xref: /freebsd/sys/net/if.c (revision cef367e6a15b6e3e2aca88d453c74962cf1f94e7)
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 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
104 
105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106     &ifqmaxlen, 0, "max send queue size");
107 
108 /* Log link state change events */
109 static int log_link_state_change = 1;
110 
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 	&log_link_state_change, 0,
113 	"log interface link state change events");
114 
115 /* Log promiscuous mode change events */
116 static int log_promisc_mode_change = 1;
117 
118 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RW,
119 	&log_promisc_mode_change, 1,
120 	"log promiscuous mode change events");
121 
122 /* Interface description */
123 static unsigned int ifdescr_maxlen = 1024;
124 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
125 	&ifdescr_maxlen, 0,
126 	"administrative maximum length for interface description");
127 
128 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
129 
130 /* global sx for non-critical path ifdescr */
131 static struct sx ifdescr_sx;
132 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
133 
134 void	(*bridge_linkstate_p)(struct ifnet *ifp);
135 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
136 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
137 /* These are external hooks for CARP. */
138 void	(*carp_linkstate_p)(struct ifnet *ifp);
139 void	(*carp_demote_adj_p)(int, char *);
140 int	(*carp_master_p)(struct ifaddr *);
141 #if defined(INET) || defined(INET6)
142 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
143 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
144     const struct sockaddr *sa);
145 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
146 int	(*carp_attach_p)(struct ifaddr *, int);
147 void	(*carp_detach_p)(struct ifaddr *);
148 #endif
149 #ifdef INET
150 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
151 #endif
152 #ifdef INET6
153 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
154 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
155     const struct in6_addr *taddr);
156 #endif
157 
158 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
159 
160 /*
161  * XXX: Style; these should be sorted alphabetically, and unprototyped
162  * static functions should be prototyped. Currently they are sorted by
163  * declaration order.
164  */
165 static void	if_attachdomain(void *);
166 static void	if_attachdomain1(struct ifnet *);
167 static int	ifconf(u_long, caddr_t);
168 static void	if_freemulti(struct ifmultiaddr *);
169 static void	if_grow(void);
170 static void	if_input_default(struct ifnet *, struct mbuf *);
171 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
172 static void	if_route(struct ifnet *, int flag, int fam);
173 static int	if_setflag(struct ifnet *, int, int, int *, int);
174 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
175 static void	if_unroute(struct ifnet *, int flag, int fam);
176 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
177 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
178 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
179 static void	do_link_state_change(void *, int);
180 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
181 static int	if_getgroupmembers(struct ifgroupreq *);
182 static void	if_delgroups(struct ifnet *);
183 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
184 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
185 
186 #ifdef INET6
187 /*
188  * XXX: declare here to avoid to include many inet6 related files..
189  * should be more generalized?
190  */
191 extern void	nd6_setmtu(struct ifnet *);
192 #endif
193 
194 /* ipsec helper hooks */
195 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
196 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
197 
198 VNET_DEFINE(int, if_index);
199 int	ifqmaxlen = IFQ_MAXLEN;
200 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
201 VNET_DEFINE(struct ifgrouphead, ifg_head);
202 
203 static VNET_DEFINE(int, if_indexlim) = 8;
204 
205 /* Table of ifnet by index. */
206 VNET_DEFINE(struct ifnet **, ifindex_table);
207 
208 #define	V_if_indexlim		VNET(if_indexlim)
209 #define	V_ifindex_table		VNET(ifindex_table)
210 
211 /*
212  * The global network interface list (V_ifnet) and related state (such as
213  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
214  * an rwlock.  Either may be acquired shared to stablize the list, but both
215  * must be acquired writable to modify the list.  This model allows us to
216  * both stablize the interface list during interrupt thread processing, but
217  * also to stablize it over long-running ioctls, without introducing priority
218  * inversions and deadlocks.
219  */
220 struct rwlock ifnet_rwlock;
221 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
222 struct sx ifnet_sxlock;
223 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
224 
225 /*
226  * The allocation of network interfaces is a rather non-atomic affair; we
227  * need to select an index before we are ready to expose the interface for
228  * use, so will use this pointer value to indicate reservation.
229  */
230 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
231 
232 static	if_com_alloc_t *if_com_alloc[256];
233 static	if_com_free_t *if_com_free[256];
234 
235 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
236 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
237 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
238 
239 struct ifnet *
240 ifnet_byindex_locked(u_short idx)
241 {
242 
243 	if (idx > V_if_index)
244 		return (NULL);
245 	if (V_ifindex_table[idx] == IFNET_HOLD)
246 		return (NULL);
247 	return (V_ifindex_table[idx]);
248 }
249 
250 struct ifnet *
251 ifnet_byindex(u_short idx)
252 {
253 	struct ifnet *ifp;
254 
255 	IFNET_RLOCK_NOSLEEP();
256 	ifp = ifnet_byindex_locked(idx);
257 	IFNET_RUNLOCK_NOSLEEP();
258 	return (ifp);
259 }
260 
261 struct ifnet *
262 ifnet_byindex_ref(u_short idx)
263 {
264 	struct ifnet *ifp;
265 
266 	IFNET_RLOCK_NOSLEEP();
267 	ifp = ifnet_byindex_locked(idx);
268 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
269 		IFNET_RUNLOCK_NOSLEEP();
270 		return (NULL);
271 	}
272 	if_ref(ifp);
273 	IFNET_RUNLOCK_NOSLEEP();
274 	return (ifp);
275 }
276 
277 /*
278  * Allocate an ifindex array entry; return 0 on success or an error on
279  * failure.
280  */
281 static u_short
282 ifindex_alloc(void)
283 {
284 	u_short idx;
285 
286 	IFNET_WLOCK_ASSERT();
287 retry:
288 	/*
289 	 * Try to find an empty slot below V_if_index.  If we fail, take the
290 	 * next slot.
291 	 */
292 	for (idx = 1; idx <= V_if_index; idx++) {
293 		if (V_ifindex_table[idx] == NULL)
294 			break;
295 	}
296 
297 	/* Catch if_index overflow. */
298 	if (idx >= V_if_indexlim) {
299 		if_grow();
300 		goto retry;
301 	}
302 	if (idx > V_if_index)
303 		V_if_index = idx;
304 	return (idx);
305 }
306 
307 static void
308 ifindex_free_locked(u_short idx)
309 {
310 
311 	IFNET_WLOCK_ASSERT();
312 
313 	V_ifindex_table[idx] = NULL;
314 	while (V_if_index > 0 &&
315 	    V_ifindex_table[V_if_index] == NULL)
316 		V_if_index--;
317 }
318 
319 static void
320 ifindex_free(u_short idx)
321 {
322 
323 	IFNET_WLOCK();
324 	ifindex_free_locked(idx);
325 	IFNET_WUNLOCK();
326 }
327 
328 static void
329 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
330 {
331 
332 	IFNET_WLOCK_ASSERT();
333 
334 	V_ifindex_table[idx] = ifp;
335 }
336 
337 static void
338 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
339 {
340 
341 	IFNET_WLOCK();
342 	ifnet_setbyindex_locked(idx, ifp);
343 	IFNET_WUNLOCK();
344 }
345 
346 struct ifaddr *
347 ifaddr_byindex(u_short idx)
348 {
349 	struct ifnet *ifp;
350 	struct ifaddr *ifa = NULL;
351 
352 	IFNET_RLOCK_NOSLEEP();
353 	ifp = ifnet_byindex_locked(idx);
354 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
355 		ifa_ref(ifa);
356 	IFNET_RUNLOCK_NOSLEEP();
357 	return (ifa);
358 }
359 
360 /*
361  * Network interface utility routines.
362  *
363  * Routines with ifa_ifwith* names take sockaddr *'s as
364  * parameters.
365  */
366 
367 static void
368 vnet_if_init(const void *unused __unused)
369 {
370 
371 	TAILQ_INIT(&V_ifnet);
372 	TAILQ_INIT(&V_ifg_head);
373 	IFNET_WLOCK();
374 	if_grow();				/* create initial table */
375 	IFNET_WUNLOCK();
376 	vnet_if_clone_init();
377 }
378 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
379     NULL);
380 
381 #ifdef VIMAGE
382 static void
383 vnet_if_uninit(const void *unused __unused)
384 {
385 
386 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
387 	    "not empty", __func__, __LINE__, &V_ifnet));
388 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
389 	    "not empty", __func__, __LINE__, &V_ifg_head));
390 
391 	free((caddr_t)V_ifindex_table, M_IFNET);
392 }
393 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
394     vnet_if_uninit, NULL);
395 #endif
396 
397 static void
398 if_grow(void)
399 {
400 	int oldlim;
401 	u_int n;
402 	struct ifnet **e;
403 
404 	IFNET_WLOCK_ASSERT();
405 	oldlim = V_if_indexlim;
406 	IFNET_WUNLOCK();
407 	n = (oldlim << 1) * sizeof(*e);
408 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
409 	IFNET_WLOCK();
410 	if (V_if_indexlim != oldlim) {
411 		free(e, M_IFNET);
412 		return;
413 	}
414 	if (V_ifindex_table != NULL) {
415 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
416 		free((caddr_t)V_ifindex_table, M_IFNET);
417 	}
418 	V_if_indexlim <<= 1;
419 	V_ifindex_table = e;
420 }
421 
422 /*
423  * Allocate a struct ifnet and an index for an interface.  A layer 2
424  * common structure will also be allocated if an allocation routine is
425  * registered for the passed type.
426  */
427 struct ifnet *
428 if_alloc(u_char type)
429 {
430 	struct ifnet *ifp;
431 	u_short idx;
432 
433 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
434 	IFNET_WLOCK();
435 	idx = ifindex_alloc();
436 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
437 	IFNET_WUNLOCK();
438 	ifp->if_index = idx;
439 	ifp->if_type = type;
440 	ifp->if_alloctype = type;
441 	if (if_com_alloc[type] != NULL) {
442 		ifp->if_l2com = if_com_alloc[type](type, ifp);
443 		if (ifp->if_l2com == NULL) {
444 			free(ifp, M_IFNET);
445 			ifindex_free(idx);
446 			return (NULL);
447 		}
448 	}
449 
450 	IF_ADDR_LOCK_INIT(ifp);
451 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
452 	ifp->if_afdata_initialized = 0;
453 	IF_AFDATA_LOCK_INIT(ifp);
454 	TAILQ_INIT(&ifp->if_addrhead);
455 	TAILQ_INIT(&ifp->if_multiaddrs);
456 	TAILQ_INIT(&ifp->if_groups);
457 #ifdef MAC
458 	mac_ifnet_init(ifp);
459 #endif
460 	ifq_init(&ifp->if_snd, ifp);
461 
462 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
463 	for (int i = 0; i < IFCOUNTERS; i++)
464 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
465 	ifp->if_get_counter = if_get_counter_default;
466 	ifnet_setbyindex(ifp->if_index, ifp);
467 	return (ifp);
468 }
469 
470 /*
471  * Do the actual work of freeing a struct ifnet, and layer 2 common
472  * structure.  This call is made when the last reference to an
473  * interface is released.
474  */
475 static void
476 if_free_internal(struct ifnet *ifp)
477 {
478 
479 	KASSERT((ifp->if_flags & IFF_DYING),
480 	    ("if_free_internal: interface not dying"));
481 
482 	if (if_com_free[ifp->if_alloctype] != NULL)
483 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
484 		    ifp->if_alloctype);
485 
486 #ifdef MAC
487 	mac_ifnet_destroy(ifp);
488 #endif /* MAC */
489 	if (ifp->if_description != NULL)
490 		free(ifp->if_description, M_IFDESCR);
491 	IF_AFDATA_DESTROY(ifp);
492 	IF_ADDR_LOCK_DESTROY(ifp);
493 	ifq_delete(&ifp->if_snd);
494 
495 	for (int i = 0; i < IFCOUNTERS; i++)
496 		counter_u64_free(ifp->if_counters[i]);
497 
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 initialization 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  * Note that if_detach_internal() removes group membership unconditionally
575  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
576  * Thus, when if_vmove() is applied to a cloned interface, group membership
577  * is lost while a cloned one always joins a group whose name is
578  * ifc->ifc_name.  To recover this after if_detach_internal() and
579  * if_attach_internal(), the cloner should be specified to
580  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
581  * attempts to join a group whose name is ifc->ifc_name.
582  *
583  * XXX:
584  *  - The decision to return void and thus require this function to
585  *    succeed is questionable.
586  *  - We should probably do more sanity checking.  For instance we don't
587  *    do anything to insure if_xname is unique or non-empty.
588  */
589 void
590 if_attach(struct ifnet *ifp)
591 {
592 
593 	if_attach_internal(ifp, 0, NULL);
594 }
595 
596 /*
597  * Compute the least common TSO limit.
598  */
599 void
600 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
601 {
602 	/*
603 	 * 1) If there is no limit currently, take the limit from
604 	 * the network adapter.
605 	 *
606 	 * 2) If the network adapter has a limit below the current
607 	 * limit, apply it.
608 	 */
609 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
610 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
611 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
612 	}
613 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
614 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
615 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
616 	}
617 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
618 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
619 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
620 	}
621 }
622 
623 /*
624  * Update TSO limit of a network adapter.
625  *
626  * Returns zero if no change. Else non-zero.
627  */
628 int
629 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
630 {
631 	int retval = 0;
632 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
633 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
634 		retval++;
635 	}
636 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
637 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
638 		retval++;
639 	}
640 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
641 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
642 		retval++;
643 	}
644 	return (retval);
645 }
646 
647 static void
648 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
649 {
650 	unsigned socksize, ifasize;
651 	int namelen, masklen;
652 	struct sockaddr_dl *sdl;
653 	struct ifaddr *ifa;
654 
655 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
656 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
657 		    ifp->if_xname);
658 
659 #ifdef VIMAGE
660 	ifp->if_vnet = curvnet;
661 	if (ifp->if_home_vnet == NULL)
662 		ifp->if_home_vnet = curvnet;
663 #endif
664 
665 	if_addgroup(ifp, IFG_ALL);
666 
667 	/* Restore group membership for cloned interfaces. */
668 	if (vmove && ifc != NULL)
669 		if_clone_addgroup(ifp, ifc);
670 
671 	getmicrotime(&ifp->if_lastchange);
672 	ifp->if_epoch = time_uptime;
673 
674 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
675 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
676 	    ("transmit and qflush must both either be set or both be NULL"));
677 	if (ifp->if_transmit == NULL) {
678 		ifp->if_transmit = if_transmit;
679 		ifp->if_qflush = if_qflush;
680 	}
681 	if (ifp->if_input == NULL)
682 		ifp->if_input = if_input_default;
683 
684 	if (ifp->if_requestencap == NULL)
685 		ifp->if_requestencap = if_requestencap_default;
686 
687 	if (!vmove) {
688 #ifdef MAC
689 		mac_ifnet_create(ifp);
690 #endif
691 
692 		/*
693 		 * Create a Link Level name for this device.
