xref: /freebsd/sys/net/if.c (revision 586f63035fbe5e45cfc971037fd76375661ece26)
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 <sys/priv.h>
62 
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #include <net/if.h>
67 #include <net/if_arp.h>
68 #include <net/if_clone.h>
69 #include <net/if_dl.h>
70 #include <net/if_types.h>
71 #include <net/if_var.h>
72 #include <net/radix.h>
73 #include <net/route.h>
74 #include <net/vnet.h>
75 
76 #if defined(INET) || defined(INET6)
77 /*XXX*/
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_carp.h>
81 #ifdef INET6
82 #include <netinet6/in6_var.h>
83 #include <netinet6/in6_ifattach.h>
84 #endif
85 #endif
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif
89 
90 #include <security/mac/mac_framework.h>
91 
92 #ifdef COMPAT_FREEBSD32
93 #include <sys/mount.h>
94 #include <compat/freebsd32/freebsd32.h>
95 #endif
96 
97 struct ifindex_entry {
98 	struct  ifnet *ife_ifnet;
99 };
100 
101 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
102 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
103 
104 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106     &ifqmaxlen, 0, "max send queue size");
107 
108 /* Log link state change events */
109 static int log_link_state_change = 1;
110 
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 	&log_link_state_change, 0,
113 	"log interface link state change events");
114 
115 /* Interface description */
116 static unsigned int ifdescr_maxlen = 1024;
117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
118 	&ifdescr_maxlen, 0,
119 	"administrative maximum length for interface description");
120 
121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
122 
123 /* global sx for non-critical path ifdescr */
124 static struct sx ifdescr_sx;
125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
126 
127 void	(*bstp_linkstate_p)(struct ifnet *ifp, int state);
128 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
129 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
130 /* These are external hooks for CARP. */
131 void	(*carp_linkstate_p)(struct ifnet *ifp);
132 #if defined(INET) || defined(INET6)
133 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
134 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
135     struct sockaddr *sa);
136 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
137 int	(*carp_attach_p)(struct ifaddr *, int);
138 void	(*carp_detach_p)(struct ifaddr *);
139 #endif
140 #ifdef INET
141 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
142 #endif
143 #ifdef INET6
144 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
145 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
146     const struct in6_addr *taddr);
147 #endif
148 
149 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
150 
151 /*
152  * XXX: Style; these should be sorted alphabetically, and unprototyped
153  * static functions should be prototyped. Currently they are sorted by
154  * declaration order.
155  */
156 static void	if_attachdomain(void *);
157 static void	if_attachdomain1(struct ifnet *);
158 static int	ifconf(u_long, caddr_t);
159 static void	if_freemulti(struct ifmultiaddr *);
160 static void	if_init(void *);
161 static void	if_grow(void);
162 static void	if_route(struct ifnet *, int flag, int fam);
163 static int	if_setflag(struct ifnet *, int, int, int *, int);
164 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
165 static void	if_unroute(struct ifnet *, int flag, int fam);
166 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
167 static int	if_rtdel(struct radix_node *, void *);
168 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
169 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
170 static void	do_link_state_change(void *, int);
171 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
172 static int	if_getgroupmembers(struct ifgroupreq *);
173 static void	if_delgroups(struct ifnet *);
174 static void	if_attach_internal(struct ifnet *, int);
175 static void	if_detach_internal(struct ifnet *, int);
176 
177 #ifdef INET6
178 /*
179  * XXX: declare here to avoid to include many inet6 related files..
180  * should be more generalized?
181  */
182 extern void	nd6_setmtu(struct ifnet *);
183 #endif
184 
185 VNET_DEFINE(int, if_index);
186 int	ifqmaxlen = IFQ_MAXLEN;
187 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
188 VNET_DEFINE(struct ifgrouphead, ifg_head);
189 
190 static VNET_DEFINE(int, if_indexlim) = 8;
191 
192 /* Table of ifnet by index. */
193 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
194 
195 #define	V_if_indexlim		VNET(if_indexlim)
196 #define	V_ifindex_table		VNET(ifindex_table)
197 
198 /*
199  * The global network interface list (V_ifnet) and related state (such as
200  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
201  * an rwlock.  Either may be acquired shared to stablize the list, but both
202  * must be acquired writable to modify the list.  This model allows us to
203  * both stablize the interface list during interrupt thread processing, but
204  * also to stablize it over long-running ioctls, without introducing priority
205  * inversions and deadlocks.
206  */
207 struct rwlock ifnet_rwlock;
208 struct sx ifnet_sxlock;
209 
210 /*
211  * The allocation of network interfaces is a rather non-atomic affair; we
212  * need to select an index before we are ready to expose the interface for
213  * use, so will use this pointer value to indicate reservation.
214  */
215 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
216 
217 static	if_com_alloc_t *if_com_alloc[256];
218 static	if_com_free_t *if_com_free[256];
219 
220 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
221 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
222 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
223 
224 struct ifnet *
225 ifnet_byindex_locked(u_short idx)
226 {
227 
228 	if (idx > V_if_index)
229 		return (NULL);
230 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
231 		return (NULL);
232 	return (V_ifindex_table[idx].ife_ifnet);
233 }
234 
235 struct ifnet *
236 ifnet_byindex(u_short idx)
237 {
238 	struct ifnet *ifp;
239 
240 	IFNET_RLOCK_NOSLEEP();
241 	ifp = ifnet_byindex_locked(idx);
242 	IFNET_RUNLOCK_NOSLEEP();
243 	return (ifp);
244 }
245 
246 struct ifnet *
247 ifnet_byindex_ref(u_short idx)
248 {
249 	struct ifnet *ifp;
250 
251 	IFNET_RLOCK_NOSLEEP();
252 	ifp = ifnet_byindex_locked(idx);
253 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
254 		IFNET_RUNLOCK_NOSLEEP();
255 		return (NULL);
256 	}
257 	if_ref(ifp);
258 	IFNET_RUNLOCK_NOSLEEP();
259 	return (ifp);
260 }
261 
262 /*
263  * Allocate an ifindex array entry; return 0 on success or an error on
264  * failure.
265  */
266 static int
267 ifindex_alloc_locked(u_short *idxp)
268 {
269 	u_short idx;
270 
271 	IFNET_WLOCK_ASSERT();
272 
273 retry:
274 	/*
275 	 * Try to find an empty slot below V_if_index.  If we fail, take the
276 	 * next slot.
277 	 */
278 	for (idx = 1; idx <= V_if_index; idx++) {
279 		if (V_ifindex_table[idx].ife_ifnet == NULL)
280 			break;
281 	}
282 
283 	/* Catch if_index overflow. */
284 	if (idx < 1)
285 		return (ENOSPC);
286 	if (idx >= V_if_indexlim) {
287 		if_grow();
288 		goto retry;
289 	}
290 	if (idx > V_if_index)
291 		V_if_index = idx;
292 	*idxp = idx;
293 	return (0);
294 }
295 
296 static void
297 ifindex_free_locked(u_short idx)
298 {
299 
300 	IFNET_WLOCK_ASSERT();
301 
302 	V_ifindex_table[idx].ife_ifnet = NULL;
303 	while (V_if_index > 0 &&
304 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
305 		V_if_index--;
306 }
307 
308 static void
309 ifindex_free(u_short idx)
310 {
311 
312 	IFNET_WLOCK();
313 	ifindex_free_locked(idx);
314 	IFNET_WUNLOCK();
315 }
316 
317 static void
318 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
319 {
320 
321 	IFNET_WLOCK_ASSERT();
322 
323 	V_ifindex_table[idx].ife_ifnet = ifp;
324 }
325 
326 static void
327 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
328 {
329 
330 	IFNET_WLOCK();
331 	ifnet_setbyindex_locked(idx, ifp);
332 	IFNET_WUNLOCK();
333 }
334 
335 struct ifaddr *
336 ifaddr_byindex(u_short idx)
337 {
338 	struct ifaddr *ifa;
339 
340 	IFNET_RLOCK_NOSLEEP();
341 	ifa = ifnet_byindex_locked(idx)->if_addr;
342 	if (ifa != NULL)
343 		ifa_ref(ifa);
344 	IFNET_RUNLOCK_NOSLEEP();
345 	return (ifa);
346 }
347 
348 /*
349  * Network interface utility routines.
350  *
351  * Routines with ifa_ifwith* names take sockaddr *'s as
352  * parameters.
353  */
354 
355 static void
356 vnet_if_init(const void *unused __unused)
357 {
358 
359 	TAILQ_INIT(&V_ifnet);
360 	TAILQ_INIT(&V_ifg_head);
361 	IFNET_WLOCK();
362 	if_grow();				/* create initial table */
363 	IFNET_WUNLOCK();
364 	vnet_if_clone_init();
365 }
366 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
367     NULL);
368 
369 /* ARGSUSED*/
370 static void
371 if_init(void *dummy __unused)
372 {
373 
374 	IFNET_LOCK_INIT();
375 	if_clone_init();
376 }
377 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
378 
379 
380 #ifdef VIMAGE
381 static void
382 vnet_if_uninit(const void *unused __unused)
383 {
384 
385 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
386 	    "not empty", __func__, __LINE__, &V_ifnet));
387 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
388 	    "not empty", __func__, __LINE__, &V_ifg_head));
389 
390 	free((caddr_t)V_ifindex_table, M_IFNET);
391 }
392 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
393     vnet_if_uninit, NULL);
394 #endif
395 
396 static void
397 if_grow(void)
398 {
399 	int oldlim;
400 	u_int n;
401 	struct ifindex_entry *e;
402 
403 	IFNET_WLOCK_ASSERT();
404 	oldlim = V_if_indexlim;
405 	IFNET_WUNLOCK();
406 	n = (oldlim << 1) * sizeof(*e);
407 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
408 	IFNET_WLOCK();
409 	if (V_if_indexlim != oldlim) {
410 		free(e, M_IFNET);
411 		return;
412 	}
413 	if (V_ifindex_table != NULL) {
414 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
415 		free((caddr_t)V_ifindex_table, M_IFNET);
416 	}
417 	V_if_indexlim <<= 1;
418 	V_ifindex_table = e;
419 }
420 
421 /*
422  * Allocate a struct ifnet and an index for an interface.  A layer 2
423  * common structure will also be allocated if an allocation routine is
424  * registered for the passed type.
425  */
426 struct ifnet *
427 if_alloc(u_char type)
428 {
429 	struct ifnet *ifp;
430 	u_short idx;
431 
432 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
433 	IFNET_WLOCK();
434 	if (ifindex_alloc_locked(&idx) != 0) {
435 		IFNET_WUNLOCK();
436 		free(ifp, M_IFNET);
437 		return (NULL);
438 	}
439 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
440 	IFNET_WUNLOCK();
441 	ifp->if_index = idx;
442 	ifp->if_type = type;
443 	ifp->if_alloctype = type;
444 	if (if_com_alloc[type] != NULL) {
445 		ifp->if_l2com = if_com_alloc[type](type, ifp);
446 		if (ifp->if_l2com == NULL) {
447 			free(ifp, M_IFNET);
448 			ifindex_free(idx);
449 			return (NULL);
450 		}
451 	}
452 
453 	IF_ADDR_LOCK_INIT(ifp);
454 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
455 	ifp->if_afdata_initialized = 0;
456 	IF_AFDATA_LOCK_INIT(ifp);
457 	TAILQ_INIT(&ifp->if_addrhead);
458 	TAILQ_INIT(&ifp->if_prefixhead);
459 	TAILQ_INIT(&ifp->if_multiaddrs);
460 	TAILQ_INIT(&ifp->if_groups);
461 #ifdef MAC
462 	mac_ifnet_init(ifp);
463 #endif
464 	ifq_init(&ifp->if_snd, ifp);
465 
466 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
467 	ifnet_setbyindex(ifp->if_index, ifp);
468 	return (ifp);
469 }
470 
471 /*
472  * Do the actual work of freeing a struct ifnet, and layer 2 common
473  * structure.  This call is made when the last reference to an
474  * interface is released.
475  */
476 static void
477 if_free_internal(struct ifnet *ifp)
478 {
479 
480 	KASSERT((ifp->if_flags & IFF_DYING),
481 	    ("if_free_internal: interface not dying"));
482 
483 	if (if_com_free[ifp->if_alloctype] != NULL)
484 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
485 		    ifp->if_alloctype);
486 
487 #ifdef MAC
488 	mac_ifnet_destroy(ifp);
489 #endif /* MAC */
490 	if (ifp->if_description != NULL)
491 		free(ifp->if_description, M_IFDESCR);
492 	IF_AFDATA_DESTROY(ifp);
493 	IF_ADDR_LOCK_DESTROY(ifp);
494 	ifq_delete(&ifp->if_snd);
495 	free(ifp, M_IFNET);
496 }
497 
498 /*
499  * Deregister an interface and free the associated storage.
