xref: /freebsd/sys/net/if.c (revision 4465557ac2231bcaa4f82da5278949c313491970)
1 /*-
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62 
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #include <net/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/if.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif /* INET */
89 #ifdef INET6
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
92 #endif /* INET6 */
93 #endif /* INET || INET6 */
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
100 #endif
101 
102 struct ifindex_entry {
103 	struct  ifnet *ife_ifnet;
104 };
105 
106 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
107 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
108 
109 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
110     &ifqmaxlen, 0, "max send queue size");
111 
112 /* Log link state change events */
113 static int log_link_state_change = 1;
114 
115 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
116 	&log_link_state_change, 0,
117 	"log interface link state change events");
118 
119 /* Interface description */
120 static unsigned int ifdescr_maxlen = 1024;
121 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
122 	&ifdescr_maxlen, 0,
123 	"administrative maximum length for interface description");
124 
125 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
126 
127 /* global sx for non-critical path ifdescr */
128 static struct sx ifdescr_sx;
129 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
130 
131 void	(*bridge_linkstate_p)(struct ifnet *ifp);
132 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
133 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
134 /* These are external hooks for CARP. */
135 void	(*carp_linkstate_p)(struct ifnet *ifp);
136 void	(*carp_demote_adj_p)(int, char *);
137 int	(*carp_master_p)(struct ifaddr *);
138 #if defined(INET) || defined(INET6)
139 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
140 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
141     const struct sockaddr *sa);
142 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
143 int	(*carp_attach_p)(struct ifaddr *, int);
144 void	(*carp_detach_p)(struct ifaddr *);
145 #endif
146 #ifdef INET
147 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
148 #endif
149 #ifdef INET6
150 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
151 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
152     const struct in6_addr *taddr);
153 #endif
154 
155 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
156 
157 /*
158  * XXX: Style; these should be sorted alphabetically, and unprototyped
159  * static functions should be prototyped. Currently they are sorted by
160  * declaration order.
161  */
162 static void	if_attachdomain(void *);
163 static void	if_attachdomain1(struct ifnet *);
164 static int	ifconf(u_long, caddr_t);
165 static void	if_freemulti(struct ifmultiaddr *);
166 static void	if_init(void *);
167 static void	if_grow(void);
168 static void	if_route(struct ifnet *, int flag, int fam);
169 static int	if_setflag(struct ifnet *, int, int, int *, int);
170 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
171 static void	if_unroute(struct ifnet *, int flag, int fam);
172 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
173 static int	if_rtdel(struct radix_node *, void *);
174 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
175 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
176 static void	do_link_state_change(void *, int);
177 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
178 static int	if_getgroupmembers(struct ifgroupreq *);
179 static void	if_delgroups(struct ifnet *);
180 static void	if_attach_internal(struct ifnet *, int);
181 static void	if_detach_internal(struct ifnet *, int);
182 
183 #ifdef INET6
184 /*
185  * XXX: declare here to avoid to include many inet6 related files..
186  * should be more generalized?
187  */
188 extern void	nd6_setmtu(struct ifnet *);
189 #endif
190 
191 VNET_DEFINE(int, if_index);
192 int	ifqmaxlen = IFQ_MAXLEN;
193 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
194 VNET_DEFINE(struct ifgrouphead, ifg_head);
195 
196 static VNET_DEFINE(int, if_indexlim) = 8;
197 
198 /* Table of ifnet by index. */
199 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
200 
201 #define	V_if_indexlim		VNET(if_indexlim)
202 #define	V_ifindex_table		VNET(ifindex_table)
203 
204 /*
205  * The global network interface list (V_ifnet) and related state (such as
206  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
207  * an rwlock.  Either may be acquired shared to stablize the list, but both
208  * must be acquired writable to modify the list.  This model allows us to
209  * both stablize the interface list during interrupt thread processing, but
210  * also to stablize it over long-running ioctls, without introducing priority
211  * inversions and deadlocks.
212  */
213 struct rwlock ifnet_rwlock;
214 struct sx ifnet_sxlock;
215 
216 /*
217  * The allocation of network interfaces is a rather non-atomic affair; we
218  * need to select an index before we are ready to expose the interface for
219  * use, so will use this pointer value to indicate reservation.
220  */
221 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
222 
223 static	if_com_alloc_t *if_com_alloc[256];
224 static	if_com_free_t *if_com_free[256];
225 
226 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
227 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
228 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
229 
230 struct ifnet *
231 ifnet_byindex_locked(u_short idx)
232 {
233 
234 	if (idx > V_if_index)
235 		return (NULL);
236 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
237 		return (NULL);
238 	return (V_ifindex_table[idx].ife_ifnet);
239 }
240 
241 struct ifnet *
242 ifnet_byindex(u_short idx)
243 {
244 	struct ifnet *ifp;
245 
246 	IFNET_RLOCK_NOSLEEP();
247 	ifp = ifnet_byindex_locked(idx);
248 	IFNET_RUNLOCK_NOSLEEP();
249 	return (ifp);
250 }
251 
252 struct ifnet *
253 ifnet_byindex_ref(u_short idx)
254 {
255 	struct ifnet *ifp;
256 
257 	IFNET_RLOCK_NOSLEEP();
258 	ifp = ifnet_byindex_locked(idx);
259 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
260 		IFNET_RUNLOCK_NOSLEEP();
261 		return (NULL);
262 	}
263 	if_ref(ifp);
264 	IFNET_RUNLOCK_NOSLEEP();
265 	return (ifp);
266 }
267 
268 /*
269  * Allocate an ifindex array entry; return 0 on success or an error on
270  * failure.
271  */
272 static int
273 ifindex_alloc_locked(u_short *idxp)
274 {
275 	u_short idx;
276 
277 	IFNET_WLOCK_ASSERT();
278 
279 retry:
280 	/*
281 	 * Try to find an empty slot below V_if_index.  If we fail, take the
282 	 * next slot.
283 	 */
284 	for (idx = 1; idx <= V_if_index; idx++) {
285 		if (V_ifindex_table[idx].ife_ifnet == NULL)
286 			break;
287 	}
288 
289 	/* Catch if_index overflow. */
290 	if (idx >= V_if_indexlim) {
291 		if_grow();
292 		goto retry;
293 	}
294 	if (idx > V_if_index)
295 		V_if_index = idx;
296 	*idxp = idx;
297 	return (0);
298 }
299 
300 static void
301 ifindex_free_locked(u_short idx)
302 {
303 
304 	IFNET_WLOCK_ASSERT();
305 
306 	V_ifindex_table[idx].ife_ifnet = NULL;
307 	while (V_if_index > 0 &&
308 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
309 		V_if_index--;
310 }
311 
312 static void
313 ifindex_free(u_short idx)
314 {
315 
316 	IFNET_WLOCK();
317 	ifindex_free_locked(idx);
318 	IFNET_WUNLOCK();
319 }
320 
321 static void
322 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
323 {
324 
325 	IFNET_WLOCK_ASSERT();
326 
327 	V_ifindex_table[idx].ife_ifnet = ifp;
328 }
329 
330 static void
331 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
332 {
333 
334 	IFNET_WLOCK();
335 	ifnet_setbyindex_locked(idx, ifp);
336 	IFNET_WUNLOCK();
337 }
338 
339 struct ifaddr *
340 ifaddr_byindex(u_short idx)
341 {
342 	struct ifaddr *ifa;
343 
344 	IFNET_RLOCK_NOSLEEP();
345 	ifa = ifnet_byindex_locked(idx)->if_addr;
346 	if (ifa != NULL)
347 		ifa_ref(ifa);
348 	IFNET_RUNLOCK_NOSLEEP();
349 	return (ifa);
350 }
351 
352 /*
353  * Network interface utility routines.
354  *
355  * Routines with ifa_ifwith* names take sockaddr *'s as
356  * parameters.
357  */
358 
359 static void
360 vnet_if_init(const void *unused __unused)
361 {
362 
363 	TAILQ_INIT(&V_ifnet);
364 	TAILQ_INIT(&V_ifg_head);
365 	IFNET_WLOCK();
366 	if_grow();				/* create initial table */
367 	IFNET_WUNLOCK();
368 	vnet_if_clone_init();
369 }
370 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
371     NULL);
372 
373 /* ARGSUSED*/
374 static void
375 if_init(void *dummy __unused)
376 {
377 
378 	IFNET_LOCK_INIT();
379 	if_clone_init();
380 }
381 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
382 
383 
384 #ifdef VIMAGE
385 static void
386 vnet_if_uninit(const void *unused __unused)
387 {
388 
389 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
390 	    "not empty", __func__, __LINE__, &V_ifnet));
391 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
392 	    "not empty", __func__, __LINE__, &V_ifg_head));
393 
394 	free((caddr_t)V_ifindex_table, M_IFNET);
395 }
396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
397     vnet_if_uninit, NULL);
398 #endif
399 
400 static void
401 if_grow(void)
402 {
403 	int oldlim;
404 	u_int n;
405 	struct ifindex_entry *e;
406 
407 	IFNET_WLOCK_ASSERT();
408 	oldlim = V_if_indexlim;
409 	IFNET_WUNLOCK();
410 	n = (oldlim << 1) * sizeof(*e);
411 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
412 	IFNET_WLOCK();
413 	if (V_if_indexlim != oldlim) {
414 		free(e, M_IFNET);
415 		return;
416 	}
417 	if (V_ifindex_table != NULL) {
418 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
419 		free((caddr_t)V_ifindex_table, M_IFNET);
420 	}
421 	V_if_indexlim <<= 1;
422 	V_ifindex_table = e;
423 }
424 
425 /*
426  * Allocate a struct ifnet and an index for an interface.  A layer 2
427  * common structure will also be allocated if an allocation routine is
428  * registered for the passed type.
429  */
430 struct ifnet *
431 if_alloc(u_char type)
432 {
433 	struct ifnet *ifp;
434 	u_short idx;
435 
436 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
437 	IFNET_WLOCK();
438 	if (ifindex_alloc_locked(&idx) != 0) {
439 		IFNET_WUNLOCK();
440 		free(ifp, M_IFNET);
441 		return (NULL);
442 	}
443 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
444 	IFNET_WUNLOCK();
445 	ifp->if_index = idx;
446 	ifp->if_type = type;
447 	ifp->if_alloctype = type;
448 	if (if_com_alloc[type] != NULL) {
449 		ifp->if_l2com = if_com_alloc[type](type, ifp);
450 		if (ifp->if_l2com == NULL) {
451 			free(ifp, M_IFNET);
452 			ifindex_free(idx);
453 			return (NULL);
454 		}
455 	}
456 
457 	IF_ADDR_LOCK_INIT(ifp);
458 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
459 	ifp->if_afdata_initialized = 0;
460 	IF_AFDATA_LOCK_INIT(ifp);
461 	TAILQ_INIT(&ifp->if_addrhead);
462 	TAILQ_INIT(&ifp->if_multiaddrs);
463 	TAILQ_INIT(&ifp->if_groups);
464 #ifdef MAC
465 	mac_ifnet_init(ifp);
466 #endif
467 	ifq_init(&ifp->if_snd, ifp);
468 
469 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
470 	for (int i = 0; i < IFCOUNTERS; i++)
471 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
472 	ifp->if_get_counter = if_get_counter_default;
473 	ifnet_setbyindex(ifp->if_index, ifp);
474 	return (ifp);
475 }
476 
477 /*
478  * Do the actual work of freeing a struct ifnet, and layer 2 common
479  * structure.  This call is made when the last reference to an
480  * interface is released.
481  */
482 static void
483 if_free_internal(struct ifnet *ifp)
484 {
485 
486 	KASSERT((ifp->if_flags & IFF_DYING),
487 	    ("if_free_internal: interface not dying"));
488 
489 	if (if_com_free[ifp->if_alloctype] != NULL)
490 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
491 		    ifp->if_alloctype);
492 
493 #ifdef MAC
494 	mac_ifnet_destroy(ifp);
495 #endif /* MAC */
496 	if (ifp->if_description != NULL)
497 		free(ifp->if_description, M_IFDESCR);
498 	IF_AFDATA_DESTROY(ifp);
499 	IF_ADDR_LOCK_DESTROY(ifp);
500 	ifq_delete(&ifp->if_snd);
501 
502 	for (int i = 0; i < IFCOUNTERS; i++)
503 		counter_u64_free(ifp->if_counters[i]);
504 
505 	free(ifp, M_IFNET);
506 }
507 
508 /*
509  * Deregister an interface and free the associated storage.
510  */
511 void
512 if_free(struct ifnet *ifp)
513 {
514 
515 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
516 
517 	CURVNET_SET_QUIET(ifp->if_vnet);
518 	IFNET_WLOCK();
519 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
520 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
521 
522 	ifindex_free_locked(ifp->if_index);
523 	IFNET_WUNLOCK();
524 
525 	if (refcount_release(&ifp->if_refcount))
526 		if_free_internal(ifp);
527 	CURVNET_RESTORE();
528 }
529 
530 /*
531  * Interfaces to keep an ifnet type-stable despite the possibility of the
532  * driver calling if_free().  If there are additional references, we defer
533  * freeing the underlying data structure.
534  */
535 void
536 if_ref(struct ifnet *ifp)
537 {
538 
539 	/* We don't assert the ifnet list lock here, but arguably should. */
540 	refcount_acquire(&ifp->if_refcount);
541 }
542 
543 void
544 if_rele(struct ifnet *ifp)
545 {
546 
547 	if (!refcount_release(&ifp->if_refcount))
548 		return;
549 	if_free_internal(ifp);
550 }
551 
552 void
553 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
554 {
555 
556 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
557 
558 	if (ifq->ifq_maxlen == 0)
559 		ifq->ifq_maxlen = ifqmaxlen;
560 
561 	ifq->altq_type = 0;
562 	ifq->altq_disc = NULL;
563 	ifq->altq_flags &= ALTQF_CANTCHANGE;
564 	ifq->altq_tbr  = NULL;
565 	ifq->altq_ifp  = ifp;
566 }
567 
568 void
569 ifq_delete(struct ifaltq *ifq)
570 {
571 	mtx_destroy(&ifq->ifq_mtx);
572 }
573 
574 /*
575  * Perform generic interface initalization tasks and attach the interface
576  * to the list of "active" interfaces.  If vmove flag is set on entry
577  * to if_attach_internal(), perform only a limited subset of initialization
578  * tasks, given that we are moving from one vnet to another an ifnet which
579  * has already been fully initialized.
580  *
581  * XXX:
582  *  - The decision to return void and thus require this function to
583  *    succeed is questionable.
584  *  - We should probably do more sanity checking.  For instance we don't
585  *    do anything to insure if_xname is unique or non-empty.
