xref: /freebsd/sys/net/if.c (revision a0e793cbf1951d07fc47a0d9ea389d7dacba5213)
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/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/if.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif /* INET */
89 #ifdef INET6
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
92 #endif /* INET6 */
93 #endif /* INET || INET6 */
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
100 #endif
101 
102 struct ifindex_entry {
103 	struct  ifnet *ife_ifnet;
104 };
105 
106 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
107 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
108 
109 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
110     &ifqmaxlen, 0, "max send queue size");
111 
112 /* Log link state change events */
113 static int log_link_state_change = 1;
114 
115 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
116 	&log_link_state_change, 0,
117 	"log interface link state change events");
118 
119 /* Interface description */
120 static unsigned int ifdescr_maxlen = 1024;
121 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
122 	&ifdescr_maxlen, 0,
123 	"administrative maximum length for interface description");
124 
125 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
126 
127 /* global sx for non-critical path ifdescr */
128 static struct sx ifdescr_sx;
129 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
130 
131 void	(*bridge_linkstate_p)(struct ifnet *ifp);
132 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
133 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
134 /* These are external hooks for CARP. */
135 void	(*carp_linkstate_p)(struct ifnet *ifp);
136 void	(*carp_demote_adj_p)(int, char *);
137 int	(*carp_master_p)(struct ifaddr *);
138 #if defined(INET) || defined(INET6)
139 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
140 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
141     const struct sockaddr *sa);
142 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
143 int	(*carp_attach_p)(struct ifaddr *, int);
144 void	(*carp_detach_p)(struct ifaddr *);
145 #endif
146 #ifdef INET
147 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
148 #endif
149 #ifdef INET6
150 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
151 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
152     const struct in6_addr *taddr);
153 #endif
154 
155 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
156 
157 /*
158  * XXX: Style; these should be sorted alphabetically, and unprototyped
159  * static functions should be prototyped. Currently they are sorted by
160  * declaration order.
161  */
162 static void	if_attachdomain(void *);
163 static void	if_attachdomain1(struct ifnet *);
164 static int	ifconf(u_long, caddr_t);
165 static void	if_freemulti(struct ifmultiaddr *);
166 static void	if_init(void *);
167 static void	if_grow(void);
168 static void	if_route(struct ifnet *, int flag, int fam);
169 static int	if_setflag(struct ifnet *, int, int, int *, int);
170 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
171 static void	if_unroute(struct ifnet *, int flag, int fam);
172 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
173 static int	if_rtdel(struct radix_node *, void *);
174 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
175 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
176 static void	do_link_state_change(void *, int);
177 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
178 static int	if_getgroupmembers(struct ifgroupreq *);
179 static void	if_delgroups(struct ifnet *);
180 static void	if_attach_internal(struct ifnet *, int);
181 static void	if_detach_internal(struct ifnet *, int);
182 
183 #ifdef INET6
184 /*
185  * XXX: declare here to avoid to include many inet6 related files..
186  * should be more generalized?
187  */
188 extern void	nd6_setmtu(struct ifnet *);
189 #endif
190 
191 VNET_DEFINE(int, if_index);
192 int	ifqmaxlen = IFQ_MAXLEN;
193 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
194 VNET_DEFINE(struct ifgrouphead, ifg_head);
195 
196 static VNET_DEFINE(int, if_indexlim) = 8;
197 
198 /* Table of ifnet by index. */
199 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
200 
201 #define	V_if_indexlim		VNET(if_indexlim)
202 #define	V_ifindex_table		VNET(ifindex_table)
203 
204 /*
205  * The global network interface list (V_ifnet) and related state (such as
206  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
207  * an rwlock.  Either may be acquired shared to stablize the list, but both
208  * must be acquired writable to modify the list.  This model allows us to
209  * both stablize the interface list during interrupt thread processing, but
210  * also to stablize it over long-running ioctls, without introducing priority
211  * inversions and deadlocks.
212  */
213 struct rwlock ifnet_rwlock;
214 struct sx ifnet_sxlock;
215 
216 /*
217  * The allocation of network interfaces is a rather non-atomic affair; we
218  * need to select an index before we are ready to expose the interface for
219  * use, so will use this pointer value to indicate reservation.
220  */
221 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
222 
223 static	if_com_alloc_t *if_com_alloc[256];
224 static	if_com_free_t *if_com_free[256];
225 
226 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
227 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
228 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
229 
230 struct ifnet *
231 ifnet_byindex_locked(u_short idx)
232 {
233 
234 	if (idx > V_if_index)
235 		return (NULL);
236 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
237 		return (NULL);
238 	return (V_ifindex_table[idx].ife_ifnet);
239 }
240 
241 struct ifnet *
242 ifnet_byindex(u_short idx)
243 {
244 	struct ifnet *ifp;
245 
246 	IFNET_RLOCK_NOSLEEP();
247 	ifp = ifnet_byindex_locked(idx);
248 	IFNET_RUNLOCK_NOSLEEP();
249 	return (ifp);
250 }
251 
252 struct ifnet *
253 ifnet_byindex_ref(u_short idx)
254 {
255 	struct ifnet *ifp;
256 
257 	IFNET_RLOCK_NOSLEEP();
258 	ifp = ifnet_byindex_locked(idx);
259 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
260 		IFNET_RUNLOCK_NOSLEEP();
261 		return (NULL);
262 	}
263 	if_ref(ifp);
264 	IFNET_RUNLOCK_NOSLEEP();
265 	return (ifp);
266 }
267 
268 /*
269  * Allocate an ifindex array entry; return 0 on success or an error on
270  * failure.
271  */
272 static int
273 ifindex_alloc_locked(u_short *idxp)
274 {
275 	u_short idx;
276 
277 	IFNET_WLOCK_ASSERT();
278 
279 retry:
280 	/*
281 	 * Try to find an empty slot below V_if_index.  If we fail, take the
282 	 * next slot.
283 	 */
284 	for (idx = 1; idx <= V_if_index; idx++) {
285 		if (V_ifindex_table[idx].ife_ifnet == NULL)
286 			break;
287 	}
288 
289 	/* Catch if_index overflow. */
290 	if (idx >= V_if_indexlim) {
291 		if_grow();
292 		goto retry;
293 	}
294 	if (idx > V_if_index)
295 		V_if_index = idx;
296 	*idxp = idx;
297 	return (0);
298 }
299 
300 static void
301 ifindex_free_locked(u_short idx)
302 {
303 
304 	IFNET_WLOCK_ASSERT();
305 
306 	V_ifindex_table[idx].ife_ifnet = NULL;
307 	while (V_if_index > 0 &&
308 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
309 		V_if_index--;
310 }
311 
312 static void
313 ifindex_free(u_short idx)
314 {
315 
316 	IFNET_WLOCK();
317 	ifindex_free_locked(idx);
318 	IFNET_WUNLOCK();
319 }
320 
321 static void
322 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
323 {
324 
325 	IFNET_WLOCK_ASSERT();
326 
327 	V_ifindex_table[idx].ife_ifnet = ifp;
328 }
329 
330 static void
331 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
332 {
333 
334 	IFNET_WLOCK();
335 	ifnet_setbyindex_locked(idx, ifp);
336 	IFNET_WUNLOCK();
337 }
338 
339 struct ifaddr *
340 ifaddr_byindex(u_short idx)
341 {
342 	struct ifaddr *ifa;
343 
344 	IFNET_RLOCK_NOSLEEP();
345 	ifa = ifnet_byindex_locked(idx)->if_addr;
346 	if (ifa != NULL)
347 		ifa_ref(ifa);
348 	IFNET_RUNLOCK_NOSLEEP();
349 	return (ifa);
350 }
351 
352 /*
353  * Network interface utility routines.
354  *
355  * Routines with ifa_ifwith* names take sockaddr *'s as
356  * parameters.
357  */
358 
359 static void
360 vnet_if_init(const void *unused __unused)
361 {
362 
363 	TAILQ_INIT(&V_ifnet);
364 	TAILQ_INIT(&V_ifg_head);
365 	IFNET_WLOCK();
366 	if_grow();				/* create initial table */
367 	IFNET_WUNLOCK();
368 	vnet_if_clone_init();
369 }
370 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
371     NULL);
372 
373 /* ARGSUSED*/
374 static void
375 if_init(void *dummy __unused)
376 {
377 
378 	IFNET_LOCK_INIT();
379 	if_clone_init();
380 }
381 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
382 
383 
384 #ifdef VIMAGE
385 static void
386 vnet_if_uninit(const void *unused __unused)
387 {
388 
389 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
390 	    "not empty", __func__, __LINE__, &V_ifnet));
391 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
392 	    "not empty", __func__, __LINE__, &V_ifg_head));
393 
394 	free((caddr_t)V_ifindex_table, M_IFNET);
395 }
396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
397     vnet_if_uninit, NULL);
398 #endif
399 
400 static void
401 if_grow(void)
402 {
403 	int oldlim;
404 	u_int n;
405 	struct ifindex_entry *e;
406 
407 	IFNET_WLOCK_ASSERT();
408 	oldlim = V_if_indexlim;
409 	IFNET_WUNLOCK();
410 	n = (oldlim << 1) * sizeof(*e);
411 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
412 	IFNET_WLOCK();
413 	if (V_if_indexlim != oldlim) {
414 		free(e, M_IFNET);
415 		return;
416 	}
417 	if (V_ifindex_table != NULL) {
418 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
419 		free((caddr_t)V_ifindex_table, M_IFNET);
420 	}
421 	V_if_indexlim <<= 1;
422 	V_ifindex_table = e;
423 }
424 
425 /*
426  * Allocate a struct ifnet and an index for an interface.  A layer 2
427  * common structure will also be allocated if an allocation routine is
428  * registered for the passed type.
429  */
430 struct ifnet *
431 if_alloc(u_char type)
432 {
433 	struct ifnet *ifp;
434 	u_short idx;
435 
436 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
437 	IFNET_WLOCK();
438 	if (ifindex_alloc_locked(&idx) != 0) {
439 		IFNET_WUNLOCK();
440 		free(ifp, M_IFNET);
441 		return (NULL);
442 	}
443 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
444 	IFNET_WUNLOCK();
445 	ifp->if_index = idx;
446 	ifp->if_type = type;
447 	ifp->if_alloctype = type;
448 	if (if_com_alloc[type] != NULL) {
449 		ifp->if_l2com = if_com_alloc[type](type, ifp);
450 		if (ifp->if_l2com == NULL) {
451 			free(ifp, M_IFNET);
452 			ifindex_free(idx);
453 			return (NULL);
454 		}
455 	}
456 
457 	IF_ADDR_LOCK_INIT(ifp);
458 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
459 	ifp->if_afdata_initialized = 0;
460 	IF_AFDATA_LOCK_INIT(ifp);
461 	TAILQ_INIT(&ifp->if_addrhead);
462 	TAILQ_INIT(&ifp->if_multiaddrs);
463 	TAILQ_INIT(&ifp->if_groups);
464 #ifdef MAC
465 	mac_ifnet_init(ifp);
466 #endif
467 	ifq_init(&ifp->if_snd, ifp);
468 
469 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
470 	ifnet_setbyindex(ifp->if_index, ifp);
471 	return (ifp);
472 }
473 
474 /*
475  * Do the actual work of freeing a struct ifnet, and layer 2 common
476  * structure.  This call is made when the last reference to an
477  * interface is released.
478  */
479 static void
480 if_free_internal(struct ifnet *ifp)
481 {
482 
483 	KASSERT((ifp->if_flags & IFF_DYING),
484 	    ("if_free_internal: interface not dying"));
485 
486 	if (if_com_free[ifp->if_alloctype] != NULL)
487 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
488 		    ifp->if_alloctype);
489 
490 #ifdef MAC
491 	mac_ifnet_destroy(ifp);
492 #endif /* MAC */
493 	if (ifp->if_description != NULL)
494 		free(ifp->if_description, M_IFDESCR);
495 	IF_AFDATA_DESTROY(ifp);
496 	IF_ADDR_LOCK_DESTROY(ifp);
497 	ifq_delete(&ifp->if_snd);
498 	free(ifp, M_IFNET);
499 }
500 
501 /*
502  * Deregister an interface and free the associated storage.
503  */
504 void
505 if_free(struct ifnet *ifp)
506 {
507 
508 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
509 
510 	CURVNET_SET_QUIET(ifp->if_vnet);
511 	IFNET_WLOCK();
512 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
513 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
514 
515 	ifindex_free_locked(ifp->if_index);
516 	IFNET_WUNLOCK();
517 
518 	if (refcount_release(&ifp->if_refcount))
519 		if_free_internal(ifp);
520 	CURVNET_RESTORE();
521 }
522 
523 /*
524  * Interfaces to keep an ifnet type-stable despite the possibility of the
525  * driver calling if_free().  If there are additional references, we defer
526  * freeing the underlying data structure.
527  */
528 void
529 if_ref(struct ifnet *ifp)
530 {
531 
532 	/* We don't assert the ifnet list lock here, but arguably should. */
533 	refcount_acquire(&ifp->if_refcount);
534 }
535 
536 void
537 if_rele(struct ifnet *ifp)
538 {
539 
540 	if (!refcount_release(&ifp->if_refcount))
541 		return;
542 	if_free_internal(ifp);
543 }
544 
545 void
546 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
547 {
548 
549 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
550 
551 	if (ifq->ifq_maxlen == 0)
552 		ifq->ifq_maxlen = ifqmaxlen;
553 
554 	ifq->altq_type = 0;
555 	ifq->altq_disc = NULL;
556 	ifq->altq_flags &= ALTQF_CANTCHANGE;
557 	ifq->altq_tbr  = NULL;
558 	ifq->altq_ifp  = ifp;
559 }
560 
561 void
562 ifq_delete(struct ifaltq *ifq)
563 {
564 	mtx_destroy(&ifq->ifq_mtx);
565 }
566 
567 /*
568  * Perform generic interface initalization tasks and attach the interface
569  * to the list of "active" interfaces.  If vmove flag is set on entry
570  * to if_attach_internal(), perform only a limited subset of initialization
571  * tasks, given that we are moving from one vnet to another an ifnet which
572  * has already been fully initialized.
573  *
574  * XXX:
575  *  - The decision to return void and thus require this function to
576  *    succeed is questionable.
577  *  - We should probably do more sanity checking.  For instance we don't
578  *    do anything to insure if_xname is unique or non-empty.
579  */
580 void
581 if_attach(struct ifnet *ifp)
582 {
583 
584 	if_attach_internal(ifp, 0);
585 }
586 
587 static void
588 if_attach_internal(struct ifnet *ifp, int vmove)
589 {
590 	unsigned socksize, ifasize;
591 	int namelen, masklen;
592 	struct sockaddr_dl *sdl;
593 	struct ifaddr *ifa;
594 
595 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
596 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
597 		    ifp->if_xname);
598 
599 #ifdef VIMAGE
600 	ifp->if_vnet = curvnet;
601 	if (ifp->if_home_vnet == NULL)
602 		ifp->if_home_vnet = curvnet;
603 #endif
604 
605 	if_addgroup(ifp, IFG_ALL);
606 
607 	getmicrotime(&ifp->if_lastchange);
608 	ifp->if_data.ifi_epoch = time_uptime;
609 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
610 
611 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
612 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
613 	    ("transmit and qflush must both either be set or both be NULL"));
614 	if (ifp->if_transmit == NULL) {
615 		ifp->if_transmit = if_transmit;
616 		ifp->if_qflush = if_qflush;
617 	}
618 
619 	if (!vmove) {
620 #ifdef MAC
621 		mac_ifnet_create(ifp);
622 #endif
623 
624 		/*
625 		 * Create a Link Level name for this device.
