuous mode status has changed, log a message */
2653 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2654 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2655 		    ifp->if_xname,
2656 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2657 	return (error);
2658 }
2659 
2660 /*
2661  * Return interface configuration
2662  * of system.  List may be used
2663  * in later ioctl's (above) to get
2664  * other information.
2665  */
2666 /*ARGSUSED*/
2667 static int
2668 ifconf(u_long cmd, caddr_t data)
2669 {
2670 	struct ifconf *ifc = (struct ifconf *)data;
2671 	struct ifnet *ifp;
2672 	struct ifaddr *ifa;
2673 	struct ifreq ifr;
2674 	struct sbuf *sb;
2675 	int error, full = 0, valid_len, max_len;
2676 
2677 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2678 	max_len = MAXPHYS - 1;
2679 
2680 	/* Prevent hostile input from being able to crash the system */
2681 	if (ifc->ifc_len <= 0)
2682 		return (EINVAL);
2683 
2684 again:
2685 	if (ifc->ifc_len <= max_len) {
2686 		max_len = ifc->ifc_len;
2687 		full = 1;
2688 	}
2689 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2690 	max_len = 0;
2691 	valid_len = 0;
2692 
2693 	IFNET_RLOCK();
2694 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2695 		int addrs;
2696 
2697 		/*
2698 		 * Zero the ifr_name buffer to make sure we don't
2699 		 * disclose the contents of the stack.
2700 		 */
2701 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2702 
2703 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2704 		    >= sizeof(ifr.ifr_name)) {
2705 			sbuf_delete(sb);
2706 			IFNET_RUNLOCK();
2707 			return (ENAMETOOLONG);
2708 		}
2709 
2710 		addrs = 0;
2711 		IF_ADDR_LOCK(ifp);
2712 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2713 			struct sockaddr *sa = ifa->ifa_addr;
2714 
2715 			if (prison_if(curthread->td_ucred, sa) != 0)
2716 				continue;
2717 			addrs++;
2718 #ifdef COMPAT_43
2719 			if (cmd == OSIOCGIFCONF) {
2720 				struct osockaddr *osa =
2721 					 (struct osockaddr *)&ifr.ifr_addr;
2722 				ifr.ifr_addr = *sa;
2723 				osa->sa_family = sa->sa_family;
2724 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2725 				max_len += sizeof(ifr);
2726 			} else
2727 #endif
2728 			if (sa->sa_len <= sizeof(*sa)) {
2729 				ifr.ifr_addr = *sa;
2730 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2731 				max_len += sizeof(ifr);
2732 			} else {
2733 				sbuf_bcat(sb, &ifr,
2734 				    offsetof(struct ifreq, ifr_addr));
2735 				max_len += offsetof(struct ifreq, ifr_addr);
2736 				sbuf_bcat(sb, sa, sa->sa_len);
2737 				max_len += sa->sa_len;
2738 			}
2739 
2740 			if (sbuf_error(sb) == 0)
2741 				valid_len = sbuf_len(sb);
2742 		}
2743 		IF_ADDR_UNLOCK(ifp);
2744 		if (addrs == 0) {
2745 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2746 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2747 			max_len += sizeof(ifr);
2748 
2749 			if (sbuf_error(sb) == 0)
2750 				valid_len = sbuf_len(sb);
2751 		}
2752 	}
2753 	IFNET_RUNLOCK();
2754 
2755 	/*
2756 	 * If we didn't allocate enough space (uncommon), try again.  If
2757 	 * we have already allocated as much space as we are allowed,
2758 	 * return what we've got.
2759 	 */
2760 	if (valid_len != max_len && !full) {
2761 		sbuf_delete(sb);
2762 		goto again;
2763 	}
2764 
2765 	ifc->ifc_len = valid_len;
2766 	sbuf_finish(sb);
2767 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2768 	sbuf_delete(sb);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Just like ifpromisc(), but for all-multicast-reception mode.
2774  */
2775 int
2776 if_allmulti(struct ifnet *ifp, int onswitch)
2777 {
2778 
2779 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2780 }
2781 
2782 struct ifmultiaddr *
2783 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2784 {
2785 	struct ifmultiaddr *ifma;
2786 
2787 	IF_ADDR_LOCK_ASSERT(ifp);
2788 
2789 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2790 		if (sa->sa_family == AF_LINK) {
2791 			if (sa_dl_equal(ifma->ifma_addr, sa))
2792 				break;
2793 		} else {
2794 			if (sa_equal(ifma->ifma_addr, sa))
2795 				break;
2796 		}
2797 	}
2798 
2799 	return ifma;
2800 }
2801 
2802 /*
2803  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2804  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2805  * the ifnet multicast address list here, so the caller must do that and
2806  * other setup work (such as notifying the device driver).  The reference
2807  * count is initialized to 1.
