xref: /freebsd/sys/net/if.c (revision ee5cf11617a9b7f034d95c639bd4d27d1f09e848)
1 /*-
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62 
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #include <net/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/if.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif /* INET */
89 #ifdef INET6
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
92 #endif /* INET6 */
93 #endif /* INET || INET6 */
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
100 #endif
101 
102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
104 
105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106     &ifqmaxlen, 0, "max send queue size");
107 
108 /* Log link state change events */
109 static int log_link_state_change = 1;
110 
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 	&log_link_state_change, 0,
113 	"log interface link state change events");
114 
115 /* Log promiscuous mode change events */
116 static int log_promisc_mode_change = 1;
117 
118 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
119 	&log_promisc_mode_change, 1,
120 	"log promiscuous mode change events");
121 
122 /* Interface description */
123 static unsigned int ifdescr_maxlen = 1024;
124 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
125 	&ifdescr_maxlen, 0,
126 	"administrative maximum length for interface description");
127 
128 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
129 
130 /* global sx for non-critical path ifdescr */
131 static struct sx ifdescr_sx;
132 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
133 
134 void	(*bridge_linkstate_p)(struct ifnet *ifp);
135 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
136 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
137 /* These are external hooks for CARP. */
138 void	(*carp_linkstate_p)(struct ifnet *ifp);
139 void	(*carp_demote_adj_p)(int, char *);
140 int	(*carp_master_p)(struct ifaddr *);
141 #if defined(INET) || defined(INET6)
142 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
143 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
144     const struct sockaddr *sa);
145 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
146 int	(*carp_attach_p)(struct ifaddr *, int);
147 void	(*carp_detach_p)(struct ifaddr *);
148 #endif
149 #ifdef INET
150 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
151 #endif
152 #ifdef INET6
153 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
154 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
155     const struct in6_addr *taddr);
156 #endif
157 
158 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
159 
160 /*
161  * XXX: Style; these should be sorted alphabetically, and unprototyped
162  * static functions should be prototyped. Currently they are sorted by
163  * declaration order.
164  */
165 static void	if_attachdomain(void *);
166 static void	if_attachdomain1(struct ifnet *);
167 static int	ifconf(u_long, caddr_t);
168 static void	if_freemulti(struct ifmultiaddr *);
169 static void	if_grow(void);
170 static void	if_input_default(struct ifnet *, struct mbuf *);
171 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
172 static void	if_route(struct ifnet *, int flag, int fam);
173 static int	if_setflag(struct ifnet *, int, int, int *, int);
174 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
175 static void	if_unroute(struct ifnet *, int flag, int fam);
176 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
177 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
178 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
179 static void	do_link_state_change(void *, int);
180 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
181 static int	if_getgroupmembers(struct ifgroupreq *);
182 static void	if_delgroups(struct ifnet *);
183 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
184 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
185 #ifdef VIMAGE
186 static void	if_vmove(struct ifnet *, struct vnet *);
187 #endif
188 
189 #ifdef INET6
190 /*
191  * XXX: declare here to avoid to include many inet6 related files..
192  * should be more generalized?
193  */
194 extern void	nd6_setmtu(struct ifnet *);
195 #endif
196 
197 /* ipsec helper hooks */
198 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
199 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
200 
201 VNET_DEFINE(int, if_index);
202 int	ifqmaxlen = IFQ_MAXLEN;
203 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
204 VNET_DEFINE(struct ifgrouphead, ifg_head);
205 
206 static VNET_DEFINE(int, if_indexlim) = 8;
207 
208 /* Table of ifnet by index. */
209 VNET_DEFINE(struct ifnet **, ifindex_table);
210 
211 #define	V_if_indexlim		VNET(if_indexlim)
212 #define	V_ifindex_table		VNET(ifindex_table)
213 
214 /*
215  * The global network interface list (V_ifnet) and related state (such as
216  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
217  * an rwlock.  Either may be acquired shared to stablize the list, but both
218  * must be acquired writable to modify the list.  This model allows us to
219  * both stablize the interface list during interrupt thread processing, but
220  * also to stablize it over long-running ioctls, without introducing priority
221  * inversions and deadlocks.
222  */
223 struct rwlock ifnet_rwlock;
224 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
225 struct sx ifnet_sxlock;
226 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
227 
228 /*
229  * The allocation of network interfaces is a rather non-atomic affair; we
230  * need to select an index before we are ready to expose the interface for
231  * use, so will use this pointer value to indicate reservation.
232  */
233 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
234 
235 static	if_com_alloc_t *if_com_alloc[256];
236 static	if_com_free_t *if_com_free[256];
237 
238 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
239 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
240 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
241 
242 struct ifnet *
243 ifnet_byindex_locked(u_short idx)
244 {
245 
246 	if (idx > V_if_index)
247 		return (NULL);
248 	if (V_ifindex_table[idx] == IFNET_HOLD)
249 		return (NULL);
250 	return (V_ifindex_table[idx]);
251 }
252 
253 struct ifnet *
254 ifnet_byindex(u_short idx)
255 {
256 	struct ifnet *ifp;
257 
258 	IFNET_RLOCK_NOSLEEP();
259 	ifp = ifnet_byindex_locked(idx);
260 	IFNET_RUNLOCK_NOSLEEP();
261 	return (ifp);
262 }
263 
264 struct ifnet *
265 ifnet_byindex_ref(u_short idx)
266 {
267 	struct ifnet *ifp;
268 
269 	IFNET_RLOCK_NOSLEEP();
270 	ifp = ifnet_byindex_locked(idx);
271 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
272 		IFNET_RUNLOCK_NOSLEEP();
273 		return (NULL);
274 	}
275 	if_ref(ifp);
276 	IFNET_RUNLOCK_NOSLEEP();
277 	return (ifp);
278 }
279 
280 /*
281  * Allocate an ifindex array entry; return 0 on success or an error on
282  * failure.
283  */
284 static u_short
285 ifindex_alloc(void)
286 {
287 	u_short idx;
288 
289 	IFNET_WLOCK_ASSERT();
290 retry:
291 	/*
292 	 * Try to find an empty slot below V_if_index.  If we fail, take the
293 	 * next slot.
294 	 */
295 	for (idx = 1; idx <= V_if_index; idx++) {
296 		if (V_ifindex_table[idx] == NULL)
297 			break;
298 	}
299 
300 	/* Catch if_index overflow. */
301 	if (idx >= V_if_indexlim) {
302 		if_grow();
303 		goto retry;
304 	}
305 	if (idx > V_if_index)
306 		V_if_index = idx;
307 	return (idx);
308 }
309 
310 static void
311 ifindex_free_locked(u_short idx)
312 {
313 
314 	IFNET_WLOCK_ASSERT();
315 
316 	V_ifindex_table[idx] = NULL;
317 	while (V_if_index > 0 &&
318 	    V_ifindex_table[V_if_index] == NULL)
319 		V_if_index--;
320 }
321 
322 static void
323 ifindex_free(u_short idx)
324 {
325 
326 	IFNET_WLOCK();
327 	ifindex_free_locked(idx);
328 	IFNET_WUNLOCK();
329 }
330 
331 static void
332 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
333 {
334 
335 	IFNET_WLOCK_ASSERT();
336 
337 	V_ifindex_table[idx] = ifp;
338 }
339 
340 static void
341 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
342 {
343 
344 	IFNET_WLOCK();
345 	ifnet_setbyindex_locked(idx, ifp);
346 	IFNET_WUNLOCK();
347 }
348 
349 struct ifaddr *
350 ifaddr_byindex(u_short idx)
351 {
352 	struct ifnet *ifp;
353 	struct ifaddr *ifa = NULL;
354 
355 	IFNET_RLOCK_NOSLEEP();
356 	ifp = ifnet_byindex_locked(idx);
357 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
358 		ifa_ref(ifa);
359 	IFNET_RUNLOCK_NOSLEEP();
360 	return (ifa);
361 }
362 
363 /*
364  * Network interface utility routines.
365  *
366  * Routines with ifa_ifwith* names take sockaddr *'s as
367  * parameters.
368  */
369 
370 static void
371 vnet_if_init(const void *unused __unused)
372 {
373 
374 	TAILQ_INIT(&V_ifnet);
375 	TAILQ_INIT(&V_ifg_head);
376 	IFNET_WLOCK();
377 	if_grow();				/* create initial table */
378 	IFNET_WUNLOCK();
379 	vnet_if_clone_init();
380 }
381 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
382     NULL);
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 
399 static void
400 vnet_if_return(const void *unused __unused)
401 {
402 	struct ifnet *ifp, *nifp;
403 
404 	/* Return all inherited interfaces to their parent vnets. */
405 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
406 		if (ifp->if_home_vnet != ifp->if_vnet)
407 			if_vmove(ifp, ifp->if_home_vnet);
408 	}
409 }
410 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
411     vnet_if_return, NULL);
412 #endif
413 
414 static void
415 if_grow(void)
416 {
417 	int oldlim;
418 	u_int n;
419 	struct ifnet **e;
420 
421 	IFNET_WLOCK_ASSERT();
422 	oldlim = V_if_indexlim;
423 	IFNET_WUNLOCK();
424 	n = (oldlim << 1) * sizeof(*e);
425 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
426 	IFNET_WLOCK();
427 	if (V_if_indexlim != oldlim) {
428 		free(e, M_IFNET);
429 		return;
430 	}
431 	if (V_ifindex_table != NULL) {
432 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
433 		free((caddr_t)V_ifindex_table, M_IFNET);
434 	}
435 	V_if_indexlim <<= 1;
436 	V_ifindex_table = e;
437 }
438 
439 /*
440  * Allocate a struct ifnet and an index for an interface.  A layer 2
441  * common structure will also be allocated if an allocation routine is
442  * registered for the passed type.
443  */
444 struct ifnet *
445 if_alloc(u_char type)
446 {
447 	struct ifnet *ifp;
448 	u_short idx;
449 
450 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
451 	IFNET_WLOCK();
452 	idx = ifindex_alloc();
453 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
454 	IFNET_WUNLOCK();
455 	ifp->if_index = idx;
456 	ifp->if_type = type;
457 	ifp->if_alloctype = type;
458 	if (if_com_alloc[type] != NULL) {
459 		ifp->if_l2com = if_com_alloc[type](type, ifp);
460 		if (ifp->if_l2com == NULL) {
461 			free(ifp, M_IFNET);
462 			ifindex_free(idx);
463 			return (NULL);
464 		}
465 	}
466 
467 	IF_ADDR_LOCK_INIT(ifp);
468 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
469 	ifp->if_afdata_initialized = 0;
470 	IF_AFDATA_LOCK_INIT(ifp);
471 	TAILQ_INIT(&ifp->if_addrhead);
472 	TAILQ_INIT(&ifp->if_multiaddrs);
473 	TAILQ_INIT(&ifp->if_groups);
474 #ifdef MAC
475 	mac_ifnet_init(ifp);
476 #endif
477 	ifq_init(&ifp->if_snd, ifp);
478 
479 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
480 	for (int i = 0; i < IFCOUNTERS; i++)
481 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
482 	ifp->if_get_counter = if_get_counter_default;
483 	ifnet_setbyindex(ifp->if_index, ifp);
484 	return (ifp);
485 }
486 
487 /*
488  * Do the actual work of freeing a struct ifnet, and layer 2 common
489  * structure.  This call is made when the last reference to an
490  * interface is released.
491  */
492 static void
493 if_free_internal(struct ifnet *ifp)
494 {
495 
496 	KASSERT((ifp->if_flags & IFF_DYING),
497 	    ("if_free_internal: interface not dying"));
498 
499 	if (if_com_free[ifp->if_alloctype] != NULL)
500 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
501 		    ifp->if_alloctype);
502 
503 #ifdef MAC
504 	mac_ifnet_destroy(ifp);
505 #endif /* MAC */
506 	if (ifp->if_description != NULL)
507 		free(ifp->if_description, M_IFDESCR);
508 	IF_AFDATA_DESTROY(ifp);
509 	IF_ADDR_LOCK_DESTROY(ifp);
510 	ifq_delete(&ifp->if_snd);
511 
512 	for (int i = 0; i < IFCOUNTERS; i++)
513 		counter_u64_free(ifp->if_counters[i]);
514 
515 	free(ifp, M_IFNET);
516 }
517 
518 /*
519  * Deregister an interface and free the associated storage.
520  */
521 void
522 if_free(struct ifnet *ifp)
523 {
524 
525 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
526 
527 	CURVNET_SET_QUIET(ifp->if_vnet);
528 	IFNET_WLOCK();
529 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
530 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
531 
532 	ifindex_free_locked(ifp->if_index);
533 	IFNET_WUNLOCK();
534 
535 	if (refcount_release(&ifp->if_refcount))
536 		if_free_internal(ifp);
537 	CURVNET_RESTORE();
538 }
539 
540 /*
541  * Interfaces to keep an ifnet type-stable despite the possibility of the
542  * driver calling if_free().  If there are additional references, we defer
543  * freeing the underlying data structure.
544  */
545 void
546 if_ref(struct ifnet *ifp)
547 {
548 
549 	/* We don't assert the ifnet list lock here, but arguably should. */
550 	refcount_acquire(&ifp->if_refcount);
551 }
552 
553 void
554 if_rele(struct ifnet *ifp)
555 {
556 
557 	if (!refcount_release(&ifp->if_refcount))
558 		return;
559 	if_free_internal(ifp);
560 }
561 
562 void
563 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
564 {
565 
566 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
567 
568 	if (ifq->ifq_maxlen == 0)
569 		ifq->ifq_maxlen = ifqmaxlen;
570 
571 	ifq->altq_type = 0;
572 	ifq->altq_disc = NULL;
573 	ifq->altq_flags &= ALTQF_CANTCHANGE;
574 	ifq->altq_tbr  = NULL;
575 	ifq->altq_ifp  = ifp;
576 }
577 
578 void
579 ifq_delete(struct ifaltq *ifq)
580 {
581 	mtx_destroy(&ifq->ifq_mtx);
582 }
583 
584 /*
585  * Perform generic interface initialization tasks and attach the interface
586  * to the list of "active" interfaces.  If vmove flag is set on entry
587  * to if_attach_internal(), perform only a limited subset of initialization
588  * tasks, given that we are moving from one vnet to another an ifnet which
589  * has already been fully initialized.
590  *
591  * Note that if_detach_internal() removes group membership unconditionally
592  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
593  * Thus, when if_vmove() is applied to a cloned interface, group membership
594  * is lost while a cloned one always joins a group whose name is
595  * ifc->ifc_name.  To recover this after if_detach_internal() and
596  * if_attach_internal(), the cloner should be specified to
597  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
598  * attempts to join a group whose name is ifc->ifc_name.