500  */
501 void
502 if_free(struct ifnet *ifp)
503 {
504 
505 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
506 
507 	IFNET_WLOCK();
508 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
509 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
510 
511 	ifindex_free_locked(ifp->if_index);
512 	IFNET_WUNLOCK();
513 
514 	if (!refcount_release(&ifp->if_refcount))
515 		return;
516 	if_free_internal(ifp);
517 }
518 
519 /*
520  * Interfaces to keep an ifnet type-stable despite the possibility of the
521  * driver calling if_free().  If there are additional references, we defer
522  * freeing the underlying data structure.
523  */
524 void
525 if_ref(struct ifnet *ifp)
526 {
527 
528 	/* We don't assert the ifnet list lock here, but arguably should. */
529 	refcount_acquire(&ifp->if_refcount);
530 }
531 
532 void
533 if_rele(struct ifnet *ifp)
534 {
535 
536 	if (!refcount_release(&ifp->if_refcount))
537 		return;
538 	if_free_internal(ifp);
539 }
540 
541 void
542 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
543 {
544 
545 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
546 
547 	if (ifq->ifq_maxlen == 0)
548 		ifq->ifq_maxlen = ifqmaxlen;
549 
550 	ifq->altq_type = 0;
551 	ifq->altq_disc = NULL;
552 	ifq->altq_flags &= ALTQF_CANTCHANGE;
553 	ifq->altq_tbr  = NULL;
554 	ifq->altq_ifp  = ifp;
555 }
556 
557 void
558 ifq_delete(struct ifaltq *ifq)
559 {
560 	mtx_destroy(&ifq->ifq_mtx);
561 }
562 
563 /*
564  * Perform generic interface initalization tasks and attach the interface
565  * to the list of "active" interfaces.  If vmove flag is set on entry
566  * to if_attach_internal(), perform only a limited subset of initialization
567  * tasks, given that we are moving from one vnet to another an ifnet which
568  * has already been fully initialized.
569  *
570  * XXX:
571  *  - The decision to return void and thus require this function to
572  *    succeed is questionable.
573  *  - We should probably do more sanity checking.  For instance we don't
574  *    do anything to insure if_xname is unique or non-empty.
575  */
576 void
577 if_attach(struct ifnet *ifp)
578 {
579 
580 	if_attach_internal(ifp, 0);
581 }
582 
583 static void
584 if_attach_internal(struct ifnet *ifp, int vmove)
585 {
586 	unsigned socksize, ifasize;
587 	int namelen, masklen;
588 	struct sockaddr_dl *sdl;
589 	struct ifaddr *ifa;
590 
591 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
592 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
593 		    ifp->if_xname);
594 
595 #ifdef VIMAGE
596 	ifp->if_vnet = curvnet;
597 	if (ifp->if_home_vnet == NULL)
598 		ifp->if_home_vnet = curvnet;
599 #endif
600 
601 	if_addgroup(ifp, IFG_ALL);
602 
603 	getmicrotime(&ifp->if_lastchange);
604 	ifp->if_data.ifi_epoch = time_uptime;
605 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
606 
607 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
608 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
609 	    ("transmit and qflush must both either be set or both be NULL"));
610 	if (ifp->if_transmit == NULL) {
611 		ifp->if_transmit = if_transmit;
612 		ifp->if_qflush = if_qflush;
613 	}
614 
615 	if (!vmove) {
616 #ifdef MAC
617 		mac_ifnet_create(ifp);
618 #endif
619 
620 		/*
621 		 * Create a Link Level name for this device.
622 		 */
623 		namelen = strlen(ifp->if_xname);
624 		/*
625 		 * Always save enough space for any possiable name so we
626 		 * can do a rename in place later.
627 		 */
628 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
629 		socksize = masklen + ifp->if_addrlen;
630 		if (socksize < sizeof(*sdl))
631 			socksize = sizeof(*sdl);
632 		socksize = roundup2(socksize, sizeof(long));
633 		ifasize = sizeof(*ifa) + 2 * socksize;
634 		ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
635 		ifa_init(ifa);
636 		sdl = (struct sockaddr_dl *)(ifa + 1);
637 		sdl->sdl_len = socksize;
638 		sdl->sdl_family = AF_LINK;
639 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
640 		sdl->sdl_nlen = namelen;
641 		sdl->sdl_index = ifp->if_index;
642 		sdl->sdl_type = ifp->if_type;
643 		ifp->if_addr = ifa;
644 		ifa->ifa_ifp = ifp;
645 		ifa->ifa_rtrequest = link_rtrequest;
646 		ifa->ifa_addr = (struct sockaddr *)sdl;
647 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
648 		ifa->ifa_netmask = (struct sockaddr *)sdl;
649 		sdl->sdl_len = masklen;
650 		while (namelen != 0)
651 			sdl->sdl_data[--namelen] = 0xff;
652 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
653 		/* Reliably crash if used uninitialized. */
654 		ifp->if_broadcastaddr = NULL;
655 	}
656 #ifdef VIMAGE
657 	else {
658 		/*
659 		 * Update the interface index in the link layer address
660 		 * of the interface.
661 		 */
662 		for (ifa = ifp->if_addr; ifa != NULL;
663 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
664 			if (ifa->ifa_addr->sa_family == AF_LINK) {
665 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
666 				sdl->sdl_index = ifp->if_index;
667 			}
668 		}
669 	}
670 #endif
671 
672 	IFNET_WLOCK();
673 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
674 #ifdef VIMAGE
675 	curvnet->vnet_ifcnt++;
676 #endif
677 	IFNET_WUNLOCK();
678 
679 	if (domain_init_status >= 2)
680 		if_attachdomain1(ifp);
681 
682 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
683 	if (IS_DEFAULT_VNET(curvnet))
684 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
685 
686 	/* Announce the interface. */
687 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
688 }
689 
690 static void
691 if_attachdomain(void *dummy)
692 {
693 	struct ifnet *ifp;
694 	int s;
695 
696 	s = splnet();
697 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
698 		if_attachdomain1(ifp);
699 	splx(s);
700 }
701 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
702     if_attachdomain, NULL);
703 
704 static void
705 if_attachdomain1(struct ifnet *ifp)
706 {
707 	struct domain *dp;
708 	int s;
709 
710 	s = splnet();
711 
712 	/*
713 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
714 	 * cannot lock ifp->if_afdata initialization, entirely.
715 	 */
716 	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
717 		splx(s);
718 		return;
719 	}
720 	if (ifp->if_afdata_initialized >= domain_init_status) {
721 		IF_AFDATA_UNLOCK(ifp);
722 		splx(s);
723 		printf("if_attachdomain called more than once on %s\n",
724 		    ifp->if_xname);
725 		return;
726 	}
727 	ifp->if_afdata_initialized = domain_init_status;
728 	IF_AFDATA_UNLOCK(ifp);
729 
730 	/* address family dependent data region */
731 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
732 	for (dp = domains; dp; dp = dp->dom_next) {
733 		if (dp->dom_ifattach)
734 			ifp->if_afdata[dp->dom_family] =
735 			    (*dp->dom_ifattach)(ifp);
736 	}
737 
738 	splx(s);
739 }
740 
741 /*
742  * Remove any unicast or broadcast network addresses from an interface.
743  */
744 void
745 if_purgeaddrs(struct ifnet *ifp)
746 {
747 	struct ifaddr *ifa, *next;
748 
749 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
750 		if (ifa->ifa_addr->sa_family == AF_LINK)
751 			continue;
752 #ifdef INET
753 		/* XXX: Ugly!! ad hoc just for INET */
754 		if (ifa->ifa_addr->sa_family == AF_INET) {
755 			struct ifaliasreq ifr;
756 
757 			bzero(&ifr, sizeof(ifr));
758 			ifr.ifra_addr = *ifa->ifa_addr;
759 			if (ifa->ifa_dstaddr)
760 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
761 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
762 			    NULL) == 0)
763 				continue;
764 		}
765 #endif /* INET */
766 #ifdef INET6
767 		if (ifa->ifa_addr->sa_family == AF_INET6) {
768 			in6_purgeaddr(ifa);
769 			/* ifp_addrhead is already updated */
770 			continue;
771 		}
772 #endif /* INET6 */
773 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
774 		ifa_free(ifa);
775 	}
776 }
777 
778 /*
779  * Remove any multicast network addresses from an interface when an ifnet
780  * is going away.
781  */
782 static void
783 if_purgemaddrs(struct ifnet *ifp)
784 {
785 	struct ifmultiaddr *ifma;
786 	struct ifmultiaddr *next;
787 
788 	IF_ADDR_LOCK(ifp);
789 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
790 		if_delmulti_locked(ifp, ifma, 1);
791 	IF_ADDR_UNLOCK(ifp);
792 }
793 
794 /*
795  * Detach an interface, removing it from the list of "active" interfaces.
796  * If vmove flag is set on entry to if_detach_internal(), perform only a
797  * limited subset of cleanup tasks, given that we are moving an ifnet from
798  * one vnet to another, where it must be fully operational.
799  *
800  * XXXRW: There are some significant questions about event ordering, and
801  * how to prevent things from starting to use the interface during detach.
802  */
803 void
804 if_detach(struct ifnet *ifp)
805 {
806 
807 	if_detach_internal(ifp, 0);
808 }
809 
810 static void
811 if_detach_internal(struct ifnet *ifp, int vmove)
812 {
813 	struct ifaddr *ifa;
814 	struct radix_node_head	*rnh;
815 	int i, j;
816 	struct domain *dp;
817  	struct ifnet *iter;
818  	int found = 0;
819 
820 	IFNET_WLOCK();
821 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
822 		if (iter == ifp) {
823 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
824 			found = 1;
825 			break;
826 		}
827 #ifdef VIMAGE
828 	if (found)
829 		curvnet->vnet_ifcnt--;
830 #endif
831 	IFNET_WUNLOCK();
832 	if (!found) {
833 		if (vmove)
834 			panic("%s: ifp=%p not on the ifnet tailq %p",
835 			    __func__, ifp, &V_ifnet);
836 		else
837 			return; /* XXX this should panic as well? */
838 	}
839 
840 	/*
841 	 * Remove/wait for pending events.
842 	 */
843 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
844 
845 	/*
846 	 * Remove routes and flush queues.
847 	 */
848 	if_down(ifp);
849 #ifdef ALTQ
850 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
851 		altq_disable(&ifp->if_snd);
852 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
853 		altq_detach(&ifp->if_snd);
854 #endif
855 
856 	if_purgeaddrs(ifp);
857 
858 #ifdef INET
859 	in_ifdetach(ifp);
860 #endif
861 
862 #ifdef INET6
863 	/*
864 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
865 	 * before removing routing entries below, since IPv6 interface direct
866 	 * routes are expected to be removed by the IPv6-specific kernel API.
867 	 * Otherwise, the kernel will detect some inconsistency and bark it.
868 	 */
869 	in6_ifdetach(ifp);
870 #endif
871 	if_purgemaddrs(ifp);
872 
873 	if (!vmove) {
874 		/*
875 		 * Prevent further calls into the device driver via ifnet.
876 		 */
877 		if_dead(ifp);
878 
879 		/*
880 		 * Remove link ifaddr pointer and maybe decrement if_index.
881 		 * Clean up all addresses.
882 		 */
883 		ifp->if_addr = NULL;
884 
885 		/* We can now free link ifaddr. */
886 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
887 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
888 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
889 			ifa_free(ifa);
890 		}
891 	}
892 
893 	/*
894 	 * Delete all remaining routes using this interface
895 	 * Unfortuneatly the only way to do this is to slog through
896 	 * the entire routing table looking for routes which point
897 	 * to this interface...oh well...
898 	 */
899 	for (i = 1; i <= AF_MAX; i++) {
900 		for (j = 0; j < rt_numfibs; j++) {
901 			rnh = rt_tables_get_rnh(j, i);
902 			if (rnh == NULL)
903 				continue;
904 			RADIX_NODE_HEAD_LOCK(rnh);
905 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
906 			RADIX_NODE_HEAD_UNLOCK(rnh);
907 		}
908 	}
909 
910 	/* Announce that the interface is gone. */
911 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
912 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
913 	if (IS_DEFAULT_VNET(curvnet))
914 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
915 	if_delgroups(ifp);
916 
917 	/*
918 	 * We cannot hold the lock over dom_ifdetach calls as they might
919 	 * sleep, for example trying to drain a callout, thus open up the
920 	 * theoretical race with re-attaching.
921 	 */
922 	IF_AFDATA_LOCK(ifp);
923 	i = ifp->if_afdata_initialized;
924 	ifp->if_afdata_initialized = 0;
925 	IF_AFDATA_UNLOCK(ifp);
926 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
927 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
928 			(*dp->dom_ifdetach)(ifp,
929 			    ifp->if_afdata[dp->dom_family]);
930 	}
931 }
932 
933 #ifdef VIMAGE
934 /*
935  * if_vmove() performs a limited version of if_detach() in current
936  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
937  * An attempt is made to shrink if_index in current vnet, find an
938  * unused if_index in target vnet and calls if_grow() if necessary,
939  * and finally find an unused if_xname for the target vnet.