626 		 */
627 		namelen = strlen(ifp->if_xname);
628 		/*
629 		 * Always save enough space for any possiable name so we
630 		 * can do a rename in place later.
631 		 */
632 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
633 		socksize = masklen + ifp->if_addrlen;
634 		if (socksize < sizeof(*sdl))
635 			socksize = sizeof(*sdl);
636 		socksize = roundup2(socksize, sizeof(long));
637 		ifasize = sizeof(*ifa) + 2 * socksize;
638 		ifa = ifa_alloc(ifasize, M_WAITOK);
639 		sdl = (struct sockaddr_dl *)(ifa + 1);
640 		sdl->sdl_len = socksize;
641 		sdl->sdl_family = AF_LINK;
642 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
643 		sdl->sdl_nlen = namelen;
644 		sdl->sdl_index = ifp->if_index;
645 		sdl->sdl_type = ifp->if_type;
646 		ifp->if_addr = ifa;
647 		ifa->ifa_ifp = ifp;
648 		ifa->ifa_rtrequest = link_rtrequest;
649 		ifa->ifa_addr = (struct sockaddr *)sdl;
650 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
651 		ifa->ifa_netmask = (struct sockaddr *)sdl;
652 		sdl->sdl_len = masklen;
653 		while (namelen != 0)
654 			sdl->sdl_data[--namelen] = 0xff;
655 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
656 		/* Reliably crash if used uninitialized. */
657 		ifp->if_broadcastaddr = NULL;
658 
659 #if defined(INET) || defined(INET6)
660 		/* Initialize to max value. */
661 		if (ifp->if_hw_tsomax == 0)
662 			ifp->if_hw_tsomax = min(IP_MAXPACKET, 32 * MCLBYTES -
663 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
664 		KASSERT(ifp->if_hw_tsomax <= IP_MAXPACKET &&
665 		    ifp->if_hw_tsomax >= IP_MAXPACKET / 8,
666 		    ("%s: tsomax outside of range", __func__));
667 #endif
668 	}
669 #ifdef VIMAGE
670 	else {
671 		/*
672 		 * Update the interface index in the link layer address
673 		 * of the interface.
674 		 */
675 		for (ifa = ifp->if_addr; ifa != NULL;
676 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
677 			if (ifa->ifa_addr->sa_family == AF_LINK) {
678 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
679 				sdl->sdl_index = ifp->if_index;
680 			}
681 		}
682 	}
683 #endif
684 
685 	IFNET_WLOCK();
686 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
687 #ifdef VIMAGE
688 	curvnet->vnet_ifcnt++;
689 #endif
690 	IFNET_WUNLOCK();
691 
692 	if (domain_init_status >= 2)
693 		if_attachdomain1(ifp);
694 
695 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
696 	if (IS_DEFAULT_VNET(curvnet))
697 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
698 
699 	/* Announce the interface. */
700 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
701 }
702 
703 static void
704 if_attachdomain(void *dummy)
705 {
706 	struct ifnet *ifp;
707 
708 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
709 		if_attachdomain1(ifp);
710 }
711 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
712     if_attachdomain, NULL);
713 
714 static void
715 if_attachdomain1(struct ifnet *ifp)
716 {
717 	struct domain *dp;
718 
719 	/*
720 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
721 	 * cannot lock ifp->if_afdata initialization, entirely.
722 	 */
723 	if (IF_AFDATA_TRYLOCK(ifp) == 0)
724 		return;
725 	if (ifp->if_afdata_initialized >= domain_init_status) {
726 		IF_AFDATA_UNLOCK(ifp);
727 		log(LOG_WARNING, "%s called more than once on %s\n",
728 		    __func__, ifp->if_xname);
729 		return;
730 	}
731 	ifp->if_afdata_initialized = domain_init_status;
732 	IF_AFDATA_UNLOCK(ifp);
733 
734 	/* address family dependent data region */
735 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
736 	for (dp = domains; dp; dp = dp->dom_next) {
737 		if (dp->dom_ifattach)
738 			ifp->if_afdata[dp->dom_family] =
739 			    (*dp->dom_ifattach)(ifp);
740 	}
741 }
742 
743 /*
744  * Remove any unicast or broadcast network addresses from an interface.
745  */
746 void
747 if_purgeaddrs(struct ifnet *ifp)
748 {
749 	struct ifaddr *ifa, *next;
750 
751 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
752 		if (ifa->ifa_addr->sa_family == AF_LINK)
753 			continue;
754 #ifdef INET
755 		/* XXX: Ugly!! ad hoc just for INET */
756 		if (ifa->ifa_addr->sa_family == AF_INET) {
757 			struct ifaliasreq ifr;
758 
759 			bzero(&ifr, sizeof(ifr));
760 			ifr.ifra_addr = *ifa->ifa_addr;
761 			if (ifa->ifa_dstaddr)
762 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
763 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
764 			    NULL) == 0)
765 				continue;
766 		}
767 #endif /* INET */
768 #ifdef INET6
769 		if (ifa->ifa_addr->sa_family == AF_INET6) {
770 			in6_purgeaddr(ifa);
771 			/* ifp_addrhead is already updated */
772 			continue;
773 		}
774 #endif /* INET6 */
775 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
776 		ifa_free(ifa);
777 	}
778 }
779 
780 /*
781  * Remove any multicast network addresses from an interface when an ifnet
782  * is going away.
783  */
784 static void
785 if_purgemaddrs(struct ifnet *ifp)
786 {
787 	struct ifmultiaddr *ifma;
788 	struct ifmultiaddr *next;
789 
790 	IF_ADDR_WLOCK(ifp);
791 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
792 		if_delmulti_locked(ifp, ifma, 1);
793 	IF_ADDR_WUNLOCK(ifp);
794 }
795 
796 /*
797  * Detach an interface, removing it from the list of "active" interfaces.
798  * If vmove flag is set on entry to if_detach_internal(), perform only a
799  * limited subset of cleanup tasks, given that we are moving an ifnet from
800  * one vnet to another, where it must be fully operational.
801  *
802  * XXXRW: There are some significant questions about event ordering, and
803  * how to prevent things from starting to use the interface during detach.
804  */
805 void
806 if_detach(struct ifnet *ifp)
807 {
808 
809 	CURVNET_SET_QUIET(ifp->if_vnet);
810 	if_detach_internal(ifp, 0);
811 	CURVNET_RESTORE();
812 }
813 
814 static void
815 if_detach_internal(struct ifnet *ifp, int vmove)
816 {
817 	struct ifaddr *ifa;
818 	struct radix_node_head	*rnh;
819 	int i, j;
820 	struct domain *dp;
821  	struct ifnet *iter;
822  	int found = 0;
823 
824 	IFNET_WLOCK();
825 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
826 		if (iter == ifp) {
827 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
828 			found = 1;
829 			break;
830 		}
831 #ifdef VIMAGE
832 	if (found)
833 		curvnet->vnet_ifcnt--;
834 #endif
835 	IFNET_WUNLOCK();
836 	if (!found) {
837 		if (vmove)
838 			panic("%s: ifp=%p not on the ifnet tailq %p",
839 			    __func__, ifp, &V_ifnet);
840 		else
841 			return; /* XXX this should panic as well? */
842 	}
843 
844 	/*
845 	 * Remove/wait for pending events.
846 	 */
847 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
848 
849 	/*
850 	 * Remove routes and flush queues.
851 	 */
852 	if_down(ifp);
853 #ifdef ALTQ
854 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
855 		altq_disable(&ifp->if_snd);
856 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
857 		altq_detach(&ifp->if_snd);
858 #endif
859 
860 	if_purgeaddrs(ifp);
861 
862 #ifdef INET
863 	in_ifdetach(ifp);
864 #endif
865 
866 #ifdef INET6
867 	/*
868 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
869 	 * before removing routing entries below, since IPv6 interface direct
870 	 * routes are expected to be removed by the IPv6-specific kernel API.
871 	 * Otherwise, the kernel will detect some inconsistency and bark it.
872 	 */
873 	in6_ifdetach(ifp);
874 #endif
875 	if_purgemaddrs(ifp);
876 
877 	if (!vmove) {
878 		/*
879 		 * Prevent further calls into the device driver via ifnet.
880 		 */
881 		if_dead(ifp);
882 
883 		/*
884 		 * Remove link ifaddr pointer and maybe decrement if_index.
885 		 * Clean up all addresses.
886 		 */
887 		ifp->if_addr = NULL;
888 
889 		/* We can now free link ifaddr. */
890 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
891 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
892 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
893 			ifa_free(ifa);
894 		}
895 	}
896 
897 	/*
898 	 * Delete all remaining routes using this interface
899 	 * Unfortuneatly the only way to do this is to slog through
900 	 * the entire routing table looking for routes which point
901 	 * to this interface...oh well...
902 	 */
903 	for (i = 1; i <= AF_MAX; i++) {
904 		for (j = 0; j < rt_numfibs; j++) {
905 			rnh = rt_tables_get_rnh(j, i);
906 			if (rnh == NULL)
907 				continue;
908 			RADIX_NODE_HEAD_LOCK(rnh);
909 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
910 			RADIX_NODE_HEAD_UNLOCK(rnh);
911 		}
912 	}
913 
914 	/* Announce that the interface is gone. */
915 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
916 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
917 	if (IS_DEFAULT_VNET(curvnet))
918 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
919 	if_delgroups(ifp);
920 
921 	/*
922 	 * We cannot hold the lock over dom_ifdetach calls as they might
923 	 * sleep, for example trying to drain a callout, thus open up the
924 	 * theoretical race with re-attaching.
925 	 */
926 	IF_AFDATA_LOCK(ifp);
927 	i = ifp->if_afdata_initialized;
928 	ifp->if_afdata_initialized = 0;
929 	IF_AFDATA_UNLOCK(ifp);
930 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
931 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
932 			(*dp->dom_ifdetach)(ifp,
933 			    ifp->if_afdata[dp->dom_family]);
934 	}
935 }
936 
937 #ifdef VIMAGE
938 /*
939  * if_vmove() performs a limited version of if_detach() in current
940  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
941  * An attempt is made to shrink if_index in current vnet, find an
942  * unused if_index in target vnet and calls if_grow() if necessary,
943  * and finally find an unused if_xname for the target vnet.
944  */
945 void
946 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
947 {
948 	u_short idx;
949 
950 	/*
951 	 * Detach from current vnet, but preserve LLADDR info, do not
952 	 * mark as dead etc. so that the ifnet can be reattached later.
953 	 */
954 	if_detach_internal(ifp, 1);
955 
956 	/*
957 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
958 	 * the if_index for that vnet if possible.
959 	 *
960 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
961 	 * or we'd lock on one vnet and unlock on another.
962 	 */
963 	IFNET_WLOCK();
964 	ifindex_free_locked(ifp->if_index);
965 	IFNET_WUNLOCK();
966 
967 	/*
968 	 * Perform interface-specific reassignment tasks, if provided by
969 	 * the driver.
970 	 */
971 	if (ifp->if_reassign != NULL)
972 		ifp->if_reassign(ifp, new_vnet, NULL);
973 
974 	/*
975 	 * Switch to the context of the target vnet.