2808  */
2809 static struct ifmultiaddr *
2810 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2811     int mflags)
2812 {
2813 	struct ifmultiaddr *ifma;
2814 	struct sockaddr *dupsa;
2815 
2816 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2817 	    M_ZERO);
2818 	if (ifma == NULL)
2819 		return (NULL);
2820 
2821 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2822 	if (dupsa == NULL) {
2823 		free(ifma, M_IFMADDR);
2824 		return (NULL);
2825 	}
2826 	bcopy(sa, dupsa, sa->sa_len);
2827 	ifma->ifma_addr = dupsa;
2828 
2829 	ifma->ifma_ifp = ifp;
2830 	ifma->ifma_refcount = 1;
2831 	ifma->ifma_protospec = NULL;
2832 
2833 	if (llsa == NULL) {
2834 		ifma->ifma_lladdr = NULL;
2835 		return (ifma);
2836 	}
2837 
2838 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2839 	if (dupsa == NULL) {
2840 		free(ifma->ifma_addr, M_IFMADDR);
2841 		free(ifma, M_IFMADDR);
2842 		return (NULL);
2843 	}
2844 	bcopy(llsa, dupsa, llsa->sa_len);
2845 	ifma->ifma_lladdr = dupsa;
2846 
2847 	return (ifma);
2848 }
2849 
2850 /*
2851  * if_freemulti: free ifmultiaddr structure and possibly attached related
2852  * addresses.  The caller is responsible for implementing reference
2853  * counting, notifying the driver, handling routing messages, and releasing
2854  * any dependent link layer state.
2855  */
2856 static void
2857 if_freemulti(struct ifmultiaddr *ifma)
2858 {
2859 
2860 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2861 	    ifma->ifma_refcount));
2862 	KASSERT(ifma->ifma_protospec == NULL,
2863 	    ("if_freemulti: protospec not NULL"));
2864 
2865 	if (ifma->ifma_lladdr != NULL)
2866 		free(ifma->ifma_lladdr, M_IFMADDR);
2867 	free(ifma->ifma_addr, M_IFMADDR);
2868 	free(ifma, M_IFMADDR);
2869 }
2870 
2871 /*
2872  * Register an additional multicast address with a network interface.
2873  *
2874  * - If the address is already present, bump the reference count on the
2875  *   address and return.
2876  * - If the address is not link-layer, look up a link layer address.
2877  * - Allocate address structures for one or both addresses, and attach to the
2878  *   multicast address list on the interface.  If automatically adding a link
2879  *   layer address, the protocol address will own a reference to the link
2880  *   layer address, to be freed when it is freed.
2881  * - Notify the network device driver of an addition to the multicast address
2882  *   list.
2883  *
2884  * 'sa' points to caller-owned memory with the desired multicast address.
2885  *
2886  * 'retifma' will be used to return a pointer to the resulting multicast
2887  * address reference, if desired.
2888  */
2889 int
2890 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2891     struct ifmultiaddr **retifma)
2892 {
2893 	struct ifmultiaddr *ifma, *ll_ifma;
2894 	struct sockaddr *llsa;
2895 	int error;
2896 
2897 	/*
2898 	 * If the address is already present, return a new reference to it;
2899 	 * otherwise, allocate storage and set up a new address.
2900 	 */
2901 	IF_ADDR_LOCK(ifp);
2902 	ifma = if_findmulti(ifp, sa);
2903 	if (ifma != NULL) {
2904 		ifma->ifma_refcount++;
2905 		if (retifma != NULL)
2906 			*retifma = ifma;
2907 		IF_ADDR_UNLOCK(ifp);
2908 		return (0);
2909 	}
2910 
2911 	/*
2912 	 * The address isn't already present; resolve the protocol address
2913 	 * into a link layer address, and then look that up, bump its
2914 	 * refcount or allocate an ifma for that also.  If 'llsa' was
2915 	 * returned, we will need to free it later.
2916 	 */
2917 	llsa = NULL;
2918 	ll_ifma = NULL;
2919 	if (ifp->if_resolvemulti != NULL) {
2920 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2921 		if (error)
2922 			goto unlock_out;
2923 	}
2924 
2925 	/*
2926 	 * Allocate the new address.  Don't hook it up yet, as we may also
2927 	 * need to allocate a link layer multicast address.