599  *
600  * XXX:
601  *  - The decision to return void and thus require this function to
602  *    succeed is questionable.
603  *  - We should probably do more sanity checking.  For instance we don't
604  *    do anything to insure if_xname is unique or non-empty.
605  */
606 void
607 if_attach(struct ifnet *ifp)
608 {
609 
610 	if_attach_internal(ifp, 0, NULL);
611 }
612 
613 /*
614  * Compute the least common TSO limit.
615  */
616 void
617 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
618 {
619 	/*
620 	 * 1) If there is no limit currently, take the limit from
621 	 * the network adapter.
622 	 *
623 	 * 2) If the network adapter has a limit below the current
624 	 * limit, apply it.
625 	 */
626 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
627 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
628 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
629 	}
630 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
631 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
632 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
633 	}
634 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
635 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
636 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
637 	}
638 }
639 
640 /*
641  * Update TSO limit of a network adapter.
642  *
643  * Returns zero if no change. Else non-zero.
644  */
645 int
646 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
647 {
648 	int retval = 0;
649 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
650 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
651 		retval++;
652 	}
653 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
654 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
655 		retval++;
656 	}
657 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
658 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
659 		retval++;
660 	}
661 	return (retval);
662 }
663 
664 static void
665 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
666 {
667 	unsigned socksize, ifasize;
668 	int namelen, masklen;
669 	struct sockaddr_dl *sdl;
670 	struct ifaddr *ifa;
671 
672 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
673 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
674 		    ifp->if_xname);
675 
676 #ifdef VIMAGE
677 	ifp->if_vnet = curvnet;
678 	if (ifp->if_home_vnet == NULL)
679 		ifp->if_home_vnet = curvnet;
680 #endif
681 
682 	if_addgroup(ifp, IFG_ALL);
683 
684 	/* Restore group membership for cloned interfaces. */
685 	if (vmove && ifc != NULL)
686 		if_clone_addgroup(ifp, ifc);
687 
688 	getmicrotime(&ifp->if_lastchange);
689 	ifp->if_epoch = time_uptime;
690 
691 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
692 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
693 	    ("transmit and qflush must both either be set or both be NULL"));
694 	if (ifp->if_transmit == NULL) {
695 		ifp->if_transmit = if_transmit;
696 		ifp->if_qflush = if_qflush;
697 	}
698 	if (ifp->if_input == NULL)
699 		ifp->if_input = if_input_default;
700 
701 	if (ifp->if_requestencap == NULL)
702 		ifp->if_requestencap = if_requestencap_default;
703 
704 	if (!vmove) {
705 #ifdef MAC
706 		mac_ifnet_create(ifp);
707 #endif
708 
709 		/*
710 		 * Create a Link Level name for this device.
711 		 */
712 		namelen = strlen(ifp->if_xname);
713 		/*
714 		 * Always save enough space for any possiable name so we
715 		 * can do a rename in place later.
716 		 */
717 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
718 		socksize = masklen + ifp->if_addrlen;
719 		if (socksize < sizeof(*sdl))
720 			socksize = sizeof(*sdl);
721 		socksize = roundup2(socksize, sizeof(long));
722 		ifasize = sizeof(*ifa) + 2 * socksize;
723 		ifa = ifa_alloc(ifasize, M_WAITOK);
724 		sdl = (struct sockaddr_dl *)(ifa + 1);
725 		sdl->sdl_len = socksize;
726 		sdl->sdl_family = AF_LINK;
727 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
728 		sdl->sdl_nlen = namelen;
729 		sdl->sdl_index = ifp->if_index;
730 		sdl->sdl_type = ifp->if_type;
731 		ifp->if_addr = ifa;
732 		ifa->ifa_ifp = ifp;
733 		ifa->ifa_rtrequest = link_rtrequest;
734 		ifa->ifa_addr = (struct sockaddr *)sdl;
735 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
736 		ifa->ifa_netmask = (struct sockaddr *)sdl;
737 		sdl->sdl_len = masklen;
738 		while (namelen != 0)
739 			sdl->sdl_data[--namelen] = 0xff;
740 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
741 		/* Reliably crash if used uninitialized. */
742 		ifp->if_broadcastaddr = NULL;
743 
744 #if defined(INET) || defined(INET6)
745 		/* Use defaults for TSO, if nothing is set */
746 		if (ifp->if_hw_tsomax == 0 &&
747 		    ifp->if_hw_tsomaxsegcount == 0 &&
748 		    ifp->if_hw_tsomaxsegsize == 0) {
749 			/*
750 			 * The TSO defaults needs to be such that an
751 			 * NFS mbuf list of 35 mbufs totalling just
752 			 * below 64K works and that a chain of mbufs
753 			 * can be defragged into at most 32 segments:
754 			 */
755 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
756 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
757 			ifp->if_hw_tsomaxsegcount = 35;
758 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
759 
760 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
761 			if (ifp->if_capabilities & IFCAP_TSO) {
762 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
763 				    ifp->if_hw_tsomax,
764 				    ifp->if_hw_tsomaxsegcount,
765 				    ifp->if_hw_tsomaxsegsize);
766 			}
767 		}
768 #endif
769 	}
770 #ifdef VIMAGE
771 	else {
772 		/*
773 		 * Update the interface index in the link layer address
774 		 * of the interface.
775 		 */
776 		for (ifa = ifp->if_addr; ifa != NULL;
777 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
778 			if (ifa->ifa_addr->sa_family == AF_LINK) {
779 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
780 				sdl->sdl_index = ifp->if_index;
781 			}
782 		}
783 	}
784 #endif
785 
786 	IFNET_WLOCK();
787 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
788 #ifdef VIMAGE
789 	curvnet->vnet_ifcnt++;
790 #endif
791 	IFNET_WUNLOCK();
792 
793 	if (domain_init_status >= 2)
794 		if_attachdomain1(ifp);
795 
796 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
797 	if (IS_DEFAULT_VNET(curvnet))
798 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
799 
800 	/* Announce the interface. */
801 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
802 }
803 
804 static void
805 if_attachdomain(void *dummy)
806 {
807 	struct ifnet *ifp;
808 
809 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
810 		if_attachdomain1(ifp);
811 }
812 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
813     if_attachdomain, NULL);
814 
815 static void
816 if_attachdomain1(struct ifnet *ifp)
817 {
818 	struct domain *dp;
819 
820 	/*
821 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
822 	 * cannot lock ifp->if_afdata initialization, entirely.
823 	 */
824 	IF_AFDATA_LOCK(ifp);
825 	if (ifp->if_afdata_initialized >= domain_init_status) {
826 		IF_AFDATA_UNLOCK(ifp);
827 		log(LOG_WARNING, "%s called more than once on %s\n",
828 		    __func__, ifp->if_xname);
829 		return;
830 	}
831 	ifp->if_afdata_initialized = domain_init_status;
832 	IF_AFDATA_UNLOCK(ifp);
833 
834 	/* address family dependent data region */
835 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
836 	for (dp = domains; dp; dp = dp->dom_next) {
837 		if (dp->dom_ifattach)
838 			ifp->if_afdata[dp->dom_family] =
839 			    (*dp->dom_ifattach)(ifp);
840 	}
841 }
842 
843 /*
844  * Remove any unicast or broadcast network addresses from an interface.
845  */
846 void
847 if_purgeaddrs(struct ifnet *ifp)
848 {
849 	struct ifaddr *ifa, *next;
850 
851 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
852 		if (ifa->ifa_addr->sa_family == AF_LINK)
853 			continue;
854 #ifdef INET
855 		/* XXX: Ugly!! ad hoc just for INET */
856 		if (ifa->ifa_addr->sa_family == AF_INET) {
857 			struct ifaliasreq ifr;
858 
859 			bzero(&ifr, sizeof(ifr));
860 			ifr.ifra_addr = *ifa->ifa_addr;
861 			if (ifa->ifa_dstaddr)
862 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
863 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
864 			    NULL) == 0)
865 				continue;
866 		}
867 #endif /* INET */
868 #ifdef INET6
869 		if (ifa->ifa_addr->sa_family == AF_INET6) {
870 			in6_purgeaddr(ifa);
871 			/* ifp_addrhead is already updated */
872 			continue;
873 		}
874 #endif /* INET6 */
875 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
876 		ifa_free(ifa);
877 	}
878 }
879 
880 /*
881  * Remove any multicast network addresses from an interface when an ifnet
882  * is going away.
883  */
884 static void
885 if_purgemaddrs(struct ifnet *ifp)
886 {
887 	struct ifmultiaddr *ifma;
888 	struct ifmultiaddr *next;
889 
890 	IF_ADDR_WLOCK(ifp);
891 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
892 		if_delmulti_locked(ifp, ifma, 1);
893 	IF_ADDR_WUNLOCK(ifp);
894 }
895 
896 /*
897  * Detach an interface, removing it from the list of "active" interfaces.
898  * If vmove flag is set on entry to if_detach_internal(), perform only a
899  * limited subset of cleanup tasks, given that we are moving an ifnet from
900  * one vnet to another, where it must be fully operational.
901  *
902  * XXXRW: There are some significant questions about event ordering, and
903  * how to prevent things from starting to use the interface during detach.
904  */
905 void
906 if_detach(struct ifnet *ifp)
907 {
908 
909 	CURVNET_SET_QUIET(ifp->if_vnet);
910 	if_detach_internal(ifp, 0, NULL);
911 	CURVNET_RESTORE();
912 }
913 
914 static int
915 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
916 {
917 	struct ifaddr *ifa;
918 	int i;
919 	struct domain *dp;
920  	struct ifnet *iter;
921  	int found = 0;
922 
923 	IFNET_WLOCK();
924 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
925 		if (iter == ifp) {
926 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
927 			found = 1;
928 			break;
929 		}
930 #ifdef VIMAGE
931 	if (found)
932 		curvnet->vnet_ifcnt--;
933 #endif
934 	IFNET_WUNLOCK();
935 	if (!found) {
936 		/*
937 		 * While we would want to panic here, we cannot
938 		 * guarantee that the interface is indeed still on
939 		 * the list given we don't hold locks all the way.
940 		 */
941 		return (ENOENT);
942 #if 0
943 		if (vmove)
944 			panic("%s: ifp=%p not on the ifnet tailq %p",
945 			    __func__, ifp, &V_ifnet);
946 		else
947 			return; /* XXX this should panic as well? */
948 #endif
949 	}
950 
951 	/* Check if this is a cloned interface or not. */
952 	if (vmove && ifcp != NULL)
953 		*ifcp = if_clone_findifc(ifp);
954 
955 	/*
956 	 * Remove/wait for pending events.
957 	 */
958 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
959 
960 	/*
961 	 * Remove routes and flush queues.
962 	 */
963 	if_down(ifp);
964 #ifdef ALTQ
965 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
966 		altq_disable(&ifp->if_snd);
967 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
968 		altq_detach(&ifp->if_snd);
969 #endif
970 
971 	if_purgeaddrs(ifp);
972 
973 #ifdef INET
974 	in_ifdetach(ifp);
975 #endif
976 
977 #ifdef INET6
978 	/*
979 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
980 	 * before removing routing entries below, since IPv6 interface direct
981 	 * routes are expected to be removed by the IPv6-specific kernel API.
982 	 * Otherwise, the kernel will detect some inconsistency and bark it.
983 	 */
984 	in6_ifdetach(ifp);
985 #endif
986 	if_purgemaddrs(ifp);
987 
988 	/* Announce that the interface is gone. */
989 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
990 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
991 	if (IS_DEFAULT_VNET(curvnet))
992 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
993 
994 	if (!vmove) {
995 		/*
996 		 * Prevent further calls into the device driver via ifnet.
997 		 */
998 		if_dead(ifp);
999 
1000 		/*
1001 		 * Remove link ifaddr pointer and maybe decrement if_index.
1002 		 * Clean up all addresses.
1003 		 */
1004 		ifp->if_addr = NULL;
1005 
1006 		/* We can now free link ifaddr. */
1007 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1008 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1009 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1010 			ifa_free(ifa);
1011 		}
1012 	}
1013 
1014 	rt_flushifroutes(ifp);
1015 	if_delgroups(ifp);
1016 
1017 	/*
1018 	 * We cannot hold the lock over dom_ifdetach calls as they might
1019 	 * sleep, for example trying to drain a callout, thus open up the
1020 	 * theoretical race with re-attaching.
1021 	 */
1022 	IF_AFDATA_LOCK(ifp);
1023 	i = ifp->if_afdata_initialized;
1024 	ifp->if_afdata_initialized = 0;
1025 	IF_AFDATA_UNLOCK(ifp);
1026 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1027 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1028 			(*dp->dom_ifdetach)(ifp,
1029 			    ifp->if_afdata[dp->dom_family]);
1030 	}
1031 
1032 	return (0);
1033 }
1034 
1035 #ifdef VIMAGE
1036 /*
1037  * if_vmove() performs a limited version of if_detach() in current
1038  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1039  * An attempt is made to shrink if_index in current vnet, find an
1040  * unused if_index in target vnet and calls if_grow() if necessary,
1041  * and finally find an unused if_xname for the target vnet.
1042  */
1043 void
1044 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1045 {
1046 	struct if_clone *ifc;
1047 	u_int bif_dlt, bif_hdrlen;
1048 	int rc;
1049 
1050  	/*
1051 	 * if_detach_internal() will call the eventhandler to notify
1052 	 * interface departure.  That will detach if_bpf.  We need to
1053 	 * safe the dlt and hdrlen so we can re-attach it later.
1054 	 */
1055 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1056 
1057 	/*
1058 	 * Detach from current vnet, but preserve LLADDR info, do not
1059 	 * mark as dead etc. so that the ifnet can be reattached later.
1060 	 * If we cannot find it, we lost the race to someone else.
1061 	 */
1062 	rc = if_detach_internal(ifp, 1, &ifc);
1063 	if (rc != 0)
1064 		return;
1065 
1066 	/*
1067 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1068 	 * the if_index for that vnet if possible.
1069 	 *
1070 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1071 	 * or we'd lock on one vnet and unlock on another.
1072 	 */
1073 	IFNET_WLOCK();
1074 	ifindex_free_locked(ifp->if_index);
1075 	IFNET_WUNLOCK();
1076 
1077 	/*
1078 	 * Perform interface-specific reassignment tasks, if provided by
1079 	 * the driver.
1080 	 */
1081 	if (ifp->if_reassign != NULL)
1082 		ifp->if_reassign(ifp, new_vnet, NULL);
1083 
1084 	/*
1085 	 * Switch to the context of the target vnet.