694 		 */
695 		namelen = strlen(ifp->if_xname);
696 		/*
697 		 * Always save enough space for any possiable name so we
698 		 * can do a rename in place later.
699 		 */
700 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
701 		socksize = masklen + ifp->if_addrlen;
702 		if (socksize < sizeof(*sdl))
703 			socksize = sizeof(*sdl);
704 		socksize = roundup2(socksize, sizeof(long));
705 		ifasize = sizeof(*ifa) + 2 * socksize;
706 		ifa = ifa_alloc(ifasize, M_WAITOK);
707 		sdl = (struct sockaddr_dl *)(ifa + 1);
708 		sdl->sdl_len = socksize;
709 		sdl->sdl_family = AF_LINK;
710 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
711 		sdl->sdl_nlen = namelen;
712 		sdl->sdl_index = ifp->if_index;
713 		sdl->sdl_type = ifp->if_type;
714 		ifp->if_addr = ifa;
715 		ifa->ifa_ifp = ifp;
716 		ifa->ifa_rtrequest = link_rtrequest;
717 		ifa->ifa_addr = (struct sockaddr *)sdl;
718 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
719 		ifa->ifa_netmask = (struct sockaddr *)sdl;
720 		sdl->sdl_len = masklen;
721 		while (namelen != 0)
722 			sdl->sdl_data[--namelen] = 0xff;
723 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
724 		/* Reliably crash if used uninitialized. */
725 		ifp->if_broadcastaddr = NULL;
726 
727 #if defined(INET) || defined(INET6)
728 		/* Use defaults for TSO, if nothing is set */
729 		if (ifp->if_hw_tsomax == 0 &&
730 		    ifp->if_hw_tsomaxsegcount == 0 &&
731 		    ifp->if_hw_tsomaxsegsize == 0) {
732 			/*
733 			 * The TSO defaults needs to be such that an
734 			 * NFS mbuf list of 35 mbufs totalling just
735 			 * below 64K works and that a chain of mbufs
736 			 * can be defragged into at most 32 segments:
737 			 */
738 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
739 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
740 			ifp->if_hw_tsomaxsegcount = 35;
741 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
742 
743 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
744 			if (ifp->if_capabilities & IFCAP_TSO) {
745 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
746 				    ifp->if_hw_tsomax,
747 				    ifp->if_hw_tsomaxsegcount,
748 				    ifp->if_hw_tsomaxsegsize);
749 			}
750 		}
751 #endif
752 	}
753 #ifdef VIMAGE
754 	else {
755 		/*
756 		 * Update the interface index in the link layer address
757 		 * of the interface.
758 		 */
759 		for (ifa = ifp->if_addr; ifa != NULL;
760 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
761 			if (ifa->ifa_addr->sa_family == AF_LINK) {
762 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
763 				sdl->sdl_index = ifp->if_index;
764 			}
765 		}
766 	}
767 #endif
768 
769 	IFNET_WLOCK();
770 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
771 #ifdef VIMAGE
772 	curvnet->vnet_ifcnt++;
773 #endif
774 	IFNET_WUNLOCK();
775 
776 	if (domain_init_status >= 2)
777 		if_attachdomain1(ifp);
778 
779 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
780 	if (IS_DEFAULT_VNET(curvnet))
781 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
782 
783 	/* Announce the interface. */
784 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
785 }
786 
787 static void
788 if_attachdomain(void *dummy)
789 {
790 	struct ifnet *ifp;
791 
792 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
793 		if_attachdomain1(ifp);
794 }
795 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
796     if_attachdomain, NULL);
797 
798 static void
799 if_attachdomain1(struct ifnet *ifp)
800 {
801 	struct domain *dp;
802 
803 	/*
804 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
805 	 * cannot lock ifp->if_afdata initialization, entirely.
806 	 */
807 	if (IF_AFDATA_TRYLOCK(ifp) == 0)
808 		return;
809 	if (ifp->if_afdata_initialized >= domain_init_status) {
810 		IF_AFDATA_UNLOCK(ifp);
811 		log(LOG_WARNING, "%s called more than once on %s\n",
812 		    __func__, ifp->if_xname);
813 		return;
814 	}
815 	ifp->if_afdata_initialized = domain_init_status;
816 	IF_AFDATA_UNLOCK(ifp);
817 
818 	/* address family dependent data region */
819 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
820 	for (dp = domains; dp; dp = dp->dom_next) {
821 		if (dp->dom_ifattach)
822 			ifp->if_afdata[dp->dom_family] =
823 			    (*dp->dom_ifattach)(ifp);
824 	}
825 }
826 
827 /*
828  * Remove any unicast or broadcast network addresses from an interface.
829  */
830 void
831 if_purgeaddrs(struct ifnet *ifp)
832 {
833 	struct ifaddr *ifa, *next;
834 
835 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
836 		if (ifa->ifa_addr->sa_family == AF_LINK)
837 			continue;
838 #ifdef INET
839 		/* XXX: Ugly!! ad hoc just for INET */
840 		if (ifa->ifa_addr->sa_family == AF_INET) {
841 			struct ifaliasreq ifr;
842 
843 			bzero(&ifr, sizeof(ifr));
844 			ifr.ifra_addr = *ifa->ifa_addr;
845 			if (ifa->ifa_dstaddr)
846 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
847 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
848 			    NULL) == 0)
849 				continue;
850 		}
851 #endif /* INET */
852 #ifdef INET6
853 		if (ifa->ifa_addr->sa_family == AF_INET6) {
854 			in6_purgeaddr(ifa);
855 			/* ifp_addrhead is already updated */
856 			continue;
857 		}
858 #endif /* INET6 */
859 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
860 		ifa_free(ifa);
861 	}
862 }
863 
864 /*
865  * Remove any multicast network addresses from an interface when an ifnet
866  * is going away.
867  */
868 static void
869 if_purgemaddrs(struct ifnet *ifp)
870 {
871 	struct ifmultiaddr *ifma;
872 	struct ifmultiaddr *next;
873 
874 	IF_ADDR_WLOCK(ifp);
875 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
876 		if_delmulti_locked(ifp, ifma, 1);
877 	IF_ADDR_WUNLOCK(ifp);
878 }
879 
880 /*
881  * Detach an interface, removing it from the list of "active" interfaces.
882  * If vmove flag is set on entry to if_detach_internal(), perform only a
883  * limited subset of cleanup tasks, given that we are moving an ifnet from
884  * one vnet to another, where it must be fully operational.
885  *
886  * XXXRW: There are some significant questions about event ordering, and
887  * how to prevent things from starting to use the interface during detach.
888  */
889 void
890 if_detach(struct ifnet *ifp)
891 {
892 
893 	CURVNET_SET_QUIET(ifp->if_vnet);
894 	if_detach_internal(ifp, 0, NULL);
895 	CURVNET_RESTORE();
896 }
897 
898 static int
899 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
900 {
901 	struct ifaddr *ifa;
902 	int i;
903 	struct domain *dp;
904  	struct ifnet *iter;
905  	int found = 0;
906 
907 	IFNET_WLOCK();
908 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
909 		if (iter == ifp) {
910 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
911 			found = 1;
912 			break;
913 		}
914 #ifdef VIMAGE
915 	if (found)
916 		curvnet->vnet_ifcnt--;
917 #endif
918 	IFNET_WUNLOCK();
919 	if (!found) {
920 		/*
921 		 * While we would want to panic here, we cannot
922 		 * guarantee that the interface is indeed still on
923 		 * the list given we don't hold locks all the way.
924 		 */
925 		return (ENOENT);
926 #if 0
927 		if (vmove)
928 			panic("%s: ifp=%p not on the ifnet tailq %p",
929 			    __func__, ifp, &V_ifnet);
930 		else
931 			return; /* XXX this should panic as well? */
932 #endif
933 	}
934 
935 	/* Check if this is a cloned interface or not. */
936 	if (vmove && ifcp != NULL)
937 		*ifcp = if_clone_findifc(ifp);
938 
939 	/*
940 	 * Remove/wait for pending events.
941 	 */
942 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
943 
944 	/*
945 	 * Remove routes and flush queues.
946 	 */
947 	if_down(ifp);
948 #ifdef ALTQ
949 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
950 		altq_disable(&ifp->if_snd);
951 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
952 		altq_detach(&ifp->if_snd);
953 #endif
954 
955 	if_purgeaddrs(ifp);
956 
957 #ifdef INET
958 	in_ifdetach(ifp);
959 #endif
960 
961 #ifdef INET6
962 	/*
963 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
964 	 * before removing routing entries below, since IPv6 interface direct
965 	 * routes are expected to be removed by the IPv6-specific kernel API.
966 	 * Otherwise, the kernel will detect some inconsistency and bark it.
967 	 */
968 	in6_ifdetach(ifp);
969 #endif
970 	if_purgemaddrs(ifp);
971 
972 	/* Announce that the interface is gone. */
973 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
974 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
975 	if (IS_DEFAULT_VNET(curvnet))
976 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
977 
978 	if (!vmove) {
979 		/*
980 		 * Prevent further calls into the device driver via ifnet.
981 		 */
982 		if_dead(ifp);
983 
984 		/*
985 		 * Remove link ifaddr pointer and maybe decrement if_index.
986 		 * Clean up all addresses.
987 		 */
988 		ifp->if_addr = NULL;
989 
990 		/* We can now free link ifaddr. */
991 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
992 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
993 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
994 			ifa_free(ifa);
995 		}
996 	}
997 
998 	rt_flushifroutes(ifp);
999 	if_delgroups(ifp);
1000 
1001 	/*
1002 	 * We cannot hold the lock over dom_ifdetach calls as they might
1003 	 * sleep, for example trying to drain a callout, thus open up the
1004 	 * theoretical race with re-attaching.
1005 	 */
1006 	IF_AFDATA_LOCK(ifp);
1007 	i = ifp->if_afdata_initialized;
1008 	ifp->if_afdata_initialized = 0;
1009 	IF_AFDATA_UNLOCK(ifp);
1010 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1011 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1012 			(*dp->dom_ifdetach)(ifp,
1013 			    ifp->if_afdata[dp->dom_family]);
1014 	}
1015 
1016 	return (0);
1017 }
1018 
1019 #ifdef VIMAGE
1020 /*
1021  * if_vmove() performs a limited version of if_detach() in current
1022  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1023  * An attempt is made to shrink if_index in current vnet, find an
1024  * unused if_index in target vnet and calls if_grow() if necessary,
1025  * and finally find an unused if_xname for the target vnet.
1026  */
1027 void
1028 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1029 {
1030 	struct if_clone *ifc;
1031 	u_int bif_dlt, bif_hdrlen;
1032 	int rc;
1033 
1034  	/*
1035 	 * if_detach_internal() will call the eventhandler to notify
1036 	 * interface departure.  That will detach if_bpf.  We need to
1037 	 * safe the dlt and hdrlen so we can re-attach it later.
1038 	 */
1039 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1040 
1041 	/*
1042 	 * Detach from current vnet, but preserve LLADDR info, do not
1043 	 * mark as dead etc. so that the ifnet can be reattached later.
1044 	 * If we cannot find it, we lost the race to someone else.
1045 	 */
1046 	rc = if_detach_internal(ifp, 1, &ifc);
1047 	if (rc != 0)
1048 		return;
1049 
1050 	/*
1051 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1052 	 * the if_index for that vnet if possible.
1053 	 *
1054 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1055 	 * or we'd lock on one vnet and unlock on another.
1056 	 */
1057 	IFNET_WLOCK();
1058 	ifindex_free_locked(ifp->if_index);
1059 	IFNET_WUNLOCK();
1060 
1061 	/*
1062 	 * Perform interface-specific reassignment tasks, if provided by
1063 	 * the driver.
1064 	 */
1065 	if (ifp->if_reassign != NULL)
1066 		ifp->if_reassign(ifp, new_vnet, NULL);
1067 
1068 	/*
1069 	 * Switch to the context of the target vnet.