586  */
587 void
588 if_attach(struct ifnet *ifp)
589 {
590 
591 	if_attach_internal(ifp, 0);
592 }
593 
594 /*
595  * Compute the least common TSO limit.
596  */
597 void
598 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
599 {
600 	/*
601 	 * 1) If there is no limit currently, take the limit from
602 	 * the network adapter.
603 	 *
604 	 * 2) If the network adapter has a limit below the current
605 	 * limit, apply it.
606 	 */
607 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
608 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
609 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
610 	}
611 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
612 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
613 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
614 	}
615 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
616 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
617 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
618 	}
619 }
620 
621 /*
622  * Update TSO limit of a network adapter.
623  *
624  * Returns zero if no change. Else non-zero.
625  */
626 int
627 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
628 {
629 	int retval = 0;
630 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
631 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
632 		retval++;
633 	}
634 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
635 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
636 		retval++;
637 	}
638 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
639 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
640 		retval++;
641 	}
642 	return (retval);
643 }
644 
645 static void
646 if_attach_internal(struct ifnet *ifp, int vmove)
647 {
648 	unsigned socksize, ifasize;
649 	int namelen, masklen;
650 	struct sockaddr_dl *sdl;
651 	struct ifaddr *ifa;
652 
653 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
654 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
655 		    ifp->if_xname);
656 
657 #ifdef VIMAGE
658 	ifp->if_vnet = curvnet;
659 	if (ifp->if_home_vnet == NULL)
660 		ifp->if_home_vnet = curvnet;
661 #endif
662 
663 	if_addgroup(ifp, IFG_ALL);
664 
665 	getmicrotime(&ifp->if_lastchange);
666 	ifp->if_epoch = time_uptime;
667 
668 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
669 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
670 	    ("transmit and qflush must both either be set or both be NULL"));
671 	if (ifp->if_transmit == NULL) {
672 		ifp->if_transmit = if_transmit;
673 		ifp->if_qflush = if_qflush;
674 	}
675 
676 	if (!vmove) {
677 #ifdef MAC
678 		mac_ifnet_create(ifp);
679 #endif
680 
681 		/*
682 		 * Create a Link Level name for this device.
683 		 */
684 		namelen = strlen(ifp->if_xname);
685 		/*
686 		 * Always save enough space for any possiable name so we
687 		 * can do a rename in place later.
688 		 */
689 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
690 		socksize = masklen + ifp->if_addrlen;
691 		if (socksize < sizeof(*sdl))
692 			socksize = sizeof(*sdl);
693 		socksize = roundup2(socksize, sizeof(long));
694 		ifasize = sizeof(*ifa) + 2 * socksize;
695 		ifa = ifa_alloc(ifasize, M_WAITOK);
696 		sdl = (struct sockaddr_dl *)(ifa + 1);
697 		sdl->sdl_len = socksize;
698 		sdl->sdl_family = AF_LINK;
699 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
700 		sdl->sdl_nlen = namelen;
701 		sdl->sdl_index = ifp->if_index;
702 		sdl->sdl_type = ifp->if_type;
703 		ifp->if_addr = ifa;
704 		ifa->ifa_ifp = ifp;
705 		ifa->ifa_rtrequest = link_rtrequest;
706 		ifa->ifa_addr = (struct sockaddr *)sdl;
707 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
708 		ifa->ifa_netmask = (struct sockaddr *)sdl;
709 		sdl->sdl_len = masklen;
710 		while (namelen != 0)
711 			sdl->sdl_data[--namelen] = 0xff;
712 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
713 		/* Reliably crash if used uninitialized. */
714 		ifp->if_broadcastaddr = NULL;
715 
716 #if defined(INET) || defined(INET6)
717 		/* Use defaults for TSO, if nothing is set */
718 		if (ifp->if_hw_tsomax == 0 &&
719 		    ifp->if_hw_tsomaxsegcount == 0 &&
720 		    ifp->if_hw_tsomaxsegsize == 0) {
721 			/*
722 			 * The TSO defaults needs to be such that an
723 			 * NFS mbuf list of 35 mbufs totalling just
724 			 * below 64K works and that a chain of mbufs
725 			 * can be defragged into at most 32 segments:
726 			 */
727 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
728 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
729 			ifp->if_hw_tsomaxsegcount = 35;
730 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
731 
732 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
733 			if (ifp->if_capabilities & IFCAP_TSO) {
734 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
735 				    ifp->if_hw_tsomax,
736 				    ifp->if_hw_tsomaxsegcount,
737 				    ifp->if_hw_tsomaxsegsize);
738 			}
739 		}
740 		/*
741 		 * If the "if_hw_tsomax" limit is set, check if it is
742 		 * too small:
743 		 */
744 		KASSERT(ifp->if_hw_tsomax == 0 ||
745 		    ifp->if_hw_tsomax >= (IP_MAXPACKET / 8),
746 		    ("%s: if_hw_tsomax is outside of range", __func__));
747 #endif
748 	}
749 #ifdef VIMAGE
750 	else {
751 		/*
752 		 * Update the interface index in the link layer address
753 		 * of the interface.
754 		 */
755 		for (ifa = ifp->if_addr; ifa != NULL;
756 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
757 			if (ifa->ifa_addr->sa_family == AF_LINK) {
758 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
759 				sdl->sdl_index = ifp->if_index;
760 			}
761 		}
762 	}
763 #endif
764 
765 	IFNET_WLOCK();
766 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
767 #ifdef VIMAGE
768 	curvnet->vnet_ifcnt++;
769 #endif
770 	IFNET_WUNLOCK();
771 
772 	if (domain_init_status >= 2)
773 		if_attachdomain1(ifp);
774 
775 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
776 	if (IS_DEFAULT_VNET(curvnet))
777 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
778 
779 	/* Announce the interface. */
780 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
781 }
782 
783 static void
784 if_attachdomain(void *dummy)
785 {
786 	struct ifnet *ifp;
787 
788 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
789 		if_attachdomain1(ifp);
790 }
791 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
792     if_attachdomain, NULL);
793 
794 static void
795 if_attachdomain1(struct ifnet *ifp)
796 {
797 	struct domain *dp;
798 
799 	/*
800 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
801 	 * cannot lock ifp->if_afdata initialization, entirely.
802 	 */
803 	if (IF_AFDATA_TRYLOCK(ifp) == 0)
804 		return;
805 	if (ifp->if_afdata_initialized >= domain_init_status) {
806 		IF_AFDATA_UNLOCK(ifp);
807 		log(LOG_WARNING, "%s called more than once on %s\n",
808 		    __func__, ifp->if_xname);
809 		return;
810 	}
811 	ifp->if_afdata_initialized = domain_init_status;
812 	IF_AFDATA_UNLOCK(ifp);
813 
814 	/* address family dependent data region */
815 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
816 	for (dp = domains; dp; dp = dp->dom_next) {
817 		if (dp->dom_ifattach)
818 			ifp->if_afdata[dp->dom_family] =
819 			    (*dp->dom_ifattach)(ifp);
820 	}
821 }
822 
823 /*
824  * Remove any unicast or broadcast network addresses from an interface.
825  */
826 void
827 if_purgeaddrs(struct ifnet *ifp)
828 {
829 	struct ifaddr *ifa, *next;
830 
831 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
832 		if (ifa->ifa_addr->sa_family == AF_LINK)
833 			continue;
834 #ifdef INET
835 		/* XXX: Ugly!! ad hoc just for INET */
836 		if (ifa->ifa_addr->sa_family == AF_INET) {
837 			struct ifaliasreq ifr;
838 
839 			bzero(&ifr, sizeof(ifr));
840 			ifr.ifra_addr = *ifa->ifa_addr;
841 			if (ifa->ifa_dstaddr)
842 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
843 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
844 			    NULL) == 0)
845 				continue;
846 		}
847 #endif /* INET */
848 #ifdef INET6
849 		if (ifa->ifa_addr->sa_family == AF_INET6) {
850 			in6_purgeaddr(ifa);
851 			/* ifp_addrhead is already updated */
852 			continue;
853 		}
854 #endif /* INET6 */
855 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
856 		ifa_free(ifa);
857 	}
858 }
859 
860 /*
861  * Remove any multicast network addresses from an interface when an ifnet
862  * is going away.
863  */
864 static void
865 if_purgemaddrs(struct ifnet *ifp)
866 {
867 	struct ifmultiaddr *ifma;
868 	struct ifmultiaddr *next;
869 
870 	IF_ADDR_WLOCK(ifp);
871 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
872 		if_delmulti_locked(ifp, ifma, 1);
873 	IF_ADDR_WUNLOCK(ifp);
874 }
875 
876 /*
877  * Detach an interface, removing it from the list of "active" interfaces.
878  * If vmove flag is set on entry to if_detach_internal(), perform only a
879  * limited subset of cleanup tasks, given that we are moving an ifnet from
880  * one vnet to another, where it must be fully operational.
881  *
882  * XXXRW: There are some significant questions about event ordering, and
883  * how to prevent things from starting to use the interface during detach.
884  */
885 void
886 if_detach(struct ifnet *ifp)
887 {
888 
889 	CURVNET_SET_QUIET(ifp->if_vnet);
890 	if_detach_internal(ifp, 0);
891 	CURVNET_RESTORE();
892 }
893 
894 static void
895 if_detach_internal(struct ifnet *ifp, int vmove)
896 {
897 	struct ifaddr *ifa;
898 	struct radix_node_head	*rnh;
899 	int i, j;
900 	struct domain *dp;
901  	struct ifnet *iter;
902  	int found = 0;
903 
904 	IFNET_WLOCK();
905 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
906 		if (iter == ifp) {
907 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
908 			found = 1;
909 			break;
910 		}
911 #ifdef VIMAGE
912 	if (found)
913 		curvnet->vnet_ifcnt--;
914 #endif
915 	IFNET_WUNLOCK();
916 	if (!found) {
917 		if (vmove)
918 			panic("%s: ifp=%p not on the ifnet tailq %p",
919 			    __func__, ifp, &V_ifnet);
920 		else
921 			return; /* XXX this should panic as well? */
922 	}
923 
924 	/*
925 	 * Remove/wait for pending events.
926 	 */
927 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
928 
929 	/*
930 	 * Remove routes and flush queues.
931 	 */
932 	if_down(ifp);
933 #ifdef ALTQ
934 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
935 		altq_disable(&ifp->if_snd);
936 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
937 		altq_detach(&ifp->if_snd);
938 #endif
939 
940 	if_purgeaddrs(ifp);
941 
942 #ifdef INET
943 	in_ifdetach(ifp);
944 #endif
945 
946 #ifdef INET6
947 	/*
948 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
949 	 * before removing routing entries below, since IPv6 interface direct
950 	 * routes are expected to be removed by the IPv6-specific kernel API.
951 	 * Otherwise, the kernel will detect some inconsistency and bark it.
952 	 */
953 	in6_ifdetach(ifp);
954 #endif
955 	if_purgemaddrs(ifp);
956 
957 	/* Announce that the interface is gone. */
958 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
959 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
960 	if (IS_DEFAULT_VNET(curvnet))
961 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
962 
963 	if (!vmove) {
964 		/*
965 		 * Prevent further calls into the device driver via ifnet.
966 		 */
967 		if_dead(ifp);
968 
969 		/*
970 		 * Remove link ifaddr pointer and maybe decrement if_index.
971 		 * Clean up all addresses.
972 		 */
973 		ifp->if_addr = NULL;
974 
975 		/* We can now free link ifaddr. */
976 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
977 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
978 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
979 			ifa_free(ifa);
980 		}
981 	}
982 
983 	/*
984 	 * Delete all remaining routes using this interface
985 	 * Unfortuneatly the only way to do this is to slog through
986 	 * the entire routing table looking for routes which point
987 	 * to this interface...oh well...
988 	 */
989 	for (i = 1; i <= AF_MAX; i++) {
990 		for (j = 0; j < rt_numfibs; j++) {
991 			rnh = rt_tables_get_rnh(j, i);
992 			if (rnh == NULL)
993 				continue;
994 			RADIX_NODE_HEAD_LOCK(rnh);
995 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
996 			RADIX_NODE_HEAD_UNLOCK(rnh);
997 		}
998 	}
999 
1000 	if_delgroups(ifp);
1001 
1002 	/*
1003 	 * We cannot hold the lock over dom_ifdetach calls as they might
1004 	 * sleep, for example trying to drain a callout, thus open up the
1005 	 * theoretical race with re-attaching.
1006 	 */
1007 	IF_AFDATA_LOCK(ifp);
1008 	i = ifp->if_afdata_initialized;
1009 	ifp->if_afdata_initialized = 0;
1010 	IF_AFDATA_UNLOCK(ifp);
1011 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1012 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1013 			(*dp->dom_ifdetach)(ifp,
1014 			    ifp->if_afdata[dp->dom_family]);
1015 	}
1016 }
1017 
1018 #ifdef VIMAGE
1019 /*
1020  * if_vmove() performs a limited version of if_detach() in current
1021  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1022  * An attempt is made to shrink if_index in current vnet, find an
1023  * unused if_index in target vnet and calls if_grow() if necessary,
1024  * and finally find an unused if_xname for the target vnet.
1025  */
1026 void
1027 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1028 {
1029 	u_short idx;
1030 
1031 	/*
1032 	 * Detach from current vnet, but preserve LLADDR info, do not
1033 	 * mark as dead etc. so that the ifnet can be reattached later.
1034 	 */
1035 	if_detach_internal(ifp, 1);
1036 
1037 	/*
1038 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1039 	 * the if_index for that vnet if possible.
1040 	 *
1041 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1042 	 * or we'd lock on one vnet and unlock on another.
1043 	 */
1044 	IFNET_WLOCK();
1045 	ifindex_free_locked(ifp->if_index);
1046 	IFNET_WUNLOCK();
1047 
1048 	/*
1049 	 * Perform interface-specific reassignment tasks, if provided by
1050 	 * the driver.
1051 	 */
1052 	if (ifp->if_reassign != NULL)
1053 		ifp->if_reassign(ifp, new_vnet, NULL);
1054 
1055 	/*
1056 	 * Switch to the context of the target vnet.
