xref: /freebsd/sys/net/if.c (revision ff0ba87247820afbdfdc1b307c803f7923d0e4d3)
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 /* Interface description */
116 static unsigned int ifdescr_maxlen = 1024;
117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
118 	&ifdescr_maxlen, 0,
119 	"administrative maximum length for interface description");
120 
121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
122 
123 /* global sx for non-critical path ifdescr */
124 static struct sx ifdescr_sx;
125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
126 
127 void	(*bridge_linkstate_p)(struct ifnet *ifp);
128 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
129 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
130 /* These are external hooks for CARP. */
131 void	(*carp_linkstate_p)(struct ifnet *ifp);
132 void	(*carp_demote_adj_p)(int, char *);
133 int	(*carp_master_p)(struct ifaddr *);
134 #if defined(INET) || defined(INET6)
135 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
136 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
137     const struct sockaddr *sa);
138 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
139 int	(*carp_attach_p)(struct ifaddr *, int);
140 void	(*carp_detach_p)(struct ifaddr *);
141 #endif
142 #ifdef INET
143 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
144 #endif
145 #ifdef INET6
146 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
147 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
148     const struct in6_addr *taddr);
149 #endif
150 
151 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
152 
153 /*
154  * XXX: Style; these should be sorted alphabetically, and unprototyped
155  * static functions should be prototyped. Currently they are sorted by
156  * declaration order.
157  */
158 static void	if_attachdomain(void *);
159 static void	if_attachdomain1(struct ifnet *);
160 static int	ifconf(u_long, caddr_t);
161 static void	if_freemulti(struct ifmultiaddr *);
162 static void	if_grow(void);
163 static void	if_route(struct ifnet *, int flag, int fam);
164 static int	if_setflag(struct ifnet *, int, int, int *, int);
165 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
166 static void	if_unroute(struct ifnet *, int flag, int fam);
167 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
168 static int	if_rtdel(struct radix_node *, void *);
169 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
170 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
171 static void	do_link_state_change(void *, int);
172 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
173 static int	if_getgroupmembers(struct ifgroupreq *);
174 static void	if_delgroups(struct ifnet *);
175 static void	if_attach_internal(struct ifnet *, int);
176 static void	if_detach_internal(struct ifnet *, int);
177 
178 #ifdef INET6
179 /*
180  * XXX: declare here to avoid to include many inet6 related files..
181  * should be more generalized?
182  */
183 extern void	nd6_setmtu(struct ifnet *);
184 #endif
185 
186 VNET_DEFINE(int, if_index);
187 int	ifqmaxlen = IFQ_MAXLEN;
188 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
189 VNET_DEFINE(struct ifgrouphead, ifg_head);
190 
191 static VNET_DEFINE(int, if_indexlim) = 8;
192 
193 /* Table of ifnet by index. */
194 VNET_DEFINE(struct ifnet **, ifindex_table);
195 
196 #define	V_if_indexlim		VNET(if_indexlim)
197 #define	V_ifindex_table		VNET(ifindex_table)
198 
199 /*
200  * The global network interface list (V_ifnet) and related state (such as
201  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
202  * an rwlock.  Either may be acquired shared to stablize the list, but both
203  * must be acquired writable to modify the list.  This model allows us to
204  * both stablize the interface list during interrupt thread processing, but
205  * also to stablize it over long-running ioctls, without introducing priority
206  * inversions and deadlocks.
207  */
208 struct rwlock ifnet_rwlock;
209 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
210 struct sx ifnet_sxlock;
211 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
212 
213 /*
214  * The allocation of network interfaces is a rather non-atomic affair; we
215  * need to select an index before we are ready to expose the interface for
216  * use, so will use this pointer value to indicate reservation.
217  */
218 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
219 
220 static	if_com_alloc_t *if_com_alloc[256];
221 static	if_com_free_t *if_com_free[256];
222 
223 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
224 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
225 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
226 
227 struct ifnet *
228 ifnet_byindex_locked(u_short idx)
229 {
230 
231 	if (idx > V_if_index)
232 		return (NULL);
233 	if (V_ifindex_table[idx] == IFNET_HOLD)
234 		return (NULL);
235 	return (V_ifindex_table[idx]);
236 }
237 
238 struct ifnet *
239 ifnet_byindex(u_short idx)
240 {
241 	struct ifnet *ifp;
242 
243 	IFNET_RLOCK_NOSLEEP();
244 	ifp = ifnet_byindex_locked(idx);
245 	IFNET_RUNLOCK_NOSLEEP();
246 	return (ifp);
247 }
248 
249 struct ifnet *
250 ifnet_byindex_ref(u_short idx)
251 {
252 	struct ifnet *ifp;
253 
254 	IFNET_RLOCK_NOSLEEP();
255 	ifp = ifnet_byindex_locked(idx);
256 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
257 		IFNET_RUNLOCK_NOSLEEP();
258 		return (NULL);
259 	}
260 	if_ref(ifp);
261 	IFNET_RUNLOCK_NOSLEEP();
262 	return (ifp);
263 }
264 
265 /*
266  * Allocate an ifindex array entry; return 0 on success or an error on
267  * failure.
268  */
269 static u_short
270 ifindex_alloc(void)
271 {
272 	u_short idx;
273 
274 	IFNET_WLOCK_ASSERT();
275 retry:
276 	/*
277 	 * Try to find an empty slot below V_if_index.  If we fail, take the
278 	 * next slot.
279 	 */
280 	for (idx = 1; idx <= V_if_index; idx++) {
281 		if (V_ifindex_table[idx] == NULL)
282 			break;
283 	}
284 
285 	/* Catch if_index overflow. */
286 	if (idx >= V_if_indexlim) {
287 		if_grow();
288 		goto retry;
289 	}
290 	if (idx > V_if_index)
291 		V_if_index = idx;
292 	return (idx);
293 }
294 
295 static void
296 ifindex_free_locked(u_short idx)
297 {
298 
299 	IFNET_WLOCK_ASSERT();
300 
301 	V_ifindex_table[idx] = NULL;
302 	while (V_if_index > 0 &&
303 	    V_ifindex_table[V_if_index] == NULL)
304 		V_if_index--;
305 }
306 
307 static void
308 ifindex_free(u_short idx)
309 {
310 
311 	IFNET_WLOCK();
312 	ifindex_free_locked(idx);
313 	IFNET_WUNLOCK();
314 }
315 
316 static void
317 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
318 {
319 
320 	IFNET_WLOCK_ASSERT();
321 
322 	V_ifindex_table[idx] = ifp;
323 }
324 
325 static void
326 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
327 {
328 
329 	IFNET_WLOCK();
330 	ifnet_setbyindex_locked(idx, ifp);
331 	IFNET_WUNLOCK();
332 }
333 
334 struct ifaddr *
335 ifaddr_byindex(u_short idx)
336 {
337 	struct ifaddr *ifa;
338 
339 	IFNET_RLOCK_NOSLEEP();
340 	ifa = ifnet_byindex_locked(idx)->if_addr;
341 	if (ifa != NULL)
342 		ifa_ref(ifa);
343 	IFNET_RUNLOCK_NOSLEEP();
344 	return (ifa);
345 }
346 
347 /*
348  * Network interface utility routines.
349  *
350  * Routines with ifa_ifwith* names take sockaddr *'s as
351  * parameters.
352  */
353 
354 static void
355 vnet_if_init(const void *unused __unused)
356 {
357 
358 	TAILQ_INIT(&V_ifnet);
359 	TAILQ_INIT(&V_ifg_head);
360 	IFNET_WLOCK();
361 	if_grow();				/* create initial table */
362 	IFNET_WUNLOCK();
363 	vnet_if_clone_init();
364 }
365 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
366     NULL);
367 
368 #ifdef VIMAGE
369 static void
370 vnet_if_uninit(const void *unused __unused)
371 {
372 
373 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
374 	    "not empty", __func__, __LINE__, &V_ifnet));
375 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
376 	    "not empty", __func__, __LINE__, &V_ifg_head));
377 
378 	free((caddr_t)V_ifindex_table, M_IFNET);
379 }
380 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
381     vnet_if_uninit, NULL);
382 #endif
383 
384 static void
385 if_grow(void)
386 {
387 	int oldlim;
388 	u_int n;
389 	struct ifnet **e;
390 
391 	IFNET_WLOCK_ASSERT();
392 	oldlim = V_if_indexlim;
393 	IFNET_WUNLOCK();
394 	n = (oldlim << 1) * sizeof(*e);
395 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
396 	IFNET_WLOCK();
397 	if (V_if_indexlim != oldlim) {
398 		free(e, M_IFNET);
399 		return;
400 	}
401 	if (V_ifindex_table != NULL) {
402 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
403 		free((caddr_t)V_ifindex_table, M_IFNET);
404 	}
405 	V_if_indexlim <<= 1;
406 	V_ifindex_table = e;
407 }
408 
409 /*
410  * Allocate a struct ifnet and an index for an interface.  A layer 2
411  * common structure will also be allocated if an allocation routine is
412  * registered for the passed type.
413  */
414 struct ifnet *
415 if_alloc(u_char type)
416 {
417 	struct ifnet *ifp;
418 	u_short idx;
419 
420 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
421 	IFNET_WLOCK();
422 	idx = ifindex_alloc();
423 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
424 	IFNET_WUNLOCK();
425 	ifp->if_index = idx;
426 	ifp->if_type = type;
427 	ifp->if_alloctype = type;
428 	if (if_com_alloc[type] != NULL) {
429 		ifp->if_l2com = if_com_alloc[type](type, ifp);
430 		if (ifp->if_l2com == NULL) {
431 			free(ifp, M_IFNET);
432 			ifindex_free(idx);
433 			return (NULL);
434 		}
435 	}
436 
437 	IF_ADDR_LOCK_INIT(ifp);
438 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
439 	ifp->if_afdata_initialized = 0;
440 	IF_AFDATA_LOCK_INIT(ifp);
441 	TAILQ_INIT(&ifp->if_addrhead);
442 	TAILQ_INIT(&ifp->if_multiaddrs);
443 	TAILQ_INIT(&ifp->if_groups);
444 #ifdef MAC
445 	mac_ifnet_init(ifp);
446 #endif
447 	ifq_init(&ifp->if_snd, ifp);
448 
449 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
450 	for (int i = 0; i < IFCOUNTERS; i++)
451 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
452 	ifp->if_get_counter = if_get_counter_default;
453 	ifnet_setbyindex(ifp->if_index, ifp);
454 	return (ifp);
455 }
456 
457 /*
458  * Do the actual work of freeing a struct ifnet, and layer 2 common
459  * structure.  This call is made when the last reference to an
460  * interface is released.
461  */
462 static void
463 if_free_internal(struct ifnet *ifp)
464 {
465 
466 	KASSERT((ifp->if_flags & IFF_DYING),
467 	    ("if_free_internal: interface not dying"));
468 
469 	if (if_com_free[ifp->if_alloctype] != NULL)
470 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
471 		    ifp->if_alloctype);
472 
473 #ifdef MAC
474 	mac_ifnet_destroy(ifp);
475 #endif /* MAC */
476 	if (ifp->if_description != NULL)
477 		free(ifp->if_description, M_IFDESCR);
478 	IF_AFDATA_DESTROY(ifp);
479 	IF_ADDR_LOCK_DESTROY(ifp);
480 	ifq_delete(&ifp->if_snd);
481 
482 	for (int i = 0; i < IFCOUNTERS; i++)
483 		counter_u64_free(ifp->if_counters[i]);
484 
485 	free(ifp, M_IFNET);
486 }
487 
488 /*
489  * Deregister an interface and free the associated storage.
490  */
491 void
492 if_free(struct ifnet *ifp)
493 {
494 
495 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
496 
497 	CURVNET_SET_QUIET(ifp->if_vnet);
498 	IFNET_WLOCK();
499 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
500 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
501 
502 	ifindex_free_locked(ifp->if_index);
503 	IFNET_WUNLOCK();
504 
505 	if (refcount_release(&ifp->if_refcount))
506 		if_free_internal(ifp);
507 	CURVNET_RESTORE();
508 }
509 
510 /*
511  * Interfaces to keep an ifnet type-stable despite the possibility of the
512  * driver calling if_free().  If there are additional references, we defer
513  * freeing the underlying data structure.
514  */
515 void
516 if_ref(struct ifnet *ifp)
517 {
518 
519 	/* We don't assert the ifnet list lock here, but arguably should. */
520 	refcount_acquire(&ifp->if_refcount);
521 }
522 
523 void
524 if_rele(struct ifnet *ifp)
525 {
526 
527 	if (!refcount_release(&ifp->if_refcount))
528 		return;
529 	if_free_internal(ifp);
530 }
531 
532 void
533 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
534 {
535 
536 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
537 
538 	if (ifq->ifq_maxlen == 0)
539 		ifq->ifq_maxlen = ifqmaxlen;
540 
541 	ifq->altq_type = 0;
542 	ifq->altq_disc = NULL;
543 	ifq->altq_flags &= ALTQF_CANTCHANGE;
544 	ifq->altq_tbr  = NULL;
545 	ifq->altq_ifp  = ifp;
546 }
547 
548 void
549 ifq_delete(struct ifaltq *ifq)
550 {
551 	mtx_destroy(&ifq->ifq_mtx);
552 }
553 
554 /*
555  * Perform generic interface initalization tasks and attach the interface
556  * to the list of "active" interfaces.  If vmove flag is set on entry
557  * to if_attach_internal(), perform only a limited subset of initialization
558  * tasks, given that we are moving from one vnet to another an ifnet which
559  * has already been fully initialized.
560  *
561  * XXX:
562  *  - The decision to return void and thus require this function to
563  *    succeed is questionable.
564  *  - We should probably do more sanity checking.  For instance we don't
565  *    do anything to insure if_xname is unique or non-empty.
566  */
567 void
568 if_attach(struct ifnet *ifp)
569 {
570 
571 	if_attach_internal(ifp, 0);
572 }
573 
574 /*
575  * Compute the least common TSO limit.
576  */
577 void
578 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
579 {
580 	/*
581 	 * 1) If there is no limit currently, take the limit from
582 	 * the network adapter.
583 	 *
584 	 * 2) If the network adapter has a limit below the current
585 	 * limit, apply it.
586 	 */
587 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
588 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
589 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
590 	}
591 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
592 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
593 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
594 	}
595 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
596 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
597 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
598 	}
599 }
600 
601 /*
602  * Update TSO limit of a network adapter.
603  *
604  * Returns zero if no change. Else non-zero.