2928 	 */
2929 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2930 	if (ifma == NULL) {
2931 		error = ENOMEM;
2932 		goto free_llsa_out;
2933 	}
2934 
2935 	/*
2936 	 * If a link layer address is found, we'll need to see if it's
2937 	 * already present in the address list, or allocate is as well.
2938 	 * When this block finishes, the link layer address will be on the
2939 	 * list.
2940 	 */
2941 	if (llsa != NULL) {
2942 		ll_ifma = if_findmulti(ifp, llsa);
2943 		if (ll_ifma == NULL) {
2944 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2945 			if (ll_ifma == NULL) {
2946 				--ifma->ifma_refcount;
2947 				if_freemulti(ifma);
2948 				error = ENOMEM;
2949 				goto free_llsa_out;
2950 			}
2951 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2952 			    ifma_link);
2953 		} else
2954 			ll_ifma->ifma_refcount++;
2955 		ifma->ifma_llifma = ll_ifma;
2956 	}
2957 
2958 	/*
2959 	 * We now have a new multicast address, ifma, and possibly a new or
2960 	 * referenced link layer address.  Add the primary address to the
2961 	 * ifnet address list.
2962 	 */
2963 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2964 
2965 	if (retifma != NULL)
2966 		*retifma = ifma;
2967 
2968 	/*
2969 	 * Must generate the message while holding the lock so that 'ifma'
2970 	 * pointer is still valid.
2971 	 */
2972 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2973 	IF_ADDR_UNLOCK(ifp);
2974 
2975 	/*
2976 	 * We are certain we have added something, so call down to the
2977 	 * interface to let them know about it.
2978 	 */
2979 	if (ifp->if_ioctl != NULL) {
2980 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2981 	}
2982 
2983 	if (llsa != NULL)
2984 		free(llsa, M_IFMADDR);
2985 
2986 	return (0);
2987 
2988 free_llsa_out:
2989 	if (llsa != NULL)
2990 		free(llsa, M_IFMADDR);
2991 
2992 unlock_out:
2993 	IF_ADDR_UNLOCK(ifp);
2994 	return (error);
2995 }
2996 
2997 /*
2998  * Delete a multicast group membership by network-layer group address.
2999  *
3000  * Returns ENOENT if the entry could not be found. If ifp no longer
3001  * exists, results are undefined. This entry point should only be used
3002  * from subsystems which do appropriate locking to hold ifp for the
3003  * duration of the call.
3004  * Network-layer protocol domains must use if_delmulti_ifma().
3005  */
3006 int
3007 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3008 {
3009 	struct ifmultiaddr *ifma;
3010 	int lastref;
3011 #ifdef INVARIANTS
3012 	struct ifnet *oifp;
3013 
3014 	IFNET_RLOCK_NOSLEEP();
3015 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3016 		if (ifp == oifp)
3017 			break;
3018 	if (ifp != oifp)
3019 		ifp = NULL;
3020 	IFNET_RUNLOCK_NOSLEEP();
3021 
3022 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3023 #endif
3024 	if (ifp == NULL)
3025 		return (ENOENT);
3026 
3027 	IF_ADDR_LOCK(ifp);
3028 	lastref = 0;
3029 	ifma = if_findmulti(ifp, sa);
3030 	if (ifma != NULL)
3031 		lastref = if_delmulti_locked(ifp, ifma, 0);
3032 	IF_ADDR_UNLOCK(ifp);
3033 
3034 	if (ifma == NULL)
3035 		return (ENOENT);
3036 
3037 	if (lastref && ifp->if_ioctl != NULL) {
3038 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3039 	}
3040 
3041 	return (0);
3042 }
3043 
3044 /*
3045  * Delete all multicast group membership for an interface.
3046  * Should be used to quickly flush all multicast filters.
3047  */
3048 void
3049 if_delallmulti(struct ifnet *ifp)
3050 {
3051 	struct ifmultiaddr *ifma;
3052 	struct ifmultiaddr *next;
3053 
3054 	IF_ADDR_LOCK(ifp);
3055 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3056 		if_delmulti_locked(ifp, ifma, 0);
3057 	IF_ADDR_UNLOCK(ifp);
3058 }
3059 
3060 /*
3061  * Delete a multicast group membership by group membership pointer.
3062  * Network-layer protocol domains must use this routine.
3063  *
3064  * It is safe to call this routine if the ifp disappeared.