940  */
941 void
942 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
943 {
944 	u_short idx;
945 
946 	/*
947 	 * Detach from current vnet, but preserve LLADDR info, do not
948 	 * mark as dead etc. so that the ifnet can be reattached later.
949 	 */
950 	if_detach_internal(ifp, 1);
951 
952 	/*
953 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
954 	 * the if_index for that vnet if possible.
955 	 *
956 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
957 	 * or we'd lock on one vnet and unlock on another.
958 	 */
959 	IFNET_WLOCK();
960 	ifindex_free_locked(ifp->if_index);
961 	IFNET_WUNLOCK();
962 
963 	/*
964 	 * Perform interface-specific reassignment tasks, if provided by
965 	 * the driver.
966 	 */
967 	if (ifp->if_reassign != NULL)
968 		ifp->if_reassign(ifp, new_vnet, NULL);
969 
970 	/*
971 	 * Switch to the context of the target vnet.
972 	 */
973 	CURVNET_SET_QUIET(new_vnet);
974 
975 	IFNET_WLOCK();
976 	if (ifindex_alloc_locked(&idx) != 0) {
977 		IFNET_WUNLOCK();
978 		panic("if_index overflow");
979 	}
980 	ifp->if_index = idx;
981 	ifnet_setbyindex_locked(ifp->if_index, ifp);
982 	IFNET_WUNLOCK();
983 
984 	if_attach_internal(ifp, 1);
985 
986 	CURVNET_RESTORE();
987 }
988 
989 /*
990  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
991  */
992 static int
993 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
994 {
995 	struct prison *pr;
996 	struct ifnet *difp;
997 
998 	/* Try to find the prison within our visibility. */
999 	sx_slock(&allprison_lock);
1000 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1001 	sx_sunlock(&allprison_lock);
1002 	if (pr == NULL)
1003 		return (ENXIO);
1004 	prison_hold_locked(pr);
1005 	mtx_unlock(&pr->pr_mtx);
1006 
1007 	/* Do not try to move the iface from and to the same prison. */
1008 	if (pr->pr_vnet == ifp->if_vnet) {
1009 		prison_free(pr);
1010 		return (EEXIST);
1011 	}
1012 
1013 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1014 	/* XXX Lock interfaces to avoid races. */
1015 	CURVNET_SET_QUIET(pr->pr_vnet);
1016 	difp = ifunit(ifname);
1017 	CURVNET_RESTORE();
1018 	if (difp != NULL) {
1019 		prison_free(pr);
1020 		return (EEXIST);
1021 	}
1022 
1023 	/* Move the interface into the child jail/vnet. */
1024 	if_vmove(ifp, pr->pr_vnet);
1025 
1026 	/* Report the new if_xname back to the userland. */
1027 	sprintf(ifname, "%s", ifp->if_xname);
1028 
1029 	prison_free(pr);
1030 	return (0);
1031 }
1032 
1033 static int
1034 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1035 {
1036 	struct prison *pr;
1037 	struct vnet *vnet_dst;
1038 	struct ifnet *ifp;
1039 
1040 	/* Try to find the prison within our visibility. */
1041 	sx_slock(&allprison_lock);
1042 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1043 	sx_sunlock(&allprison_lock);
1044 	if (pr == NULL)
1045 		return (ENXIO);
1046 	prison_hold_locked(pr);
1047 	mtx_unlock(&pr->pr_mtx);
1048 
1049 	/* Make sure the named iface exists in the source prison/vnet. */
1050 	CURVNET_SET(pr->pr_vnet);
1051 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1052 	if (ifp == NULL) {
1053 		CURVNET_RESTORE();
1054 		prison_free(pr);
1055 		return (ENXIO);
1056 	}
1057 
1058 	/* Do not try to move the iface from and to the same prison. */
1059 	vnet_dst = TD_TO_VNET(td);
1060 	if (vnet_dst == ifp->if_vnet) {
1061 		CURVNET_RESTORE();
1062 		prison_free(pr);
1063 		return (EEXIST);
1064 	}
1065 
1066 	/* Get interface back from child jail/vnet. */
1067 	if_vmove(ifp, vnet_dst);
1068 	CURVNET_RESTORE();
1069 
1070 	/* Report the new if_xname back to the userland. */
1071 	sprintf(ifname, "%s", ifp->if_xname);
1072 
1073 	prison_free(pr);
1074 	return (0);
1075 }
1076 #endif /* VIMAGE */
1077 
1078 /*
1079  * Add a group to an interface
1080  */
1081 int
1082 if_addgroup(struct ifnet *ifp, const char *groupname)
1083 {
1084 	struct ifg_list		*ifgl;
1085 	struct ifg_group	*ifg = NULL;
1086 	struct ifg_member	*ifgm;
1087 
1088 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1089 	    groupname[strlen(groupname) - 1] <= '9')
1090 		return (EINVAL);
1091 
1092 	IFNET_WLOCK();
1093 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1094 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1095 			IFNET_WUNLOCK();
1096 			return (EEXIST);
1097 		}
1098 
1099 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1100 	    M_NOWAIT)) == NULL) {
1101 	    	IFNET_WUNLOCK();
1102 		return (ENOMEM);
1103 	}
1104 
1105 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1106 	    M_TEMP, M_NOWAIT)) == NULL) {
1107 		free(ifgl, M_TEMP);
1108 		IFNET_WUNLOCK();
1109 		return (ENOMEM);
1110 	}
1111 
1112 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1113 		if (!strcmp(ifg->ifg_group, groupname))
1114 			break;
1115 
1116 	if (ifg == NULL) {
1117 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1118 		    M_TEMP, M_NOWAIT)) == NULL) {
1119 			free(ifgl, M_TEMP);
1120 			free(ifgm, M_TEMP);
1121 			IFNET_WUNLOCK();
1122 			return (ENOMEM);
1123 		}
1124 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1125 		ifg->ifg_refcnt = 0;
1126 		TAILQ_INIT(&ifg->ifg_members);
1127 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1128 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1129 	}
1130 
1131 	ifg->ifg_refcnt++;
1132 	ifgl->ifgl_group = ifg;
1133 	ifgm->ifgm_ifp = ifp;
1134 
1135 	IF_ADDR_LOCK(ifp);
1136 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1137 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1138 	IF_ADDR_UNLOCK(ifp);
1139 
1140 	IFNET_WUNLOCK();
1141 
1142 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1143 
1144 	return (0);
1145 }
1146 
1147 /*
1148  * Remove a group from an interface
1149  */
1150 int
1151 if_delgroup(struct ifnet *ifp, const char *groupname)
1152 {
1153 	struct ifg_list		*ifgl;
1154 	struct ifg_member	*ifgm;
1155 
1156 	IFNET_WLOCK();
1157 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1158 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1159 			break;
1160 	if (ifgl == NULL) {
1161 		IFNET_WUNLOCK();
1162 		return (ENOENT);
1163 	}
1164 
1165 	IF_ADDR_LOCK(ifp);
1166 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1167 	IF_ADDR_UNLOCK(ifp);
1168 
1169 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1170 		if (ifgm->ifgm_ifp == ifp)
1171 			break;
1172 
1173 	if (ifgm != NULL) {
1174 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1175 		free(ifgm, M_TEMP);
1176 	}
1177 
1178 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1179 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1180 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1181 		free(ifgl->ifgl_group, M_TEMP);
1182 	}
1183 	IFNET_WUNLOCK();
1184 
1185 	free(ifgl, M_TEMP);
1186 
1187 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1188 
1189 	return (0);
1190 }
1191 
1192 /*
1193  * Remove an interface from all groups
1194  */
1195 static void
1196 if_delgroups(struct ifnet *ifp)
1197 {
1198 	struct ifg_list		*ifgl;
1199 	struct ifg_member	*ifgm;
1200 	char groupname[IFNAMSIZ];
1201 
1202 	IFNET_WLOCK();
1203 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1204 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1205 
1206 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1207 
1208 		IF_ADDR_LOCK(ifp);
1209 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1210 		IF_ADDR_UNLOCK(ifp);
1211 
1212 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1213 			if (ifgm->ifgm_ifp == ifp)
1214 				break;
1215 
1216 		if (ifgm != NULL) {
1217 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1218 			    ifgm_next);
1219 			free(ifgm, M_TEMP);
1220 		}
1221 
1222 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1223 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1224 			EVENTHANDLER_INVOKE(group_detach_event,
1225 			    ifgl->ifgl_group);
1226 			free(ifgl->ifgl_group, M_TEMP);
1227 		}
1228 		IFNET_WUNLOCK();
1229 
1230 		free(ifgl, M_TEMP);
1231 
1232 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1233 
1234 		IFNET_WLOCK();
1235 	}
1236 	IFNET_WUNLOCK();
1237 }
1238 
1239 /*
1240  * Stores all groups from an interface in memory pointed
1241  * to by data
1242  */
1243 static int
1244 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1245 {
1246 	int			 len, error;
1247 	struct ifg_list		*ifgl;
1248 	struct ifg_req		 ifgrq, *ifgp;
1249 	struct ifgroupreq	*ifgr = data;
1250 
1251 	if (ifgr->ifgr_len == 0) {
1252 		IF_ADDR_LOCK(ifp);
1253 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1254 			ifgr->ifgr_len += sizeof(struct ifg_req);
1255 		IF_ADDR_UNLOCK(ifp);
1256 		return (0);
1257 	}
1258 
1259 	len = ifgr->ifgr_len;
1260 	ifgp = ifgr->ifgr_groups;
1261 	/* XXX: wire */
1262 	IF_ADDR_LOCK(ifp);
1263 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1264 		if (len < sizeof(ifgrq)) {
1265 			IF_ADDR_UNLOCK(ifp);
1266 			return (EINVAL);
1267 		}
1268 		bzero(&ifgrq, sizeof ifgrq);
1269 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1270 		    sizeof(ifgrq.ifgrq_group));
1271 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1272 		    	IF_ADDR_UNLOCK(ifp);
1273 			return (error);
1274 		}
1275 		len -= sizeof(ifgrq);
1276 		ifgp++;
1277 	}
1278 	IF_ADDR_UNLOCK(ifp);
1279 
1280 	return (0);
1281 }
1282 
1283 /*
1284  * Stores all members of a group in memory pointed to by data
1285  */
1286 static int
1287 if_getgroupmembers(struct ifgroupreq *data)
1288 {
1289 	struct ifgroupreq	*ifgr = data;
1290 	struct ifg_group	*ifg;
1291 	struct ifg_member	*ifgm;
1292 	struct ifg_req		 ifgrq, *ifgp;
1293 	int			 len, error;
1294 
1295 	IFNET_RLOCK();
1296 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1297 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1298 			break;
1299 	if (ifg == NULL) {
1300 		IFNET_RUNLOCK();
1301 		return (ENOENT);
1302 	}
1303 
1304 	if (ifgr->ifgr_len == 0) {
1305 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1306 			ifgr->ifgr_len += sizeof(ifgrq);
1307 		IFNET_RUNLOCK();
1308 		return (0);
1309 	}
1310 
1311 	len = ifgr->ifgr_len;
1312 	ifgp = ifgr->ifgr_groups;
1313 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1314 		if (len < sizeof(ifgrq)) {
1315 			IFNET_RUNLOCK();
1316 			return (EINVAL);
1317 		}
1318 		bzero(&ifgrq, sizeof ifgrq);
1319 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1320 		    sizeof(ifgrq.ifgrq_member));
1321 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1322 			IFNET_RUNLOCK();
1323 			return (error);
1324 		}
1325 		len -= sizeof(ifgrq);
1326 		ifgp++;
1327 	}
1328 	IFNET_RUNLOCK();
1329 
1330 	return (0);
1331 }
1332 
1333 /*
1334  * Delete Routes for a Network Interface
1335  *
1336  * Called for each routing entry via the rnh->rnh_walktree() call above
1337  * to delete all route entries referencing a detaching network interface.
1338  *
1339  * Arguments:
1340  *	rn	pointer to node in the routing table
1341  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1342  *
1343  * Returns:
1344  *	0	successful
1345  *	errno	failed - reason indicated
1346  *
1347  */
1348 static int
1349 if_rtdel(struct radix_node *rn, void *arg)
1350 {
1351 	struct rtentry	*rt = (struct rtentry *)rn;
1352 	struct ifnet	*ifp = arg;
1353 	int		err;
1354 
1355 	if (rt->rt_ifp == ifp) {
1356 
1357 		/*
1358 		 * Protect (sorta) against walktree recursion problems
1359 		 * with cloned routes
1360 		 */
1361 		if ((rt->rt_flags & RTF_UP) == 0)
1362 			return (0);
1363 
1364 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1365 				rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
1366 				(struct rtentry **) NULL, rt->rt_fibnum);
1367 		if (err) {
1368 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1369 		}
1370 	}
1371 
1372 	return (0);
1373 }
1374 
1375 /*
1376  * Wrapper functions for struct ifnet address list locking macros.  These are
1377  * used by kernel modules to avoid encoding programming interface or binary
1378  * interface assumptions that may be violated when kernel-internal locking
1379  * approaches change.