605  */
606 int
607 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
608 {
609 	int retval = 0;
610 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
611 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
612 		retval++;
613 	}
614 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
615 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
616 		retval++;
617 	}
618 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
619 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
620 		retval++;
621 	}
622 	return (retval);
623 }
624 
625 static void
626 if_attach_internal(struct ifnet *ifp, int vmove)
627 {
628 	unsigned socksize, ifasize;
629 	int namelen, masklen;
630 	struct sockaddr_dl *sdl;
631 	struct ifaddr *ifa;
632 
633 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
634 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
635 		    ifp->if_xname);
636 
637 #ifdef VIMAGE
638 	ifp->if_vnet = curvnet;
639 	if (ifp->if_home_vnet == NULL)
640 		ifp->if_home_vnet = curvnet;
641 #endif
642 
643 	if_addgroup(ifp, IFG_ALL);
644 
645 	getmicrotime(&ifp->if_lastchange);
646 	ifp->if_epoch = time_uptime;
647 
648 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
649 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
650 	    ("transmit and qflush must both either be set or both be NULL"));
651 	if (ifp->if_transmit == NULL) {
652 		ifp->if_transmit = if_transmit;
653 		ifp->if_qflush = if_qflush;
654 	}
655 
656 	if (!vmove) {
657 #ifdef MAC
658 		mac_ifnet_create(ifp);
659 #endif
660 
661 		/*
662 		 * Create a Link Level name for this device.
663 		 */
664 		namelen = strlen(ifp->if_xname);
665 		/*
666 		 * Always save enough space for any possiable name so we
667 		 * can do a rename in place later.
668 		 */
669 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
670 		socksize = masklen + ifp->if_addrlen;
671 		if (socksize < sizeof(*sdl))
672 			socksize = sizeof(*sdl);
673 		socksize = roundup2(socksize, sizeof(long));
674 		ifasize = sizeof(*ifa) + 2 * socksize;
675 		ifa = ifa_alloc(ifasize, M_WAITOK);
676 		sdl = (struct sockaddr_dl *)(ifa + 1);
677 		sdl->sdl_len = socksize;
678 		sdl->sdl_family = AF_LINK;
679 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
680 		sdl->sdl_nlen = namelen;
681 		sdl->sdl_index = ifp->if_index;
682 		sdl->sdl_type = ifp->if_type;
683 		ifp->if_addr = ifa;
684 		ifa->ifa_ifp = ifp;
685 		ifa->ifa_rtrequest = link_rtrequest;
686 		ifa->ifa_addr = (struct sockaddr *)sdl;
687 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
688 		ifa->ifa_netmask = (struct sockaddr *)sdl;
689 		sdl->sdl_len = masklen;
690 		while (namelen != 0)
691 			sdl->sdl_data[--namelen] = 0xff;
692 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
693 		/* Reliably crash if used uninitialized. */
694 		ifp->if_broadcastaddr = NULL;
695 
696 #if defined(INET) || defined(INET6)
697 		/* Use defaults for TSO, if nothing is set */
698 		if (ifp->if_hw_tsomax == 0 &&
699 		    ifp->if_hw_tsomaxsegcount == 0 &&
700 		    ifp->if_hw_tsomaxsegsize == 0) {
701 			/*
702 			 * The TSO defaults needs to be such that an
703 			 * NFS mbuf list of 35 mbufs totalling just
704 			 * below 64K works and that a chain of mbufs
705 			 * can be defragged into at most 32 segments:
706 			 */
707 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
708 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
709 			ifp->if_hw_tsomaxsegcount = 35;
710 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
711 
712 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
713 			if (ifp->if_capabilities & IFCAP_TSO) {
714 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
715 				    ifp->if_hw_tsomax,
716 				    ifp->if_hw_tsomaxsegcount,
717 				    ifp->if_hw_tsomaxsegsize);
718 			}
719 		}
720 #endif
721 	}
722 #ifdef VIMAGE
723 	else {
724 		/*
725 		 * Update the interface index in the link layer address
726 		 * of the interface.
727 		 */
728 		for (ifa = ifp->if_addr; ifa != NULL;
729 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
730 			if (ifa->ifa_addr->sa_family == AF_LINK) {
731 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
732 				sdl->sdl_index = ifp->if_index;
733 			}
734 		}
735 	}
736 #endif
737 
738 	IFNET_WLOCK();
739 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
740 #ifdef VIMAGE
741 	curvnet->vnet_ifcnt++;
742 #endif
743 	IFNET_WUNLOCK();
744 
745 	if (domain_init_status >= 2)
746 		if_attachdomain1(ifp);
747 
748 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
749 	if (IS_DEFAULT_VNET(curvnet))
750 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
751 
752 	/* Announce the interface. */
753 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
754 }
755 
756 static void
757 if_attachdomain(void *dummy)
758 {
759 	struct ifnet *ifp;
760 
761 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
762 		if_attachdomain1(ifp);
763 }
764 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
765     if_attachdomain, NULL);
766 
767 static void
768 if_attachdomain1(struct ifnet *ifp)
769 {
770 	struct domain *dp;
771 
772 	/*
773 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
774 	 * cannot lock ifp->if_afdata initialization, entirely.
775 	 */
776 	if (IF_AFDATA_TRYLOCK(ifp) == 0)
777 		return;
778 	if (ifp->if_afdata_initialized >= domain_init_status) {
779 		IF_AFDATA_UNLOCK(ifp);
780 		log(LOG_WARNING, "%s called more than once on %s\n",
781 		    __func__, ifp->if_xname);
782 		return;
783 	}
784 	ifp->if_afdata_initialized = domain_init_status;
785 	IF_AFDATA_UNLOCK(ifp);
786 
787 	/* address family dependent data region */
788 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
789 	for (dp = domains; dp; dp = dp->dom_next) {
790 		if (dp->dom_ifattach)
791 			ifp->if_afdata[dp->dom_family] =
792 			    (*dp->dom_ifattach)(ifp);
793 	}
794 }
795 
796 /*
797  * Remove any unicast or broadcast network addresses from an interface.
798  */
799 void
800 if_purgeaddrs(struct ifnet *ifp)
801 {
802 	struct ifaddr *ifa, *next;
803 
804 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
805 		if (ifa->ifa_addr->sa_family == AF_LINK)
806 			continue;
807 #ifdef INET
808 		/* XXX: Ugly!! ad hoc just for INET */
809 		if (ifa->ifa_addr->sa_family == AF_INET) {
810 			struct ifaliasreq ifr;
811 
812 			bzero(&ifr, sizeof(ifr));
813 			ifr.ifra_addr = *ifa->ifa_addr;
814 			if (ifa->ifa_dstaddr)
815 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
816 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
817 			    NULL) == 0)
818 				continue;
819 		}
820 #endif /* INET */
821 #ifdef INET6
822 		if (ifa->ifa_addr->sa_family == AF_INET6) {
823 			in6_purgeaddr(ifa);
824 			/* ifp_addrhead is already updated */
825 			continue;
826 		}
827 #endif /* INET6 */
828 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
829 		ifa_free(ifa);
830 	}
831 }
832 
833 /*
834  * Remove any multicast network addresses from an interface when an ifnet
835  * is going away.
836  */
837 static void
838 if_purgemaddrs(struct ifnet *ifp)
839 {
840 	struct ifmultiaddr *ifma;
841 	struct ifmultiaddr *next;
842 
843 	IF_ADDR_WLOCK(ifp);
844 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
845 		if_delmulti_locked(ifp, ifma, 1);
846 	IF_ADDR_WUNLOCK(ifp);
847 }
848 
849 /*
850  * Detach an interface, removing it from the list of "active" interfaces.
851  * If vmove flag is set on entry to if_detach_internal(), perform only a
852  * limited subset of cleanup tasks, given that we are moving an ifnet from
853  * one vnet to another, where it must be fully operational.
854  *
855  * XXXRW: There are some significant questions about event ordering, and
856  * how to prevent things from starting to use the interface during detach.
857  */
858 void
859 if_detach(struct ifnet *ifp)
860 {
861 
862 	CURVNET_SET_QUIET(ifp->if_vnet);
863 	if_detach_internal(ifp, 0);
864 	CURVNET_RESTORE();
865 }
866 
867 static void
868 if_detach_internal(struct ifnet *ifp, int vmove)
869 {
870 	struct ifaddr *ifa;
871 	struct radix_node_head	*rnh;
872 	int i, j;
873 	struct domain *dp;
874  	struct ifnet *iter;
875  	int found = 0;
876 
877 	IFNET_WLOCK();
878 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
879 		if (iter == ifp) {
880 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
881 			found = 1;
882 			break;
883 		}
884 #ifdef VIMAGE
885 	if (found)
886 		curvnet->vnet_ifcnt--;
887 #endif
888 	IFNET_WUNLOCK();
889 	if (!found) {
890 		if (vmove)
891 			panic("%s: ifp=%p not on the ifnet tailq %p",
892 			    __func__, ifp, &V_ifnet);
893 		else
894 			return; /* XXX this should panic as well? */
895 	}
896 
897 	/*
898 	 * Remove/wait for pending events.
899 	 */
900 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
901 
902 	/*
903 	 * Remove routes and flush queues.
904 	 */
905 	if_down(ifp);
906 #ifdef ALTQ
907 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
908 		altq_disable(&ifp->if_snd);
909 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
910 		altq_detach(&ifp->if_snd);
911 #endif
912 
913 	if_purgeaddrs(ifp);
914 
915 #ifdef INET
916 	in_ifdetach(ifp);
917 #endif
918 
919 #ifdef INET6
920 	/*
921 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
922 	 * before removing routing entries below, since IPv6 interface direct
923 	 * routes are expected to be removed by the IPv6-specific kernel API.
924 	 * Otherwise, the kernel will detect some inconsistency and bark it.
925 	 */
926 	in6_ifdetach(ifp);
927 #endif
928 	if_purgemaddrs(ifp);
929 
930 	/* Announce that the interface is gone. */
931 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
932 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
933 	if (IS_DEFAULT_VNET(curvnet))
934 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
935 
936 	if (!vmove) {
937 		/*
938 		 * Prevent further calls into the device driver via ifnet.
939 		 */
940 		if_dead(ifp);
941 
942 		/*
943 		 * Remove link ifaddr pointer and maybe decrement if_index.
944 		 * Clean up all addresses.
945 		 */
946 		ifp->if_addr = NULL;
947 
948 		/* We can now free link ifaddr. */
949 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
950 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
951 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
952 			ifa_free(ifa);
953 		}
954 	}
955 
956 	/*
957 	 * Delete all remaining routes using this interface
958 	 * Unfortuneatly the only way to do this is to slog through
959 	 * the entire routing table looking for routes which point
960 	 * to this interface...oh well...
961 	 */
962 	for (i = 1; i <= AF_MAX; i++) {
963 		for (j = 0; j < rt_numfibs; j++) {
964 			rnh = rt_tables_get_rnh(j, i);
965 			if (rnh == NULL)
966 				continue;
967 			RADIX_NODE_HEAD_LOCK(rnh);
968 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
969 			RADIX_NODE_HEAD_UNLOCK(rnh);
970 		}
971 	}
972 
973 	if_delgroups(ifp);
974 
975 	/*
976 	 * We cannot hold the lock over dom_ifdetach calls as they might
977 	 * sleep, for example trying to drain a callout, thus open up the
978 	 * theoretical race with re-attaching.
979 	 */
980 	IF_AFDATA_LOCK(ifp);
981 	i = ifp->if_afdata_initialized;
982 	ifp->if_afdata_initialized = 0;
983 	IF_AFDATA_UNLOCK(ifp);
984 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
985 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
986 			(*dp->dom_ifdetach)(ifp,
987 			    ifp->if_afdata[dp->dom_family]);
988 	}
989 }
990 
991 #ifdef VIMAGE
992 /*
993  * if_vmove() performs a limited version of if_detach() in current
994  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
995  * An attempt is made to shrink if_index in current vnet, find an
996  * unused if_index in target vnet and calls if_grow() if necessary,
997  * and finally find an unused if_xname for the target vnet.
998  */
999 void
1000 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1001 {
1002 
1003 	/*
1004 	 * Detach from current vnet, but preserve LLADDR info, do not
1005 	 * mark as dead etc. so that the ifnet can be reattached later.
1006 	 */
1007 	if_detach_internal(ifp, 1);
1008 
1009 	/*
1010 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1011 	 * the if_index for that vnet if possible.
1012 	 *
1013 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1014 	 * or we'd lock on one vnet and unlock on another.
1015 	 */
1016 	IFNET_WLOCK();
1017 	ifindex_free_locked(ifp->if_index);
1018 	IFNET_WUNLOCK();
1019 
1020 	/*
1021 	 * Perform interface-specific reassignment tasks, if provided by
1022 	 * the driver.
1023 	 */
1024 	if (ifp->if_reassign != NULL)
1025 		ifp->if_reassign(ifp, new_vnet, NULL);
1026 
1027 	/*
1028 	 * Switch to the context of the target vnet.