1086 	 */
1087 	CURVNET_SET_QUIET(new_vnet);
1088 
1089 	IFNET_WLOCK();
1090 	ifp->if_index = ifindex_alloc();
1091 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1092 	IFNET_WUNLOCK();
1093 
1094 	if_attach_internal(ifp, 1, ifc);
1095 
1096 	if (ifp->if_bpf == NULL)
1097 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1098 
1099 	CURVNET_RESTORE();
1100 }
1101 
1102 /*
1103  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1104  */
1105 static int
1106 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1107 {
1108 	struct prison *pr;
1109 	struct ifnet *difp;
1110 
1111 	/* Try to find the prison within our visibility. */
1112 	sx_slock(&allprison_lock);
1113 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1114 	sx_sunlock(&allprison_lock);
1115 	if (pr == NULL)
1116 		return (ENXIO);
1117 	prison_hold_locked(pr);
1118 	mtx_unlock(&pr->pr_mtx);
1119 
1120 	/* Do not try to move the iface from and to the same prison. */
1121 	if (pr->pr_vnet == ifp->if_vnet) {
1122 		prison_free(pr);
1123 		return (EEXIST);
1124 	}
1125 
1126 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1127 	/* XXX Lock interfaces to avoid races. */
1128 	CURVNET_SET_QUIET(pr->pr_vnet);
1129 	difp = ifunit(ifname);
1130 	CURVNET_RESTORE();
1131 	if (difp != NULL) {
1132 		prison_free(pr);
1133 		return (EEXIST);
1134 	}
1135 
1136 	/* Move the interface into the child jail/vnet. */
1137 	if_vmove(ifp, pr->pr_vnet);
1138 
1139 	/* Report the new if_xname back to the userland. */
1140 	sprintf(ifname, "%s", ifp->if_xname);
1141 
1142 	prison_free(pr);
1143 	return (0);
1144 }
1145 
1146 static int
1147 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1148 {
1149 	struct prison *pr;
1150 	struct vnet *vnet_dst;
1151 	struct ifnet *ifp;
1152 
1153 	/* Try to find the prison within our visibility. */
1154 	sx_slock(&allprison_lock);
1155 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1156 	sx_sunlock(&allprison_lock);
1157 	if (pr == NULL)
1158 		return (ENXIO);
1159 	prison_hold_locked(pr);
1160 	mtx_unlock(&pr->pr_mtx);
1161 
1162 	/* Make sure the named iface exists in the source prison/vnet. */
1163 	CURVNET_SET(pr->pr_vnet);
1164 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1165 	if (ifp == NULL) {
1166 		CURVNET_RESTORE();
1167 		prison_free(pr);
1168 		return (ENXIO);
1169 	}
1170 
1171 	/* Do not try to move the iface from and to the same prison. */
1172 	vnet_dst = TD_TO_VNET(td);
1173 	if (vnet_dst == ifp->if_vnet) {
1174 		CURVNET_RESTORE();
1175 		prison_free(pr);
1176 		return (EEXIST);
1177 	}
1178 
1179 	/* Get interface back from child jail/vnet. */
1180 	if_vmove(ifp, vnet_dst);
1181 	CURVNET_RESTORE();
1182 
1183 	/* Report the new if_xname back to the userland. */
1184 	sprintf(ifname, "%s", ifp->if_xname);
1185 
1186 	prison_free(pr);
1187 	return (0);
1188 }
1189 #endif /* VIMAGE */
1190 
1191 /*
1192  * Add a group to an interface
1193  */
1194 int
1195 if_addgroup(struct ifnet *ifp, const char *groupname)
1196 {
1197 	struct ifg_list		*ifgl;
1198 	struct ifg_group	*ifg = NULL;
1199 	struct ifg_member	*ifgm;
1200 	int 			 new = 0;
1201 
1202 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1203 	    groupname[strlen(groupname) - 1] <= '9')
1204 		return (EINVAL);
1205 
1206 	IFNET_WLOCK();
1207 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1208 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1209 			IFNET_WUNLOCK();
1210 			return (EEXIST);
1211 		}
1212 
1213 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1214 	    M_NOWAIT)) == NULL) {
1215 	    	IFNET_WUNLOCK();
1216 		return (ENOMEM);
1217 	}
1218 
1219 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1220 	    M_TEMP, M_NOWAIT)) == NULL) {
1221 		free(ifgl, M_TEMP);
1222 		IFNET_WUNLOCK();
1223 		return (ENOMEM);
1224 	}
1225 
1226 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1227 		if (!strcmp(ifg->ifg_group, groupname))
1228 			break;
1229 
1230 	if (ifg == NULL) {
1231 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1232 		    M_TEMP, M_NOWAIT)) == NULL) {
1233 			free(ifgl, M_TEMP);
1234 			free(ifgm, M_TEMP);
1235 			IFNET_WUNLOCK();
1236 			return (ENOMEM);
1237 		}
1238 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1239 		ifg->ifg_refcnt = 0;
1240 		TAILQ_INIT(&ifg->ifg_members);
1241 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1242 		new = 1;
1243 	}
1244 
1245 	ifg->ifg_refcnt++;
1246 	ifgl->ifgl_group = ifg;
1247 	ifgm->ifgm_ifp = ifp;
1248 
1249 	IF_ADDR_WLOCK(ifp);
1250 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1251 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1252 	IF_ADDR_WUNLOCK(ifp);
1253 
1254 	IFNET_WUNLOCK();
1255 
1256 	if (new)
1257 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1258 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1259 
1260 	return (0);
1261 }
1262 
1263 /*
1264  * Remove a group from an interface
1265  */
1266 int
1267 if_delgroup(struct ifnet *ifp, const char *groupname)
1268 {
1269 	struct ifg_list		*ifgl;
1270 	struct ifg_member	*ifgm;
1271 
1272 	IFNET_WLOCK();
1273 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1274 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1275 			break;
1276 	if (ifgl == NULL) {
1277 		IFNET_WUNLOCK();
1278 		return (ENOENT);
1279 	}
1280 
1281 	IF_ADDR_WLOCK(ifp);
1282 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1283 	IF_ADDR_WUNLOCK(ifp);
1284 
1285 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1286 		if (ifgm->ifgm_ifp == ifp)
1287 			break;
1288 
1289 	if (ifgm != NULL) {
1290 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1291 		free(ifgm, M_TEMP);
1292 	}
1293 
1294 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1295 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1296 		IFNET_WUNLOCK();
1297 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1298 		free(ifgl->ifgl_group, M_TEMP);
1299 	} else
1300 		IFNET_WUNLOCK();
1301 
1302 	free(ifgl, M_TEMP);
1303 
1304 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1305 
1306 	return (0);
1307 }
1308 
1309 /*
1310  * Remove an interface from all groups
1311  */
1312 static void
1313 if_delgroups(struct ifnet *ifp)
1314 {
1315 	struct ifg_list		*ifgl;
1316 	struct ifg_member	*ifgm;
1317 	char groupname[IFNAMSIZ];
1318 
1319 	IFNET_WLOCK();
1320 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1321 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1322 
1323 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1324 
1325 		IF_ADDR_WLOCK(ifp);
1326 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1327 		IF_ADDR_WUNLOCK(ifp);
1328 
1329 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1330 			if (ifgm->ifgm_ifp == ifp)
1331 				break;
1332 
1333 		if (ifgm != NULL) {
1334 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1335 			    ifgm_next);
1336 			free(ifgm, M_TEMP);
1337 		}
1338 
1339 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1340 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1341 			IFNET_WUNLOCK();
1342 			EVENTHANDLER_INVOKE(group_detach_event,
1343 			    ifgl->ifgl_group);
1344 			free(ifgl->ifgl_group, M_TEMP);
1345 		} else
1346 			IFNET_WUNLOCK();
1347 
1348 		free(ifgl, M_TEMP);
1349 
1350 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1351 
1352 		IFNET_WLOCK();
1353 	}
1354 	IFNET_WUNLOCK();
1355 }
1356 
1357 /*
1358  * Stores all groups from an interface in memory pointed
1359  * to by data
1360  */
1361 static int
1362 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1363 {
1364 	int			 len, error;
1365 	struct ifg_list		*ifgl;
1366 	struct ifg_req		 ifgrq, *ifgp;
1367 	struct ifgroupreq	*ifgr = data;
1368 
1369 	if (ifgr->ifgr_len == 0) {
1370 		IF_ADDR_RLOCK(ifp);
1371 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1372 			ifgr->ifgr_len += sizeof(struct ifg_req);
1373 		IF_ADDR_RUNLOCK(ifp);
1374 		return (0);
1375 	}
1376 
1377 	len = ifgr->ifgr_len;
1378 	ifgp = ifgr->ifgr_groups;
1379 	/* XXX: wire */
1380 	IF_ADDR_RLOCK(ifp);
1381 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1382 		if (len < sizeof(ifgrq)) {
1383 			IF_ADDR_RUNLOCK(ifp);
1384 			return (EINVAL);
1385 		}
1386 		bzero(&ifgrq, sizeof ifgrq);
1387 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1388 		    sizeof(ifgrq.ifgrq_group));
1389 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1390 		    	IF_ADDR_RUNLOCK(ifp);
1391 			return (error);
1392 		}
1393 		len -= sizeof(ifgrq);
1394 		ifgp++;
1395 	}
1396 	IF_ADDR_RUNLOCK(ifp);
1397 
1398 	return (0);
1399 }
1400 
1401 /*
1402  * Stores all members of a group in memory pointed to by data
1403  */
1404 static int
1405 if_getgroupmembers(struct ifgroupreq *data)
1406 {
1407 	struct ifgroupreq	*ifgr = data;
1408 	struct ifg_group	*ifg;
1409 	struct ifg_member	*ifgm;
1410 	struct ifg_req		 ifgrq, *ifgp;
1411 	int			 len, error;
1412 
1413 	IFNET_RLOCK();
1414 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1415 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1416 			break;
1417 	if (ifg == NULL) {
1418 		IFNET_RUNLOCK();
1419 		return (ENOENT);
1420 	}
1421 
1422 	if (ifgr->ifgr_len == 0) {
1423 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1424 			ifgr->ifgr_len += sizeof(ifgrq);
1425 		IFNET_RUNLOCK();
1426 		return (0);
1427 	}
1428 
1429 	len = ifgr->ifgr_len;
1430 	ifgp = ifgr->ifgr_groups;
1431 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1432 		if (len < sizeof(ifgrq)) {
1433 			IFNET_RUNLOCK();
1434 			return (EINVAL);
1435 		}
1436 		bzero(&ifgrq, sizeof ifgrq);
1437 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1438 		    sizeof(ifgrq.ifgrq_member));
1439 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1440 			IFNET_RUNLOCK();
1441 			return (error);
1442 		}
1443 		len -= sizeof(ifgrq);
1444 		ifgp++;
1445 	}
1446 	IFNET_RUNLOCK();
1447 
1448 	return (0);
1449 }
1450 
1451 /*
1452  * Return counter values from counter(9)s stored in ifnet.
1453  */
1454 uint64_t
1455 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1456 {
1457 
1458 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1459 
1460 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1461 }
1462 
1463 /*
1464  * Increase an ifnet counter. Usually used for counters shared
1465  * between the stack and a driver, but function supports them all.
1466  */
1467 void
1468 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1469 {
1470 
1471 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1472 
1473 	counter_u64_add(ifp->if_counters[cnt], inc);
1474 }
1475 
1476 /*
1477  * Copy data from ifnet to userland API structure if_data.
1478  */
1479 void
1480 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1481 {
1482 
1483 	ifd->ifi_type = ifp->if_type;
1484 	ifd->ifi_physical = 0;
1485 	ifd->ifi_addrlen = ifp->if_addrlen;
1486 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1487 	ifd->ifi_link_state = ifp->if_link_state;
1488 	ifd->ifi_vhid = 0;
1489 	ifd->ifi_datalen = sizeof(struct if_data);
1490 	ifd->ifi_mtu = ifp->if_mtu;
1491 	ifd->ifi_metric = ifp->if_metric;
1492 	ifd->ifi_baudrate = ifp->if_baudrate;
1493 	ifd->ifi_hwassist = ifp->if_hwassist;
1494 	ifd->ifi_epoch = ifp->if_epoch;
1495 	ifd->ifi_lastchange = ifp->if_lastchange;
1496 
1497 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1498 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1499 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1500 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1501 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1502 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1503 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1504 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1505 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1506 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1507 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1508 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1509 }
1510 
1511 /*
1512  * Wrapper functions for struct ifnet address list locking macros.  These are
1513  * used by kernel modules to avoid encoding programming interface or binary
1514  * interface assumptions that may be violated when kernel-internal locking
1515  * approaches change.
1516  */
1517 void
1518 if_addr_rlock(struct ifnet *ifp)
1519 {
1520 
1521 	IF_ADDR_RLOCK(ifp);
1522 }
1523 
1524 void
1525 if_addr_runlock(struct ifnet *ifp)
1526 {
1527 
1528 	IF_ADDR_RUNLOCK(ifp);
1529 }
1530 
1531 void
1532 if_maddr_rlock(if_t ifp)
1533 {
1534 
1535 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1536 }
1537 
1538 void
1539 if_maddr_runlock(if_t ifp)
1540 {
1541 
1542 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1543 }
1544 
1545 /*
1546  * Initialization, destruction and refcounting functions for ifaddrs.
1547  */
1548 struct ifaddr *
1549 ifa_alloc(size_t size, int flags)
1550 {
1551 	struct ifaddr *ifa;
1552 
1553 	KASSERT(size >= sizeof(struct ifaddr),
1554 	    ("%s: invalid size %zu", __func__, size));
1555 
1556 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1557 	if (ifa == NULL)
1558 		return (NULL);
1559 
1560 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1561 		goto fail;
1562 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1563 		goto fail;
1564 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1565 		goto fail;
1566 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1567 		goto fail;
1568 
1569 	refcount_init(&ifa->ifa_refcnt, 1);
1570 
1571 	return (ifa);
1572 
1573 fail:
1574 	/* free(NULL) is okay */
1575 	counter_u64_free(ifa->ifa_opackets);
1576 	counter_u64_free(ifa->ifa_ipackets);
1577 	counter_u64_free(ifa->ifa_obytes);
1578 	counter_u64_free(ifa->ifa_ibytes);
1579 	free(ifa, M_IFADDR);
1580 
1581 	return (NULL);
1582 }
1583 
1584 void
1585 ifa_ref(struct ifaddr *ifa)
1586 {
1587 
1588 	refcount_acquire(&ifa->ifa_refcnt);
1589 }
1590 
1591 void
1592 ifa_free(struct ifaddr *ifa)
1593 {
1594 
1595 	if (refcount_release(&ifa->ifa_refcnt)) {
1596 		counter_u64_free(ifa->ifa_opackets);
1597 		counter_u64_free(ifa->ifa_ipackets);
1598 		counter_u64_free(ifa->ifa_obytes);
1599 		counter_u64_free(ifa->ifa_ibytes);
1600 		free(ifa, M_IFADDR);
1601 	}
1602 }
1603 
1604 static int
1605 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1606     struct sockaddr *ia)
1607 {
1608 	int error;
1609 	struct rt_addrinfo info;
1610 	struct sockaddr_dl null_sdl;
1611 	struct ifnet *ifp;
1612 
1613 	ifp = ifa->ifa_ifp;
1614 
1615 	bzero(&info, sizeof(info));
1616 	if (cmd != RTM_DELETE)
1617 		info.rti_ifp = V_loif;
1618 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1619 	info.rti_info[RTAX_DST] = ia;
1620 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1621 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1622 
1623 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1624 
1625 	if (error != 0)
1626 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1627 		    __func__, otype, if_name(ifp), error);
1628 
1629 	return (error);
1630 }
1631 
1632 int
1633 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1634 {
1635 
1636 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1637 }
1638 
1639 int
1640 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1641 {
1642 
1643 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1644 }
1645 
1646 int
1647 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1648 {
1649 
1650 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1651 }
1652 
1653 /*
1654  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1655  * structs used to represent other address families, it is necessary
1656  * to perform a different comparison.
