xref: /freebsd/sys/net/if.c (revision dd48af360fdbbb9552f9fc6de7abe50d68ad5331)
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 #include "opt_carp.h"
37 #include "opt_ddb.h"
38 
39 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/conf.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #ifdef DDB
67 #include <ddb/ddb.h>
68 #endif
69 
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_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 /*XXX*/
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #ifdef INET6
85 #include <netinet6/in6_var.h>
86 #include <netinet6/in6_ifattach.h>
87 #endif
88 #endif
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #endif
92 #if defined(INET) || defined(INET6)
93 #ifdef DEV_CARP
94 #include <netinet/ip_carp.h>
95 #endif
96 #endif
97 
98 #include <security/mac/mac_framework.h>
99 
100 struct ifindex_entry {
101 	struct  ifnet *ife_ifnet;
102 };
103 
104 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
105 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
106 
107 /* Log link state change events */
108 static int log_link_state_change = 1;
109 
110 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
111 	&log_link_state_change, 0,
112 	"log interface link state change events");
113 
114 /* Interface description */
115 static unsigned int ifdescr_maxlen = 1024;
116 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
117 	&ifdescr_maxlen, 0,
118 	"administrative maximum length for interface description");
119 
120 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
121 
122 /* global sx for non-critical path ifdescr */
123 static struct sx ifdescr_sx;
124 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
125 
126 void	(*bstp_linkstate_p)(struct ifnet *ifp, int state);
127 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
128 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
129 
130 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
131 
132 /*
133  * XXX: Style; these should be sorted alphabetically, and unprototyped
134  * static functions should be prototyped. Currently they are sorted by
135  * declaration order.
136  */
137 static void	if_attachdomain(void *);
138 static void	if_attachdomain1(struct ifnet *);
139 static int	ifconf(u_long, caddr_t);
140 static void	if_freemulti(struct ifmultiaddr *);
141 static void	if_init(void *);
142 static void	if_grow(void);
143 static void	if_route(struct ifnet *, int flag, int fam);
144 static int	if_setflag(struct ifnet *, int, int, int *, int);
145 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
146 static void	if_unroute(struct ifnet *, int flag, int fam);
147 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
148 static int	if_rtdel(struct radix_node *, void *);
149 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
150 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
151 static void	do_link_state_change(void *, int);
152 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
153 static int	if_getgroupmembers(struct ifgroupreq *);
154 static void	if_delgroups(struct ifnet *);
155 static void	if_attach_internal(struct ifnet *, int);
156 static void	if_detach_internal(struct ifnet *, int);
157 
158 #ifdef INET6
159 /*
160  * XXX: declare here to avoid to include many inet6 related files..
161  * should be more generalized?
162  */
163 extern void	nd6_setmtu(struct ifnet *);
164 #endif
165 
166 VNET_DEFINE(int, if_index);
167 int	ifqmaxlen = IFQ_MAXLEN;
168 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
169 VNET_DEFINE(struct ifgrouphead, ifg_head);
170 
171 static VNET_DEFINE(int, if_indexlim) = 8;
172 
173 /* Table of ifnet by index. */
174 static VNET_DEFINE(struct ifindex_entry *, ifindex_table);
175 
176 #define	V_if_indexlim		VNET(if_indexlim)
177 #define	V_ifindex_table		VNET(ifindex_table)
178 
179 /*
180  * The global network interface list (V_ifnet) and related state (such as
181  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
182  * an rwlock.  Either may be acquired shared to stablize the list, but both
183  * must be acquired writable to modify the list.  This model allows us to
184  * both stablize the interface list during interrupt thread processing, but
185  * also to stablize it over long-running ioctls, without introducing priority
186  * inversions and deadlocks.
187  */
188 struct rwlock ifnet_rwlock;
189 struct sx ifnet_sxlock;
190 
191 /*
192  * The allocation of network interfaces is a rather non-atomic affair; we
193  * need to select an index before we are ready to expose the interface for
194  * use, so will use this pointer value to indicate reservation.
195  */
196 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
197 
198 static	if_com_alloc_t *if_com_alloc[256];
199 static	if_com_free_t *if_com_free[256];
200 
201 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
202 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
203 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
204 
205 struct ifnet *
206 ifnet_byindex_locked(u_short idx)
207 {
208 
209 	if (idx > V_if_index)
210 		return (NULL);
211 	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
212 		return (NULL);
213 	return (V_ifindex_table[idx].ife_ifnet);
214 }
215 
216 struct ifnet *
217 ifnet_byindex(u_short idx)
218 {
219 	struct ifnet *ifp;
220 
221 	IFNET_RLOCK_NOSLEEP();
222 	ifp = ifnet_byindex_locked(idx);
223 	IFNET_RUNLOCK_NOSLEEP();
224 	return (ifp);
225 }
226 
227 struct ifnet *
228 ifnet_byindex_ref(u_short idx)
229 {
230 	struct ifnet *ifp;
231 
232 	IFNET_RLOCK_NOSLEEP();
233 	ifp = ifnet_byindex_locked(idx);
234 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
235 		IFNET_RUNLOCK_NOSLEEP();
236 		return (NULL);
237 	}
238 	if_ref(ifp);
239 	IFNET_RUNLOCK_NOSLEEP();
240 	return (ifp);
241 }
242 
243 /*
244  * Allocate an ifindex array entry; return 0 on success or an error on
245  * failure.
246  */
247 static int
248 ifindex_alloc_locked(u_short *idxp)
249 {
250 	u_short idx;
251 
252 	IFNET_WLOCK_ASSERT();
253 
254 	/*
255 	 * Try to find an empty slot below V_if_index.  If we fail, take the
256 	 * next slot.
257 	 */
258 	for (idx = 1; idx <= V_if_index; idx++) {
259 		if (V_ifindex_table[idx].ife_ifnet == NULL)
260 			break;
261 	}
262 
263 	/* Catch if_index overflow. */
264 	if (idx < 1)
265 		return (ENOSPC);
266 	if (idx > V_if_index)
267 		V_if_index = idx;
268 	if (V_if_index >= V_if_indexlim)
269 		if_grow();
270 	*idxp = idx;
271 	return (0);
272 }
273 
274 static void
275 ifindex_free_locked(u_short idx)
276 {
277 
278 	IFNET_WLOCK_ASSERT();
279 
280 	V_ifindex_table[idx].ife_ifnet = NULL;
281 	while (V_if_index > 0 &&
282 	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
283 		V_if_index--;
284 }
285 
286 static void
287 ifindex_free(u_short idx)
288 {
289 
290 	IFNET_WLOCK();
291 	ifindex_free_locked(idx);
292 	IFNET_WUNLOCK();
293 }
294 
295 static void
296 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
297 {
298 
299 	IFNET_WLOCK_ASSERT();
300 
301 	V_ifindex_table[idx].ife_ifnet = ifp;
302 }
303 
304 static void
305 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
306 {
307 
308 	IFNET_WLOCK();
309 	ifnet_setbyindex_locked(idx, ifp);
310 	IFNET_WUNLOCK();
311 }
312 
313 struct ifaddr *
314 ifaddr_byindex(u_short idx)
315 {
316 	struct ifaddr *ifa;
317 
318 	IFNET_RLOCK_NOSLEEP();
319 	ifa = ifnet_byindex_locked(idx)->if_addr;
320 	if (ifa != NULL)
321 		ifa_ref(ifa);
322 	IFNET_RUNLOCK_NOSLEEP();
323 	return (ifa);
324 }
325 
326 /*
327  * Network interface utility routines.
328  *
329  * Routines with ifa_ifwith* names take sockaddr *'s as
330  * parameters.
331  */
332 
333 static void
334 vnet_if_init(const void *unused __unused)
335 {
336 
337 	TAILQ_INIT(&V_ifnet);
338 	TAILQ_INIT(&V_ifg_head);
339 	if_grow();				/* create initial table */
340 	vnet_if_clone_init();
341 }
342 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
343     NULL);
344 
345 /* ARGSUSED*/
346 static void
347 if_init(void *dummy __unused)
348 {
349 
350 	IFNET_LOCK_INIT();
351 	if_clone_init();
352 }
353 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
354 
355 
356 #ifdef VIMAGE
357 static void
358 vnet_if_uninit(const void *unused __unused)
359 {
360 
361 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
362 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
363 
364 	free((caddr_t)V_ifindex_table, M_IFNET);
365 }
366 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
367     vnet_if_uninit, NULL);
368 #endif
369 
370 static void
371 if_grow(void)
372 {
373 	u_int n;
374 	struct ifindex_entry *e;
375 
376 	V_if_indexlim <<= 1;
377 	n = V_if_indexlim * sizeof(*e);
378 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
379 	if (V_ifindex_table != NULL) {
380 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
381 		free((caddr_t)V_ifindex_table, M_IFNET);
382 	}
383 	V_ifindex_table = e;
384 }
385 
386 /*
387  * Allocate a struct ifnet and an index for an interface.  A layer 2
388  * common structure will also be allocated if an allocation routine is
389  * registered for the passed type.
390  */
391 struct ifnet *
392 if_alloc(u_char type)
393 {
394 	struct ifnet *ifp;
395 	u_short idx;
396 
397 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
398 	IFNET_WLOCK();
399 	if (ifindex_alloc_locked(&idx) != 0) {
400 		IFNET_WUNLOCK();
401 		free(ifp, M_IFNET);
402 		return (NULL);
403 	}
404 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
405 	IFNET_WUNLOCK();
406 	ifp->if_index = idx;
407 	ifp->if_type = type;
408 	ifp->if_alloctype = type;
409 	if (if_com_alloc[type] != NULL) {
410 		ifp->if_l2com = if_com_alloc[type](type, ifp);
411 		if (ifp->if_l2com == NULL) {
412 			free(ifp, M_IFNET);
413 			ifindex_free(idx);
414 			return (NULL);
415 		}
416 	}
417 
418 	IF_ADDR_LOCK_INIT(ifp);
419 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
420 	ifp->if_afdata_initialized = 0;
421 	IF_AFDATA_LOCK_INIT(ifp);
422 	TAILQ_INIT(&ifp->if_addrhead);
423 	TAILQ_INIT(&ifp->if_prefixhead);
424 	TAILQ_INIT(&ifp->if_multiaddrs);
425 	TAILQ_INIT(&ifp->if_groups);
426 #ifdef MAC
427 	mac_ifnet_init(ifp);
428 #endif
429 	ifq_init(&ifp->if_snd, ifp);
430 
431 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
432 	ifnet_setbyindex(ifp->if_index, ifp);
433 	return (ifp);
434 }
435 
436 /*
437  * Do the actual work of freeing a struct ifnet, associated index, and layer
438  * 2 common structure.  This call is made when the last reference to an
439  * interface is released.
440  */
441 static void
442 if_free_internal(struct ifnet *ifp)
443 {
444 
445 	KASSERT((ifp->if_flags & IFF_DYING),
446 	    ("if_free_internal: interface not dying"));
447 
448 	IFNET_WLOCK();
449 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
450 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
451 
452 	ifindex_free_locked(ifp->if_index);
453 	IFNET_WUNLOCK();
454 
455 	if (if_com_free[ifp->if_alloctype] != NULL)
456 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
457 		    ifp->if_alloctype);
458 
459 #ifdef MAC
460 	mac_ifnet_destroy(ifp);
461 #endif /* MAC */
462 	if (ifp->if_description != NULL)
463 		free(ifp->if_description, M_IFDESCR);
464 	IF_AFDATA_DESTROY(ifp);
465 	IF_ADDR_LOCK_DESTROY(ifp);
466 	ifq_delete(&ifp->if_snd);
467 	free(ifp, M_IFNET);
468 }
469 
470 /*
471  * This version should only be called by intefaces that switch their type
472  * after calling if_alloc().  if_free_type() will go away again now that we
473  * have if_alloctype to cache the original allocation type.  For now, assert
474  * that they match, since we require that in practice.
475  */
476 void
477 if_free_type(struct ifnet *ifp, u_char type)
478 {
479 
480 	KASSERT(ifp->if_alloctype == type,
481 	    ("if_free_type: type (%d) != alloctype (%d)", type,
482 	    ifp->if_alloctype));
483 
484 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
485 	if (!refcount_release(&ifp->if_refcount))
486 		return;
487 	if_free_internal(ifp);
488 }
489 
490 /*
491  * This is the normal version of if_free(), used by device drivers to free a
492  * detached network interface.  The contents of if_free_type() will move into
493  * here when if_free_type() goes away.
494  */
495 void
496 if_free(struct ifnet *ifp)
497 {
498 
499 	if_free_type(ifp, ifp->if_alloctype);
500 }
501 
502 /*
503  * Interfaces to keep an ifnet type-stable despite the possibility of the
504  * driver calling if_free().  If there are additional references, we defer
505  * freeing the underlying data structure.