1057 	 */
1058 	CURVNET_SET_QUIET(new_vnet);
1059 
1060 	IFNET_WLOCK();
1061 	if (ifindex_alloc_locked(&idx) != 0) {
1062 		IFNET_WUNLOCK();
1063 		panic("if_index overflow");
1064 	}
1065 	ifp->if_index = idx;
1066 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1067 	IFNET_WUNLOCK();
1068 
1069 	if_attach_internal(ifp, 1);
1070 
1071 	CURVNET_RESTORE();
1072 }
1073 
1074 /*
1075  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1076  */
1077 static int
1078 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1079 {
1080 	struct prison *pr;
1081 	struct ifnet *difp;
1082 
1083 	/* Try to find the prison within our visibility. */
1084 	sx_slock(&allprison_lock);
1085 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1086 	sx_sunlock(&allprison_lock);
1087 	if (pr == NULL)
1088 		return (ENXIO);
1089 	prison_hold_locked(pr);
1090 	mtx_unlock(&pr->pr_mtx);
1091 
1092 	/* Do not try to move the iface from and to the same prison. */
1093 	if (pr->pr_vnet == ifp->if_vnet) {
1094 		prison_free(pr);
1095 		return (EEXIST);
1096 	}
1097 
1098 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1099 	/* XXX Lock interfaces to avoid races. */
1100 	CURVNET_SET_QUIET(pr->pr_vnet);
1101 	difp = ifunit(ifname);
1102 	CURVNET_RESTORE();
1103 	if (difp != NULL) {
1104 		prison_free(pr);
1105 		return (EEXIST);
1106 	}
1107 
1108 	/* Move the interface into the child jail/vnet. */
1109 	if_vmove(ifp, pr->pr_vnet);
1110 
1111 	/* Report the new if_xname back to the userland. */
1112 	sprintf(ifname, "%s", ifp->if_xname);
1113 
1114 	prison_free(pr);
1115 	return (0);
1116 }
1117 
1118 static int
1119 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1120 {
1121 	struct prison *pr;
1122 	struct vnet *vnet_dst;
1123 	struct ifnet *ifp;
1124 
1125 	/* Try to find the prison within our visibility. */
1126 	sx_slock(&allprison_lock);
1127 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1128 	sx_sunlock(&allprison_lock);
1129 	if (pr == NULL)
1130 		return (ENXIO);
1131 	prison_hold_locked(pr);
1132 	mtx_unlock(&pr->pr_mtx);
1133 
1134 	/* Make sure the named iface exists in the source prison/vnet. */
1135 	CURVNET_SET(pr->pr_vnet);
1136 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1137 	if (ifp == NULL) {
1138 		CURVNET_RESTORE();
1139 		prison_free(pr);
1140 		return (ENXIO);
1141 	}
1142 
1143 	/* Do not try to move the iface from and to the same prison. */
1144 	vnet_dst = TD_TO_VNET(td);
1145 	if (vnet_dst == ifp->if_vnet) {
1146 		CURVNET_RESTORE();
1147 		prison_free(pr);
1148 		return (EEXIST);
1149 	}
1150 
1151 	/* Get interface back from child jail/vnet. */
1152 	if_vmove(ifp, vnet_dst);
1153 	CURVNET_RESTORE();
1154 
1155 	/* Report the new if_xname back to the userland. */
1156 	sprintf(ifname, "%s", ifp->if_xname);
1157 
1158 	prison_free(pr);
1159 	return (0);
1160 }
1161 #endif /* VIMAGE */
1162 
1163 /*
1164  * Add a group to an interface
1165  */
1166 int
1167 if_addgroup(struct ifnet *ifp, const char *groupname)
1168 {
1169 	struct ifg_list		*ifgl;
1170 	struct ifg_group	*ifg = NULL;
1171 	struct ifg_member	*ifgm;
1172 	int 			 new = 0;
1173 
1174 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1175 	    groupname[strlen(groupname) - 1] <= '9')
1176 		return (EINVAL);
1177 
1178 	IFNET_WLOCK();
1179 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1180 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1181 			IFNET_WUNLOCK();
1182 			return (EEXIST);
1183 		}
1184 
1185 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1186 	    M_NOWAIT)) == NULL) {
1187 	    	IFNET_WUNLOCK();
1188 		return (ENOMEM);
1189 	}
1190 
1191 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1192 	    M_TEMP, M_NOWAIT)) == NULL) {
1193 		free(ifgl, M_TEMP);
1194 		IFNET_WUNLOCK();
1195 		return (ENOMEM);
1196 	}
1197 
1198 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1199 		if (!strcmp(ifg->ifg_group, groupname))
1200 			break;
1201 
1202 	if (ifg == NULL) {
1203 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1204 		    M_TEMP, M_NOWAIT)) == NULL) {
1205 			free(ifgl, M_TEMP);
1206 			free(ifgm, M_TEMP);
1207 			IFNET_WUNLOCK();
1208 			return (ENOMEM);
1209 		}
1210 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1211 		ifg->ifg_refcnt = 0;
1212 		TAILQ_INIT(&ifg->ifg_members);
1213 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1214 		new = 1;
1215 	}
1216 
1217 	ifg->ifg_refcnt++;
1218 	ifgl->ifgl_group = ifg;
1219 	ifgm->ifgm_ifp = ifp;
1220 
1221 	IF_ADDR_WLOCK(ifp);
1222 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1223 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1224 	IF_ADDR_WUNLOCK(ifp);
1225 
1226 	IFNET_WUNLOCK();
1227 
1228 	if (new)
1229 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1230 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1231 
1232 	return (0);
1233 }
1234 
1235 /*
1236  * Remove a group from an interface
1237  */
1238 int
1239 if_delgroup(struct ifnet *ifp, const char *groupname)
1240 {
1241 	struct ifg_list		*ifgl;
1242 	struct ifg_member	*ifgm;
1243 
1244 	IFNET_WLOCK();
1245 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1246 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1247 			break;
1248 	if (ifgl == NULL) {
1249 		IFNET_WUNLOCK();
1250 		return (ENOENT);
1251 	}
1252 
1253 	IF_ADDR_WLOCK(ifp);
1254 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1255 	IF_ADDR_WUNLOCK(ifp);
1256 
1257 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1258 		if (ifgm->ifgm_ifp == ifp)
1259 			break;
1260 
1261 	if (ifgm != NULL) {
1262 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1263 		free(ifgm, M_TEMP);
1264 	}
1265 
1266 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1267 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1268 		IFNET_WUNLOCK();
1269 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1270 		free(ifgl->ifgl_group, M_TEMP);
1271 	} else
1272 		IFNET_WUNLOCK();
1273 
1274 	free(ifgl, M_TEMP);
1275 
1276 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1277 
1278 	return (0);
1279 }
1280 
1281 /*
1282  * Remove an interface from all groups
1283  */
1284 static void
1285 if_delgroups(struct ifnet *ifp)
1286 {
1287 	struct ifg_list		*ifgl;
1288 	struct ifg_member	*ifgm;
1289 	char groupname[IFNAMSIZ];
1290 
1291 	IFNET_WLOCK();
1292 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1293 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1294 
1295 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1296 
1297 		IF_ADDR_WLOCK(ifp);
1298 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1299 		IF_ADDR_WUNLOCK(ifp);
1300 
1301 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1302 			if (ifgm->ifgm_ifp == ifp)
1303 				break;
1304 
1305 		if (ifgm != NULL) {
1306 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1307 			    ifgm_next);
1308 			free(ifgm, M_TEMP);
1309 		}
1310 
1311 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1312 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1313 			IFNET_WUNLOCK();
1314 			EVENTHANDLER_INVOKE(group_detach_event,
1315 			    ifgl->ifgl_group);
1316 			free(ifgl->ifgl_group, M_TEMP);
1317 		} else
1318 			IFNET_WUNLOCK();
1319 
1320 		free(ifgl, M_TEMP);
1321 
1322 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1323 
1324 		IFNET_WLOCK();
1325 	}
1326 	IFNET_WUNLOCK();
1327 }
1328 
1329 /*
1330  * Stores all groups from an interface in memory pointed
1331  * to by data
1332  */
1333 static int
1334 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1335 {
1336 	int			 len, error;
1337 	struct ifg_list		*ifgl;
1338 	struct ifg_req		 ifgrq, *ifgp;
1339 	struct ifgroupreq	*ifgr = data;
1340 
1341 	if (ifgr->ifgr_len == 0) {
1342 		IF_ADDR_RLOCK(ifp);
1343 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1344 			ifgr->ifgr_len += sizeof(struct ifg_req);
1345 		IF_ADDR_RUNLOCK(ifp);
1346 		return (0);
1347 	}
1348 
1349 	len = ifgr->ifgr_len;
1350 	ifgp = ifgr->ifgr_groups;
1351 	/* XXX: wire */
1352 	IF_ADDR_RLOCK(ifp);
1353 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1354 		if (len < sizeof(ifgrq)) {
1355 			IF_ADDR_RUNLOCK(ifp);
1356 			return (EINVAL);
1357 		}
1358 		bzero(&ifgrq, sizeof ifgrq);
1359 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1360 		    sizeof(ifgrq.ifgrq_group));
1361 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1362 		    	IF_ADDR_RUNLOCK(ifp);
1363 			return (error);
1364 		}
1365 		len -= sizeof(ifgrq);
1366 		ifgp++;
1367 	}
1368 	IF_ADDR_RUNLOCK(ifp);
1369 
1370 	return (0);
1371 }
1372 
1373 /*
1374  * Stores all members of a group in memory pointed to by data
1375  */
1376 static int
1377 if_getgroupmembers(struct ifgroupreq *data)
1378 {
1379 	struct ifgroupreq	*ifgr = data;
1380 	struct ifg_group	*ifg;
1381 	struct ifg_member	*ifgm;
1382 	struct ifg_req		 ifgrq, *ifgp;
1383 	int			 len, error;
1384 
1385 	IFNET_RLOCK();
1386 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1387 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1388 			break;
1389 	if (ifg == NULL) {
1390 		IFNET_RUNLOCK();
1391 		return (ENOENT);
1392 	}
1393 
1394 	if (ifgr->ifgr_len == 0) {
1395 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1396 			ifgr->ifgr_len += sizeof(ifgrq);
1397 		IFNET_RUNLOCK();
1398 		return (0);
1399 	}
1400 
1401 	len = ifgr->ifgr_len;
1402 	ifgp = ifgr->ifgr_groups;
1403 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1404 		if (len < sizeof(ifgrq)) {
1405 			IFNET_RUNLOCK();
1406 			return (EINVAL);
1407 		}
1408 		bzero(&ifgrq, sizeof ifgrq);
1409 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1410 		    sizeof(ifgrq.ifgrq_member));
1411 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1412 			IFNET_RUNLOCK();
1413 			return (error);
1414 		}
1415 		len -= sizeof(ifgrq);
1416 		ifgp++;
1417 	}
1418 	IFNET_RUNLOCK();
1419 
1420 	return (0);
1421 }
1422 
1423 /*
1424  * Delete Routes for a Network Interface
1425  *
1426  * Called for each routing entry via the rnh->rnh_walktree() call above
1427  * to delete all route entries referencing a detaching network interface.
1428  *
1429  * Arguments:
1430  *	rn	pointer to node in the routing table
1431  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1432  *
1433  * Returns:
1434  *	0	successful
1435  *	errno	failed - reason indicated
1436  *
1437  */
1438 static int
1439 if_rtdel(struct radix_node *rn, void *arg)
1440 {
1441 	struct rtentry	*rt = (struct rtentry *)rn;
1442 	struct ifnet	*ifp = arg;
1443 	int		err;
1444 
1445 	if (rt->rt_ifp == ifp) {
1446 
1447 		/*
1448 		 * Protect (sorta) against walktree recursion problems
1449 		 * with cloned routes
1450 		 */
1451 		if ((rt->rt_flags & RTF_UP) == 0)
1452 			return (0);
1453 
1454 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1455 				rt_mask(rt),
1456 				rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1457 				(struct rtentry **) NULL, rt->rt_fibnum);
1458 		if (err) {
1459 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1460 		}
1461 	}
1462 
1463 	return (0);
1464 }
1465 
1466 /*
1467  * Return counter values from counter(9)s stored in ifnet.
1468  */
1469 uint64_t
1470 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1471 {
1472 
1473 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1474 
1475 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1476 }
1477 
1478 /*
1479  * Increase an ifnet counter. Usually used for counters shared
1480  * between the stack and a driver, but function supports them all.
1481  */
1482 void
1483 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1484 {
1485 
1486 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1487 
1488 	counter_u64_add(ifp->if_counters[cnt], inc);
1489 }
1490 
1491 /*
1492  * Copy data from ifnet to userland API structure if_data.
1493  */
1494 void
1495 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1496 {
1497 
1498 	ifd->ifi_type = ifp->if_type;
1499 	ifd->ifi_physical = 0;
1500 	ifd->ifi_addrlen = ifp->if_addrlen;
1501 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1502 	ifd->ifi_link_state = ifp->if_link_state;
1503 	ifd->ifi_vhid = 0;
1504 	ifd->ifi_datalen = sizeof(struct if_data);
1505 	ifd->ifi_mtu = ifp->if_mtu;
1506 	ifd->ifi_metric = ifp->if_metric;
1507 	ifd->ifi_baudrate = ifp->if_baudrate;
1508 	ifd->ifi_hwassist = ifp->if_hwassist;
1509 	ifd->ifi_epoch = ifp->if_epoch;
1510 	ifd->ifi_lastchange = ifp->if_lastchange;
1511 
1512 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1513 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1514 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1515 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1516 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1517 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1518 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1519 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1520 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1521 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1522 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1523 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1524 }
1525 
1526 /*
1527  * Wrapper functions for struct ifnet address list locking macros.  These are
1528  * used by kernel modules to avoid encoding programming interface or binary
1529  * interface assumptions that may be violated when kernel-internal locking
1530  * approaches change.
1531  */
1532 void
1533 if_addr_rlock(struct ifnet *ifp)
1534 {
1535 
1536 	IF_ADDR_RLOCK(ifp);
1537 }
1538 
1539 void
1540 if_addr_runlock(struct ifnet *ifp)
1541 {
1542 
1543 	IF_ADDR_RUNLOCK(ifp);
1544 }
1545 
1546 void
1547 if_maddr_rlock(if_t ifp)
1548 {
1549 
1550 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1551 }
1552 
1553 void
1554 if_maddr_runlock(if_t ifp)
1555 {
1556 
1557 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1558 }
1559 
1560 /*
1561  * Initialization, destruction and refcounting functions for ifaddrs.