1380  */
1381 void
1382 if_addr_rlock(struct ifnet *ifp)
1383 {
1384 
1385 	IF_ADDR_LOCK(ifp);
1386 }
1387 
1388 void
1389 if_addr_runlock(struct ifnet *ifp)
1390 {
1391 
1392 	IF_ADDR_UNLOCK(ifp);
1393 }
1394 
1395 void
1396 if_maddr_rlock(struct ifnet *ifp)
1397 {
1398 
1399 	IF_ADDR_LOCK(ifp);
1400 }
1401 
1402 void
1403 if_maddr_runlock(struct ifnet *ifp)
1404 {
1405 
1406 	IF_ADDR_UNLOCK(ifp);
1407 }
1408 
1409 /*
1410  * Reference count functions for ifaddrs.
1411  */
1412 void
1413 ifa_init(struct ifaddr *ifa)
1414 {
1415 
1416 	mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1417 	refcount_init(&ifa->ifa_refcnt, 1);
1418 }
1419 
1420 void
1421 ifa_ref(struct ifaddr *ifa)
1422 {
1423 
1424 	refcount_acquire(&ifa->ifa_refcnt);
1425 }
1426 
1427 void
1428 ifa_free(struct ifaddr *ifa)
1429 {
1430 
1431 	if (refcount_release(&ifa->ifa_refcnt)) {
1432 		mtx_destroy(&ifa->ifa_mtx);
1433 		free(ifa, M_IFADDR);
1434 	}
1435 }
1436 
1437 int
1438 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1439 {
1440 	int error = 0;
1441 	struct rtentry *rt = NULL;
1442 	struct rt_addrinfo info;
1443 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1444 
1445 	bzero(&info, sizeof(info));
1446 	info.rti_ifp = V_loif;
1447 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1448 	info.rti_info[RTAX_DST] = ia;
1449 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1450 	error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1451 
1452 	if (error == 0 && rt != NULL) {
1453 		RT_LOCK(rt);
1454 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1455 			ifa->ifa_ifp->if_type;
1456 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1457 			ifa->ifa_ifp->if_index;
1458 		RT_REMREF(rt);
1459 		RT_UNLOCK(rt);
1460 	} else if (error != 0)
1461 		log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1462 
1463 	return (error);
1464 }
1465 
1466 int
1467 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1468 {
1469 	int error = 0;
1470 	struct rt_addrinfo info;
1471 	struct sockaddr_dl null_sdl;
1472 
1473 	bzero(&null_sdl, sizeof(null_sdl));
1474 	null_sdl.sdl_len = sizeof(null_sdl);
1475 	null_sdl.sdl_family = AF_LINK;
1476 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1477 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1478 	bzero(&info, sizeof(info));
1479 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1480 	info.rti_info[RTAX_DST] = ia;
1481 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1482 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1483 
1484 	if (error != 0)
1485 		log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1486 
1487 	return (error);
1488 }
1489 
1490 /*
1491  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1492  * structs used to represent other address families, it is necessary
1493  * to perform a different comparison.
1494  */
1495 
1496 #define	sa_equal(a1, a2)	\
1497 	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1498 
1499 #define	sa_dl_equal(a1, a2)	\
1500 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1501 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1502 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1503 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1504 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1505 
1506 /*
1507  * Locate an interface based on a complete address.
1508  */
1509 /*ARGSUSED*/
1510 static struct ifaddr *
1511 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1512 {
1513 	struct ifnet *ifp;
1514 	struct ifaddr *ifa;
1515 
1516 	IFNET_RLOCK_NOSLEEP();
1517 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1518 		IF_ADDR_LOCK(ifp);
1519 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1520 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1521 				continue;
1522 			if (sa_equal(addr, ifa->ifa_addr)) {
1523 				if (getref)
1524 					ifa_ref(ifa);
1525 				IF_ADDR_UNLOCK(ifp);
1526 				goto done;
1527 			}
1528 			/* IP6 doesn't have broadcast */
1529 			if ((ifp->if_flags & IFF_BROADCAST) &&
1530 			    ifa->ifa_broadaddr &&
1531 			    ifa->ifa_broadaddr->sa_len != 0 &&
1532 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1533 				if (getref)
1534 					ifa_ref(ifa);
1535 				IF_ADDR_UNLOCK(ifp);
1536 				goto done;
1537 			}
1538 		}
1539 		IF_ADDR_UNLOCK(ifp);
1540 	}
1541 	ifa = NULL;
1542 done:
1543 	IFNET_RUNLOCK_NOSLEEP();
1544 	return (ifa);
1545 }
1546 
1547 struct ifaddr *
1548 ifa_ifwithaddr(struct sockaddr *addr)
1549 {
1550 
1551 	return (ifa_ifwithaddr_internal(addr, 1));
1552 }
1553 
1554 int
1555 ifa_ifwithaddr_check(struct sockaddr *addr)
1556 {
1557 
1558 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1559 }
1560 
1561 /*
1562  * Locate an interface based on the broadcast address.
1563  */
1564 /* ARGSUSED */
1565 struct ifaddr *
1566 ifa_ifwithbroadaddr(struct sockaddr *addr)
1567 {
1568 	struct ifnet *ifp;
1569 	struct ifaddr *ifa;
1570 
1571 	IFNET_RLOCK_NOSLEEP();
1572 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1573 		IF_ADDR_LOCK(ifp);
1574 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1575 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1576 				continue;
1577 			if ((ifp->if_flags & IFF_BROADCAST) &&
1578 			    ifa->ifa_broadaddr &&
1579 			    ifa->ifa_broadaddr->sa_len != 0 &&
1580 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1581 				ifa_ref(ifa);
1582 				IF_ADDR_UNLOCK(ifp);
1583 				goto done;
1584 			}
1585 		}
1586 		IF_ADDR_UNLOCK(ifp);
1587 	}
1588 	ifa = NULL;
1589 done:
1590 	IFNET_RUNLOCK_NOSLEEP();
1591 	return (ifa);
1592 }
1593 
1594 /*
1595  * Locate the point to point interface with a given destination address.
1596  */
1597 /*ARGSUSED*/
1598 struct ifaddr *
1599 ifa_ifwithdstaddr(struct sockaddr *addr)
1600 {
1601 	struct ifnet *ifp;
1602 	struct ifaddr *ifa;
1603 
1604 	IFNET_RLOCK_NOSLEEP();
1605 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1606 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1607 			continue;
1608 		IF_ADDR_LOCK(ifp);
1609 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1610 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1611 				continue;
1612 			if (ifa->ifa_dstaddr != NULL &&
1613 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1614 				ifa_ref(ifa);
1615 				IF_ADDR_UNLOCK(ifp);
1616 				goto done;
1617 			}
1618 		}
1619 		IF_ADDR_UNLOCK(ifp);
1620 	}
1621 	ifa = NULL;
1622 done:
1623 	IFNET_RUNLOCK_NOSLEEP();
1624 	return (ifa);
1625 }
1626 
1627 /*
1628  * Find an interface on a specific network.  If many, choice
1629  * is most specific found.
1630  */
1631 struct ifaddr *
1632 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1633 {
1634 	struct ifnet *ifp;
1635 	struct ifaddr *ifa;
1636 	struct ifaddr *ifa_maybe = NULL;
1637 	u_int af = addr->sa_family;
1638 	char *addr_data = addr->sa_data, *cplim;
1639 
1640 	/*
1641 	 * AF_LINK addresses can be looked up directly by their index number,
1642 	 * so do that if we can.
1643 	 */
1644 	if (af == AF_LINK) {
1645 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1646 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1647 		return (ifaddr_byindex(sdl->sdl_index));
1648 	}
1649 
1650 	/*
1651 	 * Scan though each interface, looking for ones that have addresses
1652 	 * in this address family.  Maintain a reference on ifa_maybe once
1653 	 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1654 	 * when we move onto the next interface.
1655 	 */
1656 	IFNET_RLOCK_NOSLEEP();
1657 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1658 		IF_ADDR_LOCK(ifp);
1659 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1660 			char *cp, *cp2, *cp3;
1661 
1662 			if (ifa->ifa_addr->sa_family != af)
1663 next:				continue;
1664 			if (af == AF_INET &&
1665 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1666 				/*
1667 				 * This is a bit broken as it doesn't
1668 				 * take into account that the remote end may
1669 				 * be a single node in the network we are
1670 				 * looking for.
1671 				 * The trouble is that we don't know the
1672 				 * netmask for the remote end.
1673 				 */
1674 				if (ifa->ifa_dstaddr != NULL &&
1675 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1676 					ifa_ref(ifa);
1677 					IF_ADDR_UNLOCK(ifp);
1678 					goto done;
1679 				}
1680 			} else {
1681 				/*
1682 				 * if we have a special address handler,
1683 				 * then use it instead of the generic one.
1684 				 */
1685 				if (ifa->ifa_claim_addr) {
1686 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1687 						ifa_ref(ifa);
1688 						IF_ADDR_UNLOCK(ifp);
1689 						goto done;
1690 					}
1691 					continue;
1692 				}
1693 
1694 				/*
1695 				 * Scan all the bits in the ifa's address.
1696 				 * If a bit dissagrees with what we are
1697 				 * looking for, mask it with the netmask
1698 				 * to see if it really matters.
1699 				 * (A byte at a time)
1700 				 */
1701 				if (ifa->ifa_netmask == 0)
1702 					continue;
1703 				cp = addr_data;
1704 				cp2 = ifa->ifa_addr->sa_data;
1705 				cp3 = ifa->ifa_netmask->sa_data;
1706 				cplim = ifa->ifa_netmask->sa_len
1707 					+ (char *)ifa->ifa_netmask;
1708 				while (cp3 < cplim)
1709 					if ((*cp++ ^ *cp2++) & *cp3++)
1710 						goto next; /* next address! */
1711 				/*
1712 				 * If the netmask of what we just found
1713 				 * is more specific than what we had before
1714 				 * (if we had one) then remember the new one
1715 				 * before continuing to search
1716 				 * for an even better one.
1717 				 */
1718 				if (ifa_maybe == NULL ||
1719 				    rn_refines((caddr_t)ifa->ifa_netmask,
1720 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1721 					if (ifa_maybe != NULL)
1722 						ifa_free(ifa_maybe);
1723 					ifa_maybe = ifa;
1724 					ifa_ref(ifa_maybe);
1725 				}
1726 			}
1727 		}
1728 		IF_ADDR_UNLOCK(ifp);
1729 	}
1730 	ifa = ifa_maybe;
1731 	ifa_maybe = NULL;
1732 done:
1733 	IFNET_RUNLOCK_NOSLEEP();
1734 	if (ifa_maybe != NULL)
1735 		ifa_free(ifa_maybe);
1736 	return (ifa);
1737 }
1738 
1739 /*
1740  * Find an interface address specific to an interface best matching
1741  * a given address.
1742  */
1743 struct ifaddr *
1744 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1745 {
1746 	struct ifaddr *ifa;
1747 	char *cp, *cp2, *cp3;
1748 	char *cplim;
1749 	struct ifaddr *ifa_maybe = NULL;
1750 	u_int af = addr->sa_family;
1751 
1752 	if (af >= AF_MAX)
1753 		return (NULL);
1754 	IF_ADDR_LOCK(ifp);
1755 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1756 		if (ifa->ifa_addr->sa_family != af)
1757 			continue;
1758 		if (ifa_maybe == NULL)
1759 			ifa_maybe = ifa;
1760 		if (ifa->ifa_netmask == 0) {
1761 			if (sa_equal(addr, ifa->ifa_addr) ||
1762 			    (ifa->ifa_dstaddr &&
1763 			    sa_equal(addr, ifa->ifa_dstaddr)))
1764 				goto done;
1765 			continue;
1766 		}
1767 		if (ifp->if_flags & IFF_POINTOPOINT) {
1768 			if (sa_equal(addr, ifa->ifa_dstaddr))
1769 				goto done;
1770 		} else {
1771 			cp = addr->sa_data;
1772 			cp2 = ifa->ifa_addr->sa_data;
1773 			cp3 = ifa->ifa_netmask->sa_data;
1774 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1775 			for (; cp3 < cplim; cp3++)
1776 				if ((*cp++ ^ *cp2++) & *cp3)
1777 					break;
1778 			if (cp3 == cplim)
1779 				goto done;
1780 		}
1781 	}
1782 	ifa = ifa_maybe;
1783 done:
1784 	if (ifa != NULL)
1785 		ifa_ref(ifa);
1786 	IF_ADDR_UNLOCK(ifp);
1787 	return (ifa);
1788 }
1789 
1790 #include <net/if_llatbl.h>
1791 
1792 /*
1793  * Default action when installing a route with a Link Level gateway.
1794  * Lookup an appropriate real ifa to point to.
1795  * This should be moved to /sys/net/link.c eventually.