976 	 */
977 	CURVNET_SET_QUIET(new_vnet);
978 
979 	IFNET_WLOCK();
980 	if (ifindex_alloc_locked(&idx) != 0) {
981 		IFNET_WUNLOCK();
982 		panic("if_index overflow");
983 	}
984 	ifp->if_index = idx;
985 	ifnet_setbyindex_locked(ifp->if_index, ifp);
986 	IFNET_WUNLOCK();
987 
988 	if_attach_internal(ifp, 1);
989 
990 	CURVNET_RESTORE();
991 }
992 
993 /*
994  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
995  */
996 static int
997 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
998 {
999 	struct prison *pr;
1000 	struct ifnet *difp;
1001 
1002 	/* Try to find the prison within our visibility. */
1003 	sx_slock(&allprison_lock);
1004 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1005 	sx_sunlock(&allprison_lock);
1006 	if (pr == NULL)
1007 		return (ENXIO);
1008 	prison_hold_locked(pr);
1009 	mtx_unlock(&pr->pr_mtx);
1010 
1011 	/* Do not try to move the iface from and to the same prison. */
1012 	if (pr->pr_vnet == ifp->if_vnet) {
1013 		prison_free(pr);
1014 		return (EEXIST);
1015 	}
1016 
1017 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1018 	/* XXX Lock interfaces to avoid races. */
1019 	CURVNET_SET_QUIET(pr->pr_vnet);
1020 	difp = ifunit(ifname);
1021 	CURVNET_RESTORE();
1022 	if (difp != NULL) {
1023 		prison_free(pr);
1024 		return (EEXIST);
1025 	}
1026 
1027 	/* Move the interface into the child jail/vnet. */
1028 	if_vmove(ifp, pr->pr_vnet);
1029 
1030 	/* Report the new if_xname back to the userland. */
1031 	sprintf(ifname, "%s", ifp->if_xname);
1032 
1033 	prison_free(pr);
1034 	return (0);
1035 }
1036 
1037 static int
1038 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1039 {
1040 	struct prison *pr;
1041 	struct vnet *vnet_dst;
1042 	struct ifnet *ifp;
1043 
1044 	/* Try to find the prison within our visibility. */
1045 	sx_slock(&allprison_lock);
1046 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1047 	sx_sunlock(&allprison_lock);
1048 	if (pr == NULL)
1049 		return (ENXIO);
1050 	prison_hold_locked(pr);
1051 	mtx_unlock(&pr->pr_mtx);
1052 
1053 	/* Make sure the named iface exists in the source prison/vnet. */
1054 	CURVNET_SET(pr->pr_vnet);
1055 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1056 	if (ifp == NULL) {
1057 		CURVNET_RESTORE();
1058 		prison_free(pr);
1059 		return (ENXIO);
1060 	}
1061 
1062 	/* Do not try to move the iface from and to the same prison. */
1063 	vnet_dst = TD_TO_VNET(td);
1064 	if (vnet_dst == ifp->if_vnet) {
1065 		CURVNET_RESTORE();
1066 		prison_free(pr);
1067 		return (EEXIST);
1068 	}
1069 
1070 	/* Get interface back from child jail/vnet. */
1071 	if_vmove(ifp, vnet_dst);
1072 	CURVNET_RESTORE();
1073 
1074 	/* Report the new if_xname back to the userland. */
1075 	sprintf(ifname, "%s", ifp->if_xname);
1076 
1077 	prison_free(pr);
1078 	return (0);
1079 }
1080 #endif /* VIMAGE */
1081 
1082 /*
1083  * Add a group to an interface
1084  */
1085 int
1086 if_addgroup(struct ifnet *ifp, const char *groupname)
1087 {
1088 	struct ifg_list		*ifgl;
1089 	struct ifg_group	*ifg = NULL;
1090 	struct ifg_member	*ifgm;
1091 	int 			 new = 0;
1092 
1093 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1094 	    groupname[strlen(groupname) - 1] <= '9')
1095 		return (EINVAL);
1096 
1097 	IFNET_WLOCK();
1098 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1099 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1100 			IFNET_WUNLOCK();
1101 			return (EEXIST);
1102 		}
1103 
1104 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1105 	    M_NOWAIT)) == NULL) {
1106 	    	IFNET_WUNLOCK();
1107 		return (ENOMEM);
1108 	}
1109 
1110 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1111 	    M_TEMP, M_NOWAIT)) == NULL) {
1112 		free(ifgl, M_TEMP);
1113 		IFNET_WUNLOCK();
1114 		return (ENOMEM);
1115 	}
1116 
1117 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1118 		if (!strcmp(ifg->ifg_group, groupname))
1119 			break;
1120 
1121 	if (ifg == NULL) {
1122 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1123 		    M_TEMP, M_NOWAIT)) == NULL) {
1124 			free(ifgl, M_TEMP);
1125 			free(ifgm, M_TEMP);
1126 			IFNET_WUNLOCK();
1127 			return (ENOMEM);
1128 		}
1129 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1130 		ifg->ifg_refcnt = 0;
1131 		TAILQ_INIT(&ifg->ifg_members);
1132 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1133 		new = 1;
1134 	}
1135 
1136 	ifg->ifg_refcnt++;
1137 	ifgl->ifgl_group = ifg;
1138 	ifgm->ifgm_ifp = ifp;
1139 
1140 	IF_ADDR_WLOCK(ifp);
1141 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1142 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1143 	IF_ADDR_WUNLOCK(ifp);
1144 
1145 	IFNET_WUNLOCK();
1146 
1147 	if (new)
1148 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1149 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1150 
1151 	return (0);
1152 }
1153 
1154 /*
1155  * Remove a group from an interface
1156  */
1157 int
1158 if_delgroup(struct ifnet *ifp, const char *groupname)
1159 {
1160 	struct ifg_list		*ifgl;
1161 	struct ifg_member	*ifgm;
1162 
1163 	IFNET_WLOCK();
1164 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1165 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1166 			break;
1167 	if (ifgl == NULL) {
1168 		IFNET_WUNLOCK();
1169 		return (ENOENT);
1170 	}
1171 
1172 	IF_ADDR_WLOCK(ifp);
1173 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1174 	IF_ADDR_WUNLOCK(ifp);
1175 
1176 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1177 		if (ifgm->ifgm_ifp == ifp)
1178 			break;
1179 
1180 	if (ifgm != NULL) {
1181 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1182 		free(ifgm, M_TEMP);
1183 	}
1184 
1185 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1186 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1187 		IFNET_WUNLOCK();
1188 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1189 		free(ifgl->ifgl_group, M_TEMP);
1190 	} else
1191 		IFNET_WUNLOCK();
1192 
1193 	free(ifgl, M_TEMP);
1194 
1195 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1196 
1197 	return (0);
1198 }
1199 
1200 /*
1201  * Remove an interface from all groups
1202  */
1203 static void
1204 if_delgroups(struct ifnet *ifp)
1205 {
1206 	struct ifg_list		*ifgl;
1207 	struct ifg_member	*ifgm;
1208 	char groupname[IFNAMSIZ];
1209 
1210 	IFNET_WLOCK();
1211 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1212 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1213 
1214 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1215 
1216 		IF_ADDR_WLOCK(ifp);
1217 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1218 		IF_ADDR_WUNLOCK(ifp);
1219 
1220 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1221 			if (ifgm->ifgm_ifp == ifp)
1222 				break;
1223 
1224 		if (ifgm != NULL) {
1225 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1226 			    ifgm_next);
1227 			free(ifgm, M_TEMP);
1228 		}
1229 
1230 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1231 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1232 			IFNET_WUNLOCK();
1233 			EVENTHANDLER_INVOKE(group_detach_event,
1234 			    ifgl->ifgl_group);
1235 			free(ifgl->ifgl_group, M_TEMP);
1236 		} else
1237 			IFNET_WUNLOCK();
1238 
1239 		free(ifgl, M_TEMP);
1240 
1241 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1242 
1243 		IFNET_WLOCK();
1244 	}
1245 	IFNET_WUNLOCK();
1246 }
1247 
1248 /*
1249  * Stores all groups from an interface in memory pointed
1250  * to by data
1251  */
1252 static int
1253 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1254 {
1255 	int			 len, error;
1256 	struct ifg_list		*ifgl;
1257 	struct ifg_req		 ifgrq, *ifgp;
1258 	struct ifgroupreq	*ifgr = data;
1259 
1260 	if (ifgr->ifgr_len == 0) {
1261 		IF_ADDR_RLOCK(ifp);
1262 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1263 			ifgr->ifgr_len += sizeof(struct ifg_req);
1264 		IF_ADDR_RUNLOCK(ifp);
1265 		return (0);
1266 	}
1267 
1268 	len = ifgr->ifgr_len;
1269 	ifgp = ifgr->ifgr_groups;
1270 	/* XXX: wire */
1271 	IF_ADDR_RLOCK(ifp);
1272 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1273 		if (len < sizeof(ifgrq)) {
1274 			IF_ADDR_RUNLOCK(ifp);
1275 			return (EINVAL);
1276 		}
1277 		bzero(&ifgrq, sizeof ifgrq);
1278 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1279 		    sizeof(ifgrq.ifgrq_group));
1280 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1281 		    	IF_ADDR_RUNLOCK(ifp);
1282 			return (error);
1283 		}
1284 		len -= sizeof(ifgrq);
1285 		ifgp++;
1286 	}
1287 	IF_ADDR_RUNLOCK(ifp);
1288 
1289 	return (0);
1290 }
1291 
1292 /*
1293  * Stores all members of a group in memory pointed to by data
1294  */
1295 static int
1296 if_getgroupmembers(struct ifgroupreq *data)
1297 {
1298 	struct ifgroupreq	*ifgr = data;
1299 	struct ifg_group	*ifg;
1300 	struct ifg_member	*ifgm;
1301 	struct ifg_req		 ifgrq, *ifgp;
1302 	int			 len, error;
1303 
1304 	IFNET_RLOCK();
1305 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1306 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1307 			break;
1308 	if (ifg == NULL) {
1309 		IFNET_RUNLOCK();
1310 		return (ENOENT);
1311 	}
1312 
1313 	if (ifgr->ifgr_len == 0) {
1314 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1315 			ifgr->ifgr_len += sizeof(ifgrq);
1316 		IFNET_RUNLOCK();
1317 		return (0);
1318 	}
1319 
1320 	len = ifgr->ifgr_len;
1321 	ifgp = ifgr->ifgr_groups;
1322 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1323 		if (len < sizeof(ifgrq)) {
1324 			IFNET_RUNLOCK();
1325 			return (EINVAL);
1326 		}
1327 		bzero(&ifgrq, sizeof ifgrq);
1328 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1329 		    sizeof(ifgrq.ifgrq_member));
1330 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1331 			IFNET_RUNLOCK();
1332 			return (error);
1333 		}
1334 		len -= sizeof(ifgrq);
1335 		ifgp++;
1336 	}
1337 	IFNET_RUNLOCK();
1338 
1339 	return (0);
1340 }
1341 
1342 /*
1343  * Delete Routes for a Network Interface
1344  *
1345  * Called for each routing entry via the rnh->rnh_walktree() call above
1346  * to delete all route entries referencing a detaching network interface.
1347  *
1348  * Arguments:
1349  *	rn	pointer to node in the routing table
1350  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1351  *
1352  * Returns:
1353  *	0	successful
1354  *	errno	failed - reason indicated
1355  *
1356  */
1357 static int
1358 if_rtdel(struct radix_node *rn, void *arg)
1359 {
1360 	struct rtentry	*rt = (struct rtentry *)rn;
1361 	struct ifnet	*ifp = arg;
1362 	int		err;
1363 
1364 	if (rt->rt_ifp == ifp) {
1365 
1366 		/*
1367 		 * Protect (sorta) against walktree recursion problems
1368 		 * with cloned routes
1369 		 */
1370 		if ((rt->rt_flags & RTF_UP) == 0)
1371 			return (0);
1372 
1373 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1374 				rt_mask(rt),
1375 				rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1376 				(struct rtentry **) NULL, rt->rt_fibnum);
1377 		if (err) {
1378 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1379 		}
1380 	}
1381 
1382 	return (0);
1383 }
1384 
1385 /*
1386  * Wrapper functions for struct ifnet address list locking macros.  These are
1387  * used by kernel modules to avoid encoding programming interface or binary
1388  * interface assumptions that may be violated when kernel-internal locking
1389  * approaches change.
1390  */
1391 void
1392 if_addr_rlock(struct ifnet *ifp)
1393 {
1394 
1395 	IF_ADDR_RLOCK(ifp);
1396 }
1397 
1398 void
1399 if_addr_runlock(struct ifnet *ifp)
1400 {
1401 
1402 	IF_ADDR_RUNLOCK(ifp);
1403 }
1404 
1405 void
1406 if_maddr_rlock(if_t ifp)
1407 {
1408 
1409 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1410 }
1411 
1412 void
1413 if_maddr_runlock(if_t ifp)
1414 {
1415 
1416 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1417 }
1418 
1419 /*
1420  * Initialization, destruction and refcounting functions for ifaddrs.
1421  */
1422 struct ifaddr *
1423 ifa_alloc(size_t size, int flags)
1424 {
1425 	struct ifaddr *ifa;
1426 
1427 	KASSERT(size >= sizeof(struct ifaddr),
1428 	    ("%s: invalid size %zu", __func__, size));
1429 
1430 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1431 	if (ifa == NULL)
1432 		return (NULL);
1433 
1434 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1435 		goto fail;
1436 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1437 		goto fail;
1438 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1439 		goto fail;
1440 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1441 		goto fail;
1442 
1443 	refcount_init(&ifa->ifa_refcnt, 1);
1444 
1445 	return (ifa);
1446 
1447 fail:
1448 	/* free(NULL) is okay */
1449 	counter_u64_free(ifa->ifa_opackets);
1450 	counter_u64_free(ifa->ifa_ipackets);
1451 	counter_u64_free(ifa->ifa_obytes);
1452 	counter_u64_free(ifa->ifa_ibytes);
1453 	free(ifa, M_IFADDR);
1454 
1455 	return (NULL);
1456 }
1457 
1458 void
1459 ifa_ref(struct ifaddr *ifa)
1460 {
1461 
1462 	refcount_acquire(&ifa->ifa_refcnt);
1463 }
1464 
1465 void
1466 ifa_free(struct ifaddr *ifa)
1467 {
1468 
1469 	if (refcount_release(&ifa->ifa_refcnt)) {
1470 		counter_u64_free(ifa->ifa_opackets);
1471 		counter_u64_free(ifa->ifa_ipackets);
1472 		counter_u64_free(ifa->ifa_obytes);
1473 		counter_u64_free(ifa->ifa_ibytes);
1474 		free(ifa, M_IFADDR);
1475 	}
1476 }
1477 
1478 int
1479 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1480 {
1481 	int error = 0;
1482 	struct rtentry *rt = NULL;
1483 	struct rt_addrinfo info;
1484 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1485 
1486 	bzero(&info, sizeof(info));
1487 	info.rti_ifp = V_loif;
1488 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1489 	info.rti_info[RTAX_DST] = ia;
1490 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1491 	error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1492 
1493 	if (error == 0 && rt != NULL) {
1494 		RT_LOCK(rt);
1495 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1496 			ifa->ifa_ifp->if_type;
1497 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1498 			ifa->ifa_ifp->if_index;
1499 		RT_REMREF(rt);
1500 		RT_UNLOCK(rt);
1501 	} else if (error != 0)
1502 		log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1503 
1504 	return (error);
1505 }
1506 
1507 int
1508 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1509 {
1510 	int error = 0;
1511 	struct rt_addrinfo info;
1512 	struct sockaddr_dl null_sdl;
1513 
1514 	bzero(&null_sdl, sizeof(null_sdl));
1515 	null_sdl.sdl_len = sizeof(null_sdl);
1516 	null_sdl.sdl_family = AF_LINK;
1517 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1518 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1519 	bzero(&info, sizeof(info));
1520 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1521 	info.rti_info[RTAX_DST] = ia;
1522 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1523 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1524 
1525 	if (error != 0)
1526 		log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1527 
1528 	return (error);
1529 }
1530 
1531 int
1532 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib)
1533 {
1534 	struct rtentry *rt;
1535 
1536 	rt = rtalloc1_fib(sa, 0, 0, fib);
1537 	if (rt == NULL) {
1538 		log(LOG_DEBUG, "%s: fail", __func__);
1539 		return (EHOSTUNREACH);
1540 	}
1541 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1542 	    ifa->ifa_ifp->if_type;
1543 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1544 	    ifa->ifa_ifp->if_index;
1545 	RTFREE_LOCKED(rt);
1546 
1547 	return (0);
1548 }
1549 
1550 /*
1551  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1552  * structs used to represent other address families, it is necessary
1553  * to perform a different comparison.
1554  */
1555 
1556 #define	sa_dl_equal(a1, a2)	\
1557 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1558 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1559 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1560 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1561 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1562 
1563 /*
1564  * Locate an interface based on a complete address.
1565  */
1566 /*ARGSUSED*/
1567 static struct ifaddr *
1568 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1569 {
1570 	struct ifnet *ifp;
1571 	struct ifaddr *ifa;
1572 
1573 	IFNET_RLOCK_NOSLEEP();
1574 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1575 		IF_ADDR_RLOCK(ifp);
1576 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1577 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1578 				continue;
1579 			if (sa_equal(addr, ifa->ifa_addr)) {
1580 				if (getref)
1581 					ifa_ref(ifa);
1582 				IF_ADDR_RUNLOCK(ifp);
1583 				goto done;
1584 			}
1585 			/* IP6 doesn't have broadcast */
1586 			if ((ifp->if_flags & IFF_BROADCAST) &&
1587 			    ifa->ifa_broadaddr &&
1588 			    ifa->ifa_broadaddr->sa_len != 0 &&
1589 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1590 				if (getref)
1591 					ifa_ref(ifa);
1592 				IF_ADDR_RUNLOCK(ifp);
1593 				goto done;
1594 			}
1595 		}
1596 		IF_ADDR_RUNLOCK(ifp);
1597 	}
1598 	ifa = NULL;
1599 done:
1600 	IFNET_RUNLOCK_NOSLEEP();
1601 	return (ifa);
1602 }
1603 
1604 struct ifaddr *
1605 ifa_ifwithaddr(struct sockaddr *addr)
1606 {
1607 
1608 	return (ifa_ifwithaddr_internal(addr, 1));
1609 }
1610 
1611 int
1612 ifa_ifwithaddr_check(struct sockaddr *addr)
1613 {
1614 
1615 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1616 }
1617 
1618 /*
1619  * Locate an interface based on the broadcast address.