506  */
507 void
508 if_ref(struct ifnet *ifp)
509 {
510 
511 	/* We don't assert the ifnet list lock here, but arguably should. */
512 	refcount_acquire(&ifp->if_refcount);
513 }
514 
515 void
516 if_rele(struct ifnet *ifp)
517 {
518 
519 	if (!refcount_release(&ifp->if_refcount))
520 		return;
521 	if_free_internal(ifp);
522 }
523 
524 void
525 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
526 {
527 
528 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
529 
530 	if (ifq->ifq_maxlen == 0)
531 		ifq->ifq_maxlen = ifqmaxlen;
532 
533 	ifq->altq_type = 0;
534 	ifq->altq_disc = NULL;
535 	ifq->altq_flags &= ALTQF_CANTCHANGE;
536 	ifq->altq_tbr  = NULL;
537 	ifq->altq_ifp  = ifp;
538 }
539 
540 void
541 ifq_delete(struct ifaltq *ifq)
542 {
543 	mtx_destroy(&ifq->ifq_mtx);
544 }
545 
546 /*
547  * Perform generic interface initalization tasks and attach the interface
548  * to the list of "active" interfaces.  If vmove flag is set on entry
549  * to if_attach_internal(), perform only a limited subset of initialization
550  * tasks, given that we are moving from one vnet to another an ifnet which
551  * has already been fully initialized.
552  *
553  * XXX:
554  *  - The decision to return void and thus require this function to
555  *    succeed is questionable.
556  *  - We should probably do more sanity checking.  For instance we don't
557  *    do anything to insure if_xname is unique or non-empty.
558  */
559 void
560 if_attach(struct ifnet *ifp)
561 {
562 
563 	if_attach_internal(ifp, 0);
564 }
565 
566 static void
567 if_attach_internal(struct ifnet *ifp, int vmove)
568 {
569 	unsigned socksize, ifasize;
570 	int namelen, masklen;
571 	struct sockaddr_dl *sdl;
572 	struct ifaddr *ifa;
573 
574 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
575 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
576 		    ifp->if_xname);
577 
578 #ifdef VIMAGE
579 	ifp->if_vnet = curvnet;
580 	if (ifp->if_home_vnet == NULL)
581 		ifp->if_home_vnet = curvnet;
582 #endif
583 
584 	if_addgroup(ifp, IFG_ALL);
585 
586 	getmicrotime(&ifp->if_lastchange);
587 	ifp->if_data.ifi_epoch = time_uptime;
588 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
589 
590 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
591 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
592 	    ("transmit and qflush must both either be set or both be NULL"));
593 	if (ifp->if_transmit == NULL) {
594 		ifp->if_transmit = if_transmit;
595 		ifp->if_qflush = if_qflush;
596 	}
597 
598 	if (!vmove) {
599 #ifdef MAC
600 		mac_ifnet_create(ifp);
601 #endif
602 
603 		/*
604 		 * Create a Link Level name for this device.
605 		 */
606 		namelen = strlen(ifp->if_xname);
607 		/*
608 		 * Always save enough space for any possiable name so we
609 		 * can do a rename in place later.
610 		 */
611 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
612 		socksize = masklen + ifp->if_addrlen;
613 		if (socksize < sizeof(*sdl))
614 			socksize = sizeof(*sdl);
615 		socksize = roundup2(socksize, sizeof(long));
616 		ifasize = sizeof(*ifa) + 2 * socksize;
617 		ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
618 		ifa_init(ifa);
619 		sdl = (struct sockaddr_dl *)(ifa + 1);
620 		sdl->sdl_len = socksize;
621 		sdl->sdl_family = AF_LINK;
622 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
623 		sdl->sdl_nlen = namelen;
624 		sdl->sdl_index = ifp->if_index;
625 		sdl->sdl_type = ifp->if_type;
626 		ifp->if_addr = ifa;
627 		ifa->ifa_ifp = ifp;
628 		ifa->ifa_rtrequest = link_rtrequest;
629 		ifa->ifa_addr = (struct sockaddr *)sdl;
630 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
631 		ifa->ifa_netmask = (struct sockaddr *)sdl;
632 		sdl->sdl_len = masklen;
633 		while (namelen != 0)
634 			sdl->sdl_data[--namelen] = 0xff;
635 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
636 		/* Reliably crash if used uninitialized. */
637 		ifp->if_broadcastaddr = NULL;
638 	}
639 #ifdef VIMAGE
640 	else {
641 		/*
642 		 * Update the interface index in the link layer address
643 		 * of the interface.
644 		 */
645 		for (ifa = ifp->if_addr; ifa != NULL;
646 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
647 			if (ifa->ifa_addr->sa_family == AF_LINK) {
648 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
649 				sdl->sdl_index = ifp->if_index;
650 			}
651 		}
652 	}
653 #endif
654 
655 	IFNET_WLOCK();
656 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
657 #ifdef VIMAGE
658 	curvnet->vnet_ifcnt++;
659 #endif
660 	IFNET_WUNLOCK();
661 
662 	if (domain_init_status >= 2)
663 		if_attachdomain1(ifp);
664 
665 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
666 	if (IS_DEFAULT_VNET(curvnet))
667 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
668 
669 	/* Announce the interface. */
670 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
671 }
672 
673 static void
674 if_attachdomain(void *dummy)
675 {
676 	struct ifnet *ifp;
677 	int s;
678 
679 	s = splnet();
680 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
681 		if_attachdomain1(ifp);
682 	splx(s);
683 }
684 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
685     if_attachdomain, NULL);
686 
687 static void
688 if_attachdomain1(struct ifnet *ifp)
689 {
690 	struct domain *dp;
691 	int s;
692 
693 	s = splnet();
694 
695 	/*
696 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
697 	 * cannot lock ifp->if_afdata initialization, entirely.
698 	 */
699 	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
700 		splx(s);
701 		return;
702 	}
703 	if (ifp->if_afdata_initialized >= domain_init_status) {
704 		IF_AFDATA_UNLOCK(ifp);
705 		splx(s);
706 		printf("if_attachdomain called more than once on %s\n",
707 		    ifp->if_xname);
708 		return;
709 	}
710 	ifp->if_afdata_initialized = domain_init_status;
711 	IF_AFDATA_UNLOCK(ifp);
712 
713 	/* address family dependent data region */
714 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
715 	for (dp = domains; dp; dp = dp->dom_next) {
716 		if (dp->dom_ifattach)
717 			ifp->if_afdata[dp->dom_family] =
718 			    (*dp->dom_ifattach)(ifp);
719 	}
720 
721 	splx(s);
722 }
723 
724 /*
725  * Remove any unicast or broadcast network addresses from an interface.
726  */
727 void
728 if_purgeaddrs(struct ifnet *ifp)
729 {
730 	struct ifaddr *ifa, *next;
731 
732 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
733 		if (ifa->ifa_addr->sa_family == AF_LINK)
734 			continue;
735 #ifdef INET
736 		/* XXX: Ugly!! ad hoc just for INET */
737 		if (ifa->ifa_addr->sa_family == AF_INET) {
738 			struct ifaliasreq ifr;
739 
740 			bzero(&ifr, sizeof(ifr));
741 			ifr.ifra_addr = *ifa->ifa_addr;
742 			if (ifa->ifa_dstaddr)
743 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
744 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
745 			    NULL) == 0)
746 				continue;
747 		}
748 #endif /* INET */
749 #ifdef INET6
750 		if (ifa->ifa_addr->sa_family == AF_INET6) {
751 			in6_purgeaddr(ifa);
752 			/* ifp_addrhead is already updated */
753 			continue;
754 		}
755 #endif /* INET6 */
756 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
757 		ifa_free(ifa);
758 	}
759 }
760 
761 /*
762  * Remove any multicast network addresses from an interface when an ifnet
763  * is going away.
764  */
765 static void
766 if_purgemaddrs(struct ifnet *ifp)
767 {
768 	struct ifmultiaddr *ifma;
769 	struct ifmultiaddr *next;
770 
771 	IF_ADDR_LOCK(ifp);
772 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
773 		if_delmulti_locked(ifp, ifma, 1);
774 	IF_ADDR_UNLOCK(ifp);
775 }
776 
777 /*
778  * Detach an interface, removing it from the list of "active" interfaces.
779  * If vmove flag is set on entry to if_detach_internal(), perform only a
780  * limited subset of cleanup tasks, given that we are moving an ifnet from
781  * one vnet to another, where it must be fully operational.
782  *
783  * XXXRW: There are some significant questions about event ordering, and
784  * how to prevent things from starting to use the interface during detach.
785  */
786 void
787 if_detach(struct ifnet *ifp)
788 {
789 
790 	if_detach_internal(ifp, 0);
791 }
792 
793 static void
794 if_detach_internal(struct ifnet *ifp, int vmove)
795 {
796 	struct ifaddr *ifa;
797 	struct radix_node_head	*rnh;
798 	int i, j;
799 	struct domain *dp;
800  	struct ifnet *iter;
801  	int found = 0;
802 
803 	IFNET_WLOCK();
804 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
805 		if (iter == ifp) {
806 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
807 			found = 1;
808 			break;
809 		}
810 #ifdef VIMAGE
811 	if (found)
812 		curvnet->vnet_ifcnt--;
813 #endif
814 	IFNET_WUNLOCK();
815 	if (!found) {
816 		if (vmove)
817 			panic("%s: ifp=%p not on the ifnet tailq %p",
818 			    __func__, ifp, &V_ifnet);
819 		else
820 			return; /* XXX this should panic as well? */
821 	}
822 
823 	/*
824 	 * Remove/wait for pending events.
825 	 */
826 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
827 
828 	/*
829 	 * Remove routes and flush queues.
830 	 */
831 	if_down(ifp);
832 #ifdef ALTQ
833 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
834 		altq_disable(&ifp->if_snd);
835 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
836 		altq_detach(&ifp->if_snd);
837 #endif
838 
839 	if_purgeaddrs(ifp);
840 
841 #ifdef INET
842 	in_ifdetach(ifp);
843 #endif
844 
845 #ifdef INET6
846 	/*
847 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
848 	 * before removing routing entries below, since IPv6 interface direct
849 	 * routes are expected to be removed by the IPv6-specific kernel API.
850 	 * Otherwise, the kernel will detect some inconsistency and bark it.
851 	 */
852 	in6_ifdetach(ifp);
853 #endif
854 	if_purgemaddrs(ifp);
855 
856 	if (!vmove) {
857 		/*
858 		 * Prevent further calls into the device driver via ifnet.
859 		 */
860 		if_dead(ifp);
861 
862 		/*
863 		 * Remove link ifaddr pointer and maybe decrement if_index.
864 		 * Clean up all addresses.
865 		 */
866 		ifp->if_addr = NULL;
867 
868 		/* We can now free link ifaddr. */
869 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
870 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
871 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
872 			ifa_free(ifa);
873 		}
874 	}
875 
876 	/*
877 	 * Delete all remaining routes using this interface
878 	 * Unfortuneatly the only way to do this is to slog through
879 	 * the entire routing table looking for routes which point
880 	 * to this interface...oh well...
881 	 */
882 	for (i = 1; i <= AF_MAX; i++) {
883 		for (j = 0; j < rt_numfibs; j++) {
884 			rnh = rt_tables_get_rnh(j, i);
885 			if (rnh == NULL)
886 				continue;
887 			RADIX_NODE_HEAD_LOCK(rnh);
888 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
889 			RADIX_NODE_HEAD_UNLOCK(rnh);
890 		}
891 	}
892 
893 	/* Announce that the interface is gone. */
894 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
895 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
896 	if (IS_DEFAULT_VNET(curvnet))
897 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
898 	if_delgroups(ifp);
899 
900 	/*
901 	 * We cannot hold the lock over dom_ifdetach calls as they might
902 	 * sleep, for example trying to drain a callout, thus open up the
903 	 * theoretical race with re-attaching.
904 	 */
905 	IF_AFDATA_LOCK(ifp);
906 	i = ifp->if_afdata_initialized;
907 	ifp->if_afdata_initialized = 0;
908 	IF_AFDATA_UNLOCK(ifp);
909 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
910 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
911 			(*dp->dom_ifdetach)(ifp,
912 			    ifp->if_afdata[dp->dom_family]);
913 	}
914 }
915 
916 #ifdef VIMAGE
917 /*
918  * if_vmove() performs a limited version of if_detach() in current
919  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
920  * An attempt is made to shrink if_index in current vnet, find an
921  * unused if_index in target vnet and calls if_grow() if necessary,
922  * and finally find an unused if_xname for the target vnet.
923  */
924 void
925 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
926 {
927 	u_short idx;
928 
929 	/*
930 	 * Detach from current vnet, but preserve LLADDR info, do not
931 	 * mark as dead etc. so that the ifnet can be reattached later.
932 	 */
933 	if_detach_internal(ifp, 1);
934 
935 	/*
936 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
937 	 * the if_index for that vnet if possible.
938 	 *
939 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
940 	 * or we'd lock on one vnet and unlock on another.
941 	 */
942 	IFNET_WLOCK();
943 	ifindex_free_locked(ifp->if_index);
944 
945 	/*
946 	 * Switch to the context of the target vnet.