1562  */
1563 struct ifaddr *
1564 ifa_alloc(size_t size, int flags)
1565 {
1566 	struct ifaddr *ifa;
1567 
1568 	KASSERT(size >= sizeof(struct ifaddr),
1569 	    ("%s: invalid size %zu", __func__, size));
1570 
1571 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1572 	if (ifa == NULL)
1573 		return (NULL);
1574 
1575 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1576 		goto fail;
1577 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1578 		goto fail;
1579 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1580 		goto fail;
1581 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1582 		goto fail;
1583 
1584 	refcount_init(&ifa->ifa_refcnt, 1);
1585 
1586 	return (ifa);
1587 
1588 fail:
1589 	/* free(NULL) is okay */
1590 	counter_u64_free(ifa->ifa_opackets);
1591 	counter_u64_free(ifa->ifa_ipackets);
1592 	counter_u64_free(ifa->ifa_obytes);
1593 	counter_u64_free(ifa->ifa_ibytes);
1594 	free(ifa, M_IFADDR);
1595 
1596 	return (NULL);
1597 }
1598 
1599 void
1600 ifa_ref(struct ifaddr *ifa)
1601 {
1602 
1603 	refcount_acquire(&ifa->ifa_refcnt);
1604 }
1605 
1606 void
1607 ifa_free(struct ifaddr *ifa)
1608 {
1609 
1610 	if (refcount_release(&ifa->ifa_refcnt)) {
1611 		counter_u64_free(ifa->ifa_opackets);
1612 		counter_u64_free(ifa->ifa_ipackets);
1613 		counter_u64_free(ifa->ifa_obytes);
1614 		counter_u64_free(ifa->ifa_ibytes);
1615 		free(ifa, M_IFADDR);
1616 	}
1617 }
1618 
1619 int
1620 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1621 {
1622 	int error = 0;
1623 	struct rtentry *rt = NULL;
1624 	struct rt_addrinfo info;
1625 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1626 
1627 	bzero(&info, sizeof(info));
1628 	info.rti_ifp = V_loif;
1629 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1630 	info.rti_info[RTAX_DST] = ia;
1631 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1632 	error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1633 
1634 	if (error == 0 && rt != NULL) {
1635 		RT_LOCK(rt);
1636 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1637 			ifa->ifa_ifp->if_type;
1638 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1639 			ifa->ifa_ifp->if_index;
1640 		RT_REMREF(rt);
1641 		RT_UNLOCK(rt);
1642 	} else if (error != 0)
1643 		log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1644 
1645 	return (error);
1646 }
1647 
1648 int
1649 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1650 {
1651 	int error = 0;
1652 	struct rt_addrinfo info;
1653 	struct sockaddr_dl null_sdl;
1654 
1655 	bzero(&null_sdl, sizeof(null_sdl));
1656 	null_sdl.sdl_len = sizeof(null_sdl);
1657 	null_sdl.sdl_family = AF_LINK;
1658 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1659 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1660 	bzero(&info, sizeof(info));
1661 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1662 	info.rti_info[RTAX_DST] = ia;
1663 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1664 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1665 
1666 	if (error != 0)
1667 		log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1668 
1669 	return (error);
1670 }
1671 
1672 int
1673 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib)
1674 {
1675 	struct rtentry *rt;
1676 
1677 	rt = rtalloc1_fib(sa, 0, 0, fib);
1678 	if (rt == NULL) {
1679 		log(LOG_DEBUG, "%s: fail", __func__);
1680 		return (EHOSTUNREACH);
1681 	}
1682 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1683 	    ifa->ifa_ifp->if_type;
1684 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1685 	    ifa->ifa_ifp->if_index;
1686 	RTFREE_LOCKED(rt);
1687 
1688 	return (0);
1689 }
1690 
1691 /*
1692  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1693  * structs used to represent other address families, it is necessary
1694  * to perform a different comparison.
1695  */
1696 
1697 #define	sa_dl_equal(a1, a2)	\
1698 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1699 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1700 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1701 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1702 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1703 
1704 /*
1705  * Locate an interface based on a complete address.
1706  */
1707 /*ARGSUSED*/
1708 static struct ifaddr *
1709 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1710 {
1711 	struct ifnet *ifp;
1712 	struct ifaddr *ifa;
1713 
1714 	IFNET_RLOCK_NOSLEEP();
1715 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1716 		IF_ADDR_RLOCK(ifp);
1717 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1718 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1719 				continue;
1720 			if (sa_equal(addr, ifa->ifa_addr)) {
1721 				if (getref)
1722 					ifa_ref(ifa);
1723 				IF_ADDR_RUNLOCK(ifp);
1724 				goto done;
1725 			}
1726 			/* IP6 doesn't have broadcast */
1727 			if ((ifp->if_flags & IFF_BROADCAST) &&
1728 			    ifa->ifa_broadaddr &&
1729 			    ifa->ifa_broadaddr->sa_len != 0 &&
1730 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1731 				if (getref)
1732 					ifa_ref(ifa);
1733 				IF_ADDR_RUNLOCK(ifp);
1734 				goto done;
1735 			}
1736 		}
1737 		IF_ADDR_RUNLOCK(ifp);
1738 	}
1739 	ifa = NULL;
1740 done:
1741 	IFNET_RUNLOCK_NOSLEEP();
1742 	return (ifa);
1743 }
1744 
1745 struct ifaddr *
1746 ifa_ifwithaddr(struct sockaddr *addr)
1747 {
1748 
1749 	return (ifa_ifwithaddr_internal(addr, 1));
1750 }
1751 
1752 int
1753 ifa_ifwithaddr_check(struct sockaddr *addr)
1754 {
1755 
1756 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1757 }
1758 
1759 /*
1760  * Locate an interface based on the broadcast address.
1761  */
1762 /* ARGSUSED */
1763 struct ifaddr *
1764 ifa_ifwithbroadaddr(struct sockaddr *addr, int fibnum)
1765 {
1766 	struct ifnet *ifp;
1767 	struct ifaddr *ifa;
1768 
1769 	IFNET_RLOCK_NOSLEEP();
1770 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1771 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1772 			continue;
1773 		IF_ADDR_RLOCK(ifp);
1774 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1775 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1776 				continue;
1777 			if ((ifp->if_flags & IFF_BROADCAST) &&
1778 			    ifa->ifa_broadaddr &&
1779 			    ifa->ifa_broadaddr->sa_len != 0 &&
1780 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1781 				ifa_ref(ifa);
1782 				IF_ADDR_RUNLOCK(ifp);
1783 				goto done;
1784 			}
1785 		}
1786 		IF_ADDR_RUNLOCK(ifp);
1787 	}
1788 	ifa = NULL;
1789 done:
1790 	IFNET_RUNLOCK_NOSLEEP();
1791 	return (ifa);
1792 }
1793 
1794 /*
1795  * Locate the point to point interface with a given destination address.
1796  */
1797 /*ARGSUSED*/
1798 struct ifaddr *
1799 ifa_ifwithdstaddr(struct sockaddr *addr, int fibnum)
1800 {
1801 	struct ifnet *ifp;
1802 	struct ifaddr *ifa;
1803 
1804 	IFNET_RLOCK_NOSLEEP();
1805 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1806 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1807 			continue;
1808 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1809 			continue;
1810 		IF_ADDR_RLOCK(ifp);
1811 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1812 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1813 				continue;
1814 			if (ifa->ifa_dstaddr != NULL &&
1815 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1816 				ifa_ref(ifa);
1817 				IF_ADDR_RUNLOCK(ifp);
1818 				goto done;
1819 			}
1820 		}
1821 		IF_ADDR_RUNLOCK(ifp);
1822 	}
1823 	ifa = NULL;
1824 done:
1825 	IFNET_RUNLOCK_NOSLEEP();
1826 	return (ifa);
1827 }
1828 
1829 /*
1830  * Find an interface on a specific network.  If many, choice
1831  * is most specific found.
1832  */
1833 struct ifaddr *
1834 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp, int fibnum)
1835 {
1836 	struct ifnet *ifp;
1837 	struct ifaddr *ifa;
1838 	struct ifaddr *ifa_maybe = NULL;
1839 	u_int af = addr->sa_family;
1840 	char *addr_data = addr->sa_data, *cplim;
1841 
1842 	/*
1843 	 * AF_LINK addresses can be looked up directly by their index number,
1844 	 * so do that if we can.
1845 	 */
1846 	if (af == AF_LINK) {
1847 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1848 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1849 		return (ifaddr_byindex(sdl->sdl_index));
1850 	}
1851 
1852 	/*
1853 	 * Scan though each interface, looking for ones that have addresses
1854 	 * in this address family and the requested fib.  Maintain a reference
1855 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1856 	 * kept it stable when we move onto the next interface.
1857 	 */
1858 	IFNET_RLOCK_NOSLEEP();
1859 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1860 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1861 			continue;
1862 		IF_ADDR_RLOCK(ifp);
1863 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1864 			char *cp, *cp2, *cp3;
1865 
1866 			if (ifa->ifa_addr->sa_family != af)
1867 next:				continue;
1868 			if (af == AF_INET &&
1869 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1870 				/*
1871 				 * This is a bit broken as it doesn't
1872 				 * take into account that the remote end may
1873 				 * be a single node in the network we are
1874 				 * looking for.
1875 				 * The trouble is that we don't know the
1876 				 * netmask for the remote end.
1877 				 */
1878 				if (ifa->ifa_dstaddr != NULL &&
1879 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1880 					ifa_ref(ifa);
1881 					IF_ADDR_RUNLOCK(ifp);
1882 					goto done;
1883 				}
1884 			} else {
1885 				/*
1886 				 * Scan all the bits in the ifa's address.
1887 				 * If a bit dissagrees with what we are
1888 				 * looking for, mask it with the netmask
1889 				 * to see if it really matters.
1890 				 * (A byte at a time)
1891 				 */
1892 				if (ifa->ifa_netmask == 0)
1893 					continue;
1894 				cp = addr_data;
1895 				cp2 = ifa->ifa_addr->sa_data;
1896 				cp3 = ifa->ifa_netmask->sa_data;
1897 				cplim = ifa->ifa_netmask->sa_len
1898 					+ (char *)ifa->ifa_netmask;
1899 				while (cp3 < cplim)
1900 					if ((*cp++ ^ *cp2++) & *cp3++)
1901 						goto next; /* next address! */
1902 				/*
1903 				 * If the netmask of what we just found
1904 				 * is more specific than what we had before
1905 				 * (if we had one), or if the virtual status
1906 				 * of new prefix is better than of the old one,
1907 				 * then remember the new one before continuing
1908 				 * to search for an even better one.
1909 				 */
1910 				if (ifa_maybe == NULL ||
1911 				    ifa_preferred(ifa_maybe, ifa) ||
1912 				    rn_refines((caddr_t)ifa->ifa_netmask,
1913 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1914 					if (ifa_maybe != NULL)
1915 						ifa_free(ifa_maybe);
1916 					ifa_maybe = ifa;
1917 					ifa_ref(ifa_maybe);
1918 				}
1919 			}
1920 		}
1921 		IF_ADDR_RUNLOCK(ifp);
1922 	}
1923 	ifa = ifa_maybe;
1924 	ifa_maybe = NULL;
1925 done:
1926 	IFNET_RUNLOCK_NOSLEEP();
1927 	if (ifa_maybe != NULL)
1928 		ifa_free(ifa_maybe);
1929 	return (ifa);
1930 }
1931 
1932 /*
1933  * Find an interface address specific to an interface best matching
1934  * a given address.
1935  */
1936 struct ifaddr *
1937 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1938 {
1939 	struct ifaddr *ifa;
1940 	char *cp, *cp2, *cp3;
1941 	char *cplim;
1942 	struct ifaddr *ifa_maybe = NULL;
1943 	u_int af = addr->sa_family;
1944 
1945 	if (af >= AF_MAX)
1946 		return (NULL);
1947 	IF_ADDR_RLOCK(ifp);
1948 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1949 		if (ifa->ifa_addr->sa_family != af)
1950 			continue;
1951 		if (ifa_maybe == NULL)
1952 			ifa_maybe = ifa;
1953 		if (ifa->ifa_netmask == 0) {
1954 			if (sa_equal(addr, ifa->ifa_addr) ||
1955 			    (ifa->ifa_dstaddr &&
1956 			    sa_equal(addr, ifa->ifa_dstaddr)))
1957 				goto done;
1958 			continue;
1959 		}
1960 		if (ifp->if_flags & IFF_POINTOPOINT) {
1961 			if (sa_equal(addr, ifa->ifa_dstaddr))
1962 				goto done;
1963 		} else {
1964 			cp = addr->sa_data;
1965 			cp2 = ifa->ifa_addr->sa_data;
1966 			cp3 = ifa->ifa_netmask->sa_data;
1967 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1968 			for (; cp3 < cplim; cp3++)
1969 				if ((*cp++ ^ *cp2++) & *cp3)
1970 					break;
1971 			if (cp3 == cplim)
1972 				goto done;
1973 		}
1974 	}
1975 	ifa = ifa_maybe;
1976 done:
1977 	if (ifa != NULL)
1978 		ifa_ref(ifa);
1979 	IF_ADDR_RUNLOCK(ifp);
1980 	return (ifa);
1981 }
1982 
1983 /*
1984  * See whether new ifa is better than current one:
1985  * 1) A non-virtual one is preferred over virtual.
1986  * 2) A virtual in master state preferred over any other state.
1987  *
1988  * Used in several address selecting functions.
1989  */
1990 int
1991 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1992 {
1993 
1994 	return (cur->ifa_carp && (!next->ifa_carp ||
1995 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1996 }
1997 
1998 #include <net/if_llatbl.h>
1999 
2000 /*
2001  * Default action when installing a route with a Link Level gateway.
2002  * Lookup an appropriate real ifa to point to.
2003  * This should be moved to /sys/net/link.c eventually.
2004  */
2005 static void
2006 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2007 {
2008 	struct ifaddr *ifa, *oifa;
2009 	struct sockaddr *dst;
2010 	struct ifnet *ifp;
2011 
2012 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
2013 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
2014 		return;
2015 	ifa = ifaof_ifpforaddr(dst, ifp);
2016 	if (ifa) {
2017 		oifa = rt->rt_ifa;
2018 		rt->rt_ifa = ifa;
2019 		ifa_free(oifa);
2020 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2021 			ifa->ifa_rtrequest(cmd, rt, info);
2022 	}
2023 }
2024 
2025 struct sockaddr_dl *
2026 link_alloc_sdl(size_t size, int flags)
2027 {
2028 
2029 	return (malloc(size, M_TEMP, flags));
2030 }
2031 
2032 void
2033 link_free_sdl(struct sockaddr *sa)
2034 {
2035 	free(sa, M_TEMP);
2036 }
2037 
2038 /*
2039  * Fills in given sdl with interface basic info.
2040  * Returns pointer to filled sdl.
2041  */
2042 struct sockaddr_dl *
2043 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2044 {
2045 	struct sockaddr_dl *sdl;
2046 
2047 	sdl = (struct sockaddr_dl *)paddr;
2048 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2049 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2050 	sdl->sdl_family = AF_LINK;
2051 	sdl->sdl_index = ifp->if_index;
2052 	sdl->sdl_type = iftype;
2053 
2054 	return (sdl);
2055 }
2056 
2057 /*
2058  * Mark an interface down and notify protocols of
2059  * the transition.