1620  */
1621 /* ARGSUSED */
1622 struct ifaddr *
1623 ifa_ifwithbroadaddr(struct sockaddr *addr)
1624 {
1625 	struct ifnet *ifp;
1626 	struct ifaddr *ifa;
1627 
1628 	IFNET_RLOCK_NOSLEEP();
1629 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1630 		IF_ADDR_RLOCK(ifp);
1631 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1632 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1633 				continue;
1634 			if ((ifp->if_flags & IFF_BROADCAST) &&
1635 			    ifa->ifa_broadaddr &&
1636 			    ifa->ifa_broadaddr->sa_len != 0 &&
1637 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1638 				ifa_ref(ifa);
1639 				IF_ADDR_RUNLOCK(ifp);
1640 				goto done;
1641 			}
1642 		}
1643 		IF_ADDR_RUNLOCK(ifp);
1644 	}
1645 	ifa = NULL;
1646 done:
1647 	IFNET_RUNLOCK_NOSLEEP();
1648 	return (ifa);
1649 }
1650 
1651 /*
1652  * Locate the point to point interface with a given destination address.
1653  */
1654 /*ARGSUSED*/
1655 struct ifaddr *
1656 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
1657 {
1658 	struct ifnet *ifp;
1659 	struct ifaddr *ifa;
1660 
1661 	IFNET_RLOCK_NOSLEEP();
1662 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1663 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1664 			continue;
1665 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1666 			continue;
1667 		IF_ADDR_RLOCK(ifp);
1668 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1669 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1670 				continue;
1671 			if (ifa->ifa_dstaddr != NULL &&
1672 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1673 				ifa_ref(ifa);
1674 				IF_ADDR_RUNLOCK(ifp);
1675 				goto done;
1676 			}
1677 		}
1678 		IF_ADDR_RUNLOCK(ifp);
1679 	}
1680 	ifa = NULL;
1681 done:
1682 	IFNET_RUNLOCK_NOSLEEP();
1683 	return (ifa);
1684 }
1685 
1686 struct ifaddr *
1687 ifa_ifwithdstaddr(struct sockaddr *addr)
1688 {
1689 
1690 	return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
1691 }
1692 
1693 /*
1694  * Find an interface on a specific network.  If many, choice
1695  * is most specific found.
1696  */
1697 struct ifaddr *
1698 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
1699 {
1700 	struct ifnet *ifp;
1701 	struct ifaddr *ifa;
1702 	struct ifaddr *ifa_maybe = NULL;
1703 	u_int af = addr->sa_family;
1704 	char *addr_data = addr->sa_data, *cplim;
1705 
1706 	/*
1707 	 * AF_LINK addresses can be looked up directly by their index number,
1708 	 * so do that if we can.
1709 	 */
1710 	if (af == AF_LINK) {
1711 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1712 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1713 		return (ifaddr_byindex(sdl->sdl_index));
1714 	}
1715 
1716 	/*
1717 	 * Scan though each interface, looking for ones that have addresses
1718 	 * in this address family and the requested fib.  Maintain a reference
1719 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1720 	 * kept it stable when we move onto the next interface.
1721 	 */
1722 	IFNET_RLOCK_NOSLEEP();
1723 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1724 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1725 			continue;
1726 		IF_ADDR_RLOCK(ifp);
1727 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1728 			char *cp, *cp2, *cp3;
1729 
1730 			if (ifa->ifa_addr->sa_family != af)
1731 next:				continue;
1732 			if (af == AF_INET &&
1733 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1734 				/*
1735 				 * This is a bit broken as it doesn't
1736 				 * take into account that the remote end may
1737 				 * be a single node in the network we are
1738 				 * looking for.
1739 				 * The trouble is that we don't know the
1740 				 * netmask for the remote end.
1741 				 */
1742 				if (ifa->ifa_dstaddr != NULL &&
1743 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1744 					ifa_ref(ifa);
1745 					IF_ADDR_RUNLOCK(ifp);
1746 					goto done;
1747 				}
1748 			} else {
1749 				/*
1750 				 * Scan all the bits in the ifa's address.
1751 				 * If a bit dissagrees with what we are
1752 				 * looking for, mask it with the netmask
1753 				 * to see if it really matters.
1754 				 * (A byte at a time)
1755 				 */
1756 				if (ifa->ifa_netmask == 0)
1757 					continue;
1758 				cp = addr_data;
1759 				cp2 = ifa->ifa_addr->sa_data;
1760 				cp3 = ifa->ifa_netmask->sa_data;
1761 				cplim = ifa->ifa_netmask->sa_len
1762 					+ (char *)ifa->ifa_netmask;
1763 				while (cp3 < cplim)
1764 					if ((*cp++ ^ *cp2++) & *cp3++)
1765 						goto next; /* next address! */
1766 				/*
1767 				 * If the netmask of what we just found
1768 				 * is more specific than what we had before
1769 				 * (if we had one), or if the virtual status
1770 				 * of new prefix is better than of the old one,
1771 				 * then remember the new one before continuing
1772 				 * to search for an even better one.
1773 				 */
1774 				if (ifa_maybe == NULL ||
1775 				    ifa_preferred(ifa_maybe, ifa) ||
1776 				    rn_refines((caddr_t)ifa->ifa_netmask,
1777 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1778 					if (ifa_maybe != NULL)
1779 						ifa_free(ifa_maybe);
1780 					ifa_maybe = ifa;
1781 					ifa_ref(ifa_maybe);
1782 				}
1783 			}
1784 		}
1785 		IF_ADDR_RUNLOCK(ifp);
1786 	}
1787 	ifa = ifa_maybe;
1788 	ifa_maybe = NULL;
1789 done:
1790 	IFNET_RUNLOCK_NOSLEEP();
1791 	if (ifa_maybe != NULL)
1792 		ifa_free(ifa_maybe);
1793 	return (ifa);
1794 }
1795 
1796 struct ifaddr *
1797 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1798 {
1799 
1800 	return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
1801 }
1802 
1803 /*
1804  * Find an interface address specific to an interface best matching
1805  * a given address.
1806  */
1807 struct ifaddr *
1808 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1809 {
1810 	struct ifaddr *ifa;
1811 	char *cp, *cp2, *cp3;
1812 	char *cplim;
1813 	struct ifaddr *ifa_maybe = NULL;
1814 	u_int af = addr->sa_family;
1815 
1816 	if (af >= AF_MAX)
1817 		return (NULL);
1818 	IF_ADDR_RLOCK(ifp);
1819 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1820 		if (ifa->ifa_addr->sa_family != af)
1821 			continue;
1822 		if (ifa_maybe == NULL)
1823 			ifa_maybe = ifa;
1824 		if (ifa->ifa_netmask == 0) {
1825 			if (sa_equal(addr, ifa->ifa_addr) ||
1826 			    (ifa->ifa_dstaddr &&
1827 			    sa_equal(addr, ifa->ifa_dstaddr)))
1828 				goto done;
1829 			continue;
1830 		}
1831 		if (ifp->if_flags & IFF_POINTOPOINT) {
1832 			if (sa_equal(addr, ifa->ifa_dstaddr))
1833 				goto done;
1834 		} else {
1835 			cp = addr->sa_data;
1836 			cp2 = ifa->ifa_addr->sa_data;
1837 			cp3 = ifa->ifa_netmask->sa_data;
1838 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1839 			for (; cp3 < cplim; cp3++)
1840 				if ((*cp++ ^ *cp2++) & *cp3)
1841 					break;
1842 			if (cp3 == cplim)
1843 				goto done;
1844 		}
1845 	}
1846 	ifa = ifa_maybe;
1847 done:
1848 	if (ifa != NULL)
1849 		ifa_ref(ifa);
1850 	IF_ADDR_RUNLOCK(ifp);
1851 	return (ifa);
1852 }
1853 
1854 /*
1855  * See whether new ifa is better than current one:
1856  * 1) A non-virtual one is preferred over virtual.
1857  * 2) A virtual in master state preferred over any other state.
1858  *
1859  * Used in several address selecting functions.
1860  */
1861 int
1862 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1863 {
1864 
1865 	return (cur->ifa_carp && (!next->ifa_carp ||
1866 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1867 }
1868 
1869 #include <net/if_llatbl.h>
1870 
1871 /*
1872  * Default action when installing a route with a Link Level gateway.
1873  * Lookup an appropriate real ifa to point to.
1874  * This should be moved to /sys/net/link.c eventually.
1875  */
1876 static void
1877 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1878 {
1879 	struct ifaddr *ifa, *oifa;
1880 	struct sockaddr *dst;
1881 	struct ifnet *ifp;
1882 
1883 	RT_LOCK_ASSERT(rt);
1884 
1885 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1886 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1887 		return;
1888 	ifa = ifaof_ifpforaddr(dst, ifp);
1889 	if (ifa) {
1890 		oifa = rt->rt_ifa;
1891 		rt->rt_ifa = ifa;
1892 		ifa_free(oifa);
1893 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1894 			ifa->ifa_rtrequest(cmd, rt, info);
1895 	}
1896 }
1897 
1898 struct sockaddr_dl *
1899 link_alloc_sdl(size_t size, int flags)
1900 {
1901 
1902 	return (malloc(size, M_TEMP, flags));
1903 }
1904 
1905 void
1906 link_free_sdl(struct sockaddr *sa)
1907 {
1908 	free(sa, M_TEMP);
1909 }
1910 
1911 /*
1912  * Fills in given sdl with interface basic info.
1913  * Returns pointer to filled sdl.
1914  */
1915 struct sockaddr_dl *
1916 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1917 {
1918 	struct sockaddr_dl *sdl;
1919 
1920 	sdl = (struct sockaddr_dl *)paddr;
1921 	memset(sdl, 0, sizeof(struct sockaddr_dl));
1922 	sdl->sdl_len = sizeof(struct sockaddr_dl);
1923 	sdl->sdl_family = AF_LINK;
1924 	sdl->sdl_index = ifp->if_index;
1925 	sdl->sdl_type = iftype;
1926 
1927 	return (sdl);
1928 }
1929 
1930 /*
1931  * Mark an interface down and notify protocols of
1932  * the transition.
1933  */
1934 static void
1935 if_unroute(struct ifnet *ifp, int flag, int fam)
1936 {
1937 	struct ifaddr *ifa;
1938 
1939 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1940 
1941 	ifp->if_flags &= ~flag;
1942 	getmicrotime(&ifp->if_lastchange);
1943 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1944 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1945 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1946 	ifp->if_qflush(ifp);
1947 
1948 	if (ifp->if_carp)
1949 		(*carp_linkstate_p)(ifp);
1950 	rt_ifmsg(ifp);
1951 }
1952 
1953 /*
1954  * Mark an interface up and notify protocols of
1955  * the transition.
1956  */
1957 static void
1958 if_route(struct ifnet *ifp, int flag, int fam)
1959 {
1960 	struct ifaddr *ifa;
1961 
1962 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1963 
1964 	ifp->if_flags |= flag;
1965 	getmicrotime(&ifp->if_lastchange);
1966 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1967 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1968 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1969 	if (ifp->if_carp)
1970 		(*carp_linkstate_p)(ifp);
1971 	rt_ifmsg(ifp);
1972 #ifdef INET6
1973 	in6_if_up(ifp);
1974 #endif
1975 }
1976 
1977 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1978 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1979 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
1980 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
1981 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
1982 int	(*vlan_setcookie_p)(struct ifnet *, void *);
1983 void	*(*vlan_cookie_p)(struct ifnet *);
1984 
1985 /*
1986  * Handle a change in the interface link state. To avoid LORs
1987  * between driver lock and upper layer locks, as well as possible
1988  * recursions, we post event to taskqueue, and all job
1989  * is done in static do_link_state_change().
1990  */
1991 void
1992 if_link_state_change(struct ifnet *ifp, int link_state)
1993 {
1994 	/* Return if state hasn't changed. */
1995 	if (ifp->if_link_state == link_state)
1996 		return;
1997 
1998 	ifp->if_link_state = link_state;
1999 
2000 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2001 }
2002 
2003 static void
2004 do_link_state_change(void *arg, int pending)
2005 {
2006 	struct ifnet *ifp = (struct ifnet *)arg;
2007 	int link_state = ifp->if_link_state;
2008 	CURVNET_SET(ifp->if_vnet);
2009 
2010 	/* Notify that the link state has changed. */
2011 	rt_ifmsg(ifp);
2012 	if (ifp->if_vlantrunk != NULL)
2013 		(*vlan_link_state_p)(ifp);
2014 
2015 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2016 	    IFP2AC(ifp)->ac_netgraph != NULL)
2017 		(*ng_ether_link_state_p)(ifp, link_state);
2018 	if (ifp->if_carp)
2019 		(*carp_linkstate_p)(ifp);
2020 	if (ifp->if_bridge)
2021 		(*bridge_linkstate_p)(ifp);
2022 	if (ifp->if_lagg)
2023 		(*lagg_linkstate_p)(ifp, link_state);
2024 
2025 	if (IS_DEFAULT_VNET(curvnet))
2026 		devctl_notify("IFNET", ifp->if_xname,
2027 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2028 		    NULL);
2029 	if (pending > 1)
2030 		if_printf(ifp, "%d link states coalesced\n", pending);
2031 	if (log_link_state_change)
2032 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2033 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2034 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2035 	CURVNET_RESTORE();
2036 }
2037 
2038 /*
2039  * Mark an interface down and notify protocols of
2040  * the transition.
2041  */
2042 void
2043 if_down(struct ifnet *ifp)
2044 {
2045 
2046 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2047 }
2048 
2049 /*
2050  * Mark an interface up and notify protocols of
2051  * the transition.
2052  */
2053 void
2054 if_up(struct ifnet *ifp)
2055 {
2056 
2057 	if_route(ifp, IFF_UP, AF_UNSPEC);
2058 }
2059 
2060 /*
2061  * Flush an interface queue.
2062  */
2063 void
2064 if_qflush(struct ifnet *ifp)
2065 {
2066 	struct mbuf *m, *n;
2067 	struct ifaltq *ifq;
2068 
2069 	ifq = &ifp->if_snd;
2070 	IFQ_LOCK(ifq);
2071 #ifdef ALTQ
2072 	if (ALTQ_IS_ENABLED(ifq))
2073 		ALTQ_PURGE(ifq);
2074 #endif
2075 	n = ifq->ifq_head;
2076 	while ((m = n) != 0) {
2077 		n = m->m_nextpkt;
2078 		m_freem(m);
2079 	}
2080 	ifq->ifq_head = 0;
2081 	ifq->ifq_tail = 0;
2082 	ifq->ifq_len = 0;
2083 	IFQ_UNLOCK(ifq);
2084 }
2085 
2086 /*
2087  * Map interface name to interface structure pointer, with or without
2088  * returning a reference.