3065  */
3066 void
3067 if_delmulti_ifma(struct ifmultiaddr *ifma)
3068 {
3069 	struct ifnet *ifp;
3070 	int lastref;
3071 
3072 	ifp = ifma->ifma_ifp;
3073 #ifdef DIAGNOSTIC
3074 	if (ifp == NULL) {
3075 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3076 	} else {
3077 		struct ifnet *oifp;
3078 
3079 		IFNET_RLOCK_NOSLEEP();
3080 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3081 			if (ifp == oifp)
3082 				break;
3083 		if (ifp != oifp) {
3084 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3085 			ifp = NULL;
3086 		}
3087 		IFNET_RUNLOCK_NOSLEEP();
3088 	}
3089 #endif
3090 	/*
3091 	 * If and only if the ifnet instance exists: Acquire the address lock.
3092 	 */
3093 	if (ifp != NULL)
3094 		IF_ADDR_LOCK(ifp);
3095 
3096 	lastref = if_delmulti_locked(ifp, ifma, 0);
3097 
3098 	if (ifp != NULL) {
3099 		/*
3100 		 * If and only if the ifnet instance exists:
3101 		 *  Release the address lock.
3102 		 *  If the group was left: update the hardware hash filter.
3103 		 */
3104 		IF_ADDR_UNLOCK(ifp);
3105 		if (lastref && ifp->if_ioctl != NULL) {
3106 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3107 		}
3108 	}
3109 }
3110 
3111 /*
3112  * Perform deletion of network-layer and/or link-layer multicast address.
3113  *
3114  * Return 0 if the reference count was decremented.
3115  * Return 1 if the final reference was released, indicating that the
3116  * hardware hash filter should be reprogrammed.
3117  */
3118 static int
3119 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3120 {
3121 	struct ifmultiaddr *ll_ifma;
3122 
3123 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3124 		KASSERT(ifma->ifma_ifp == ifp,
3125 		    ("%s: inconsistent ifp %p", __func__, ifp));
3126 		IF_ADDR_LOCK_ASSERT(ifp);
3127 	}
3128 
3129 	ifp = ifma->ifma_ifp;
3130 
3131 	/*
3132 	 * If the ifnet is detaching, null out references to ifnet,
3133 	 * so that upper protocol layers will notice, and not attempt
3134 	 * to obtain locks for an ifnet which no longer exists. The
3135 	 * routing socket announcement must happen before the ifnet
3136 	 * instance is detached from the system.
3137 	 */
3138 	if (detaching) {
3139 #ifdef DIAGNOSTIC
3140 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3141 #endif
3142 		/*
3143 		 * ifp may already be nulled out if we are being reentered
3144 		 * to delete the ll_ifma.
3145 		 */
3146 		if (ifp != NULL) {
3147 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3148 			ifma->ifma_ifp = NULL;
3149 		}
3150 	}
3151 
3152 	if (--ifma->ifma_refcount > 0)
3153 		return 0;
3154 
3155 	/*
3156 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3157 	 * have been associated with it. Release it first if so.
3158 	 */
3159 	ll_ifma = ifma->ifma_llifma;
3160 	if (ll_ifma != NULL) {
3161 		KASSERT(ifma->ifma_lladdr != NULL,
3162 		    ("%s: llifma w/o lladdr", __func__));
3163 		if (detaching)
3164 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3165 		if (--ll_ifma->ifma_refcount == 0) {
3166 			if (ifp != NULL) {
3167 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3168 				    ifma_link);
3169 			}
3170 			if_freemulti(ll_ifma);
3171 		}
3172 	}
3173 
3174 	if (ifp != NULL)
3175 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3176 
3177 	if_freemulti(ifma);
3178 
3179 	/*
3180 	 * The last reference to this instance of struct ifmultiaddr
3181 	 * was released; the hardware should be notified of this change.
3182 	 */
3183 	return 1;
3184 }
3185 
3186 /*
3187  * Set the link layer address on an interface.
3188  *
3189  * At this time we only support certain types of interfaces,
3190  * and we don't allow the length of the address to change.
3191  */
3192 int
3193 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3194 {
3195 	struct sockaddr_dl *sdl;
3196 	struct ifaddr *ifa;
3197 	struct ifreq ifr;
3198 
3199 	IF_ADDR_LOCK(ifp);
3200 	ifa = ifp->if_addr;
3201 	if (ifa == NULL) {
3202 		IF_ADDR_UNLOCK(ifp);
3203 		return (EINVAL);
3204 	}
3205 	ifa_ref(ifa);
3206 	IF_ADDR_UNLOCK(ifp);
3207 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3208 	if (sdl == NULL) {
3209 		ifa_free(ifa);
3210 		return (EINVAL);
3211 	}
3212 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3213 		ifa_free(ifa);
3214 		return (EINVAL);
3215 	}
3216 	switch (ifp->if_type) {
3217 	case IFT_ETHER:
3218 	case IFT_FDDI:
3219 	case IFT_XETHER:
3220 	case IFT_ISO88025:
3221 	case IFT_L2VLAN:
3222 	case IFT_BRIDGE:
3223 	case IFT_ARCNET:
3224 	case IFT_IEEE8023ADLAG:
3225 	case IFT_IEEE80211:
3226 		bcopy(lladdr, LLADDR(sdl), len);
3227 		ifa_free(ifa);
3228 		break;
3229 	default:
3230 		ifa_free(ifa);
3231 		return (ENODEV);
3232 	}
3233 
3234 	/*
3235 	 * If the interface is already up, we need
3236 	 * to re-init it in order to reprogram its
3237 	 * address filter.