1796  */
1797 static void
1798 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1799 {
1800 	struct ifaddr *ifa, *oifa;
1801 	struct sockaddr *dst;
1802 	struct ifnet *ifp;
1803 
1804 	RT_LOCK_ASSERT(rt);
1805 
1806 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1807 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1808 		return;
1809 	ifa = ifaof_ifpforaddr(dst, ifp);
1810 	if (ifa) {
1811 		oifa = rt->rt_ifa;
1812 		rt->rt_ifa = ifa;
1813 		ifa_free(oifa);
1814 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1815 			ifa->ifa_rtrequest(cmd, rt, info);
1816 	}
1817 }
1818 
1819 /*
1820  * Mark an interface down and notify protocols of
1821  * the transition.
1822  * NOTE: must be called at splnet or eqivalent.
1823  */
1824 static void
1825 if_unroute(struct ifnet *ifp, int flag, int fam)
1826 {
1827 	struct ifaddr *ifa;
1828 
1829 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1830 
1831 	ifp->if_flags &= ~flag;
1832 	getmicrotime(&ifp->if_lastchange);
1833 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1834 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1835 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1836 	ifp->if_qflush(ifp);
1837 
1838 	if (ifp->if_carp)
1839 		(*carp_linkstate_p)(ifp);
1840 	rt_ifmsg(ifp);
1841 }
1842 
1843 /*
1844  * Mark an interface up and notify protocols of
1845  * the transition.
1846  * NOTE: must be called at splnet or eqivalent.
1847  */
1848 static void
1849 if_route(struct ifnet *ifp, int flag, int fam)
1850 {
1851 	struct ifaddr *ifa;
1852 
1853 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1854 
1855 	ifp->if_flags |= flag;
1856 	getmicrotime(&ifp->if_lastchange);
1857 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1858 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1859 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1860 	if (ifp->if_carp)
1861 		(*carp_linkstate_p)(ifp);
1862 	rt_ifmsg(ifp);
1863 #ifdef INET6
1864 	in6_if_up(ifp);
1865 #endif
1866 }
1867 
1868 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1869 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1870 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
1871 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
1872 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
1873 int	(*vlan_setcookie_p)(struct ifnet *, void *);
1874 void	*(*vlan_cookie_p)(struct ifnet *);
1875 
1876 /*
1877  * Handle a change in the interface link state. To avoid LORs
1878  * between driver lock and upper layer locks, as well as possible
1879  * recursions, we post event to taskqueue, and all job
1880  * is done in static do_link_state_change().
1881  */
1882 void
1883 if_link_state_change(struct ifnet *ifp, int link_state)
1884 {
1885 	/* Return if state hasn't changed. */
1886 	if (ifp->if_link_state == link_state)
1887 		return;
1888 
1889 	ifp->if_link_state = link_state;
1890 
1891 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1892 }
1893 
1894 static void
1895 do_link_state_change(void *arg, int pending)
1896 {
1897 	struct ifnet *ifp = (struct ifnet *)arg;
1898 	int link_state = ifp->if_link_state;
1899 	CURVNET_SET(ifp->if_vnet);
1900 
1901 	/* Notify that the link state has changed. */
1902 	rt_ifmsg(ifp);
1903 	if (ifp->if_vlantrunk != NULL)
1904 		(*vlan_link_state_p)(ifp);
1905 
1906 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1907 	    IFP2AC(ifp)->ac_netgraph != NULL)
1908 		(*ng_ether_link_state_p)(ifp, link_state);
1909 	if (ifp->if_carp)
1910 		(*carp_linkstate_p)(ifp);
1911 	if (ifp->if_bridge) {
1912 		KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1913 		(*bstp_linkstate_p)(ifp, link_state);
1914 	}
1915 	if (ifp->if_lagg) {
1916 		KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1917 		(*lagg_linkstate_p)(ifp, link_state);
1918 	}
1919 
1920 	if (IS_DEFAULT_VNET(curvnet))
1921 		devctl_notify("IFNET", ifp->if_xname,
1922 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1923 		    NULL);
1924 	if (pending > 1)
1925 		if_printf(ifp, "%d link states coalesced\n", pending);
1926 	if (log_link_state_change)
1927 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1928 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1929 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
1930 	CURVNET_RESTORE();
1931 }
1932 
1933 /*
1934  * Mark an interface down and notify protocols of
1935  * the transition.
1936  * NOTE: must be called at splnet or eqivalent.
1937  */
1938 void
1939 if_down(struct ifnet *ifp)
1940 {
1941 
1942 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1943 }
1944 
1945 /*
1946  * Mark an interface up and notify protocols of
1947  * the transition.
1948  * NOTE: must be called at splnet or eqivalent.
1949  */
1950 void
1951 if_up(struct ifnet *ifp)
1952 {
1953 
1954 	if_route(ifp, IFF_UP, AF_UNSPEC);
1955 }
1956 
1957 /*
1958  * Flush an interface queue.
1959  */
1960 void
1961 if_qflush(struct ifnet *ifp)
1962 {
1963 	struct mbuf *m, *n;
1964 	struct ifaltq *ifq;
1965 
1966 	ifq = &ifp->if_snd;
1967 	IFQ_LOCK(ifq);
1968 #ifdef ALTQ
1969 	if (ALTQ_IS_ENABLED(ifq))
1970 		ALTQ_PURGE(ifq);
1971 #endif
1972 	n = ifq->ifq_head;
1973 	while ((m = n) != 0) {
1974 		n = m->m_act;
1975 		m_freem(m);
1976 	}
1977 	ifq->ifq_head = 0;
1978 	ifq->ifq_tail = 0;
1979 	ifq->ifq_len = 0;
1980 	IFQ_UNLOCK(ifq);
1981 }
1982 
1983 /*
1984  * Map interface name to interface structure pointer, with or without
1985  * returning a reference.
1986  */
1987 struct ifnet *
1988 ifunit_ref(const char *name)
1989 {
1990 	struct ifnet *ifp;
1991 
1992 	IFNET_RLOCK_NOSLEEP();
1993 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1994 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
1995 		    !(ifp->if_flags & IFF_DYING))
1996 			break;
1997 	}
1998 	if (ifp != NULL)
1999 		if_ref(ifp);
2000 	IFNET_RUNLOCK_NOSLEEP();
2001 	return (ifp);
2002 }
2003 
2004 struct ifnet *
2005 ifunit(const char *name)
2006 {
2007 	struct ifnet *ifp;
2008 
2009 	IFNET_RLOCK_NOSLEEP();
2010 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2011 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2012 			break;
2013 	}
2014 	IFNET_RUNLOCK_NOSLEEP();
2015 	return (ifp);
2016 }
2017 
2018 /*
2019  * Hardware specific interface ioctls.
2020  */
2021 static int
2022 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2023 {
2024 	struct ifreq *ifr;
2025 	struct ifstat *ifs;
2026 	int error = 0;
2027 	int new_flags, temp_flags;
2028 	size_t namelen, onamelen;
2029 	size_t descrlen;
2030 	char *descrbuf, *odescrbuf;
2031 	char new_name[IFNAMSIZ];
2032 	struct ifaddr *ifa;
2033 	struct sockaddr_dl *sdl;
2034 
2035 	ifr = (struct ifreq *)data;
2036 	switch (cmd) {
2037 	case SIOCGIFINDEX:
2038 		ifr->ifr_index = ifp->if_index;
2039 		break;
2040 
2041 	case SIOCGIFFLAGS:
2042 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2043 		ifr->ifr_flags = temp_flags & 0xffff;
2044 		ifr->ifr_flagshigh = temp_flags >> 16;
2045 		break;
2046 
2047 	case SIOCGIFCAP:
2048 		ifr->ifr_reqcap = ifp->if_capabilities;
2049 		ifr->ifr_curcap = ifp->if_capenable;
2050 		break;
2051 
2052 #ifdef MAC
2053 	case SIOCGIFMAC:
2054 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2055 		break;
2056 #endif
2057 
2058 	case SIOCGIFMETRIC:
2059 		ifr->ifr_metric = ifp->if_metric;
2060 		break;
2061 
2062 	case SIOCGIFMTU:
2063 		ifr->ifr_mtu = ifp->if_mtu;
2064 		break;
2065 
2066 	case SIOCGIFPHYS:
2067 		ifr->ifr_phys = ifp->if_physical;
2068 		break;
2069 
2070 	case SIOCGIFDESCR:
2071 		error = 0;
2072 		sx_slock(&ifdescr_sx);
2073 		if (ifp->if_description == NULL)
2074 			error = ENOMSG;
2075 		else {
2076 			/* space for terminating nul */
2077 			descrlen = strlen(ifp->if_description) + 1;
2078 			if (ifr->ifr_buffer.length < descrlen)
2079 				ifr->ifr_buffer.buffer = NULL;
2080 			else
2081 				error = copyout(ifp->if_description,
2082 				    ifr->ifr_buffer.buffer, descrlen);
2083 			ifr->ifr_buffer.length = descrlen;
2084 		}
2085 		sx_sunlock(&ifdescr_sx);
2086 		break;
2087 
2088 	case SIOCSIFDESCR:
2089 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2090 		if (error)
2091 			return (error);
2092 
2093 		/*
2094 		 * Copy only (length-1) bytes to make sure that
2095 		 * if_description is always nul terminated.  The
2096 		 * length parameter is supposed to count the
2097 		 * terminating nul in.
2098 		 */
2099 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2100 			return (ENAMETOOLONG);
2101 		else if (ifr->ifr_buffer.length == 0)
2102 			descrbuf = NULL;
2103 		else {
2104 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2105 			    M_WAITOK | M_ZERO);
2106 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2107 			    ifr->ifr_buffer.length - 1);
2108 			if (error) {
2109 				free(descrbuf, M_IFDESCR);
2110 				break;
2111 			}
2112 		}
2113 
2114 		sx_xlock(&ifdescr_sx);
2115 		odescrbuf = ifp->if_description;
2116 		ifp->if_description = descrbuf;
2117 		sx_xunlock(&ifdescr_sx);
2118 
2119 		getmicrotime(&ifp->if_lastchange);
2120 		free(odescrbuf, M_IFDESCR);
2121 		break;
2122 
2123 	case SIOCGIFFIB:
2124 		ifr->ifr_fib = ifp->if_fib;
2125 		break;
2126 
2127 	case SIOCSIFFIB:
2128 		error = priv_check(td, PRIV_NET_SETIFFIB);
2129 		if (error)
2130 			return (error);
2131 		if (ifr->ifr_fib >= rt_numfibs)
2132 			return (EINVAL);
2133 
2134 		ifp->if_fib = ifr->ifr_fib;
2135 		break;
2136 
2137 	case SIOCSIFFLAGS:
2138 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2139 		if (error)
2140 			return (error);
2141 		/*
2142 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2143 		 * check, so we don't need special handling here yet.
2144 		 */
2145 		new_flags = (ifr->ifr_flags & 0xffff) |
2146 		    (ifr->ifr_flagshigh << 16);
2147 		if (ifp->if_flags & IFF_SMART) {
2148 			/* Smart drivers twiddle their own routes */
2149 		} else if (ifp->if_flags & IFF_UP &&
2150 		    (new_flags & IFF_UP) == 0) {
2151 			int s = splimp();
2152 			if_down(ifp);
2153 			splx(s);
2154 		} else if (new_flags & IFF_UP &&
2155 		    (ifp->if_flags & IFF_UP) == 0) {
2156 			int s = splimp();
2157 			if_up(ifp);
2158 			splx(s);
2159 		}
2160 		/* See if permanently promiscuous mode bit is about to flip */
2161 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2162 			if (new_flags & IFF_PPROMISC)
2163 				ifp->if_flags |= IFF_PROMISC;
2164 			else if (ifp->if_pcount == 0)
2165 				ifp->if_flags &= ~IFF_PROMISC;
2166 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2167 			    ifp->if_xname,
2168 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2169 		}
2170 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2171 			(new_flags &~ IFF_CANTCHANGE);
2172 		if (ifp->if_ioctl) {
2173 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2174 		}
2175 		getmicrotime(&ifp->if_lastchange);
2176 		break;
2177 
2178 	case SIOCSIFCAP:
2179 		error = priv_check(td, PRIV_NET_SETIFCAP);
2180 		if (error)
2181 			return (error);
2182 		if (ifp->if_ioctl == NULL)
2183 			return (EOPNOTSUPP);
2184 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2185 			return (EINVAL);
2186 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2187 		if (error == 0)
2188 			getmicrotime(&ifp->if_lastchange);
2189 		break;
2190 
2191 #ifdef MAC
2192 	case SIOCSIFMAC:
2193 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2194 		break;
2195 #endif
2196 
2197 	case SIOCSIFNAME:
2198 		error = priv_check(td, PRIV_NET_SETIFNAME);
2199 		if (error)
2200 			return (error);
2201 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2202 		if (error != 0)
2203 			return (error);
2204 		if (new_name[0] == '\0')
2205 			return (EINVAL);
2206 		if (ifunit(new_name) != NULL)
2207 			return (EEXIST);
2208 
2209 		/*
2210 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2211 		 * and there are numerous other races with the
2212 		 * ifunit() checks not being atomic with namespace
2213 		 * changes (renames, vmoves, if_attach, etc).