2089  */
2090 struct ifnet *
2091 ifunit_ref(const char *name)
2092 {
2093 	struct ifnet *ifp;
2094 
2095 	IFNET_RLOCK_NOSLEEP();
2096 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2097 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2098 		    !(ifp->if_flags & IFF_DYING))
2099 			break;
2100 	}
2101 	if (ifp != NULL)
2102 		if_ref(ifp);
2103 	IFNET_RUNLOCK_NOSLEEP();
2104 	return (ifp);
2105 }
2106 
2107 struct ifnet *
2108 ifunit(const char *name)
2109 {
2110 	struct ifnet *ifp;
2111 
2112 	IFNET_RLOCK_NOSLEEP();
2113 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2114 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2115 			break;
2116 	}
2117 	IFNET_RUNLOCK_NOSLEEP();
2118 	return (ifp);
2119 }
2120 
2121 /*
2122  * Hardware specific interface ioctls.
2123  */
2124 static int
2125 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2126 {
2127 	struct ifreq *ifr;
2128 	int error = 0;
2129 	int new_flags, temp_flags;
2130 	size_t namelen, onamelen;
2131 	size_t descrlen;
2132 	char *descrbuf, *odescrbuf;
2133 	char new_name[IFNAMSIZ];
2134 	struct ifaddr *ifa;
2135 	struct sockaddr_dl *sdl;
2136 
2137 	ifr = (struct ifreq *)data;
2138 	switch (cmd) {
2139 	case SIOCGIFINDEX:
2140 		ifr->ifr_index = ifp->if_index;
2141 		break;
2142 
2143 	case SIOCGIFFLAGS:
2144 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2145 		ifr->ifr_flags = temp_flags & 0xffff;
2146 		ifr->ifr_flagshigh = temp_flags >> 16;
2147 		break;
2148 
2149 	case SIOCGIFCAP:
2150 		ifr->ifr_reqcap = ifp->if_capabilities;
2151 		ifr->ifr_curcap = ifp->if_capenable;
2152 		break;
2153 
2154 #ifdef MAC
2155 	case SIOCGIFMAC:
2156 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2157 		break;
2158 #endif
2159 
2160 	case SIOCGIFMETRIC:
2161 		ifr->ifr_metric = ifp->if_metric;
2162 		break;
2163 
2164 	case SIOCGIFMTU:
2165 		ifr->ifr_mtu = ifp->if_mtu;
2166 		break;
2167 
2168 	case SIOCGIFPHYS:
2169 		ifr->ifr_phys = ifp->if_physical;
2170 		break;
2171 
2172 	case SIOCGIFDESCR:
2173 		error = 0;
2174 		sx_slock(&ifdescr_sx);
2175 		if (ifp->if_description == NULL)
2176 			error = ENOMSG;
2177 		else {
2178 			/* space for terminating nul */
2179 			descrlen = strlen(ifp->if_description) + 1;
2180 			if (ifr->ifr_buffer.length < descrlen)
2181 				ifr->ifr_buffer.buffer = NULL;
2182 			else
2183 				error = copyout(ifp->if_description,
2184 				    ifr->ifr_buffer.buffer, descrlen);
2185 			ifr->ifr_buffer.length = descrlen;
2186 		}
2187 		sx_sunlock(&ifdescr_sx);
2188 		break;
2189 
2190 	case SIOCSIFDESCR:
2191 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2192 		if (error)
2193 			return (error);
2194 
2195 		/*
2196 		 * Copy only (length-1) bytes to make sure that
2197 		 * if_description is always nul terminated.  The
2198 		 * length parameter is supposed to count the
2199 		 * terminating nul in.
2200 		 */
2201 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2202 			return (ENAMETOOLONG);
2203 		else if (ifr->ifr_buffer.length == 0)
2204 			descrbuf = NULL;
2205 		else {
2206 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2207 			    M_WAITOK | M_ZERO);
2208 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2209 			    ifr->ifr_buffer.length - 1);
2210 			if (error) {
2211 				free(descrbuf, M_IFDESCR);
2212 				break;
2213 			}
2214 		}
2215 
2216 		sx_xlock(&ifdescr_sx);
2217 		odescrbuf = ifp->if_description;
2218 		ifp->if_description = descrbuf;
2219 		sx_xunlock(&ifdescr_sx);
2220 
2221 		getmicrotime(&ifp->if_lastchange);
2222 		free(odescrbuf, M_IFDESCR);
2223 		break;
2224 
2225 	case SIOCGIFFIB:
2226 		ifr->ifr_fib = ifp->if_fib;
2227 		break;
2228 
2229 	case SIOCSIFFIB:
2230 		error = priv_check(td, PRIV_NET_SETIFFIB);
2231 		if (error)
2232 			return (error);
2233 		if (ifr->ifr_fib >= rt_numfibs)
2234 			return (EINVAL);
2235 
2236 		ifp->if_fib = ifr->ifr_fib;
2237 		break;
2238 
2239 	case SIOCSIFFLAGS:
2240 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2241 		if (error)
2242 			return (error);
2243 		/*
2244 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2245 		 * check, so we don't need special handling here yet.
2246 		 */
2247 		new_flags = (ifr->ifr_flags & 0xffff) |
2248 		    (ifr->ifr_flagshigh << 16);
2249 		if (ifp->if_flags & IFF_UP &&
2250 		    (new_flags & IFF_UP) == 0) {
2251 			if_down(ifp);
2252 		} else if (new_flags & IFF_UP &&
2253 		    (ifp->if_flags & IFF_UP) == 0) {
2254 			if_up(ifp);
2255 		}
2256 		/* See if permanently promiscuous mode bit is about to flip */
2257 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2258 			if (new_flags & IFF_PPROMISC)
2259 				ifp->if_flags |= IFF_PROMISC;
2260 			else if (ifp->if_pcount == 0)
2261 				ifp->if_flags &= ~IFF_PROMISC;
2262 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2263 			    ifp->if_xname,
2264 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2265 		}
2266 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2267 			(new_flags &~ IFF_CANTCHANGE);
2268 		if (ifp->if_ioctl) {
2269 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2270 		}
2271 		getmicrotime(&ifp->if_lastchange);
2272 		break;
2273 
2274 	case SIOCSIFCAP:
2275 		error = priv_check(td, PRIV_NET_SETIFCAP);
2276 		if (error)
2277 			return (error);
2278 		if (ifp->if_ioctl == NULL)
2279 			return (EOPNOTSUPP);
2280 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2281 			return (EINVAL);
2282 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2283 		if (error == 0)
2284 			getmicrotime(&ifp->if_lastchange);
2285 		break;
2286 
2287 #ifdef MAC
2288 	case SIOCSIFMAC:
2289 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2290 		break;
2291 #endif
2292 
2293 	case SIOCSIFNAME:
2294 		error = priv_check(td, PRIV_NET_SETIFNAME);
2295 		if (error)
2296 			return (error);
2297 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2298 		if (error != 0)
2299 			return (error);
2300 		if (new_name[0] == '\0')
2301 			return (EINVAL);
2302 		if (ifunit(new_name) != NULL)
2303 			return (EEXIST);
2304 
2305 		/*
2306 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2307 		 * and there are numerous other races with the
2308 		 * ifunit() checks not being atomic with namespace
2309 		 * changes (renames, vmoves, if_attach, etc).
2310 		 */
2311 		ifp->if_flags |= IFF_RENAMING;
2312 
2313 		/* Announce the departure of the interface. */
2314 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2315 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2316 
2317 		log(LOG_INFO, "%s: changing name to '%s'\n",
2318 		    ifp->if_xname, new_name);
2319 
2320 		IF_ADDR_WLOCK(ifp);
2321 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2322 		ifa = ifp->if_addr;
2323 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2324 		namelen = strlen(new_name);
2325 		onamelen = sdl->sdl_nlen;
2326 		/*
2327 		 * Move the address if needed.  This is safe because we
2328 		 * allocate space for a name of length IFNAMSIZ when we
2329 		 * create this in if_attach().
2330 		 */
2331 		if (namelen != onamelen) {
2332 			bcopy(sdl->sdl_data + onamelen,
2333 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2334 		}
2335 		bcopy(new_name, sdl->sdl_data, namelen);
2336 		sdl->sdl_nlen = namelen;
2337 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2338 		bzero(sdl->sdl_data, onamelen);
2339 		while (namelen != 0)
2340 			sdl->sdl_data[--namelen] = 0xff;
2341 		IF_ADDR_WUNLOCK(ifp);
2342 
2343 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2344 		/* Announce the return of the interface. */
2345 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2346 
2347 		ifp->if_flags &= ~IFF_RENAMING;
2348 		break;
2349 
2350 #ifdef VIMAGE
2351 	case SIOCSIFVNET:
2352 		error = priv_check(td, PRIV_NET_SETIFVNET);
2353 		if (error)
2354 			return (error);
2355 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2356 		break;
2357 #endif
2358 
2359 	case SIOCSIFMETRIC:
2360 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2361 		if (error)
2362 			return (error);
2363 		ifp->if_metric = ifr->ifr_metric;
2364 		getmicrotime(&ifp->if_lastchange);
2365 		break;
2366 
2367 	case SIOCSIFPHYS:
2368 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2369 		if (error)
2370 			return (error);
2371 		if (ifp->if_ioctl == NULL)
2372 			return (EOPNOTSUPP);
2373 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2374 		if (error == 0)
2375 			getmicrotime(&ifp->if_lastchange);
2376 		break;
2377 
2378 	case SIOCSIFMTU:
2379 	{
2380 		u_long oldmtu = ifp->if_mtu;
2381 
2382 		error = priv_check(td, PRIV_NET_SETIFMTU);
2383 		if (error)
2384 			return (error);
2385 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2386 			return (EINVAL);
2387 		if (ifp->if_ioctl == NULL)
2388 			return (EOPNOTSUPP);
2389 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2390 		if (error == 0) {
2391 			getmicrotime(&ifp->if_lastchange);
2392 			rt_ifmsg(ifp);
2393 		}
2394 		/*
2395 		 * If the link MTU changed, do network layer specific procedure.
2396 		 */
2397 		if (ifp->if_mtu != oldmtu) {
2398 #ifdef INET6
2399 			nd6_setmtu(ifp);
2400 #endif
2401 		}
2402 		break;
2403 	}
2404 
2405 	case SIOCADDMULTI:
2406 	case SIOCDELMULTI:
2407 		if (cmd == SIOCADDMULTI)
2408 			error = priv_check(td, PRIV_NET_ADDMULTI);
2409 		else
2410 			error = priv_check(td, PRIV_NET_DELMULTI);
2411 		if (error)
2412 			return (error);
2413 
2414 		/* Don't allow group membership on non-multicast interfaces. */
2415 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2416 			return (EOPNOTSUPP);
2417 
2418 		/* Don't let users screw up protocols' entries. */
2419 		if (ifr->ifr_addr.sa_family != AF_LINK)
2420 			return (EINVAL);
2421 
2422 		if (cmd == SIOCADDMULTI) {
2423 			struct ifmultiaddr *ifma;
2424 
2425 			/*
2426 			 * Userland is only permitted to join groups once
2427 			 * via the if_addmulti() KPI, because it cannot hold
2428 			 * struct ifmultiaddr * between calls. It may also
2429 			 * lose a race while we check if the membership
2430 			 * already exists.
2431 			 */
2432 			IF_ADDR_RLOCK(ifp);
2433 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2434 			IF_ADDR_RUNLOCK(ifp);
2435 			if (ifma != NULL)
2436 				error = EADDRINUSE;
2437 			else
2438 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2439 		} else {
2440 			error = if_delmulti(ifp, &ifr->ifr_addr);
2441 		}
2442 		if (error == 0)
2443 			getmicrotime(&ifp->if_lastchange);
2444 		break;
2445 
2446 	case SIOCSIFPHYADDR:
2447 	case SIOCDIFPHYADDR:
2448 #ifdef INET6
2449 	case SIOCSIFPHYADDR_IN6:
2450 #endif
2451 	case SIOCSIFMEDIA:
2452 	case SIOCSIFGENERIC:
2453 		error = priv_check(td, PRIV_NET_HWIOCTL);
2454 		if (error)
2455 			return (error);
2456 		if (ifp->if_ioctl == NULL)
2457 			return (EOPNOTSUPP);
2458 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2459 		if (error == 0)
2460 			getmicrotime(&ifp->if_lastchange);
2461 		break;
2462 
2463 	case SIOCGIFSTATUS:
2464 	case SIOCGIFPSRCADDR:
2465 	case SIOCGIFPDSTADDR:
2466 	case SIOCGIFMEDIA:
2467 	case SIOCGIFGENERIC:
2468 		if (ifp->if_ioctl == NULL)
2469 			return (EOPNOTSUPP);
2470 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2471 		break;
2472 
2473 	case SIOCSIFLLADDR:
2474 		error = priv_check(td, PRIV_NET_SETLLADDR);
2475 		if (error)
2476 			return (error);
2477 		error = if_setlladdr(ifp,
2478 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2479 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2480 		break;
2481 
2482 	case SIOCAIFGROUP:
2483 	{
2484 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2485 
2486 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2487 		if (error)
2488 			return (error);
2489 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2490 			return (error);
2491 		break;
2492 	}
2493 
2494 	case SIOCGIFGROUP:
2495 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2496 			return (error);
2497 		break;
2498 
2499 	case SIOCDIFGROUP:
2500 	{
2501 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2502 
2503 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2504 		if (error)
2505 			return (error);
2506 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2507 			return (error);
2508 		break;
2509 	}
2510 
2511 	default:
2512 		error = ENOIOCTL;
2513 		break;
2514 	}
2515 	return (error);
2516 }
2517 
2518 #ifdef COMPAT_FREEBSD32
2519 struct ifconf32 {
2520 	int32_t	ifc_len;
2521 	union {
2522 		uint32_t	ifcu_buf;
2523 		uint32_t	ifcu_req;
2524 	} ifc_ifcu;
2525 };
2526 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2527 #endif
2528 
2529 /*
2530  * Interface ioctls.