947 	 */
948 	CURVNET_SET_QUIET(new_vnet);
949 
950 	if (ifindex_alloc_locked(&idx) != 0) {
951 		IFNET_WUNLOCK();
952 		panic("if_index overflow");
953 	}
954 	ifp->if_index = idx;
955 	ifnet_setbyindex_locked(ifp->if_index, ifp);
956 	IFNET_WUNLOCK();
957 
958 	if_attach_internal(ifp, 1);
959 
960 	CURVNET_RESTORE();
961 }
962 
963 /*
964  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
965  */
966 static int
967 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
968 {
969 	struct prison *pr;
970 	struct ifnet *difp;
971 
972 	/* Try to find the prison within our visibility. */
973 	sx_slock(&allprison_lock);
974 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
975 	sx_sunlock(&allprison_lock);
976 	if (pr == NULL)
977 		return (ENXIO);
978 	prison_hold_locked(pr);
979 	mtx_unlock(&pr->pr_mtx);
980 
981 	/* Do not try to move the iface from and to the same prison. */
982 	if (pr->pr_vnet == ifp->if_vnet) {
983 		prison_free(pr);
984 		return (EEXIST);
985 	}
986 
987 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
988 	/* XXX Lock interfaces to avoid races. */
989 	CURVNET_SET_QUIET(pr->pr_vnet);
990 	difp = ifunit(ifname);
991 	CURVNET_RESTORE();
992 	if (difp != NULL) {
993 		prison_free(pr);
994 		return (EEXIST);
995 	}
996 
997 	/* Move the interface into the child jail/vnet. */
998 	if_vmove(ifp, pr->pr_vnet);
999 
1000 	/* Report the new if_xname back to the userland. */
1001 	sprintf(ifname, "%s", ifp->if_xname);
1002 
1003 	prison_free(pr);
1004 	return (0);
1005 }
1006 
1007 static int
1008 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1009 {
1010 	struct prison *pr;
1011 	struct vnet *vnet_dst;
1012 	struct ifnet *ifp;
1013 
1014 	/* Try to find the prison within our visibility. */
1015 	sx_slock(&allprison_lock);
1016 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1017 	sx_sunlock(&allprison_lock);
1018 	if (pr == NULL)
1019 		return (ENXIO);
1020 	prison_hold_locked(pr);
1021 	mtx_unlock(&pr->pr_mtx);
1022 
1023 	/* Make sure the named iface exists in the source prison/vnet. */
1024 	CURVNET_SET(pr->pr_vnet);
1025 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1026 	if (ifp == NULL) {
1027 		CURVNET_RESTORE();
1028 		prison_free(pr);
1029 		return (ENXIO);
1030 	}
1031 
1032 	/* Do not try to move the iface from and to the same prison. */
1033 	vnet_dst = TD_TO_VNET(td);
1034 	if (vnet_dst == ifp->if_vnet) {
1035 		CURVNET_RESTORE();
1036 		prison_free(pr);
1037 		return (EEXIST);
1038 	}
1039 
1040 	/* Get interface back from child jail/vnet. */
1041 	if_vmove(ifp, vnet_dst);
1042 	CURVNET_RESTORE();
1043 
1044 	/* Report the new if_xname back to the userland. */
1045 	sprintf(ifname, "%s", ifp->if_xname);
1046 
1047 	prison_free(pr);
1048 	return (0);
1049 }
1050 #endif /* VIMAGE */
1051 
1052 /*
1053  * Add a group to an interface
1054  */
1055 int
1056 if_addgroup(struct ifnet *ifp, const char *groupname)
1057 {
1058 	struct ifg_list		*ifgl;
1059 	struct ifg_group	*ifg = NULL;
1060 	struct ifg_member	*ifgm;
1061 
1062 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1063 	    groupname[strlen(groupname) - 1] <= '9')
1064 		return (EINVAL);
1065 
1066 	IFNET_WLOCK();
1067 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1068 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1069 			IFNET_WUNLOCK();
1070 			return (EEXIST);
1071 		}
1072 
1073 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1074 	    M_NOWAIT)) == NULL) {
1075 	    	IFNET_WUNLOCK();
1076 		return (ENOMEM);
1077 	}
1078 
1079 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1080 	    M_TEMP, M_NOWAIT)) == NULL) {
1081 		free(ifgl, M_TEMP);
1082 		IFNET_WUNLOCK();
1083 		return (ENOMEM);
1084 	}
1085 
1086 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1087 		if (!strcmp(ifg->ifg_group, groupname))
1088 			break;
1089 
1090 	if (ifg == NULL) {
1091 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1092 		    M_TEMP, M_NOWAIT)) == NULL) {
1093 			free(ifgl, M_TEMP);
1094 			free(ifgm, M_TEMP);
1095 			IFNET_WUNLOCK();
1096 			return (ENOMEM);
1097 		}
1098 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1099 		ifg->ifg_refcnt = 0;
1100 		TAILQ_INIT(&ifg->ifg_members);
1101 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1102 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1103 	}
1104 
1105 	ifg->ifg_refcnt++;
1106 	ifgl->ifgl_group = ifg;
1107 	ifgm->ifgm_ifp = ifp;
1108 
1109 	IF_ADDR_LOCK(ifp);
1110 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1111 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1112 	IF_ADDR_UNLOCK(ifp);
1113 
1114 	IFNET_WUNLOCK();
1115 
1116 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1117 
1118 	return (0);
1119 }
1120 
1121 /*
1122  * Remove a group from an interface
1123  */
1124 int
1125 if_delgroup(struct ifnet *ifp, const char *groupname)
1126 {
1127 	struct ifg_list		*ifgl;
1128 	struct ifg_member	*ifgm;
1129 
1130 	IFNET_WLOCK();
1131 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1132 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1133 			break;
1134 	if (ifgl == NULL) {
1135 		IFNET_WUNLOCK();
1136 		return (ENOENT);
1137 	}
1138 
1139 	IF_ADDR_LOCK(ifp);
1140 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1141 	IF_ADDR_UNLOCK(ifp);
1142 
1143 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1144 		if (ifgm->ifgm_ifp == ifp)
1145 			break;
1146 
1147 	if (ifgm != NULL) {
1148 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1149 		free(ifgm, M_TEMP);
1150 	}
1151 
1152 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1153 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1154 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1155 		free(ifgl->ifgl_group, M_TEMP);
1156 	}
1157 	IFNET_WUNLOCK();
1158 
1159 	free(ifgl, M_TEMP);
1160 
1161 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1162 
1163 	return (0);
1164 }
1165 
1166 /*
1167  * Remove an interface from all groups
1168  */
1169 static void
1170 if_delgroups(struct ifnet *ifp)
1171 {
1172 	struct ifg_list		*ifgl;
1173 	struct ifg_member	*ifgm;
1174 	char groupname[IFNAMSIZ];
1175 
1176 	IFNET_WLOCK();
1177 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1178 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1179 
1180 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1181 
1182 		IF_ADDR_LOCK(ifp);
1183 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1184 		IF_ADDR_UNLOCK(ifp);
1185 
1186 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1187 			if (ifgm->ifgm_ifp == ifp)
1188 				break;
1189 
1190 		if (ifgm != NULL) {
1191 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1192 			    ifgm_next);
1193 			free(ifgm, M_TEMP);
1194 		}
1195 
1196 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1197 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1198 			EVENTHANDLER_INVOKE(group_detach_event,
1199 			    ifgl->ifgl_group);
1200 			free(ifgl->ifgl_group, M_TEMP);
1201 		}
1202 		IFNET_WUNLOCK();
1203 
1204 		free(ifgl, M_TEMP);
1205 
1206 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1207 
1208 		IFNET_WLOCK();
1209 	}
1210 	IFNET_WUNLOCK();
1211 }
1212 
1213 /*
1214  * Stores all groups from an interface in memory pointed
1215  * to by data
1216  */
1217 static int
1218 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1219 {
1220 	int			 len, error;
1221 	struct ifg_list		*ifgl;
1222 	struct ifg_req		 ifgrq, *ifgp;
1223 	struct ifgroupreq	*ifgr = data;
1224 
1225 	if (ifgr->ifgr_len == 0) {
1226 		IF_ADDR_LOCK(ifp);
1227 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1228 			ifgr->ifgr_len += sizeof(struct ifg_req);
1229 		IF_ADDR_UNLOCK(ifp);
1230 		return (0);
1231 	}
1232 
1233 	len = ifgr->ifgr_len;
1234 	ifgp = ifgr->ifgr_groups;
1235 	/* XXX: wire */
1236 	IF_ADDR_LOCK(ifp);
1237 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1238 		if (len < sizeof(ifgrq)) {
1239 			IF_ADDR_UNLOCK(ifp);
1240 			return (EINVAL);
1241 		}
1242 		bzero(&ifgrq, sizeof ifgrq);
1243 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1244 		    sizeof(ifgrq.ifgrq_group));
1245 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1246 		    	IF_ADDR_UNLOCK(ifp);
1247 			return (error);
1248 		}
1249 		len -= sizeof(ifgrq);
1250 		ifgp++;
1251 	}
1252 	IF_ADDR_UNLOCK(ifp);
1253 
1254 	return (0);
1255 }
1256 
1257 /*
1258  * Stores all members of a group in memory pointed to by data
1259  */
1260 static int
1261 if_getgroupmembers(struct ifgroupreq *data)
1262 {
1263 	struct ifgroupreq	*ifgr = data;
1264 	struct ifg_group	*ifg;
1265 	struct ifg_member	*ifgm;
1266 	struct ifg_req		 ifgrq, *ifgp;
1267 	int			 len, error;
1268 
1269 	IFNET_RLOCK();
1270 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1271 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1272 			break;
1273 	if (ifg == NULL) {
1274 		IFNET_RUNLOCK();
1275 		return (ENOENT);
1276 	}
1277 
1278 	if (ifgr->ifgr_len == 0) {
1279 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1280 			ifgr->ifgr_len += sizeof(ifgrq);
1281 		IFNET_RUNLOCK();
1282 		return (0);
1283 	}
1284 
1285 	len = ifgr->ifgr_len;
1286 	ifgp = ifgr->ifgr_groups;
1287 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1288 		if (len < sizeof(ifgrq)) {
1289 			IFNET_RUNLOCK();
1290 			return (EINVAL);
1291 		}
1292 		bzero(&ifgrq, sizeof ifgrq);
1293 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1294 		    sizeof(ifgrq.ifgrq_member));
1295 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1296 			IFNET_RUNLOCK();
1297 			return (error);
1298 		}
1299 		len -= sizeof(ifgrq);
1300 		ifgp++;
1301 	}
1302 	IFNET_RUNLOCK();
1303 
1304 	return (0);
1305 }
1306 
1307 /*
1308  * Delete Routes for a Network Interface
1309  *
1310  * Called for each routing entry via the rnh->rnh_walktree() call above
1311  * to delete all route entries referencing a detaching network interface.
1312  *
1313  * Arguments:
1314  *	rn	pointer to node in the routing table
1315  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1316  *
1317  * Returns:
1318  *	0	successful
1319  *	errno	failed - reason indicated
1320  *
1321  */
1322 static int
1323 if_rtdel(struct radix_node *rn, void *arg)
1324 {
1325 	struct rtentry	*rt = (struct rtentry *)rn;
1326 	struct ifnet	*ifp = arg;
1327 	int		err;
1328 
1329 	if (rt->rt_ifp == ifp) {
1330 
1331 		/*
1332 		 * Protect (sorta) against walktree recursion problems
1333 		 * with cloned routes
1334 		 */
1335 		if ((rt->rt_flags & RTF_UP) == 0)
1336 			return (0);
1337 
1338 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1339 				rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
1340 				(struct rtentry **) NULL, rt->rt_fibnum);
1341 		if (err) {
1342 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1343 		}
1344 	}
1345 
1346 	return (0);
1347 }
1348 
1349 /*
1350  * Wrapper functions for struct ifnet address list locking macros.  These are
1351  * used by kernel modules to avoid encoding programming interface or binary
1352  * interface assumptions that may be violated when kernel-internal locking
1353  * approaches change.
1354  */
1355 void
1356 if_addr_rlock(struct ifnet *ifp)
1357 {
1358 
1359 	IF_ADDR_LOCK(ifp);
1360 }
1361 
1362 void
1363 if_addr_runlock(struct ifnet *ifp)
1364 {
1365 
1366 	IF_ADDR_UNLOCK(ifp);
1367 }
1368 
1369 void
1370 if_maddr_rlock(struct ifnet *ifp)
1371 {
1372 
1373 	IF_ADDR_LOCK(ifp);
1374 }
1375 
1376 void
1377 if_maddr_runlock(struct ifnet *ifp)
1378 {
1379 
1380 	IF_ADDR_UNLOCK(ifp);
1381 }
1382 
1383 /*
1384  * Reference count functions for ifaddrs.