2214 		 */
2215 		ifp->if_flags |= IFF_RENAMING;
2216 
2217 		/* Announce the departure of the interface. */
2218 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2219 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2220 
2221 		log(LOG_INFO, "%s: changing name to '%s'\n",
2222 		    ifp->if_xname, new_name);
2223 
2224 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2225 		ifa = ifp->if_addr;
2226 		IFA_LOCK(ifa);
2227 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2228 		namelen = strlen(new_name);
2229 		onamelen = sdl->sdl_nlen;
2230 		/*
2231 		 * Move the address if needed.  This is safe because we
2232 		 * allocate space for a name of length IFNAMSIZ when we
2233 		 * create this in if_attach().
2234 		 */
2235 		if (namelen != onamelen) {
2236 			bcopy(sdl->sdl_data + onamelen,
2237 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2238 		}
2239 		bcopy(new_name, sdl->sdl_data, namelen);
2240 		sdl->sdl_nlen = namelen;
2241 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2242 		bzero(sdl->sdl_data, onamelen);
2243 		while (namelen != 0)
2244 			sdl->sdl_data[--namelen] = 0xff;
2245 		IFA_UNLOCK(ifa);
2246 
2247 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2248 		/* Announce the return of the interface. */
2249 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2250 
2251 		ifp->if_flags &= ~IFF_RENAMING;
2252 		break;
2253 
2254 #ifdef VIMAGE
2255 	case SIOCSIFVNET:
2256 		error = priv_check(td, PRIV_NET_SETIFVNET);
2257 		if (error)
2258 			return (error);
2259 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2260 		break;
2261 #endif
2262 
2263 	case SIOCSIFMETRIC:
2264 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2265 		if (error)
2266 			return (error);
2267 		ifp->if_metric = ifr->ifr_metric;
2268 		getmicrotime(&ifp->if_lastchange);
2269 		break;
2270 
2271 	case SIOCSIFPHYS:
2272 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2273 		if (error)
2274 			return (error);
2275 		if (ifp->if_ioctl == NULL)
2276 			return (EOPNOTSUPP);
2277 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2278 		if (error == 0)
2279 			getmicrotime(&ifp->if_lastchange);
2280 		break;
2281 
2282 	case SIOCSIFMTU:
2283 	{
2284 		u_long oldmtu = ifp->if_mtu;
2285 
2286 		error = priv_check(td, PRIV_NET_SETIFMTU);
2287 		if (error)
2288 			return (error);
2289 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2290 			return (EINVAL);
2291 		if (ifp->if_ioctl == NULL)
2292 			return (EOPNOTSUPP);
2293 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2294 		if (error == 0) {
2295 			getmicrotime(&ifp->if_lastchange);
2296 			rt_ifmsg(ifp);
2297 		}
2298 		/*
2299 		 * If the link MTU changed, do network layer specific procedure.
2300 		 */
2301 		if (ifp->if_mtu != oldmtu) {
2302 #ifdef INET6
2303 			nd6_setmtu(ifp);
2304 #endif
2305 		}
2306 		break;
2307 	}
2308 
2309 	case SIOCADDMULTI:
2310 	case SIOCDELMULTI:
2311 		if (cmd == SIOCADDMULTI)
2312 			error = priv_check(td, PRIV_NET_ADDMULTI);
2313 		else
2314 			error = priv_check(td, PRIV_NET_DELMULTI);
2315 		if (error)
2316 			return (error);
2317 
2318 		/* Don't allow group membership on non-multicast interfaces. */
2319 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2320 			return (EOPNOTSUPP);
2321 
2322 		/* Don't let users screw up protocols' entries. */
2323 		if (ifr->ifr_addr.sa_family != AF_LINK)
2324 			return (EINVAL);
2325 
2326 		if (cmd == SIOCADDMULTI) {
2327 			struct ifmultiaddr *ifma;
2328 
2329 			/*
2330 			 * Userland is only permitted to join groups once
2331 			 * via the if_addmulti() KPI, because it cannot hold
2332 			 * struct ifmultiaddr * between calls. It may also
2333 			 * lose a race while we check if the membership
2334 			 * already exists.
2335 			 */
2336 			IF_ADDR_LOCK(ifp);
2337 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2338 			IF_ADDR_UNLOCK(ifp);
2339 			if (ifma != NULL)
2340 				error = EADDRINUSE;
2341 			else
2342 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2343 		} else {
2344 			error = if_delmulti(ifp, &ifr->ifr_addr);
2345 		}
2346 		if (error == 0)
2347 			getmicrotime(&ifp->if_lastchange);
2348 		break;
2349 
2350 	case SIOCSIFPHYADDR:
2351 	case SIOCDIFPHYADDR:
2352 #ifdef INET6
2353 	case SIOCSIFPHYADDR_IN6:
2354 #endif
2355 	case SIOCSLIFPHYADDR:
2356 	case SIOCSIFMEDIA:
2357 	case SIOCSIFGENERIC:
2358 		error = priv_check(td, PRIV_NET_HWIOCTL);
2359 		if (error)
2360 			return (error);
2361 		if (ifp->if_ioctl == NULL)
2362 			return (EOPNOTSUPP);
2363 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2364 		if (error == 0)
2365 			getmicrotime(&ifp->if_lastchange);
2366 		break;
2367 
2368 	case SIOCGIFSTATUS:
2369 		ifs = (struct ifstat *)data;
2370 		ifs->ascii[0] = '\0';
2371 
2372 	case SIOCGIFPSRCADDR:
2373 	case SIOCGIFPDSTADDR:
2374 	case SIOCGLIFPHYADDR:
2375 	case SIOCGIFMEDIA:
2376 	case SIOCGIFGENERIC:
2377 		if (ifp->if_ioctl == NULL)
2378 			return (EOPNOTSUPP);
2379 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2380 		break;
2381 
2382 	case SIOCSIFLLADDR:
2383 		error = priv_check(td, PRIV_NET_SETLLADDR);
2384 		if (error)
2385 			return (error);
2386 		error = if_setlladdr(ifp,
2387 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2388 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2389 		break;
2390 
2391 	case SIOCAIFGROUP:
2392 	{
2393 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2394 
2395 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2396 		if (error)
2397 			return (error);
2398 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2399 			return (error);
2400 		break;
2401 	}
2402 
2403 	case SIOCGIFGROUP:
2404 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2405 			return (error);
2406 		break;
2407 
2408 	case SIOCDIFGROUP:
2409 	{
2410 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2411 
2412 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2413 		if (error)
2414 			return (error);
2415 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2416 			return (error);
2417 		break;
2418 	}
2419 
2420 	default:
2421 		error = ENOIOCTL;
2422 		break;
2423 	}
2424 	return (error);
2425 }
2426 
2427 #ifdef COMPAT_FREEBSD32
2428 struct ifconf32 {
2429 	int32_t	ifc_len;
2430 	union {
2431 		uint32_t	ifcu_buf;
2432 		uint32_t	ifcu_req;
2433 	} ifc_ifcu;
2434 };
2435 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2436 #endif
2437 
2438 /*
2439  * Interface ioctls.
2440  */
2441 int
2442 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2443 {
2444 	struct ifnet *ifp;
2445 	struct ifreq *ifr;
2446 	int error;
2447 	int oif_flags;
2448 
2449 	CURVNET_SET(so->so_vnet);
2450 	switch (cmd) {
2451 	case SIOCGIFCONF:
2452 	case OSIOCGIFCONF:
2453 		error = ifconf(cmd, data);
2454 		CURVNET_RESTORE();
2455 		return (error);
2456 
2457 #ifdef COMPAT_FREEBSD32
2458 	case SIOCGIFCONF32:
2459 		{
2460 			struct ifconf32 *ifc32;
2461 			struct ifconf ifc;
2462 
2463 			ifc32 = (struct ifconf32 *)data;
2464 			ifc.ifc_len = ifc32->ifc_len;
2465 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2466 
2467 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2468 			CURVNET_RESTORE();
2469 			if (error == 0)
2470 				ifc32->ifc_len = ifc.ifc_len;
2471 			return (error);
2472 		}
2473 #endif
2474 	}
2475 	ifr = (struct ifreq *)data;
2476 
2477 	switch (cmd) {
2478 #ifdef VIMAGE
2479 	case SIOCSIFRVNET:
2480 		error = priv_check(td, PRIV_NET_SETIFVNET);
2481 		if (error == 0)
2482 			error = if_vmove_reclaim(td, ifr->ifr_name,
2483 			    ifr->ifr_jid);
2484 		CURVNET_RESTORE();
2485 		return (error);
2486 #endif
2487 	case SIOCIFCREATE:
2488 	case SIOCIFCREATE2:
2489 		error = priv_check(td, PRIV_NET_IFCREATE);
2490 		if (error == 0)
2491 			error = if_clone_create(ifr->ifr_name,
2492 			    sizeof(ifr->ifr_name),
2493 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2494 		CURVNET_RESTORE();
2495 		return (error);
2496 	case SIOCIFDESTROY:
2497 		error = priv_check(td, PRIV_NET_IFDESTROY);
2498 		if (error == 0)
2499 			error = if_clone_destroy(ifr->ifr_name);
2500 		CURVNET_RESTORE();
2501 		return (error);
2502 
2503 	case SIOCIFGCLONERS:
2504 		error = if_clone_list((struct if_clonereq *)data);
2505 		CURVNET_RESTORE();
2506 		return (error);
2507 	case SIOCGIFGMEMB:
2508 		error = if_getgroupmembers((struct ifgroupreq *)data);
2509 		CURVNET_RESTORE();
2510 		return (error);
2511 #if defined(INET) || defined(INET6)
2512 	case SIOCSVH:
2513 	case SIOCGVH:
2514 		if (carp_ioctl_p == NULL)
2515 			error = EPROTONOSUPPORT;
2516 		else
2517 			error = (*carp_ioctl_p)(ifr, cmd, td);
2518 		CURVNET_RESTORE();
2519 		return (error);
2520 #endif
2521 	}
2522 
2523 	ifp = ifunit_ref(ifr->ifr_name);
2524 	if (ifp == NULL) {
2525 		CURVNET_RESTORE();
2526 		return (ENXIO);
2527 	}
2528 
2529 	error = ifhwioctl(cmd, ifp, data, td);
2530 	if (error != ENOIOCTL) {
2531 		if_rele(ifp);
2532 		CURVNET_RESTORE();
2533 		return (error);
2534 	}
2535 
2536 	oif_flags = ifp->if_flags;
2537 	if (so->so_proto == NULL) {
2538 		if_rele(ifp);
2539 		CURVNET_RESTORE();
2540 		return (EOPNOTSUPP);
2541 	}
2542 #ifndef COMPAT_43
2543 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2544 								 data,
2545 								 ifp, td));
2546 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2547 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2548 #else
2549 	{
2550 		u_long ocmd = cmd;
2551 
2552 		switch (cmd) {
2553 
2554 		case SIOCSIFDSTADDR:
2555 		case SIOCSIFADDR:
2556 		case SIOCSIFBRDADDR:
2557 		case SIOCSIFNETMASK:
2558 #if BYTE_ORDER != BIG_ENDIAN
2559 			if (ifr->ifr_addr.sa_family == 0 &&
2560 			    ifr->ifr_addr.sa_len < 16) {
2561 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2562 				ifr->ifr_addr.sa_len = 16;
2563 			}
2564 #else
2565 			if (ifr->ifr_addr.sa_len == 0)
2566 				ifr->ifr_addr.sa_len = 16;
2567 #endif
2568 			break;
2569 
2570 		case OSIOCGIFADDR:
2571 			cmd = SIOCGIFADDR;
2572 			break;
2573 
2574 		case OSIOCGIFDSTADDR:
2575 			cmd = SIOCGIFDSTADDR;
2576 			break;
2577 
2578 		case OSIOCGIFBRDADDR:
2579 			cmd = SIOCGIFBRDADDR;
2580 			break;
2581 
2582 		case OSIOCGIFNETMASK:
2583 			cmd = SIOCGIFNETMASK;
2584 		}
2585 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
2586 								   cmd,
2587 								   data,
2588 								   ifp, td));
2589 		if (error == EOPNOTSUPP && ifp != NULL &&
2590 		    ifp->if_ioctl != NULL)
2591 			error = (*ifp->if_ioctl)(ifp, cmd, data);
2592 		switch (ocmd) {
2593 
2594 		case OSIOCGIFADDR:
2595 		case OSIOCGIFDSTADDR:
2596 		case OSIOCGIFBRDADDR:
2597 		case OSIOCGIFNETMASK:
2598 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2599 
2600 		}
2601 	}
2602 #endif /* COMPAT_43 */
2603 
2604 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2605 #ifdef INET6
2606 		if (ifp->if_flags & IFF_UP) {
2607 			int s = splimp();
2608 			in6_if_up(ifp);
2609 			splx(s);
2610 		}
2611 #endif
2612 	}
2613 	if_rele(ifp);
2614 	CURVNET_RESTORE();
2615 	return (error);
2616 }
2617 
2618 /*
2619  * The code common to handling reference counted flags,
2620  * e.g., in ifpromisc() and if_allmulti().