2531  */
2532 int
2533 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2534 {
2535 	struct ifnet *ifp;
2536 	struct ifreq *ifr;
2537 	int error;
2538 	int oif_flags;
2539 
2540 	CURVNET_SET(so->so_vnet);
2541 	switch (cmd) {
2542 	case SIOCGIFCONF:
2543 		error = ifconf(cmd, data);
2544 		CURVNET_RESTORE();
2545 		return (error);
2546 
2547 #ifdef COMPAT_FREEBSD32
2548 	case SIOCGIFCONF32:
2549 		{
2550 			struct ifconf32 *ifc32;
2551 			struct ifconf ifc;
2552 
2553 			ifc32 = (struct ifconf32 *)data;
2554 			ifc.ifc_len = ifc32->ifc_len;
2555 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2556 
2557 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2558 			CURVNET_RESTORE();
2559 			if (error == 0)
2560 				ifc32->ifc_len = ifc.ifc_len;
2561 			return (error);
2562 		}
2563 #endif
2564 	}
2565 	ifr = (struct ifreq *)data;
2566 
2567 	switch (cmd) {
2568 #ifdef VIMAGE
2569 	case SIOCSIFRVNET:
2570 		error = priv_check(td, PRIV_NET_SETIFVNET);
2571 		if (error == 0)
2572 			error = if_vmove_reclaim(td, ifr->ifr_name,
2573 			    ifr->ifr_jid);
2574 		CURVNET_RESTORE();
2575 		return (error);
2576 #endif
2577 	case SIOCIFCREATE:
2578 	case SIOCIFCREATE2:
2579 		error = priv_check(td, PRIV_NET_IFCREATE);
2580 		if (error == 0)
2581 			error = if_clone_create(ifr->ifr_name,
2582 			    sizeof(ifr->ifr_name),
2583 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2584 		CURVNET_RESTORE();
2585 		return (error);
2586 	case SIOCIFDESTROY:
2587 		error = priv_check(td, PRIV_NET_IFDESTROY);
2588 		if (error == 0)
2589 			error = if_clone_destroy(ifr->ifr_name);
2590 		CURVNET_RESTORE();
2591 		return (error);
2592 
2593 	case SIOCIFGCLONERS:
2594 		error = if_clone_list((struct if_clonereq *)data);
2595 		CURVNET_RESTORE();
2596 		return (error);
2597 	case SIOCGIFGMEMB:
2598 		error = if_getgroupmembers((struct ifgroupreq *)data);
2599 		CURVNET_RESTORE();
2600 		return (error);
2601 #if defined(INET) || defined(INET6)
2602 	case SIOCSVH:
2603 	case SIOCGVH:
2604 		if (carp_ioctl_p == NULL)
2605 			error = EPROTONOSUPPORT;
2606 		else
2607 			error = (*carp_ioctl_p)(ifr, cmd, td);
2608 		CURVNET_RESTORE();
2609 		return (error);
2610 #endif
2611 	}
2612 
2613 	ifp = ifunit_ref(ifr->ifr_name);
2614 	if (ifp == NULL) {
2615 		CURVNET_RESTORE();
2616 		return (ENXIO);
2617 	}
2618 
2619 	error = ifhwioctl(cmd, ifp, data, td);
2620 	if (error != ENOIOCTL) {
2621 		if_rele(ifp);
2622 		CURVNET_RESTORE();
2623 		return (error);
2624 	}
2625 
2626 	oif_flags = ifp->if_flags;
2627 	if (so->so_proto == NULL) {
2628 		if_rele(ifp);
2629 		CURVNET_RESTORE();
2630 		return (EOPNOTSUPP);
2631 	}
2632 
2633 	/*
2634 	 * Pass the request on to the socket control method, and if the
2635 	 * latter returns EOPNOTSUPP, directly to the interface.
2636 	 *
2637 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2638 	 * trust SIOCSIFADDR et al to come from an already privileged
2639 	 * layer, and do not perform any credentials checks or input
2640 	 * validation.
2641 	 */
2642 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2643 	    ifp, td));
2644 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2645 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2646 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2647 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2648 
2649 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2650 #ifdef INET6
2651 		if (ifp->if_flags & IFF_UP)
2652 			in6_if_up(ifp);
2653 #endif
2654 	}
2655 	if_rele(ifp);
2656 	CURVNET_RESTORE();
2657 	return (error);
2658 }
2659 
2660 /*
2661  * The code common to handling reference counted flags,
2662  * e.g., in ifpromisc() and if_allmulti().
2663  * The "pflag" argument can specify a permanent mode flag to check,
2664  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2665  *
2666  * Only to be used on stack-owned flags, not driver-owned flags.
2667  */
2668 static int
2669 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2670 {
2671 	struct ifreq ifr;
2672 	int error;
2673 	int oldflags, oldcount;
2674 
2675 	/* Sanity checks to catch programming errors */
2676 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2677 	    ("%s: setting driver-owned flag %d", __func__, flag));
2678 
2679 	if (onswitch)
2680 		KASSERT(*refcount >= 0,
2681 		    ("%s: increment negative refcount %d for flag %d",
2682 		    __func__, *refcount, flag));
2683 	else
2684 		KASSERT(*refcount > 0,
2685 		    ("%s: decrement non-positive refcount %d for flag %d",
2686 		    __func__, *refcount, flag));
2687 
2688 	/* In case this mode is permanent, just touch refcount */
2689 	if (ifp->if_flags & pflag) {
2690 		*refcount += onswitch ? 1 : -1;
2691 		return (0);
2692 	}
2693 
2694 	/* Save ifnet parameters for if_ioctl() may fail */
2695 	oldcount = *refcount;
2696 	oldflags = ifp->if_flags;
2697 
2698 	/*
2699 	 * See if we aren't the only and touching refcount is enough.
2700 	 * Actually toggle interface flag if we are the first or last.
2701 	 */
2702 	if (onswitch) {
2703 		if ((*refcount)++)
2704 			return (0);
2705 		ifp->if_flags |= flag;
2706 	} else {
2707 		if (--(*refcount))
2708 			return (0);
2709 		ifp->if_flags &= ~flag;
2710 	}
2711 
2712 	/* Call down the driver since we've changed interface flags */
2713 	if (ifp->if_ioctl == NULL) {
2714 		error = EOPNOTSUPP;
2715 		goto recover;
2716 	}
2717 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2718 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2719 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2720 	if (error)
2721 		goto recover;
2722 	/* Notify userland that interface flags have changed */
2723 	rt_ifmsg(ifp);
2724 	return (0);
2725 
2726 recover:
2727 	/* Recover after driver error */
2728 	*refcount = oldcount;
2729 	ifp->if_flags = oldflags;
2730 	return (error);
2731 }
2732 
2733 /*
2734  * Set/clear promiscuous mode on interface ifp based on the truth value
2735  * of pswitch.  The calls are reference counted so that only the first
2736  * "on" request actually has an effect, as does the final "off" request.
2737  * Results are undefined if the "off" and "on" requests are not matched.
2738  */
2739 int
2740 ifpromisc(struct ifnet *ifp, int pswitch)
2741 {
2742 	int error;
2743 	int oldflags = ifp->if_flags;
2744 
2745 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2746 			   &ifp->if_pcount, pswitch);
2747 	/* If promiscuous mode status has changed, log a message */
2748 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2749 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2750 		    ifp->if_xname,
2751 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2752 	return (error);
2753 }
2754 
2755 /*
2756  * Return interface configuration
2757  * of system.  List may be used
2758  * in later ioctl's (above) to get
2759  * other information.
2760  */
2761 /*ARGSUSED*/
2762 static int
2763 ifconf(u_long cmd, caddr_t data)
2764 {
2765 	struct ifconf *ifc = (struct ifconf *)data;
2766 	struct ifnet *ifp;
2767 	struct ifaddr *ifa;
2768 	struct ifreq ifr;
2769 	struct sbuf *sb;
2770 	int error, full = 0, valid_len, max_len;
2771 
2772 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2773 	max_len = MAXPHYS - 1;
2774 
2775 	/* Prevent hostile input from being able to crash the system */
2776 	if (ifc->ifc_len <= 0)
2777 		return (EINVAL);
2778 
2779 again:
2780 	if (ifc->ifc_len <= max_len) {
2781 		max_len = ifc->ifc_len;
2782 		full = 1;
2783 	}
2784 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2785 	max_len = 0;
2786 	valid_len = 0;
2787 
2788 	IFNET_RLOCK();
2789 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2790 		int addrs;
2791 
2792 		/*
2793 		 * Zero the ifr_name buffer to make sure we don't
2794 		 * disclose the contents of the stack.
2795 		 */
2796 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2797 
2798 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2799 		    >= sizeof(ifr.ifr_name)) {
2800 			sbuf_delete(sb);
2801 			IFNET_RUNLOCK();
2802 			return (ENAMETOOLONG);
2803 		}
2804 
2805 		addrs = 0;
2806 		IF_ADDR_RLOCK(ifp);
2807 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2808 			struct sockaddr *sa = ifa->ifa_addr;
2809 
2810 			if (prison_if(curthread->td_ucred, sa) != 0)
2811 				continue;
2812 			addrs++;
2813 			if (sa->sa_len <= sizeof(*sa)) {
2814 				ifr.ifr_addr = *sa;
2815 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2816 				max_len += sizeof(ifr);
2817 			} else {
2818 				sbuf_bcat(sb, &ifr,
2819 				    offsetof(struct ifreq, ifr_addr));
2820 				max_len += offsetof(struct ifreq, ifr_addr);
2821 				sbuf_bcat(sb, sa, sa->sa_len);
2822 				max_len += sa->sa_len;
2823 			}
2824 
2825 			if (sbuf_error(sb) == 0)
2826 				valid_len = sbuf_len(sb);
2827 		}
2828 		IF_ADDR_RUNLOCK(ifp);
2829 		if (addrs == 0) {
2830 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2831 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2832 			max_len += sizeof(ifr);
2833 
2834 			if (sbuf_error(sb) == 0)
2835 				valid_len = sbuf_len(sb);
2836 		}
2837 	}
2838 	IFNET_RUNLOCK();
2839 
2840 	/*
2841 	 * If we didn't allocate enough space (uncommon), try again.  If
2842 	 * we have already allocated as much space as we are allowed,
2843 	 * return what we've got.
2844 	 */
2845 	if (valid_len != max_len && !full) {
2846 		sbuf_delete(sb);
2847 		goto again;
2848 	}
2849 
2850 	ifc->ifc_len = valid_len;
2851 	sbuf_finish(sb);
2852 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2853 	sbuf_delete(sb);
2854 	return (error);
2855 }
2856 
2857 /*
2858  * Just like ifpromisc(), but for all-multicast-reception mode.
2859  */
2860 int
2861 if_allmulti(struct ifnet *ifp, int onswitch)
2862 {
2863 
2864 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2865 }
2866 
2867 struct ifmultiaddr *
2868 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2869 {
2870 	struct ifmultiaddr *ifma;
2871 
2872 	IF_ADDR_LOCK_ASSERT(ifp);
2873 
2874 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2875 		if (sa->sa_family == AF_LINK) {
2876 			if (sa_dl_equal(ifma->ifma_addr, sa))
2877 				break;
2878 		} else {
2879 			if (sa_equal(ifma->ifma_addr, sa))
2880 				break;
2881 		}
2882 	}
2883 
2884 	return ifma;
2885 }
2886 
2887 /*
2888  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2889  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2890  * the ifnet multicast address list here, so the caller must do that and
2891  * other setup work (such as notifying the device driver).  The reference
2892  * count is initialized to 1.
2893  */
2894 static struct ifmultiaddr *
2895 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2896     int mflags)
2897 {
2898 	struct ifmultiaddr *ifma;
2899 	struct sockaddr *dupsa;
2900 
2901 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2902 	    M_ZERO);
2903 	if (ifma == NULL)
2904 		return (NULL);
2905 
2906 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2907 	if (dupsa == NULL) {
2908 		free(ifma, M_IFMADDR);
2909 		return (NULL);
2910 	}
2911 	bcopy(sa, dupsa, sa->sa_len);
2912 	ifma->ifma_addr = dupsa;
2913 
2914 	ifma->ifma_ifp = ifp;
2915 	ifma->ifma_refcount = 1;
2916 	ifma->ifma_protospec = NULL;
2917 
2918 	if (llsa == NULL) {
2919 		ifma->ifma_lladdr = NULL;
2920 		return (ifma);
2921 	}
2922 
2923 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2924 	if (dupsa == NULL) {
2925 		free(ifma->ifma_addr, M_IFMADDR);
2926 		free(ifma, M_IFMADDR);
2927 		return (NULL);
2928 	}
2929 	bcopy(llsa, dupsa, llsa->sa_len);
2930 	ifma->ifma_lladdr = dupsa;
2931 
2932 	return (ifma);
2933 }
2934 
2935 /*
2936  * if_freemulti: free ifmultiaddr structure and possibly attached related
2937  * addresses.  The caller is responsible for implementing reference
2938  * counting, notifying the driver, handling routing messages, and releasing
2939  * any dependent link layer state.
2940  */
2941 static void
2942 if_freemulti(struct ifmultiaddr *ifma)
2943 {
2944 
2945 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2946 	    ifma->ifma_refcount));
2947 
2948 	if (ifma->ifma_lladdr != NULL)
2949 		free(ifma->ifma_lladdr, M_IFMADDR);
2950 	free(ifma->ifma_addr, M_IFMADDR);
2951 	free(ifma, M_IFMADDR);
2952 }
2953 
2954 /*
2955  * Register an additional multicast address with a network interface.
2956  *
2957  * - If the address is already present, bump the reference count on the
2958  *   address and return.
2959  * - If the address is not link-layer, look up a link layer address.
2960  * - Allocate address structures for one or both addresses, and attach to the
2961  *   multicast address list on the interface.  If automatically adding a link
2962  *   layer address, the protocol address will own a reference to the link
2963  *   layer address, to be freed when it is freed.
2964  * - Notify the network device driver of an addition to the multicast address
2965  *   list.
2966  *
2967  * 'sa' points to caller-owned memory with the desired multicast address.
2968  *
2969  * 'retifma' will be used to return a pointer to the resulting multicast
2970  * address reference, if desired.
2971  */
2972 int
2973 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2974     struct ifmultiaddr **retifma)
2975 {
2976 	struct ifmultiaddr *ifma, *ll_ifma;
2977 	struct sockaddr *llsa;
2978 	struct sockaddr_dl sdl;
2979 	int error;
2980 
2981 	/*
2982 	 * If the address is already present, return a new reference to it;
2983 	 * otherwise, allocate storage and set up a new address.
2984 	 */
2985 	IF_ADDR_WLOCK(ifp);
2986 	ifma = if_findmulti(ifp, sa);
2987 	if (ifma != NULL) {
2988 		ifma->ifma_refcount++;
2989 		if (retifma != NULL)
2990 			*retifma = ifma;
2991 		IF_ADDR_WUNLOCK(ifp);
2992 		return (0);
2993 	}
2994 
2995 	/*
2996 	 * The address isn't already present; resolve the protocol address
2997 	 * into a link layer address, and then look that up, bump its
2998 	 * refcount or allocate an ifma for that also.
2999 	 * Most link layer resolving functions returns address data which
3000 	 * fits inside default sockaddr_dl structure. However callback
3001 	 * can allocate another sockaddr structure, in that case we need to
3002 	 * free it later.
3003 	 */
3004 	llsa = NULL;
3005 	ll_ifma = NULL;
3006 	if (ifp->if_resolvemulti != NULL) {
3007 		/* Provide called function with buffer size information */
3008 		sdl.sdl_len = sizeof(sdl);
3009 		llsa = (struct sockaddr *)&sdl;
3010 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3011 		if (error)
3012 			goto unlock_out;
3013 	}
3014 
3015 	/*
3016 	 * Allocate the new address.  Don't hook it up yet, as we may also
3017 	 * need to allocate a link layer multicast address.
3018 	 */
3019 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3020 	if (ifma == NULL) {
3021 		error = ENOMEM;
3022 		goto free_llsa_out;
3023 	}
3024 
3025 	/*
3026 	 * If a link layer address is found, we'll need to see if it's
3027 	 * already present in the address list, or allocate is as well.