1657  */
1658 
1659 #define	sa_dl_equal(a1, a2)	\
1660 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1661 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1662 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1663 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1664 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1665 
1666 /*
1667  * Locate an interface based on a complete address.
1668  */
1669 /*ARGSUSED*/
1670 static struct ifaddr *
1671 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1672 {
1673 	struct ifnet *ifp;
1674 	struct ifaddr *ifa;
1675 
1676 	IFNET_RLOCK_NOSLEEP();
1677 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1678 		IF_ADDR_RLOCK(ifp);
1679 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1680 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1681 				continue;
1682 			if (sa_equal(addr, ifa->ifa_addr)) {
1683 				if (getref)
1684 					ifa_ref(ifa);
1685 				IF_ADDR_RUNLOCK(ifp);
1686 				goto done;
1687 			}
1688 			/* IP6 doesn't have broadcast */
1689 			if ((ifp->if_flags & IFF_BROADCAST) &&
1690 			    ifa->ifa_broadaddr &&
1691 			    ifa->ifa_broadaddr->sa_len != 0 &&
1692 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1693 				if (getref)
1694 					ifa_ref(ifa);
1695 				IF_ADDR_RUNLOCK(ifp);
1696 				goto done;
1697 			}
1698 		}
1699 		IF_ADDR_RUNLOCK(ifp);
1700 	}
1701 	ifa = NULL;
1702 done:
1703 	IFNET_RUNLOCK_NOSLEEP();
1704 	return (ifa);
1705 }
1706 
1707 struct ifaddr *
1708 ifa_ifwithaddr(const struct sockaddr *addr)
1709 {
1710 
1711 	return (ifa_ifwithaddr_internal(addr, 1));
1712 }
1713 
1714 int
1715 ifa_ifwithaddr_check(const struct sockaddr *addr)
1716 {
1717 
1718 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1719 }
1720 
1721 /*
1722  * Locate an interface based on the broadcast address.
1723  */
1724 /* ARGSUSED */
1725 struct ifaddr *
1726 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1727 {
1728 	struct ifnet *ifp;
1729 	struct ifaddr *ifa;
1730 
1731 	IFNET_RLOCK_NOSLEEP();
1732 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1733 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1734 			continue;
1735 		IF_ADDR_RLOCK(ifp);
1736 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1737 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1738 				continue;
1739 			if ((ifp->if_flags & IFF_BROADCAST) &&
1740 			    ifa->ifa_broadaddr &&
1741 			    ifa->ifa_broadaddr->sa_len != 0 &&
1742 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1743 				ifa_ref(ifa);
1744 				IF_ADDR_RUNLOCK(ifp);
1745 				goto done;
1746 			}
1747 		}
1748 		IF_ADDR_RUNLOCK(ifp);
1749 	}
1750 	ifa = NULL;
1751 done:
1752 	IFNET_RUNLOCK_NOSLEEP();
1753 	return (ifa);
1754 }
1755 
1756 /*
1757  * Locate the point to point interface with a given destination address.
1758  */
1759 /*ARGSUSED*/
1760 struct ifaddr *
1761 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1762 {
1763 	struct ifnet *ifp;
1764 	struct ifaddr *ifa;
1765 
1766 	IFNET_RLOCK_NOSLEEP();
1767 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1768 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1769 			continue;
1770 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1771 			continue;
1772 		IF_ADDR_RLOCK(ifp);
1773 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1774 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1775 				continue;
1776 			if (ifa->ifa_dstaddr != NULL &&
1777 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1778 				ifa_ref(ifa);
1779 				IF_ADDR_RUNLOCK(ifp);
1780 				goto done;
1781 			}
1782 		}
1783 		IF_ADDR_RUNLOCK(ifp);
1784 	}
1785 	ifa = NULL;
1786 done:
1787 	IFNET_RUNLOCK_NOSLEEP();
1788 	return (ifa);
1789 }
1790 
1791 /*
1792  * Find an interface on a specific network.  If many, choice
1793  * is most specific found.
1794  */
1795 struct ifaddr *
1796 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1797 {
1798 	struct ifnet *ifp;
1799 	struct ifaddr *ifa;
1800 	struct ifaddr *ifa_maybe = NULL;
1801 	u_int af = addr->sa_family;
1802 	const char *addr_data = addr->sa_data, *cplim;
1803 
1804 	/*
1805 	 * AF_LINK addresses can be looked up directly by their index number,
1806 	 * so do that if we can.
1807 	 */
1808 	if (af == AF_LINK) {
1809 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1810 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1811 		return (ifaddr_byindex(sdl->sdl_index));
1812 	}
1813 
1814 	/*
1815 	 * Scan though each interface, looking for ones that have addresses
1816 	 * in this address family and the requested fib.  Maintain a reference
1817 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1818 	 * kept it stable when we move onto the next interface.
1819 	 */
1820 	IFNET_RLOCK_NOSLEEP();
1821 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1822 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1823 			continue;
1824 		IF_ADDR_RLOCK(ifp);
1825 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1826 			const char *cp, *cp2, *cp3;
1827 
1828 			if (ifa->ifa_addr->sa_family != af)
1829 next:				continue;
1830 			if (af == AF_INET &&
1831 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1832 				/*
1833 				 * This is a bit broken as it doesn't
1834 				 * take into account that the remote end may
1835 				 * be a single node in the network we are
1836 				 * looking for.
1837 				 * The trouble is that we don't know the
1838 				 * netmask for the remote end.
1839 				 */
1840 				if (ifa->ifa_dstaddr != NULL &&
1841 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1842 					ifa_ref(ifa);
1843 					IF_ADDR_RUNLOCK(ifp);
1844 					goto done;
1845 				}
1846 			} else {
1847 				/*
1848 				 * Scan all the bits in the ifa's address.
1849 				 * If a bit dissagrees with what we are
1850 				 * looking for, mask it with the netmask
1851 				 * to see if it really matters.
1852 				 * (A byte at a time)
1853 				 */
1854 				if (ifa->ifa_netmask == 0)
1855 					continue;
1856 				cp = addr_data;
1857 				cp2 = ifa->ifa_addr->sa_data;
1858 				cp3 = ifa->ifa_netmask->sa_data;
1859 				cplim = ifa->ifa_netmask->sa_len
1860 					+ (char *)ifa->ifa_netmask;
1861 				while (cp3 < cplim)
1862 					if ((*cp++ ^ *cp2++) & *cp3++)
1863 						goto next; /* next address! */
1864 				/*
1865 				 * If the netmask of what we just found
1866 				 * is more specific than what we had before
1867 				 * (if we had one), or if the virtual status
1868 				 * of new prefix is better than of the old one,
1869 				 * then remember the new one before continuing
1870 				 * to search for an even better one.
1871 				 */
1872 				if (ifa_maybe == NULL ||
1873 				    ifa_preferred(ifa_maybe, ifa) ||
1874 				    rn_refines((caddr_t)ifa->ifa_netmask,
1875 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1876 					if (ifa_maybe != NULL)
1877 						ifa_free(ifa_maybe);
1878 					ifa_maybe = ifa;
1879 					ifa_ref(ifa_maybe);
1880 				}
1881 			}
1882 		}
1883 		IF_ADDR_RUNLOCK(ifp);
1884 	}
1885 	ifa = ifa_maybe;
1886 	ifa_maybe = NULL;
1887 done:
1888 	IFNET_RUNLOCK_NOSLEEP();
1889 	if (ifa_maybe != NULL)
1890 		ifa_free(ifa_maybe);
1891 	return (ifa);
1892 }
1893 
1894 /*
1895  * Find an interface address specific to an interface best matching
1896  * a given address.
1897  */
1898 struct ifaddr *
1899 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1900 {
1901 	struct ifaddr *ifa;
1902 	const char *cp, *cp2, *cp3;
1903 	char *cplim;
1904 	struct ifaddr *ifa_maybe = NULL;
1905 	u_int af = addr->sa_family;
1906 
1907 	if (af >= AF_MAX)
1908 		return (NULL);
1909 	IF_ADDR_RLOCK(ifp);
1910 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1911 		if (ifa->ifa_addr->sa_family != af)
1912 			continue;
1913 		if (ifa_maybe == NULL)
1914 			ifa_maybe = ifa;
1915 		if (ifa->ifa_netmask == 0) {
1916 			if (sa_equal(addr, ifa->ifa_addr) ||
1917 			    (ifa->ifa_dstaddr &&
1918 			    sa_equal(addr, ifa->ifa_dstaddr)))
1919 				goto done;
1920 			continue;
1921 		}
1922 		if (ifp->if_flags & IFF_POINTOPOINT) {
1923 			if (sa_equal(addr, ifa->ifa_dstaddr))
1924 				goto done;
1925 		} else {
1926 			cp = addr->sa_data;
1927 			cp2 = ifa->ifa_addr->sa_data;
1928 			cp3 = ifa->ifa_netmask->sa_data;
1929 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1930 			for (; cp3 < cplim; cp3++)
1931 				if ((*cp++ ^ *cp2++) & *cp3)
1932 					break;
1933 			if (cp3 == cplim)
1934 				goto done;
1935 		}
1936 	}
1937 	ifa = ifa_maybe;
1938 done:
1939 	if (ifa != NULL)
1940 		ifa_ref(ifa);
1941 	IF_ADDR_RUNLOCK(ifp);
1942 	return (ifa);
1943 }
1944 
1945 /*
1946  * See whether new ifa is better than current one:
1947  * 1) A non-virtual one is preferred over virtual.
1948  * 2) A virtual in master state preferred over any other state.
1949  *
1950  * Used in several address selecting functions.
1951  */
1952 int
1953 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1954 {
1955 
1956 	return (cur->ifa_carp && (!next->ifa_carp ||
1957 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1958 }
1959 
1960 #include <net/if_llatbl.h>
1961 
1962 /*
1963  * Default action when installing a route with a Link Level gateway.
1964  * Lookup an appropriate real ifa to point to.
1965  * This should be moved to /sys/net/link.c eventually.
1966  */
1967 static void
1968 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1969 {
1970 	struct ifaddr *ifa, *oifa;
1971 	struct sockaddr *dst;
1972 	struct ifnet *ifp;
1973 
1974 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
1975 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
1976 		return;
1977 	ifa = ifaof_ifpforaddr(dst, ifp);
1978 	if (ifa) {
1979 		oifa = rt->rt_ifa;
1980 		rt->rt_ifa = ifa;
1981 		ifa_free(oifa);
1982 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1983 			ifa->ifa_rtrequest(cmd, rt, info);
1984 	}
1985 }
1986 
1987 struct sockaddr_dl *
1988 link_alloc_sdl(size_t size, int flags)
1989 {
1990 
1991 	return (malloc(size, M_TEMP, flags));
1992 }
1993 
1994 void
1995 link_free_sdl(struct sockaddr *sa)
1996 {
1997 	free(sa, M_TEMP);
1998 }
1999 
2000 /*
2001  * Fills in given sdl with interface basic info.
2002  * Returns pointer to filled sdl.
2003  */
2004 struct sockaddr_dl *
2005 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2006 {
2007 	struct sockaddr_dl *sdl;
2008 
2009 	sdl = (struct sockaddr_dl *)paddr;
2010 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2011 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2012 	sdl->sdl_family = AF_LINK;
2013 	sdl->sdl_index = ifp->if_index;
2014 	sdl->sdl_type = iftype;
2015 
2016 	return (sdl);
2017 }
2018 
2019 /*
2020  * Mark an interface down and notify protocols of
2021  * the transition.
2022  */
2023 static void
2024 if_unroute(struct ifnet *ifp, int flag, int fam)
2025 {
2026 	struct ifaddr *ifa;
2027 
2028 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2029 
2030 	ifp->if_flags &= ~flag;
2031 	getmicrotime(&ifp->if_lastchange);
2032 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2033 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2034 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2035 	ifp->if_qflush(ifp);
2036 
2037 	if (ifp->if_carp)
2038 		(*carp_linkstate_p)(ifp);
2039 	rt_ifmsg(ifp);
2040 }
2041 
2042 /*
2043  * Mark an interface up and notify protocols of
2044  * the transition.
2045  */
2046 static void
2047 if_route(struct ifnet *ifp, int flag, int fam)
2048 {
2049 	struct ifaddr *ifa;
2050 
2051 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2052 
2053 	ifp->if_flags |= flag;
2054 	getmicrotime(&ifp->if_lastchange);
2055 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2056 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2057 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2058 	if (ifp->if_carp)
2059 		(*carp_linkstate_p)(ifp);
2060 	rt_ifmsg(ifp);
2061 #ifdef INET6
2062 	in6_if_up(ifp);
2063 #endif
2064 }
2065 
2066 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2067 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2068 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2069 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2070 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2071 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2072 void	*(*vlan_cookie_p)(struct ifnet *);
2073 
2074 /*
2075  * Handle a change in the interface link state. To avoid LORs
2076  * between driver lock and upper layer locks, as well as possible
2077  * recursions, we post event to taskqueue, and all job
2078  * is done in static do_link_state_change().