1385  */
1386 void
1387 ifa_init(struct ifaddr *ifa)
1388 {
1389 
1390 	mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1391 	refcount_init(&ifa->ifa_refcnt, 1);
1392 }
1393 
1394 void
1395 ifa_ref(struct ifaddr *ifa)
1396 {
1397 
1398 	refcount_acquire(&ifa->ifa_refcnt);
1399 }
1400 
1401 void
1402 ifa_free(struct ifaddr *ifa)
1403 {
1404 
1405 	if (refcount_release(&ifa->ifa_refcnt)) {
1406 		mtx_destroy(&ifa->ifa_mtx);
1407 		free(ifa, M_IFADDR);
1408 	}
1409 }
1410 
1411 int
1412 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1413 {
1414 	int error = 0;
1415 	struct rtentry *rt = NULL;
1416 	struct rt_addrinfo info;
1417 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1418 
1419 	bzero(&info, sizeof(info));
1420 	info.rti_ifp = V_loif;
1421 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1422 	info.rti_info[RTAX_DST] = ia;
1423 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1424 	error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1425 
1426 	if (error == 0 && rt != NULL) {
1427 		RT_LOCK(rt);
1428 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1429 			ifa->ifa_ifp->if_type;
1430 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1431 			ifa->ifa_ifp->if_index;
1432 		RT_REMREF(rt);
1433 		RT_UNLOCK(rt);
1434 	} else if (error != 0)
1435 		log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1436 
1437 	return (error);
1438 }
1439 
1440 int
1441 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1442 {
1443 	int error = 0;
1444 	struct rt_addrinfo info;
1445 	struct sockaddr_dl null_sdl;
1446 
1447 	bzero(&null_sdl, sizeof(null_sdl));
1448 	null_sdl.sdl_len = sizeof(null_sdl);
1449 	null_sdl.sdl_family = AF_LINK;
1450 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1451 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1452 	bzero(&info, sizeof(info));
1453 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1454 	info.rti_info[RTAX_DST] = ia;
1455 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1456 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1457 
1458 	if (error != 0)
1459 		log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1460 
1461 	return (error);
1462 }
1463 
1464 /*
1465  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1466  * structs used to represent other address families, it is necessary
1467  * to perform a different comparison.
1468  */
1469 
1470 #define	sa_equal(a1, a2)	\
1471 	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1472 
1473 #define	sa_dl_equal(a1, a2)	\
1474 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1475 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1476 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1477 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1478 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1479 
1480 /*
1481  * Locate an interface based on a complete address.
1482  */
1483 /*ARGSUSED*/
1484 static struct ifaddr *
1485 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1486 {
1487 	struct ifnet *ifp;
1488 	struct ifaddr *ifa;
1489 
1490 	IFNET_RLOCK_NOSLEEP();
1491 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1492 		IF_ADDR_LOCK(ifp);
1493 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1494 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1495 				continue;
1496 			if (sa_equal(addr, ifa->ifa_addr)) {
1497 				if (getref)
1498 					ifa_ref(ifa);
1499 				IF_ADDR_UNLOCK(ifp);
1500 				goto done;
1501 			}
1502 			/* IP6 doesn't have broadcast */
1503 			if ((ifp->if_flags & IFF_BROADCAST) &&
1504 			    ifa->ifa_broadaddr &&
1505 			    ifa->ifa_broadaddr->sa_len != 0 &&
1506 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1507 				if (getref)
1508 					ifa_ref(ifa);
1509 				IF_ADDR_UNLOCK(ifp);
1510 				goto done;
1511 			}
1512 		}
1513 		IF_ADDR_UNLOCK(ifp);
1514 	}
1515 	ifa = NULL;
1516 done:
1517 	IFNET_RUNLOCK_NOSLEEP();
1518 	return (ifa);
1519 }
1520 
1521 struct ifaddr *
1522 ifa_ifwithaddr(struct sockaddr *addr)
1523 {
1524 
1525 	return (ifa_ifwithaddr_internal(addr, 1));
1526 }
1527 
1528 int
1529 ifa_ifwithaddr_check(struct sockaddr *addr)
1530 {
1531 
1532 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1533 }
1534 
1535 /*
1536  * Locate an interface based on the broadcast address.
1537  */
1538 /* ARGSUSED */
1539 struct ifaddr *
1540 ifa_ifwithbroadaddr(struct sockaddr *addr)
1541 {
1542 	struct ifnet *ifp;
1543 	struct ifaddr *ifa;
1544 
1545 	IFNET_RLOCK_NOSLEEP();
1546 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1547 		IF_ADDR_LOCK(ifp);
1548 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1549 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1550 				continue;
1551 			if ((ifp->if_flags & IFF_BROADCAST) &&
1552 			    ifa->ifa_broadaddr &&
1553 			    ifa->ifa_broadaddr->sa_len != 0 &&
1554 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1555 				ifa_ref(ifa);
1556 				IF_ADDR_UNLOCK(ifp);
1557 				goto done;
1558 			}
1559 		}
1560 		IF_ADDR_UNLOCK(ifp);
1561 	}
1562 	ifa = NULL;
1563 done:
1564 	IFNET_RUNLOCK_NOSLEEP();
1565 	return (ifa);
1566 }
1567 
1568 /*
1569  * Locate the point to point interface with a given destination address.
1570  */
1571 /*ARGSUSED*/
1572 struct ifaddr *
1573 ifa_ifwithdstaddr(struct sockaddr *addr)
1574 {
1575 	struct ifnet *ifp;
1576 	struct ifaddr *ifa;
1577 
1578 	IFNET_RLOCK_NOSLEEP();
1579 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1580 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1581 			continue;
1582 		IF_ADDR_LOCK(ifp);
1583 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1584 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1585 				continue;
1586 			if (ifa->ifa_dstaddr != NULL &&
1587 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1588 				ifa_ref(ifa);
1589 				IF_ADDR_UNLOCK(ifp);
1590 				goto done;
1591 			}
1592 		}
1593 		IF_ADDR_UNLOCK(ifp);
1594 	}
1595 	ifa = NULL;
1596 done:
1597 	IFNET_RUNLOCK_NOSLEEP();
1598 	return (ifa);
1599 }
1600 
1601 /*
1602  * Find an interface on a specific network.  If many, choice
1603  * is most specific found.
1604  */
1605 struct ifaddr *
1606 ifa_ifwithnet(struct sockaddr *addr)
1607 {
1608 	struct ifnet *ifp;
1609 	struct ifaddr *ifa;
1610 	struct ifaddr *ifa_maybe = NULL;
1611 	u_int af = addr->sa_family;
1612 	char *addr_data = addr->sa_data, *cplim;
1613 
1614 	/*
1615 	 * AF_LINK addresses can be looked up directly by their index number,
1616 	 * so do that if we can.
1617 	 */
1618 	if (af == AF_LINK) {
1619 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1620 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1621 		return (ifaddr_byindex(sdl->sdl_index));
1622 	}
1623 
1624 	/*
1625 	 * Scan though each interface, looking for ones that have addresses
1626 	 * in this address family.  Maintain a reference on ifa_maybe once
1627 	 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1628 	 * when we move onto the next interface.
1629 	 */
1630 	IFNET_RLOCK_NOSLEEP();
1631 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1632 		IF_ADDR_LOCK(ifp);
1633 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1634 			char *cp, *cp2, *cp3;
1635 
1636 			if (ifa->ifa_addr->sa_family != af)
1637 next:				continue;
1638 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
1639 				/*
1640 				 * This is a bit broken as it doesn't
1641 				 * take into account that the remote end may
1642 				 * be a single node in the network we are
1643 				 * looking for.
1644 				 * The trouble is that we don't know the
1645 				 * netmask for the remote end.
1646 				 */
1647 				if (ifa->ifa_dstaddr != NULL &&
1648 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1649 					ifa_ref(ifa);
1650 					IF_ADDR_UNLOCK(ifp);
1651 					goto done;
1652 				}
1653 			} else {
1654 				/*
1655 				 * if we have a special address handler,
1656 				 * then use it instead of the generic one.
1657 				 */
1658 				if (ifa->ifa_claim_addr) {
1659 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1660 						ifa_ref(ifa);
1661 						IF_ADDR_UNLOCK(ifp);
1662 						goto done;
1663 					}
1664 					continue;
1665 				}
1666 
1667 				/*
1668 				 * Scan all the bits in the ifa's address.
1669 				 * If a bit dissagrees with what we are
1670 				 * looking for, mask it with the netmask
1671 				 * to see if it really matters.
1672 				 * (A byte at a time)
1673 				 */
1674 				if (ifa->ifa_netmask == 0)
1675 					continue;
1676 				cp = addr_data;
1677 				cp2 = ifa->ifa_addr->sa_data;
1678 				cp3 = ifa->ifa_netmask->sa_data;
1679 				cplim = ifa->ifa_netmask->sa_len
1680 					+ (char *)ifa->ifa_netmask;
1681 				while (cp3 < cplim)
1682 					if ((*cp++ ^ *cp2++) & *cp3++)
1683 						goto next; /* next address! */
1684 				/*
1685 				 * If the netmask of what we just found
1686 				 * is more specific than what we had before
1687 				 * (if we had one) then remember the new one
1688 				 * before continuing to search
1689 				 * for an even better one.
1690 				 */
1691 				if (ifa_maybe == NULL ||
1692 				    rn_refines((caddr_t)ifa->ifa_netmask,
1693 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1694 					if (ifa_maybe != NULL)
1695 						ifa_free(ifa_maybe);
1696 					ifa_maybe = ifa;
1697 					ifa_ref(ifa_maybe);
1698 				}
1699 			}
1700 		}
1701 		IF_ADDR_UNLOCK(ifp);
1702 	}
1703 	ifa = ifa_maybe;
1704 	ifa_maybe = NULL;
1705 done:
1706 	IFNET_RUNLOCK_NOSLEEP();
1707 	if (ifa_maybe != NULL)
1708 		ifa_free(ifa_maybe);
1709 	return (ifa);
1710 }
1711 
1712 /*
1713  * Find an interface address specific to an interface best matching
1714  * a given address.
1715  */
1716 struct ifaddr *
1717 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1718 {
1719 	struct ifaddr *ifa;
1720 	char *cp, *cp2, *cp3;
1721 	char *cplim;
1722 	struct ifaddr *ifa_maybe = NULL;
1723 	u_int af = addr->sa_family;
1724 
1725 	if (af >= AF_MAX)
1726 		return (0);
1727 	IF_ADDR_LOCK(ifp);
1728 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1729 		if (ifa->ifa_addr->sa_family != af)
1730 			continue;
1731 		if (ifa_maybe == NULL)
1732 			ifa_maybe = ifa;
1733 		if (ifa->ifa_netmask == 0) {
1734 			if (sa_equal(addr, ifa->ifa_addr) ||
1735 			    (ifa->ifa_dstaddr &&
1736 			    sa_equal(addr, ifa->ifa_dstaddr)))
1737 				goto done;
1738 			continue;
1739 		}
1740 		if (ifp->if_flags & IFF_POINTOPOINT) {
1741 			if (sa_equal(addr, ifa->ifa_dstaddr))
1742 				goto done;
1743 		} else {
1744 			cp = addr->sa_data;
1745 			cp2 = ifa->ifa_addr->sa_data;
1746 			cp3 = ifa->ifa_netmask->sa_data;
1747 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1748 			for (; cp3 < cplim; cp3++)
1749 				if ((*cp++ ^ *cp2++) & *cp3)
1750 					break;
1751 			if (cp3 == cplim)
1752 				goto done;
1753 		}
1754 	}
1755 	ifa = ifa_maybe;
1756 done:
1757 	if (ifa != NULL)
1758 		ifa_ref(ifa);
1759 	IF_ADDR_UNLOCK(ifp);
1760 	return (ifa);
1761 }
1762 
1763 #include <net/if_llatbl.h>
1764 
1765 /*
1766  * Default action when installing a route with a Link Level gateway.
1767  * Lookup an appropriate real ifa to point to.
1768  * This should be moved to /sys/net/link.c eventually.
1769  */
1770 static void
1771 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1772 {
1773 	struct ifaddr *ifa, *oifa;
1774 	struct sockaddr *dst;
1775 	struct ifnet *ifp;
1776 
1777 	RT_LOCK_ASSERT(rt);
1778 
1779 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1780 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1781 		return;
1782 	ifa = ifaof_ifpforaddr(dst, ifp);
1783 	if (ifa) {
1784 		oifa = rt->rt_ifa;
1785 		rt->rt_ifa = ifa;
1786 		ifa_free(oifa);
1787 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1788 			ifa->ifa_rtrequest(cmd, rt, info);
1789 	}
1790 }
1791 
1792 /*
1793  * Mark an interface down and notify protocols of
1794  * the transition.
1795  * NOTE: must be called at splnet or eqivalent.