1029 	 */
1030 	CURVNET_SET_QUIET(new_vnet);
1031 
1032 	IFNET_WLOCK();
1033 	ifp->if_index = ifindex_alloc();
1034 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1035 	IFNET_WUNLOCK();
1036 
1037 	if_attach_internal(ifp, 1);
1038 
1039 	CURVNET_RESTORE();
1040 }
1041 
1042 /*
1043  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1044  */
1045 static int
1046 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1047 {
1048 	struct prison *pr;
1049 	struct ifnet *difp;
1050 
1051 	/* Try to find the prison within our visibility. */
1052 	sx_slock(&allprison_lock);
1053 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1054 	sx_sunlock(&allprison_lock);
1055 	if (pr == NULL)
1056 		return (ENXIO);
1057 	prison_hold_locked(pr);
1058 	mtx_unlock(&pr->pr_mtx);
1059 
1060 	/* Do not try to move the iface from and to the same prison. */
1061 	if (pr->pr_vnet == ifp->if_vnet) {
1062 		prison_free(pr);
1063 		return (EEXIST);
1064 	}
1065 
1066 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1067 	/* XXX Lock interfaces to avoid races. */
1068 	CURVNET_SET_QUIET(pr->pr_vnet);
1069 	difp = ifunit(ifname);
1070 	CURVNET_RESTORE();
1071 	if (difp != NULL) {
1072 		prison_free(pr);
1073 		return (EEXIST);
1074 	}
1075 
1076 	/* Move the interface into the child jail/vnet. */
1077 	if_vmove(ifp, pr->pr_vnet);
1078 
1079 	/* Report the new if_xname back to the userland. */
1080 	sprintf(ifname, "%s", ifp->if_xname);
1081 
1082 	prison_free(pr);
1083 	return (0);
1084 }
1085 
1086 static int
1087 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1088 {
1089 	struct prison *pr;
1090 	struct vnet *vnet_dst;
1091 	struct ifnet *ifp;
1092 
1093 	/* Try to find the prison within our visibility. */
1094 	sx_slock(&allprison_lock);
1095 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1096 	sx_sunlock(&allprison_lock);
1097 	if (pr == NULL)
1098 		return (ENXIO);
1099 	prison_hold_locked(pr);
1100 	mtx_unlock(&pr->pr_mtx);
1101 
1102 	/* Make sure the named iface exists in the source prison/vnet. */
1103 	CURVNET_SET(pr->pr_vnet);
1104 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1105 	if (ifp == NULL) {
1106 		CURVNET_RESTORE();
1107 		prison_free(pr);
1108 		return (ENXIO);
1109 	}
1110 
1111 	/* Do not try to move the iface from and to the same prison. */
1112 	vnet_dst = TD_TO_VNET(td);
1113 	if (vnet_dst == ifp->if_vnet) {
1114 		CURVNET_RESTORE();
1115 		prison_free(pr);
1116 		return (EEXIST);
1117 	}
1118 
1119 	/* Get interface back from child jail/vnet. */
1120 	if_vmove(ifp, vnet_dst);
1121 	CURVNET_RESTORE();
1122 
1123 	/* Report the new if_xname back to the userland. */
1124 	sprintf(ifname, "%s", ifp->if_xname);
1125 
1126 	prison_free(pr);
1127 	return (0);
1128 }
1129 #endif /* VIMAGE */
1130 
1131 /*
1132  * Add a group to an interface
1133  */
1134 int
1135 if_addgroup(struct ifnet *ifp, const char *groupname)
1136 {
1137 	struct ifg_list		*ifgl;
1138 	struct ifg_group	*ifg = NULL;
1139 	struct ifg_member	*ifgm;
1140 	int 			 new = 0;
1141 
1142 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1143 	    groupname[strlen(groupname) - 1] <= '9')
1144 		return (EINVAL);
1145 
1146 	IFNET_WLOCK();
1147 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1148 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1149 			IFNET_WUNLOCK();
1150 			return (EEXIST);
1151 		}
1152 
1153 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1154 	    M_NOWAIT)) == NULL) {
1155 	    	IFNET_WUNLOCK();
1156 		return (ENOMEM);
1157 	}
1158 
1159 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1160 	    M_TEMP, M_NOWAIT)) == NULL) {
1161 		free(ifgl, M_TEMP);
1162 		IFNET_WUNLOCK();
1163 		return (ENOMEM);
1164 	}
1165 
1166 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1167 		if (!strcmp(ifg->ifg_group, groupname))
1168 			break;
1169 
1170 	if (ifg == NULL) {
1171 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1172 		    M_TEMP, M_NOWAIT)) == NULL) {
1173 			free(ifgl, M_TEMP);
1174 			free(ifgm, M_TEMP);
1175 			IFNET_WUNLOCK();
1176 			return (ENOMEM);
1177 		}
1178 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1179 		ifg->ifg_refcnt = 0;
1180 		TAILQ_INIT(&ifg->ifg_members);
1181 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1182 		new = 1;
1183 	}
1184 
1185 	ifg->ifg_refcnt++;
1186 	ifgl->ifgl_group = ifg;
1187 	ifgm->ifgm_ifp = ifp;
1188 
1189 	IF_ADDR_WLOCK(ifp);
1190 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1191 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1192 	IF_ADDR_WUNLOCK(ifp);
1193 
1194 	IFNET_WUNLOCK();
1195 
1196 	if (new)
1197 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1198 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1199 
1200 	return (0);
1201 }
1202 
1203 /*
1204  * Remove a group from an interface
1205  */
1206 int
1207 if_delgroup(struct ifnet *ifp, const char *groupname)
1208 {
1209 	struct ifg_list		*ifgl;
1210 	struct ifg_member	*ifgm;
1211 
1212 	IFNET_WLOCK();
1213 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1214 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1215 			break;
1216 	if (ifgl == NULL) {
1217 		IFNET_WUNLOCK();
1218 		return (ENOENT);
1219 	}
1220 
1221 	IF_ADDR_WLOCK(ifp);
1222 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1223 	IF_ADDR_WUNLOCK(ifp);
1224 
1225 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1226 		if (ifgm->ifgm_ifp == ifp)
1227 			break;
1228 
1229 	if (ifgm != NULL) {
1230 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1231 		free(ifgm, M_TEMP);
1232 	}
1233 
1234 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1235 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1236 		IFNET_WUNLOCK();
1237 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1238 		free(ifgl->ifgl_group, M_TEMP);
1239 	} else
1240 		IFNET_WUNLOCK();
1241 
1242 	free(ifgl, M_TEMP);
1243 
1244 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1245 
1246 	return (0);
1247 }
1248 
1249 /*
1250  * Remove an interface from all groups
1251  */
1252 static void
1253 if_delgroups(struct ifnet *ifp)
1254 {
1255 	struct ifg_list		*ifgl;
1256 	struct ifg_member	*ifgm;
1257 	char groupname[IFNAMSIZ];
1258 
1259 	IFNET_WLOCK();
1260 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1261 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1262 
1263 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1264 
1265 		IF_ADDR_WLOCK(ifp);
1266 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1267 		IF_ADDR_WUNLOCK(ifp);
1268 
1269 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1270 			if (ifgm->ifgm_ifp == ifp)
1271 				break;
1272 
1273 		if (ifgm != NULL) {
1274 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1275 			    ifgm_next);
1276 			free(ifgm, M_TEMP);
1277 		}
1278 
1279 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1280 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1281 			IFNET_WUNLOCK();
1282 			EVENTHANDLER_INVOKE(group_detach_event,
1283 			    ifgl->ifgl_group);
1284 			free(ifgl->ifgl_group, M_TEMP);
1285 		} else
1286 			IFNET_WUNLOCK();
1287 
1288 		free(ifgl, M_TEMP);
1289 
1290 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1291 
1292 		IFNET_WLOCK();
1293 	}
1294 	IFNET_WUNLOCK();
1295 }
1296 
1297 /*
1298  * Stores all groups from an interface in memory pointed
1299  * to by data
1300  */
1301 static int
1302 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1303 {
1304 	int			 len, error;
1305 	struct ifg_list		*ifgl;
1306 	struct ifg_req		 ifgrq, *ifgp;
1307 	struct ifgroupreq	*ifgr = data;
1308 
1309 	if (ifgr->ifgr_len == 0) {
1310 		IF_ADDR_RLOCK(ifp);
1311 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1312 			ifgr->ifgr_len += sizeof(struct ifg_req);
1313 		IF_ADDR_RUNLOCK(ifp);
1314 		return (0);
1315 	}
1316 
1317 	len = ifgr->ifgr_len;
1318 	ifgp = ifgr->ifgr_groups;
1319 	/* XXX: wire */
1320 	IF_ADDR_RLOCK(ifp);
1321 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1322 		if (len < sizeof(ifgrq)) {
1323 			IF_ADDR_RUNLOCK(ifp);
1324 			return (EINVAL);
1325 		}
1326 		bzero(&ifgrq, sizeof ifgrq);
1327 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1328 		    sizeof(ifgrq.ifgrq_group));
1329 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1330 		    	IF_ADDR_RUNLOCK(ifp);
1331 			return (error);
1332 		}
1333 		len -= sizeof(ifgrq);
1334 		ifgp++;
1335 	}
1336 	IF_ADDR_RUNLOCK(ifp);
1337 
1338 	return (0);
1339 }
1340 
1341 /*
1342  * Stores all members of a group in memory pointed to by data
1343  */
1344 static int
1345 if_getgroupmembers(struct ifgroupreq *data)
1346 {
1347 	struct ifgroupreq	*ifgr = data;
1348 	struct ifg_group	*ifg;
1349 	struct ifg_member	*ifgm;
1350 	struct ifg_req		 ifgrq, *ifgp;
1351 	int			 len, error;
1352 
1353 	IFNET_RLOCK();
1354 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1355 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1356 			break;
1357 	if (ifg == NULL) {
1358 		IFNET_RUNLOCK();
1359 		return (ENOENT);
1360 	}
1361 
1362 	if (ifgr->ifgr_len == 0) {
1363 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1364 			ifgr->ifgr_len += sizeof(ifgrq);
1365 		IFNET_RUNLOCK();
1366 		return (0);
1367 	}
1368 
1369 	len = ifgr->ifgr_len;
1370 	ifgp = ifgr->ifgr_groups;
1371 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1372 		if (len < sizeof(ifgrq)) {
1373 			IFNET_RUNLOCK();
1374 			return (EINVAL);
1375 		}
1376 		bzero(&ifgrq, sizeof ifgrq);
1377 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1378 		    sizeof(ifgrq.ifgrq_member));
1379 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1380 			IFNET_RUNLOCK();
1381 			return (error);
1382 		}
1383 		len -= sizeof(ifgrq);
1384 		ifgp++;
1385 	}
1386 	IFNET_RUNLOCK();
1387 
1388 	return (0);
1389 }
1390 
1391 /*
1392  * Delete Routes for a Network Interface
1393  *
1394  * Called for each routing entry via the rnh->rnh_walktree() call above
1395  * to delete all route entries referencing a detaching network interface.
1396  *
1397  * Arguments:
1398  *	rn	pointer to node in the routing table
1399  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1400  *
1401  * Returns:
1402  *	0	successful
1403  *	errno	failed - reason indicated
1404  *
1405  */
1406 static int
1407 if_rtdel(struct radix_node *rn, void *arg)
1408 {
1409 	struct rtentry	*rt = (struct rtentry *)rn;
1410 	struct ifnet	*ifp = arg;
1411 	int		err;
1412 
1413 	if (rt->rt_ifp == ifp) {
1414 
1415 		/*
1416 		 * Protect (sorta) against walktree recursion problems
1417 		 * with cloned routes
1418 		 */
1419 		if ((rt->rt_flags & RTF_UP) == 0)
1420 			return (0);
1421 
1422 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1423 				rt_mask(rt),
1424 				rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1425 				(struct rtentry **) NULL, rt->rt_fibnum);
1426 		if (err) {
1427 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1428 		}
1429 	}
1430 
1431 	return (0);
1432 }
1433 
1434 /*
1435  * Return counter values from counter(9)s stored in ifnet.
1436  */
1437 uint64_t
1438 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1439 {
1440 
1441 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1442 
1443 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1444 }
1445 
1446 /*
1447  * Increase an ifnet counter. Usually used for counters shared
1448  * between the stack and a driver, but function supports them all.
1449  */
1450 void
1451 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1452 {
1453 
1454 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1455 
1456 	counter_u64_add(ifp->if_counters[cnt], inc);
1457 }
1458 
1459 /*
1460  * Copy data from ifnet to userland API structure if_data.
1461  */
1462 void
1463 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1464 {
1465 
1466 	ifd->ifi_type = ifp->if_type;
1467 	ifd->ifi_physical = 0;
1468 	ifd->ifi_addrlen = ifp->if_addrlen;
1469 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1470 	ifd->ifi_link_state = ifp->if_link_state;
1471 	ifd->ifi_vhid = 0;
1472 	ifd->ifi_datalen = sizeof(struct if_data);
1473 	ifd->ifi_mtu = ifp->if_mtu;
1474 	ifd->ifi_metric = ifp->if_metric;
1475 	ifd->ifi_baudrate = ifp->if_baudrate;
1476 	ifd->ifi_hwassist = ifp->if_hwassist;
1477 	ifd->ifi_epoch = ifp->if_epoch;
1478 	ifd->ifi_lastchange = ifp->if_lastchange;
1479 
1480 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1481 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1482 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1483 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1484 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1485 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1486 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1487 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1488 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1489 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1490 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1491 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1492 }
1493 
1494 /*
1495  * Wrapper functions for struct ifnet address list locking macros.  These are
1496  * used by kernel modules to avoid encoding programming interface or binary
1497  * interface assumptions that may be violated when kernel-internal locking
1498  * approaches change.
1499  */
1500 void
1501 if_addr_rlock(struct ifnet *ifp)
1502 {
1503 
1504 	IF_ADDR_RLOCK(ifp);
1505 }
1506 
1507 void
1508 if_addr_runlock(struct ifnet *ifp)
1509 {
1510 
1511 	IF_ADDR_RUNLOCK(ifp);
1512 }
1513 
1514 void
1515 if_maddr_rlock(if_t ifp)
1516 {
1517 
1518 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1519 }
1520 
1521 void
1522 if_maddr_runlock(if_t ifp)
1523 {
1524 
1525 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1526 }
1527 
1528 /*
1529  * Initialization, destruction and refcounting functions for ifaddrs.