3238 	 */
3239 	if ((ifp->if_flags & IFF_UP) != 0) {
3240 		if (ifp->if_ioctl) {
3241 			ifp->if_flags &= ~IFF_UP;
3242 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3243 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3244 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3245 			ifp->if_flags |= IFF_UP;
3246 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3247 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3248 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3249 		}
3250 #ifdef INET
3251 		/*
3252 		 * Also send gratuitous ARPs to notify other nodes about
3253 		 * the address change.
3254 		 */
3255 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3256 			if (ifa->ifa_addr->sa_family == AF_INET)
3257 				arp_ifinit(ifp, ifa);
3258 		}
3259 #endif
3260 	}
3261 	return (0);
3262 }
3263 
3264 /*
3265  * The name argument must be a pointer to storage which will last as
3266  * long as the interface does.  For physical devices, the result of
3267  * device_get_name(dev) is a good choice and for pseudo-devices a
3268  * static string works well.
3269  */
3270 void
3271 if_initname(struct ifnet *ifp, const char *name, int unit)
3272 {
3273 	ifp->if_dname = name;
3274 	ifp->if_dunit = unit;
3275 	if (unit != IF_DUNIT_NONE)
3276 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3277 	else
3278 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3279 }
3280 
3281 int
3282 if_printf(struct ifnet *ifp, const char * fmt, ...)
3283 {
3284 	va_list ap;
3285 	int retval;
3286 
3287 	retval = printf("%s: ", ifp->if_xname);
3288 	va_start(ap, fmt);
3289 	retval += vprintf(fmt, ap);
3290 	va_end(ap);
3291 	return (retval);
3292 }
3293 
3294 void
3295 if_start(struct ifnet *ifp)
3296 {
3297 
3298 	(*(ifp)->if_start)(ifp);
3299 }
3300 
3301 /*
3302  * Backwards compatibility interface for drivers
3303  * that have not implemented it
3304  */
3305 static int
3306 if_transmit(struct ifnet *ifp, struct mbuf *m)
3307 {
3308 	int error;
3309 
3310 	IFQ_HANDOFF(ifp, m, error);
3311 	return (error);
3312 }
3313 
3314 int
3315 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3316 {
3317 	int active = 0;
3318 
3319 	IF_LOCK(ifq);
3320 	if (_IF_QFULL(ifq)) {
3321 		_IF_DROP(ifq);
3322 		IF_UNLOCK(ifq);
3323 		m_freem(m);
3324 		return (0);
3325 	}
3326 	if (ifp != NULL) {
3327 		ifp->if_obytes += m->m_pkthdr.len + adjust;
3328 		if (m->m_flags & (M_BCAST|M_MCAST))
3329 			ifp->if_omcasts++;
3330 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3331 	}
3332 	_IF_ENQUEUE(ifq, m);
3333 	IF_UNLOCK(ifq);
3334 	if (ifp != NULL && !active)
3335 		(*(ifp)->if_start)(ifp);
3336 	return (1);
3337 }
3338 
3339 void
3340 if_register_com_alloc(u_char type,
3341     if_com_alloc_t *a, if_com_free_t *f)
3342 {
3343 
3344 	KASSERT(if_com_alloc[type] == NULL,
3345 	    ("if_register_com_alloc: %d already registered", type));
3346 	KASSERT(if_com_free[type] == NULL,
3347 	    ("if_register_com_alloc: %d free already registered", type));
3348 
3349 	if_com_alloc[type] = a;
3350 	if_com_free[type] = f;
3351 }
3352 
3353 void
3354 if_deregister_com_alloc(u_char type)
3355 {
3356 
3357 	KASSERT(if_com_alloc[type] != NULL,
3358 	    ("if_deregister_com_alloc: %d not registered", type));
3359 	KASSERT(if_com_free[type] != NULL,
3360 	    ("if_deregister_com_alloc: %d free not registered", type));
3361 	if_com_alloc[type] = NULL;
3362 	if_com_free[type] = NULL;
3363 }
3364