1070 	 */
1071 	CURVNET_SET_QUIET(new_vnet);
1072 
1073 	IFNET_WLOCK();
1074 	ifp->if_index = ifindex_alloc();
1075 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1076 	IFNET_WUNLOCK();
1077 
1078 	if_attach_internal(ifp, 1, ifc);
1079 
1080 	if (ifp->if_bpf == NULL)
1081 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1082 
1083 	CURVNET_RESTORE();
1084 }
1085 
1086 /*
1087  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1088  */
1089 static int
1090 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1091 {
1092 	struct prison *pr;
1093 	struct ifnet *difp;
1094 
1095 	/* Try to find the prison within our visibility. */
1096 	sx_slock(&allprison_lock);
1097 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1098 	sx_sunlock(&allprison_lock);
1099 	if (pr == NULL)
1100 		return (ENXIO);
1101 	prison_hold_locked(pr);
1102 	mtx_unlock(&pr->pr_mtx);
1103 
1104 	/* Do not try to move the iface from and to the same prison. */
1105 	if (pr->pr_vnet == ifp->if_vnet) {
1106 		prison_free(pr);
1107 		return (EEXIST);
1108 	}
1109 
1110 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1111 	/* XXX Lock interfaces to avoid races. */
1112 	CURVNET_SET_QUIET(pr->pr_vnet);
1113 	difp = ifunit(ifname);
1114 	CURVNET_RESTORE();
1115 	if (difp != NULL) {
1116 		prison_free(pr);
1117 		return (EEXIST);
1118 	}
1119 
1120 	/* Move the interface into the child jail/vnet. */
1121 	if_vmove(ifp, pr->pr_vnet);
1122 
1123 	/* Report the new if_xname back to the userland. */
1124 	sprintf(ifname, "%s", ifp->if_xname);
1125 
1126 	prison_free(pr);
1127 	return (0);
1128 }
1129 
1130 static int
1131 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1132 {
1133 	struct prison *pr;
1134 	struct vnet *vnet_dst;
1135 	struct ifnet *ifp;
1136 
1137 	/* Try to find the prison within our visibility. */
1138 	sx_slock(&allprison_lock);
1139 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1140 	sx_sunlock(&allprison_lock);
1141 	if (pr == NULL)
1142 		return (ENXIO);
1143 	prison_hold_locked(pr);
1144 	mtx_unlock(&pr->pr_mtx);
1145 
1146 	/* Make sure the named iface exists in the source prison/vnet. */
1147 	CURVNET_SET(pr->pr_vnet);
1148 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1149 	if (ifp == NULL) {
1150 		CURVNET_RESTORE();
1151 		prison_free(pr);
1152 		return (ENXIO);
1153 	}
1154 
1155 	/* Do not try to move the iface from and to the same prison. */
1156 	vnet_dst = TD_TO_VNET(td);
1157 	if (vnet_dst == ifp->if_vnet) {
1158 		CURVNET_RESTORE();
1159 		prison_free(pr);
1160 		return (EEXIST);
1161 	}
1162 
1163 	/* Get interface back from child jail/vnet. */
1164 	if_vmove(ifp, vnet_dst);
1165 	CURVNET_RESTORE();
1166 
1167 	/* Report the new if_xname back to the userland. */
1168 	sprintf(ifname, "%s", ifp->if_xname);
1169 
1170 	prison_free(pr);
1171 	return (0);
1172 }
1173 #endif /* VIMAGE */
1174 
1175 /*
1176  * Add a group to an interface
1177  */
1178 int
1179 if_addgroup(struct ifnet *ifp, const char *groupname)
1180 {
1181 	struct ifg_list		*ifgl;
1182 	struct ifg_group	*ifg = NULL;
1183 	struct ifg_member	*ifgm;
1184 	int 			 new = 0;
1185 
1186 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1187 	    groupname[strlen(groupname) - 1] <= '9')
1188 		return (EINVAL);
1189 
1190 	IFNET_WLOCK();
1191 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1192 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1193 			IFNET_WUNLOCK();
1194 			return (EEXIST);
1195 		}
1196 
1197 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1198 	    M_NOWAIT)) == NULL) {
1199 	    	IFNET_WUNLOCK();
1200 		return (ENOMEM);
1201 	}
1202 
1203 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1204 	    M_TEMP, M_NOWAIT)) == NULL) {
1205 		free(ifgl, M_TEMP);
1206 		IFNET_WUNLOCK();
1207 		return (ENOMEM);
1208 	}
1209 
1210 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1211 		if (!strcmp(ifg->ifg_group, groupname))
1212 			break;
1213 
1214 	if (ifg == NULL) {
1215 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1216 		    M_TEMP, M_NOWAIT)) == NULL) {
1217 			free(ifgl, M_TEMP);
1218 			free(ifgm, M_TEMP);
1219 			IFNET_WUNLOCK();
1220 			return (ENOMEM);
1221 		}
1222 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1223 		ifg->ifg_refcnt = 0;
1224 		TAILQ_INIT(&ifg->ifg_members);
1225 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1226 		new = 1;
1227 	}
1228 
1229 	ifg->ifg_refcnt++;
1230 	ifgl->ifgl_group = ifg;
1231 	ifgm->ifgm_ifp = ifp;
1232 
1233 	IF_ADDR_WLOCK(ifp);
1234 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1235 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1236 	IF_ADDR_WUNLOCK(ifp);
1237 
1238 	IFNET_WUNLOCK();
1239 
1240 	if (new)
1241 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1242 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1243 
1244 	return (0);
1245 }
1246 
1247 /*
1248  * Remove a group from an interface
1249  */
1250 int
1251 if_delgroup(struct ifnet *ifp, const char *groupname)
1252 {
1253 	struct ifg_list		*ifgl;
1254 	struct ifg_member	*ifgm;
1255 
1256 	IFNET_WLOCK();
1257 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1258 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1259 			break;
1260 	if (ifgl == NULL) {
1261 		IFNET_WUNLOCK();
1262 		return (ENOENT);
1263 	}
1264 
1265 	IF_ADDR_WLOCK(ifp);
1266 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1267 	IF_ADDR_WUNLOCK(ifp);
1268 
1269 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1270 		if (ifgm->ifgm_ifp == ifp)
1271 			break;
1272 
1273 	if (ifgm != NULL) {
1274 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1275 		free(ifgm, M_TEMP);
1276 	}
1277 
1278 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1279 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1280 		IFNET_WUNLOCK();
1281 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1282 		free(ifgl->ifgl_group, M_TEMP);
1283 	} else
1284 		IFNET_WUNLOCK();
1285 
1286 	free(ifgl, M_TEMP);
1287 
1288 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1289 
1290 	return (0);
1291 }
1292 
1293 /*
1294  * Remove an interface from all groups
1295  */
1296 static void
1297 if_delgroups(struct ifnet *ifp)
1298 {
1299 	struct ifg_list		*ifgl;
1300 	struct ifg_member	*ifgm;
1301 	char groupname[IFNAMSIZ];
1302 
1303 	IFNET_WLOCK();
1304 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1305 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1306 
1307 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1308 
1309 		IF_ADDR_WLOCK(ifp);
1310 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1311 		IF_ADDR_WUNLOCK(ifp);
1312 
1313 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1314 			if (ifgm->ifgm_ifp == ifp)
1315 				break;
1316 
1317 		if (ifgm != NULL) {
1318 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1319 			    ifgm_next);
1320 			free(ifgm, M_TEMP);
1321 		}
1322 
1323 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1324 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1325 			IFNET_WUNLOCK();
1326 			EVENTHANDLER_INVOKE(group_detach_event,
1327 			    ifgl->ifgl_group);
1328 			free(ifgl->ifgl_group, M_TEMP);
1329 		} else
1330 			IFNET_WUNLOCK();
1331 
1332 		free(ifgl, M_TEMP);
1333 
1334 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1335 
1336 		IFNET_WLOCK();
1337 	}
1338 	IFNET_WUNLOCK();
1339 }
1340 
1341 /*
1342  * Stores all groups from an interface in memory pointed
1343  * to by data
1344  */
1345 static int
1346 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1347 {
1348 	int			 len, error;
1349 	struct ifg_list		*ifgl;
1350 	struct ifg_req		 ifgrq, *ifgp;
1351 	struct ifgroupreq	*ifgr = data;
1352 
1353 	if (ifgr->ifgr_len == 0) {
1354 		IF_ADDR_RLOCK(ifp);
1355 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1356 			ifgr->ifgr_len += sizeof(struct ifg_req);
1357 		IF_ADDR_RUNLOCK(ifp);
1358 		return (0);
1359 	}
1360 
1361 	len = ifgr->ifgr_len;
1362 	ifgp = ifgr->ifgr_groups;
1363 	/* XXX: wire */
1364 	IF_ADDR_RLOCK(ifp);
1365 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1366 		if (len < sizeof(ifgrq)) {
1367 			IF_ADDR_RUNLOCK(ifp);
1368 			return (EINVAL);
1369 		}
1370 		bzero(&ifgrq, sizeof ifgrq);
1371 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1372 		    sizeof(ifgrq.ifgrq_group));
1373 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1374 		    	IF_ADDR_RUNLOCK(ifp);
1375 			return (error);
1376 		}
1377 		len -= sizeof(ifgrq);
1378 		ifgp++;
1379 	}
1380 	IF_ADDR_RUNLOCK(ifp);
1381 
1382 	return (0);
1383 }
1384 
1385 /*
1386  * Stores all members of a group in memory pointed to by data
1387  */
1388 static int
1389 if_getgroupmembers(struct ifgroupreq *data)
1390 {
1391 	struct ifgroupreq	*ifgr = data;
1392 	struct ifg_group	*ifg;
1393 	struct ifg_member	*ifgm;
1394 	struct ifg_req		 ifgrq, *ifgp;
1395 	int			 len, error;
1396 
1397 	IFNET_RLOCK();
1398 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1399 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1400 			break;
1401 	if (ifg == NULL) {
1402 		IFNET_RUNLOCK();
1403 		return (ENOENT);
1404 	}
1405 
1406 	if (ifgr->ifgr_len == 0) {
1407 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1408 			ifgr->ifgr_len += sizeof(ifgrq);
1409 		IFNET_RUNLOCK();
1410 		return (0);
1411 	}
1412 
1413 	len = ifgr->ifgr_len;
1414 	ifgp = ifgr->ifgr_groups;
1415 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1416 		if (len < sizeof(ifgrq)) {
1417 			IFNET_RUNLOCK();
1418 			return (EINVAL);
1419 		}
1420 		bzero(&ifgrq, sizeof ifgrq);
1421 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1422 		    sizeof(ifgrq.ifgrq_member));
1423 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1424 			IFNET_RUNLOCK();
1425 			return (error);
1426 		}
1427 		len -= sizeof(ifgrq);
1428 		ifgp++;
1429 	}
1430 	IFNET_RUNLOCK();
1431 
1432 	return (0);
1433 }
1434 
1435 /*
1436  * Return counter values from counter(9)s stored in ifnet.
1437  */
1438 uint64_t
1439 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1440 {
1441 
1442 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1443 
1444 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1445 }
1446 
1447 /*
1448  * Increase an ifnet counter. Usually used for counters shared
1449  * between the stack and a driver, but function supports them all.
1450  */
1451 void
1452 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1453 {
1454 
1455 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1456 
1457 	counter_u64_add(ifp->if_counters[cnt], inc);
1458 }
1459 
1460 /*
1461  * Copy data from ifnet to userland API structure if_data.
1462  */
1463 void
1464 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1465 {
1466 
1467 	ifd->ifi_type = ifp->if_type;
1468 	ifd->ifi_physical = 0;
1469 	ifd->ifi_addrlen = ifp->if_addrlen;
1470 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1471 	ifd->ifi_link_state = ifp->if_link_state;
1472 	ifd->ifi_vhid = 0;
1473 	ifd->ifi_datalen = sizeof(struct if_data);
1474 	ifd->ifi_mtu = ifp->if_mtu;
1475 	ifd->ifi_metric = ifp->if_metric;
1476 	ifd->ifi_baudrate = ifp->if_baudrate;
1477 	ifd->ifi_hwassist = ifp->if_hwassist;
1478 	ifd->ifi_epoch = ifp->if_epoch;
1479 	ifd->ifi_lastchange = ifp->if_lastchange;
1480 
1481 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1482 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1483 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1484 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1485 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1486 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1487 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1488 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1489 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1490 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1491 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1492 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1493 }
1494 
1495 /*
1496  * Wrapper functions for struct ifnet address list locking macros.  These are
1497  * used by kernel modules to avoid encoding programming interface or binary
1498  * interface assumptions that may be violated when kernel-internal locking
1499  * approaches change.
1500  */
1501 void
1502 if_addr_rlock(struct ifnet *ifp)
1503 {
1504 
1505 	IF_ADDR_RLOCK(ifp);
1506 }
1507 
1508 void
1509 if_addr_runlock(struct ifnet *ifp)
1510 {
1511 
1512 	IF_ADDR_RUNLOCK(ifp);
1513 }
1514 
1515 void
1516 if_maddr_rlock(if_t ifp)
1517 {
1518 
1519 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1520 }
1521 
1522 void
1523 if_maddr_runlock(if_t ifp)
1524 {
1525 
1526 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1527 }
1528 
1529 /*
1530  * Initialization, destruction and refcounting functions for ifaddrs.
1531  */
1532 struct ifaddr *
1533 ifa_alloc(size_t size, int flags)
1534 {
1535 	struct ifaddr *ifa;
1536 
1537 	KASSERT(size >= sizeof(struct ifaddr),
1538 	    ("%s: invalid size %zu", __func__, size));
1539 
1540 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1541 	if (ifa == NULL)
1542 		return (NULL);
1543 
1544 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1545 		goto fail;
1546 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1547 		goto fail;
1548 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1549 		goto fail;
1550 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1551 		goto fail;
1552 
1553 	refcount_init(&ifa->ifa_refcnt, 1);
1554 
1555 	return (ifa);
1556 
1557 fail:
1558 	/* free(NULL) is okay */
1559 	counter_u64_free(ifa->ifa_opackets);
1560 	counter_u64_free(ifa->ifa_ipackets);
1561 	counter_u64_free(ifa->ifa_obytes);
1562 	counter_u64_free(ifa->ifa_ibytes);
1563 	free(ifa, M_IFADDR);
1564 
1565 	return (NULL);
1566 }
1567 
1568 void
1569 ifa_ref(struct ifaddr *ifa)
1570 {
1571 
1572 	refcount_acquire(&ifa->ifa_refcnt);
1573 }
1574 
1575 void
1576 ifa_free(struct ifaddr *ifa)
1577 {
1578 
1579 	if (refcount_release(&ifa->ifa_refcnt)) {
1580 		counter_u64_free(ifa->ifa_opackets);
1581 		counter_u64_free(ifa->ifa_ipackets);
1582 		counter_u64_free(ifa->ifa_obytes);
1583 		counter_u64_free(ifa->ifa_ibytes);
1584 		free(ifa, M_IFADDR);
1585 	}
1586 }
1587 
1588 static int
1589 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1590     struct sockaddr *ia)
1591 {
1592 	int error;
1593 	struct rt_addrinfo info;
1594 	struct sockaddr_dl null_sdl;
1595 	struct ifnet *ifp;
1596 
1597 	ifp = ifa->ifa_ifp;
1598 
1599 	bzero(&info, sizeof(info));
1600 	if (cmd != RTM_DELETE)
1601 		info.rti_ifp = V_loif;
1602 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1603 	info.rti_info[RTAX_DST] = ia;
1604 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1605 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1606 
1607 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1608 
1609 	if (error != 0)
1610 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1611 		    __func__, otype, if_name(ifp), error);
1612 
1613 	return (error);
1614 }
1615 
1616 int
1617 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1618 {
1619 
1620 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1621 }
1622 
1623 int
1624 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1625 {
1626 
1627 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1628 }
1629 
1630 int
1631 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1632 {
1633 
1634 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1635 }
1636 
1637 /*
1638  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1639  * structs used to represent other address families, it is necessary
1640  * to perform a different comparison.