2079  */
2080 void
2081 if_link_state_change(struct ifnet *ifp, int link_state)
2082 {
2083 	/* Return if state hasn't changed. */
2084 	if (ifp->if_link_state == link_state)
2085 		return;
2086 
2087 	ifp->if_link_state = link_state;
2088 
2089 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2090 }
2091 
2092 static void
2093 do_link_state_change(void *arg, int pending)
2094 {
2095 	struct ifnet *ifp = (struct ifnet *)arg;
2096 	int link_state = ifp->if_link_state;
2097 	CURVNET_SET(ifp->if_vnet);
2098 
2099 	/* Notify that the link state has changed. */
2100 	rt_ifmsg(ifp);
2101 	if (ifp->if_vlantrunk != NULL)
2102 		(*vlan_link_state_p)(ifp);
2103 
2104 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2105 	    ifp->if_l2com != NULL)
2106 		(*ng_ether_link_state_p)(ifp, link_state);
2107 	if (ifp->if_carp)
2108 		(*carp_linkstate_p)(ifp);
2109 	if (ifp->if_bridge)
2110 		(*bridge_linkstate_p)(ifp);
2111 	if (ifp->if_lagg)
2112 		(*lagg_linkstate_p)(ifp, link_state);
2113 
2114 	if (IS_DEFAULT_VNET(curvnet))
2115 		devctl_notify("IFNET", ifp->if_xname,
2116 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2117 		    NULL);
2118 	if (pending > 1)
2119 		if_printf(ifp, "%d link states coalesced\n", pending);
2120 	if (log_link_state_change)
2121 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2122 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2123 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2124 	CURVNET_RESTORE();
2125 }
2126 
2127 /*
2128  * Mark an interface down and notify protocols of
2129  * the transition.
2130  */
2131 void
2132 if_down(struct ifnet *ifp)
2133 {
2134 
2135 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2136 }
2137 
2138 /*
2139  * Mark an interface up and notify protocols of
2140  * the transition.
2141  */
2142 void
2143 if_up(struct ifnet *ifp)
2144 {
2145 
2146 	if_route(ifp, IFF_UP, AF_UNSPEC);
2147 }
2148 
2149 /*
2150  * Flush an interface queue.
2151  */
2152 void
2153 if_qflush(struct ifnet *ifp)
2154 {
2155 	struct mbuf *m, *n;
2156 	struct ifaltq *ifq;
2157 
2158 	ifq = &ifp->if_snd;
2159 	IFQ_LOCK(ifq);
2160 #ifdef ALTQ
2161 	if (ALTQ_IS_ENABLED(ifq))
2162 		ALTQ_PURGE(ifq);
2163 #endif
2164 	n = ifq->ifq_head;
2165 	while ((m = n) != NULL) {
2166 		n = m->m_nextpkt;
2167 		m_freem(m);
2168 	}
2169 	ifq->ifq_head = 0;
2170 	ifq->ifq_tail = 0;
2171 	ifq->ifq_len = 0;
2172 	IFQ_UNLOCK(ifq);
2173 }
2174 
2175 /*
2176  * Map interface name to interface structure pointer, with or without
2177  * returning a reference.
2178  */
2179 struct ifnet *
2180 ifunit_ref(const char *name)
2181 {
2182 	struct ifnet *ifp;
2183 
2184 	IFNET_RLOCK_NOSLEEP();
2185 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2186 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2187 		    !(ifp->if_flags & IFF_DYING))
2188 			break;
2189 	}
2190 	if (ifp != NULL)
2191 		if_ref(ifp);
2192 	IFNET_RUNLOCK_NOSLEEP();
2193 	return (ifp);
2194 }
2195 
2196 struct ifnet *
2197 ifunit(const char *name)
2198 {
2199 	struct ifnet *ifp;
2200 
2201 	IFNET_RLOCK_NOSLEEP();
2202 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2203 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2204 			break;
2205 	}
2206 	IFNET_RUNLOCK_NOSLEEP();
2207 	return (ifp);
2208 }
2209 
2210 /*
2211  * Hardware specific interface ioctls.
2212  */
2213 static int
2214 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2215 {
2216 	struct ifreq *ifr;
2217 	int error = 0;
2218 	int new_flags, temp_flags;
2219 	size_t namelen, onamelen;
2220 	size_t descrlen;
2221 	char *descrbuf, *odescrbuf;
2222 	char new_name[IFNAMSIZ];
2223 	struct ifaddr *ifa;
2224 	struct sockaddr_dl *sdl;
2225 
2226 	ifr = (struct ifreq *)data;
2227 	switch (cmd) {
2228 	case SIOCGIFINDEX:
2229 		ifr->ifr_index = ifp->if_index;
2230 		break;
2231 
2232 	case SIOCGIFFLAGS:
2233 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2234 		ifr->ifr_flags = temp_flags & 0xffff;
2235 		ifr->ifr_flagshigh = temp_flags >> 16;
2236 		break;
2237 
2238 	case SIOCGIFCAP:
2239 		ifr->ifr_reqcap = ifp->if_capabilities;
2240 		ifr->ifr_curcap = ifp->if_capenable;
2241 		break;
2242 
2243 #ifdef MAC
2244 	case SIOCGIFMAC:
2245 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2246 		break;
2247 #endif
2248 
2249 	case SIOCGIFMETRIC:
2250 		ifr->ifr_metric = ifp->if_metric;
2251 		break;
2252 
2253 	case SIOCGIFMTU:
2254 		ifr->ifr_mtu = ifp->if_mtu;
2255 		break;
2256 
2257 	case SIOCGIFPHYS:
2258 		/* XXXGL: did this ever worked? */
2259 		ifr->ifr_phys = 0;
2260 		break;
2261 
2262 	case SIOCGIFDESCR:
2263 		error = 0;
2264 		sx_slock(&ifdescr_sx);
2265 		if (ifp->if_description == NULL)
2266 			error = ENOMSG;
2267 		else {
2268 			/* space for terminating nul */
2269 			descrlen = strlen(ifp->if_description) + 1;
2270 			if (ifr->ifr_buffer.length < descrlen)
2271 				ifr->ifr_buffer.buffer = NULL;
2272 			else
2273 				error = copyout(ifp->if_description,
2274 				    ifr->ifr_buffer.buffer, descrlen);
2275 			ifr->ifr_buffer.length = descrlen;
2276 		}
2277 		sx_sunlock(&ifdescr_sx);
2278 		break;
2279 
2280 	case SIOCSIFDESCR:
2281 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2282 		if (error)
2283 			return (error);
2284 
2285 		/*
2286 		 * Copy only (length-1) bytes to make sure that
2287 		 * if_description is always nul terminated.  The
2288 		 * length parameter is supposed to count the
2289 		 * terminating nul in.
2290 		 */
2291 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2292 			return (ENAMETOOLONG);
2293 		else if (ifr->ifr_buffer.length == 0)
2294 			descrbuf = NULL;
2295 		else {
2296 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2297 			    M_WAITOK | M_ZERO);
2298 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2299 			    ifr->ifr_buffer.length - 1);
2300 			if (error) {
2301 				free(descrbuf, M_IFDESCR);
2302 				break;
2303 			}
2304 		}
2305 
2306 		sx_xlock(&ifdescr_sx);
2307 		odescrbuf = ifp->if_description;
2308 		ifp->if_description = descrbuf;
2309 		sx_xunlock(&ifdescr_sx);
2310 
2311 		getmicrotime(&ifp->if_lastchange);
2312 		free(odescrbuf, M_IFDESCR);
2313 		break;
2314 
2315 	case SIOCGIFFIB:
2316 		ifr->ifr_fib = ifp->if_fib;
2317 		break;
2318 
2319 	case SIOCSIFFIB:
2320 		error = priv_check(td, PRIV_NET_SETIFFIB);
2321 		if (error)
2322 			return (error);
2323 		if (ifr->ifr_fib >= rt_numfibs)
2324 			return (EINVAL);
2325 
2326 		ifp->if_fib = ifr->ifr_fib;
2327 		break;
2328 
2329 	case SIOCSIFFLAGS:
2330 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2331 		if (error)
2332 			return (error);
2333 		/*
2334 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2335 		 * check, so we don't need special handling here yet.
2336 		 */
2337 		new_flags = (ifr->ifr_flags & 0xffff) |
2338 		    (ifr->ifr_flagshigh << 16);
2339 		if (ifp->if_flags & IFF_UP &&
2340 		    (new_flags & IFF_UP) == 0) {
2341 			if_down(ifp);
2342 		} else if (new_flags & IFF_UP &&
2343 		    (ifp->if_flags & IFF_UP) == 0) {
2344 			if_up(ifp);
2345 		}
2346 		/* See if permanently promiscuous mode bit is about to flip */
2347 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2348 			if (new_flags & IFF_PPROMISC)
2349 				ifp->if_flags |= IFF_PROMISC;
2350 			else if (ifp->if_pcount == 0)
2351 				ifp->if_flags &= ~IFF_PROMISC;
2352 			if (log_promisc_mode_change)
2353                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2354                                     ifp->if_xname,
2355                                     ((new_flags & IFF_PPROMISC) ?
2356                                      "enabled" : "disabled"));
2357 		}
2358 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2359 			(new_flags &~ IFF_CANTCHANGE);
2360 		if (ifp->if_ioctl) {
2361 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2362 		}
2363 		getmicrotime(&ifp->if_lastchange);
2364 		break;
2365 
2366 	case SIOCSIFCAP:
2367 		error = priv_check(td, PRIV_NET_SETIFCAP);
2368 		if (error)
2369 			return (error);
2370 		if (ifp->if_ioctl == NULL)
2371 			return (EOPNOTSUPP);
2372 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2373 			return (EINVAL);
2374 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2375 		if (error == 0)
2376 			getmicrotime(&ifp->if_lastchange);
2377 		break;
2378 
2379 #ifdef MAC
2380 	case SIOCSIFMAC:
2381 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2382 		break;
2383 #endif
2384 
2385 	case SIOCSIFNAME:
2386 		error = priv_check(td, PRIV_NET_SETIFNAME);
2387 		if (error)
2388 			return (error);
2389 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2390 		if (error != 0)
2391 			return (error);
2392 		if (new_name[0] == '\0')
2393 			return (EINVAL);
2394 		if (new_name[IFNAMSIZ-1] != '\0') {
2395 			new_name[IFNAMSIZ-1] = '\0';
2396 			if (strlen(new_name) == IFNAMSIZ-1)
2397 				return (EINVAL);
2398 		}
2399 		if (ifunit(new_name) != NULL)
2400 			return (EEXIST);
2401 
2402 		/*
2403 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2404 		 * and there are numerous other races with the
2405 		 * ifunit() checks not being atomic with namespace
2406 		 * changes (renames, vmoves, if_attach, etc).
2407 		 */
2408 		ifp->if_flags |= IFF_RENAMING;
2409 
2410 		/* Announce the departure of the interface. */
2411 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2412 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2413 
2414 		log(LOG_INFO, "%s: changing name to '%s'\n",
2415 		    ifp->if_xname, new_name);
2416 
2417 		IF_ADDR_WLOCK(ifp);
2418 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2419 		ifa = ifp->if_addr;
2420 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2421 		namelen = strlen(new_name);
2422 		onamelen = sdl->sdl_nlen;
2423 		/*
2424 		 * Move the address if needed.  This is safe because we
2425 		 * allocate space for a name of length IFNAMSIZ when we
2426 		 * create this in if_attach().
2427 		 */
2428 		if (namelen != onamelen) {
2429 			bcopy(sdl->sdl_data + onamelen,
2430 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2431 		}
2432 		bcopy(new_name, sdl->sdl_data, namelen);
2433 		sdl->sdl_nlen = namelen;
2434 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2435 		bzero(sdl->sdl_data, onamelen);
2436 		while (namelen != 0)
2437 			sdl->sdl_data[--namelen] = 0xff;
2438 		IF_ADDR_WUNLOCK(ifp);
2439 
2440 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2441 		/* Announce the return of the interface. */
2442 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2443 
2444 		ifp->if_flags &= ~IFF_RENAMING;
2445 		break;
2446 
2447 #ifdef VIMAGE
2448 	case SIOCSIFVNET:
2449 		error = priv_check(td, PRIV_NET_SETIFVNET);
2450 		if (error)
2451 			return (error);
2452 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2453 		break;
2454 #endif
2455 
2456 	case SIOCSIFMETRIC:
2457 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2458 		if (error)
2459 			return (error);
2460 		ifp->if_metric = ifr->ifr_metric;
2461 		getmicrotime(&ifp->if_lastchange);
2462 		break;
2463 
2464 	case SIOCSIFPHYS:
2465 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2466 		if (error)
2467 			return (error);
2468 		if (ifp->if_ioctl == NULL)
2469 			return (EOPNOTSUPP);
2470 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2471 		if (error == 0)
2472 			getmicrotime(&ifp->if_lastchange);
2473 		break;
2474 
2475 	case SIOCSIFMTU:
2476 	{
2477 		u_long oldmtu = ifp->if_mtu;
2478 
2479 		error = priv_check(td, PRIV_NET_SETIFMTU);
2480 		if (error)
2481 			return (error);
2482 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2483 			return (EINVAL);
2484 		if (ifp->if_ioctl == NULL)
2485 			return (EOPNOTSUPP);
2486 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2487 		if (error == 0) {
2488 			getmicrotime(&ifp->if_lastchange);
2489 			rt_ifmsg(ifp);
2490 		}
2491 		/*
2492 		 * If the link MTU changed, do network layer specific procedure.
2493 		 */
2494 		if (ifp->if_mtu != oldmtu) {
2495 #ifdef INET6
2496 			nd6_setmtu(ifp);
2497 #endif
2498 			rt_updatemtu(ifp);
2499 		}
2500 		break;
2501 	}
2502 
2503 	case SIOCADDMULTI:
2504 	case SIOCDELMULTI:
2505 		if (cmd == SIOCADDMULTI)
2506 			error = priv_check(td, PRIV_NET_ADDMULTI);
2507 		else
2508 			error = priv_check(td, PRIV_NET_DELMULTI);
2509 		if (error)
2510 			return (error);
2511 
2512 		/* Don't allow group membership on non-multicast interfaces. */
2513 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2514 			return (EOPNOTSUPP);
2515 
2516 		/* Don't let users screw up protocols' entries. */
2517 		if (ifr->ifr_addr.sa_family != AF_LINK)
2518 			return (EINVAL);
2519 
2520 		if (cmd == SIOCADDMULTI) {
2521 			struct ifmultiaddr *ifma;
2522 
2523 			/*
2524 			 * Userland is only permitted to join groups once
2525 			 * via the if_addmulti() KPI, because it cannot hold
2526 			 * struct ifmultiaddr * between calls. It may also
2527 			 * lose a race while we check if the membership
2528 			 * already exists.