2060  */
2061 static void
2062 if_unroute(struct ifnet *ifp, int flag, int fam)
2063 {
2064 	struct ifaddr *ifa;
2065 
2066 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2067 
2068 	ifp->if_flags &= ~flag;
2069 	getmicrotime(&ifp->if_lastchange);
2070 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2071 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2072 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2073 	ifp->if_qflush(ifp);
2074 
2075 	if (ifp->if_carp)
2076 		(*carp_linkstate_p)(ifp);
2077 	rt_ifmsg(ifp);
2078 }
2079 
2080 /*
2081  * Mark an interface up and notify protocols of
2082  * the transition.
2083  */
2084 static void
2085 if_route(struct ifnet *ifp, int flag, int fam)
2086 {
2087 	struct ifaddr *ifa;
2088 
2089 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2090 
2091 	ifp->if_flags |= flag;
2092 	getmicrotime(&ifp->if_lastchange);
2093 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2094 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2095 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2096 	if (ifp->if_carp)
2097 		(*carp_linkstate_p)(ifp);
2098 	rt_ifmsg(ifp);
2099 #ifdef INET6
2100 	in6_if_up(ifp);
2101 #endif
2102 }
2103 
2104 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2105 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2106 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2107 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2108 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2109 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2110 void	*(*vlan_cookie_p)(struct ifnet *);
2111 
2112 /*
2113  * Handle a change in the interface link state. To avoid LORs
2114  * between driver lock and upper layer locks, as well as possible
2115  * recursions, we post event to taskqueue, and all job
2116  * is done in static do_link_state_change().
2117  */
2118 void
2119 if_link_state_change(struct ifnet *ifp, int link_state)
2120 {
2121 	/* Return if state hasn't changed. */
2122 	if (ifp->if_link_state == link_state)
2123 		return;
2124 
2125 	ifp->if_link_state = link_state;
2126 
2127 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2128 }
2129 
2130 static void
2131 do_link_state_change(void *arg, int pending)
2132 {
2133 	struct ifnet *ifp = (struct ifnet *)arg;
2134 	int link_state = ifp->if_link_state;
2135 	CURVNET_SET(ifp->if_vnet);
2136 
2137 	/* Notify that the link state has changed. */
2138 	rt_ifmsg(ifp);
2139 	if (ifp->if_vlantrunk != NULL)
2140 		(*vlan_link_state_p)(ifp);
2141 
2142 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2143 	    IFP2AC(ifp)->ac_netgraph != NULL)
2144 		(*ng_ether_link_state_p)(ifp, link_state);
2145 	if (ifp->if_carp)
2146 		(*carp_linkstate_p)(ifp);
2147 	if (ifp->if_bridge)
2148 		(*bridge_linkstate_p)(ifp);
2149 	if (ifp->if_lagg)
2150 		(*lagg_linkstate_p)(ifp, link_state);
2151 
2152 	if (IS_DEFAULT_VNET(curvnet))
2153 		devctl_notify("IFNET", ifp->if_xname,
2154 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2155 		    NULL);
2156 	if (pending > 1)
2157 		if_printf(ifp, "%d link states coalesced\n", pending);
2158 	if (log_link_state_change)
2159 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2160 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2161 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2162 	CURVNET_RESTORE();
2163 }
2164 
2165 /*
2166  * Mark an interface down and notify protocols of
2167  * the transition.
2168  */
2169 void
2170 if_down(struct ifnet *ifp)
2171 {
2172 
2173 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2174 }
2175 
2176 /*
2177  * Mark an interface up and notify protocols of
2178  * the transition.
2179  */
2180 void
2181 if_up(struct ifnet *ifp)
2182 {
2183 
2184 	if_route(ifp, IFF_UP, AF_UNSPEC);
2185 }
2186 
2187 /*
2188  * Flush an interface queue.
2189  */
2190 void
2191 if_qflush(struct ifnet *ifp)
2192 {
2193 	struct mbuf *m, *n;
2194 	struct ifaltq *ifq;
2195 
2196 	ifq = &ifp->if_snd;
2197 	IFQ_LOCK(ifq);
2198 #ifdef ALTQ
2199 	if (ALTQ_IS_ENABLED(ifq))
2200 		ALTQ_PURGE(ifq);
2201 #endif
2202 	n = ifq->ifq_head;
2203 	while ((m = n) != 0) {
2204 		n = m->m_nextpkt;
2205 		m_freem(m);
2206 	}
2207 	ifq->ifq_head = 0;
2208 	ifq->ifq_tail = 0;
2209 	ifq->ifq_len = 0;
2210 	IFQ_UNLOCK(ifq);
2211 }
2212 
2213 /*
2214  * Map interface name to interface structure pointer, with or without
2215  * returning a reference.
2216  */
2217 struct ifnet *
2218 ifunit_ref(const char *name)
2219 {
2220 	struct ifnet *ifp;
2221 
2222 	IFNET_RLOCK_NOSLEEP();
2223 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2224 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2225 		    !(ifp->if_flags & IFF_DYING))
2226 			break;
2227 	}
2228 	if (ifp != NULL)
2229 		if_ref(ifp);
2230 	IFNET_RUNLOCK_NOSLEEP();
2231 	return (ifp);
2232 }
2233 
2234 struct ifnet *
2235 ifunit(const char *name)
2236 {
2237 	struct ifnet *ifp;
2238 
2239 	IFNET_RLOCK_NOSLEEP();
2240 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2241 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2242 			break;
2243 	}
2244 	IFNET_RUNLOCK_NOSLEEP();
2245 	return (ifp);
2246 }
2247 
2248 /*
2249  * Hardware specific interface ioctls.
2250  */
2251 static int
2252 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2253 {
2254 	struct ifreq *ifr;
2255 	int error = 0;
2256 	int new_flags, temp_flags;
2257 	size_t namelen, onamelen;
2258 	size_t descrlen;
2259 	char *descrbuf, *odescrbuf;
2260 	char new_name[IFNAMSIZ];
2261 	struct ifaddr *ifa;
2262 	struct sockaddr_dl *sdl;
2263 
2264 	ifr = (struct ifreq *)data;
2265 	switch (cmd) {
2266 	case SIOCGIFINDEX:
2267 		ifr->ifr_index = ifp->if_index;
2268 		break;
2269 
2270 	case SIOCGIFFLAGS:
2271 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2272 		ifr->ifr_flags = temp_flags & 0xffff;
2273 		ifr->ifr_flagshigh = temp_flags >> 16;
2274 		break;
2275 
2276 	case SIOCGIFCAP:
2277 		ifr->ifr_reqcap = ifp->if_capabilities;
2278 		ifr->ifr_curcap = ifp->if_capenable;
2279 		break;
2280 
2281 #ifdef MAC
2282 	case SIOCGIFMAC:
2283 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2284 		break;
2285 #endif
2286 
2287 	case SIOCGIFMETRIC:
2288 		ifr->ifr_metric = ifp->if_metric;
2289 		break;
2290 
2291 	case SIOCGIFMTU:
2292 		ifr->ifr_mtu = ifp->if_mtu;
2293 		break;
2294 
2295 	case SIOCGIFPHYS:
2296 		/* XXXGL: did this ever worked? */
2297 		ifr->ifr_phys = 0;
2298 		break;
2299 
2300 	case SIOCGIFDESCR:
2301 		error = 0;
2302 		sx_slock(&ifdescr_sx);
2303 		if (ifp->if_description == NULL)
2304 			error = ENOMSG;
2305 		else {
2306 			/* space for terminating nul */
2307 			descrlen = strlen(ifp->if_description) + 1;
2308 			if (ifr->ifr_buffer.length < descrlen)
2309 				ifr->ifr_buffer.buffer = NULL;
2310 			else
2311 				error = copyout(ifp->if_description,
2312 				    ifr->ifr_buffer.buffer, descrlen);
2313 			ifr->ifr_buffer.length = descrlen;
2314 		}
2315 		sx_sunlock(&ifdescr_sx);
2316 		break;
2317 
2318 	case SIOCSIFDESCR:
2319 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2320 		if (error)
2321 			return (error);
2322 
2323 		/*
2324 		 * Copy only (length-1) bytes to make sure that
2325 		 * if_description is always nul terminated.  The
2326 		 * length parameter is supposed to count the
2327 		 * terminating nul in.
2328 		 */
2329 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2330 			return (ENAMETOOLONG);
2331 		else if (ifr->ifr_buffer.length == 0)
2332 			descrbuf = NULL;
2333 		else {
2334 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2335 			    M_WAITOK | M_ZERO);
2336 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2337 			    ifr->ifr_buffer.length - 1);
2338 			if (error) {
2339 				free(descrbuf, M_IFDESCR);
2340 				break;
2341 			}
2342 		}
2343 
2344 		sx_xlock(&ifdescr_sx);
2345 		odescrbuf = ifp->if_description;
2346 		ifp->if_description = descrbuf;
2347 		sx_xunlock(&ifdescr_sx);
2348 
2349 		getmicrotime(&ifp->if_lastchange);
2350 		free(odescrbuf, M_IFDESCR);
2351 		break;
2352 
2353 	case SIOCGIFFIB:
2354 		ifr->ifr_fib = ifp->if_fib;
2355 		break;
2356 
2357 	case SIOCSIFFIB:
2358 		error = priv_check(td, PRIV_NET_SETIFFIB);
2359 		if (error)
2360 			return (error);
2361 		if (ifr->ifr_fib >= rt_numfibs)
2362 			return (EINVAL);
2363 
2364 		ifp->if_fib = ifr->ifr_fib;
2365 		break;
2366 
2367 	case SIOCSIFFLAGS:
2368 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2369 		if (error)
2370 			return (error);
2371 		/*
2372 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2373 		 * check, so we don't need special handling here yet.
2374 		 */
2375 		new_flags = (ifr->ifr_flags & 0xffff) |
2376 		    (ifr->ifr_flagshigh << 16);
2377 		if (ifp->if_flags & IFF_UP &&
2378 		    (new_flags & IFF_UP) == 0) {
2379 			if_down(ifp);
2380 		} else if (new_flags & IFF_UP &&
2381 		    (ifp->if_flags & IFF_UP) == 0) {
2382 			if_up(ifp);
2383 		}
2384 		/* See if permanently promiscuous mode bit is about to flip */
2385 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2386 			if (new_flags & IFF_PPROMISC)
2387 				ifp->if_flags |= IFF_PROMISC;
2388 			else if (ifp->if_pcount == 0)
2389 				ifp->if_flags &= ~IFF_PROMISC;
2390 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2391 			    ifp->if_xname,
2392 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2393 		}
2394 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2395 			(new_flags &~ IFF_CANTCHANGE);
2396 		if (ifp->if_ioctl) {
2397 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2398 		}
2399 		getmicrotime(&ifp->if_lastchange);
2400 		break;
2401 
2402 	case SIOCSIFCAP:
2403 		error = priv_check(td, PRIV_NET_SETIFCAP);
2404 		if (error)
2405 			return (error);
2406 		if (ifp->if_ioctl == NULL)
2407 			return (EOPNOTSUPP);
2408 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2409 			return (EINVAL);
2410 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2411 		if (error == 0)
2412 			getmicrotime(&ifp->if_lastchange);
2413 		break;
2414 
2415 #ifdef MAC
2416 	case SIOCSIFMAC:
2417 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2418 		break;
2419 #endif
2420 
2421 	case SIOCSIFNAME:
2422 		error = priv_check(td, PRIV_NET_SETIFNAME);
2423 		if (error)
2424 			return (error);
2425 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2426 		if (error != 0)
2427 			return (error);
2428 		if (new_name[0] == '\0')
2429 			return (EINVAL);
2430 		if (ifunit(new_name) != NULL)
2431 			return (EEXIST);
2432 
2433 		/*
2434 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2435 		 * and there are numerous other races with the
2436 		 * ifunit() checks not being atomic with namespace
2437 		 * changes (renames, vmoves, if_attach, etc).
2438 		 */
2439 		ifp->if_flags |= IFF_RENAMING;
2440 
2441 		/* Announce the departure of the interface. */
2442 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2443 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2444 
2445 		log(LOG_INFO, "%s: changing name to '%s'\n",
2446 		    ifp->if_xname, new_name);
2447 
2448 		IF_ADDR_WLOCK(ifp);
2449 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2450 		ifa = ifp->if_addr;
2451 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2452 		namelen = strlen(new_name);
2453 		onamelen = sdl->sdl_nlen;
2454 		/*
2455 		 * Move the address if needed.  This is safe because we
2456 		 * allocate space for a name of length IFNAMSIZ when we
2457 		 * create this in if_attach().
2458 		 */
2459 		if (namelen != onamelen) {
2460 			bcopy(sdl->sdl_data + onamelen,
2461 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2462 		}
2463 		bcopy(new_name, sdl->sdl_data, namelen);
2464 		sdl->sdl_nlen = namelen;
2465 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2466 		bzero(sdl->sdl_data, onamelen);
2467 		while (namelen != 0)
2468 			sdl->sdl_data[--namelen] = 0xff;
2469 		IF_ADDR_WUNLOCK(ifp);
2470 
2471 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2472 		/* Announce the return of the interface. */
2473 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2474 
2475 		ifp->if_flags &= ~IFF_RENAMING;
2476 		break;
2477 
2478 #ifdef VIMAGE
2479 	case SIOCSIFVNET:
2480 		error = priv_check(td, PRIV_NET_SETIFVNET);
2481 		if (error)
2482 			return (error);
2483 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2484 		break;
2485 #endif
2486 
2487 	case SIOCSIFMETRIC:
2488 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2489 		if (error)
2490 			return (error);
2491 		ifp->if_metric = ifr->ifr_metric;
2492 		getmicrotime(&ifp->if_lastchange);
2493 		break;
2494 
2495 	case SIOCSIFPHYS:
2496 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2497 		if (error)
2498 			return (error);
2499 		if (ifp->if_ioctl == NULL)
2500 			return (EOPNOTSUPP);
2501 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2502 		if (error == 0)
2503 			getmicrotime(&ifp->if_lastchange);
2504 		break;
2505 
2506 	case SIOCSIFMTU:
2507 	{
2508 		u_long oldmtu = ifp->if_mtu;
2509 
2510 		error = priv_check(td, PRIV_NET_SETIFMTU);
2511 		if (error)
2512 			return (error);
2513 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2514 			return (EINVAL);
2515 		if (ifp->if_ioctl == NULL)
2516 			return (EOPNOTSUPP);
2517 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2518 		if (error == 0) {
2519 			getmicrotime(&ifp->if_lastchange);
2520 			rt_ifmsg(ifp);
2521 		}
2522 		/*
2523 		 * If the link MTU changed, do network layer specific procedure.