1641  */
1642 
1643 #define	sa_dl_equal(a1, a2)	\
1644 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1645 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1646 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1647 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1648 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1649 
1650 /*
1651  * Locate an interface based on a complete address.
1652  */
1653 /*ARGSUSED*/
1654 static struct ifaddr *
1655 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1656 {
1657 	struct ifnet *ifp;
1658 	struct ifaddr *ifa;
1659 
1660 	IFNET_RLOCK_NOSLEEP();
1661 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1662 		IF_ADDR_RLOCK(ifp);
1663 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1664 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1665 				continue;
1666 			if (sa_equal(addr, ifa->ifa_addr)) {
1667 				if (getref)
1668 					ifa_ref(ifa);
1669 				IF_ADDR_RUNLOCK(ifp);
1670 				goto done;
1671 			}
1672 			/* IP6 doesn't have broadcast */
1673 			if ((ifp->if_flags & IFF_BROADCAST) &&
1674 			    ifa->ifa_broadaddr &&
1675 			    ifa->ifa_broadaddr->sa_len != 0 &&
1676 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1677 				if (getref)
1678 					ifa_ref(ifa);
1679 				IF_ADDR_RUNLOCK(ifp);
1680 				goto done;
1681 			}
1682 		}
1683 		IF_ADDR_RUNLOCK(ifp);
1684 	}
1685 	ifa = NULL;
1686 done:
1687 	IFNET_RUNLOCK_NOSLEEP();
1688 	return (ifa);
1689 }
1690 
1691 struct ifaddr *
1692 ifa_ifwithaddr(const struct sockaddr *addr)
1693 {
1694 
1695 	return (ifa_ifwithaddr_internal(addr, 1));
1696 }
1697 
1698 int
1699 ifa_ifwithaddr_check(const struct sockaddr *addr)
1700 {
1701 
1702 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1703 }
1704 
1705 /*
1706  * Locate an interface based on the broadcast address.
1707  */
1708 /* ARGSUSED */
1709 struct ifaddr *
1710 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1711 {
1712 	struct ifnet *ifp;
1713 	struct ifaddr *ifa;
1714 
1715 	IFNET_RLOCK_NOSLEEP();
1716 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1717 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1718 			continue;
1719 		IF_ADDR_RLOCK(ifp);
1720 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1721 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1722 				continue;
1723 			if ((ifp->if_flags & IFF_BROADCAST) &&
1724 			    ifa->ifa_broadaddr &&
1725 			    ifa->ifa_broadaddr->sa_len != 0 &&
1726 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1727 				ifa_ref(ifa);
1728 				IF_ADDR_RUNLOCK(ifp);
1729 				goto done;
1730 			}
1731 		}
1732 		IF_ADDR_RUNLOCK(ifp);
1733 	}
1734 	ifa = NULL;
1735 done:
1736 	IFNET_RUNLOCK_NOSLEEP();
1737 	return (ifa);
1738 }
1739 
1740 /*
1741  * Locate the point to point interface with a given destination address.
1742  */
1743 /*ARGSUSED*/
1744 struct ifaddr *
1745 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1746 {
1747 	struct ifnet *ifp;
1748 	struct ifaddr *ifa;
1749 
1750 	IFNET_RLOCK_NOSLEEP();
1751 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1752 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1753 			continue;
1754 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1755 			continue;
1756 		IF_ADDR_RLOCK(ifp);
1757 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1758 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1759 				continue;
1760 			if (ifa->ifa_dstaddr != NULL &&
1761 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1762 				ifa_ref(ifa);
1763 				IF_ADDR_RUNLOCK(ifp);
1764 				goto done;
1765 			}
1766 		}
1767 		IF_ADDR_RUNLOCK(ifp);
1768 	}
1769 	ifa = NULL;
1770 done:
1771 	IFNET_RUNLOCK_NOSLEEP();
1772 	return (ifa);
1773 }
1774 
1775 /*
1776  * Find an interface on a specific network.  If many, choice
1777  * is most specific found.
1778  */
1779 struct ifaddr *
1780 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1781 {
1782 	struct ifnet *ifp;
1783 	struct ifaddr *ifa;
1784 	struct ifaddr *ifa_maybe = NULL;
1785 	u_int af = addr->sa_family;
1786 	const char *addr_data = addr->sa_data, *cplim;
1787 
1788 	/*
1789 	 * AF_LINK addresses can be looked up directly by their index number,
1790 	 * so do that if we can.
1791 	 */
1792 	if (af == AF_LINK) {
1793 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1794 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1795 		return (ifaddr_byindex(sdl->sdl_index));
1796 	}
1797 
1798 	/*
1799 	 * Scan though each interface, looking for ones that have addresses
1800 	 * in this address family and the requested fib.  Maintain a reference
1801 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1802 	 * kept it stable when we move onto the next interface.
1803 	 */
1804 	IFNET_RLOCK_NOSLEEP();
1805 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1806 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1807 			continue;
1808 		IF_ADDR_RLOCK(ifp);
1809 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1810 			const char *cp, *cp2, *cp3;
1811 
1812 			if (ifa->ifa_addr->sa_family != af)
1813 next:				continue;
1814 			if (af == AF_INET &&
1815 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1816 				/*
1817 				 * This is a bit broken as it doesn't
1818 				 * take into account that the remote end may
1819 				 * be a single node in the network we are
1820 				 * looking for.
1821 				 * The trouble is that we don't know the
1822 				 * netmask for the remote end.
1823 				 */
1824 				if (ifa->ifa_dstaddr != NULL &&
1825 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1826 					ifa_ref(ifa);
1827 					IF_ADDR_RUNLOCK(ifp);
1828 					goto done;
1829 				}
1830 			} else {
1831 				/*
1832 				 * Scan all the bits in the ifa's address.
1833 				 * If a bit dissagrees with what we are
1834 				 * looking for, mask it with the netmask
1835 				 * to see if it really matters.
1836 				 * (A byte at a time)
1837 				 */
1838 				if (ifa->ifa_netmask == 0)
1839 					continue;
1840 				cp = addr_data;
1841 				cp2 = ifa->ifa_addr->sa_data;
1842 				cp3 = ifa->ifa_netmask->sa_data;
1843 				cplim = ifa->ifa_netmask->sa_len
1844 					+ (char *)ifa->ifa_netmask;
1845 				while (cp3 < cplim)
1846 					if ((*cp++ ^ *cp2++) & *cp3++)
1847 						goto next; /* next address! */
1848 				/*
1849 				 * If the netmask of what we just found
1850 				 * is more specific than what we had before
1851 				 * (if we had one), or if the virtual status
1852 				 * of new prefix is better than of the old one,
1853 				 * then remember the new one before continuing
1854 				 * to search for an even better one.
1855 				 */
1856 				if (ifa_maybe == NULL ||
1857 				    ifa_preferred(ifa_maybe, ifa) ||
1858 				    rn_refines((caddr_t)ifa->ifa_netmask,
1859 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1860 					if (ifa_maybe != NULL)
1861 						ifa_free(ifa_maybe);
1862 					ifa_maybe = ifa;
1863 					ifa_ref(ifa_maybe);
1864 				}
1865 			}
1866 		}
1867 		IF_ADDR_RUNLOCK(ifp);
1868 	}
1869 	ifa = ifa_maybe;
1870 	ifa_maybe = NULL;
1871 done:
1872 	IFNET_RUNLOCK_NOSLEEP();
1873 	if (ifa_maybe != NULL)
1874 		ifa_free(ifa_maybe);
1875 	return (ifa);
1876 }
1877 
1878 /*
1879  * Find an interface address specific to an interface best matching
1880  * a given address.
1881  */
1882 struct ifaddr *
1883 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1884 {
1885 	struct ifaddr *ifa;
1886 	const char *cp, *cp2, *cp3;
1887 	char *cplim;
1888 	struct ifaddr *ifa_maybe = NULL;
1889 	u_int af = addr->sa_family;
1890 
1891 	if (af >= AF_MAX)
1892 		return (NULL);
1893 	IF_ADDR_RLOCK(ifp);
1894 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1895 		if (ifa->ifa_addr->sa_family != af)
1896 			continue;
1897 		if (ifa_maybe == NULL)
1898 			ifa_maybe = ifa;
1899 		if (ifa->ifa_netmask == 0) {
1900 			if (sa_equal(addr, ifa->ifa_addr) ||
1901 			    (ifa->ifa_dstaddr &&
1902 			    sa_equal(addr, ifa->ifa_dstaddr)))
1903 				goto done;
1904 			continue;
1905 		}
1906 		if (ifp->if_flags & IFF_POINTOPOINT) {
1907 			if (sa_equal(addr, ifa->ifa_dstaddr))
1908 				goto done;
1909 		} else {
1910 			cp = addr->sa_data;
1911 			cp2 = ifa->ifa_addr->sa_data;
1912 			cp3 = ifa->ifa_netmask->sa_data;
1913 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1914 			for (; cp3 < cplim; cp3++)
1915 				if ((*cp++ ^ *cp2++) & *cp3)
1916 					break;
1917 			if (cp3 == cplim)
1918 				goto done;
1919 		}
1920 	}
1921 	ifa = ifa_maybe;
1922 done:
1923 	if (ifa != NULL)
1924 		ifa_ref(ifa);
1925 	IF_ADDR_RUNLOCK(ifp);
1926 	return (ifa);
1927 }
1928 
1929 /*
1930  * See whether new ifa is better than current one:
1931  * 1) A non-virtual one is preferred over virtual.
1932  * 2) A virtual in master state preferred over any other state.
1933  *
1934  * Used in several address selecting functions.
1935  */
1936 int
1937 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1938 {
1939 
1940 	return (cur->ifa_carp && (!next->ifa_carp ||
1941 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1942 }
1943 
1944 #include <net/if_llatbl.h>
1945 
1946 /*
1947  * Default action when installing a route with a Link Level gateway.
1948  * Lookup an appropriate real ifa to point to.
1949  * This should be moved to /sys/net/link.c eventually.
1950  */
1951 static void
1952 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1953 {
1954 	struct ifaddr *ifa, *oifa;
1955 	struct sockaddr *dst;
1956 	struct ifnet *ifp;
1957 
1958 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
1959 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
1960 		return;
1961 	ifa = ifaof_ifpforaddr(dst, ifp);
1962 	if (ifa) {
1963 		oifa = rt->rt_ifa;
1964 		rt->rt_ifa = ifa;
1965 		ifa_free(oifa);
1966 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1967 			ifa->ifa_rtrequest(cmd, rt, info);
1968 	}
1969 }
1970 
1971 struct sockaddr_dl *
1972 link_alloc_sdl(size_t size, int flags)
1973 {
1974 
1975 	return (malloc(size, M_TEMP, flags));
1976 }
1977 
1978 void
1979 link_free_sdl(struct sockaddr *sa)
1980 {
1981 	free(sa, M_TEMP);
1982 }
1983 
1984 /*
1985  * Fills in given sdl with interface basic info.
1986  * Returns pointer to filled sdl.
1987  */
1988 struct sockaddr_dl *
1989 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1990 {
1991 	struct sockaddr_dl *sdl;
1992 
1993 	sdl = (struct sockaddr_dl *)paddr;
1994 	memset(sdl, 0, sizeof(struct sockaddr_dl));
1995 	sdl->sdl_len = sizeof(struct sockaddr_dl);
1996 	sdl->sdl_family = AF_LINK;
1997 	sdl->sdl_index = ifp->if_index;
1998 	sdl->sdl_type = iftype;
1999 
2000 	return (sdl);
2001 }
2002 
2003 /*
2004  * Mark an interface down and notify protocols of
2005  * the transition.
2006  */
2007 static void
2008 if_unroute(struct ifnet *ifp, int flag, int fam)
2009 {
2010 	struct ifaddr *ifa;
2011 
2012 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2013 
2014 	ifp->if_flags &= ~flag;
2015 	getmicrotime(&ifp->if_lastchange);
2016 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2017 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2018 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2019 	ifp->if_qflush(ifp);
2020 
2021 	if (ifp->if_carp)
2022 		(*carp_linkstate_p)(ifp);
2023 	rt_ifmsg(ifp);
2024 }
2025 
2026 /*
2027  * Mark an interface up and notify protocols of
2028  * the transition.
2029  */
2030 static void
2031 if_route(struct ifnet *ifp, int flag, int fam)
2032 {
2033 	struct ifaddr *ifa;
2034 
2035 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2036 
2037 	ifp->if_flags |= flag;
2038 	getmicrotime(&ifp->if_lastchange);
2039 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2040 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2041 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2042 	if (ifp->if_carp)
2043 		(*carp_linkstate_p)(ifp);
2044 	rt_ifmsg(ifp);
2045 #ifdef INET6
2046 	in6_if_up(ifp);
2047 #endif
2048 }
2049 
2050 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2051 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2052 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2053 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2054 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2055 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2056 void	*(*vlan_cookie_p)(struct ifnet *);
2057 
2058 /*
2059  * Handle a change in the interface link state. To avoid LORs
2060  * between driver lock and upper layer locks, as well as possible
2061  * recursions, we post event to taskqueue, and all job
2062  * is done in static do_link_state_change().
2063  */
2064 void
2065 if_link_state_change(struct ifnet *ifp, int link_state)
2066 {
2067 	/* Return if state hasn't changed. */
2068 	if (ifp->if_link_state == link_state)
2069 		return;
2070 
2071 	ifp->if_link_state = link_state;
2072 
2073 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2074 }
2075 
2076 static void
2077 do_link_state_change(void *arg, int pending)
2078 {
2079 	struct ifnet *ifp = (struct ifnet *)arg;
2080 	int link_state = ifp->if_link_state;
2081 	CURVNET_SET(ifp->if_vnet);
2082 
2083 	/* Notify that the link state has changed. */
2084 	rt_ifmsg(ifp);
2085 	if (ifp->if_vlantrunk != NULL)
2086 		(*vlan_link_state_p)(ifp);
2087 
2088 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2089 	    ifp->if_l2com != NULL)
2090 		(*ng_ether_link_state_p)(ifp, link_state);
2091 	if (ifp->if_carp)
2092 		(*carp_linkstate_p)(ifp);
2093 	if (ifp->if_bridge)
2094 		(*bridge_linkstate_p)(ifp);
2095 	if (ifp->if_lagg)
2096 		(*lagg_linkstate_p)(ifp, link_state);
2097 
2098 	if (IS_DEFAULT_VNET(curvnet))
2099 		devctl_notify("IFNET", ifp->if_xname,
2100 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2101 		    NULL);
2102 	if (pending > 1)
2103 		if_printf(ifp, "%d link states coalesced\n", pending);
2104 	if (log_link_state_change)
2105 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2106 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2107 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2108 	CURVNET_RESTORE();
2109 }
2110 
2111 /*
2112  * Mark an interface down and notify protocols of
2113  * the transition.