3028 	 * When this block finishes, the link layer address will be on the
3029 	 * list.
3030 	 */
3031 	if (llsa != NULL) {
3032 		ll_ifma = if_findmulti(ifp, llsa);
3033 		if (ll_ifma == NULL) {
3034 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3035 			if (ll_ifma == NULL) {
3036 				--ifma->ifma_refcount;
3037 				if_freemulti(ifma);
3038 				error = ENOMEM;
3039 				goto free_llsa_out;
3040 			}
3041 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3042 			    ifma_link);
3043 		} else
3044 			ll_ifma->ifma_refcount++;
3045 		ifma->ifma_llifma = ll_ifma;
3046 	}
3047 
3048 	/*
3049 	 * We now have a new multicast address, ifma, and possibly a new or
3050 	 * referenced link layer address.  Add the primary address to the
3051 	 * ifnet address list.
3052 	 */
3053 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3054 
3055 	if (retifma != NULL)
3056 		*retifma = ifma;
3057 
3058 	/*
3059 	 * Must generate the message while holding the lock so that 'ifma'
3060 	 * pointer is still valid.
3061 	 */
3062 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3063 	IF_ADDR_WUNLOCK(ifp);
3064 
3065 	/*
3066 	 * We are certain we have added something, so call down to the
3067 	 * interface to let them know about it.
3068 	 */
3069 	if (ifp->if_ioctl != NULL) {
3070 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3071 	}
3072 
3073 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3074 		link_free_sdl(llsa);
3075 
3076 	return (0);
3077 
3078 free_llsa_out:
3079 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3080 		link_free_sdl(llsa);
3081 
3082 unlock_out:
3083 	IF_ADDR_WUNLOCK(ifp);
3084 	return (error);
3085 }
3086 
3087 /*
3088  * Delete a multicast group membership by network-layer group address.
3089  *
3090  * Returns ENOENT if the entry could not be found. If ifp no longer
3091  * exists, results are undefined. This entry point should only be used
3092  * from subsystems which do appropriate locking to hold ifp for the
3093  * duration of the call.
3094  * Network-layer protocol domains must use if_delmulti_ifma().
3095  */
3096 int
3097 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3098 {
3099 	struct ifmultiaddr *ifma;
3100 	int lastref;
3101 #ifdef INVARIANTS
3102 	struct ifnet *oifp;
3103 
3104 	IFNET_RLOCK_NOSLEEP();
3105 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3106 		if (ifp == oifp)
3107 			break;
3108 	if (ifp != oifp)
3109 		ifp = NULL;
3110 	IFNET_RUNLOCK_NOSLEEP();
3111 
3112 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3113 #endif
3114 	if (ifp == NULL)
3115 		return (ENOENT);
3116 
3117 	IF_ADDR_WLOCK(ifp);
3118 	lastref = 0;
3119 	ifma = if_findmulti(ifp, sa);
3120 	if (ifma != NULL)
3121 		lastref = if_delmulti_locked(ifp, ifma, 0);
3122 	IF_ADDR_WUNLOCK(ifp);
3123 
3124 	if (ifma == NULL)
3125 		return (ENOENT);
3126 
3127 	if (lastref && ifp->if_ioctl != NULL) {
3128 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3129 	}
3130 
3131 	return (0);
3132 }
3133 
3134 /*
3135  * Delete all multicast group membership for an interface.
3136  * Should be used to quickly flush all multicast filters.
3137  */
3138 void
3139 if_delallmulti(struct ifnet *ifp)
3140 {
3141 	struct ifmultiaddr *ifma;
3142 	struct ifmultiaddr *next;
3143 
3144 	IF_ADDR_WLOCK(ifp);
3145 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3146 		if_delmulti_locked(ifp, ifma, 0);
3147 	IF_ADDR_WUNLOCK(ifp);
3148 }
3149 
3150 /*
3151  * Delete a multicast group membership by group membership pointer.
3152  * Network-layer protocol domains must use this routine.
3153  *
3154  * It is safe to call this routine if the ifp disappeared.
3155  */
3156 void
3157 if_delmulti_ifma(struct ifmultiaddr *ifma)
3158 {
3159 	struct ifnet *ifp;
3160 	int lastref;
3161 
3162 	ifp = ifma->ifma_ifp;
3163 #ifdef DIAGNOSTIC
3164 	if (ifp == NULL) {
3165 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3166 	} else {
3167 		struct ifnet *oifp;
3168 
3169 		IFNET_RLOCK_NOSLEEP();
3170 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3171 			if (ifp == oifp)
3172 				break;
3173 		if (ifp != oifp) {
3174 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3175 			ifp = NULL;
3176 		}
3177 		IFNET_RUNLOCK_NOSLEEP();
3178 	}
3179 #endif
3180 	/*
3181 	 * If and only if the ifnet instance exists: Acquire the address lock.
3182 	 */
3183 	if (ifp != NULL)
3184 		IF_ADDR_WLOCK(ifp);
3185 
3186 	lastref = if_delmulti_locked(ifp, ifma, 0);
3187 
3188 	if (ifp != NULL) {
3189 		/*
3190 		 * If and only if the ifnet instance exists:
3191 		 *  Release the address lock.
3192 		 *  If the group was left: update the hardware hash filter.
3193 		 */
3194 		IF_ADDR_WUNLOCK(ifp);
3195 		if (lastref && ifp->if_ioctl != NULL) {
3196 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3197 		}
3198 	}
3199 }
3200 
3201 /*
3202  * Perform deletion of network-layer and/or link-layer multicast address.
3203  *
3204  * Return 0 if the reference count was decremented.
3205  * Return 1 if the final reference was released, indicating that the
3206  * hardware hash filter should be reprogrammed.
3207  */
3208 static int
3209 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3210 {
3211 	struct ifmultiaddr *ll_ifma;
3212 
3213 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3214 		KASSERT(ifma->ifma_ifp == ifp,
3215 		    ("%s: inconsistent ifp %p", __func__, ifp));
3216 		IF_ADDR_WLOCK_ASSERT(ifp);
3217 	}
3218 
3219 	ifp = ifma->ifma_ifp;
3220 
3221 	/*
3222 	 * If the ifnet is detaching, null out references to ifnet,
3223 	 * so that upper protocol layers will notice, and not attempt
3224 	 * to obtain locks for an ifnet which no longer exists. The
3225 	 * routing socket announcement must happen before the ifnet
3226 	 * instance is detached from the system.
3227 	 */
3228 	if (detaching) {
3229 #ifdef DIAGNOSTIC
3230 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3231 #endif
3232 		/*
3233 		 * ifp may already be nulled out if we are being reentered
3234 		 * to delete the ll_ifma.
3235 		 */
3236 		if (ifp != NULL) {
3237 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3238 			ifma->ifma_ifp = NULL;
3239 		}
3240 	}
3241 
3242 	if (--ifma->ifma_refcount > 0)
3243 		return 0;
3244 
3245 	/*
3246 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3247 	 * have been associated with it. Release it first if so.
3248 	 */
3249 	ll_ifma = ifma->ifma_llifma;
3250 	if (ll_ifma != NULL) {
3251 		KASSERT(ifma->ifma_lladdr != NULL,
3252 		    ("%s: llifma w/o lladdr", __func__));
3253 		if (detaching)
3254 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3255 		if (--ll_ifma->ifma_refcount == 0) {
3256 			if (ifp != NULL) {
3257 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3258 				    ifma_link);
3259 			}
3260 			if_freemulti(ll_ifma);
3261 		}
3262 	}
3263 
3264 	if (ifp != NULL)
3265 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3266 
3267 	if_freemulti(ifma);
3268 
3269 	/*
3270 	 * The last reference to this instance of struct ifmultiaddr
3271 	 * was released; the hardware should be notified of this change.
3272 	 */
3273 	return 1;
3274 }
3275 
3276 /*
3277  * Set the link layer address on an interface.
3278  *
3279  * At this time we only support certain types of interfaces,
3280  * and we don't allow the length of the address to change.
3281  */
3282 int
3283 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3284 {
3285 	struct sockaddr_dl *sdl;
3286 	struct ifaddr *ifa;
3287 	struct ifreq ifr;
3288 
3289 	IF_ADDR_RLOCK(ifp);
3290 	ifa = ifp->if_addr;
3291 	if (ifa == NULL) {
3292 		IF_ADDR_RUNLOCK(ifp);
3293 		return (EINVAL);
3294 	}
3295 	ifa_ref(ifa);
3296 	IF_ADDR_RUNLOCK(ifp);
3297 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3298 	if (sdl == NULL) {
3299 		ifa_free(ifa);
3300 		return (EINVAL);
3301 	}
3302 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3303 		ifa_free(ifa);
3304 		return (EINVAL);
3305 	}
3306 	switch (ifp->if_type) {
3307 	case IFT_ETHER:
3308 	case IFT_FDDI:
3309 	case IFT_XETHER:
3310 	case IFT_ISO88025:
3311 	case IFT_L2VLAN:
3312 	case IFT_BRIDGE:
3313 	case IFT_ARCNET:
3314 	case IFT_IEEE8023ADLAG:
3315 	case IFT_IEEE80211:
3316 		bcopy(lladdr, LLADDR(sdl), len);
3317 		ifa_free(ifa);
3318 		break;
3319 	default:
3320 		ifa_free(ifa);
3321 		return (ENODEV);
3322 	}
3323 
3324 	/*
3325 	 * If the interface is already up, we need
3326 	 * to re-init it in order to reprogram its
3327 	 * address filter.
3328 	 */
3329 	if ((ifp->if_flags & IFF_UP) != 0) {
3330 		if (ifp->if_ioctl) {
3331 			ifp->if_flags &= ~IFF_UP;
3332 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3333 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3334 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3335 			ifp->if_flags |= IFF_UP;
3336 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3337 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3338 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3339 		}
3340 #ifdef INET
3341 		/*
3342 		 * Also send gratuitous ARPs to notify other nodes about
3343 		 * the address change.
3344 		 */
3345 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3346 			if (ifa->ifa_addr->sa_family == AF_INET)
3347 				arp_ifinit(ifp, ifa);
3348 		}
3349 #endif
3350 	}
3351 	return (0);
3352 }
3353 
3354 /*
3355  * The name argument must be a pointer to storage which will last as
3356  * long as the interface does.  For physical devices, the result of
3357  * device_get_name(dev) is a good choice and for pseudo-devices a
3358  * static string works well.
3359  */
3360 void
3361 if_initname(struct ifnet *ifp, const char *name, int unit)
3362 {
3363 	ifp->if_dname = name;
3364 	ifp->if_dunit = unit;
3365 	if (unit != IF_DUNIT_NONE)
3366 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3367 	else
3368 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3369 }
3370 
3371 int
3372 if_printf(struct ifnet *ifp, const char * fmt, ...)
3373 {
3374 	va_list ap;
3375 	int retval;
3376 
3377 	retval = printf("%s: ", ifp->if_xname);
3378 	va_start(ap, fmt);
3379 	retval += vprintf(fmt, ap);
3380 	va_end(ap);
3381 	return (retval);
3382 }
3383 
3384 void
3385 if_start(struct ifnet *ifp)
3386 {
3387 
3388 	(*(ifp)->if_start)(ifp);
3389 }
3390 
3391 /*
3392  * Backwards compatibility interface for drivers
3393  * that have not implemented it
3394  */
3395 static int
3396 if_transmit(struct ifnet *ifp, struct mbuf *m)
3397 {
3398 	int error;
3399 
3400 	IFQ_HANDOFF(ifp, m, error);
3401 	return (error);
3402 }
3403 
3404 int
3405 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3406 {
3407 	int active = 0;
3408 
3409 	IF_LOCK(ifq);
3410 	if (_IF_QFULL(ifq)) {
3411 		_IF_DROP(ifq);
3412 		IF_UNLOCK(ifq);
3413 		m_freem(m);
3414 		return (0);
3415 	}
3416 	if (ifp != NULL) {
3417 		ifp->if_obytes += m->m_pkthdr.len + adjust;
3418 		if (m->m_flags & (M_BCAST|M_MCAST))
3419 			ifp->if_omcasts++;
3420 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3421 	}
3422 	_IF_ENQUEUE(ifq, m);
3423 	IF_UNLOCK(ifq);
3424 	if (ifp != NULL && !active)
3425 		(*(ifp)->if_start)(ifp);
3426 	return (1);
3427 }
3428 
3429 void
3430 if_register_com_alloc(u_char type,
3431     if_com_alloc_t *a, if_com_free_t *f)
3432 {
3433 
3434 	KASSERT(if_com_alloc[type] == NULL,
3435 	    ("if_register_com_alloc: %d already registered", type));
3436 	KASSERT(if_com_free[type] == NULL,
3437 	    ("if_register_com_alloc: %d free already registered", type));
3438 
3439 	if_com_alloc[type] = a;
3440 	if_com_free[type] = f;
3441 }
3442 
3443 void
3444 if_deregister_com_alloc(u_char type)
3445 {
3446 
3447 	KASSERT(if_com_alloc[type] != NULL,
3448 	    ("if_deregister_com_alloc: %d not registered", type));
3449 	KASSERT(if_com_free[type] != NULL,
3450 	    ("if_deregister_com_alloc: %d free not registered", type));
3451 	if_com_alloc[type] = NULL;
3452 	if_com_free[type] = NULL;
3453 }
3454 
3455 /* API for driver access to network stack owned ifnet.*/
3456 uint64_t
3457 if_setbaudrate(void *arg, uint64_t baudrate)
3458 {
3459 	struct ifnet *ifp = arg;
3460 	uint64_t oldbrate;
3461 
3462 	oldbrate = ifp->if_baudrate;
3463 	ifp->if_baudrate = baudrate;
3464 	return (oldbrate);
3465 }
3466 
3467 uint64_t
3468 if_getbaudrate(if_t ifp)
3469 {
3470 
3471 	return (((struct ifnet *)ifp)->if_baudrate);
3472 }
3473 
3474 int
3475 if_setcapabilities(if_t ifp, int capabilities)
3476 {
3477 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3478 	return (0);
3479 }
3480 
3481 int
3482 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3483 {
3484 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3485 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3486 
3487 	return (0);
3488 }
3489 
3490 int
3491 if_getcapabilities(if_t ifp)
3492 {
3493 	return ((struct ifnet *)ifp)->if_capabilities;
3494 }
3495 
3496 int
3497 if_setcapenable(if_t ifp, int capabilities)
3498 {
3499 	((struct ifnet *)ifp)->if_capenable = capabilities;
3500 	return (0);
3501 }
3502 
3503 int
3504 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3505 {
3506 	if(setcap)
3507 		((struct ifnet *)ifp)->if_capenable |= setcap;
3508 	if(clearcap)
3509 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3510 
3511 	return (0);
3512 }
3513 
3514 const char *
3515 if_getdname(if_t ifp)
3516 {
3517 	return ((struct ifnet *)ifp)->if_dname;
3518 }
3519 
3520 int
3521 if_togglecapenable(if_t ifp, int togglecap)
3522 {
3523 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3524 	return (0);
3525 }
3526 
3527 int
3528 if_getcapenable(if_t ifp)
3529 {
3530 	return ((struct ifnet *)ifp)->if_capenable;
3531 }
3532 
3533 /*
3534  * This is largely undesirable because it ties ifnet to a device, but does
3535  * provide flexiblity for an embedded product vendor. Should be used with
3536  * the understanding that it violates the interface boundaries, and should be
3537  * a last resort only.