1530  */
1531 struct ifaddr *
1532 ifa_alloc(size_t size, int flags)
1533 {
1534 	struct ifaddr *ifa;
1535 
1536 	KASSERT(size >= sizeof(struct ifaddr),
1537 	    ("%s: invalid size %zu", __func__, size));
1538 
1539 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1540 	if (ifa == NULL)
1541 		return (NULL);
1542 
1543 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1544 		goto fail;
1545 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1546 		goto fail;
1547 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1548 		goto fail;
1549 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1550 		goto fail;
1551 
1552 	refcount_init(&ifa->ifa_refcnt, 1);
1553 
1554 	return (ifa);
1555 
1556 fail:
1557 	/* free(NULL) is okay */
1558 	counter_u64_free(ifa->ifa_opackets);
1559 	counter_u64_free(ifa->ifa_ipackets);
1560 	counter_u64_free(ifa->ifa_obytes);
1561 	counter_u64_free(ifa->ifa_ibytes);
1562 	free(ifa, M_IFADDR);
1563 
1564 	return (NULL);
1565 }
1566 
1567 void
1568 ifa_ref(struct ifaddr *ifa)
1569 {
1570 
1571 	refcount_acquire(&ifa->ifa_refcnt);
1572 }
1573 
1574 void
1575 ifa_free(struct ifaddr *ifa)
1576 {
1577 
1578 	if (refcount_release(&ifa->ifa_refcnt)) {
1579 		counter_u64_free(ifa->ifa_opackets);
1580 		counter_u64_free(ifa->ifa_ipackets);
1581 		counter_u64_free(ifa->ifa_obytes);
1582 		counter_u64_free(ifa->ifa_ibytes);
1583 		free(ifa, M_IFADDR);
1584 	}
1585 }
1586 
1587 int
1588 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1589 {
1590 	int error = 0;
1591 	struct rtentry *rt = NULL;
1592 	struct rt_addrinfo info;
1593 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1594 
1595 	bzero(&info, sizeof(info));
1596 	info.rti_ifp = V_loif;
1597 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1598 	info.rti_info[RTAX_DST] = ia;
1599 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1600 	error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1601 
1602 	if (error == 0 && rt != NULL) {
1603 		RT_LOCK(rt);
1604 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1605 			ifa->ifa_ifp->if_type;
1606 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1607 			ifa->ifa_ifp->if_index;
1608 		RT_REMREF(rt);
1609 		RT_UNLOCK(rt);
1610 	} else if (error != 0)
1611 		log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1612 
1613 	return (error);
1614 }
1615 
1616 int
1617 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1618 {
1619 	int error = 0;
1620 	struct rt_addrinfo info;
1621 	struct sockaddr_dl null_sdl;
1622 
1623 	bzero(&null_sdl, sizeof(null_sdl));
1624 	null_sdl.sdl_len = sizeof(null_sdl);
1625 	null_sdl.sdl_family = AF_LINK;
1626 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1627 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1628 	bzero(&info, sizeof(info));
1629 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1630 	info.rti_info[RTAX_DST] = ia;
1631 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1632 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1633 
1634 	if (error != 0)
1635 		log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1636 
1637 	return (error);
1638 }
1639 
1640 int
1641 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib)
1642 {
1643 	struct rtentry *rt;
1644 
1645 	rt = rtalloc1_fib(sa, 0, 0, fib);
1646 	if (rt == NULL) {
1647 		log(LOG_DEBUG, "%s: fail", __func__);
1648 		return (EHOSTUNREACH);
1649 	}
1650 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1651 	    ifa->ifa_ifp->if_type;
1652 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1653 	    ifa->ifa_ifp->if_index;
1654 	RTFREE_LOCKED(rt);
1655 
1656 	return (0);
1657 }
1658 
1659 /*
1660  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1661  * structs used to represent other address families, it is necessary
1662  * to perform a different comparison.
1663  */
1664 
1665 #define	sa_dl_equal(a1, a2)	\
1666 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1667 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1668 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1669 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1670 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1671 
1672 /*
1673  * Locate an interface based on a complete address.
1674  */
1675 /*ARGSUSED*/
1676 static struct ifaddr *
1677 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1678 {
1679 	struct ifnet *ifp;
1680 	struct ifaddr *ifa;
1681 
1682 	IFNET_RLOCK_NOSLEEP();
1683 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1684 		IF_ADDR_RLOCK(ifp);
1685 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1686 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1687 				continue;
1688 			if (sa_equal(addr, ifa->ifa_addr)) {
1689 				if (getref)
1690 					ifa_ref(ifa);
1691 				IF_ADDR_RUNLOCK(ifp);
1692 				goto done;
1693 			}
1694 			/* IP6 doesn't have broadcast */
1695 			if ((ifp->if_flags & IFF_BROADCAST) &&
1696 			    ifa->ifa_broadaddr &&
1697 			    ifa->ifa_broadaddr->sa_len != 0 &&
1698 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1699 				if (getref)
1700 					ifa_ref(ifa);
1701 				IF_ADDR_RUNLOCK(ifp);
1702 				goto done;
1703 			}
1704 		}
1705 		IF_ADDR_RUNLOCK(ifp);
1706 	}
1707 	ifa = NULL;
1708 done:
1709 	IFNET_RUNLOCK_NOSLEEP();
1710 	return (ifa);
1711 }
1712 
1713 struct ifaddr *
1714 ifa_ifwithaddr(struct sockaddr *addr)
1715 {
1716 
1717 	return (ifa_ifwithaddr_internal(addr, 1));
1718 }
1719 
1720 int
1721 ifa_ifwithaddr_check(struct sockaddr *addr)
1722 {
1723 
1724 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1725 }
1726 
1727 /*
1728  * Locate an interface based on the broadcast address.
1729  */
1730 /* ARGSUSED */
1731 struct ifaddr *
1732 ifa_ifwithbroadaddr(struct sockaddr *addr, int fibnum)
1733 {
1734 	struct ifnet *ifp;
1735 	struct ifaddr *ifa;
1736 
1737 	IFNET_RLOCK_NOSLEEP();
1738 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1739 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1740 			continue;
1741 		IF_ADDR_RLOCK(ifp);
1742 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1743 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1744 				continue;
1745 			if ((ifp->if_flags & IFF_BROADCAST) &&
1746 			    ifa->ifa_broadaddr &&
1747 			    ifa->ifa_broadaddr->sa_len != 0 &&
1748 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1749 				ifa_ref(ifa);
1750 				IF_ADDR_RUNLOCK(ifp);
1751 				goto done;
1752 			}
1753 		}
1754 		IF_ADDR_RUNLOCK(ifp);
1755 	}
1756 	ifa = NULL;
1757 done:
1758 	IFNET_RUNLOCK_NOSLEEP();
1759 	return (ifa);
1760 }
1761 
1762 /*
1763  * Locate the point to point interface with a given destination address.
1764  */
1765 /*ARGSUSED*/
1766 struct ifaddr *
1767 ifa_ifwithdstaddr(struct sockaddr *addr, int fibnum)
1768 {
1769 	struct ifnet *ifp;
1770 	struct ifaddr *ifa;
1771 
1772 	IFNET_RLOCK_NOSLEEP();
1773 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1774 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1775 			continue;
1776 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1777 			continue;
1778 		IF_ADDR_RLOCK(ifp);
1779 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1780 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1781 				continue;
1782 			if (ifa->ifa_dstaddr != NULL &&
1783 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1784 				ifa_ref(ifa);
1785 				IF_ADDR_RUNLOCK(ifp);
1786 				goto done;
1787 			}
1788 		}
1789 		IF_ADDR_RUNLOCK(ifp);
1790 	}
1791 	ifa = NULL;
1792 done:
1793 	IFNET_RUNLOCK_NOSLEEP();
1794 	return (ifa);
1795 }
1796 
1797 /*
1798  * Find an interface on a specific network.  If many, choice
1799  * is most specific found.
1800  */
1801 struct ifaddr *
1802 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp, int fibnum)
1803 {
1804 	struct ifnet *ifp;
1805 	struct ifaddr *ifa;
1806 	struct ifaddr *ifa_maybe = NULL;
1807 	u_int af = addr->sa_family;
1808 	char *addr_data = addr->sa_data, *cplim;
1809 
1810 	/*
1811 	 * AF_LINK addresses can be looked up directly by their index number,
1812 	 * so do that if we can.
1813 	 */
1814 	if (af == AF_LINK) {
1815 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1816 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1817 		return (ifaddr_byindex(sdl->sdl_index));
1818 	}
1819 
1820 	/*
1821 	 * Scan though each interface, looking for ones that have addresses
1822 	 * in this address family and the requested fib.  Maintain a reference
1823 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1824 	 * kept it stable when we move onto the next interface.
1825 	 */
1826 	IFNET_RLOCK_NOSLEEP();
1827 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1828 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1829 			continue;
1830 		IF_ADDR_RLOCK(ifp);
1831 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1832 			char *cp, *cp2, *cp3;
1833 
1834 			if (ifa->ifa_addr->sa_family != af)
1835 next:				continue;
1836 			if (af == AF_INET &&
1837 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1838 				/*
1839 				 * This is a bit broken as it doesn't
1840 				 * take into account that the remote end may
1841 				 * be a single node in the network we are
1842 				 * looking for.
1843 				 * The trouble is that we don't know the
1844 				 * netmask for the remote end.
1845 				 */
1846 				if (ifa->ifa_dstaddr != NULL &&
1847 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1848 					ifa_ref(ifa);
1849 					IF_ADDR_RUNLOCK(ifp);
1850 					goto done;
1851 				}
1852 			} else {
1853 				/*
1854 				 * Scan all the bits in the ifa's address.
1855 				 * If a bit dissagrees with what we are
1856 				 * looking for, mask it with the netmask
1857 				 * to see if it really matters.
1858 				 * (A byte at a time)
1859 				 */
1860 				if (ifa->ifa_netmask == 0)
1861 					continue;
1862 				cp = addr_data;
1863 				cp2 = ifa->ifa_addr->sa_data;
1864 				cp3 = ifa->ifa_netmask->sa_data;
1865 				cplim = ifa->ifa_netmask->sa_len
1866 					+ (char *)ifa->ifa_netmask;
1867 				while (cp3 < cplim)
1868 					if ((*cp++ ^ *cp2++) & *cp3++)
1869 						goto next; /* next address! */
1870 				/*
1871 				 * If the netmask of what we just found
1872 				 * is more specific than what we had before
1873 				 * (if we had one), or if the virtual status
1874 				 * of new prefix is better than of the old one,
1875 				 * then remember the new one before continuing
1876 				 * to search for an even better one.
1877 				 */
1878 				if (ifa_maybe == NULL ||
1879 				    ifa_preferred(ifa_maybe, ifa) ||
1880 				    rn_refines((caddr_t)ifa->ifa_netmask,
1881 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1882 					if (ifa_maybe != NULL)
1883 						ifa_free(ifa_maybe);
1884 					ifa_maybe = ifa;
1885 					ifa_ref(ifa_maybe);
1886 				}
1887 			}
1888 		}
1889 		IF_ADDR_RUNLOCK(ifp);
1890 	}
1891 	ifa = ifa_maybe;
1892 	ifa_maybe = NULL;
1893 done:
1894 	IFNET_RUNLOCK_NOSLEEP();
1895 	if (ifa_maybe != NULL)
1896 		ifa_free(ifa_maybe);
1897 	return (ifa);
1898 }
1899 
1900 /*
1901  * Find an interface address specific to an interface best matching
1902  * a given address.
1903  */
1904 struct ifaddr *
1905 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1906 {
1907 	struct ifaddr *ifa;
1908 	char *cp, *cp2, *cp3;
1909 	char *cplim;
1910 	struct ifaddr *ifa_maybe = NULL;
1911 	u_int af = addr->sa_family;
1912 
1913 	if (af >= AF_MAX)
1914 		return (NULL);
1915 	IF_ADDR_RLOCK(ifp);
1916 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1917 		if (ifa->ifa_addr->sa_family != af)
1918 			continue;
1919 		if (ifa_maybe == NULL)
1920 			ifa_maybe = ifa;
1921 		if (ifa->ifa_netmask == 0) {
1922 			if (sa_equal(addr, ifa->ifa_addr) ||
1923 			    (ifa->ifa_dstaddr &&
1924 			    sa_equal(addr, ifa->ifa_dstaddr)))
1925 				goto done;
1926 			continue;
1927 		}
1928 		if (ifp->if_flags & IFF_POINTOPOINT) {
1929 			if (sa_equal(addr, ifa->ifa_dstaddr))
1930 				goto done;
1931 		} else {
1932 			cp = addr->sa_data;
1933 			cp2 = ifa->ifa_addr->sa_data;
1934 			cp3 = ifa->ifa_netmask->sa_data;
1935 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1936 			for (; cp3 < cplim; cp3++)
1937 				if ((*cp++ ^ *cp2++) & *cp3)
1938 					break;
1939 			if (cp3 == cplim)
1940 				goto done;
1941 		}
1942 	}
1943 	ifa = ifa_maybe;
1944 done:
1945 	if (ifa != NULL)
1946 		ifa_ref(ifa);
1947 	IF_ADDR_RUNLOCK(ifp);
1948 	return (ifa);
1949 }
1950 
1951 /*
1952  * See whether new ifa is better than current one:
1953  * 1) A non-virtual one is preferred over virtual.
1954  * 2) A virtual in master state preferred over any other state.
1955  *
1956  * Used in several address selecting functions.
1957  */
1958 int
1959 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1960 {
1961 
1962 	return (cur->ifa_carp && (!next->ifa_carp ||
1963 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1964 }
1965 
1966 #include <net/if_llatbl.h>
1967 
1968 /*
1969  * Default action when installing a route with a Link Level gateway.
1970  * Lookup an appropriate real ifa to point to.
1971  * This should be moved to /sys/net/link.c eventually.
1972  */
1973 static void
1974 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1975 {
1976 	struct ifaddr *ifa, *oifa;
1977 	struct sockaddr *dst;
1978 	struct ifnet *ifp;
1979 
1980 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1981 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1982 		return;
1983 	ifa = ifaof_ifpforaddr(dst, ifp);
1984 	if (ifa) {
1985 		oifa = rt->rt_ifa;
1986 		rt->rt_ifa = ifa;
1987 		ifa_free(oifa);
1988 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1989 			ifa->ifa_rtrequest(cmd, rt, info);
1990 	}
1991 }
1992 
1993 struct sockaddr_dl *
1994 link_alloc_sdl(size_t size, int flags)
1995 {
1996 
1997 	return (malloc(size, M_TEMP, flags));
1998 }
1999 
2000 void
2001 link_free_sdl(struct sockaddr *sa)
2002 {
2003 	free(sa, M_TEMP);
2004 }
2005 
2006 /*
2007  * Fills in given sdl with interface basic info.
2008  * Returns pointer to filled sdl.
2009  */
2010 struct sockaddr_dl *
2011 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2012 {
2013 	struct sockaddr_dl *sdl;
2014 
2015 	sdl = (struct sockaddr_dl *)paddr;
2016 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2017 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2018 	sdl->sdl_family = AF_LINK;
2019 	sdl->sdl_index = ifp->if_index;
2020 	sdl->sdl_type = iftype;
2021 
2022 	return (sdl);
2023 }
2024 
2025 /*
2026  * Mark an interface down and notify protocols of
2027  * the transition.
2028  */
2029 static void
2030 if_unroute(struct ifnet *ifp, int flag, int fam)
2031 {
2032 	struct ifaddr *ifa;
2033 
2034 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2035 
2036 	ifp->if_flags &= ~flag;
2037 	getmicrotime(&ifp->if_lastchange);
2038 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2039 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2040 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2041 	ifp->if_qflush(ifp);
2042 
2043 	if (ifp->if_carp)
2044 		(*carp_linkstate_p)(ifp);
2045 	rt_ifmsg(ifp);
2046 }
2047 
2048 /*
2049  * Mark an interface up and notify protocols of
2050  * the transition.