2114  */
2115 void
2116 if_down(struct ifnet *ifp)
2117 {
2118 
2119 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2120 }
2121 
2122 /*
2123  * Mark an interface up and notify protocols of
2124  * the transition.
2125  */
2126 void
2127 if_up(struct ifnet *ifp)
2128 {
2129 
2130 	if_route(ifp, IFF_UP, AF_UNSPEC);
2131 }
2132 
2133 /*
2134  * Flush an interface queue.
2135  */
2136 void
2137 if_qflush(struct ifnet *ifp)
2138 {
2139 	struct mbuf *m, *n;
2140 	struct ifaltq *ifq;
2141 
2142 	ifq = &ifp->if_snd;
2143 	IFQ_LOCK(ifq);
2144 #ifdef ALTQ
2145 	if (ALTQ_IS_ENABLED(ifq))
2146 		ALTQ_PURGE(ifq);
2147 #endif
2148 	n = ifq->ifq_head;
2149 	while ((m = n) != NULL) {
2150 		n = m->m_nextpkt;
2151 		m_freem(m);
2152 	}
2153 	ifq->ifq_head = 0;
2154 	ifq->ifq_tail = 0;
2155 	ifq->ifq_len = 0;
2156 	IFQ_UNLOCK(ifq);
2157 }
2158 
2159 /*
2160  * Map interface name to interface structure pointer, with or without
2161  * returning a reference.
2162  */
2163 struct ifnet *
2164 ifunit_ref(const char *name)
2165 {
2166 	struct ifnet *ifp;
2167 
2168 	IFNET_RLOCK_NOSLEEP();
2169 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2170 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2171 		    !(ifp->if_flags & IFF_DYING))
2172 			break;
2173 	}
2174 	if (ifp != NULL)
2175 		if_ref(ifp);
2176 	IFNET_RUNLOCK_NOSLEEP();
2177 	return (ifp);
2178 }
2179 
2180 struct ifnet *
2181 ifunit(const char *name)
2182 {
2183 	struct ifnet *ifp;
2184 
2185 	IFNET_RLOCK_NOSLEEP();
2186 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2187 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2188 			break;
2189 	}
2190 	IFNET_RUNLOCK_NOSLEEP();
2191 	return (ifp);
2192 }
2193 
2194 /*
2195  * Hardware specific interface ioctls.
2196  */
2197 static int
2198 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2199 {
2200 	struct ifreq *ifr;
2201 	int error = 0;
2202 	int new_flags, temp_flags;
2203 	size_t namelen, onamelen;
2204 	size_t descrlen;
2205 	char *descrbuf, *odescrbuf;
2206 	char new_name[IFNAMSIZ];
2207 	struct ifaddr *ifa;
2208 	struct sockaddr_dl *sdl;
2209 
2210 	ifr = (struct ifreq *)data;
2211 	switch (cmd) {
2212 	case SIOCGIFINDEX:
2213 		ifr->ifr_index = ifp->if_index;
2214 		break;
2215 
2216 	case SIOCGIFFLAGS:
2217 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2218 		ifr->ifr_flags = temp_flags & 0xffff;
2219 		ifr->ifr_flagshigh = temp_flags >> 16;
2220 		break;
2221 
2222 	case SIOCGIFCAP:
2223 		ifr->ifr_reqcap = ifp->if_capabilities;
2224 		ifr->ifr_curcap = ifp->if_capenable;
2225 		break;
2226 
2227 #ifdef MAC
2228 	case SIOCGIFMAC:
2229 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2230 		break;
2231 #endif
2232 
2233 	case SIOCGIFMETRIC:
2234 		ifr->ifr_metric = ifp->if_metric;
2235 		break;
2236 
2237 	case SIOCGIFMTU:
2238 		ifr->ifr_mtu = ifp->if_mtu;
2239 		break;
2240 
2241 	case SIOCGIFPHYS:
2242 		/* XXXGL: did this ever worked? */
2243 		ifr->ifr_phys = 0;
2244 		break;
2245 
2246 	case SIOCGIFDESCR:
2247 		error = 0;
2248 		sx_slock(&ifdescr_sx);
2249 		if (ifp->if_description == NULL)
2250 			error = ENOMSG;
2251 		else {
2252 			/* space for terminating nul */
2253 			descrlen = strlen(ifp->if_description) + 1;
2254 			if (ifr->ifr_buffer.length < descrlen)
2255 				ifr->ifr_buffer.buffer = NULL;
2256 			else
2257 				error = copyout(ifp->if_description,
2258 				    ifr->ifr_buffer.buffer, descrlen);
2259 			ifr->ifr_buffer.length = descrlen;
2260 		}
2261 		sx_sunlock(&ifdescr_sx);
2262 		break;
2263 
2264 	case SIOCSIFDESCR:
2265 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2266 		if (error)
2267 			return (error);
2268 
2269 		/*
2270 		 * Copy only (length-1) bytes to make sure that
2271 		 * if_description is always nul terminated.  The
2272 		 * length parameter is supposed to count the
2273 		 * terminating nul in.
2274 		 */
2275 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2276 			return (ENAMETOOLONG);
2277 		else if (ifr->ifr_buffer.length == 0)
2278 			descrbuf = NULL;
2279 		else {
2280 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2281 			    M_WAITOK | M_ZERO);
2282 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2283 			    ifr->ifr_buffer.length - 1);
2284 			if (error) {
2285 				free(descrbuf, M_IFDESCR);
2286 				break;
2287 			}
2288 		}
2289 
2290 		sx_xlock(&ifdescr_sx);
2291 		odescrbuf = ifp->if_description;
2292 		ifp->if_description = descrbuf;
2293 		sx_xunlock(&ifdescr_sx);
2294 
2295 		getmicrotime(&ifp->if_lastchange);
2296 		free(odescrbuf, M_IFDESCR);
2297 		break;
2298 
2299 	case SIOCGIFFIB:
2300 		ifr->ifr_fib = ifp->if_fib;
2301 		break;
2302 
2303 	case SIOCSIFFIB:
2304 		error = priv_check(td, PRIV_NET_SETIFFIB);
2305 		if (error)
2306 			return (error);
2307 		if (ifr->ifr_fib >= rt_numfibs)
2308 			return (EINVAL);
2309 
2310 		ifp->if_fib = ifr->ifr_fib;
2311 		break;
2312 
2313 	case SIOCSIFFLAGS:
2314 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2315 		if (error)
2316 			return (error);
2317 		/*
2318 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2319 		 * check, so we don't need special handling here yet.
2320 		 */
2321 		new_flags = (ifr->ifr_flags & 0xffff) |
2322 		    (ifr->ifr_flagshigh << 16);
2323 		if (ifp->if_flags & IFF_UP &&
2324 		    (new_flags & IFF_UP) == 0) {
2325 			if_down(ifp);
2326 		} else if (new_flags & IFF_UP &&
2327 		    (ifp->if_flags & IFF_UP) == 0) {
2328 			if_up(ifp);
2329 		}
2330 		/* See if permanently promiscuous mode bit is about to flip */
2331 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2332 			if (new_flags & IFF_PPROMISC)
2333 				ifp->if_flags |= IFF_PROMISC;
2334 			else if (ifp->if_pcount == 0)
2335 				ifp->if_flags &= ~IFF_PROMISC;
2336 			if (log_promisc_mode_change)
2337                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2338                                     ifp->if_xname,
2339                                     ((new_flags & IFF_PPROMISC) ?
2340                                      "enabled" : "disabled"));
2341 		}
2342 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2343 			(new_flags &~ IFF_CANTCHANGE);
2344 		if (ifp->if_ioctl) {
2345 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2346 		}
2347 		getmicrotime(&ifp->if_lastchange);
2348 		break;
2349 
2350 	case SIOCSIFCAP:
2351 		error = priv_check(td, PRIV_NET_SETIFCAP);
2352 		if (error)
2353 			return (error);
2354 		if (ifp->if_ioctl == NULL)
2355 			return (EOPNOTSUPP);
2356 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2357 			return (EINVAL);
2358 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2359 		if (error == 0)
2360 			getmicrotime(&ifp->if_lastchange);
2361 		break;
2362 
2363 #ifdef MAC
2364 	case SIOCSIFMAC:
2365 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2366 		break;
2367 #endif
2368 
2369 	case SIOCSIFNAME:
2370 		error = priv_check(td, PRIV_NET_SETIFNAME);
2371 		if (error)
2372 			return (error);
2373 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2374 		if (error != 0)
2375 			return (error);
2376 		if (new_name[0] == '\0')
2377 			return (EINVAL);
2378 		if (new_name[IFNAMSIZ-1] != '\0') {
2379 			new_name[IFNAMSIZ-1] = '\0';
2380 			if (strlen(new_name) == IFNAMSIZ-1)
2381 				return (EINVAL);
2382 		}
2383 		if (ifunit(new_name) != NULL)
2384 			return (EEXIST);
2385 
2386 		/*
2387 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2388 		 * and there are numerous other races with the
2389 		 * ifunit() checks not being atomic with namespace
2390 		 * changes (renames, vmoves, if_attach, etc).
2391 		 */
2392 		ifp->if_flags |= IFF_RENAMING;
2393 
2394 		/* Announce the departure of the interface. */
2395 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2396 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2397 
2398 		log(LOG_INFO, "%s: changing name to '%s'\n",
2399 		    ifp->if_xname, new_name);
2400 
2401 		IF_ADDR_WLOCK(ifp);
2402 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2403 		ifa = ifp->if_addr;
2404 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2405 		namelen = strlen(new_name);
2406 		onamelen = sdl->sdl_nlen;
2407 		/*
2408 		 * Move the address if needed.  This is safe because we
2409 		 * allocate space for a name of length IFNAMSIZ when we
2410 		 * create this in if_attach().
2411 		 */
2412 		if (namelen != onamelen) {
2413 			bcopy(sdl->sdl_data + onamelen,
2414 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2415 		}
2416 		bcopy(new_name, sdl->sdl_data, namelen);
2417 		sdl->sdl_nlen = namelen;
2418 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2419 		bzero(sdl->sdl_data, onamelen);
2420 		while (namelen != 0)
2421 			sdl->sdl_data[--namelen] = 0xff;
2422 		IF_ADDR_WUNLOCK(ifp);
2423 
2424 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2425 		/* Announce the return of the interface. */
2426 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2427 
2428 		ifp->if_flags &= ~IFF_RENAMING;
2429 		break;
2430 
2431 #ifdef VIMAGE
2432 	case SIOCSIFVNET:
2433 		error = priv_check(td, PRIV_NET_SETIFVNET);
2434 		if (error)
2435 			return (error);
2436 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2437 		break;
2438 #endif
2439 
2440 	case SIOCSIFMETRIC:
2441 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2442 		if (error)
2443 			return (error);
2444 		ifp->if_metric = ifr->ifr_metric;
2445 		getmicrotime(&ifp->if_lastchange);
2446 		break;
2447 
2448 	case SIOCSIFPHYS:
2449 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2450 		if (error)
2451 			return (error);
2452 		if (ifp->if_ioctl == NULL)
2453 			return (EOPNOTSUPP);
2454 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2455 		if (error == 0)
2456 			getmicrotime(&ifp->if_lastchange);
2457 		break;
2458 
2459 	case SIOCSIFMTU:
2460 	{
2461 		u_long oldmtu = ifp->if_mtu;
2462 
2463 		error = priv_check(td, PRIV_NET_SETIFMTU);
2464 		if (error)
2465 			return (error);
2466 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2467 			return (EINVAL);
2468 		if (ifp->if_ioctl == NULL)
2469 			return (EOPNOTSUPP);
2470 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2471 		if (error == 0) {
2472 			getmicrotime(&ifp->if_lastchange);
2473 			rt_ifmsg(ifp);
2474 		}
2475 		/*
2476 		 * If the link MTU changed, do network layer specific procedure.
2477 		 */
2478 		if (ifp->if_mtu != oldmtu) {
2479 #ifdef INET6
2480 			nd6_setmtu(ifp);
2481 #endif
2482 			rt_updatemtu(ifp);
2483 		}
2484 		break;
2485 	}
2486 
2487 	case SIOCADDMULTI:
2488 	case SIOCDELMULTI:
2489 		if (cmd == SIOCADDMULTI)
2490 			error = priv_check(td, PRIV_NET_ADDMULTI);
2491 		else
2492 			error = priv_check(td, PRIV_NET_DELMULTI);
2493 		if (error)
2494 			return (error);
2495 
2496 		/* Don't allow group membership on non-multicast interfaces. */
2497 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2498 			return (EOPNOTSUPP);
2499 
2500 		/* Don't let users screw up protocols' entries. */
2501 		if (ifr->ifr_addr.sa_family != AF_LINK)
2502 			return (EINVAL);
2503 
2504 		if (cmd == SIOCADDMULTI) {
2505 			struct ifmultiaddr *ifma;
2506 
2507 			/*
2508 			 * Userland is only permitted to join groups once
2509 			 * via the if_addmulti() KPI, because it cannot hold
2510 			 * struct ifmultiaddr * between calls. It may also
2511 			 * lose a race while we check if the membership
2512 			 * already exists.