2621  * The "pflag" argument can specify a permanent mode flag to check,
2622  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2623  *
2624  * Only to be used on stack-owned flags, not driver-owned flags.
2625  */
2626 static int
2627 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2628 {
2629 	struct ifreq ifr;
2630 	int error;
2631 	int oldflags, oldcount;
2632 
2633 	/* Sanity checks to catch programming errors */
2634 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2635 	    ("%s: setting driver-owned flag %d", __func__, flag));
2636 
2637 	if (onswitch)
2638 		KASSERT(*refcount >= 0,
2639 		    ("%s: increment negative refcount %d for flag %d",
2640 		    __func__, *refcount, flag));
2641 	else
2642 		KASSERT(*refcount > 0,
2643 		    ("%s: decrement non-positive refcount %d for flag %d",
2644 		    __func__, *refcount, flag));
2645 
2646 	/* In case this mode is permanent, just touch refcount */
2647 	if (ifp->if_flags & pflag) {
2648 		*refcount += onswitch ? 1 : -1;
2649 		return (0);
2650 	}
2651 
2652 	/* Save ifnet parameters for if_ioctl() may fail */
2653 	oldcount = *refcount;
2654 	oldflags = ifp->if_flags;
2655 
2656 	/*
2657 	 * See if we aren't the only and touching refcount is enough.
2658 	 * Actually toggle interface flag if we are the first or last.
2659 	 */
2660 	if (onswitch) {
2661 		if ((*refcount)++)
2662 			return (0);
2663 		ifp->if_flags |= flag;
2664 	} else {
2665 		if (--(*refcount))
2666 			return (0);
2667 		ifp->if_flags &= ~flag;
2668 	}
2669 
2670 	/* Call down the driver since we've changed interface flags */
2671 	if (ifp->if_ioctl == NULL) {
2672 		error = EOPNOTSUPP;
2673 		goto recover;
2674 	}
2675 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2676 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2677 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2678 	if (error)
2679 		goto recover;
2680 	/* Notify userland that interface flags have changed */
2681 	rt_ifmsg(ifp);
2682 	return (0);
2683 
2684 recover:
2685 	/* Recover after driver error */
2686 	*refcount = oldcount;
2687 	ifp->if_flags = oldflags;
2688 	return (error);
2689 }
2690 
2691 /*
2692  * Set/clear promiscuous mode on interface ifp based on the truth value
2693  * of pswitch.  The calls are reference counted so that only the first
2694  * "on" request actually has an effect, as does the final "off" request.
2695  * Results are undefined if the "off" and "on" requests are not matched.
2696  */
2697 int
2698 ifpromisc(struct ifnet *ifp, int pswitch)
2699 {
2700 	int error;
2701 	int oldflags = ifp->if_flags;
2702 
2703 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2704 			   &ifp->if_pcount, pswitch);
2705 	/* If promiscuous mode status has changed, log a message */
2706 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2707 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2708 		    ifp->if_xname,
2709 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2710 	return (error);
2711 }
2712 
2713 /*
2714  * Return interface configuration
2715  * of system.  List may be used
2716  * in later ioctl's (above) to get
2717  * other information.
2718  */
2719 /*ARGSUSED*/
2720 static int
2721 ifconf(u_long cmd, caddr_t data)
2722 {
2723 	struct ifconf *ifc = (struct ifconf *)data;
2724 	struct ifnet *ifp;
2725 	struct ifaddr *ifa;
2726 	struct ifreq ifr;
2727 	struct sbuf *sb;
2728 	int error, full = 0, valid_len, max_len;
2729 
2730 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2731 	max_len = MAXPHYS - 1;
2732 
2733 	/* Prevent hostile input from being able to crash the system */
2734 	if (ifc->ifc_len <= 0)
2735 		return (EINVAL);
2736 
2737 again:
2738 	if (ifc->ifc_len <= max_len) {
2739 		max_len = ifc->ifc_len;
2740 		full = 1;
2741 	}
2742 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2743 	max_len = 0;
2744 	valid_len = 0;
2745 
2746 	IFNET_RLOCK();
2747 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2748 		int addrs;
2749 
2750 		/*
2751 		 * Zero the ifr_name buffer to make sure we don't
2752 		 * disclose the contents of the stack.
2753 		 */
2754 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2755 
2756 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2757 		    >= sizeof(ifr.ifr_name)) {
2758 			sbuf_delete(sb);
2759 			IFNET_RUNLOCK();
2760 			return (ENAMETOOLONG);
2761 		}
2762 
2763 		addrs = 0;
2764 		IF_ADDR_LOCK(ifp);
2765 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2766 			struct sockaddr *sa = ifa->ifa_addr;
2767 
2768 			if (prison_if(curthread->td_ucred, sa) != 0)
2769 				continue;
2770 			addrs++;
2771 #ifdef COMPAT_43
2772 			if (cmd == OSIOCGIFCONF) {
2773 				struct osockaddr *osa =
2774 					 (struct osockaddr *)&ifr.ifr_addr;
2775 				ifr.ifr_addr = *sa;
2776 				osa->sa_family = sa->sa_family;
2777 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2778 				max_len += sizeof(ifr);
2779 			} else
2780 #endif
2781 			if (sa->sa_len <= sizeof(*sa)) {
2782 				ifr.ifr_addr = *sa;
2783 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2784 				max_len += sizeof(ifr);
2785 			} else {
2786 				sbuf_bcat(sb, &ifr,
2787 				    offsetof(struct ifreq, ifr_addr));
2788 				max_len += offsetof(struct ifreq, ifr_addr);
2789 				sbuf_bcat(sb, sa, sa->sa_len);
2790 				max_len += sa->sa_len;
2791 			}
2792 
2793 			if (sbuf_error(sb) == 0)
2794 				valid_len = sbuf_len(sb);
2795 		}
2796 		IF_ADDR_UNLOCK(ifp);
2797 		if (addrs == 0) {
2798 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2799 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2800 			max_len += sizeof(ifr);
2801 
2802 			if (sbuf_error(sb) == 0)
2803 				valid_len = sbuf_len(sb);
2804 		}
2805 	}
2806 	IFNET_RUNLOCK();
2807 
2808 	/*
2809 	 * If we didn't allocate enough space (uncommon), try again.  If
2810 	 * we have already allocated as much space as we are allowed,
2811 	 * return what we've got.
2812 	 */
2813 	if (valid_len != max_len && !full) {
2814 		sbuf_delete(sb);
2815 		goto again;
2816 	}
2817 
2818 	ifc->ifc_len = valid_len;
2819 	sbuf_finish(sb);
2820 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2821 	sbuf_delete(sb);
2822 	return (error);
2823 }
2824 
2825 /*
2826  * Just like ifpromisc(), but for all-multicast-reception mode.
2827  */
2828 int
2829 if_allmulti(struct ifnet *ifp, int onswitch)
2830 {
2831 
2832 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2833 }
2834 
2835 struct ifmultiaddr *
2836 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2837 {
2838 	struct ifmultiaddr *ifma;
2839 
2840 	IF_ADDR_LOCK_ASSERT(ifp);
2841 
2842 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2843 		if (sa->sa_family == AF_LINK) {
2844 			if (sa_dl_equal(ifma->ifma_addr, sa))
2845 				break;
2846 		} else {
2847 			if (sa_equal(ifma->ifma_addr, sa))
2848 				break;
2849 		}
2850 	}
2851 
2852 	return ifma;
2853 }
2854 
2855 /*
2856  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2857  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2858  * the ifnet multicast address list here, so the caller must do that and
2859  * other setup work (such as notifying the device driver).  The reference
2860  * count is initialized to 1.
2861  */
2862 static struct ifmultiaddr *
2863 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2864     int mflags)
2865 {
2866 	struct ifmultiaddr *ifma;
2867 	struct sockaddr *dupsa;
2868 
2869 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2870 	    M_ZERO);
2871 	if (ifma == NULL)
2872 		return (NULL);
2873 
2874 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2875 	if (dupsa == NULL) {
2876 		free(ifma, M_IFMADDR);
2877 		return (NULL);
2878 	}
2879 	bcopy(sa, dupsa, sa->sa_len);
2880 	ifma->ifma_addr = dupsa;
2881 
2882 	ifma->ifma_ifp = ifp;
2883 	ifma->ifma_refcount = 1;
2884 	ifma->ifma_protospec = NULL;
2885 
2886 	if (llsa == NULL) {
2887 		ifma->ifma_lladdr = NULL;
2888 		return (ifma);
2889 	}
2890 
2891 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2892 	if (dupsa == NULL) {
2893 		free(ifma->ifma_addr, M_IFMADDR);
2894 		free(ifma, M_IFMADDR);
2895 		return (NULL);
2896 	}
2897 	bcopy(llsa, dupsa, llsa->sa_len);
2898 	ifma->ifma_lladdr = dupsa;
2899 
2900 	return (ifma);
2901 }
2902 
2903 /*
2904  * if_freemulti: free ifmultiaddr structure and possibly attached related
2905  * addresses.  The caller is responsible for implementing reference
2906  * counting, notifying the driver, handling routing messages, and releasing
2907  * any dependent link layer state.
2908  */
2909 static void
2910 if_freemulti(struct ifmultiaddr *ifma)
2911 {
2912 
2913 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2914 	    ifma->ifma_refcount));
2915 	KASSERT(ifma->ifma_protospec == NULL,
2916 	    ("if_freemulti: protospec not NULL"));
2917 
2918 	if (ifma->ifma_lladdr != NULL)
2919 		free(ifma->ifma_lladdr, M_IFMADDR);
2920 	free(ifma->ifma_addr, M_IFMADDR);
2921 	free(ifma, M_IFMADDR);
2922 }
2923 
2924 /*
2925  * Register an additional multicast address with a network interface.
2926  *
2927  * - If the address is already present, bump the reference count on the
2928  *   address and return.
2929  * - If the address is not link-layer, look up a link layer address.
2930  * - Allocate address structures for one or both addresses, and attach to the
2931  *   multicast address list on the interface.  If automatically adding a link
2932  *   layer address, the protocol address will own a reference to the link
2933  *   layer address, to be freed when it is freed.
2934  * - Notify the network device driver of an addition to the multicast address
2935  *   list.
2936  *
2937  * 'sa' points to caller-owned memory with the desired multicast address.
2938  *
2939  * 'retifma' will be used to return a pointer to the resulting multicast
2940  * address reference, if desired.
2941  */
2942 int
2943 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2944     struct ifmultiaddr **retifma)
2945 {
2946 	struct ifmultiaddr *ifma, *ll_ifma;
2947 	struct sockaddr *llsa;
2948 	int error;
2949 
2950 	/*
2951 	 * If the address is already present, return a new reference to it;
2952 	 * otherwise, allocate storage and set up a new address.
2953 	 */
2954 	IF_ADDR_LOCK(ifp);
2955 	ifma = if_findmulti(ifp, sa);
2956 	if (ifma != NULL) {
2957 		ifma->ifma_refcount++;
2958 		if (retifma != NULL)
2959 			*retifma = ifma;
2960 		IF_ADDR_UNLOCK(ifp);
2961 		return (0);
2962 	}
2963 
2964 	/*
2965 	 * The address isn't already present; resolve the protocol address
2966 	 * into a link layer address, and then look that up, bump its
2967 	 * refcount or allocate an ifma for that also.  If 'llsa' was
2968 	 * returned, we will need to free it later.
2969 	 */
2970 	llsa = NULL;
2971 	ll_ifma = NULL;
2972 	if (ifp->if_resolvemulti != NULL) {
2973 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2974 		if (error)
2975 			goto unlock_out;
2976 	}
2977 
2978 	/*
2979 	 * Allocate the new address.  Don't hook it up yet, as we may also
2980 	 * need to allocate a link layer multicast address.
2981 	 */
2982 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2983 	if (ifma == NULL) {
2984 		error = ENOMEM;
2985 		goto free_llsa_out;
2986 	}
2987 
2988 	/*
2989 	 * If a link layer address is found, we'll need to see if it's
2990 	 * already present in the address list, or allocate is as well.
2991 	 * When this block finishes, the link layer address will be on the
2992 	 * list.