2524 		 */
2525 		if (ifp->if_mtu != oldmtu) {
2526 #ifdef INET6
2527 			nd6_setmtu(ifp);
2528 #endif
2529 		}
2530 		break;
2531 	}
2532 
2533 	case SIOCADDMULTI:
2534 	case SIOCDELMULTI:
2535 		if (cmd == SIOCADDMULTI)
2536 			error = priv_check(td, PRIV_NET_ADDMULTI);
2537 		else
2538 			error = priv_check(td, PRIV_NET_DELMULTI);
2539 		if (error)
2540 			return (error);
2541 
2542 		/* Don't allow group membership on non-multicast interfaces. */
2543 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2544 			return (EOPNOTSUPP);
2545 
2546 		/* Don't let users screw up protocols' entries. */
2547 		if (ifr->ifr_addr.sa_family != AF_LINK)
2548 			return (EINVAL);
2549 
2550 		if (cmd == SIOCADDMULTI) {
2551 			struct ifmultiaddr *ifma;
2552 
2553 			/*
2554 			 * Userland is only permitted to join groups once
2555 			 * via the if_addmulti() KPI, because it cannot hold
2556 			 * struct ifmultiaddr * between calls. It may also
2557 			 * lose a race while we check if the membership
2558 			 * already exists.
2559 			 */
2560 			IF_ADDR_RLOCK(ifp);
2561 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2562 			IF_ADDR_RUNLOCK(ifp);
2563 			if (ifma != NULL)
2564 				error = EADDRINUSE;
2565 			else
2566 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2567 		} else {
2568 			error = if_delmulti(ifp, &ifr->ifr_addr);
2569 		}
2570 		if (error == 0)
2571 			getmicrotime(&ifp->if_lastchange);
2572 		break;
2573 
2574 	case SIOCSIFPHYADDR:
2575 	case SIOCDIFPHYADDR:
2576 #ifdef INET6
2577 	case SIOCSIFPHYADDR_IN6:
2578 #endif
2579 	case SIOCSIFMEDIA:
2580 	case SIOCSIFGENERIC:
2581 		error = priv_check(td, PRIV_NET_HWIOCTL);
2582 		if (error)
2583 			return (error);
2584 		if (ifp->if_ioctl == NULL)
2585 			return (EOPNOTSUPP);
2586 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2587 		if (error == 0)
2588 			getmicrotime(&ifp->if_lastchange);
2589 		break;
2590 
2591 	case SIOCGIFSTATUS:
2592 	case SIOCGIFPSRCADDR:
2593 	case SIOCGIFPDSTADDR:
2594 	case SIOCGIFMEDIA:
2595 	case SIOCGIFGENERIC:
2596 		if (ifp->if_ioctl == NULL)
2597 			return (EOPNOTSUPP);
2598 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2599 		break;
2600 
2601 	case SIOCSIFLLADDR:
2602 		error = priv_check(td, PRIV_NET_SETLLADDR);
2603 		if (error)
2604 			return (error);
2605 		error = if_setlladdr(ifp,
2606 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2607 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2608 		break;
2609 
2610 	case SIOCAIFGROUP:
2611 	{
2612 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2613 
2614 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2615 		if (error)
2616 			return (error);
2617 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2618 			return (error);
2619 		break;
2620 	}
2621 
2622 	case SIOCGIFGROUP:
2623 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2624 			return (error);
2625 		break;
2626 
2627 	case SIOCDIFGROUP:
2628 	{
2629 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2630 
2631 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2632 		if (error)
2633 			return (error);
2634 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2635 			return (error);
2636 		break;
2637 	}
2638 
2639 	default:
2640 		error = ENOIOCTL;
2641 		break;
2642 	}
2643 	return (error);
2644 }
2645 
2646 #ifdef COMPAT_FREEBSD32
2647 struct ifconf32 {
2648 	int32_t	ifc_len;
2649 	union {
2650 		uint32_t	ifcu_buf;
2651 		uint32_t	ifcu_req;
2652 	} ifc_ifcu;
2653 };
2654 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2655 #endif
2656 
2657 /*
2658  * Interface ioctls.
2659  */
2660 int
2661 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2662 {
2663 	struct ifnet *ifp;
2664 	struct ifreq *ifr;
2665 	int error;
2666 	int oif_flags;
2667 
2668 	CURVNET_SET(so->so_vnet);
2669 	switch (cmd) {
2670 	case SIOCGIFCONF:
2671 		error = ifconf(cmd, data);
2672 		CURVNET_RESTORE();
2673 		return (error);
2674 
2675 #ifdef COMPAT_FREEBSD32
2676 	case SIOCGIFCONF32:
2677 		{
2678 			struct ifconf32 *ifc32;
2679 			struct ifconf ifc;
2680 
2681 			ifc32 = (struct ifconf32 *)data;
2682 			ifc.ifc_len = ifc32->ifc_len;
2683 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2684 
2685 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2686 			CURVNET_RESTORE();
2687 			if (error == 0)
2688 				ifc32->ifc_len = ifc.ifc_len;
2689 			return (error);
2690 		}
2691 #endif
2692 	}
2693 	ifr = (struct ifreq *)data;
2694 
2695 	switch (cmd) {
2696 #ifdef VIMAGE
2697 	case SIOCSIFRVNET:
2698 		error = priv_check(td, PRIV_NET_SETIFVNET);
2699 		if (error == 0)
2700 			error = if_vmove_reclaim(td, ifr->ifr_name,
2701 			    ifr->ifr_jid);
2702 		CURVNET_RESTORE();
2703 		return (error);
2704 #endif
2705 	case SIOCIFCREATE:
2706 	case SIOCIFCREATE2:
2707 		error = priv_check(td, PRIV_NET_IFCREATE);
2708 		if (error == 0)
2709 			error = if_clone_create(ifr->ifr_name,
2710 			    sizeof(ifr->ifr_name),
2711 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2712 		CURVNET_RESTORE();
2713 		return (error);
2714 	case SIOCIFDESTROY:
2715 		error = priv_check(td, PRIV_NET_IFDESTROY);
2716 		if (error == 0)
2717 			error = if_clone_destroy(ifr->ifr_name);
2718 		CURVNET_RESTORE();
2719 		return (error);
2720 
2721 	case SIOCIFGCLONERS:
2722 		error = if_clone_list((struct if_clonereq *)data);
2723 		CURVNET_RESTORE();
2724 		return (error);
2725 	case SIOCGIFGMEMB:
2726 		error = if_getgroupmembers((struct ifgroupreq *)data);
2727 		CURVNET_RESTORE();
2728 		return (error);
2729 #if defined(INET) || defined(INET6)
2730 	case SIOCSVH:
2731 	case SIOCGVH:
2732 		if (carp_ioctl_p == NULL)
2733 			error = EPROTONOSUPPORT;
2734 		else
2735 			error = (*carp_ioctl_p)(ifr, cmd, td);
2736 		CURVNET_RESTORE();
2737 		return (error);
2738 #endif
2739 	}
2740 
2741 	ifp = ifunit_ref(ifr->ifr_name);
2742 	if (ifp == NULL) {
2743 		CURVNET_RESTORE();
2744 		return (ENXIO);
2745 	}
2746 
2747 	error = ifhwioctl(cmd, ifp, data, td);
2748 	if (error != ENOIOCTL) {
2749 		if_rele(ifp);
2750 		CURVNET_RESTORE();
2751 		return (error);
2752 	}
2753 
2754 	oif_flags = ifp->if_flags;
2755 	if (so->so_proto == NULL) {
2756 		if_rele(ifp);
2757 		CURVNET_RESTORE();
2758 		return (EOPNOTSUPP);
2759 	}
2760 
2761 	/*
2762 	 * Pass the request on to the socket control method, and if the
2763 	 * latter returns EOPNOTSUPP, directly to the interface.
2764 	 *
2765 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2766 	 * trust SIOCSIFADDR et al to come from an already privileged
2767 	 * layer, and do not perform any credentials checks or input
2768 	 * validation.
2769 	 */
2770 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2771 	    ifp, td));
2772 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2773 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2774 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2775 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2776 
2777 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2778 #ifdef INET6
2779 		if (ifp->if_flags & IFF_UP)
2780 			in6_if_up(ifp);
2781 #endif
2782 	}
2783 	if_rele(ifp);
2784 	CURVNET_RESTORE();
2785 	return (error);
2786 }
2787 
2788 /*
2789  * The code common to handling reference counted flags,
2790  * e.g., in ifpromisc() and if_allmulti().
2791  * The "pflag" argument can specify a permanent mode flag to check,
2792  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2793  *
2794  * Only to be used on stack-owned flags, not driver-owned flags.
2795  */
2796 static int
2797 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2798 {
2799 	struct ifreq ifr;
2800 	int error;
2801 	int oldflags, oldcount;
2802 
2803 	/* Sanity checks to catch programming errors */
2804 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2805 	    ("%s: setting driver-owned flag %d", __func__, flag));
2806 
2807 	if (onswitch)
2808 		KASSERT(*refcount >= 0,
2809 		    ("%s: increment negative refcount %d for flag %d",
2810 		    __func__, *refcount, flag));
2811 	else
2812 		KASSERT(*refcount > 0,
2813 		    ("%s: decrement non-positive refcount %d for flag %d",
2814 		    __func__, *refcount, flag));
2815 
2816 	/* In case this mode is permanent, just touch refcount */
2817 	if (ifp->if_flags & pflag) {
2818 		*refcount += onswitch ? 1 : -1;
2819 		return (0);
2820 	}
2821 
2822 	/* Save ifnet parameters for if_ioctl() may fail */
2823 	oldcount = *refcount;
2824 	oldflags = ifp->if_flags;
2825 
2826 	/*
2827 	 * See if we aren't the only and touching refcount is enough.
2828 	 * Actually toggle interface flag if we are the first or last.
2829 	 */
2830 	if (onswitch) {
2831 		if ((*refcount)++)
2832 			return (0);
2833 		ifp->if_flags |= flag;
2834 	} else {
2835 		if (--(*refcount))
2836 			return (0);
2837 		ifp->if_flags &= ~flag;
2838 	}
2839 
2840 	/* Call down the driver since we've changed interface flags */
2841 	if (ifp->if_ioctl == NULL) {
2842 		error = EOPNOTSUPP;
2843 		goto recover;
2844 	}
2845 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2846 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2847 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2848 	if (error)
2849 		goto recover;
2850 	/* Notify userland that interface flags have changed */
2851 	rt_ifmsg(ifp);
2852 	return (0);
2853 
2854 recover:
2855 	/* Recover after driver error */
2856 	*refcount = oldcount;
2857 	ifp->if_flags = oldflags;
2858 	return (error);
2859 }
2860 
2861 /*
2862  * Set/clear promiscuous mode on interface ifp based on the truth value
2863  * of pswitch.  The calls are reference counted so that only the first
2864  * "on" request actually has an effect, as does the final "off" request.
2865  * Results are undefined if the "off" and "on" requests are not matched.
2866  */
2867 int
2868 ifpromisc(struct ifnet *ifp, int pswitch)
2869 {
2870 	int error;
2871 	int oldflags = ifp->if_flags;
2872 
2873 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2874 			   &ifp->if_pcount, pswitch);
2875 	/* If promiscuous mode status has changed, log a message */
2876 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2877 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2878 		    ifp->if_xname,
2879 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2880 	return (error);
2881 }
2882 
2883 /*
2884  * Return interface configuration
2885  * of system.  List may be used
2886  * in later ioctl's (above) to get
2887  * other information.
2888  */
2889 /*ARGSUSED*/
2890 static int
2891 ifconf(u_long cmd, caddr_t data)
2892 {
2893 	struct ifconf *ifc = (struct ifconf *)data;
2894 	struct ifnet *ifp;
2895 	struct ifaddr *ifa;
2896 	struct ifreq ifr;
2897 	struct sbuf *sb;
2898 	int error, full = 0, valid_len, max_len;
2899 
2900 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2901 	max_len = MAXPHYS - 1;
2902 
2903 	/* Prevent hostile input from being able to crash the system */
2904 	if (ifc->ifc_len <= 0)
2905 		return (EINVAL);
2906 
2907 again:
2908 	if (ifc->ifc_len <= max_len) {
2909 		max_len = ifc->ifc_len;
2910 		full = 1;
2911 	}
2912 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2913 	max_len = 0;
2914 	valid_len = 0;
2915 
2916 	IFNET_RLOCK();
2917 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2918 		int addrs;
2919 
2920 		/*
2921 		 * Zero the ifr_name buffer to make sure we don't
2922 		 * disclose the contents of the stack.
2923 		 */
2924 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2925 
2926 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2927 		    >= sizeof(ifr.ifr_name)) {
2928 			sbuf_delete(sb);
2929 			IFNET_RUNLOCK();
2930 			return (ENAMETOOLONG);
2931 		}
2932 
2933 		addrs = 0;
2934 		IF_ADDR_RLOCK(ifp);
2935 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2936 			struct sockaddr *sa = ifa->ifa_addr;
2937 
2938 			if (prison_if(curthread->td_ucred, sa) != 0)
2939 				continue;
2940 			addrs++;
2941 			if (sa->sa_len <= sizeof(*sa)) {
2942 				ifr.ifr_addr = *sa;
2943 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2944 				max_len += sizeof(ifr);
2945 			} else {
2946 				sbuf_bcat(sb, &ifr,
2947 				    offsetof(struct ifreq, ifr_addr));
2948 				max_len += offsetof(struct ifreq, ifr_addr);
2949 				sbuf_bcat(sb, sa, sa->sa_len);
2950 				max_len += sa->sa_len;
2951 			}
2952 
2953 			if (sbuf_error(sb) == 0)
2954 				valid_len = sbuf_len(sb);
2955 		}
2956 		IF_ADDR_RUNLOCK(ifp);
2957 		if (addrs == 0) {
2958 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2959 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2960 			max_len += sizeof(ifr);
2961 
2962 			if (sbuf_error(sb) == 0)
2963 				valid_len = sbuf_len(sb);
2964 		}
2965 	}
2966 	IFNET_RUNLOCK();
2967 
2968 	/*
2969 	 * If we didn't allocate enough space (uncommon), try again.  If
2970 	 * we have already allocated as much space as we are allowed,
2971 	 * return what we've got.
2972 	 */
2973 	if (valid_len != max_len && !full) {
2974 		sbuf_delete(sb);
2975 		goto again;
2976 	}
2977 
2978 	ifc->ifc_len = valid_len;
2979 	sbuf_finish(sb);
2980 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2981 	sbuf_delete(sb);
2982 	return (error);
2983 }
2984 
2985 /*
2986  * Just like ifpromisc(), but for all-multicast-reception mode.