3538  */
3539 int
3540 if_setdev(if_t ifp, void *dev)
3541 {
3542 	return (0);
3543 }
3544 
3545 int
3546 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3547 {
3548 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3549 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3550 
3551 	return (0);
3552 }
3553 
3554 int
3555 if_getdrvflags(if_t ifp)
3556 {
3557 	return ((struct ifnet *)ifp)->if_drv_flags;
3558 }
3559 
3560 int
3561 if_setdrvflags(if_t ifp, int flags)
3562 {
3563 	((struct ifnet *)ifp)->if_drv_flags = flags;
3564 	return (0);
3565 }
3566 
3567 
3568 int
3569 if_setflags(if_t ifp, int flags)
3570 {
3571 	((struct ifnet *)ifp)->if_flags = flags;
3572 	return (0);
3573 }
3574 
3575 int
3576 if_setflagbits(if_t ifp, int set, int clear)
3577 {
3578 	((struct ifnet *)ifp)->if_flags |= set;
3579 	((struct ifnet *)ifp)->if_flags &= ~clear;
3580 
3581 	return (0);
3582 }
3583 
3584 int
3585 if_getflags(if_t ifp)
3586 {
3587 	return ((struct ifnet *)ifp)->if_flags;
3588 }
3589 
3590 int
3591 if_clearhwassist(if_t ifp)
3592 {
3593 	((struct ifnet *)ifp)->if_hwassist = 0;
3594 	return (0);
3595 }
3596 
3597 int
3598 if_sethwassistbits(if_t ifp, int toset, int toclear)
3599 {
3600 	((struct ifnet *)ifp)->if_hwassist |= toset;
3601 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3602 
3603 	return (0);
3604 }
3605 
3606 int
3607 if_sethwassist(if_t ifp, int hwassist_bit)
3608 {
3609 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3610 	return (0);
3611 }
3612 
3613 int
3614 if_gethwassist(if_t ifp)
3615 {
3616 	return ((struct ifnet *)ifp)->if_hwassist;
3617 }
3618 
3619 int
3620 if_setmtu(if_t ifp, int mtu)
3621 {
3622 	((struct ifnet *)ifp)->if_mtu = mtu;
3623 	return (0);
3624 }
3625 
3626 int
3627 if_getmtu(if_t ifp)
3628 {
3629 	return ((struct ifnet *)ifp)->if_mtu;
3630 }
3631 
3632 int
3633 if_setsoftc(if_t ifp, void *softc)
3634 {
3635 	((struct ifnet *)ifp)->if_softc = softc;
3636 	return (0);
3637 }
3638 
3639 void *
3640 if_getsoftc(if_t ifp)
3641 {
3642 	return ((struct ifnet *)ifp)->if_softc;
3643 }
3644 
3645 void
3646 if_setrcvif(struct mbuf *m, if_t ifp)
3647 {
3648 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3649 }
3650 
3651 void
3652 if_setvtag(struct mbuf *m, uint16_t tag)
3653 {
3654 	m->m_pkthdr.ether_vtag = tag;
3655 }
3656 
3657 uint16_t
3658 if_getvtag(struct mbuf *m)
3659 {
3660 
3661 	return (m->m_pkthdr.ether_vtag);
3662 }
3663 
3664 /* Statistics */
3665 int
3666 if_incipackets(if_t ifp, int pkts)
3667 {
3668 	((struct ifnet *)ifp)->if_ipackets += pkts;
3669 	return (0);
3670 }
3671 
3672 int
3673 if_incopackets(if_t ifp, int pkts)
3674 {
3675 	((struct ifnet *)ifp)->if_opackets += pkts;
3676 	return (0);
3677 }
3678 
3679 int
3680 if_incierrors(if_t ifp, int ierrors)
3681 {
3682 	((struct ifnet *)ifp)->if_ierrors += ierrors;
3683 	return (0);
3684 }
3685 
3686 
3687 int
3688 if_setierrors(if_t ifp, int ierrors)
3689 {
3690 	((struct ifnet *)ifp)->if_ierrors = ierrors;
3691 	return (0);
3692 }
3693 
3694 int
3695 if_setoerrors(if_t ifp, int oerrors)
3696 {
3697 	((struct ifnet *)ifp)->if_oerrors = oerrors;
3698 	return (0);
3699 }
3700 
3701 int if_incoerrors(if_t ifp, int oerrors)
3702 {
3703 	((struct ifnet *)ifp)->if_oerrors += oerrors;
3704 	return (0);
3705 }
3706 
3707 int if_inciqdrops(if_t ifp, int val)
3708 {
3709 	((struct ifnet *)ifp)->if_iqdrops += val;
3710 	return (0);
3711 }
3712 
3713 int
3714 if_setcollisions(if_t ifp, int collisions)
3715 {
3716 	((struct ifnet *)ifp)->if_collisions = collisions;
3717 	return (0);
3718 }
3719 
3720 int
3721 if_inccollisions(if_t ifp, int collisions)
3722 {
3723 	((struct ifnet *)ifp)->if_collisions += collisions;
3724 	return (0);
3725 }
3726 
3727 int
3728 if_setipackets(if_t ifp, int pkts)
3729 {
3730 	((struct ifnet *)ifp)->if_ipackets = pkts;
3731 	return (0);
3732 }
3733 
3734 int
3735 if_setopackets(if_t ifp, int pkts)
3736 {
3737 	((struct ifnet *)ifp)->if_opackets = pkts;
3738 	return (0);
3739 }
3740 
3741 int
3742 if_incobytes(if_t ifp, int bytes)
3743 {
3744 	((struct ifnet *)ifp)->if_obytes += bytes;
3745 	return (0);
3746 }
3747 
3748 int
3749 if_setibytes(if_t ifp, int bytes)
3750 {
3751 	((struct ifnet *)ifp)->if_ibytes = bytes;
3752 	return (0);
3753 }
3754 
3755 int
3756 if_setobytes(if_t ifp, int bytes)
3757 {
3758 	((struct ifnet *)ifp)->if_obytes = bytes;
3759 	return (0);
3760 }
3761 
3762 
3763 int
3764 if_sendq_empty(if_t ifp)
3765 {
3766 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3767 }
3768 
3769 int if_getiqdrops(if_t ifp)
3770 {
3771 	return ((struct ifnet *)ifp)->if_iqdrops;
3772 }
3773 
3774 int
3775 if_incimcasts(if_t ifp, int mcast)
3776 {
3777 	((struct ifnet *)ifp)->if_imcasts += mcast;
3778 	return (0);
3779 }
3780 
3781 
3782 int
3783 if_incomcasts(if_t ifp, int mcast)
3784 {
3785 	((struct ifnet *)ifp)->if_omcasts += mcast;
3786 	return (0);
3787 }
3788 
3789 int
3790 if_setimcasts(if_t ifp, int mcast)
3791 {
3792 	((struct ifnet *)ifp)->if_imcasts = mcast;
3793 	return (0);
3794 }
3795 
3796 
3797 struct ifaddr *
3798 if_getifaddr(if_t ifp)
3799 {
3800 	return ((struct ifnet *)ifp)->if_addr;
3801 }
3802 
3803 int
3804 if_getamcount(if_t ifp)
3805 {
3806 	return ((struct ifnet *)ifp)->if_amcount;
3807 }
3808 
3809 
3810 int
3811 if_setsendqready(if_t ifp)
3812 {
3813 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3814 	return (0);
3815 }
3816 
3817 int
3818 if_setsendqlen(if_t ifp, int tx_desc_count)
3819 {
3820 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3821 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3822 
3823 	return (0);
3824 }
3825 
3826 int
3827 if_vlantrunkinuse(if_t ifp)
3828 {
3829 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3830 }
3831 
3832 int
3833 if_input(if_t ifp, struct mbuf* sendmp)
3834 {
3835 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3836 	return (0);
3837 
3838 }
3839 
3840 /* XXX */
3841 #ifndef ETH_ADDR_LEN
3842 #define ETH_ADDR_LEN 6
3843 #endif
3844 
3845 int
3846 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3847 {
3848 	struct ifmultiaddr *ifma;
3849 	uint8_t *lmta = (uint8_t *)mta;
3850 	int mcnt = 0;
3851 
3852 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3853 		if (ifma->ifma_addr->sa_family != AF_LINK)
3854 			continue;
3855 
3856 		if (mcnt == max)
3857 			break;
3858 
3859 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3860 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3861 		mcnt++;
3862 	}
3863 	*cnt = mcnt;
3864 
3865 	return (0);
3866 }
3867 
3868 int
3869 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3870 {
3871 	int error;
3872 
3873 	if_maddr_rlock(ifp);
3874 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3875 	if_maddr_runlock(ifp);
3876 	return (error);
3877 }
3878 
3879 int
3880 if_multiaddr_count(if_t ifp, int max)
3881 {
3882 	struct ifmultiaddr *ifma;
3883 	int count;
3884 
3885 	count = 0;
3886 	if_maddr_rlock(ifp);
3887 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3888 		if (ifma->ifma_addr->sa_family != AF_LINK)
3889 			continue;
3890 		count++;
3891 		if (count == max)
3892 			break;
3893 	}
3894 	if_maddr_runlock(ifp);
3895 	return (count);
3896 }
3897 
3898 struct mbuf *
3899 if_dequeue(if_t ifp)
3900 {
3901 	struct mbuf *m;
3902 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3903 
3904 	return (m);
3905 }
3906 
3907 int
3908 if_sendq_prepend(if_t ifp, struct mbuf *m)
3909 {
3910 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3911 	return (0);
3912 }
3913 
3914 int
3915 if_setifheaderlen(if_t ifp, int len)
3916 {
3917 	((struct ifnet *)ifp)->if_data.ifi_hdrlen = len;
3918 	return (0);
3919 }
3920 
3921 caddr_t
3922 if_getlladdr(if_t ifp)
3923 {
3924 	return (IF_LLADDR((struct ifnet *)ifp));
3925 }
3926 
3927 void *
3928 if_gethandle(u_char type)
3929 {
3930 	return (if_alloc(type));
3931 }
3932 
3933 void
3934 if_bpfmtap(if_t ifh, struct mbuf *m)
3935 {
3936 	struct ifnet *ifp = (struct ifnet *)ifh;
3937 
3938 	BPF_MTAP(ifp, m);
3939 }
3940 
3941 void
3942 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3943 {
3944 	struct ifnet *ifp = (struct ifnet *)ifh;
3945 
3946 	ETHER_BPF_MTAP(ifp, m);
3947 }
3948 
3949 void
3950 if_vlancap(if_t ifh)
3951 {
3952 	struct ifnet *ifp = (struct ifnet *)ifh;
3953 	VLAN_CAPABILITIES(ifp);
3954 }
3955 
3956 void
3957 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3958 {
3959 	((struct ifnet *)ifp)->if_init = init_fn;
3960 }
3961 
3962 void
3963 if_setioctlfn(if_t ifp, int (*ioctl_fn)(void *, u_long, caddr_t))
3964 {
3965 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3966 }
3967 
3968 void
3969 if_setstartfn(if_t ifp, void (*start_fn)(void *))
3970 {
3971 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
3972 }
3973 
3974 void
3975 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3976 {
3977 	((struct ifnet *)ifp)->if_transmit = start_fn;
3978 }
3979 
3980 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3981 {
3982 	((struct ifnet *)ifp)->if_qflush = flush_fn;
3983 
3984 }
3985 
3986 /* These wrappers are hopefully temporary, till all drivers use drvapi */
3987 #ifdef INET
3988 void
3989 arp_ifinit_drv(if_t ifh, struct ifaddr *ifa)
3990 {
3991 	arp_ifinit((struct ifnet *)ifh, ifa);
3992 }
3993 #endif
3994 
3995 void
3996 ether_ifattach_drv(if_t ifh, const u_int8_t *lla)
3997 {
3998 	ether_ifattach((struct ifnet *)ifh, lla);
3999 }
4000 
4001 void
4002 ether_ifdetach_drv(if_t ifh)
4003 {
4004 	ether_ifdetach((struct ifnet *)ifh);
4005 }
4006 
4007 int
4008 ether_ioctl_drv(if_t ifh, u_long cmd, caddr_t data)
4009 {
4010 	struct ifnet *ifp = (struct ifnet *)ifh;
4011 
4012 	return (ether_ioctl(ifp, cmd, data));
4013 }
4014 
4015 int
4016 ifmedia_ioctl_drv(if_t ifh, struct ifreq *ifr, struct ifmedia *ifm,
4017     u_long cmd)
4018 {
4019 	struct ifnet *ifp = (struct ifnet *)ifh;
4020 
4021 	return (ifmedia_ioctl(ifp, ifr, ifm, cmd));
4022 }
4023 
4024 void
4025 if_free_drv(if_t ifh)
4026 {
4027 	if_free((struct ifnet *)ifh);
4028 }
4029 
4030 void
4031 if_initname_drv(if_t ifh, const char *name, int unit)
4032 {
4033 	if_initname((struct ifnet *)ifh, name, unit);
4034 }
4035 
4036 void
4037 if_linkstate_change_drv(if_t ifh, int link_state)
4038 {
4039 	if_link_state_change((struct ifnet *)ifh, link_state);
4040 }
4041 
4042 void
4043 ifmedia_init_drv(struct ifmedia *ifm, int ncmask, int (*chg_cb)(void *),
4044     void (*sts_cb)(void *, struct ifmediareq *))
4045 {
4046 	ifmedia_init(ifm, ncmask, (ifm_change_cb_t)chg_cb,
4047 	    (ifm_stat_cb_t)sts_cb);
4048 }
4049 
4050 void
4051 if_addr_rlock_drv(if_t ifh)
4052 {
4053 
4054 	if_addr_runlock((struct ifnet *)ifh);
4055 }
4056 
4057 void
4058 if_addr_runlock_drv(if_t ifh)
4059 {
4060 	if_addr_runlock((struct ifnet *)ifh);
4061 }
4062 
4063 void
4064 if_qflush_drv(if_t ifh)
4065 {
4066 	if_qflush((struct ifnet *)ifh);
4067 
4068 }
4069 
4070 /* Revisit these - These are inline functions originally. */
4071 int
4072 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4073 {
4074 	return drbr_inuse_drv(ifh, br);
4075 }
4076 
4077 struct mbuf*
4078 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4079 {
4080 	return drbr_dequeue(ifh, br);
4081 }
4082 
4083 int
4084 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4085 {
4086 	return drbr_needs_enqueue(ifh, br);
4087 }
4088 
4089 int
4090 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4091 {
4092 	return drbr_enqueue(ifh, br, m);
4093 
4094 }
4095