2051  */
2052 static void
2053 if_route(struct ifnet *ifp, int flag, int fam)
2054 {
2055 	struct ifaddr *ifa;
2056 
2057 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2058 
2059 	ifp->if_flags |= flag;
2060 	getmicrotime(&ifp->if_lastchange);
2061 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2062 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2063 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2064 	if (ifp->if_carp)
2065 		(*carp_linkstate_p)(ifp);
2066 	rt_ifmsg(ifp);
2067 #ifdef INET6
2068 	in6_if_up(ifp);
2069 #endif
2070 }
2071 
2072 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2073 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2074 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2075 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2076 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2077 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2078 void	*(*vlan_cookie_p)(struct ifnet *);
2079 
2080 /*
2081  * Handle a change in the interface link state. To avoid LORs
2082  * between driver lock and upper layer locks, as well as possible
2083  * recursions, we post event to taskqueue, and all job
2084  * is done in static do_link_state_change().
2085  */
2086 void
2087 if_link_state_change(struct ifnet *ifp, int link_state)
2088 {
2089 	/* Return if state hasn't changed. */
2090 	if (ifp->if_link_state == link_state)
2091 		return;
2092 
2093 	ifp->if_link_state = link_state;
2094 
2095 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2096 }
2097 
2098 static void
2099 do_link_state_change(void *arg, int pending)
2100 {
2101 	struct ifnet *ifp = (struct ifnet *)arg;
2102 	int link_state = ifp->if_link_state;
2103 	CURVNET_SET(ifp->if_vnet);
2104 
2105 	/* Notify that the link state has changed. */
2106 	rt_ifmsg(ifp);
2107 	if (ifp->if_vlantrunk != NULL)
2108 		(*vlan_link_state_p)(ifp);
2109 
2110 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2111 	    ifp->if_l2com != NULL)
2112 		(*ng_ether_link_state_p)(ifp, link_state);
2113 	if (ifp->if_carp)
2114 		(*carp_linkstate_p)(ifp);
2115 	if (ifp->if_bridge)
2116 		(*bridge_linkstate_p)(ifp);
2117 	if (ifp->if_lagg)
2118 		(*lagg_linkstate_p)(ifp, link_state);
2119 
2120 	if (IS_DEFAULT_VNET(curvnet))
2121 		devctl_notify("IFNET", ifp->if_xname,
2122 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2123 		    NULL);
2124 	if (pending > 1)
2125 		if_printf(ifp, "%d link states coalesced\n", pending);
2126 	if (log_link_state_change)
2127 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2128 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2129 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2130 	CURVNET_RESTORE();
2131 }
2132 
2133 /*
2134  * Mark an interface down and notify protocols of
2135  * the transition.
2136  */
2137 void
2138 if_down(struct ifnet *ifp)
2139 {
2140 
2141 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2142 }
2143 
2144 /*
2145  * Mark an interface up and notify protocols of
2146  * the transition.
2147  */
2148 void
2149 if_up(struct ifnet *ifp)
2150 {
2151 
2152 	if_route(ifp, IFF_UP, AF_UNSPEC);
2153 }
2154 
2155 /*
2156  * Flush an interface queue.
2157  */
2158 void
2159 if_qflush(struct ifnet *ifp)
2160 {
2161 	struct mbuf *m, *n;
2162 	struct ifaltq *ifq;
2163 
2164 	ifq = &ifp->if_snd;
2165 	IFQ_LOCK(ifq);
2166 #ifdef ALTQ
2167 	if (ALTQ_IS_ENABLED(ifq))
2168 		ALTQ_PURGE(ifq);
2169 #endif
2170 	n = ifq->ifq_head;
2171 	while ((m = n) != 0) {
2172 		n = m->m_nextpkt;
2173 		m_freem(m);
2174 	}
2175 	ifq->ifq_head = 0;
2176 	ifq->ifq_tail = 0;
2177 	ifq->ifq_len = 0;
2178 	IFQ_UNLOCK(ifq);
2179 }
2180 
2181 /*
2182  * Map interface name to interface structure pointer, with or without
2183  * returning a reference.
2184  */
2185 struct ifnet *
2186 ifunit_ref(const char *name)
2187 {
2188 	struct ifnet *ifp;
2189 
2190 	IFNET_RLOCK_NOSLEEP();
2191 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2192 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2193 		    !(ifp->if_flags & IFF_DYING))
2194 			break;
2195 	}
2196 	if (ifp != NULL)
2197 		if_ref(ifp);
2198 	IFNET_RUNLOCK_NOSLEEP();
2199 	return (ifp);
2200 }
2201 
2202 struct ifnet *
2203 ifunit(const char *name)
2204 {
2205 	struct ifnet *ifp;
2206 
2207 	IFNET_RLOCK_NOSLEEP();
2208 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2209 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2210 			break;
2211 	}
2212 	IFNET_RUNLOCK_NOSLEEP();
2213 	return (ifp);
2214 }
2215 
2216 /*
2217  * Hardware specific interface ioctls.
2218  */
2219 static int
2220 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2221 {
2222 	struct ifreq *ifr;
2223 	int error = 0;
2224 	int new_flags, temp_flags;
2225 	size_t namelen, onamelen;
2226 	size_t descrlen;
2227 	char *descrbuf, *odescrbuf;
2228 	char new_name[IFNAMSIZ];
2229 	struct ifaddr *ifa;
2230 	struct sockaddr_dl *sdl;
2231 
2232 	ifr = (struct ifreq *)data;
2233 	switch (cmd) {
2234 	case SIOCGIFINDEX:
2235 		ifr->ifr_index = ifp->if_index;
2236 		break;
2237 
2238 	case SIOCGIFFLAGS:
2239 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2240 		ifr->ifr_flags = temp_flags & 0xffff;
2241 		ifr->ifr_flagshigh = temp_flags >> 16;
2242 		break;
2243 
2244 	case SIOCGIFCAP:
2245 		ifr->ifr_reqcap = ifp->if_capabilities;
2246 		ifr->ifr_curcap = ifp->if_capenable;
2247 		break;
2248 
2249 #ifdef MAC
2250 	case SIOCGIFMAC:
2251 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2252 		break;
2253 #endif
2254 
2255 	case SIOCGIFMETRIC:
2256 		ifr->ifr_metric = ifp->if_metric;
2257 		break;
2258 
2259 	case SIOCGIFMTU:
2260 		ifr->ifr_mtu = ifp->if_mtu;
2261 		break;
2262 
2263 	case SIOCGIFPHYS:
2264 		/* XXXGL: did this ever worked? */
2265 		ifr->ifr_phys = 0;
2266 		break;
2267 
2268 	case SIOCGIFDESCR:
2269 		error = 0;
2270 		sx_slock(&ifdescr_sx);
2271 		if (ifp->if_description == NULL)
2272 			error = ENOMSG;
2273 		else {
2274 			/* space for terminating nul */
2275 			descrlen = strlen(ifp->if_description) + 1;
2276 			if (ifr->ifr_buffer.length < descrlen)
2277 				ifr->ifr_buffer.buffer = NULL;
2278 			else
2279 				error = copyout(ifp->if_description,
2280 				    ifr->ifr_buffer.buffer, descrlen);
2281 			ifr->ifr_buffer.length = descrlen;
2282 		}
2283 		sx_sunlock(&ifdescr_sx);
2284 		break;
2285 
2286 	case SIOCSIFDESCR:
2287 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2288 		if (error)
2289 			return (error);
2290 
2291 		/*
2292 		 * Copy only (length-1) bytes to make sure that
2293 		 * if_description is always nul terminated.  The
2294 		 * length parameter is supposed to count the
2295 		 * terminating nul in.
2296 		 */
2297 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2298 			return (ENAMETOOLONG);
2299 		else if (ifr->ifr_buffer.length == 0)
2300 			descrbuf = NULL;
2301 		else {
2302 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2303 			    M_WAITOK | M_ZERO);
2304 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2305 			    ifr->ifr_buffer.length - 1);
2306 			if (error) {
2307 				free(descrbuf, M_IFDESCR);
2308 				break;
2309 			}
2310 		}
2311 
2312 		sx_xlock(&ifdescr_sx);
2313 		odescrbuf = ifp->if_description;
2314 		ifp->if_description = descrbuf;
2315 		sx_xunlock(&ifdescr_sx);
2316 
2317 		getmicrotime(&ifp->if_lastchange);
2318 		free(odescrbuf, M_IFDESCR);
2319 		break;
2320 
2321 	case SIOCGIFFIB:
2322 		ifr->ifr_fib = ifp->if_fib;
2323 		break;
2324 
2325 	case SIOCSIFFIB:
2326 		error = priv_check(td, PRIV_NET_SETIFFIB);
2327 		if (error)
2328 			return (error);
2329 		if (ifr->ifr_fib >= rt_numfibs)
2330 			return (EINVAL);
2331 
2332 		ifp->if_fib = ifr->ifr_fib;
2333 		break;
2334 
2335 	case SIOCSIFFLAGS:
2336 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2337 		if (error)
2338 			return (error);
2339 		/*
2340 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2341 		 * check, so we don't need special handling here yet.
2342 		 */
2343 		new_flags = (ifr->ifr_flags & 0xffff) |
2344 		    (ifr->ifr_flagshigh << 16);
2345 		if (ifp->if_flags & IFF_UP &&
2346 		    (new_flags & IFF_UP) == 0) {
2347 			if_down(ifp);
2348 		} else if (new_flags & IFF_UP &&
2349 		    (ifp->if_flags & IFF_UP) == 0) {
2350 			if_up(ifp);
2351 		}
2352 		/* See if permanently promiscuous mode bit is about to flip */
2353 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2354 			if (new_flags & IFF_PPROMISC)
2355 				ifp->if_flags |= IFF_PROMISC;
2356 			else if (ifp->if_pcount == 0)
2357 				ifp->if_flags &= ~IFF_PROMISC;
2358 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2359 			    ifp->if_xname,
2360 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2361 		}
2362 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2363 			(new_flags &~ IFF_CANTCHANGE);
2364 		if (ifp->if_ioctl) {
2365 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2366 		}
2367 		getmicrotime(&ifp->if_lastchange);
2368 		break;
2369 
2370 	case SIOCSIFCAP:
2371 		error = priv_check(td, PRIV_NET_SETIFCAP);
2372 		if (error)
2373 			return (error);
2374 		if (ifp->if_ioctl == NULL)
2375 			return (EOPNOTSUPP);
2376 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2377 			return (EINVAL);
2378 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2379 		if (error == 0)
2380 			getmicrotime(&ifp->if_lastchange);
2381 		break;
2382 
2383 #ifdef MAC
2384 	case SIOCSIFMAC:
2385 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2386 		break;
2387 #endif
2388 
2389 	case SIOCSIFNAME:
2390 		error = priv_check(td, PRIV_NET_SETIFNAME);
2391 		if (error)
2392 			return (error);
2393 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2394 		if (error != 0)
2395 			return (error);
2396 		if (new_name[0] == '\0')
2397 			return (EINVAL);
2398 		if (ifunit(new_name) != NULL)
2399 			return (EEXIST);
2400 
2401 		/*
2402 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2403 		 * and there are numerous other races with the
2404 		 * ifunit() checks not being atomic with namespace
2405 		 * changes (renames, vmoves, if_attach, etc).
2406 		 */
2407 		ifp->if_flags |= IFF_RENAMING;
2408 
2409 		/* Announce the departure of the interface. */
2410 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2411 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2412 
2413 		log(LOG_INFO, "%s: changing name to '%s'\n",
2414 		    ifp->if_xname, new_name);
2415 
2416 		IF_ADDR_WLOCK(ifp);
2417 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2418 		ifa = ifp->if_addr;
2419 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2420 		namelen = strlen(new_name);
2421 		onamelen = sdl->sdl_nlen;
2422 		/*
2423 		 * Move the address if needed.  This is safe because we
2424 		 * allocate space for a name of length IFNAMSIZ when we
2425 		 * create this in if_attach().
2426 		 */
2427 		if (namelen != onamelen) {
2428 			bcopy(sdl->sdl_data + onamelen,
2429 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2430 		}
2431 		bcopy(new_name, sdl->sdl_data, namelen);
2432 		sdl->sdl_nlen = namelen;
2433 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2434 		bzero(sdl->sdl_data, onamelen);
2435 		while (namelen != 0)
2436 			sdl->sdl_data[--namelen] = 0xff;
2437 		IF_ADDR_WUNLOCK(ifp);
2438 
2439 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2440 		/* Announce the return of the interface. */
2441 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2442 
2443 		ifp->if_flags &= ~IFF_RENAMING;
2444 		break;
2445 
2446 #ifdef VIMAGE
2447 	case SIOCSIFVNET:
2448 		error = priv_check(td, PRIV_NET_SETIFVNET);
2449 		if (error)
2450 			return (error);
2451 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2452 		break;
2453 #endif
2454 
2455 	case SIOCSIFMETRIC:
2456 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2457 		if (error)
2458 			return (error);
2459 		ifp->if_metric = ifr->ifr_metric;
2460 		getmicrotime(&ifp->if_lastchange);
2461 		break;
2462 
2463 	case SIOCSIFPHYS:
2464 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2465 		if (error)
2466 			return (error);
2467 		if (ifp->if_ioctl == NULL)
2468 			return (EOPNOTSUPP);
2469 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2470 		if (error == 0)
2471 			getmicrotime(&ifp->if_lastchange);
2472 		break;
2473 
2474 	case SIOCSIFMTU:
2475 	{
2476 		u_long oldmtu = ifp->if_mtu;
2477 
2478 		error = priv_check(td, PRIV_NET_SETIFMTU);
2479 		if (error)
2480 			return (error);
2481 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2482 			return (EINVAL);
2483 		if (ifp->if_ioctl == NULL)
2484 			return (EOPNOTSUPP);
2485 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2486 		if (error == 0) {
2487 			getmicrotime(&ifp->if_lastchange);
2488 			rt_ifmsg(ifp);
2489 		}
2490 		/*
2491 		 * If the link MTU changed, do network layer specific procedure.