2987  */
2988 int
2989 if_allmulti(struct ifnet *ifp, int onswitch)
2990 {
2991 
2992 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2993 }
2994 
2995 struct ifmultiaddr *
2996 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2997 {
2998 	struct ifmultiaddr *ifma;
2999 
3000 	IF_ADDR_LOCK_ASSERT(ifp);
3001 
3002 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3003 		if (sa->sa_family == AF_LINK) {
3004 			if (sa_dl_equal(ifma->ifma_addr, sa))
3005 				break;
3006 		} else {
3007 			if (sa_equal(ifma->ifma_addr, sa))
3008 				break;
3009 		}
3010 	}
3011 
3012 	return ifma;
3013 }
3014 
3015 /*
3016  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3017  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3018  * the ifnet multicast address list here, so the caller must do that and
3019  * other setup work (such as notifying the device driver).  The reference
3020  * count is initialized to 1.
3021  */
3022 static struct ifmultiaddr *
3023 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3024     int mflags)
3025 {
3026 	struct ifmultiaddr *ifma;
3027 	struct sockaddr *dupsa;
3028 
3029 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3030 	    M_ZERO);
3031 	if (ifma == NULL)
3032 		return (NULL);
3033 
3034 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3035 	if (dupsa == NULL) {
3036 		free(ifma, M_IFMADDR);
3037 		return (NULL);
3038 	}
3039 	bcopy(sa, dupsa, sa->sa_len);
3040 	ifma->ifma_addr = dupsa;
3041 
3042 	ifma->ifma_ifp = ifp;
3043 	ifma->ifma_refcount = 1;
3044 	ifma->ifma_protospec = NULL;
3045 
3046 	if (llsa == NULL) {
3047 		ifma->ifma_lladdr = NULL;
3048 		return (ifma);
3049 	}
3050 
3051 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3052 	if (dupsa == NULL) {
3053 		free(ifma->ifma_addr, M_IFMADDR);
3054 		free(ifma, M_IFMADDR);
3055 		return (NULL);
3056 	}
3057 	bcopy(llsa, dupsa, llsa->sa_len);
3058 	ifma->ifma_lladdr = dupsa;
3059 
3060 	return (ifma);
3061 }
3062 
3063 /*
3064  * if_freemulti: free ifmultiaddr structure and possibly attached related
3065  * addresses.  The caller is responsible for implementing reference
3066  * counting, notifying the driver, handling routing messages, and releasing
3067  * any dependent link layer state.
3068  */
3069 static void
3070 if_freemulti(struct ifmultiaddr *ifma)
3071 {
3072 
3073 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3074 	    ifma->ifma_refcount));
3075 
3076 	if (ifma->ifma_lladdr != NULL)
3077 		free(ifma->ifma_lladdr, M_IFMADDR);
3078 	free(ifma->ifma_addr, M_IFMADDR);
3079 	free(ifma, M_IFMADDR);
3080 }
3081 
3082 /*
3083  * Register an additional multicast address with a network interface.
3084  *
3085  * - If the address is already present, bump the reference count on the
3086  *   address and return.
3087  * - If the address is not link-layer, look up a link layer address.
3088  * - Allocate address structures for one or both addresses, and attach to the
3089  *   multicast address list on the interface.  If automatically adding a link
3090  *   layer address, the protocol address will own a reference to the link
3091  *   layer address, to be freed when it is freed.
3092  * - Notify the network device driver of an addition to the multicast address
3093  *   list.
3094  *
3095  * 'sa' points to caller-owned memory with the desired multicast address.
3096  *
3097  * 'retifma' will be used to return a pointer to the resulting multicast
3098  * address reference, if desired.
3099  */
3100 int
3101 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3102     struct ifmultiaddr **retifma)
3103 {
3104 	struct ifmultiaddr *ifma, *ll_ifma;
3105 	struct sockaddr *llsa;
3106 	struct sockaddr_dl sdl;
3107 	int error;
3108 
3109 	/*
3110 	 * If the address is already present, return a new reference to it;
3111 	 * otherwise, allocate storage and set up a new address.
3112 	 */
3113 	IF_ADDR_WLOCK(ifp);
3114 	ifma = if_findmulti(ifp, sa);
3115 	if (ifma != NULL) {
3116 		ifma->ifma_refcount++;
3117 		if (retifma != NULL)
3118 			*retifma = ifma;
3119 		IF_ADDR_WUNLOCK(ifp);
3120 		return (0);
3121 	}
3122 
3123 	/*
3124 	 * The address isn't already present; resolve the protocol address
3125 	 * into a link layer address, and then look that up, bump its
3126 	 * refcount or allocate an ifma for that also.
3127 	 * Most link layer resolving functions returns address data which
3128 	 * fits inside default sockaddr_dl structure. However callback
3129 	 * can allocate another sockaddr structure, in that case we need to
3130 	 * free it later.
3131 	 */
3132 	llsa = NULL;
3133 	ll_ifma = NULL;
3134 	if (ifp->if_resolvemulti != NULL) {
3135 		/* Provide called function with buffer size information */
3136 		sdl.sdl_len = sizeof(sdl);
3137 		llsa = (struct sockaddr *)&sdl;
3138 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3139 		if (error)
3140 			goto unlock_out;
3141 	}
3142 
3143 	/*
3144 	 * Allocate the new address.  Don't hook it up yet, as we may also
3145 	 * need to allocate a link layer multicast address.
3146 	 */
3147 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3148 	if (ifma == NULL) {
3149 		error = ENOMEM;
3150 		goto free_llsa_out;
3151 	}
3152 
3153 	/*
3154 	 * If a link layer address is found, we'll need to see if it's
3155 	 * already present in the address list, or allocate is as well.
3156 	 * When this block finishes, the link layer address will be on the
3157 	 * list.
3158 	 */
3159 	if (llsa != NULL) {
3160 		ll_ifma = if_findmulti(ifp, llsa);
3161 		if (ll_ifma == NULL) {
3162 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3163 			if (ll_ifma == NULL) {
3164 				--ifma->ifma_refcount;
3165 				if_freemulti(ifma);
3166 				error = ENOMEM;
3167 				goto free_llsa_out;
3168 			}
3169 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3170 			    ifma_link);
3171 		} else
3172 			ll_ifma->ifma_refcount++;
3173 		ifma->ifma_llifma = ll_ifma;
3174 	}
3175 
3176 	/*
3177 	 * We now have a new multicast address, ifma, and possibly a new or
3178 	 * referenced link layer address.  Add the primary address to the
3179 	 * ifnet address list.
3180 	 */
3181 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3182 
3183 	if (retifma != NULL)
3184 		*retifma = ifma;
3185 
3186 	/*
3187 	 * Must generate the message while holding the lock so that 'ifma'
3188 	 * pointer is still valid.
3189 	 */
3190 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3191 	IF_ADDR_WUNLOCK(ifp);
3192 
3193 	/*
3194 	 * We are certain we have added something, so call down to the
3195 	 * interface to let them know about it.
3196 	 */
3197 	if (ifp->if_ioctl != NULL) {
3198 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3199 	}
3200 
3201 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3202 		link_free_sdl(llsa);
3203 
3204 	return (0);
3205 
3206 free_llsa_out:
3207 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3208 		link_free_sdl(llsa);
3209 
3210 unlock_out:
3211 	IF_ADDR_WUNLOCK(ifp);
3212 	return (error);
3213 }
3214 
3215 /*
3216  * Delete a multicast group membership by network-layer group address.
3217  *
3218  * Returns ENOENT if the entry could not be found. If ifp no longer
3219  * exists, results are undefined. This entry point should only be used
3220  * from subsystems which do appropriate locking to hold ifp for the
3221  * duration of the call.
3222  * Network-layer protocol domains must use if_delmulti_ifma().
3223  */
3224 int
3225 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3226 {
3227 	struct ifmultiaddr *ifma;
3228 	int lastref;
3229 #ifdef INVARIANTS
3230 	struct ifnet *oifp;
3231 
3232 	IFNET_RLOCK_NOSLEEP();
3233 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3234 		if (ifp == oifp)
3235 			break;
3236 	if (ifp != oifp)
3237 		ifp = NULL;
3238 	IFNET_RUNLOCK_NOSLEEP();
3239 
3240 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3241 #endif
3242 	if (ifp == NULL)
3243 		return (ENOENT);
3244 
3245 	IF_ADDR_WLOCK(ifp);
3246 	lastref = 0;
3247 	ifma = if_findmulti(ifp, sa);
3248 	if (ifma != NULL)
3249 		lastref = if_delmulti_locked(ifp, ifma, 0);
3250 	IF_ADDR_WUNLOCK(ifp);
3251 
3252 	if (ifma == NULL)
3253 		return (ENOENT);
3254 
3255 	if (lastref && ifp->if_ioctl != NULL) {
3256 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3257 	}
3258 
3259 	return (0);
3260 }
3261 
3262 /*
3263  * Delete all multicast group membership for an interface.
3264  * Should be used to quickly flush all multicast filters.
3265  */
3266 void
3267 if_delallmulti(struct ifnet *ifp)
3268 {
3269 	struct ifmultiaddr *ifma;
3270 	struct ifmultiaddr *next;
3271 
3272 	IF_ADDR_WLOCK(ifp);
3273 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3274 		if_delmulti_locked(ifp, ifma, 0);
3275 	IF_ADDR_WUNLOCK(ifp);
3276 }
3277 
3278 /*
3279  * Delete a multicast group membership by group membership pointer.
3280  * Network-layer protocol domains must use this routine.
3281  *
3282  * It is safe to call this routine if the ifp disappeared.
3283  */
3284 void
3285 if_delmulti_ifma(struct ifmultiaddr *ifma)
3286 {
3287 	struct ifnet *ifp;
3288 	int lastref;
3289 
3290 	ifp = ifma->ifma_ifp;
3291 #ifdef DIAGNOSTIC
3292 	if (ifp == NULL) {
3293 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3294 	} else {
3295 		struct ifnet *oifp;
3296 
3297 		IFNET_RLOCK_NOSLEEP();
3298 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3299 			if (ifp == oifp)
3300 				break;
3301 		if (ifp != oifp) {
3302 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3303 			ifp = NULL;
3304 		}
3305 		IFNET_RUNLOCK_NOSLEEP();
3306 	}
3307 #endif
3308 	/*
3309 	 * If and only if the ifnet instance exists: Acquire the address lock.
3310 	 */
3311 	if (ifp != NULL)
3312 		IF_ADDR_WLOCK(ifp);
3313 
3314 	lastref = if_delmulti_locked(ifp, ifma, 0);
3315 
3316 	if (ifp != NULL) {
3317 		/*
3318 		 * If and only if the ifnet instance exists:
3319 		 *  Release the address lock.
3320 		 *  If the group was left: update the hardware hash filter.
3321 		 */
3322 		IF_ADDR_WUNLOCK(ifp);
3323 		if (lastref && ifp->if_ioctl != NULL) {
3324 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3325 		}
3326 	}
3327 }
3328 
3329 /*
3330  * Perform deletion of network-layer and/or link-layer multicast address.
3331  *
3332  * Return 0 if the reference count was decremented.
3333  * Return 1 if the final reference was released, indicating that the
3334  * hardware hash filter should be reprogrammed.
3335  */
3336 static int
3337 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3338 {
3339 	struct ifmultiaddr *ll_ifma;
3340 
3341 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3342 		KASSERT(ifma->ifma_ifp == ifp,
3343 		    ("%s: inconsistent ifp %p", __func__, ifp));
3344 		IF_ADDR_WLOCK_ASSERT(ifp);
3345 	}
3346 
3347 	ifp = ifma->ifma_ifp;
3348 
3349 	/*
3350 	 * If the ifnet is detaching, null out references to ifnet,
3351 	 * so that upper protocol layers will notice, and not attempt
3352 	 * to obtain locks for an ifnet which no longer exists. The
3353 	 * routing socket announcement must happen before the ifnet
3354 	 * instance is detached from the system.
3355 	 */
3356 	if (detaching) {
3357 #ifdef DIAGNOSTIC
3358 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3359 #endif
3360 		/*
3361 		 * ifp may already be nulled out if we are being reentered
3362 		 * to delete the ll_ifma.
3363 		 */
3364 		if (ifp != NULL) {
3365 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3366 			ifma->ifma_ifp = NULL;
3367 		}
3368 	}
3369 
3370 	if (--ifma->ifma_refcount > 0)
3371 		return 0;
3372 
3373 	/*
3374 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3375 	 * have been associated with it. Release it first if so.
3376 	 */
3377 	ll_ifma = ifma->ifma_llifma;
3378 	if (ll_ifma != NULL) {
3379 		KASSERT(ifma->ifma_lladdr != NULL,
3380 		    ("%s: llifma w/o lladdr", __func__));
3381 		if (detaching)
3382 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3383 		if (--ll_ifma->ifma_refcount == 0) {
3384 			if (ifp != NULL) {
3385 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3386 				    ifma_link);
3387 			}
3388 			if_freemulti(ll_ifma);
3389 		}
3390 	}
3391 
3392 	if (ifp != NULL)
3393 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3394 
3395 	if_freemulti(ifma);
3396 
3397 	/*
3398 	 * The last reference to this instance of struct ifmultiaddr
3399 	 * was released; the hardware should be notified of this change.
3400 	 */
3401 	return 1;
3402 }
3403 
3404 /*
3405  * Set the link layer address on an interface.
3406  *
3407  * At this time we only support certain types of interfaces,
3408  * and we don't allow the length of the address to change.
3409  */
3410 int
3411 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3412 {
3413 	struct sockaddr_dl *sdl;
3414 	struct ifaddr *ifa;
3415 	struct ifreq ifr;
3416 
3417 	IF_ADDR_RLOCK(ifp);
3418 	ifa = ifp->if_addr;
3419 	if (ifa == NULL) {
3420 		IF_ADDR_RUNLOCK(ifp);
3421 		return (EINVAL);
3422 	}
3423 	ifa_ref(ifa);
3424 	IF_ADDR_RUNLOCK(ifp);
3425 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3426 	if (sdl == NULL) {
3427 		ifa_free(ifa);
3428 		return (EINVAL);
3429 	}
3430 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3431 		ifa_free(ifa);
3432 		return (EINVAL);
3433 	}
3434 	switch (ifp->if_type) {
3435 	case IFT_ETHER:
3436 	case IFT_FDDI:
3437 	case IFT_XETHER:
3438 	case IFT_ISO88025:
3439 	case IFT_L2VLAN:
3440 	case IFT_BRIDGE:
3441 	case IFT_ARCNET:
3442 	case IFT_IEEE8023ADLAG:
3443 	case IFT_IEEE80211:
3444 		bcopy(lladdr, LLADDR(sdl), len);
3445 		ifa_free(ifa);
3446 		break;
3447 	default:
3448 		ifa_free(ifa);
3449 		return (ENODEV);
3450 	}
3451 
3452 	/*
3453 	 * If the interface is already up, we need
3454 	 * to re-init it in order to reprogram its
3455 	 * address filter.