2529 			 */
2530 			IF_ADDR_RLOCK(ifp);
2531 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2532 			IF_ADDR_RUNLOCK(ifp);
2533 			if (ifma != NULL)
2534 				error = EADDRINUSE;
2535 			else
2536 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2537 		} else {
2538 			error = if_delmulti(ifp, &ifr->ifr_addr);
2539 		}
2540 		if (error == 0)
2541 			getmicrotime(&ifp->if_lastchange);
2542 		break;
2543 
2544 	case SIOCSIFPHYADDR:
2545 	case SIOCDIFPHYADDR:
2546 #ifdef INET6
2547 	case SIOCSIFPHYADDR_IN6:
2548 #endif
2549 	case SIOCSIFMEDIA:
2550 	case SIOCSIFGENERIC:
2551 		error = priv_check(td, PRIV_NET_HWIOCTL);
2552 		if (error)
2553 			return (error);
2554 		if (ifp->if_ioctl == NULL)
2555 			return (EOPNOTSUPP);
2556 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2557 		if (error == 0)
2558 			getmicrotime(&ifp->if_lastchange);
2559 		break;
2560 
2561 	case SIOCGIFSTATUS:
2562 	case SIOCGIFPSRCADDR:
2563 	case SIOCGIFPDSTADDR:
2564 	case SIOCGIFMEDIA:
2565 	case SIOCGIFXMEDIA:
2566 	case SIOCGIFGENERIC:
2567 		if (ifp->if_ioctl == NULL)
2568 			return (EOPNOTSUPP);
2569 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2570 		break;
2571 
2572 	case SIOCSIFLLADDR:
2573 		error = priv_check(td, PRIV_NET_SETLLADDR);
2574 		if (error)
2575 			return (error);
2576 		error = if_setlladdr(ifp,
2577 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2578 		break;
2579 
2580 	case SIOCAIFGROUP:
2581 	{
2582 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2583 
2584 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2585 		if (error)
2586 			return (error);
2587 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2588 			return (error);
2589 		break;
2590 	}
2591 
2592 	case SIOCGIFGROUP:
2593 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2594 			return (error);
2595 		break;
2596 
2597 	case SIOCDIFGROUP:
2598 	{
2599 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2600 
2601 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2602 		if (error)
2603 			return (error);
2604 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2605 			return (error);
2606 		break;
2607 	}
2608 
2609 	default:
2610 		error = ENOIOCTL;
2611 		break;
2612 	}
2613 	return (error);
2614 }
2615 
2616 #ifdef COMPAT_FREEBSD32
2617 struct ifconf32 {
2618 	int32_t	ifc_len;
2619 	union {
2620 		uint32_t	ifcu_buf;
2621 		uint32_t	ifcu_req;
2622 	} ifc_ifcu;
2623 };
2624 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2625 #endif
2626 
2627 /*
2628  * Interface ioctls.
2629  */
2630 int
2631 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2632 {
2633 	struct ifnet *ifp;
2634 	struct ifreq *ifr;
2635 	int error;
2636 	int oif_flags;
2637 
2638 	CURVNET_SET(so->so_vnet);
2639 	switch (cmd) {
2640 	case SIOCGIFCONF:
2641 		error = ifconf(cmd, data);
2642 		CURVNET_RESTORE();
2643 		return (error);
2644 
2645 #ifdef COMPAT_FREEBSD32
2646 	case SIOCGIFCONF32:
2647 		{
2648 			struct ifconf32 *ifc32;
2649 			struct ifconf ifc;
2650 
2651 			ifc32 = (struct ifconf32 *)data;
2652 			ifc.ifc_len = ifc32->ifc_len;
2653 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2654 
2655 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2656 			CURVNET_RESTORE();
2657 			if (error == 0)
2658 				ifc32->ifc_len = ifc.ifc_len;
2659 			return (error);
2660 		}
2661 #endif
2662 	}
2663 	ifr = (struct ifreq *)data;
2664 
2665 	switch (cmd) {
2666 #ifdef VIMAGE
2667 	case SIOCSIFRVNET:
2668 		error = priv_check(td, PRIV_NET_SETIFVNET);
2669 		if (error == 0)
2670 			error = if_vmove_reclaim(td, ifr->ifr_name,
2671 			    ifr->ifr_jid);
2672 		CURVNET_RESTORE();
2673 		return (error);
2674 #endif
2675 	case SIOCIFCREATE:
2676 	case SIOCIFCREATE2:
2677 		error = priv_check(td, PRIV_NET_IFCREATE);
2678 		if (error == 0)
2679 			error = if_clone_create(ifr->ifr_name,
2680 			    sizeof(ifr->ifr_name),
2681 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2682 		CURVNET_RESTORE();
2683 		return (error);
2684 	case SIOCIFDESTROY:
2685 		error = priv_check(td, PRIV_NET_IFDESTROY);
2686 		if (error == 0)
2687 			error = if_clone_destroy(ifr->ifr_name);
2688 		CURVNET_RESTORE();
2689 		return (error);
2690 
2691 	case SIOCIFGCLONERS:
2692 		error = if_clone_list((struct if_clonereq *)data);
2693 		CURVNET_RESTORE();
2694 		return (error);
2695 	case SIOCGIFGMEMB:
2696 		error = if_getgroupmembers((struct ifgroupreq *)data);
2697 		CURVNET_RESTORE();
2698 		return (error);
2699 #if defined(INET) || defined(INET6)
2700 	case SIOCSVH:
2701 	case SIOCGVH:
2702 		if (carp_ioctl_p == NULL)
2703 			error = EPROTONOSUPPORT;
2704 		else
2705 			error = (*carp_ioctl_p)(ifr, cmd, td);
2706 		CURVNET_RESTORE();
2707 		return (error);
2708 #endif
2709 	}
2710 
2711 	ifp = ifunit_ref(ifr->ifr_name);
2712 	if (ifp == NULL) {
2713 		CURVNET_RESTORE();
2714 		return (ENXIO);
2715 	}
2716 
2717 	error = ifhwioctl(cmd, ifp, data, td);
2718 	if (error != ENOIOCTL) {
2719 		if_rele(ifp);
2720 		CURVNET_RESTORE();
2721 		return (error);
2722 	}
2723 
2724 	oif_flags = ifp->if_flags;
2725 	if (so->so_proto == NULL) {
2726 		if_rele(ifp);
2727 		CURVNET_RESTORE();
2728 		return (EOPNOTSUPP);
2729 	}
2730 
2731 	/*
2732 	 * Pass the request on to the socket control method, and if the
2733 	 * latter returns EOPNOTSUPP, directly to the interface.
2734 	 *
2735 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2736 	 * trust SIOCSIFADDR et al to come from an already privileged
2737 	 * layer, and do not perform any credentials checks or input
2738 	 * validation.
2739 	 */
2740 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2741 	    ifp, td));
2742 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2743 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2744 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2745 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2746 
2747 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2748 #ifdef INET6
2749 		if (ifp->if_flags & IFF_UP)
2750 			in6_if_up(ifp);
2751 #endif
2752 	}
2753 	if_rele(ifp);
2754 	CURVNET_RESTORE();
2755 	return (error);
2756 }
2757 
2758 /*
2759  * The code common to handling reference counted flags,
2760  * e.g., in ifpromisc() and if_allmulti().
2761  * The "pflag" argument can specify a permanent mode flag to check,
2762  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2763  *
2764  * Only to be used on stack-owned flags, not driver-owned flags.
2765  */
2766 static int
2767 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2768 {
2769 	struct ifreq ifr;
2770 	int error;
2771 	int oldflags, oldcount;
2772 
2773 	/* Sanity checks to catch programming errors */
2774 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2775 	    ("%s: setting driver-owned flag %d", __func__, flag));
2776 
2777 	if (onswitch)
2778 		KASSERT(*refcount >= 0,
2779 		    ("%s: increment negative refcount %d for flag %d",
2780 		    __func__, *refcount, flag));
2781 	else
2782 		KASSERT(*refcount > 0,
2783 		    ("%s: decrement non-positive refcount %d for flag %d",
2784 		    __func__, *refcount, flag));
2785 
2786 	/* In case this mode is permanent, just touch refcount */
2787 	if (ifp->if_flags & pflag) {
2788 		*refcount += onswitch ? 1 : -1;
2789 		return (0);
2790 	}
2791 
2792 	/* Save ifnet parameters for if_ioctl() may fail */
2793 	oldcount = *refcount;
2794 	oldflags = ifp->if_flags;
2795 
2796 	/*
2797 	 * See if we aren't the only and touching refcount is enough.
2798 	 * Actually toggle interface flag if we are the first or last.
2799 	 */
2800 	if (onswitch) {
2801 		if ((*refcount)++)
2802 			return (0);
2803 		ifp->if_flags |= flag;
2804 	} else {
2805 		if (--(*refcount))
2806 			return (0);
2807 		ifp->if_flags &= ~flag;
2808 	}
2809 
2810 	/* Call down the driver since we've changed interface flags */
2811 	if (ifp->if_ioctl == NULL) {
2812 		error = EOPNOTSUPP;
2813 		goto recover;
2814 	}
2815 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2816 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2817 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2818 	if (error)
2819 		goto recover;
2820 	/* Notify userland that interface flags have changed */
2821 	rt_ifmsg(ifp);
2822 	return (0);
2823 
2824 recover:
2825 	/* Recover after driver error */
2826 	*refcount = oldcount;
2827 	ifp->if_flags = oldflags;
2828 	return (error);
2829 }
2830 
2831 /*
2832  * Set/clear promiscuous mode on interface ifp based on the truth value
2833  * of pswitch.  The calls are reference counted so that only the first
2834  * "on" request actually has an effect, as does the final "off" request.
2835  * Results are undefined if the "off" and "on" requests are not matched.
2836  */
2837 int
2838 ifpromisc(struct ifnet *ifp, int pswitch)
2839 {
2840 	int error;
2841 	int oldflags = ifp->if_flags;
2842 
2843 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2844 			   &ifp->if_pcount, pswitch);
2845 	/* If promiscuous mode status has changed, log a message */
2846 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2847             log_promisc_mode_change)
2848 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2849 		    ifp->if_xname,
2850 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2851 	return (error);
2852 }
2853 
2854 /*
2855  * Return interface configuration
2856  * of system.  List may be used
2857  * in later ioctl's (above) to get
2858  * other information.
2859  */
2860 /*ARGSUSED*/
2861 static int
2862 ifconf(u_long cmd, caddr_t data)
2863 {
2864 	struct ifconf *ifc = (struct ifconf *)data;
2865 	struct ifnet *ifp;
2866 	struct ifaddr *ifa;
2867 	struct ifreq ifr;
2868 	struct sbuf *sb;
2869 	int error, full = 0, valid_len, max_len;
2870 
2871 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2872 	max_len = MAXPHYS - 1;
2873 
2874 	/* Prevent hostile input from being able to crash the system */
2875 	if (ifc->ifc_len <= 0)
2876 		return (EINVAL);
2877 
2878 again:
2879 	if (ifc->ifc_len <= max_len) {
2880 		max_len = ifc->ifc_len;
2881 		full = 1;
2882 	}
2883 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2884 	max_len = 0;
2885 	valid_len = 0;
2886 
2887 	IFNET_RLOCK();
2888 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2889 		int addrs;
2890 
2891 		/*
2892 		 * Zero the ifr_name buffer to make sure we don't
2893 		 * disclose the contents of the stack.
2894 		 */
2895 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2896 
2897 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2898 		    >= sizeof(ifr.ifr_name)) {
2899 			sbuf_delete(sb);
2900 			IFNET_RUNLOCK();
2901 			return (ENAMETOOLONG);
2902 		}
2903 
2904 		addrs = 0;
2905 		IF_ADDR_RLOCK(ifp);
2906 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2907 			struct sockaddr *sa = ifa->ifa_addr;
2908 
2909 			if (prison_if(curthread->td_ucred, sa) != 0)
2910 				continue;
2911 			addrs++;
2912 			if (sa->sa_len <= sizeof(*sa)) {
2913 				ifr.ifr_addr = *sa;
2914 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2915 				max_len += sizeof(ifr);
2916 			} else {
2917 				sbuf_bcat(sb, &ifr,
2918 				    offsetof(struct ifreq, ifr_addr));
2919 				max_len += offsetof(struct ifreq, ifr_addr);
2920 				sbuf_bcat(sb, sa, sa->sa_len);
2921 				max_len += sa->sa_len;
2922 			}
2923 
2924 			if (sbuf_error(sb) == 0)
2925 				valid_len = sbuf_len(sb);
2926 		}
2927 		IF_ADDR_RUNLOCK(ifp);
2928 		if (addrs == 0) {
2929 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2930 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2931 			max_len += sizeof(ifr);
2932 
2933 			if (sbuf_error(sb) == 0)
2934 				valid_len = sbuf_len(sb);
2935 		}
2936 	}
2937 	IFNET_RUNLOCK();
2938 
2939 	/*
2940 	 * If we didn't allocate enough space (uncommon), try again.  If
2941 	 * we have already allocated as much space as we are allowed,
2942 	 * return what we've got.
2943 	 */
2944 	if (valid_len != max_len && !full) {
2945 		sbuf_delete(sb);
2946 		goto again;
2947 	}
2948 
2949 	ifc->ifc_len = valid_len;
2950 	sbuf_finish(sb);
2951 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2952 	sbuf_delete(sb);
2953 	return (error);
2954 }
2955 
2956 /*
2957  * Just like ifpromisc(), but for all-multicast-reception mode.
2958  */
2959 int
2960 if_allmulti(struct ifnet *ifp, int onswitch)
2961 {
2962 
2963 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2964 }
2965 
2966 struct ifmultiaddr *
2967 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
2968 {
2969 	struct ifmultiaddr *ifma;
2970 
2971 	IF_ADDR_LOCK_ASSERT(ifp);
2972 
2973 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2974 		if (sa->sa_family == AF_LINK) {
2975 			if (sa_dl_equal(ifma->ifma_addr, sa))
2976 				break;
2977 		} else {
2978 			if (sa_equal(ifma->ifma_addr, sa))
2979 				break;
2980 		}
2981 	}
2982 
2983 	return ifma;
2984 }
2985 
2986 /*
2987  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2988  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2989  * the ifnet multicast address list here, so the caller must do that and
2990  * other setup work (such as notifying the device driver).  The reference
2991  * count is initialized to 1.