2492 		 */
2493 		if (ifp->if_mtu != oldmtu) {
2494 #ifdef INET6
2495 			nd6_setmtu(ifp);
2496 #endif
2497 			rt_updatemtu(ifp);
2498 		}
2499 		break;
2500 	}
2501 
2502 	case SIOCADDMULTI:
2503 	case SIOCDELMULTI:
2504 		if (cmd == SIOCADDMULTI)
2505 			error = priv_check(td, PRIV_NET_ADDMULTI);
2506 		else
2507 			error = priv_check(td, PRIV_NET_DELMULTI);
2508 		if (error)
2509 			return (error);
2510 
2511 		/* Don't allow group membership on non-multicast interfaces. */
2512 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2513 			return (EOPNOTSUPP);
2514 
2515 		/* Don't let users screw up protocols' entries. */
2516 		if (ifr->ifr_addr.sa_family != AF_LINK)
2517 			return (EINVAL);
2518 
2519 		if (cmd == SIOCADDMULTI) {
2520 			struct ifmultiaddr *ifma;
2521 
2522 			/*
2523 			 * Userland is only permitted to join groups once
2524 			 * via the if_addmulti() KPI, because it cannot hold
2525 			 * struct ifmultiaddr * between calls. It may also
2526 			 * lose a race while we check if the membership
2527 			 * already exists.
2528 			 */
2529 			IF_ADDR_RLOCK(ifp);
2530 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2531 			IF_ADDR_RUNLOCK(ifp);
2532 			if (ifma != NULL)
2533 				error = EADDRINUSE;
2534 			else
2535 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2536 		} else {
2537 			error = if_delmulti(ifp, &ifr->ifr_addr);
2538 		}
2539 		if (error == 0)
2540 			getmicrotime(&ifp->if_lastchange);
2541 		break;
2542 
2543 	case SIOCSIFPHYADDR:
2544 	case SIOCDIFPHYADDR:
2545 #ifdef INET6
2546 	case SIOCSIFPHYADDR_IN6:
2547 #endif
2548 	case SIOCSIFMEDIA:
2549 	case SIOCSIFGENERIC:
2550 		error = priv_check(td, PRIV_NET_HWIOCTL);
2551 		if (error)
2552 			return (error);
2553 		if (ifp->if_ioctl == NULL)
2554 			return (EOPNOTSUPP);
2555 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2556 		if (error == 0)
2557 			getmicrotime(&ifp->if_lastchange);
2558 		break;
2559 
2560 	case SIOCGIFSTATUS:
2561 	case SIOCGIFPSRCADDR:
2562 	case SIOCGIFPDSTADDR:
2563 	case SIOCGIFMEDIA:
2564 	case SIOCGIFGENERIC:
2565 		if (ifp->if_ioctl == NULL)
2566 			return (EOPNOTSUPP);
2567 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2568 		break;
2569 
2570 	case SIOCSIFLLADDR:
2571 		error = priv_check(td, PRIV_NET_SETLLADDR);
2572 		if (error)
2573 			return (error);
2574 		error = if_setlladdr(ifp,
2575 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2576 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2577 		break;
2578 
2579 	case SIOCAIFGROUP:
2580 	{
2581 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2582 
2583 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2584 		if (error)
2585 			return (error);
2586 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2587 			return (error);
2588 		break;
2589 	}
2590 
2591 	case SIOCGIFGROUP:
2592 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2593 			return (error);
2594 		break;
2595 
2596 	case SIOCDIFGROUP:
2597 	{
2598 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2599 
2600 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2601 		if (error)
2602 			return (error);
2603 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2604 			return (error);
2605 		break;
2606 	}
2607 
2608 	default:
2609 		error = ENOIOCTL;
2610 		break;
2611 	}
2612 	return (error);
2613 }
2614 
2615 #ifdef COMPAT_FREEBSD32
2616 struct ifconf32 {
2617 	int32_t	ifc_len;
2618 	union {
2619 		uint32_t	ifcu_buf;
2620 		uint32_t	ifcu_req;
2621 	} ifc_ifcu;
2622 };
2623 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2624 #endif
2625 
2626 /*
2627  * Interface ioctls.
2628  */
2629 int
2630 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2631 {
2632 	struct ifnet *ifp;
2633 	struct ifreq *ifr;
2634 	int error;
2635 	int oif_flags;
2636 
2637 	CURVNET_SET(so->so_vnet);
2638 	switch (cmd) {
2639 	case SIOCGIFCONF:
2640 		error = ifconf(cmd, data);
2641 		CURVNET_RESTORE();
2642 		return (error);
2643 
2644 #ifdef COMPAT_FREEBSD32
2645 	case SIOCGIFCONF32:
2646 		{
2647 			struct ifconf32 *ifc32;
2648 			struct ifconf ifc;
2649 
2650 			ifc32 = (struct ifconf32 *)data;
2651 			ifc.ifc_len = ifc32->ifc_len;
2652 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2653 
2654 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2655 			CURVNET_RESTORE();
2656 			if (error == 0)
2657 				ifc32->ifc_len = ifc.ifc_len;
2658 			return (error);
2659 		}
2660 #endif
2661 	}
2662 	ifr = (struct ifreq *)data;
2663 
2664 	switch (cmd) {
2665 #ifdef VIMAGE
2666 	case SIOCSIFRVNET:
2667 		error = priv_check(td, PRIV_NET_SETIFVNET);
2668 		if (error == 0)
2669 			error = if_vmove_reclaim(td, ifr->ifr_name,
2670 			    ifr->ifr_jid);
2671 		CURVNET_RESTORE();
2672 		return (error);
2673 #endif
2674 	case SIOCIFCREATE:
2675 	case SIOCIFCREATE2:
2676 		error = priv_check(td, PRIV_NET_IFCREATE);
2677 		if (error == 0)
2678 			error = if_clone_create(ifr->ifr_name,
2679 			    sizeof(ifr->ifr_name),
2680 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2681 		CURVNET_RESTORE();
2682 		return (error);
2683 	case SIOCIFDESTROY:
2684 		error = priv_check(td, PRIV_NET_IFDESTROY);
2685 		if (error == 0)
2686 			error = if_clone_destroy(ifr->ifr_name);
2687 		CURVNET_RESTORE();
2688 		return (error);
2689 
2690 	case SIOCIFGCLONERS:
2691 		error = if_clone_list((struct if_clonereq *)data);
2692 		CURVNET_RESTORE();
2693 		return (error);
2694 	case SIOCGIFGMEMB:
2695 		error = if_getgroupmembers((struct ifgroupreq *)data);
2696 		CURVNET_RESTORE();
2697 		return (error);
2698 #if defined(INET) || defined(INET6)
2699 	case SIOCSVH:
2700 	case SIOCGVH:
2701 		if (carp_ioctl_p == NULL)
2702 			error = EPROTONOSUPPORT;
2703 		else
2704 			error = (*carp_ioctl_p)(ifr, cmd, td);
2705 		CURVNET_RESTORE();
2706 		return (error);
2707 #endif
2708 	}
2709 
2710 	ifp = ifunit_ref(ifr->ifr_name);
2711 	if (ifp == NULL) {
2712 		CURVNET_RESTORE();
2713 		return (ENXIO);
2714 	}
2715 
2716 	error = ifhwioctl(cmd, ifp, data, td);
2717 	if (error != ENOIOCTL) {
2718 		if_rele(ifp);
2719 		CURVNET_RESTORE();
2720 		return (error);
2721 	}
2722 
2723 	oif_flags = ifp->if_flags;
2724 	if (so->so_proto == NULL) {
2725 		if_rele(ifp);
2726 		CURVNET_RESTORE();
2727 		return (EOPNOTSUPP);
2728 	}
2729 
2730 	/*
2731 	 * Pass the request on to the socket control method, and if the
2732 	 * latter returns EOPNOTSUPP, directly to the interface.
2733 	 *
2734 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2735 	 * trust SIOCSIFADDR et al to come from an already privileged
2736 	 * layer, and do not perform any credentials checks or input
2737 	 * validation.
2738 	 */
2739 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2740 	    ifp, td));
2741 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2742 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2743 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2744 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2745 
2746 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2747 #ifdef INET6
2748 		if (ifp->if_flags & IFF_UP)
2749 			in6_if_up(ifp);
2750 #endif
2751 	}
2752 	if_rele(ifp);
2753 	CURVNET_RESTORE();
2754 	return (error);
2755 }
2756 
2757 /*
2758  * The code common to handling reference counted flags,
2759  * e.g., in ifpromisc() and if_allmulti().
2760  * The "pflag" argument can specify a permanent mode flag to check,
2761  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2762  *
2763  * Only to be used on stack-owned flags, not driver-owned flags.
2764  */
2765 static int
2766 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2767 {
2768 	struct ifreq ifr;
2769 	int error;
2770 	int oldflags, oldcount;
2771 
2772 	/* Sanity checks to catch programming errors */
2773 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2774 	    ("%s: setting driver-owned flag %d", __func__, flag));
2775 
2776 	if (onswitch)
2777 		KASSERT(*refcount >= 0,
2778 		    ("%s: increment negative refcount %d for flag %d",
2779 		    __func__, *refcount, flag));
2780 	else
2781 		KASSERT(*refcount > 0,
2782 		    ("%s: decrement non-positive refcount %d for flag %d",
2783 		    __func__, *refcount, flag));
2784 
2785 	/* In case this mode is permanent, just touch refcount */
2786 	if (ifp->if_flags & pflag) {
2787 		*refcount += onswitch ? 1 : -1;
2788 		return (0);
2789 	}
2790 
2791 	/* Save ifnet parameters for if_ioctl() may fail */
2792 	oldcount = *refcount;
2793 	oldflags = ifp->if_flags;
2794 
2795 	/*
2796 	 * See if we aren't the only and touching refcount is enough.
2797 	 * Actually toggle interface flag if we are the first or last.
2798 	 */
2799 	if (onswitch) {
2800 		if ((*refcount)++)
2801 			return (0);
2802 		ifp->if_flags |= flag;
2803 	} else {
2804 		if (--(*refcount))
2805 			return (0);
2806 		ifp->if_flags &= ~flag;
2807 	}
2808 
2809 	/* Call down the driver since we've changed interface flags */
2810 	if (ifp->if_ioctl == NULL) {
2811 		error = EOPNOTSUPP;
2812 		goto recover;
2813 	}
2814 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2815 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2816 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2817 	if (error)
2818 		goto recover;
2819 	/* Notify userland that interface flags have changed */
2820 	rt_ifmsg(ifp);
2821 	return (0);
2822 
2823 recover:
2824 	/* Recover after driver error */
2825 	*refcount = oldcount;
2826 	ifp->if_flags = oldflags;
2827 	return (error);
2828 }
2829 
2830 /*
2831  * Set/clear promiscuous mode on interface ifp based on the truth value
2832  * of pswitch.  The calls are reference counted so that only the first
2833  * "on" request actually has an effect, as does the final "off" request.
2834  * Results are undefined if the "off" and "on" requests are not matched.
2835  */
2836 int
2837 ifpromisc(struct ifnet *ifp, int pswitch)
2838 {
2839 	int error;
2840 	int oldflags = ifp->if_flags;
2841 
2842 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2843 			   &ifp->if_pcount, pswitch);
2844 	/* If promiscuous mode status has changed, log a message */
2845 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2846 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2847 		    ifp->if_xname,
2848 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2849 	return (error);
2850 }
2851 
2852 /*
2853  * Return interface configuration
2854  * of system.  List may be used
2855  * in later ioctl's (above) to get
2856  * other information.
2857  */
2858 /*ARGSUSED*/
2859 static int
2860 ifconf(u_long cmd, caddr_t data)
2861 {
2862 	struct ifconf *ifc = (struct ifconf *)data;
2863 	struct ifnet *ifp;
2864 	struct ifaddr *ifa;
2865 	struct ifreq ifr;
2866 	struct sbuf *sb;
2867 	int error, full = 0, valid_len, max_len;
2868 
2869 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2870 	max_len = MAXPHYS - 1;
2871 
2872 	/* Prevent hostile input from being able to crash the system */
2873 	if (ifc->ifc_len <= 0)
2874 		return (EINVAL);
2875 
2876 again:
2877 	if (ifc->ifc_len <= max_len) {
2878 		max_len = ifc->ifc_len;
2879 		full = 1;
2880 	}
2881 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2882 	max_len = 0;
2883 	valid_len = 0;
2884 
2885 	IFNET_RLOCK();
2886 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2887 		int addrs;
2888 
2889 		/*
2890 		 * Zero the ifr_name buffer to make sure we don't
2891 		 * disclose the contents of the stack.
2892 		 */
2893 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2894 
2895 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2896 		    >= sizeof(ifr.ifr_name)) {
2897 			sbuf_delete(sb);
2898 			IFNET_RUNLOCK();
2899 			return (ENAMETOOLONG);
2900 		}
2901 
2902 		addrs = 0;
2903 		IF_ADDR_RLOCK(ifp);
2904 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2905 			struct sockaddr *sa = ifa->ifa_addr;
2906 
2907 			if (prison_if(curthread->td_ucred, sa) != 0)
2908 				continue;
2909 			addrs++;
2910 			if (sa->sa_len <= sizeof(*sa)) {
2911 				ifr.ifr_addr = *sa;
2912 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2913 				max_len += sizeof(ifr);
2914 			} else {
2915 				sbuf_bcat(sb, &ifr,
2916 				    offsetof(struct ifreq, ifr_addr));
2917 				max_len += offsetof(struct ifreq, ifr_addr);
2918 				sbuf_bcat(sb, sa, sa->sa_len);
2919 				max_len += sa->sa_len;
2920 			}
2921 
2922 			if (sbuf_error(sb) == 0)
2923 				valid_len = sbuf_len(sb);
2924 		}
2925 		IF_ADDR_RUNLOCK(ifp);
2926 		if (addrs == 0) {
2927 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2928 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2929 			max_len += sizeof(ifr);
2930 
2931 			if (sbuf_error(sb) == 0)
2932 				valid_len = sbuf_len(sb);
2933 		}
2934 	}
2935 	IFNET_RUNLOCK();
2936 
2937 	/*
2938 	 * If we didn't allocate enough space (uncommon), try again.  If
2939 	 * we have already allocated as much space as we are allowed,
2940 	 * return what we've got.
2941 	 */
2942 	if (valid_len != max_len && !full) {
2943 		sbuf_delete(sb);
2944 		goto again;
2945 	}
2946 
2947 	ifc->ifc_len = valid_len;
2948 	sbuf_finish(sb);
2949 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2950 	sbuf_delete(sb);
2951 	return (error);
2952 }
2953 
2954 /*
2955  * Just like ifpromisc(), but for all-multicast-reception mode.