1796  */
1797 static void
1798 if_unroute(struct ifnet *ifp, int flag, int fam)
1799 {
1800 	struct ifaddr *ifa;
1801 
1802 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1803 
1804 	ifp->if_flags &= ~flag;
1805 	getmicrotime(&ifp->if_lastchange);
1806 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1807 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1808 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1809 	ifp->if_qflush(ifp);
1810 
1811 #if defined(INET) || defined(INET6)
1812 #ifdef DEV_CARP
1813 	if (ifp->if_carp)
1814 		carp_carpdev_state(ifp->if_carp);
1815 #endif
1816 #endif
1817 	rt_ifmsg(ifp);
1818 }
1819 
1820 /*
1821  * Mark an interface up and notify protocols of
1822  * the transition.
1823  * NOTE: must be called at splnet or eqivalent.
1824  */
1825 static void
1826 if_route(struct ifnet *ifp, int flag, int fam)
1827 {
1828 	struct ifaddr *ifa;
1829 
1830 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1831 
1832 	ifp->if_flags |= flag;
1833 	getmicrotime(&ifp->if_lastchange);
1834 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1835 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1836 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1837 #if defined(INET) || defined(INET6)
1838 #ifdef DEV_CARP
1839 	if (ifp->if_carp)
1840 		carp_carpdev_state(ifp->if_carp);
1841 #endif
1842 #endif
1843 	rt_ifmsg(ifp);
1844 #ifdef INET6
1845 	in6_if_up(ifp);
1846 #endif
1847 }
1848 
1849 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1850 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1851 
1852 /*
1853  * Handle a change in the interface link state. To avoid LORs
1854  * between driver lock and upper layer locks, as well as possible
1855  * recursions, we post event to taskqueue, and all job
1856  * is done in static do_link_state_change().
1857  */
1858 void
1859 if_link_state_change(struct ifnet *ifp, int link_state)
1860 {
1861 	/* Return if state hasn't changed. */
1862 	if (ifp->if_link_state == link_state)
1863 		return;
1864 
1865 	ifp->if_link_state = link_state;
1866 
1867 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1868 }
1869 
1870 static void
1871 do_link_state_change(void *arg, int pending)
1872 {
1873 	struct ifnet *ifp = (struct ifnet *)arg;
1874 	int link_state = ifp->if_link_state;
1875 	CURVNET_SET(ifp->if_vnet);
1876 
1877 	/* Notify that the link state has changed. */
1878 	rt_ifmsg(ifp);
1879 	if (ifp->if_vlantrunk != NULL)
1880 		(*vlan_link_state_p)(ifp);
1881 
1882 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1883 	    IFP2AC(ifp)->ac_netgraph != NULL)
1884 		(*ng_ether_link_state_p)(ifp, link_state);
1885 #if defined(INET) || defined(INET6)
1886 #ifdef DEV_CARP
1887 	if (ifp->if_carp)
1888 		carp_carpdev_state(ifp->if_carp);
1889 #endif
1890 #endif
1891 	if (ifp->if_bridge) {
1892 		KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1893 		(*bstp_linkstate_p)(ifp, link_state);
1894 	}
1895 	if (ifp->if_lagg) {
1896 		KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1897 		(*lagg_linkstate_p)(ifp, link_state);
1898 	}
1899 
1900 	if (IS_DEFAULT_VNET(curvnet))
1901 		devctl_notify("IFNET", ifp->if_xname,
1902 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1903 		    NULL);
1904 	if (pending > 1)
1905 		if_printf(ifp, "%d link states coalesced\n", pending);
1906 	if (log_link_state_change)
1907 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1908 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1909 	CURVNET_RESTORE();
1910 }
1911 
1912 /*
1913  * Mark an interface down and notify protocols of
1914  * the transition.
1915  * NOTE: must be called at splnet or eqivalent.
1916  */
1917 void
1918 if_down(struct ifnet *ifp)
1919 {
1920 
1921 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1922 }
1923 
1924 /*
1925  * Mark an interface up and notify protocols of
1926  * the transition.
1927  * NOTE: must be called at splnet or eqivalent.
1928  */
1929 void
1930 if_up(struct ifnet *ifp)
1931 {
1932 
1933 	if_route(ifp, IFF_UP, AF_UNSPEC);
1934 }
1935 
1936 /*
1937  * Flush an interface queue.
1938  */
1939 void
1940 if_qflush(struct ifnet *ifp)
1941 {
1942 	struct mbuf *m, *n;
1943 	struct ifaltq *ifq;
1944 
1945 	ifq = &ifp->if_snd;
1946 	IFQ_LOCK(ifq);
1947 #ifdef ALTQ
1948 	if (ALTQ_IS_ENABLED(ifq))
1949 		ALTQ_PURGE(ifq);
1950 #endif
1951 	n = ifq->ifq_head;
1952 	while ((m = n) != 0) {
1953 		n = m->m_act;
1954 		m_freem(m);
1955 	}
1956 	ifq->ifq_head = 0;
1957 	ifq->ifq_tail = 0;
1958 	ifq->ifq_len = 0;
1959 	IFQ_UNLOCK(ifq);
1960 }
1961 
1962 /*
1963  * Map interface name to interface structure pointer, with or without
1964  * returning a reference.
1965  */
1966 struct ifnet *
1967 ifunit_ref(const char *name)
1968 {
1969 	struct ifnet *ifp;
1970 
1971 	IFNET_RLOCK_NOSLEEP();
1972 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1973 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
1974 		    !(ifp->if_flags & IFF_DYING))
1975 			break;
1976 	}
1977 	if (ifp != NULL)
1978 		if_ref(ifp);
1979 	IFNET_RUNLOCK_NOSLEEP();
1980 	return (ifp);
1981 }
1982 
1983 struct ifnet *
1984 ifunit(const char *name)
1985 {
1986 	struct ifnet *ifp;
1987 
1988 	IFNET_RLOCK_NOSLEEP();
1989 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1990 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1991 			break;
1992 	}
1993 	IFNET_RUNLOCK_NOSLEEP();
1994 	return (ifp);
1995 }
1996 
1997 /*
1998  * Hardware specific interface ioctls.
1999  */
2000 static int
2001 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2002 {
2003 	struct ifreq *ifr;
2004 	struct ifstat *ifs;
2005 	int error = 0;
2006 	int new_flags, temp_flags;
2007 	size_t namelen, onamelen;
2008 	size_t descrlen;
2009 	char *descrbuf, *odescrbuf;
2010 	char new_name[IFNAMSIZ];
2011 	struct ifaddr *ifa;
2012 	struct sockaddr_dl *sdl;
2013 
2014 	ifr = (struct ifreq *)data;
2015 	switch (cmd) {
2016 	case SIOCGIFINDEX:
2017 		ifr->ifr_index = ifp->if_index;
2018 		break;
2019 
2020 	case SIOCGIFFLAGS:
2021 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2022 		ifr->ifr_flags = temp_flags & 0xffff;
2023 		ifr->ifr_flagshigh = temp_flags >> 16;
2024 		break;
2025 
2026 	case SIOCGIFCAP:
2027 		ifr->ifr_reqcap = ifp->if_capabilities;
2028 		ifr->ifr_curcap = ifp->if_capenable;
2029 		break;
2030 
2031 #ifdef MAC
2032 	case SIOCGIFMAC:
2033 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2034 		break;
2035 #endif
2036 
2037 	case SIOCGIFMETRIC:
2038 		ifr->ifr_metric = ifp->if_metric;
2039 		break;
2040 
2041 	case SIOCGIFMTU:
2042 		ifr->ifr_mtu = ifp->if_mtu;
2043 		break;
2044 
2045 	case SIOCGIFPHYS:
2046 		ifr->ifr_phys = ifp->if_physical;
2047 		break;
2048 
2049 	case SIOCGIFDESCR:
2050 		error = 0;
2051 		sx_slock(&ifdescr_sx);
2052 		if (ifp->if_description == NULL)
2053 			error = ENOMSG;
2054 		else {
2055 			/* space for terminating nul */
2056 			descrlen = strlen(ifp->if_description) + 1;
2057 			if (ifr->ifr_buffer.length < descrlen)
2058 				ifr->ifr_buffer.buffer = NULL;
2059 			else
2060 				error = copyout(ifp->if_description,
2061 				    ifr->ifr_buffer.buffer, descrlen);
2062 			ifr->ifr_buffer.length = descrlen;
2063 		}
2064 		sx_sunlock(&ifdescr_sx);
2065 		break;
2066 
2067 	case SIOCSIFDESCR:
2068 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2069 		if (error)
2070 			return (error);
2071 
2072 		/*
2073 		 * Copy only (length-1) bytes to make sure that
2074 		 * if_description is always nul terminated.  The
2075 		 * length parameter is supposed to count the
2076 		 * terminating nul in.
2077 		 */
2078 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2079 			return (ENAMETOOLONG);
2080 		else if (ifr->ifr_buffer.length == 0)
2081 			descrbuf = NULL;
2082 		else {
2083 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2084 			    M_WAITOK | M_ZERO);
2085 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2086 			    ifr->ifr_buffer.length - 1);
2087 			if (error) {
2088 				free(descrbuf, M_IFDESCR);
2089 				break;
2090 			}
2091 		}
2092 
2093 		sx_xlock(&ifdescr_sx);
2094 		odescrbuf = ifp->if_description;
2095 		ifp->if_description = descrbuf;
2096 		sx_xunlock(&ifdescr_sx);
2097 
2098 		getmicrotime(&ifp->if_lastchange);
2099 		free(odescrbuf, M_IFDESCR);
2100 		break;
2101 
2102 	case SIOCSIFFLAGS:
2103 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2104 		if (error)
2105 			return (error);
2106 		/*
2107 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2108 		 * check, so we don't need special handling here yet.
2109 		 */
2110 		new_flags = (ifr->ifr_flags & 0xffff) |
2111 		    (ifr->ifr_flagshigh << 16);
2112 		if (ifp->if_flags & IFF_SMART) {
2113 			/* Smart drivers twiddle their own routes */
2114 		} else if (ifp->if_flags & IFF_UP &&
2115 		    (new_flags & IFF_UP) == 0) {
2116 			int s = splimp();
2117 			if_down(ifp);
2118 			splx(s);
2119 		} else if (new_flags & IFF_UP &&
2120 		    (ifp->if_flags & IFF_UP) == 0) {
2121 			int s = splimp();
2122 			if_up(ifp);
2123 			splx(s);
2124 		}
2125 		/* See if permanently promiscuous mode bit is about to flip */
2126 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2127 			if (new_flags & IFF_PPROMISC)
2128 				ifp->if_flags |= IFF_PROMISC;
2129 			else if (ifp->if_pcount == 0)
2130 				ifp->if_flags &= ~IFF_PROMISC;
2131 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2132 			    ifp->if_xname,
2133 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2134 		}
2135 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2136 			(new_flags &~ IFF_CANTCHANGE);
2137 		if (ifp->if_ioctl) {
2138 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2139 		}
2140 		getmicrotime(&ifp->if_lastchange);
2141 		break;
2142 
2143 	case SIOCSIFCAP:
2144 		error = priv_check(td, PRIV_NET_SETIFCAP);
2145 		if (error)
2146 			return (error);
2147 		if (ifp->if_ioctl == NULL)
2148 			return (EOPNOTSUPP);
2149 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2150 			return (EINVAL);
2151 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2152 		if (error == 0)
2153 			getmicrotime(&ifp->if_lastchange);
2154 		break;
2155 
2156 #ifdef MAC
2157 	case SIOCSIFMAC:
2158 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2159 		break;
2160 #endif
2161 
2162 	case SIOCSIFNAME:
2163 		error = priv_check(td, PRIV_NET_SETIFNAME);
2164 		if (error)
2165 			return (error);
2166 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2167 		if (error != 0)
2168 			return (error);
2169 		if (new_name[0] == '\0')
2170 			return (EINVAL);
2171 		if (ifunit(new_name) != NULL)
2172 			return (EEXIST);
2173 
2174 		/*
2175 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2176 		 * and there are numerous other races with the
2177 		 * ifunit() checks not being atomic with namespace
2178 		 * changes (renames, vmoves, if_attach, etc).
2179 		 */
2180 		ifp->if_flags |= IFF_RENAMING;
2181 
2182 		/* Announce the departure of the interface. */
2183 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2184 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2185 
2186 		log(LOG_INFO, "%s: changing name to '%s'\n",
2187 		    ifp->if_xname, new_name);
2188 
2189 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2190 		ifa = ifp->if_addr;
2191 		IFA_LOCK(ifa);
2192 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2193 		namelen = strlen(new_name);
2194 		onamelen = sdl->sdl_nlen;
2195 		/*
2196 		 * Move the address if needed.  This is safe because we
2197 		 * allocate space for a name of length IFNAMSIZ when we
2198 		 * create this in if_attach().