2992  */
2993 static struct ifmultiaddr *
2994 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2995     int mflags)
2996 {
2997 	struct ifmultiaddr *ifma;
2998 	struct sockaddr *dupsa;
2999 
3000 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3001 	    M_ZERO);
3002 	if (ifma == NULL)
3003 		return (NULL);
3004 
3005 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3006 	if (dupsa == NULL) {
3007 		free(ifma, M_IFMADDR);
3008 		return (NULL);
3009 	}
3010 	bcopy(sa, dupsa, sa->sa_len);
3011 	ifma->ifma_addr = dupsa;
3012 
3013 	ifma->ifma_ifp = ifp;
3014 	ifma->ifma_refcount = 1;
3015 	ifma->ifma_protospec = NULL;
3016 
3017 	if (llsa == NULL) {
3018 		ifma->ifma_lladdr = NULL;
3019 		return (ifma);
3020 	}
3021 
3022 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3023 	if (dupsa == NULL) {
3024 		free(ifma->ifma_addr, M_IFMADDR);
3025 		free(ifma, M_IFMADDR);
3026 		return (NULL);
3027 	}
3028 	bcopy(llsa, dupsa, llsa->sa_len);
3029 	ifma->ifma_lladdr = dupsa;
3030 
3031 	return (ifma);
3032 }
3033 
3034 /*
3035  * if_freemulti: free ifmultiaddr structure and possibly attached related
3036  * addresses.  The caller is responsible for implementing reference
3037  * counting, notifying the driver, handling routing messages, and releasing
3038  * any dependent link layer state.
3039  */
3040 static void
3041 if_freemulti(struct ifmultiaddr *ifma)
3042 {
3043 
3044 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3045 	    ifma->ifma_refcount));
3046 
3047 	if (ifma->ifma_lladdr != NULL)
3048 		free(ifma->ifma_lladdr, M_IFMADDR);
3049 	free(ifma->ifma_addr, M_IFMADDR);
3050 	free(ifma, M_IFMADDR);
3051 }
3052 
3053 /*
3054  * Register an additional multicast address with a network interface.
3055  *
3056  * - If the address is already present, bump the reference count on the
3057  *   address and return.
3058  * - If the address is not link-layer, look up a link layer address.
3059  * - Allocate address structures for one or both addresses, and attach to the
3060  *   multicast address list on the interface.  If automatically adding a link
3061  *   layer address, the protocol address will own a reference to the link
3062  *   layer address, to be freed when it is freed.
3063  * - Notify the network device driver of an addition to the multicast address
3064  *   list.
3065  *
3066  * 'sa' points to caller-owned memory with the desired multicast address.
3067  *
3068  * 'retifma' will be used to return a pointer to the resulting multicast
3069  * address reference, if desired.
3070  */
3071 int
3072 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3073     struct ifmultiaddr **retifma)
3074 {
3075 	struct ifmultiaddr *ifma, *ll_ifma;
3076 	struct sockaddr *llsa;
3077 	struct sockaddr_dl sdl;
3078 	int error;
3079 
3080 	/*
3081 	 * If the address is already present, return a new reference to it;
3082 	 * otherwise, allocate storage and set up a new address.
3083 	 */
3084 	IF_ADDR_WLOCK(ifp);
3085 	ifma = if_findmulti(ifp, sa);
3086 	if (ifma != NULL) {
3087 		ifma->ifma_refcount++;
3088 		if (retifma != NULL)
3089 			*retifma = ifma;
3090 		IF_ADDR_WUNLOCK(ifp);
3091 		return (0);
3092 	}
3093 
3094 	/*
3095 	 * The address isn't already present; resolve the protocol address
3096 	 * into a link layer address, and then look that up, bump its
3097 	 * refcount or allocate an ifma for that also.
3098 	 * Most link layer resolving functions returns address data which
3099 	 * fits inside default sockaddr_dl structure. However callback
3100 	 * can allocate another sockaddr structure, in that case we need to
3101 	 * free it later.
3102 	 */
3103 	llsa = NULL;
3104 	ll_ifma = NULL;
3105 	if (ifp->if_resolvemulti != NULL) {
3106 		/* Provide called function with buffer size information */
3107 		sdl.sdl_len = sizeof(sdl);
3108 		llsa = (struct sockaddr *)&sdl;
3109 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3110 		if (error)
3111 			goto unlock_out;
3112 	}
3113 
3114 	/*
3115 	 * Allocate the new address.  Don't hook it up yet, as we may also
3116 	 * need to allocate a link layer multicast address.
3117 	 */
3118 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3119 	if (ifma == NULL) {
3120 		error = ENOMEM;
3121 		goto free_llsa_out;
3122 	}
3123 
3124 	/*
3125 	 * If a link layer address is found, we'll need to see if it's
3126 	 * already present in the address list, or allocate is as well.
3127 	 * When this block finishes, the link layer address will be on the
3128 	 * list.
3129 	 */
3130 	if (llsa != NULL) {
3131 		ll_ifma = if_findmulti(ifp, llsa);
3132 		if (ll_ifma == NULL) {
3133 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3134 			if (ll_ifma == NULL) {
3135 				--ifma->ifma_refcount;
3136 				if_freemulti(ifma);
3137 				error = ENOMEM;
3138 				goto free_llsa_out;
3139 			}
3140 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3141 			    ifma_link);
3142 		} else
3143 			ll_ifma->ifma_refcount++;
3144 		ifma->ifma_llifma = ll_ifma;
3145 	}
3146 
3147 	/*
3148 	 * We now have a new multicast address, ifma, and possibly a new or
3149 	 * referenced link layer address.  Add the primary address to the
3150 	 * ifnet address list.
3151 	 */
3152 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3153 
3154 	if (retifma != NULL)
3155 		*retifma = ifma;
3156 
3157 	/*
3158 	 * Must generate the message while holding the lock so that 'ifma'
3159 	 * pointer is still valid.
3160 	 */
3161 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3162 	IF_ADDR_WUNLOCK(ifp);
3163 
3164 	/*
3165 	 * We are certain we have added something, so call down to the
3166 	 * interface to let them know about it.
3167 	 */
3168 	if (ifp->if_ioctl != NULL) {
3169 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3170 	}
3171 
3172 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3173 		link_free_sdl(llsa);
3174 
3175 	return (0);
3176 
3177 free_llsa_out:
3178 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3179 		link_free_sdl(llsa);
3180 
3181 unlock_out:
3182 	IF_ADDR_WUNLOCK(ifp);
3183 	return (error);
3184 }
3185 
3186 /*
3187  * Delete a multicast group membership by network-layer group address.
3188  *
3189  * Returns ENOENT if the entry could not be found. If ifp no longer
3190  * exists, results are undefined. This entry point should only be used
3191  * from subsystems which do appropriate locking to hold ifp for the
3192  * duration of the call.
3193  * Network-layer protocol domains must use if_delmulti_ifma().
3194  */
3195 int
3196 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3197 {
3198 	struct ifmultiaddr *ifma;
3199 	int lastref;
3200 #ifdef INVARIANTS
3201 	struct ifnet *oifp;
3202 
3203 	IFNET_RLOCK_NOSLEEP();
3204 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3205 		if (ifp == oifp)
3206 			break;
3207 	if (ifp != oifp)
3208 		ifp = NULL;
3209 	IFNET_RUNLOCK_NOSLEEP();
3210 
3211 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3212 #endif
3213 	if (ifp == NULL)
3214 		return (ENOENT);
3215 
3216 	IF_ADDR_WLOCK(ifp);
3217 	lastref = 0;
3218 	ifma = if_findmulti(ifp, sa);
3219 	if (ifma != NULL)
3220 		lastref = if_delmulti_locked(ifp, ifma, 0);
3221 	IF_ADDR_WUNLOCK(ifp);
3222 
3223 	if (ifma == NULL)
3224 		return (ENOENT);
3225 
3226 	if (lastref && ifp->if_ioctl != NULL) {
3227 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3228 	}
3229 
3230 	return (0);
3231 }
3232 
3233 /*
3234  * Delete all multicast group membership for an interface.
3235  * Should be used to quickly flush all multicast filters.
3236  */
3237 void
3238 if_delallmulti(struct ifnet *ifp)
3239 {
3240 	struct ifmultiaddr *ifma;
3241 	struct ifmultiaddr *next;
3242 
3243 	IF_ADDR_WLOCK(ifp);
3244 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3245 		if_delmulti_locked(ifp, ifma, 0);
3246 	IF_ADDR_WUNLOCK(ifp);
3247 }
3248 
3249 /*
3250  * Delete a multicast group membership by group membership pointer.
3251  * Network-layer protocol domains must use this routine.
3252  *
3253  * It is safe to call this routine if the ifp disappeared.
3254  */
3255 void
3256 if_delmulti_ifma(struct ifmultiaddr *ifma)
3257 {
3258 	struct ifnet *ifp;
3259 	int lastref;
3260 
3261 	ifp = ifma->ifma_ifp;
3262 #ifdef DIAGNOSTIC
3263 	if (ifp == NULL) {
3264 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3265 	} else {
3266 		struct ifnet *oifp;
3267 
3268 		IFNET_RLOCK_NOSLEEP();
3269 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3270 			if (ifp == oifp)
3271 				break;
3272 		if (ifp != oifp) {
3273 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3274 			ifp = NULL;
3275 		}
3276 		IFNET_RUNLOCK_NOSLEEP();
3277 	}
3278 #endif
3279 	/*
3280 	 * If and only if the ifnet instance exists: Acquire the address lock.
3281 	 */
3282 	if (ifp != NULL)
3283 		IF_ADDR_WLOCK(ifp);
3284 
3285 	lastref = if_delmulti_locked(ifp, ifma, 0);
3286 
3287 	if (ifp != NULL) {
3288 		/*
3289 		 * If and only if the ifnet instance exists:
3290 		 *  Release the address lock.
3291 		 *  If the group was left: update the hardware hash filter.
3292 		 */
3293 		IF_ADDR_WUNLOCK(ifp);
3294 		if (lastref && ifp->if_ioctl != NULL) {
3295 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3296 		}
3297 	}
3298 }
3299 
3300 /*
3301  * Perform deletion of network-layer and/or link-layer multicast address.
3302  *
3303  * Return 0 if the reference count was decremented.
3304  * Return 1 if the final reference was released, indicating that the
3305  * hardware hash filter should be reprogrammed.
3306  */
3307 static int
3308 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3309 {
3310 	struct ifmultiaddr *ll_ifma;
3311 
3312 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3313 		KASSERT(ifma->ifma_ifp == ifp,
3314 		    ("%s: inconsistent ifp %p", __func__, ifp));
3315 		IF_ADDR_WLOCK_ASSERT(ifp);
3316 	}
3317 
3318 	ifp = ifma->ifma_ifp;
3319 
3320 	/*
3321 	 * If the ifnet is detaching, null out references to ifnet,
3322 	 * so that upper protocol layers will notice, and not attempt
3323 	 * to obtain locks for an ifnet which no longer exists. The
3324 	 * routing socket announcement must happen before the ifnet
3325 	 * instance is detached from the system.
3326 	 */
3327 	if (detaching) {
3328 #ifdef DIAGNOSTIC
3329 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3330 #endif
3331 		/*
3332 		 * ifp may already be nulled out if we are being reentered
3333 		 * to delete the ll_ifma.
3334 		 */
3335 		if (ifp != NULL) {
3336 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3337 			ifma->ifma_ifp = NULL;
3338 		}
3339 	}
3340 
3341 	if (--ifma->ifma_refcount > 0)
3342 		return 0;
3343 
3344 	/*
3345 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3346 	 * have been associated with it. Release it first if so.
3347 	 */
3348 	ll_ifma = ifma->ifma_llifma;
3349 	if (ll_ifma != NULL) {
3350 		KASSERT(ifma->ifma_lladdr != NULL,
3351 		    ("%s: llifma w/o lladdr", __func__));
3352 		if (detaching)
3353 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3354 		if (--ll_ifma->ifma_refcount == 0) {
3355 			if (ifp != NULL) {
3356 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3357 				    ifma_link);
3358 			}
3359 			if_freemulti(ll_ifma);
3360 		}
3361 	}
3362 
3363 	if (ifp != NULL)
3364 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3365 
3366 	if_freemulti(ifma);
3367 
3368 	/*
3369 	 * The last reference to this instance of struct ifmultiaddr
3370 	 * was released; the hardware should be notified of this change.
3371 	 */
3372 	return 1;
3373 }
3374 
3375 /*
3376  * Set the link layer address on an interface.
3377  *
3378  * At this time we only support certain types of interfaces,
3379  * and we don't allow the length of the address to change.
3380  *
3381  * Set noinline to be dtrace-friendly
3382  */
3383 __noinline int
3384 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3385 {
3386 	struct sockaddr_dl *sdl;
3387 	struct ifaddr *ifa;
3388 	struct ifreq ifr;
3389 
3390 	IF_ADDR_RLOCK(ifp);
3391 	ifa = ifp->if_addr;
3392 	if (ifa == NULL) {
3393 		IF_ADDR_RUNLOCK(ifp);
3394 		return (EINVAL);
3395 	}
3396 	ifa_ref(ifa);
3397 	IF_ADDR_RUNLOCK(ifp);
3398 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3399 	if (sdl == NULL) {
3400 		ifa_free(ifa);
3401 		return (EINVAL);
3402 	}
3403 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3404 		ifa_free(ifa);
3405 		return (EINVAL);
3406 	}
3407 	switch (ifp->if_type) {
3408 	case IFT_ETHER:
3409 	case IFT_FDDI:
3410 	case IFT_XETHER:
3411 	case IFT_ISO88025:
3412 	case IFT_L2VLAN:
3413 	case IFT_BRIDGE:
3414 	case IFT_ARCNET:
3415 	case IFT_IEEE8023ADLAG:
3416 	case IFT_IEEE80211:
3417 		bcopy(lladdr, LLADDR(sdl), len);
3418 		ifa_free(ifa);
3419 		break;
3420 	default:
3421 		ifa_free(ifa);
3422 		return (ENODEV);
3423 	}
3424 
3425 	/*
3426 	 * If the interface is already up, we need
3427 	 * to re-init it in order to reprogram its
3428 	 * address filter.
3429 	 */
3430 	if ((ifp->if_flags & IFF_UP) != 0) {
3431 		if (ifp->if_ioctl) {
3432 			ifp->if_flags &= ~IFF_UP;
3433 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3434 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3435 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3436 			ifp->if_flags |= IFF_UP;
3437 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3438 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3439 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3440 		}
3441 	}
3442 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3443 	return (0);
3444 }
3445 
3446 /*
3447  * Compat function for handling basic encapsulation requests.
3448  * Not converted stacks (FDDI, IB, ..) supports traditional
3449  * output model: ARP (and other similar L2 protocols) are handled
3450  * inside output routine, arpresolve/nd6_resolve() returns MAC
3451  * address instead of full prepend.
3452  *
3453  * This function creates calculated header==MAC for IPv4/IPv6 and
3454  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3455  * address families.