2956  */
2957 int
2958 if_allmulti(struct ifnet *ifp, int onswitch)
2959 {
2960 
2961 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2962 }
2963 
2964 struct ifmultiaddr *
2965 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2966 {
2967 	struct ifmultiaddr *ifma;
2968 
2969 	IF_ADDR_LOCK_ASSERT(ifp);
2970 
2971 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2972 		if (sa->sa_family == AF_LINK) {
2973 			if (sa_dl_equal(ifma->ifma_addr, sa))
2974 				break;
2975 		} else {
2976 			if (sa_equal(ifma->ifma_addr, sa))
2977 				break;
2978 		}
2979 	}
2980 
2981 	return ifma;
2982 }
2983 
2984 /*
2985  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2986  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2987  * the ifnet multicast address list here, so the caller must do that and
2988  * other setup work (such as notifying the device driver).  The reference
2989  * count is initialized to 1.
2990  */
2991 static struct ifmultiaddr *
2992 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2993     int mflags)
2994 {
2995 	struct ifmultiaddr *ifma;
2996 	struct sockaddr *dupsa;
2997 
2998 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2999 	    M_ZERO);
3000 	if (ifma == NULL)
3001 		return (NULL);
3002 
3003 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3004 	if (dupsa == NULL) {
3005 		free(ifma, M_IFMADDR);
3006 		return (NULL);
3007 	}
3008 	bcopy(sa, dupsa, sa->sa_len);
3009 	ifma->ifma_addr = dupsa;
3010 
3011 	ifma->ifma_ifp = ifp;
3012 	ifma->ifma_refcount = 1;
3013 	ifma->ifma_protospec = NULL;
3014 
3015 	if (llsa == NULL) {
3016 		ifma->ifma_lladdr = NULL;
3017 		return (ifma);
3018 	}
3019 
3020 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3021 	if (dupsa == NULL) {
3022 		free(ifma->ifma_addr, M_IFMADDR);
3023 		free(ifma, M_IFMADDR);
3024 		return (NULL);
3025 	}
3026 	bcopy(llsa, dupsa, llsa->sa_len);
3027 	ifma->ifma_lladdr = dupsa;
3028 
3029 	return (ifma);
3030 }
3031 
3032 /*
3033  * if_freemulti: free ifmultiaddr structure and possibly attached related
3034  * addresses.  The caller is responsible for implementing reference
3035  * counting, notifying the driver, handling routing messages, and releasing
3036  * any dependent link layer state.
3037  */
3038 static void
3039 if_freemulti(struct ifmultiaddr *ifma)
3040 {
3041 
3042 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3043 	    ifma->ifma_refcount));
3044 
3045 	if (ifma->ifma_lladdr != NULL)
3046 		free(ifma->ifma_lladdr, M_IFMADDR);
3047 	free(ifma->ifma_addr, M_IFMADDR);
3048 	free(ifma, M_IFMADDR);
3049 }
3050 
3051 /*
3052  * Register an additional multicast address with a network interface.
3053  *
3054  * - If the address is already present, bump the reference count on the
3055  *   address and return.
3056  * - If the address is not link-layer, look up a link layer address.
3057  * - Allocate address structures for one or both addresses, and attach to the
3058  *   multicast address list on the interface.  If automatically adding a link
3059  *   layer address, the protocol address will own a reference to the link
3060  *   layer address, to be freed when it is freed.
3061  * - Notify the network device driver of an addition to the multicast address
3062  *   list.
3063  *
3064  * 'sa' points to caller-owned memory with the desired multicast address.
3065  *
3066  * 'retifma' will be used to return a pointer to the resulting multicast
3067  * address reference, if desired.
3068  */
3069 int
3070 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3071     struct ifmultiaddr **retifma)
3072 {
3073 	struct ifmultiaddr *ifma, *ll_ifma;
3074 	struct sockaddr *llsa;
3075 	struct sockaddr_dl sdl;
3076 	int error;
3077 
3078 	/*
3079 	 * If the address is already present, return a new reference to it;
3080 	 * otherwise, allocate storage and set up a new address.
3081 	 */
3082 	IF_ADDR_WLOCK(ifp);
3083 	ifma = if_findmulti(ifp, sa);
3084 	if (ifma != NULL) {
3085 		ifma->ifma_refcount++;
3086 		if (retifma != NULL)
3087 			*retifma = ifma;
3088 		IF_ADDR_WUNLOCK(ifp);
3089 		return (0);
3090 	}
3091 
3092 	/*
3093 	 * The address isn't already present; resolve the protocol address
3094 	 * into a link layer address, and then look that up, bump its
3095 	 * refcount or allocate an ifma for that also.
3096 	 * Most link layer resolving functions returns address data which
3097 	 * fits inside default sockaddr_dl structure. However callback
3098 	 * can allocate another sockaddr structure, in that case we need to
3099 	 * free it later.
3100 	 */
3101 	llsa = NULL;
3102 	ll_ifma = NULL;
3103 	if (ifp->if_resolvemulti != NULL) {
3104 		/* Provide called function with buffer size information */
3105 		sdl.sdl_len = sizeof(sdl);
3106 		llsa = (struct sockaddr *)&sdl;
3107 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3108 		if (error)
3109 			goto unlock_out;
3110 	}
3111 
3112 	/*
3113 	 * Allocate the new address.  Don't hook it up yet, as we may also
3114 	 * need to allocate a link layer multicast address.
3115 	 */
3116 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3117 	if (ifma == NULL) {
3118 		error = ENOMEM;
3119 		goto free_llsa_out;
3120 	}
3121 
3122 	/*
3123 	 * If a link layer address is found, we'll need to see if it's
3124 	 * already present in the address list, or allocate is as well.
3125 	 * When this block finishes, the link layer address will be on the
3126 	 * list.
3127 	 */
3128 	if (llsa != NULL) {
3129 		ll_ifma = if_findmulti(ifp, llsa);
3130 		if (ll_ifma == NULL) {
3131 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3132 			if (ll_ifma == NULL) {
3133 				--ifma->ifma_refcount;
3134 				if_freemulti(ifma);
3135 				error = ENOMEM;
3136 				goto free_llsa_out;
3137 			}
3138 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3139 			    ifma_link);
3140 		} else
3141 			ll_ifma->ifma_refcount++;
3142 		ifma->ifma_llifma = ll_ifma;
3143 	}
3144 
3145 	/*
3146 	 * We now have a new multicast address, ifma, and possibly a new or
3147 	 * referenced link layer address.  Add the primary address to the
3148 	 * ifnet address list.
3149 	 */
3150 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3151 
3152 	if (retifma != NULL)
3153 		*retifma = ifma;
3154 
3155 	/*
3156 	 * Must generate the message while holding the lock so that 'ifma'
3157 	 * pointer is still valid.
3158 	 */
3159 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3160 	IF_ADDR_WUNLOCK(ifp);
3161 
3162 	/*
3163 	 * We are certain we have added something, so call down to the
3164 	 * interface to let them know about it.
3165 	 */
3166 	if (ifp->if_ioctl != NULL) {
3167 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3168 	}
3169 
3170 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3171 		link_free_sdl(llsa);
3172 
3173 	return (0);
3174 
3175 free_llsa_out:
3176 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3177 		link_free_sdl(llsa);
3178 
3179 unlock_out:
3180 	IF_ADDR_WUNLOCK(ifp);
3181 	return (error);
3182 }
3183 
3184 /*
3185  * Delete a multicast group membership by network-layer group address.
3186  *
3187  * Returns ENOENT if the entry could not be found. If ifp no longer
3188  * exists, results are undefined. This entry point should only be used
3189  * from subsystems which do appropriate locking to hold ifp for the
3190  * duration of the call.
3191  * Network-layer protocol domains must use if_delmulti_ifma().
3192  */
3193 int
3194 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3195 {
3196 	struct ifmultiaddr *ifma;
3197 	int lastref;
3198 #ifdef INVARIANTS
3199 	struct ifnet *oifp;
3200 
3201 	IFNET_RLOCK_NOSLEEP();
3202 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3203 		if (ifp == oifp)
3204 			break;
3205 	if (ifp != oifp)
3206 		ifp = NULL;
3207 	IFNET_RUNLOCK_NOSLEEP();
3208 
3209 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3210 #endif
3211 	if (ifp == NULL)
3212 		return (ENOENT);
3213 
3214 	IF_ADDR_WLOCK(ifp);
3215 	lastref = 0;
3216 	ifma = if_findmulti(ifp, sa);
3217 	if (ifma != NULL)
3218 		lastref = if_delmulti_locked(ifp, ifma, 0);
3219 	IF_ADDR_WUNLOCK(ifp);
3220 
3221 	if (ifma == NULL)
3222 		return (ENOENT);
3223 
3224 	if (lastref && ifp->if_ioctl != NULL) {
3225 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3226 	}
3227 
3228 	return (0);
3229 }
3230 
3231 /*
3232  * Delete all multicast group membership for an interface.
3233  * Should be used to quickly flush all multicast filters.
3234  */
3235 void
3236 if_delallmulti(struct ifnet *ifp)
3237 {
3238 	struct ifmultiaddr *ifma;
3239 	struct ifmultiaddr *next;
3240 
3241 	IF_ADDR_WLOCK(ifp);
3242 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3243 		if_delmulti_locked(ifp, ifma, 0);
3244 	IF_ADDR_WUNLOCK(ifp);
3245 }
3246 
3247 /*
3248  * Delete a multicast group membership by group membership pointer.
3249  * Network-layer protocol domains must use this routine.
3250  *
3251  * It is safe to call this routine if the ifp disappeared.
3252  */
3253 void
3254 if_delmulti_ifma(struct ifmultiaddr *ifma)
3255 {
3256 	struct ifnet *ifp;
3257 	int lastref;
3258 
3259 	ifp = ifma->ifma_ifp;
3260 #ifdef DIAGNOSTIC
3261 	if (ifp == NULL) {
3262 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3263 	} else {
3264 		struct ifnet *oifp;
3265 
3266 		IFNET_RLOCK_NOSLEEP();
3267 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3268 			if (ifp == oifp)
3269 				break;
3270 		if (ifp != oifp) {
3271 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3272 			ifp = NULL;
3273 		}
3274 		IFNET_RUNLOCK_NOSLEEP();
3275 	}
3276 #endif
3277 	/*
3278 	 * If and only if the ifnet instance exists: Acquire the address lock.
3279 	 */
3280 	if (ifp != NULL)
3281 		IF_ADDR_WLOCK(ifp);
3282 
3283 	lastref = if_delmulti_locked(ifp, ifma, 0);
3284 
3285 	if (ifp != NULL) {
3286 		/*
3287 		 * If and only if the ifnet instance exists:
3288 		 *  Release the address lock.
3289 		 *  If the group was left: update the hardware hash filter.
3290 		 */
3291 		IF_ADDR_WUNLOCK(ifp);
3292 		if (lastref && ifp->if_ioctl != NULL) {
3293 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3294 		}
3295 	}
3296 }
3297 
3298 /*
3299  * Perform deletion of network-layer and/or link-layer multicast address.
3300  *
3301  * Return 0 if the reference count was decremented.
3302  * Return 1 if the final reference was released, indicating that the
3303  * hardware hash filter should be reprogrammed.
3304  */
3305 static int
3306 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3307 {
3308 	struct ifmultiaddr *ll_ifma;
3309 
3310 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3311 		KASSERT(ifma->ifma_ifp == ifp,
3312 		    ("%s: inconsistent ifp %p", __func__, ifp));
3313 		IF_ADDR_WLOCK_ASSERT(ifp);
3314 	}
3315 
3316 	ifp = ifma->ifma_ifp;
3317 
3318 	/*
3319 	 * If the ifnet is detaching, null out references to ifnet,
3320 	 * so that upper protocol layers will notice, and not attempt
3321 	 * to obtain locks for an ifnet which no longer exists. The
3322 	 * routing socket announcement must happen before the ifnet
3323 	 * instance is detached from the system.
3324 	 */
3325 	if (detaching) {
3326 #ifdef DIAGNOSTIC
3327 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3328 #endif
3329 		/*
3330 		 * ifp may already be nulled out if we are being reentered
3331 		 * to delete the ll_ifma.
3332 		 */
3333 		if (ifp != NULL) {
3334 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3335 			ifma->ifma_ifp = NULL;
3336 		}
3337 	}
3338 
3339 	if (--ifma->ifma_refcount > 0)
3340 		return 0;
3341 
3342 	/*
3343 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3344 	 * have been associated with it. Release it first if so.
3345 	 */
3346 	ll_ifma = ifma->ifma_llifma;
3347 	if (ll_ifma != NULL) {
3348 		KASSERT(ifma->ifma_lladdr != NULL,
3349 		    ("%s: llifma w/o lladdr", __func__));
3350 		if (detaching)
3351 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3352 		if (--ll_ifma->ifma_refcount == 0) {
3353 			if (ifp != NULL) {
3354 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3355 				    ifma_link);
3356 			}
3357 			if_freemulti(ll_ifma);
3358 		}
3359 	}
3360 
3361 	if (ifp != NULL)
3362 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3363 
3364 	if_freemulti(ifma);
3365 
3366 	/*
3367 	 * The last reference to this instance of struct ifmultiaddr
3368 	 * was released; the hardware should be notified of this change.
3369 	 */
3370 	return 1;
3371 }
3372 
3373 /*
3374  * Set the link layer address on an interface.
3375  *
3376  * At this time we only support certain types of interfaces,
3377  * and we don't allow the length of the address to change.
3378  */
3379 int
3380 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3381 {
3382 	struct sockaddr_dl *sdl;
3383 	struct ifaddr *ifa;
3384 	struct ifreq ifr;
3385 
3386 	IF_ADDR_RLOCK(ifp);
3387 	ifa = ifp->if_addr;
3388 	if (ifa == NULL) {
3389 		IF_ADDR_RUNLOCK(ifp);
3390 		return (EINVAL);
3391 	}
3392 	ifa_ref(ifa);
3393 	IF_ADDR_RUNLOCK(ifp);
3394 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3395 	if (sdl == NULL) {
3396 		ifa_free(ifa);
3397 		return (EINVAL);
3398 	}
3399 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3400 		ifa_free(ifa);
3401 		return (EINVAL);
3402 	}
3403 	switch (ifp->if_type) {
3404 	case IFT_ETHER:
3405 	case IFT_FDDI:
3406 	case IFT_XETHER:
3407 	case IFT_ISO88025:
3408 	case IFT_L2VLAN:
3409 	case IFT_BRIDGE:
3410 	case IFT_ARCNET:
3411 	case IFT_IEEE8023ADLAG:
3412 	case IFT_IEEE80211:
3413 		bcopy(lladdr, LLADDR(sdl), len);
3414 		ifa_free(ifa);
3415 		break;
3416 	default:
3417 		ifa_free(ifa);
3418 		return (ENODEV);
3419 	}
3420 
3421 	/*
3422 	 * If the interface is already up, we need
3423 	 * to re-init it in order to reprogram its
3424 	 * address filter.