2199 		 */
2200 		if (namelen != onamelen) {
2201 			bcopy(sdl->sdl_data + onamelen,
2202 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2203 		}
2204 		bcopy(new_name, sdl->sdl_data, namelen);
2205 		sdl->sdl_nlen = namelen;
2206 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2207 		bzero(sdl->sdl_data, onamelen);
2208 		while (namelen != 0)
2209 			sdl->sdl_data[--namelen] = 0xff;
2210 		IFA_UNLOCK(ifa);
2211 
2212 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2213 		/* Announce the return of the interface. */
2214 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2215 
2216 		ifp->if_flags &= ~IFF_RENAMING;
2217 		break;
2218 
2219 #ifdef VIMAGE
2220 	case SIOCSIFVNET:
2221 		error = priv_check(td, PRIV_NET_SETIFVNET);
2222 		if (error)
2223 			return (error);
2224 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2225 		break;
2226 #endif
2227 
2228 	case SIOCSIFMETRIC:
2229 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2230 		if (error)
2231 			return (error);
2232 		ifp->if_metric = ifr->ifr_metric;
2233 		getmicrotime(&ifp->if_lastchange);
2234 		break;
2235 
2236 	case SIOCSIFPHYS:
2237 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2238 		if (error)
2239 			return (error);
2240 		if (ifp->if_ioctl == NULL)
2241 			return (EOPNOTSUPP);
2242 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2243 		if (error == 0)
2244 			getmicrotime(&ifp->if_lastchange);
2245 		break;
2246 
2247 	case SIOCSIFMTU:
2248 	{
2249 		u_long oldmtu = ifp->if_mtu;
2250 
2251 		error = priv_check(td, PRIV_NET_SETIFMTU);
2252 		if (error)
2253 			return (error);
2254 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2255 			return (EINVAL);
2256 		if (ifp->if_ioctl == NULL)
2257 			return (EOPNOTSUPP);
2258 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2259 		if (error == 0) {
2260 			getmicrotime(&ifp->if_lastchange);
2261 			rt_ifmsg(ifp);
2262 		}
2263 		/*
2264 		 * If the link MTU changed, do network layer specific procedure.
2265 		 */
2266 		if (ifp->if_mtu != oldmtu) {
2267 #ifdef INET6
2268 			nd6_setmtu(ifp);
2269 #endif
2270 		}
2271 		break;
2272 	}
2273 
2274 	case SIOCADDMULTI:
2275 	case SIOCDELMULTI:
2276 		if (cmd == SIOCADDMULTI)
2277 			error = priv_check(td, PRIV_NET_ADDMULTI);
2278 		else
2279 			error = priv_check(td, PRIV_NET_DELMULTI);
2280 		if (error)
2281 			return (error);
2282 
2283 		/* Don't allow group membership on non-multicast interfaces. */
2284 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2285 			return (EOPNOTSUPP);
2286 
2287 		/* Don't let users screw up protocols' entries. */
2288 		if (ifr->ifr_addr.sa_family != AF_LINK)
2289 			return (EINVAL);
2290 
2291 		if (cmd == SIOCADDMULTI) {
2292 			struct ifmultiaddr *ifma;
2293 
2294 			/*
2295 			 * Userland is only permitted to join groups once
2296 			 * via the if_addmulti() KPI, because it cannot hold
2297 			 * struct ifmultiaddr * between calls. It may also
2298 			 * lose a race while we check if the membership
2299 			 * already exists.
2300 			 */
2301 			IF_ADDR_LOCK(ifp);
2302 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2303 			IF_ADDR_UNLOCK(ifp);
2304 			if (ifma != NULL)
2305 				error = EADDRINUSE;
2306 			else
2307 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2308 		} else {
2309 			error = if_delmulti(ifp, &ifr->ifr_addr);
2310 		}
2311 		if (error == 0)
2312 			getmicrotime(&ifp->if_lastchange);
2313 		break;
2314 
2315 	case SIOCSIFPHYADDR:
2316 	case SIOCDIFPHYADDR:
2317 #ifdef INET6
2318 	case SIOCSIFPHYADDR_IN6:
2319 #endif
2320 	case SIOCSLIFPHYADDR:
2321 	case SIOCSIFMEDIA:
2322 	case SIOCSIFGENERIC:
2323 		error = priv_check(td, PRIV_NET_HWIOCTL);
2324 		if (error)
2325 			return (error);
2326 		if (ifp->if_ioctl == NULL)
2327 			return (EOPNOTSUPP);
2328 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2329 		if (error == 0)
2330 			getmicrotime(&ifp->if_lastchange);
2331 		break;
2332 
2333 	case SIOCGIFSTATUS:
2334 		ifs = (struct ifstat *)data;
2335 		ifs->ascii[0] = '\0';
2336 
2337 	case SIOCGIFPSRCADDR:
2338 	case SIOCGIFPDSTADDR:
2339 	case SIOCGLIFPHYADDR:
2340 	case SIOCGIFMEDIA:
2341 	case SIOCGIFGENERIC:
2342 		if (ifp->if_ioctl == NULL)
2343 			return (EOPNOTSUPP);
2344 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2345 		break;
2346 
2347 	case SIOCSIFLLADDR:
2348 		error = priv_check(td, PRIV_NET_SETLLADDR);
2349 		if (error)
2350 			return (error);
2351 		error = if_setlladdr(ifp,
2352 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2353 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2354 		break;
2355 
2356 	case SIOCAIFGROUP:
2357 	{
2358 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2359 
2360 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2361 		if (error)
2362 			return (error);
2363 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2364 			return (error);
2365 		break;
2366 	}
2367 
2368 	case SIOCGIFGROUP:
2369 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2370 			return (error);
2371 		break;
2372 
2373 	case SIOCDIFGROUP:
2374 	{
2375 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2376 
2377 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2378 		if (error)
2379 			return (error);
2380 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2381 			return (error);
2382 		break;
2383 	}
2384 
2385 	default:
2386 		error = ENOIOCTL;
2387 		break;
2388 	}
2389 	return (error);
2390 }
2391 
2392 /*
2393  * Interface ioctls.
2394  */
2395 int
2396 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2397 {
2398 	struct ifnet *ifp;
2399 	struct ifreq *ifr;
2400 	int error;
2401 	int oif_flags;
2402 
2403 	switch (cmd) {
2404 	case SIOCGIFCONF:
2405 	case OSIOCGIFCONF:
2406 #ifdef __amd64__
2407 	case SIOCGIFCONF32:
2408 #endif
2409 		return (ifconf(cmd, data));
2410 	}
2411 	ifr = (struct ifreq *)data;
2412 
2413 	switch (cmd) {
2414 #ifdef VIMAGE
2415 	case SIOCSIFRVNET:
2416 		error = priv_check(td, PRIV_NET_SETIFVNET);
2417 		if (error)
2418 			return (error);
2419 		return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
2420 #endif
2421 	case SIOCIFCREATE:
2422 	case SIOCIFCREATE2:
2423 		error = priv_check(td, PRIV_NET_IFCREATE);
2424 		if (error)
2425 			return (error);
2426 		return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
2427 			cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
2428 	case SIOCIFDESTROY:
2429 		error = priv_check(td, PRIV_NET_IFDESTROY);
2430 		if (error)
2431 			return (error);
2432 		return if_clone_destroy(ifr->ifr_name);
2433 
2434 	case SIOCIFGCLONERS:
2435 		return (if_clone_list((struct if_clonereq *)data));
2436 	case SIOCGIFGMEMB:
2437 		return (if_getgroupmembers((struct ifgroupreq *)data));
2438 	}
2439 
2440 	ifp = ifunit_ref(ifr->ifr_name);
2441 	if (ifp == NULL)
2442 		return (ENXIO);
2443 
2444 	error = ifhwioctl(cmd, ifp, data, td);
2445 	if (error != ENOIOCTL) {
2446 		if_rele(ifp);
2447 		return (error);
2448 	}
2449 
2450 	oif_flags = ifp->if_flags;
2451 	if (so->so_proto == NULL) {
2452 		if_rele(ifp);
2453 		return (EOPNOTSUPP);
2454 	}
2455 #ifndef COMPAT_43
2456 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2457 								 data,
2458 								 ifp, td));
2459 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2460 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2461 #else
2462 	{
2463 		u_long ocmd = cmd;
2464 
2465 		switch (cmd) {
2466 
2467 		case SIOCSIFDSTADDR:
2468 		case SIOCSIFADDR:
2469 		case SIOCSIFBRDADDR:
2470 		case SIOCSIFNETMASK:
2471 #if BYTE_ORDER != BIG_ENDIAN
2472 			if (ifr->ifr_addr.sa_family == 0 &&
2473 			    ifr->ifr_addr.sa_len < 16) {
2474 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2475 				ifr->ifr_addr.sa_len = 16;
2476 			}
2477 #else
2478 			if (ifr->ifr_addr.sa_len == 0)
2479 				ifr->ifr_addr.sa_len = 16;
2480 #endif
2481 			break;
2482 
2483 		case OSIOCGIFADDR:
2484 			cmd = SIOCGIFADDR;
2485 			break;
2486 
2487 		case OSIOCGIFDSTADDR:
2488 			cmd = SIOCGIFDSTADDR;
2489 			break;
2490 
2491 		case OSIOCGIFBRDADDR:
2492 			cmd = SIOCGIFBRDADDR;
2493 			break;
2494 
2495 		case OSIOCGIFNETMASK:
2496 			cmd = SIOCGIFNETMASK;
2497 		}
2498 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
2499 								   cmd,
2500 								   data,
2501 								   ifp, td));
2502 		if (error == EOPNOTSUPP && ifp != NULL &&
2503 		    ifp->if_ioctl != NULL)
2504 			error = (*ifp->if_ioctl)(ifp, cmd, data);
2505 		switch (ocmd) {
2506 
2507 		case OSIOCGIFADDR:
2508 		case OSIOCGIFDSTADDR:
2509 		case OSIOCGIFBRDADDR:
2510 		case OSIOCGIFNETMASK:
2511 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2512 
2513 		}
2514 	}
2515 #endif /* COMPAT_43 */
2516 
2517 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2518 #ifdef INET6
2519 		if (ifp->if_flags & IFF_UP) {
2520 			int s = splimp();
2521 			in6_if_up(ifp);
2522 			splx(s);
2523 		}
2524 #endif
2525 	}
2526 	if_rele(ifp);
2527 	return (error);
2528 }
2529 
2530 /*
2531  * The code common to handling reference counted flags,
2532  * e.g., in ifpromisc() and if_allmulti().
2533  * The "pflag" argument can specify a permanent mode flag to check,
2534  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2535  *
2536  * Only to be used on stack-owned flags, not driver-owned flags.
2537  */
2538 static int
2539 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2540 {
2541 	struct ifreq ifr;
2542 	int error;
2543 	int oldflags, oldcount;
2544 
2545 	/* Sanity checks to catch programming errors */
2546 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2547 	    ("%s: setting driver-owned flag %d", __func__, flag));
2548 
2549 	if (onswitch)
2550 		KASSERT(*refcount >= 0,
2551 		    ("%s: increment negative refcount %d for flag %d",
2552 		    __func__, *refcount, flag));
2553 	else
2554 		KASSERT(*refcount > 0,
2555 		    ("%s: decrement non-positive refcount %d for flag %d",
2556 		    __func__, *refcount, flag));
2557 
2558 	/* In case this mode is permanent, just touch refcount */
2559 	if (ifp->if_flags & pflag) {
2560 		*refcount += onswitch ? 1 : -1;
2561 		return (0);
2562 	}
2563 
2564 	/* Save ifnet parameters for if_ioctl() may fail */
2565 	oldcount = *refcount;
2566 	oldflags = ifp->if_flags;
2567 
2568 	/*
2569 	 * See if we aren't the only and touching refcount is enough.
2570 	 * Actually toggle interface flag if we are the first or last.
2571 	 */
2572 	if (onswitch) {
2573 		if ((*refcount)++)
2574 			return (0);
2575 		ifp->if_flags |= flag;
2576 	} else {
2577 		if (--(*refcount))
2578 			return (0);
2579 		ifp->if_flags &= ~flag;
2580 	}
2581 
2582 	/* Call down the driver since we've changed interface flags */
2583 	if (ifp->if_ioctl == NULL) {
2584 		error = EOPNOTSUPP;
2585 		goto recover;
2586 	}
2587 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2588 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2589 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2590 	if (error)
2591 		goto recover;
2592 	/* Notify userland that interface flags have changed */
2593 	rt_ifmsg(ifp);
2594 	return (0);
2595 
2596 recover:
2597 	/* Recover after driver error */
2598 	*refcount = oldcount;
2599 	ifp->if_flags = oldflags;
2600 	return (error);
2601 }
2602 
2603 /*
2604  * Set/clear promiscuous mode on interface ifp based on the truth value
2605  * of pswitch.  The calls are reference counted so that only the first
2606  * "on" request actually has an effect, as does the final "off" request.
2607  * Results are undefined if the "off" and "on" requests are not matched.
2608  */
2609 int
2610 ifpromisc(struct ifnet *ifp, int pswitch)
2611 {
2612 	int error;
2613 	int oldflags = ifp->if_flags;
2614 
2615 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2616 			   &ifp->if_pcount, pswitch);
2617 	/* If promiscuous mode status has changed, log a message */
2618 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2619 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2620 		    ifp->if_xname,
2621 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2622 	return (error);
2623 }
2624 
2625 /*
2626  * Return interface configuration
2627  * of system.  List may be used
2628  * in later ioctl's (above) to get
2629  * other information.