3456  */
3457 static int
3458 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3459 {
3460 
3461 	if (req->rtype != IFENCAP_LL)
3462 		return (EOPNOTSUPP);
3463 
3464 	if (req->bufsize < req->lladdr_len)
3465 		return (ENOMEM);
3466 
3467 	switch (req->family) {
3468 	case AF_INET:
3469 	case AF_INET6:
3470 		break;
3471 	default:
3472 		return (EAFNOSUPPORT);
3473 	}
3474 
3475 	/* Copy lladdr to storage as is */
3476 	memmove(req->buf, req->lladdr, req->lladdr_len);
3477 	req->bufsize = req->lladdr_len;
3478 	req->lladdr_off = 0;
3479 
3480 	return (0);
3481 }
3482 
3483 /*
3484  * The name argument must be a pointer to storage which will last as
3485  * long as the interface does.  For physical devices, the result of
3486  * device_get_name(dev) is a good choice and for pseudo-devices a
3487  * static string works well.
3488  */
3489 void
3490 if_initname(struct ifnet *ifp, const char *name, int unit)
3491 {
3492 	ifp->if_dname = name;
3493 	ifp->if_dunit = unit;
3494 	if (unit != IF_DUNIT_NONE)
3495 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3496 	else
3497 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3498 }
3499 
3500 int
3501 if_printf(struct ifnet *ifp, const char * fmt, ...)
3502 {
3503 	va_list ap;
3504 	int retval;
3505 
3506 	retval = printf("%s: ", ifp->if_xname);
3507 	va_start(ap, fmt);
3508 	retval += vprintf(fmt, ap);
3509 	va_end(ap);
3510 	return (retval);
3511 }
3512 
3513 void
3514 if_start(struct ifnet *ifp)
3515 {
3516 
3517 	(*(ifp)->if_start)(ifp);
3518 }
3519 
3520 /*
3521  * Backwards compatibility interface for drivers
3522  * that have not implemented it
3523  */
3524 static int
3525 if_transmit(struct ifnet *ifp, struct mbuf *m)
3526 {
3527 	int error;
3528 
3529 	IFQ_HANDOFF(ifp, m, error);
3530 	return (error);
3531 }
3532 
3533 static void
3534 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3535 {
3536 
3537 	m_freem(m);
3538 }
3539 
3540 int
3541 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3542 {
3543 	int active = 0;
3544 
3545 	IF_LOCK(ifq);
3546 	if (_IF_QFULL(ifq)) {
3547 		IF_UNLOCK(ifq);
3548 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3549 		m_freem(m);
3550 		return (0);
3551 	}
3552 	if (ifp != NULL) {
3553 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3554 		if (m->m_flags & (M_BCAST|M_MCAST))
3555 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3556 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3557 	}
3558 	_IF_ENQUEUE(ifq, m);
3559 	IF_UNLOCK(ifq);
3560 	if (ifp != NULL && !active)
3561 		(*(ifp)->if_start)(ifp);
3562 	return (1);
3563 }
3564 
3565 void
3566 if_register_com_alloc(u_char type,
3567     if_com_alloc_t *a, if_com_free_t *f)
3568 {
3569 
3570 	KASSERT(if_com_alloc[type] == NULL,
3571 	    ("if_register_com_alloc: %d already registered", type));
3572 	KASSERT(if_com_free[type] == NULL,
3573 	    ("if_register_com_alloc: %d free already registered", type));
3574 
3575 	if_com_alloc[type] = a;
3576 	if_com_free[type] = f;
3577 }
3578 
3579 void
3580 if_deregister_com_alloc(u_char type)
3581 {
3582 
3583 	KASSERT(if_com_alloc[type] != NULL,
3584 	    ("if_deregister_com_alloc: %d not registered", type));
3585 	KASSERT(if_com_free[type] != NULL,
3586 	    ("if_deregister_com_alloc: %d free not registered", type));
3587 	if_com_alloc[type] = NULL;
3588 	if_com_free[type] = NULL;
3589 }
3590 
3591 /* API for driver access to network stack owned ifnet.*/
3592 uint64_t
3593 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3594 {
3595 	uint64_t oldbrate;
3596 
3597 	oldbrate = ifp->if_baudrate;
3598 	ifp->if_baudrate = baudrate;
3599 	return (oldbrate);
3600 }
3601 
3602 uint64_t
3603 if_getbaudrate(if_t ifp)
3604 {
3605 
3606 	return (((struct ifnet *)ifp)->if_baudrate);
3607 }
3608 
3609 int
3610 if_setcapabilities(if_t ifp, int capabilities)
3611 {
3612 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3613 	return (0);
3614 }
3615 
3616 int
3617 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3618 {
3619 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3620 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3621 
3622 	return (0);
3623 }
3624 
3625 int
3626 if_getcapabilities(if_t ifp)
3627 {
3628 	return ((struct ifnet *)ifp)->if_capabilities;
3629 }
3630 
3631 int
3632 if_setcapenable(if_t ifp, int capabilities)
3633 {
3634 	((struct ifnet *)ifp)->if_capenable = capabilities;
3635 	return (0);
3636 }
3637 
3638 int
3639 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3640 {
3641 	if(setcap)
3642 		((struct ifnet *)ifp)->if_capenable |= setcap;
3643 	if(clearcap)
3644 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3645 
3646 	return (0);
3647 }
3648 
3649 const char *
3650 if_getdname(if_t ifp)
3651 {
3652 	return ((struct ifnet *)ifp)->if_dname;
3653 }
3654 
3655 int
3656 if_togglecapenable(if_t ifp, int togglecap)
3657 {
3658 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3659 	return (0);
3660 }
3661 
3662 int
3663 if_getcapenable(if_t ifp)
3664 {
3665 	return ((struct ifnet *)ifp)->if_capenable;
3666 }
3667 
3668 /*
3669  * This is largely undesirable because it ties ifnet to a device, but does
3670  * provide flexiblity for an embedded product vendor. Should be used with
3671  * the understanding that it violates the interface boundaries, and should be
3672  * a last resort only.
3673  */
3674 int
3675 if_setdev(if_t ifp, void *dev)
3676 {
3677 	return (0);
3678 }
3679 
3680 int
3681 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3682 {
3683 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3684 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3685 
3686 	return (0);
3687 }
3688 
3689 int
3690 if_getdrvflags(if_t ifp)
3691 {
3692 	return ((struct ifnet *)ifp)->if_drv_flags;
3693 }
3694 
3695 int
3696 if_setdrvflags(if_t ifp, int flags)
3697 {
3698 	((struct ifnet *)ifp)->if_drv_flags = flags;
3699 	return (0);
3700 }
3701 
3702 
3703 int
3704 if_setflags(if_t ifp, int flags)
3705 {
3706 	((struct ifnet *)ifp)->if_flags = flags;
3707 	return (0);
3708 }
3709 
3710 int
3711 if_setflagbits(if_t ifp, int set, int clear)
3712 {
3713 	((struct ifnet *)ifp)->if_flags |= set;
3714 	((struct ifnet *)ifp)->if_flags &= ~clear;
3715 
3716 	return (0);
3717 }
3718 
3719 int
3720 if_getflags(if_t ifp)
3721 {
3722 	return ((struct ifnet *)ifp)->if_flags;
3723 }
3724 
3725 int
3726 if_clearhwassist(if_t ifp)
3727 {
3728 	((struct ifnet *)ifp)->if_hwassist = 0;
3729 	return (0);
3730 }
3731 
3732 int
3733 if_sethwassistbits(if_t ifp, int toset, int toclear)
3734 {
3735 	((struct ifnet *)ifp)->if_hwassist |= toset;
3736 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3737 
3738 	return (0);
3739 }
3740 
3741 int
3742 if_sethwassist(if_t ifp, int hwassist_bit)
3743 {
3744 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3745 	return (0);
3746 }
3747 
3748 int
3749 if_gethwassist(if_t ifp)
3750 {
3751 	return ((struct ifnet *)ifp)->if_hwassist;
3752 }
3753 
3754 int
3755 if_setmtu(if_t ifp, int mtu)
3756 {
3757 	((struct ifnet *)ifp)->if_mtu = mtu;
3758 	return (0);
3759 }
3760 
3761 int
3762 if_getmtu(if_t ifp)
3763 {
3764 	return ((struct ifnet *)ifp)->if_mtu;
3765 }
3766 
3767 int
3768 if_getmtu_family(if_t ifp, int family)
3769 {
3770 	struct domain *dp;
3771 
3772 	for (dp = domains; dp; dp = dp->dom_next) {
3773 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3774 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3775 	}
3776 
3777 	return (((struct ifnet *)ifp)->if_mtu);
3778 }
3779 
3780 int
3781 if_setsoftc(if_t ifp, void *softc)
3782 {
3783 	((struct ifnet *)ifp)->if_softc = softc;
3784 	return (0);
3785 }
3786 
3787 void *
3788 if_getsoftc(if_t ifp)
3789 {
3790 	return ((struct ifnet *)ifp)->if_softc;
3791 }
3792 
3793 void
3794 if_setrcvif(struct mbuf *m, if_t ifp)
3795 {
3796 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3797 }
3798 
3799 void
3800 if_setvtag(struct mbuf *m, uint16_t tag)
3801 {
3802 	m->m_pkthdr.ether_vtag = tag;
3803 }
3804 
3805 uint16_t
3806 if_getvtag(struct mbuf *m)
3807 {
3808 
3809 	return (m->m_pkthdr.ether_vtag);
3810 }
3811 
3812 int
3813 if_sendq_empty(if_t ifp)
3814 {
3815 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3816 }
3817 
3818 struct ifaddr *
3819 if_getifaddr(if_t ifp)
3820 {
3821 	return ((struct ifnet *)ifp)->if_addr;
3822 }
3823 
3824 int
3825 if_getamcount(if_t ifp)
3826 {
3827 	return ((struct ifnet *)ifp)->if_amcount;
3828 }
3829 
3830 
3831 int
3832 if_setsendqready(if_t ifp)
3833 {
3834 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3835 	return (0);
3836 }
3837 
3838 int
3839 if_setsendqlen(if_t ifp, int tx_desc_count)
3840 {
3841 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3842 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3843 
3844 	return (0);
3845 }
3846 
3847 int
3848 if_vlantrunkinuse(if_t ifp)
3849 {
3850 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3851 }
3852 
3853 int
3854 if_input(if_t ifp, struct mbuf* sendmp)
3855 {
3856 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3857 	return (0);
3858 
3859 }
3860 
3861 /* XXX */
3862 #ifndef ETH_ADDR_LEN
3863 #define ETH_ADDR_LEN 6
3864 #endif
3865 
3866 int
3867 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3868 {
3869 	struct ifmultiaddr *ifma;
3870 	uint8_t *lmta = (uint8_t *)mta;
3871 	int mcnt = 0;
3872 
3873 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3874 		if (ifma->ifma_addr->sa_family != AF_LINK)
3875 			continue;
3876 
3877 		if (mcnt == max)
3878 			break;
3879 
3880 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3881 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3882 		mcnt++;
3883 	}
3884 	*cnt = mcnt;
3885 
3886 	return (0);
3887 }
3888 
3889 int
3890 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3891 {
3892 	int error;
3893 
3894 	if_maddr_rlock(ifp);
3895 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3896 	if_maddr_runlock(ifp);
3897 	return (error);
3898 }
3899 
3900 int
3901 if_multiaddr_count(if_t ifp, int max)
3902 {
3903 	struct ifmultiaddr *ifma;
3904 	int count;
3905 
3906 	count = 0;
3907 	if_maddr_rlock(ifp);
3908 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3909 		if (ifma->ifma_addr->sa_family != AF_LINK)
3910 			continue;
3911 		count++;
3912 		if (count == max)
3913 			break;
3914 	}
3915 	if_maddr_runlock(ifp);
3916 	return (count);
3917 }
3918 
3919 int
3920 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
3921 {
3922 	struct ifmultiaddr *ifma;
3923 	int cnt = 0;
3924 
3925 	if_maddr_rlock(ifp);
3926 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
3927 		cnt += filter(arg, ifma, cnt);
3928 	if_maddr_runlock(ifp);
3929 	return (cnt);
3930 }
3931 
3932 struct mbuf *
3933 if_dequeue(if_t ifp)
3934 {
3935 	struct mbuf *m;
3936 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3937 
3938 	return (m);
3939 }
3940 
3941 int
3942 if_sendq_prepend(if_t ifp, struct mbuf *m)
3943 {
3944 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3945 	return (0);
3946 }
3947 
3948 int
3949 if_setifheaderlen(if_t ifp, int len)
3950 {
3951 	((struct ifnet *)ifp)->if_hdrlen = len;
3952 	return (0);
3953 }
3954 
3955 caddr_t
3956 if_getlladdr(if_t ifp)
3957 {
3958 	return (IF_LLADDR((struct ifnet *)ifp));
3959 }
3960 
3961 void *
3962 if_gethandle(u_char type)
3963 {
3964 	return (if_alloc(type));
3965 }
3966 
3967 void
3968 if_bpfmtap(if_t ifh, struct mbuf *m)
3969 {
3970 	struct ifnet *ifp = (struct ifnet *)ifh;
3971 
3972 	BPF_MTAP(ifp, m);
3973 }
3974 
3975 void
3976 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3977 {
3978 	struct ifnet *ifp = (struct ifnet *)ifh;
3979 
3980 	ETHER_BPF_MTAP(ifp, m);
3981 }
3982 
3983 void
3984 if_vlancap(if_t ifh)
3985 {
3986 	struct ifnet *ifp = (struct ifnet *)ifh;
3987 	VLAN_CAPABILITIES(ifp);
3988 }
3989 
3990 void
3991 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3992 {
3993 	((struct ifnet *)ifp)->if_init = init_fn;
3994 }
3995 
3996 void
3997 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3998 {
3999 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4000 }
4001 
4002 void
4003 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4004 {
4005 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4006 }
4007 
4008 void
4009 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4010 {
4011 	((struct ifnet *)ifp)->if_transmit = start_fn;
4012 }
4013 
4014 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4015 {
4016 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4017 
4018 }
4019 
4020 void
4021 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4022 {
4023 
4024 	ifp->if_get_counter = fn;
4025 }
4026 
4027 /* Revisit these - These are inline functions originally. */
4028 int
4029 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4030 {
4031 	return drbr_inuse(ifh, br);
4032 }
4033 
4034 struct mbuf*
4035 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4036 {
4037 	return drbr_dequeue(ifh, br);
4038 }
4039 
4040 int
4041 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4042 {
4043 	return drbr_needs_enqueue(ifh, br);
4044 }
4045 
4046 int
4047 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4048 {
4049 	return drbr_enqueue(ifh, br, m);
4050 
4051 }
4052