3425 	 */
3426 	if ((ifp->if_flags & IFF_UP) != 0) {
3427 		if (ifp->if_ioctl) {
3428 			ifp->if_flags &= ~IFF_UP;
3429 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3430 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3431 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
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 		}
3437 #ifdef INET
3438 		/*
3439 		 * Also send gratuitous ARPs to notify other nodes about
3440 		 * the address change.
3441 		 */
3442 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3443 			if (ifa->ifa_addr->sa_family == AF_INET)
3444 				arp_ifinit(ifp, ifa);
3445 		}
3446 #endif
3447 	}
3448 	return (0);
3449 }
3450 
3451 /*
3452  * The name argument must be a pointer to storage which will last as
3453  * long as the interface does.  For physical devices, the result of
3454  * device_get_name(dev) is a good choice and for pseudo-devices a
3455  * static string works well.
3456  */
3457 void
3458 if_initname(struct ifnet *ifp, const char *name, int unit)
3459 {
3460 	ifp->if_dname = name;
3461 	ifp->if_dunit = unit;
3462 	if (unit != IF_DUNIT_NONE)
3463 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3464 	else
3465 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3466 }
3467 
3468 int
3469 if_printf(struct ifnet *ifp, const char * fmt, ...)
3470 {
3471 	va_list ap;
3472 	int retval;
3473 
3474 	retval = printf("%s: ", ifp->if_xname);
3475 	va_start(ap, fmt);
3476 	retval += vprintf(fmt, ap);
3477 	va_end(ap);
3478 	return (retval);
3479 }
3480 
3481 void
3482 if_start(struct ifnet *ifp)
3483 {
3484 
3485 	(*(ifp)->if_start)(ifp);
3486 }
3487 
3488 /*
3489  * Backwards compatibility interface for drivers
3490  * that have not implemented it
3491  */
3492 static int
3493 if_transmit(struct ifnet *ifp, struct mbuf *m)
3494 {
3495 	int error;
3496 
3497 	IFQ_HANDOFF(ifp, m, error);
3498 	return (error);
3499 }
3500 
3501 int
3502 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3503 {
3504 	int active = 0;
3505 
3506 	IF_LOCK(ifq);
3507 	if (_IF_QFULL(ifq)) {
3508 		IF_UNLOCK(ifq);
3509 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3510 		m_freem(m);
3511 		return (0);
3512 	}
3513 	if (ifp != NULL) {
3514 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3515 		if (m->m_flags & (M_BCAST|M_MCAST))
3516 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3517 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3518 	}
3519 	_IF_ENQUEUE(ifq, m);
3520 	IF_UNLOCK(ifq);
3521 	if (ifp != NULL && !active)
3522 		(*(ifp)->if_start)(ifp);
3523 	return (1);
3524 }
3525 
3526 void
3527 if_register_com_alloc(u_char type,
3528     if_com_alloc_t *a, if_com_free_t *f)
3529 {
3530 
3531 	KASSERT(if_com_alloc[type] == NULL,
3532 	    ("if_register_com_alloc: %d already registered", type));
3533 	KASSERT(if_com_free[type] == NULL,
3534 	    ("if_register_com_alloc: %d free already registered", type));
3535 
3536 	if_com_alloc[type] = a;
3537 	if_com_free[type] = f;
3538 }
3539 
3540 void
3541 if_deregister_com_alloc(u_char type)
3542 {
3543 
3544 	KASSERT(if_com_alloc[type] != NULL,
3545 	    ("if_deregister_com_alloc: %d not registered", type));
3546 	KASSERT(if_com_free[type] != NULL,
3547 	    ("if_deregister_com_alloc: %d free not registered", type));
3548 	if_com_alloc[type] = NULL;
3549 	if_com_free[type] = NULL;
3550 }
3551 
3552 /* API for driver access to network stack owned ifnet.*/
3553 uint64_t
3554 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3555 {
3556 	uint64_t oldbrate;
3557 
3558 	oldbrate = ifp->if_baudrate;
3559 	ifp->if_baudrate = baudrate;
3560 	return (oldbrate);
3561 }
3562 
3563 uint64_t
3564 if_getbaudrate(if_t ifp)
3565 {
3566 
3567 	return (((struct ifnet *)ifp)->if_baudrate);
3568 }
3569 
3570 int
3571 if_setcapabilities(if_t ifp, int capabilities)
3572 {
3573 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3574 	return (0);
3575 }
3576 
3577 int
3578 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3579 {
3580 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3581 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3582 
3583 	return (0);
3584 }
3585 
3586 int
3587 if_getcapabilities(if_t ifp)
3588 {
3589 	return ((struct ifnet *)ifp)->if_capabilities;
3590 }
3591 
3592 int
3593 if_setcapenable(if_t ifp, int capabilities)
3594 {
3595 	((struct ifnet *)ifp)->if_capenable = capabilities;
3596 	return (0);
3597 }
3598 
3599 int
3600 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3601 {
3602 	if(setcap)
3603 		((struct ifnet *)ifp)->if_capenable |= setcap;
3604 	if(clearcap)
3605 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3606 
3607 	return (0);
3608 }
3609 
3610 const char *
3611 if_getdname(if_t ifp)
3612 {
3613 	return ((struct ifnet *)ifp)->if_dname;
3614 }
3615 
3616 int
3617 if_togglecapenable(if_t ifp, int togglecap)
3618 {
3619 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3620 	return (0);
3621 }
3622 
3623 int
3624 if_getcapenable(if_t ifp)
3625 {
3626 	return ((struct ifnet *)ifp)->if_capenable;
3627 }
3628 
3629 /*
3630  * This is largely undesirable because it ties ifnet to a device, but does
3631  * provide flexiblity for an embedded product vendor. Should be used with
3632  * the understanding that it violates the interface boundaries, and should be
3633  * a last resort only.
3634  */
3635 int
3636 if_setdev(if_t ifp, void *dev)
3637 {
3638 	return (0);
3639 }
3640 
3641 int
3642 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3643 {
3644 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3645 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3646 
3647 	return (0);
3648 }
3649 
3650 int
3651 if_getdrvflags(if_t ifp)
3652 {
3653 	return ((struct ifnet *)ifp)->if_drv_flags;
3654 }
3655 
3656 int
3657 if_setdrvflags(if_t ifp, int flags)
3658 {
3659 	((struct ifnet *)ifp)->if_drv_flags = flags;
3660 	return (0);
3661 }
3662 
3663 
3664 int
3665 if_setflags(if_t ifp, int flags)
3666 {
3667 	((struct ifnet *)ifp)->if_flags = flags;
3668 	return (0);
3669 }
3670 
3671 int
3672 if_setflagbits(if_t ifp, int set, int clear)
3673 {
3674 	((struct ifnet *)ifp)->if_flags |= set;
3675 	((struct ifnet *)ifp)->if_flags &= ~clear;
3676 
3677 	return (0);
3678 }
3679 
3680 int
3681 if_getflags(if_t ifp)
3682 {
3683 	return ((struct ifnet *)ifp)->if_flags;
3684 }
3685 
3686 int
3687 if_clearhwassist(if_t ifp)
3688 {
3689 	((struct ifnet *)ifp)->if_hwassist = 0;
3690 	return (0);
3691 }
3692 
3693 int
3694 if_sethwassistbits(if_t ifp, int toset, int toclear)
3695 {
3696 	((struct ifnet *)ifp)->if_hwassist |= toset;
3697 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3698 
3699 	return (0);
3700 }
3701 
3702 int
3703 if_sethwassist(if_t ifp, int hwassist_bit)
3704 {
3705 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3706 	return (0);
3707 }
3708 
3709 int
3710 if_gethwassist(if_t ifp)
3711 {
3712 	return ((struct ifnet *)ifp)->if_hwassist;
3713 }
3714 
3715 int
3716 if_setmtu(if_t ifp, int mtu)
3717 {
3718 	((struct ifnet *)ifp)->if_mtu = mtu;
3719 	return (0);
3720 }
3721 
3722 int
3723 if_getmtu(if_t ifp)
3724 {
3725 	return ((struct ifnet *)ifp)->if_mtu;
3726 }
3727 
3728 int
3729 if_getmtu_family(if_t ifp, int family)
3730 {
3731 	struct domain *dp;
3732 
3733 	for (dp = domains; dp; dp = dp->dom_next) {
3734 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3735 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3736 	}
3737 
3738 	return (((struct ifnet *)ifp)->if_mtu);
3739 }
3740 
3741 int
3742 if_setsoftc(if_t ifp, void *softc)
3743 {
3744 	((struct ifnet *)ifp)->if_softc = softc;
3745 	return (0);
3746 }
3747 
3748 void *
3749 if_getsoftc(if_t ifp)
3750 {
3751 	return ((struct ifnet *)ifp)->if_softc;
3752 }
3753 
3754 void
3755 if_setrcvif(struct mbuf *m, if_t ifp)
3756 {
3757 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3758 }
3759 
3760 void
3761 if_setvtag(struct mbuf *m, uint16_t tag)
3762 {
3763 	m->m_pkthdr.ether_vtag = tag;
3764 }
3765 
3766 uint16_t
3767 if_getvtag(struct mbuf *m)
3768 {
3769 
3770 	return (m->m_pkthdr.ether_vtag);
3771 }
3772 
3773 int
3774 if_sendq_empty(if_t ifp)
3775 {
3776 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3777 }
3778 
3779 struct ifaddr *
3780 if_getifaddr(if_t ifp)
3781 {
3782 	return ((struct ifnet *)ifp)->if_addr;
3783 }
3784 
3785 int
3786 if_getamcount(if_t ifp)
3787 {
3788 	return ((struct ifnet *)ifp)->if_amcount;
3789 }
3790 
3791 
3792 int
3793 if_setsendqready(if_t ifp)
3794 {
3795 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3796 	return (0);
3797 }
3798 
3799 int
3800 if_setsendqlen(if_t ifp, int tx_desc_count)
3801 {
3802 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3803 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3804 
3805 	return (0);
3806 }
3807 
3808 int
3809 if_vlantrunkinuse(if_t ifp)
3810 {
3811 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3812 }
3813 
3814 int
3815 if_input(if_t ifp, struct mbuf* sendmp)
3816 {
3817 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3818 	return (0);
3819 
3820 }
3821 
3822 /* XXX */
3823 #ifndef ETH_ADDR_LEN
3824 #define ETH_ADDR_LEN 6
3825 #endif
3826 
3827 int
3828 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3829 {
3830 	struct ifmultiaddr *ifma;
3831 	uint8_t *lmta = (uint8_t *)mta;
3832 	int mcnt = 0;
3833 
3834 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3835 		if (ifma->ifma_addr->sa_family != AF_LINK)
3836 			continue;
3837 
3838 		if (mcnt == max)
3839 			break;
3840 
3841 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3842 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3843 		mcnt++;
3844 	}
3845 	*cnt = mcnt;
3846 
3847 	return (0);
3848 }
3849 
3850 int
3851 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3852 {
3853 	int error;
3854 
3855 	if_maddr_rlock(ifp);
3856 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3857 	if_maddr_runlock(ifp);
3858 	return (error);
3859 }
3860 
3861 int
3862 if_multiaddr_count(if_t ifp, int max)
3863 {
3864 	struct ifmultiaddr *ifma;
3865 	int count;
3866 
3867 	count = 0;
3868 	if_maddr_rlock(ifp);
3869 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3870 		if (ifma->ifma_addr->sa_family != AF_LINK)
3871 			continue;
3872 		count++;
3873 		if (count == max)
3874 			break;
3875 	}
3876 	if_maddr_runlock(ifp);
3877 	return (count);
3878 }
3879 
3880 struct mbuf *
3881 if_dequeue(if_t ifp)
3882 {
3883 	struct mbuf *m;
3884 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3885 
3886 	return (m);
3887 }
3888 
3889 int
3890 if_sendq_prepend(if_t ifp, struct mbuf *m)
3891 {
3892 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3893 	return (0);
3894 }
3895 
3896 int
3897 if_setifheaderlen(if_t ifp, int len)
3898 {
3899 	((struct ifnet *)ifp)->if_hdrlen = len;
3900 	return (0);
3901 }
3902 
3903 caddr_t
3904 if_getlladdr(if_t ifp)
3905 {
3906 	return (IF_LLADDR((struct ifnet *)ifp));
3907 }
3908 
3909 void *
3910 if_gethandle(u_char type)
3911 {
3912 	return (if_alloc(type));
3913 }
3914 
3915 void
3916 if_bpfmtap(if_t ifh, struct mbuf *m)
3917 {
3918 	struct ifnet *ifp = (struct ifnet *)ifh;
3919 
3920 	BPF_MTAP(ifp, m);
3921 }
3922 
3923 void
3924 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3925 {
3926 	struct ifnet *ifp = (struct ifnet *)ifh;
3927 
3928 	ETHER_BPF_MTAP(ifp, m);
3929 }
3930 
3931 void
3932 if_vlancap(if_t ifh)
3933 {
3934 	struct ifnet *ifp = (struct ifnet *)ifh;
3935 	VLAN_CAPABILITIES(ifp);
3936 }
3937 
3938 void
3939 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3940 {
3941 	((struct ifnet *)ifp)->if_init = init_fn;
3942 }
3943 
3944 void
3945 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3946 {
3947 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3948 }
3949 
3950 void
3951 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3952 {
3953 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
3954 }
3955 
3956 void
3957 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3958 {
3959 	((struct ifnet *)ifp)->if_transmit = start_fn;
3960 }
3961 
3962 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3963 {
3964 	((struct ifnet *)ifp)->if_qflush = flush_fn;
3965 
3966 }
3967 
3968 void
3969 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
3970 {
3971 
3972 	ifp->if_get_counter = fn;
3973 }
3974 
3975 /* Revisit these - These are inline functions originally. */
3976 int
3977 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
3978 {
3979 	return drbr_inuse_drv(ifh, br);
3980 }
3981 
3982 struct mbuf*
3983 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
3984 {
3985 	return drbr_dequeue(ifh, br);
3986 }
3987 
3988 int
3989 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
3990 {
3991 	return drbr_needs_enqueue(ifh, br);
3992 }
3993 
3994 int
3995 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
3996 {
3997 	return drbr_enqueue(ifh, br, m);
3998 
3999 }
4000