2513 			 */
2514 			IF_ADDR_RLOCK(ifp);
2515 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2516 			IF_ADDR_RUNLOCK(ifp);
2517 			if (ifma != NULL)
2518 				error = EADDRINUSE;
2519 			else
2520 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2521 		} else {
2522 			error = if_delmulti(ifp, &ifr->ifr_addr);
2523 		}
2524 		if (error == 0)
2525 			getmicrotime(&ifp->if_lastchange);
2526 		break;
2527 
2528 	case SIOCSIFPHYADDR:
2529 	case SIOCDIFPHYADDR:
2530 #ifdef INET6
2531 	case SIOCSIFPHYADDR_IN6:
2532 #endif
2533 	case SIOCSIFMEDIA:
2534 	case SIOCSIFGENERIC:
2535 		error = priv_check(td, PRIV_NET_HWIOCTL);
2536 		if (error)
2537 			return (error);
2538 		if (ifp->if_ioctl == NULL)
2539 			return (EOPNOTSUPP);
2540 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2541 		if (error == 0)
2542 			getmicrotime(&ifp->if_lastchange);
2543 		break;
2544 
2545 	case SIOCGIFSTATUS:
2546 	case SIOCGIFPSRCADDR:
2547 	case SIOCGIFPDSTADDR:
2548 	case SIOCGIFMEDIA:
2549 	case SIOCGIFXMEDIA:
2550 	case SIOCGIFGENERIC:
2551 		if (ifp->if_ioctl == NULL)
2552 			return (EOPNOTSUPP);
2553 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2554 		break;
2555 
2556 	case SIOCSIFLLADDR:
2557 		error = priv_check(td, PRIV_NET_SETLLADDR);
2558 		if (error)
2559 			return (error);
2560 		error = if_setlladdr(ifp,
2561 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2562 		break;
2563 
2564 	case SIOCAIFGROUP:
2565 	{
2566 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2567 
2568 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2569 		if (error)
2570 			return (error);
2571 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2572 			return (error);
2573 		break;
2574 	}
2575 
2576 	case SIOCGIFGROUP:
2577 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2578 			return (error);
2579 		break;
2580 
2581 	case SIOCDIFGROUP:
2582 	{
2583 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2584 
2585 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2586 		if (error)
2587 			return (error);
2588 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2589 			return (error);
2590 		break;
2591 	}
2592 
2593 	default:
2594 		error = ENOIOCTL;
2595 		break;
2596 	}
2597 	return (error);
2598 }
2599 
2600 #ifdef COMPAT_FREEBSD32
2601 struct ifconf32 {
2602 	int32_t	ifc_len;
2603 	union {
2604 		uint32_t	ifcu_buf;
2605 		uint32_t	ifcu_req;
2606 	} ifc_ifcu;
2607 };
2608 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2609 #endif
2610 
2611 /*
2612  * Interface ioctls.
2613  */
2614 int
2615 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2616 {
2617 	struct ifnet *ifp;
2618 	struct ifreq *ifr;
2619 	int error;
2620 	int oif_flags;
2621 
2622 	CURVNET_SET(so->so_vnet);
2623 	switch (cmd) {
2624 	case SIOCGIFCONF:
2625 		error = ifconf(cmd, data);
2626 		CURVNET_RESTORE();
2627 		return (error);
2628 
2629 #ifdef COMPAT_FREEBSD32
2630 	case SIOCGIFCONF32:
2631 		{
2632 			struct ifconf32 *ifc32;
2633 			struct ifconf ifc;
2634 
2635 			ifc32 = (struct ifconf32 *)data;
2636 			ifc.ifc_len = ifc32->ifc_len;
2637 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2638 
2639 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2640 			CURVNET_RESTORE();
2641 			if (error == 0)
2642 				ifc32->ifc_len = ifc.ifc_len;
2643 			return (error);
2644 		}
2645 #endif
2646 	}
2647 	ifr = (struct ifreq *)data;
2648 
2649 	switch (cmd) {
2650 #ifdef VIMAGE
2651 	case SIOCSIFRVNET:
2652 		error = priv_check(td, PRIV_NET_SETIFVNET);
2653 		if (error == 0)
2654 			error = if_vmove_reclaim(td, ifr->ifr_name,
2655 			    ifr->ifr_jid);
2656 		CURVNET_RESTORE();
2657 		return (error);
2658 #endif
2659 	case SIOCIFCREATE:
2660 	case SIOCIFCREATE2:
2661 		error = priv_check(td, PRIV_NET_IFCREATE);
2662 		if (error == 0)
2663 			error = if_clone_create(ifr->ifr_name,
2664 			    sizeof(ifr->ifr_name),
2665 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2666 		CURVNET_RESTORE();
2667 		return (error);
2668 	case SIOCIFDESTROY:
2669 		error = priv_check(td, PRIV_NET_IFDESTROY);
2670 		if (error == 0)
2671 			error = if_clone_destroy(ifr->ifr_name);
2672 		CURVNET_RESTORE();
2673 		return (error);
2674 
2675 	case SIOCIFGCLONERS:
2676 		error = if_clone_list((struct if_clonereq *)data);
2677 		CURVNET_RESTORE();
2678 		return (error);
2679 	case SIOCGIFGMEMB:
2680 		error = if_getgroupmembers((struct ifgroupreq *)data);
2681 		CURVNET_RESTORE();
2682 		return (error);
2683 #if defined(INET) || defined(INET6)
2684 	case SIOCSVH:
2685 	case SIOCGVH:
2686 		if (carp_ioctl_p == NULL)
2687 			error = EPROTONOSUPPORT;
2688 		else
2689 			error = (*carp_ioctl_p)(ifr, cmd, td);
2690 		CURVNET_RESTORE();
2691 		return (error);
2692 #endif
2693 	}
2694 
2695 	ifp = ifunit_ref(ifr->ifr_name);
2696 	if (ifp == NULL) {
2697 		CURVNET_RESTORE();
2698 		return (ENXIO);
2699 	}
2700 
2701 	error = ifhwioctl(cmd, ifp, data, td);
2702 	if (error != ENOIOCTL) {
2703 		if_rele(ifp);
2704 		CURVNET_RESTORE();
2705 		return (error);
2706 	}
2707 
2708 	oif_flags = ifp->if_flags;
2709 	if (so->so_proto == NULL) {
2710 		if_rele(ifp);
2711 		CURVNET_RESTORE();
2712 		return (EOPNOTSUPP);
2713 	}
2714 
2715 	/*
2716 	 * Pass the request on to the socket control method, and if the
2717 	 * latter returns EOPNOTSUPP, directly to the interface.
2718 	 *
2719 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2720 	 * trust SIOCSIFADDR et al to come from an already privileged
2721 	 * layer, and do not perform any credentials checks or input
2722 	 * validation.
2723 	 */
2724 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2725 	    ifp, td));
2726 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2727 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2728 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2729 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2730 
2731 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2732 #ifdef INET6
2733 		if (ifp->if_flags & IFF_UP)
2734 			in6_if_up(ifp);
2735 #endif
2736 	}
2737 	if_rele(ifp);
2738 	CURVNET_RESTORE();
2739 	return (error);
2740 }
2741 
2742 /*
2743  * The code common to handling reference counted flags,
2744  * e.g., in ifpromisc() and if_allmulti().
2745  * The "pflag" argument can specify a permanent mode flag to check,
2746  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2747  *
2748  * Only to be used on stack-owned flags, not driver-owned flags.
2749  */
2750 static int
2751 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2752 {
2753 	struct ifreq ifr;
2754 	int error;
2755 	int oldflags, oldcount;
2756 
2757 	/* Sanity checks to catch programming errors */
2758 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2759 	    ("%s: setting driver-owned flag %d", __func__, flag));
2760 
2761 	if (onswitch)
2762 		KASSERT(*refcount >= 0,
2763 		    ("%s: increment negative refcount %d for flag %d",
2764 		    __func__, *refcount, flag));
2765 	else
2766 		KASSERT(*refcount > 0,
2767 		    ("%s: decrement non-positive refcount %d for flag %d",
2768 		    __func__, *refcount, flag));
2769 
2770 	/* In case this mode is permanent, just touch refcount */
2771 	if (ifp->if_flags & pflag) {
2772 		*refcount += onswitch ? 1 : -1;
2773 		return (0);
2774 	}
2775 
2776 	/* Save ifnet parameters for if_ioctl() may fail */
2777 	oldcount = *refcount;
2778 	oldflags = ifp->if_flags;
2779 
2780 	/*
2781 	 * See if we aren't the only and touching refcount is enough.
2782 	 * Actually toggle interface flag if we are the first or last.
2783 	 */
2784 	if (onswitch) {
2785 		if ((*refcount)++)
2786 			return (0);
2787 		ifp->if_flags |= flag;
2788 	} else {
2789 		if (--(*refcount))
2790 			return (0);
2791 		ifp->if_flags &= ~flag;
2792 	}
2793 
2794 	/* Call down the driver since we've changed interface flags */
2795 	if (ifp->if_ioctl == NULL) {
2796 		error = EOPNOTSUPP;
2797 		goto recover;
2798 	}
2799 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2800 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2801 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2802 	if (error)
2803 		goto recover;
2804 	/* Notify userland that interface flags have changed */
2805 	rt_ifmsg(ifp);
2806 	return (0);
2807 
2808 recover:
2809 	/* Recover after driver error */
2810 	*refcount = oldcount;
2811 	ifp->if_flags = oldflags;
2812 	return (error);
2813 }
2814 
2815 /*
2816  * Set/clear promiscuous mode on interface ifp based on the truth value
2817  * of pswitch.  The calls are reference counted so that only the first
2818  * "on" request actually has an effect, as does the final "off" request.
2819  * Results are undefined if the "off" and "on" requests are not matched.
2820  */
2821 int
2822 ifpromisc(struct ifnet *ifp, int pswitch)
2823 {
2824 	int error;
2825 	int oldflags = ifp->if_flags;
2826 
2827 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2828 			   &ifp->if_pcount, pswitch);
2829 	/* If promiscuous mode status has changed, log a message */
2830 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2831             log_promisc_mode_change)
2832 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2833 		    ifp->if_xname,
2834 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2835 	return (error);
2836 }
2837 
2838 /*
2839  * Return interface configuration
2840  * of system.  List may be used
2841  * in later ioctl's (above) to get
2842  * other information.
2843  */
2844 /*ARGSUSED*/
2845 static int
2846 ifconf(u_long cmd, caddr_t data)
2847 {
2848 	struct ifconf *ifc = (struct ifconf *)data;
2849 	struct ifnet *ifp;
2850 	struct ifaddr *ifa;
2851 	struct ifreq ifr;
2852 	struct sbuf *sb;
2853 	int error, full = 0, valid_len, max_len;
2854 
2855 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2856 	max_len = MAXPHYS - 1;
2857 
2858 	/* Prevent hostile input from being able to crash the system */
2859 	if (ifc->ifc_len <= 0)
2860 		return (EINVAL);
2861 
2862 again:
2863 	if (ifc->ifc_len <= max_len) {
2864 		max_len = ifc->ifc_len;
2865 		full = 1;
2866 	}
2867 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2868 	max_len = 0;
2869 	valid_len = 0;
2870 
2871 	IFNET_RLOCK();
2872 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2873 		int addrs;
2874 
2875 		/*
2876 		 * Zero the ifr_name buffer to make sure we don't
2877 		 * disclose the contents of the stack.
2878 		 */
2879 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2880 
2881 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2882 		    >= sizeof(ifr.ifr_name)) {
2883 			sbuf_delete(sb);
2884 			IFNET_RUNLOCK();
2885 			return (ENAMETOOLONG);
2886 		}
2887 
2888 		addrs = 0;
2889 		IF_ADDR_RLOCK(ifp);
2890 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2891 			struct sockaddr *sa = ifa->ifa_addr;
2892 
2893 			if (prison_if(curthread->td_ucred, sa) != 0)
2894 				continue;
2895 			addrs++;
2896 			if (sa->sa_len <= sizeof(*sa)) {
2897 				ifr.ifr_addr = *sa;
2898 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2899 				max_len += sizeof(ifr);
2900 			} else {
2901 				sbuf_bcat(sb, &ifr,
2902 				    offsetof(struct ifreq, ifr_addr));
2903 				max_len += offsetof(struct ifreq, ifr_addr);
2904 				sbuf_bcat(sb, sa, sa->sa_len);
2905 				max_len += sa->sa_len;
2906 			}
2907 
2908 			if (sbuf_error(sb) == 0)
2909 				valid_len = sbuf_len(sb);
2910 		}
2911 		IF_ADDR_RUNLOCK(ifp);
2912 		if (addrs == 0) {
2913 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2914 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2915 			max_len += sizeof(ifr);
2916 
2917 			if (sbuf_error(sb) == 0)
2918 				valid_len = sbuf_len(sb);
2919 		}
2920 	}
2921 	IFNET_RUNLOCK();
2922 
2923 	/*
2924 	 * If we didn't allocate enough space (uncommon), try again.  If
2925 	 * we have already allocated as much space as we are allowed,
2926 	 * return what we've got.
2927 	 */
2928 	if (valid_len != max_len && !full) {
2929 		sbuf_delete(sb);
2930 		goto again;
2931 	}
2932 
2933 	ifc->ifc_len = valid_len;
2934 	sbuf_finish(sb);
2935 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2936 	sbuf_delete(sb);
2937 	return (error);
2938 }
2939 
2940 /*
2941  * Just like ifpromisc(), but for all-multicast-reception mode.
2942  */
2943 int
2944 if_allmulti(struct ifnet *ifp, int onswitch)
2945 {
2946 
2947 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2948 }
2949 
2950 struct ifmultiaddr *
2951 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
2952 {
2953 	struct ifmultiaddr *ifma;
2954 
2955 	IF_ADDR_LOCK_ASSERT(ifp);
2956 
2957 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2958 		if (sa->sa_family == AF_LINK) {
2959 			if (sa_dl_equal(ifma->ifma_addr, sa))
2960 				break;
2961 		} else {
2962 			if (sa_equal(ifma->ifma_addr, sa))
2963 				break;
2964 		}
2965 	}
2966 
2967 	return ifma;
2968 }
2969 
2970 /*
2971  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2972  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2973  * the ifnet multicast address list here, so the caller must do that and
2974  * other setup work (such as notifying the device driver).  The reference
2975  * count is initialized to 1.
2976  */
2977 static struct ifmultiaddr *
2978 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2979     int mflags)
2980 {
2981 	struct ifmultiaddr *ifma;
2982 	struct sockaddr *dupsa;
2983 
2984 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2985 	    M_ZERO);
2986 	if (ifma == NULL)
2987 		return (NULL);
2988 
2989 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2990 	if (dupsa == NULL) {
2991 		free(ifma, M_IFMADDR);
2992 		return (NULL);
2993 	}
2994 	bcopy(sa, dupsa, sa->sa_len);
2995 	ifma->ifma_addr = dupsa;
2996 
2997 	ifma->ifma_ifp = ifp;
2998 	ifma->ifma_refcount = 1;
2999 	ifma->ifma_protospec = NULL;
3000 
3001 	if (llsa == NULL) {
3002 		ifma->ifma_lladdr = NULL;
3003 		return (ifma);
3004 	}
3005 
3006 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3007 	if (dupsa == NULL) {
3008 		free(ifma->ifma_addr, M_IFMADDR);
3009 		free(ifma, M_IFMADDR);
3010 		return (NULL);
3011 	}
3012 	bcopy(llsa, dupsa, llsa->sa_len);
3013 	ifma->ifma_lladdr = dupsa;
3014 
3015 	return (ifma);
3016 }
3017 
3018 /*
3019  * if_freemulti: free ifmultiaddr structure and possibly attached related
3020  * addresses.  The caller is responsible for implementing reference
3021  * counting, notifying the driver, handling routing messages, and releasing
3022  * any dependent link layer state.