2993 	 */
2994 	if (llsa != NULL) {
2995 		ll_ifma = if_findmulti(ifp, llsa);
2996 		if (ll_ifma == NULL) {
2997 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2998 			if (ll_ifma == NULL) {
2999 				--ifma->ifma_refcount;
3000 				if_freemulti(ifma);
3001 				error = ENOMEM;
3002 				goto free_llsa_out;
3003 			}
3004 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3005 			    ifma_link);
3006 		} else
3007 			ll_ifma->ifma_refcount++;
3008 		ifma->ifma_llifma = ll_ifma;
3009 	}
3010 
3011 	/*
3012 	 * We now have a new multicast address, ifma, and possibly a new or
3013 	 * referenced link layer address.  Add the primary address to the
3014 	 * ifnet address list.
3015 	 */
3016 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3017 
3018 	if (retifma != NULL)
3019 		*retifma = ifma;
3020 
3021 	/*
3022 	 * Must generate the message while holding the lock so that 'ifma'
3023 	 * pointer is still valid.
3024 	 */
3025 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3026 	IF_ADDR_UNLOCK(ifp);
3027 
3028 	/*
3029 	 * We are certain we have added something, so call down to the
3030 	 * interface to let them know about it.
3031 	 */
3032 	if (ifp->if_ioctl != NULL) {
3033 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3034 	}
3035 
3036 	if (llsa != NULL)
3037 		free(llsa, M_IFMADDR);
3038 
3039 	return (0);
3040 
3041 free_llsa_out:
3042 	if (llsa != NULL)
3043 		free(llsa, M_IFMADDR);
3044 
3045 unlock_out:
3046 	IF_ADDR_UNLOCK(ifp);
3047 	return (error);
3048 }
3049 
3050 /*
3051  * Delete a multicast group membership by network-layer group address.
3052  *
3053  * Returns ENOENT if the entry could not be found. If ifp no longer
3054  * exists, results are undefined. This entry point should only be used
3055  * from subsystems which do appropriate locking to hold ifp for the
3056  * duration of the call.
3057  * Network-layer protocol domains must use if_delmulti_ifma().
3058  */
3059 int
3060 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3061 {
3062 	struct ifmultiaddr *ifma;
3063 	int lastref;
3064 #ifdef INVARIANTS
3065 	struct ifnet *oifp;
3066 
3067 	IFNET_RLOCK_NOSLEEP();
3068 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3069 		if (ifp == oifp)
3070 			break;
3071 	if (ifp != oifp)
3072 		ifp = NULL;
3073 	IFNET_RUNLOCK_NOSLEEP();
3074 
3075 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3076 #endif
3077 	if (ifp == NULL)
3078 		return (ENOENT);
3079 
3080 	IF_ADDR_LOCK(ifp);
3081 	lastref = 0;
3082 	ifma = if_findmulti(ifp, sa);
3083 	if (ifma != NULL)
3084 		lastref = if_delmulti_locked(ifp, ifma, 0);
3085 	IF_ADDR_UNLOCK(ifp);
3086 
3087 	if (ifma == NULL)
3088 		return (ENOENT);
3089 
3090 	if (lastref && ifp->if_ioctl != NULL) {
3091 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3092 	}
3093 
3094 	return (0);
3095 }
3096 
3097 /*
3098  * Delete all multicast group membership for an interface.
3099  * Should be used to quickly flush all multicast filters.
3100  */
3101 void
3102 if_delallmulti(struct ifnet *ifp)
3103 {
3104 	struct ifmultiaddr *ifma;
3105 	struct ifmultiaddr *next;
3106 
3107 	IF_ADDR_LOCK(ifp);
3108 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3109 		if_delmulti_locked(ifp, ifma, 0);
3110 	IF_ADDR_UNLOCK(ifp);
3111 }
3112 
3113 /*
3114  * Delete a multicast group membership by group membership pointer.
3115  * Network-layer protocol domains must use this routine.
3116  *
3117  * It is safe to call this routine if the ifp disappeared.
3118  */
3119 void
3120 if_delmulti_ifma(struct ifmultiaddr *ifma)
3121 {
3122 	struct ifnet *ifp;
3123 	int lastref;
3124 
3125 	ifp = ifma->ifma_ifp;
3126 #ifdef DIAGNOSTIC
3127 	if (ifp == NULL) {
3128 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3129 	} else {
3130 		struct ifnet *oifp;
3131 
3132 		IFNET_RLOCK_NOSLEEP();
3133 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3134 			if (ifp == oifp)
3135 				break;
3136 		if (ifp != oifp) {
3137 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3138 			ifp = NULL;
3139 		}
3140 		IFNET_RUNLOCK_NOSLEEP();
3141 	}
3142 #endif
3143 	/*
3144 	 * If and only if the ifnet instance exists: Acquire the address lock.
3145 	 */
3146 	if (ifp != NULL)
3147 		IF_ADDR_LOCK(ifp);
3148 
3149 	lastref = if_delmulti_locked(ifp, ifma, 0);
3150 
3151 	if (ifp != NULL) {
3152 		/*
3153 		 * If and only if the ifnet instance exists:
3154 		 *  Release the address lock.
3155 		 *  If the group was left: update the hardware hash filter.
3156 		 */
3157 		IF_ADDR_UNLOCK(ifp);
3158 		if (lastref && ifp->if_ioctl != NULL) {
3159 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3160 		}
3161 	}
3162 }
3163 
3164 /*
3165  * Perform deletion of network-layer and/or link-layer multicast address.
3166  *
3167  * Return 0 if the reference count was decremented.
3168  * Return 1 if the final reference was released, indicating that the
3169  * hardware hash filter should be reprogrammed.
3170  */
3171 static int
3172 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3173 {
3174 	struct ifmultiaddr *ll_ifma;
3175 
3176 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3177 		KASSERT(ifma->ifma_ifp == ifp,
3178 		    ("%s: inconsistent ifp %p", __func__, ifp));
3179 		IF_ADDR_LOCK_ASSERT(ifp);
3180 	}
3181 
3182 	ifp = ifma->ifma_ifp;
3183 
3184 	/*
3185 	 * If the ifnet is detaching, null out references to ifnet,
3186 	 * so that upper protocol layers will notice, and not attempt
3187 	 * to obtain locks for an ifnet which no longer exists. The
3188 	 * routing socket announcement must happen before the ifnet
3189 	 * instance is detached from the system.
3190 	 */
3191 	if (detaching) {
3192 #ifdef DIAGNOSTIC
3193 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3194 #endif
3195 		/*
3196 		 * ifp may already be nulled out if we are being reentered
3197 		 * to delete the ll_ifma.
3198 		 */
3199 		if (ifp != NULL) {
3200 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3201 			ifma->ifma_ifp = NULL;
3202 		}
3203 	}
3204 
3205 	if (--ifma->ifma_refcount > 0)
3206 		return 0;
3207 
3208 	/*
3209 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3210 	 * have been associated with it. Release it first if so.
3211 	 */
3212 	ll_ifma = ifma->ifma_llifma;
3213 	if (ll_ifma != NULL) {
3214 		KASSERT(ifma->ifma_lladdr != NULL,
3215 		    ("%s: llifma w/o lladdr", __func__));
3216 		if (detaching)
3217 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3218 		if (--ll_ifma->ifma_refcount == 0) {
3219 			if (ifp != NULL) {
3220 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3221 				    ifma_link);
3222 			}
3223 			if_freemulti(ll_ifma);
3224 		}
3225 	}
3226 
3227 	if (ifp != NULL)
3228 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3229 
3230 	if_freemulti(ifma);
3231 
3232 	/*
3233 	 * The last reference to this instance of struct ifmultiaddr
3234 	 * was released; the hardware should be notified of this change.
3235 	 */
3236 	return 1;
3237 }
3238 
3239 /*
3240  * Set the link layer address on an interface.
3241  *
3242  * At this time we only support certain types of interfaces,
3243  * and we don't allow the length of the address to change.
3244  */
3245 int
3246 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3247 {
3248 	struct sockaddr_dl *sdl;
3249 	struct ifaddr *ifa;
3250 	struct ifreq ifr;
3251 
3252 	IF_ADDR_LOCK(ifp);
3253 	ifa = ifp->if_addr;
3254 	if (ifa == NULL) {
3255 		IF_ADDR_UNLOCK(ifp);
3256 		return (EINVAL);
3257 	}
3258 	ifa_ref(ifa);
3259 	IF_ADDR_UNLOCK(ifp);
3260 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3261 	if (sdl == NULL) {
3262 		ifa_free(ifa);
3263 		return (EINVAL);
3264 	}
3265 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3266 		ifa_free(ifa);
3267 		return (EINVAL);
3268 	}
3269 	switch (ifp->if_type) {
3270 	case IFT_ETHER:
3271 	case IFT_FDDI:
3272 	case IFT_XETHER:
3273 	case IFT_ISO88025:
3274 	case IFT_L2VLAN:
3275 	case IFT_BRIDGE:
3276 	case IFT_ARCNET:
3277 	case IFT_IEEE8023ADLAG:
3278 	case IFT_IEEE80211:
3279 		bcopy(lladdr, LLADDR(sdl), len);
3280 		ifa_free(ifa);
3281 		break;
3282 	default:
3283 		ifa_free(ifa);
3284 		return (ENODEV);
3285 	}
3286 
3287 	/*
3288 	 * If the interface is already up, we need
3289 	 * to re-init it in order to reprogram its
3290 	 * address filter.
3291 	 */
3292 	if ((ifp->if_flags & IFF_UP) != 0) {
3293 		if (ifp->if_ioctl) {
3294 			ifp->if_flags &= ~IFF_UP;
3295 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3296 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3297 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3298 			ifp->if_flags |= IFF_UP;
3299 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3300 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3301 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3302 		}
3303 #ifdef INET
3304 		/*
3305 		 * Also send gratuitous ARPs to notify other nodes about
3306 		 * the address change.
3307 		 */
3308 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3309 			if (ifa->ifa_addr->sa_family == AF_INET)
3310 				arp_ifinit(ifp, ifa);
3311 		}
3312 #endif
3313 	}
3314 	return (0);
3315 }
3316 
3317 /*
3318  * The name argument must be a pointer to storage which will last as
3319  * long as the interface does.  For physical devices, the result of
3320  * device_get_name(dev) is a good choice and for pseudo-devices a
3321  * static string works well.
3322  */
3323 void
3324 if_initname(struct ifnet *ifp, const char *name, int unit)
3325 {
3326 	ifp->if_dname = name;
3327 	ifp->if_dunit = unit;
3328 	if (unit != IF_DUNIT_NONE)
3329 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3330 	else
3331 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3332 }
3333 
3334 int
3335 if_printf(struct ifnet *ifp, const char * fmt, ...)
3336 {
3337 	va_list ap;
3338 	int retval;
3339 
3340 	retval = printf("%s: ", ifp->if_xname);
3341 	va_start(ap, fmt);
3342 	retval += vprintf(fmt, ap);
3343 	va_end(ap);
3344 	return (retval);
3345 }
3346 
3347 void
3348 if_start(struct ifnet *ifp)
3349 {
3350 
3351 	(*(ifp)->if_start)(ifp);
3352 }
3353 
3354 /*
3355  * Backwards compatibility interface for drivers
3356  * that have not implemented it
3357  */
3358 static int
3359 if_transmit(struct ifnet *ifp, struct mbuf *m)
3360 {
3361 	int error;
3362 
3363 	IFQ_HANDOFF(ifp, m, error);
3364 	return (error);
3365 }
3366 
3367 int
3368 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3369 {
3370 	int active = 0;
3371 
3372 	IF_LOCK(ifq);
3373 	if (_IF_QFULL(ifq)) {
3374 		_IF_DROP(ifq);
3375 		IF_UNLOCK(ifq);
3376 		m_freem(m);
3377 		return (0);
3378 	}
3379 	if (ifp != NULL) {
3380 		ifp->if_obytes += m->m_pkthdr.len + adjust;
3381 		if (m->m_flags & (M_BCAST|M_MCAST))
3382 			ifp->if_omcasts++;
3383 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3384 	}
3385 	_IF_ENQUEUE(ifq, m);
3386 	IF_UNLOCK(ifq);
3387 	if (ifp != NULL && !active)
3388 		(*(ifp)->if_start)(ifp);
3389 	return (1);
3390 }
3391 
3392 void
3393 if_register_com_alloc(u_char type,
3394     if_com_alloc_t *a, if_com_free_t *f)
3395 {
3396 
3397 	KASSERT(if_com_alloc[type] == NULL,
3398 	    ("if_register_com_alloc: %d already registered", type));
3399 	KASSERT(if_com_free[type] == NULL,
3400 	    ("if_register_com_alloc: %d free already registered", type));
3401 
3402 	if_com_alloc[type] = a;
3403 	if_com_free[type] = f;
3404 }
3405 
3406 void
3407 if_deregister_com_alloc(u_char type)
3408 {
3409 
3410 	KASSERT(if_com_alloc[type] != NULL,
3411 	    ("if_deregister_com_alloc: %d not registered", type));
3412 	KASSERT(if_com_free[type] != NULL,
3413 	    ("if_deregister_com_alloc: %d free not registered", type));
3414 	if_com_alloc[type] = NULL;
3415 	if_com_free[type] = NULL;
3416 }
3417