2630  */
2631 /*ARGSUSED*/
2632 static int
2633 ifconf(u_long cmd, caddr_t data)
2634 {
2635 	struct ifconf *ifc = (struct ifconf *)data;
2636 #ifdef __amd64__
2637 	struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2638 	struct ifconf ifc_swab;
2639 #endif
2640 	struct ifnet *ifp;
2641 	struct ifaddr *ifa;
2642 	struct ifreq ifr;
2643 	struct sbuf *sb;
2644 	int error, full = 0, valid_len, max_len;
2645 
2646 #ifdef __amd64__
2647 	if (cmd == SIOCGIFCONF32) {
2648 		ifc_swab.ifc_len = ifc32->ifc_len;
2649 		ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2650 		ifc = &ifc_swab;
2651 	}
2652 #endif
2653 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2654 	max_len = MAXPHYS - 1;
2655 
2656 	/* Prevent hostile input from being able to crash the system */
2657 	if (ifc->ifc_len <= 0)
2658 		return (EINVAL);
2659 
2660 again:
2661 	if (ifc->ifc_len <= max_len) {
2662 		max_len = ifc->ifc_len;
2663 		full = 1;
2664 	}
2665 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2666 	max_len = 0;
2667 	valid_len = 0;
2668 
2669 	IFNET_RLOCK();
2670 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2671 		int addrs;
2672 
2673 		/*
2674 		 * Zero the ifr_name buffer to make sure we don't
2675 		 * disclose the contents of the stack.
2676 		 */
2677 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2678 
2679 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2680 		    >= sizeof(ifr.ifr_name)) {
2681 			sbuf_delete(sb);
2682 			IFNET_RUNLOCK();
2683 			return (ENAMETOOLONG);
2684 		}
2685 
2686 		addrs = 0;
2687 		IF_ADDR_LOCK(ifp);
2688 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2689 			struct sockaddr *sa = ifa->ifa_addr;
2690 
2691 			if (prison_if(curthread->td_ucred, sa) != 0)
2692 				continue;
2693 			addrs++;
2694 #ifdef COMPAT_43
2695 			if (cmd == OSIOCGIFCONF) {
2696 				struct osockaddr *osa =
2697 					 (struct osockaddr *)&ifr.ifr_addr;
2698 				ifr.ifr_addr = *sa;
2699 				osa->sa_family = sa->sa_family;
2700 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2701 				max_len += sizeof(ifr);
2702 			} else
2703 #endif
2704 			if (sa->sa_len <= sizeof(*sa)) {
2705 				ifr.ifr_addr = *sa;
2706 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2707 				max_len += sizeof(ifr);
2708 			} else {
2709 				sbuf_bcat(sb, &ifr,
2710 				    offsetof(struct ifreq, ifr_addr));
2711 				max_len += offsetof(struct ifreq, ifr_addr);
2712 				sbuf_bcat(sb, sa, sa->sa_len);
2713 				max_len += sa->sa_len;
2714 			}
2715 
2716 			if (!sbuf_overflowed(sb))
2717 				valid_len = sbuf_len(sb);
2718 		}
2719 		IF_ADDR_UNLOCK(ifp);
2720 		if (addrs == 0) {
2721 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2722 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2723 			max_len += sizeof(ifr);
2724 
2725 			if (!sbuf_overflowed(sb))
2726 				valid_len = sbuf_len(sb);
2727 		}
2728 	}
2729 	IFNET_RUNLOCK();
2730 
2731 	/*
2732 	 * If we didn't allocate enough space (uncommon), try again.  If
2733 	 * we have already allocated as much space as we are allowed,
2734 	 * return what we've got.
2735 	 */
2736 	if (valid_len != max_len && !full) {
2737 		sbuf_delete(sb);
2738 		goto again;
2739 	}
2740 
2741 	ifc->ifc_len = valid_len;
2742 #ifdef __amd64__
2743 	if (cmd == SIOCGIFCONF32)
2744 		ifc32->ifc_len = valid_len;
2745 #endif
2746 	sbuf_finish(sb);
2747 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2748 	sbuf_delete(sb);
2749 	return (error);
2750 }
2751 
2752 /*
2753  * Just like ifpromisc(), but for all-multicast-reception mode.
2754  */
2755 int
2756 if_allmulti(struct ifnet *ifp, int onswitch)
2757 {
2758 
2759 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2760 }
2761 
2762 struct ifmultiaddr *
2763 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2764 {
2765 	struct ifmultiaddr *ifma;
2766 
2767 	IF_ADDR_LOCK_ASSERT(ifp);
2768 
2769 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2770 		if (sa->sa_family == AF_LINK) {
2771 			if (sa_dl_equal(ifma->ifma_addr, sa))
2772 				break;
2773 		} else {
2774 			if (sa_equal(ifma->ifma_addr, sa))
2775 				break;
2776 		}
2777 	}
2778 
2779 	return ifma;
2780 }
2781 
2782 /*
2783  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2784  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2785  * the ifnet multicast address list here, so the caller must do that and
2786  * other setup work (such as notifying the device driver).  The reference
2787  * count is initialized to 1.
2788  */
2789 static struct ifmultiaddr *
2790 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2791     int mflags)
2792 {
2793 	struct ifmultiaddr *ifma;
2794 	struct sockaddr *dupsa;
2795 
2796 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2797 	    M_ZERO);
2798 	if (ifma == NULL)
2799 		return (NULL);
2800 
2801 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2802 	if (dupsa == NULL) {
2803 		free(ifma, M_IFMADDR);
2804 		return (NULL);
2805 	}
2806 	bcopy(sa, dupsa, sa->sa_len);
2807 	ifma->ifma_addr = dupsa;
2808 
2809 	ifma->ifma_ifp = ifp;
2810 	ifma->ifma_refcount = 1;
2811 	ifma->ifma_protospec = NULL;
2812 
2813 	if (llsa == NULL) {
2814 		ifma->ifma_lladdr = NULL;
2815 		return (ifma);
2816 	}
2817 
2818 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2819 	if (dupsa == NULL) {
2820 		free(ifma->ifma_addr, M_IFMADDR);
2821 		free(ifma, M_IFMADDR);
2822 		return (NULL);
2823 	}
2824 	bcopy(llsa, dupsa, llsa->sa_len);
2825 	ifma->ifma_lladdr = dupsa;
2826 
2827 	return (ifma);
2828 }
2829 
2830 /*
2831  * if_freemulti: free ifmultiaddr structure and possibly attached related
2832  * addresses.  The caller is responsible for implementing reference
2833  * counting, notifying the driver, handling routing messages, and releasing
2834  * any dependent link layer state.
2835  */
2836 static void
2837 if_freemulti(struct ifmultiaddr *ifma)
2838 {
2839 
2840 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2841 	    ifma->ifma_refcount));
2842 	KASSERT(ifma->ifma_protospec == NULL,
2843 	    ("if_freemulti: protospec not NULL"));
2844 
2845 	if (ifma->ifma_lladdr != NULL)
2846 		free(ifma->ifma_lladdr, M_IFMADDR);
2847 	free(ifma->ifma_addr, M_IFMADDR);
2848 	free(ifma, M_IFMADDR);
2849 }
2850 
2851 /*
2852  * Register an additional multicast address with a network interface.
2853  *
2854  * - If the address is already present, bump the reference count on the
2855  *   address and return.
2856  * - If the address is not link-layer, look up a link layer address.
2857  * - Allocate address structures for one or both addresses, and attach to the
2858  *   multicast address list on the interface.  If automatically adding a link
2859  *   layer address, the protocol address will own a reference to the link
2860  *   layer address, to be freed when it is freed.
2861  * - Notify the network device driver of an addition to the multicast address
2862  *   list.
2863  *
2864  * 'sa' points to caller-owned memory with the desired multicast address.
2865  *
2866  * 'retifma' will be used to return a pointer to the resulting multicast
2867  * address reference, if desired.
2868  */
2869 int
2870 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2871     struct ifmultiaddr **retifma)
2872 {
2873 	struct ifmultiaddr *ifma, *ll_ifma;
2874 	struct sockaddr *llsa;
2875 	int error;
2876 
2877 	/*
2878 	 * If the address is already present, return a new reference to it;
2879 	 * otherwise, allocate storage and set up a new address.
2880 	 */
2881 	IF_ADDR_LOCK(ifp);
2882 	ifma = if_findmulti(ifp, sa);
2883 	if (ifma != NULL) {
2884 		ifma->ifma_refcount++;
2885 		if (retifma != NULL)
2886 			*retifma = ifma;
2887 		IF_ADDR_UNLOCK(ifp);
2888 		return (0);
2889 	}
2890 
2891 	/*
2892 	 * The address isn't already present; resolve the protocol address
2893 	 * into a link layer address, and then look that up, bump its
2894 	 * refcount or allocate an ifma for that also.  If 'llsa' was
2895 	 * returned, we will need to free it later.
2896 	 */
2897 	llsa = NULL;
2898 	ll_ifma = NULL;
2899 	if (ifp->if_resolvemulti != NULL) {
2900 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2901 		if (error)
2902 			goto unlock_out;
2903 	}
2904 
2905 	/*
2906 	 * Allocate the new address.  Don't hook it up yet, as we may also
2907 	 * need to allocate a link layer multicast address.
2908 	 */
2909 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2910 	if (ifma == NULL) {
2911 		error = ENOMEM;
2912 		goto free_llsa_out;
2913 	}
2914 
2915 	/*
2916 	 * If a link layer address is found, we'll need to see if it's
2917 	 * already present in the address list, or allocate is as well.
2918 	 * When this block finishes, the link layer address will be on the
2919 	 * list.
2920 	 */
2921 	if (llsa != NULL) {
2922 		ll_ifma = if_findmulti(ifp, llsa);
2923 		if (ll_ifma == NULL) {
2924 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2925 			if (ll_ifma == NULL) {
2926 				--ifma->ifma_refcount;
2927 				if_freemulti(ifma);
2928 				error = ENOMEM;
2929 				goto free_llsa_out;
2930 			}
2931 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2932 			    ifma_link);
2933 		} else
2934 			ll_ifma->ifma_refcount++;
2935 		ifma->ifma_llifma = ll_ifma;
2936 	}
2937 
2938 	/*
2939 	 * We now have a new multicast address, ifma, and possibly a new or
2940 	 * referenced link layer address.  Add the primary address to the
2941 	 * ifnet address list.
2942 	 */
2943 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2944 
2945 	if (retifma != NULL)
2946 		*retifma = ifma;
2947 
2948 	/*
2949 	 * Must generate the message while holding the lock so that 'ifma'
2950 	 * pointer is still valid.
2951 	 */
2952 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2953 	IF_ADDR_UNLOCK(ifp);
2954 
2955 	/*
2956 	 * We are certain we have added something, so call down to the
2957 	 * interface to let them know about it.
2958 	 */
2959 	if (ifp->if_ioctl != NULL) {
2960 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2961 	}
2962 
2963 	if (llsa != NULL)
2964 		free(llsa, M_IFMADDR);
2965 
2966 	return (0);
2967 
2968 free_llsa_out:
2969 	if (llsa != NULL)
2970 		free(llsa, M_IFMADDR);
2971 
2972 unlock_out:
2973 	IF_ADDR_UNLOCK(ifp);
2974 	return (error);
2975 }
2976 
2977 /*
2978  * Delete a multicast group membership by network-layer group address.
2979  *
2980  * Returns ENOENT if the entry could not be found. If ifp no longer
2981  * exists, results are undefined. This entry point should only be used
2982  * from subsystems which do appropriate locking to hold ifp for the
2983  * duration of the call.
2984  * Network-layer protocol domains must use if_delmulti_ifma().
2985  */
2986 int
2987 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2988 {
2989 	struct ifmultiaddr *ifma;
2990 	int lastref;
2991 #ifdef INVARIANTS
2992 	struct ifnet *oifp;
2993 
2994 	IFNET_RLOCK_NOSLEEP();
2995 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2996 		if (ifp == oifp)
2997 			break;
2998 	if (ifp != oifp)
2999 		ifp = NULL;
3000 	IFNET_RUNLOCK_NOSLEEP();
3001 
3002 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3003 #endif
3004 	if (ifp == NULL)
3005 		return (ENOENT);
3006 
3007 	IF_ADDR_LOCK(ifp);
3008 	lastref = 0;
3009 	ifma = if_findmulti(ifp, sa);
3010 	if (ifma != NULL)
3011 		lastref = if_delmulti_locked(ifp, ifma, 0);
3012 	IF_ADDR_UNLOCK(ifp);
3013 
3014 	if (ifma == NULL)
3015 		return (ENOENT);
3016 
3017 	if (lastref && ifp->if_ioctl != NULL) {
3018 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3019 	}
3020 
3021 	return (0);
3022 }
3023 
3024 /*
3025  * Delete all multicast group membership for an interface.
3026  * Should be used to quickly flush all multicast filters.