3456 	 */
3457 	if ((ifp->if_flags & IFF_UP) != 0) {
3458 		if (ifp->if_ioctl) {
3459 			ifp->if_flags &= ~IFF_UP;
3460 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3461 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3462 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3463 			ifp->if_flags |= IFF_UP;
3464 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3465 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3466 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3467 		}
3468 #ifdef INET
3469 		/*
3470 		 * Also send gratuitous ARPs to notify other nodes about
3471 		 * the address change.
3472 		 */
3473 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3474 			if (ifa->ifa_addr->sa_family == AF_INET)
3475 				arp_ifinit(ifp, ifa);
3476 		}
3477 #endif
3478 	}
3479 	return (0);
3480 }
3481 
3482 /*
3483  * The name argument must be a pointer to storage which will last as
3484  * long as the interface does.  For physical devices, the result of
3485  * device_get_name(dev) is a good choice and for pseudo-devices a
3486  * static string works well.
3487  */
3488 void
3489 if_initname(struct ifnet *ifp, const char *name, int unit)
3490 {
3491 	ifp->if_dname = name;
3492 	ifp->if_dunit = unit;
3493 	if (unit != IF_DUNIT_NONE)
3494 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3495 	else
3496 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3497 }
3498 
3499 int
3500 if_printf(struct ifnet *ifp, const char * fmt, ...)
3501 {
3502 	va_list ap;
3503 	int retval;
3504 
3505 	retval = printf("%s: ", ifp->if_xname);
3506 	va_start(ap, fmt);
3507 	retval += vprintf(fmt, ap);
3508 	va_end(ap);
3509 	return (retval);
3510 }
3511 
3512 void
3513 if_start(struct ifnet *ifp)
3514 {
3515 
3516 	(*(ifp)->if_start)(ifp);
3517 }
3518 
3519 /*
3520  * Backwards compatibility interface for drivers
3521  * that have not implemented it
3522  */
3523 static int
3524 if_transmit(struct ifnet *ifp, struct mbuf *m)
3525 {
3526 	int error;
3527 
3528 	IFQ_HANDOFF(ifp, m, error);
3529 	return (error);
3530 }
3531 
3532 int
3533 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3534 {
3535 	int active = 0;
3536 
3537 	IF_LOCK(ifq);
3538 	if (_IF_QFULL(ifq)) {
3539 		IF_UNLOCK(ifq);
3540 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3541 		m_freem(m);
3542 		return (0);
3543 	}
3544 	if (ifp != NULL) {
3545 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3546 		if (m->m_flags & (M_BCAST|M_MCAST))
3547 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3548 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3549 	}
3550 	_IF_ENQUEUE(ifq, m);
3551 	IF_UNLOCK(ifq);
3552 	if (ifp != NULL && !active)
3553 		(*(ifp)->if_start)(ifp);
3554 	return (1);
3555 }
3556 
3557 void
3558 if_register_com_alloc(u_char type,
3559     if_com_alloc_t *a, if_com_free_t *f)
3560 {
3561 
3562 	KASSERT(if_com_alloc[type] == NULL,
3563 	    ("if_register_com_alloc: %d already registered", type));
3564 	KASSERT(if_com_free[type] == NULL,
3565 	    ("if_register_com_alloc: %d free already registered", type));
3566 
3567 	if_com_alloc[type] = a;
3568 	if_com_free[type] = f;
3569 }
3570 
3571 void
3572 if_deregister_com_alloc(u_char type)
3573 {
3574 
3575 	KASSERT(if_com_alloc[type] != NULL,
3576 	    ("if_deregister_com_alloc: %d not registered", type));
3577 	KASSERT(if_com_free[type] != NULL,
3578 	    ("if_deregister_com_alloc: %d free not registered", type));
3579 	if_com_alloc[type] = NULL;
3580 	if_com_free[type] = NULL;
3581 }
3582 
3583 /* API for driver access to network stack owned ifnet.*/
3584 uint64_t
3585 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3586 {
3587 	uint64_t oldbrate;
3588 
3589 	oldbrate = ifp->if_baudrate;
3590 	ifp->if_baudrate = baudrate;
3591 	return (oldbrate);
3592 }
3593 
3594 uint64_t
3595 if_getbaudrate(if_t ifp)
3596 {
3597 
3598 	return (((struct ifnet *)ifp)->if_baudrate);
3599 }
3600 
3601 int
3602 if_setcapabilities(if_t ifp, int capabilities)
3603 {
3604 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3605 	return (0);
3606 }
3607 
3608 int
3609 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3610 {
3611 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3612 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3613 
3614 	return (0);
3615 }
3616 
3617 int
3618 if_getcapabilities(if_t ifp)
3619 {
3620 	return ((struct ifnet *)ifp)->if_capabilities;
3621 }
3622 
3623 int
3624 if_setcapenable(if_t ifp, int capabilities)
3625 {
3626 	((struct ifnet *)ifp)->if_capenable = capabilities;
3627 	return (0);
3628 }
3629 
3630 int
3631 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3632 {
3633 	if(setcap)
3634 		((struct ifnet *)ifp)->if_capenable |= setcap;
3635 	if(clearcap)
3636 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3637 
3638 	return (0);
3639 }
3640 
3641 const char *
3642 if_getdname(if_t ifp)
3643 {
3644 	return ((struct ifnet *)ifp)->if_dname;
3645 }
3646 
3647 int
3648 if_togglecapenable(if_t ifp, int togglecap)
3649 {
3650 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3651 	return (0);
3652 }
3653 
3654 int
3655 if_getcapenable(if_t ifp)
3656 {
3657 	return ((struct ifnet *)ifp)->if_capenable;
3658 }
3659 
3660 /*
3661  * This is largely undesirable because it ties ifnet to a device, but does
3662  * provide flexiblity for an embedded product vendor. Should be used with
3663  * the understanding that it violates the interface boundaries, and should be
3664  * a last resort only.
3665  */
3666 int
3667 if_setdev(if_t ifp, void *dev)
3668 {
3669 	return (0);
3670 }
3671 
3672 int
3673 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3674 {
3675 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3676 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3677 
3678 	return (0);
3679 }
3680 
3681 int
3682 if_getdrvflags(if_t ifp)
3683 {
3684 	return ((struct ifnet *)ifp)->if_drv_flags;
3685 }
3686 
3687 int
3688 if_setdrvflags(if_t ifp, int flags)
3689 {
3690 	((struct ifnet *)ifp)->if_drv_flags = flags;
3691 	return (0);
3692 }
3693 
3694 
3695 int
3696 if_setflags(if_t ifp, int flags)
3697 {
3698 	((struct ifnet *)ifp)->if_flags = flags;
3699 	return (0);
3700 }
3701 
3702 int
3703 if_setflagbits(if_t ifp, int set, int clear)
3704 {
3705 	((struct ifnet *)ifp)->if_flags |= set;
3706 	((struct ifnet *)ifp)->if_flags &= ~clear;
3707 
3708 	return (0);
3709 }
3710 
3711 int
3712 if_getflags(if_t ifp)
3713 {
3714 	return ((struct ifnet *)ifp)->if_flags;
3715 }
3716 
3717 int
3718 if_clearhwassist(if_t ifp)
3719 {
3720 	((struct ifnet *)ifp)->if_hwassist = 0;
3721 	return (0);
3722 }
3723 
3724 int
3725 if_sethwassistbits(if_t ifp, int toset, int toclear)
3726 {
3727 	((struct ifnet *)ifp)->if_hwassist |= toset;
3728 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3729 
3730 	return (0);
3731 }
3732 
3733 int
3734 if_sethwassist(if_t ifp, int hwassist_bit)
3735 {
3736 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3737 	return (0);
3738 }
3739 
3740 int
3741 if_gethwassist(if_t ifp)
3742 {
3743 	return ((struct ifnet *)ifp)->if_hwassist;
3744 }
3745 
3746 int
3747 if_setmtu(if_t ifp, int mtu)
3748 {
3749 	((struct ifnet *)ifp)->if_mtu = mtu;
3750 	return (0);
3751 }
3752 
3753 int
3754 if_getmtu(if_t ifp)
3755 {
3756 	return ((struct ifnet *)ifp)->if_mtu;
3757 }
3758 
3759 int
3760 if_getmtu_family(if_t ifp, int family)
3761 {
3762 	struct domain *dp;
3763 
3764 	for (dp = domains; dp; dp = dp->dom_next) {
3765 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3766 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3767 	}
3768 
3769 	return (((struct ifnet *)ifp)->if_mtu);
3770 }
3771 
3772 int
3773 if_setsoftc(if_t ifp, void *softc)
3774 {
3775 	((struct ifnet *)ifp)->if_softc = softc;
3776 	return (0);
3777 }
3778 
3779 void *
3780 if_getsoftc(if_t ifp)
3781 {
3782 	return ((struct ifnet *)ifp)->if_softc;
3783 }
3784 
3785 void
3786 if_setrcvif(struct mbuf *m, if_t ifp)
3787 {
3788 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3789 }
3790 
3791 void
3792 if_setvtag(struct mbuf *m, uint16_t tag)
3793 {
3794 	m->m_pkthdr.ether_vtag = tag;
3795 }
3796 
3797 uint16_t
3798 if_getvtag(struct mbuf *m)
3799 {
3800 
3801 	return (m->m_pkthdr.ether_vtag);
3802 }
3803 
3804 int
3805 if_sendq_empty(if_t ifp)
3806 {
3807 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3808 }
3809 
3810 struct ifaddr *
3811 if_getifaddr(if_t ifp)
3812 {
3813 	return ((struct ifnet *)ifp)->if_addr;
3814 }
3815 
3816 int
3817 if_getamcount(if_t ifp)
3818 {
3819 	return ((struct ifnet *)ifp)->if_amcount;
3820 }
3821 
3822 
3823 int
3824 if_setsendqready(if_t ifp)
3825 {
3826 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3827 	return (0);
3828 }
3829 
3830 int
3831 if_setsendqlen(if_t ifp, int tx_desc_count)
3832 {
3833 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3834 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3835 
3836 	return (0);
3837 }
3838 
3839 int
3840 if_vlantrunkinuse(if_t ifp)
3841 {
3842 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3843 }
3844 
3845 int
3846 if_input(if_t ifp, struct mbuf* sendmp)
3847 {
3848 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3849 	return (0);
3850 
3851 }
3852 
3853 /* XXX */
3854 #ifndef ETH_ADDR_LEN
3855 #define ETH_ADDR_LEN 6
3856 #endif
3857 
3858 int
3859 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3860 {
3861 	struct ifmultiaddr *ifma;
3862 	uint8_t *lmta = (uint8_t *)mta;
3863 	int mcnt = 0;
3864 
3865 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3866 		if (ifma->ifma_addr->sa_family != AF_LINK)
3867 			continue;
3868 
3869 		if (mcnt == max)
3870 			break;
3871 
3872 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3873 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3874 		mcnt++;
3875 	}
3876 	*cnt = mcnt;
3877 
3878 	return (0);
3879 }
3880 
3881 int
3882 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3883 {
3884 	int error;
3885 
3886 	if_maddr_rlock(ifp);
3887 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3888 	if_maddr_runlock(ifp);
3889 	return (error);
3890 }
3891 
3892 int
3893 if_multiaddr_count(if_t ifp, int max)
3894 {
3895 	struct ifmultiaddr *ifma;
3896 	int count;
3897 
3898 	count = 0;
3899 	if_maddr_rlock(ifp);
3900 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3901 		if (ifma->ifma_addr->sa_family != AF_LINK)
3902 			continue;
3903 		count++;
3904 		if (count == max)
3905 			break;
3906 	}
3907 	if_maddr_runlock(ifp);
3908 	return (count);
3909 }
3910 
3911 struct mbuf *
3912 if_dequeue(if_t ifp)
3913 {
3914 	struct mbuf *m;
3915 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3916 
3917 	return (m);
3918 }
3919 
3920 int
3921 if_sendq_prepend(if_t ifp, struct mbuf *m)
3922 {
3923 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3924 	return (0);
3925 }
3926 
3927 int
3928 if_setifheaderlen(if_t ifp, int len)
3929 {
3930 	((struct ifnet *)ifp)->if_hdrlen = len;
3931 	return (0);
3932 }
3933 
3934 caddr_t
3935 if_getlladdr(if_t ifp)
3936 {
3937 	return (IF_LLADDR((struct ifnet *)ifp));
3938 }
3939 
3940 void *
3941 if_gethandle(u_char type)
3942 {
3943 	return (if_alloc(type));
3944 }
3945 
3946 void
3947 if_bpfmtap(if_t ifh, struct mbuf *m)
3948 {
3949 	struct ifnet *ifp = (struct ifnet *)ifh;
3950 
3951 	BPF_MTAP(ifp, m);
3952 }
3953 
3954 void
3955 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3956 {
3957 	struct ifnet *ifp = (struct ifnet *)ifh;
3958 
3959 	ETHER_BPF_MTAP(ifp, m);
3960 }
3961 
3962 void
3963 if_vlancap(if_t ifh)
3964 {
3965 	struct ifnet *ifp = (struct ifnet *)ifh;
3966 	VLAN_CAPABILITIES(ifp);
3967 }
3968 
3969 void
3970 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3971 {
3972 	((struct ifnet *)ifp)->if_init = init_fn;
3973 }
3974 
3975 void
3976 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3977 {
3978 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3979 }
3980 
3981 void
3982 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3983 {
3984 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
3985 }
3986 
3987 void
3988 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3989 {
3990 	((struct ifnet *)ifp)->if_transmit = start_fn;
3991 }
3992 
3993 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3994 {
3995 	((struct ifnet *)ifp)->if_qflush = flush_fn;
3996 
3997 }
3998 
3999 void
4000 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4001 {
4002 
4003 	ifp->if_get_counter = fn;
4004 }
4005 
4006 /* Revisit these - These are inline functions originally. */
4007 int
4008 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4009 {
4010 	return drbr_inuse_drv(ifh, br);
4011 }
4012 
4013 struct mbuf*
4014 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4015 {
4016 	return drbr_dequeue(ifh, br);
4017 }
4018 
4019 int
4020 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4021 {
4022 	return drbr_needs_enqueue(ifh, br);
4023 }
4024 
4025 int
4026 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4027 {
4028 	return drbr_enqueue(ifh, br, m);
4029 
4030 }
4031