3023  */
3024 static void
3025 if_freemulti(struct ifmultiaddr *ifma)
3026 {
3027 
3028 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3029 	    ifma->ifma_refcount));
3030 
3031 	if (ifma->ifma_lladdr != NULL)
3032 		free(ifma->ifma_lladdr, M_IFMADDR);
3033 	free(ifma->ifma_addr, M_IFMADDR);
3034 	free(ifma, M_IFMADDR);
3035 }
3036 
3037 /*
3038  * Register an additional multicast address with a network interface.
3039  *
3040  * - If the address is already present, bump the reference count on the
3041  *   address and return.
3042  * - If the address is not link-layer, look up a link layer address.
3043  * - Allocate address structures for one or both addresses, and attach to the
3044  *   multicast address list on the interface.  If automatically adding a link
3045  *   layer address, the protocol address will own a reference to the link
3046  *   layer address, to be freed when it is freed.
3047  * - Notify the network device driver of an addition to the multicast address
3048  *   list.
3049  *
3050  * 'sa' points to caller-owned memory with the desired multicast address.
3051  *
3052  * 'retifma' will be used to return a pointer to the resulting multicast
3053  * address reference, if desired.
3054  */
3055 int
3056 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3057     struct ifmultiaddr **retifma)
3058 {
3059 	struct ifmultiaddr *ifma, *ll_ifma;
3060 	struct sockaddr *llsa;
3061 	struct sockaddr_dl sdl;
3062 	int error;
3063 
3064 	/*
3065 	 * If the address is already present, return a new reference to it;
3066 	 * otherwise, allocate storage and set up a new address.
3067 	 */
3068 	IF_ADDR_WLOCK(ifp);
3069 	ifma = if_findmulti(ifp, sa);
3070 	if (ifma != NULL) {
3071 		ifma->ifma_refcount++;
3072 		if (retifma != NULL)
3073 			*retifma = ifma;
3074 		IF_ADDR_WUNLOCK(ifp);
3075 		return (0);
3076 	}
3077 
3078 	/*
3079 	 * The address isn't already present; resolve the protocol address
3080 	 * into a link layer address, and then look that up, bump its
3081 	 * refcount or allocate an ifma for that also.
3082 	 * Most link layer resolving functions returns address data which
3083 	 * fits inside default sockaddr_dl structure. However callback
3084 	 * can allocate another sockaddr structure, in that case we need to
3085 	 * free it later.
3086 	 */
3087 	llsa = NULL;
3088 	ll_ifma = NULL;
3089 	if (ifp->if_resolvemulti != NULL) {
3090 		/* Provide called function with buffer size information */
3091 		sdl.sdl_len = sizeof(sdl);
3092 		llsa = (struct sockaddr *)&sdl;
3093 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3094 		if (error)
3095 			goto unlock_out;
3096 	}
3097 
3098 	/*
3099 	 * Allocate the new address.  Don't hook it up yet, as we may also
3100 	 * need to allocate a link layer multicast address.
3101 	 */
3102 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3103 	if (ifma == NULL) {
3104 		error = ENOMEM;
3105 		goto free_llsa_out;
3106 	}
3107 
3108 	/*
3109 	 * If a link layer address is found, we'll need to see if it's
3110 	 * already present in the address list, or allocate is as well.
3111 	 * When this block finishes, the link layer address will be on the
3112 	 * list.
3113 	 */
3114 	if (llsa != NULL) {
3115 		ll_ifma = if_findmulti(ifp, llsa);
3116 		if (ll_ifma == NULL) {
3117 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3118 			if (ll_ifma == NULL) {
3119 				--ifma->ifma_refcount;
3120 				if_freemulti(ifma);
3121 				error = ENOMEM;
3122 				goto free_llsa_out;
3123 			}
3124 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3125 			    ifma_link);
3126 		} else
3127 			ll_ifma->ifma_refcount++;
3128 		ifma->ifma_llifma = ll_ifma;
3129 	}
3130 
3131 	/*
3132 	 * We now have a new multicast address, ifma, and possibly a new or
3133 	 * referenced link layer address.  Add the primary address to the
3134 	 * ifnet address list.
3135 	 */
3136 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3137 
3138 	if (retifma != NULL)
3139 		*retifma = ifma;
3140 
3141 	/*
3142 	 * Must generate the message while holding the lock so that 'ifma'
3143 	 * pointer is still valid.
3144 	 */
3145 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3146 	IF_ADDR_WUNLOCK(ifp);
3147 
3148 	/*
3149 	 * We are certain we have added something, so call down to the
3150 	 * interface to let them know about it.
3151 	 */
3152 	if (ifp->if_ioctl != NULL) {
3153 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3154 	}
3155 
3156 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3157 		link_free_sdl(llsa);
3158 
3159 	return (0);
3160 
3161 free_llsa_out:
3162 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3163 		link_free_sdl(llsa);
3164 
3165 unlock_out:
3166 	IF_ADDR_WUNLOCK(ifp);
3167 	return (error);
3168 }
3169 
3170 /*
3171  * Delete a multicast group membership by network-layer group address.
3172  *
3173  * Returns ENOENT if the entry could not be found. If ifp no longer
3174  * exists, results are undefined. This entry point should only be used
3175  * from subsystems which do appropriate locking to hold ifp for the
3176  * duration of the call.
3177  * Network-layer protocol domains must use if_delmulti_ifma().
3178  */
3179 int
3180 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3181 {
3182 	struct ifmultiaddr *ifma;
3183 	int lastref;
3184 #ifdef INVARIANTS
3185 	struct ifnet *oifp;
3186 
3187 	IFNET_RLOCK_NOSLEEP();
3188 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3189 		if (ifp == oifp)
3190 			break;
3191 	if (ifp != oifp)
3192 		ifp = NULL;
3193 	IFNET_RUNLOCK_NOSLEEP();
3194 
3195 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3196 #endif
3197 	if (ifp == NULL)
3198 		return (ENOENT);
3199 
3200 	IF_ADDR_WLOCK(ifp);
3201 	lastref = 0;
3202 	ifma = if_findmulti(ifp, sa);
3203 	if (ifma != NULL)
3204 		lastref = if_delmulti_locked(ifp, ifma, 0);
3205 	IF_ADDR_WUNLOCK(ifp);
3206 
3207 	if (ifma == NULL)
3208 		return (ENOENT);
3209 
3210 	if (lastref && ifp->if_ioctl != NULL) {
3211 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3212 	}
3213 
3214 	return (0);
3215 }
3216 
3217 /*
3218  * Delete all multicast group membership for an interface.
3219  * Should be used to quickly flush all multicast filters.
3220  */
3221 void
3222 if_delallmulti(struct ifnet *ifp)
3223 {
3224 	struct ifmultiaddr *ifma;
3225 	struct ifmultiaddr *next;
3226 
3227 	IF_ADDR_WLOCK(ifp);
3228 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3229 		if_delmulti_locked(ifp, ifma, 0);
3230 	IF_ADDR_WUNLOCK(ifp);
3231 }
3232 
3233 /*
3234  * Delete a multicast group membership by group membership pointer.
3235  * Network-layer protocol domains must use this routine.
3236  *
3237  * It is safe to call this routine if the ifp disappeared.
3238  */
3239 void
3240 if_delmulti_ifma(struct ifmultiaddr *ifma)
3241 {
3242 	struct ifnet *ifp;
3243 	int lastref;
3244 
3245 	ifp = ifma->ifma_ifp;
3246 #ifdef DIAGNOSTIC
3247 	if (ifp == NULL) {
3248 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3249 	} else {
3250 		struct ifnet *oifp;
3251 
3252 		IFNET_RLOCK_NOSLEEP();
3253 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3254 			if (ifp == oifp)
3255 				break;
3256 		if (ifp != oifp) {
3257 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3258 			ifp = NULL;
3259 		}
3260 		IFNET_RUNLOCK_NOSLEEP();
3261 	}
3262 #endif
3263 	/*
3264 	 * If and only if the ifnet instance exists: Acquire the address lock.
3265 	 */
3266 	if (ifp != NULL)
3267 		IF_ADDR_WLOCK(ifp);
3268 
3269 	lastref = if_delmulti_locked(ifp, ifma, 0);
3270 
3271 	if (ifp != NULL) {
3272 		/*
3273 		 * If and only if the ifnet instance exists:
3274 		 *  Release the address lock.
3275 		 *  If the group was left: update the hardware hash filter.
3276 		 */
3277 		IF_ADDR_WUNLOCK(ifp);
3278 		if (lastref && ifp->if_ioctl != NULL) {
3279 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3280 		}
3281 	}
3282 }
3283 
3284 /*
3285  * Perform deletion of network-layer and/or link-layer multicast address.
3286  *
3287  * Return 0 if the reference count was decremented.
3288  * Return 1 if the final reference was released, indicating that the
3289  * hardware hash filter should be reprogrammed.
3290  */
3291 static int
3292 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3293 {
3294 	struct ifmultiaddr *ll_ifma;
3295 
3296 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3297 		KASSERT(ifma->ifma_ifp == ifp,
3298 		    ("%s: inconsistent ifp %p", __func__, ifp));
3299 		IF_ADDR_WLOCK_ASSERT(ifp);
3300 	}
3301 
3302 	ifp = ifma->ifma_ifp;
3303 
3304 	/*
3305 	 * If the ifnet is detaching, null out references to ifnet,
3306 	 * so that upper protocol layers will notice, and not attempt
3307 	 * to obtain locks for an ifnet which no longer exists. The
3308 	 * routing socket announcement must happen before the ifnet
3309 	 * instance is detached from the system.
3310 	 */
3311 	if (detaching) {
3312 #ifdef DIAGNOSTIC
3313 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3314 #endif
3315 		/*
3316 		 * ifp may already be nulled out if we are being reentered
3317 		 * to delete the ll_ifma.
3318 		 */
3319 		if (ifp != NULL) {
3320 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3321 			ifma->ifma_ifp = NULL;
3322 		}
3323 	}
3324 
3325 	if (--ifma->ifma_refcount > 0)
3326 		return 0;
3327 
3328 	/*
3329 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3330 	 * have been associated with it. Release it first if so.
3331 	 */
3332 	ll_ifma = ifma->ifma_llifma;
3333 	if (ll_ifma != NULL) {
3334 		KASSERT(ifma->ifma_lladdr != NULL,
3335 		    ("%s: llifma w/o lladdr", __func__));
3336 		if (detaching)
3337 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3338 		if (--ll_ifma->ifma_refcount == 0) {
3339 			if (ifp != NULL) {
3340 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3341 				    ifma_link);
3342 			}
3343 			if_freemulti(ll_ifma);
3344 		}
3345 	}
3346 
3347 	if (ifp != NULL)
3348 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3349 
3350 	if_freemulti(ifma);
3351 
3352 	/*
3353 	 * The last reference to this instance of struct ifmultiaddr
3354 	 * was released; the hardware should be notified of this change.
3355 	 */
3356 	return 1;
3357 }
3358 
3359 /*
3360  * Set the link layer address on an interface.
3361  *
3362  * At this time we only support certain types of interfaces,
3363  * and we don't allow the length of the address to change.
3364  *
3365  * Set noinline to be dtrace-friendly
3366  */
3367 __noinline int
3368 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3369 {
3370 	struct sockaddr_dl *sdl;
3371 	struct ifaddr *ifa;
3372 	struct ifreq ifr;
3373 
3374 	IF_ADDR_RLOCK(ifp);
3375 	ifa = ifp->if_addr;
3376 	if (ifa == NULL) {
3377 		IF_ADDR_RUNLOCK(ifp);
3378 		return (EINVAL);
3379 	}
3380 	ifa_ref(ifa);
3381 	IF_ADDR_RUNLOCK(ifp);
3382 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3383 	if (sdl == NULL) {
3384 		ifa_free(ifa);
3385 		return (EINVAL);
3386 	}
3387 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3388 		ifa_free(ifa);
3389 		return (EINVAL);
3390 	}
3391 	switch (ifp->if_type) {
3392 	case IFT_ETHER:
3393 	case IFT_FDDI:
3394 	case IFT_XETHER:
3395 	case IFT_ISO88025:
3396 	case IFT_L2VLAN:
3397 	case IFT_BRIDGE:
3398 	case IFT_ARCNET:
3399 	case IFT_IEEE8023ADLAG:
3400 	case IFT_IEEE80211:
3401 		bcopy(lladdr, LLADDR(sdl), len);
3402 		ifa_free(ifa);
3403 		break;
3404 	default:
3405 		ifa_free(ifa);
3406 		return (ENODEV);
3407 	}
3408 
3409 	/*
3410 	 * If the interface is already up, we need
3411 	 * to re-init it in order to reprogram its
3412 	 * address filter.
3413 	 */
3414 	if ((ifp->if_flags & IFF_UP) != 0) {
3415 		if (ifp->if_ioctl) {
3416 			ifp->if_flags &= ~IFF_UP;
3417 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3418 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3419 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3420 			ifp->if_flags |= IFF_UP;
3421 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3422 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3423 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3424 		}
3425 	}
3426 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3427 	return (0);
3428 }
3429 
3430 /*
3431  * Compat function for handling basic encapsulation requests.
3432  * Not converted stacks (FDDI, IB, ..) supports traditional
3433  * output model: ARP (and other similar L2 protocols) are handled
3434  * inside output routine, arpresolve/nd6_resolve() returns MAC
3435  * address instead of full prepend.
3436  *
3437  * This function creates calculated header==MAC for IPv4/IPv6 and
3438  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3439  * address families.