3027  */
3028 void
3029 if_delallmulti(struct ifnet *ifp)
3030 {
3031 	struct ifmultiaddr *ifma;
3032 	struct ifmultiaddr *next;
3033 
3034 	IF_ADDR_LOCK(ifp);
3035 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3036 		if_delmulti_locked(ifp, ifma, 0);
3037 	IF_ADDR_UNLOCK(ifp);
3038 }
3039 
3040 /*
3041  * Delete a multicast group membership by group membership pointer.
3042  * Network-layer protocol domains must use this routine.
3043  *
3044  * It is safe to call this routine if the ifp disappeared.
3045  */
3046 void
3047 if_delmulti_ifma(struct ifmultiaddr *ifma)
3048 {
3049 	struct ifnet *ifp;
3050 	int lastref;
3051 
3052 	ifp = ifma->ifma_ifp;
3053 #ifdef DIAGNOSTIC
3054 	if (ifp == NULL) {
3055 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3056 	} else {
3057 		struct ifnet *oifp;
3058 
3059 		IFNET_RLOCK_NOSLEEP();
3060 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3061 			if (ifp == oifp)
3062 				break;
3063 		if (ifp != oifp) {
3064 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3065 			ifp = NULL;
3066 		}
3067 		IFNET_RUNLOCK_NOSLEEP();
3068 	}
3069 #endif
3070 	/*
3071 	 * If and only if the ifnet instance exists: Acquire the address lock.
3072 	 */
3073 	if (ifp != NULL)
3074 		IF_ADDR_LOCK(ifp);
3075 
3076 	lastref = if_delmulti_locked(ifp, ifma, 0);
3077 
3078 	if (ifp != NULL) {
3079 		/*
3080 		 * If and only if the ifnet instance exists:
3081 		 *  Release the address lock.
3082 		 *  If the group was left: update the hardware hash filter.
3083 		 */
3084 		IF_ADDR_UNLOCK(ifp);
3085 		if (lastref && ifp->if_ioctl != NULL) {
3086 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3087 		}
3088 	}
3089 }
3090 
3091 /*
3092  * Perform deletion of network-layer and/or link-layer multicast address.
3093  *
3094  * Return 0 if the reference count was decremented.
3095  * Return 1 if the final reference was released, indicating that the
3096  * hardware hash filter should be reprogrammed.
3097  */
3098 static int
3099 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3100 {
3101 	struct ifmultiaddr *ll_ifma;
3102 
3103 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3104 		KASSERT(ifma->ifma_ifp == ifp,
3105 		    ("%s: inconsistent ifp %p", __func__, ifp));
3106 		IF_ADDR_LOCK_ASSERT(ifp);
3107 	}
3108 
3109 	ifp = ifma->ifma_ifp;
3110 
3111 	/*
3112 	 * If the ifnet is detaching, null out references to ifnet,
3113 	 * so that upper protocol layers will notice, and not attempt
3114 	 * to obtain locks for an ifnet which no longer exists. The
3115 	 * routing socket announcement must happen before the ifnet
3116 	 * instance is detached from the system.
3117 	 */
3118 	if (detaching) {
3119 #ifdef DIAGNOSTIC
3120 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3121 #endif
3122 		/*
3123 		 * ifp may already be nulled out if we are being reentered
3124 		 * to delete the ll_ifma.
3125 		 */
3126 		if (ifp != NULL) {
3127 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3128 			ifma->ifma_ifp = NULL;
3129 		}
3130 	}
3131 
3132 	if (--ifma->ifma_refcount > 0)
3133 		return 0;
3134 
3135 	/*
3136 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3137 	 * have been associated with it. Release it first if so.
3138 	 */
3139 	ll_ifma = ifma->ifma_llifma;
3140 	if (ll_ifma != NULL) {
3141 		KASSERT(ifma->ifma_lladdr != NULL,
3142 		    ("%s: llifma w/o lladdr", __func__));
3143 		if (detaching)
3144 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3145 		if (--ll_ifma->ifma_refcount == 0) {
3146 			if (ifp != NULL) {
3147 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3148 				    ifma_link);
3149 			}
3150 			if_freemulti(ll_ifma);
3151 		}
3152 	}
3153 
3154 	if (ifp != NULL)
3155 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3156 
3157 	if_freemulti(ifma);
3158 
3159 	/*
3160 	 * The last reference to this instance of struct ifmultiaddr
3161 	 * was released; the hardware should be notified of this change.
3162 	 */
3163 	return 1;
3164 }
3165 
3166 /*
3167  * Set the link layer address on an interface.
3168  *
3169  * At this time we only support certain types of interfaces,
3170  * and we don't allow the length of the address to change.
3171  */
3172 int
3173 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3174 {
3175 	struct sockaddr_dl *sdl;
3176 	struct ifaddr *ifa;
3177 	struct ifreq ifr;
3178 
3179 	IF_ADDR_LOCK(ifp);
3180 	ifa = ifp->if_addr;
3181 	if (ifa == NULL) {
3182 		IF_ADDR_UNLOCK(ifp);
3183 		return (EINVAL);
3184 	}
3185 	ifa_ref(ifa);
3186 	IF_ADDR_UNLOCK(ifp);
3187 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3188 	if (sdl == NULL) {
3189 		ifa_free(ifa);
3190 		return (EINVAL);
3191 	}
3192 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3193 		ifa_free(ifa);
3194 		return (EINVAL);
3195 	}
3196 	switch (ifp->if_type) {
3197 	case IFT_ETHER:
3198 	case IFT_FDDI:
3199 	case IFT_XETHER:
3200 	case IFT_ISO88025:
3201 	case IFT_L2VLAN:
3202 	case IFT_BRIDGE:
3203 	case IFT_ARCNET:
3204 	case IFT_IEEE8023ADLAG:
3205 	case IFT_IEEE80211:
3206 		bcopy(lladdr, LLADDR(sdl), len);
3207 		ifa_free(ifa);
3208 		break;
3209 	default:
3210 		ifa_free(ifa);
3211 		return (ENODEV);
3212 	}
3213 
3214 	/*
3215 	 * If the interface is already up, we need
3216 	 * to re-init it in order to reprogram its
3217 	 * address filter.
3218 	 */
3219 	if ((ifp->if_flags & IFF_UP) != 0) {
3220 		if (ifp->if_ioctl) {
3221 			ifp->if_flags &= ~IFF_UP;
3222 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3223 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3224 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3225 			ifp->if_flags |= IFF_UP;
3226 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3227 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3228 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3229 		}
3230 #ifdef INET
3231 		/*
3232 		 * Also send gratuitous ARPs to notify other nodes about
3233 		 * the address change.
3234 		 */
3235 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3236 			if (ifa->ifa_addr->sa_family == AF_INET)
3237 				arp_ifinit(ifp, ifa);
3238 		}
3239 #endif
3240 	}
3241 	return (0);
3242 }
3243 
3244 /*
3245  * The name argument must be a pointer to storage which will last as
3246  * long as the interface does.  For physical devices, the result of
3247  * device_get_name(dev) is a good choice and for pseudo-devices a
3248  * static string works well.
3249  */
3250 void
3251 if_initname(struct ifnet *ifp, const char *name, int unit)
3252 {
3253 	ifp->if_dname = name;
3254 	ifp->if_dunit = unit;
3255 	if (unit != IF_DUNIT_NONE)
3256 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3257 	else
3258 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3259 }
3260 
3261 int
3262 if_printf(struct ifnet *ifp, const char * fmt, ...)
3263 {
3264 	va_list ap;
3265 	int retval;
3266 
3267 	retval = printf("%s: ", ifp->if_xname);
3268 	va_start(ap, fmt);
3269 	retval += vprintf(fmt, ap);
3270 	va_end(ap);
3271 	return (retval);
3272 }
3273 
3274 void
3275 if_start(struct ifnet *ifp)
3276 {
3277 
3278 	(*(ifp)->if_start)(ifp);
3279 }
3280 
3281 /*
3282  * Backwards compatibility interface for drivers
3283  * that have not implemented it
3284  */
3285 static int
3286 if_transmit(struct ifnet *ifp, struct mbuf *m)
3287 {
3288 	int error;
3289 
3290 	IFQ_HANDOFF(ifp, m, error);
3291 	return (error);
3292 }
3293 
3294 int
3295 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3296 {
3297 	int active = 0;
3298 
3299 	IF_LOCK(ifq);
3300 	if (_IF_QFULL(ifq)) {
3301 		_IF_DROP(ifq);
3302 		IF_UNLOCK(ifq);
3303 		m_freem(m);
3304 		return (0);
3305 	}
3306 	if (ifp != NULL) {
3307 		ifp->if_obytes += m->m_pkthdr.len + adjust;
3308 		if (m->m_flags & (M_BCAST|M_MCAST))
3309 			ifp->if_omcasts++;
3310 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3311 	}
3312 	_IF_ENQUEUE(ifq, m);
3313 	IF_UNLOCK(ifq);
3314 	if (ifp != NULL && !active)
3315 		(*(ifp)->if_start)(ifp);
3316 	return (1);
3317 }
3318 
3319 void
3320 if_register_com_alloc(u_char type,
3321     if_com_alloc_t *a, if_com_free_t *f)
3322 {
3323 
3324 	KASSERT(if_com_alloc[type] == NULL,
3325 	    ("if_register_com_alloc: %d already registered", type));
3326 	KASSERT(if_com_free[type] == NULL,
3327 	    ("if_register_com_alloc: %d free already registered", type));
3328 
3329 	if_com_alloc[type] = a;
3330 	if_com_free[type] = f;
3331 }
3332 
3333 void
3334 if_deregister_com_alloc(u_char type)
3335 {
3336 
3337 	KASSERT(if_com_alloc[type] != NULL,
3338 	    ("if_deregister_com_alloc: %d not registered", type));
3339 	KASSERT(if_com_free[type] != NULL,
3340 	    ("if_deregister_com_alloc: %d free not registered", type));
3341 	if_com_alloc[type] = NULL;
3342 	if_com_free[type] = NULL;
3343 }
3344 
3345 #ifdef DDB
3346 static void
3347 if_show_ifnet(struct ifnet *ifp)
3348 {
3349 
3350 	if (ifp == NULL)
3351 		return;
3352 	db_printf("%s:\n", ifp->if_xname);
3353 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
3354 	IF_DB_PRINTF("%s", if_dname);
3355 	IF_DB_PRINTF("%d", if_dunit);
3356 	IF_DB_PRINTF("%s", if_description);
3357 	IF_DB_PRINTF("%u", if_index);
3358 	IF_DB_PRINTF("%u", if_refcount);
3359 	IF_DB_PRINTF("%d", if_index_reserved);
3360 	IF_DB_PRINTF("%p", if_softc);
3361 	IF_DB_PRINTF("%p", if_l2com);
3362 	IF_DB_PRINTF("%p", if_vnet);
3363 	IF_DB_PRINTF("%p", if_home_vnet);
3364 	IF_DB_PRINTF("%p", if_addr);
3365 	IF_DB_PRINTF("%p", if_llsoftc);
3366 	IF_DB_PRINTF("%p", if_label);
3367 	IF_DB_PRINTF("%u", if_pcount);
3368 	IF_DB_PRINTF("0x%08x", if_flags);
3369 	IF_DB_PRINTF("0x%08x", if_drv_flags);
3370 	IF_DB_PRINTF("0x%08x", if_capabilities);
3371 	IF_DB_PRINTF("0x%08x", if_capenable);
3372 	IF_DB_PRINTF("%p", if_snd.ifq_head);
3373 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
3374 	IF_DB_PRINTF("%d", if_snd.ifq_len);
3375 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
3376 	IF_DB_PRINTF("%d", if_snd.ifq_drops);
3377 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
3378 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
3379 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
3380 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
3381 	IF_DB_PRINTF("%d", if_snd.altq_type);
3382 	IF_DB_PRINTF("%x", if_snd.altq_flags);
3383 #undef IF_DB_PRINTF
3384 }
3385 
3386 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
3387 {
3388 
3389 	if (!have_addr) {
3390 		db_printf("usage: show ifnet <struct ifnet *>\n");
3391 		return;
3392 	}
3393 
3394 	if_show_ifnet((struct ifnet *)addr);
3395 }
3396 
3397 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
3398 {
3399 	VNET_ITERATOR_DECL(vnet_iter);
3400 	struct ifnet *ifp;
3401 	u_short idx;
3402 
3403 	VNET_FOREACH(vnet_iter) {
3404 		CURVNET_SET_QUIET(vnet_iter);
3405 #ifdef VIMAGE
3406 		db_printf("vnet=%p\n", curvnet);
3407 #endif
3408 		for (idx = 1; idx <= V_if_index; idx++) {
3409 			ifp = V_ifindex_table[idx].ife_ifnet;
3410 			if (ifp == NULL)
3411 				continue;
3412 			db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
3413 			if (db_pager_quit)
3414 				break;
3415 		}
3416 		CURVNET_RESTORE();
3417 	}
3418 }
3419 #endif
3420