3440  */
3441 static int
3442 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3443 {
3444 
3445 	if (req->rtype != IFENCAP_LL)
3446 		return (EOPNOTSUPP);
3447 
3448 	if (req->bufsize < req->lladdr_len)
3449 		return (ENOMEM);
3450 
3451 	switch (req->family) {
3452 	case AF_INET:
3453 	case AF_INET6:
3454 		break;
3455 	default:
3456 		return (EAFNOSUPPORT);
3457 	}
3458 
3459 	/* Copy lladdr to storage as is */
3460 	memmove(req->buf, req->lladdr, req->lladdr_len);
3461 	req->bufsize = req->lladdr_len;
3462 	req->lladdr_off = 0;
3463 
3464 	return (0);
3465 }
3466 
3467 /*
3468  * The name argument must be a pointer to storage which will last as
3469  * long as the interface does.  For physical devices, the result of
3470  * device_get_name(dev) is a good choice and for pseudo-devices a
3471  * static string works well.
3472  */
3473 void
3474 if_initname(struct ifnet *ifp, const char *name, int unit)
3475 {
3476 	ifp->if_dname = name;
3477 	ifp->if_dunit = unit;
3478 	if (unit != IF_DUNIT_NONE)
3479 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3480 	else
3481 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3482 }
3483 
3484 int
3485 if_printf(struct ifnet *ifp, const char * fmt, ...)
3486 {
3487 	va_list ap;
3488 	int retval;
3489 
3490 	retval = printf("%s: ", ifp->if_xname);
3491 	va_start(ap, fmt);
3492 	retval += vprintf(fmt, ap);
3493 	va_end(ap);
3494 	return (retval);
3495 }
3496 
3497 void
3498 if_start(struct ifnet *ifp)
3499 {
3500 
3501 	(*(ifp)->if_start)(ifp);
3502 }
3503 
3504 /*
3505  * Backwards compatibility interface for drivers
3506  * that have not implemented it
3507  */
3508 static int
3509 if_transmit(struct ifnet *ifp, struct mbuf *m)
3510 {
3511 	int error;
3512 
3513 	IFQ_HANDOFF(ifp, m, error);
3514 	return (error);
3515 }
3516 
3517 static void
3518 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3519 {
3520 
3521 	m_freem(m);
3522 }
3523 
3524 int
3525 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3526 {
3527 	int active = 0;
3528 
3529 	IF_LOCK(ifq);
3530 	if (_IF_QFULL(ifq)) {
3531 		IF_UNLOCK(ifq);
3532 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3533 		m_freem(m);
3534 		return (0);
3535 	}
3536 	if (ifp != NULL) {
3537 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3538 		if (m->m_flags & (M_BCAST|M_MCAST))
3539 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3540 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3541 	}
3542 	_IF_ENQUEUE(ifq, m);
3543 	IF_UNLOCK(ifq);
3544 	if (ifp != NULL && !active)
3545 		(*(ifp)->if_start)(ifp);
3546 	return (1);
3547 }
3548 
3549 void
3550 if_register_com_alloc(u_char type,
3551     if_com_alloc_t *a, if_com_free_t *f)
3552 {
3553 
3554 	KASSERT(if_com_alloc[type] == NULL,
3555 	    ("if_register_com_alloc: %d already registered", type));
3556 	KASSERT(if_com_free[type] == NULL,
3557 	    ("if_register_com_alloc: %d free already registered", type));
3558 
3559 	if_com_alloc[type] = a;
3560 	if_com_free[type] = f;
3561 }
3562 
3563 void
3564 if_deregister_com_alloc(u_char type)
3565 {
3566 
3567 	KASSERT(if_com_alloc[type] != NULL,
3568 	    ("if_deregister_com_alloc: %d not registered", type));
3569 	KASSERT(if_com_free[type] != NULL,
3570 	    ("if_deregister_com_alloc: %d free not registered", type));
3571 	if_com_alloc[type] = NULL;
3572 	if_com_free[type] = NULL;
3573 }
3574 
3575 /* API for driver access to network stack owned ifnet.*/
3576 uint64_t
3577 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3578 {
3579 	uint64_t oldbrate;
3580 
3581 	oldbrate = ifp->if_baudrate;
3582 	ifp->if_baudrate = baudrate;
3583 	return (oldbrate);
3584 }
3585 
3586 uint64_t
3587 if_getbaudrate(if_t ifp)
3588 {
3589 
3590 	return (((struct ifnet *)ifp)->if_baudrate);
3591 }
3592 
3593 int
3594 if_setcapabilities(if_t ifp, int capabilities)
3595 {
3596 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3597 	return (0);
3598 }
3599 
3600 int
3601 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3602 {
3603 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3604 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3605 
3606 	return (0);
3607 }
3608 
3609 int
3610 if_getcapabilities(if_t ifp)
3611 {
3612 	return ((struct ifnet *)ifp)->if_capabilities;
3613 }
3614 
3615 int
3616 if_setcapenable(if_t ifp, int capabilities)
3617 {
3618 	((struct ifnet *)ifp)->if_capenable = capabilities;
3619 	return (0);
3620 }
3621 
3622 int
3623 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3624 {
3625 	if(setcap)
3626 		((struct ifnet *)ifp)->if_capenable |= setcap;
3627 	if(clearcap)
3628 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3629 
3630 	return (0);
3631 }
3632 
3633 const char *
3634 if_getdname(if_t ifp)
3635 {
3636 	return ((struct ifnet *)ifp)->if_dname;
3637 }
3638 
3639 int
3640 if_togglecapenable(if_t ifp, int togglecap)
3641 {
3642 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3643 	return (0);
3644 }
3645 
3646 int
3647 if_getcapenable(if_t ifp)
3648 {
3649 	return ((struct ifnet *)ifp)->if_capenable;
3650 }
3651 
3652 /*
3653  * This is largely undesirable because it ties ifnet to a device, but does
3654  * provide flexiblity for an embedded product vendor. Should be used with
3655  * the understanding that it violates the interface boundaries, and should be
3656  * a last resort only.
3657  */
3658 int
3659 if_setdev(if_t ifp, void *dev)
3660 {
3661 	return (0);
3662 }
3663 
3664 int
3665 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3666 {
3667 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3668 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3669 
3670 	return (0);
3671 }
3672 
3673 int
3674 if_getdrvflags(if_t ifp)
3675 {
3676 	return ((struct ifnet *)ifp)->if_drv_flags;
3677 }
3678 
3679 int
3680 if_setdrvflags(if_t ifp, int flags)
3681 {
3682 	((struct ifnet *)ifp)->if_drv_flags = flags;
3683 	return (0);
3684 }
3685 
3686 
3687 int
3688 if_setflags(if_t ifp, int flags)
3689 {
3690 	((struct ifnet *)ifp)->if_flags = flags;
3691 	return (0);
3692 }
3693 
3694 int
3695 if_setflagbits(if_t ifp, int set, int clear)
3696 {
3697 	((struct ifnet *)ifp)->if_flags |= set;
3698 	((struct ifnet *)ifp)->if_flags &= ~clear;
3699 
3700 	return (0);
3701 }
3702 
3703 int
3704 if_getflags(if_t ifp)
3705 {
3706 	return ((struct ifnet *)ifp)->if_flags;
3707 }
3708 
3709 int
3710 if_clearhwassist(if_t ifp)
3711 {
3712 	((struct ifnet *)ifp)->if_hwassist = 0;
3713 	return (0);
3714 }
3715 
3716 int
3717 if_sethwassistbits(if_t ifp, int toset, int toclear)
3718 {
3719 	((struct ifnet *)ifp)->if_hwassist |= toset;
3720 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3721 
3722 	return (0);
3723 }
3724 
3725 int
3726 if_sethwassist(if_t ifp, int hwassist_bit)
3727 {
3728 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3729 	return (0);
3730 }
3731 
3732 int
3733 if_gethwassist(if_t ifp)
3734 {
3735 	return ((struct ifnet *)ifp)->if_hwassist;
3736 }
3737 
3738 int
3739 if_setmtu(if_t ifp, int mtu)
3740 {
3741 	((struct ifnet *)ifp)->if_mtu = mtu;
3742 	return (0);
3743 }
3744 
3745 int
3746 if_getmtu(if_t ifp)
3747 {
3748 	return ((struct ifnet *)ifp)->if_mtu;
3749 }
3750 
3751 int
3752 if_getmtu_family(if_t ifp, int family)
3753 {
3754 	struct domain *dp;
3755 
3756 	for (dp = domains; dp; dp = dp->dom_next) {
3757 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3758 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3759 	}
3760 
3761 	return (((struct ifnet *)ifp)->if_mtu);
3762 }
3763 
3764 int
3765 if_setsoftc(if_t ifp, void *softc)
3766 {
3767 	((struct ifnet *)ifp)->if_softc = softc;
3768 	return (0);
3769 }
3770 
3771 void *
3772 if_getsoftc(if_t ifp)
3773 {
3774 	return ((struct ifnet *)ifp)->if_softc;
3775 }
3776 
3777 void
3778 if_setrcvif(struct mbuf *m, if_t ifp)
3779 {
3780 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3781 }
3782 
3783 void
3784 if_setvtag(struct mbuf *m, uint16_t tag)
3785 {
3786 	m->m_pkthdr.ether_vtag = tag;
3787 }
3788 
3789 uint16_t
3790 if_getvtag(struct mbuf *m)
3791 {
3792 
3793 	return (m->m_pkthdr.ether_vtag);
3794 }
3795 
3796 int
3797 if_sendq_empty(if_t ifp)
3798 {
3799 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3800 }
3801 
3802 struct ifaddr *
3803 if_getifaddr(if_t ifp)
3804 {
3805 	return ((struct ifnet *)ifp)->if_addr;
3806 }
3807 
3808 int
3809 if_getamcount(if_t ifp)
3810 {
3811 	return ((struct ifnet *)ifp)->if_amcount;
3812 }
3813 
3814 
3815 int
3816 if_setsendqready(if_t ifp)
3817 {
3818 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3819 	return (0);
3820 }
3821 
3822 int
3823 if_setsendqlen(if_t ifp, int tx_desc_count)
3824 {
3825 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3826 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3827 
3828 	return (0);
3829 }
3830 
3831 int
3832 if_vlantrunkinuse(if_t ifp)
3833 {
3834 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3835 }
3836 
3837 int
3838 if_input(if_t ifp, struct mbuf* sendmp)
3839 {
3840 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3841 	return (0);
3842 
3843 }
3844 
3845 /* XXX */
3846 #ifndef ETH_ADDR_LEN
3847 #define ETH_ADDR_LEN 6
3848 #endif
3849 
3850 int
3851 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3852 {
3853 	struct ifmultiaddr *ifma;
3854 	uint8_t *lmta = (uint8_t *)mta;
3855 	int mcnt = 0;
3856 
3857 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3858 		if (ifma->ifma_addr->sa_family != AF_LINK)
3859 			continue;
3860 
3861 		if (mcnt == max)
3862 			break;
3863 
3864 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3865 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3866 		mcnt++;
3867 	}
3868 	*cnt = mcnt;
3869 
3870 	return (0);
3871 }
3872 
3873 int
3874 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3875 {
3876 	int error;
3877 
3878 	if_maddr_rlock(ifp);
3879 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3880 	if_maddr_runlock(ifp);
3881 	return (error);
3882 }
3883 
3884 int
3885 if_multiaddr_count(if_t ifp, int max)
3886 {
3887 	struct ifmultiaddr *ifma;
3888 	int count;
3889 
3890 	count = 0;
3891 	if_maddr_rlock(ifp);
3892 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3893 		if (ifma->ifma_addr->sa_family != AF_LINK)
3894 			continue;
3895 		count++;
3896 		if (count == max)
3897 			break;
3898 	}
3899 	if_maddr_runlock(ifp);
3900 	return (count);
3901 }
3902 
3903 struct mbuf *
3904 if_dequeue(if_t ifp)
3905 {
3906 	struct mbuf *m;
3907 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3908 
3909 	return (m);
3910 }
3911 
3912 int
3913 if_sendq_prepend(if_t ifp, struct mbuf *m)
3914 {
3915 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3916 	return (0);
3917 }
3918 
3919 int
3920 if_setifheaderlen(if_t ifp, int len)
3921 {
3922 	((struct ifnet *)ifp)->if_hdrlen = len;
3923 	return (0);
3924 }
3925 
3926 caddr_t
3927 if_getlladdr(if_t ifp)
3928 {
3929 	return (IF_LLADDR((struct ifnet *)ifp));
3930 }
3931 
3932 void *
3933 if_gethandle(u_char type)
3934 {
3935 	return (if_alloc(type));
3936 }
3937 
3938 void
3939 if_bpfmtap(if_t ifh, struct mbuf *m)
3940 {
3941 	struct ifnet *ifp = (struct ifnet *)ifh;
3942 
3943 	BPF_MTAP(ifp, m);
3944 }
3945 
3946 void
3947 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3948 {
3949 	struct ifnet *ifp = (struct ifnet *)ifh;
3950 
3951 	ETHER_BPF_MTAP(ifp, m);
3952 }
3953 
3954 void
3955 if_vlancap(if_t ifh)
3956 {
3957 	struct ifnet *ifp = (struct ifnet *)ifh;
3958 	VLAN_CAPABILITIES(ifp);
3959 }
3960 
3961 void
3962 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3963 {
3964 	((struct ifnet *)ifp)->if_init = init_fn;
3965 }
3966 
3967 void
3968 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3969 {
3970 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3971 }
3972 
3973 void
3974 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3975 {
3976 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
3977 }
3978 
3979 void
3980 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3981 {
3982 	((struct ifnet *)ifp)->if_transmit = start_fn;
3983 }
3984 
3985 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3986 {
3987 	((struct ifnet *)ifp)->if_qflush = flush_fn;
3988 
3989 }
3990 
3991 void
3992 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
3993 {
3994 
3995 	ifp->if_get_counter = fn;
3996 }
3997 
3998 /* Revisit these - These are inline functions originally. */
3999 int
4000 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4001 {
4002 	return drbr_inuse(ifh, br);
4003 }
4004 
4005 struct mbuf*
4006 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4007 {
4008 	return drbr_dequeue(ifh, br);
4009 }
4010 
4011 int
4012 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4013 {
4014 	return drbr_needs_enqueue(ifh, br);
4015 }
4016 
4017 int
4018 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4019 {
4020 	return drbr_enqueue(ifh, br, m);
4021 
4022 }
4023