xref: /freebsd/sys/net/if.c (revision 99429157e8615dc3b7f11afbe3ed92de7476a5db)
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  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62 
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #include <net/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/if.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif /* INET */
89 #ifdef INET6
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
92 #endif /* INET6 */
93 #endif /* INET || INET6 */
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
100 #endif
101 
102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
104 
105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106     &ifqmaxlen, 0, "max send queue size");
107 
108 /* Log link state change events */
109 static int log_link_state_change = 1;
110 
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 	&log_link_state_change, 0,
113 	"log interface link state change events");
114 
115 /* Log promiscuous mode change events */
116 static int log_promisc_mode_change = 1;
117 
118 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
119 	&log_promisc_mode_change, 1,
120 	"log promiscuous mode change events");
121 
122 /* Interface description */
123 static unsigned int ifdescr_maxlen = 1024;
124 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
125 	&ifdescr_maxlen, 0,
126 	"administrative maximum length for interface description");
127 
128 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
129 
130 /* global sx for non-critical path ifdescr */
131 static struct sx ifdescr_sx;
132 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
133 
134 void	(*bridge_linkstate_p)(struct ifnet *ifp);
135 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
136 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
137 /* These are external hooks for CARP. */
138 void	(*carp_linkstate_p)(struct ifnet *ifp);
139 void	(*carp_demote_adj_p)(int, char *);
140 int	(*carp_master_p)(struct ifaddr *);
141 #if defined(INET) || defined(INET6)
142 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
143 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
144     const struct sockaddr *sa);
145 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
146 int	(*carp_attach_p)(struct ifaddr *, int);
147 void	(*carp_detach_p)(struct ifaddr *, bool);
148 #endif
149 #ifdef INET
150 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
151 #endif
152 #ifdef INET6
153 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
154 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
155     const struct in6_addr *taddr);
156 #endif
157 
158 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
159 
160 /*
161  * XXX: Style; these should be sorted alphabetically, and unprototyped
162  * static functions should be prototyped. Currently they are sorted by
163  * declaration order.
164  */
165 static void	if_attachdomain(void *);
166 static void	if_attachdomain1(struct ifnet *);
167 static int	ifconf(u_long, caddr_t);
168 static void	if_freemulti(struct ifmultiaddr *);
169 static void	if_grow(void);
170 static void	if_input_default(struct ifnet *, struct mbuf *);
171 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
172 static void	if_route(struct ifnet *, int flag, int fam);
173 static int	if_setflag(struct ifnet *, int, int, int *, int);
174 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
175 static void	if_unroute(struct ifnet *, int flag, int fam);
176 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
177 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
178 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
179 static void	do_link_state_change(void *, int);
180 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
181 static int	if_getgroupmembers(struct ifgroupreq *);
182 static void	if_delgroups(struct ifnet *);
183 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
184 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
185 #ifdef VIMAGE
186 static void	if_vmove(struct ifnet *, struct vnet *);
187 #endif
188 
189 #ifdef INET6
190 /*
191  * XXX: declare here to avoid to include many inet6 related files..
192  * should be more generalized?
193  */
194 extern void	nd6_setmtu(struct ifnet *);
195 #endif
196 
197 /* ipsec helper hooks */
198 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
199 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
200 
201 VNET_DEFINE(int, if_index);
202 int	ifqmaxlen = IFQ_MAXLEN;
203 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
204 VNET_DEFINE(struct ifgrouphead, ifg_head);
205 
206 static VNET_DEFINE(int, if_indexlim) = 8;
207 
208 /* Table of ifnet by index. */
209 VNET_DEFINE(struct ifnet **, ifindex_table);
210 
211 #define	V_if_indexlim		VNET(if_indexlim)
212 #define	V_ifindex_table		VNET(ifindex_table)
213 
214 /*
215  * The global network interface list (V_ifnet) and related state (such as
216  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
217  * an rwlock.  Either may be acquired shared to stablize the list, but both
218  * must be acquired writable to modify the list.  This model allows us to
219  * both stablize the interface list during interrupt thread processing, but
220  * also to stablize it over long-running ioctls, without introducing priority
221  * inversions and deadlocks.
222  */
223 struct rwlock ifnet_rwlock;
224 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
225 struct sx ifnet_sxlock;
226 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
227 
228 /*
229  * The allocation of network interfaces is a rather non-atomic affair; we
230  * need to select an index before we are ready to expose the interface for
231  * use, so will use this pointer value to indicate reservation.
232  */
233 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
234 
235 static	if_com_alloc_t *if_com_alloc[256];
236 static	if_com_free_t *if_com_free[256];
237 
238 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
239 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
240 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
241 
242 struct ifnet *
243 ifnet_byindex_locked(u_short idx)
244 {
245 
246 	if (idx > V_if_index)
247 		return (NULL);
248 	if (V_ifindex_table[idx] == IFNET_HOLD)
249 		return (NULL);
250 	return (V_ifindex_table[idx]);
251 }
252 
253 struct ifnet *
254 ifnet_byindex(u_short idx)
255 {
256 	struct ifnet *ifp;
257 
258 	IFNET_RLOCK_NOSLEEP();
259 	ifp = ifnet_byindex_locked(idx);
260 	IFNET_RUNLOCK_NOSLEEP();
261 	return (ifp);
262 }
263 
264 struct ifnet *
265 ifnet_byindex_ref(u_short idx)
266 {
267 	struct ifnet *ifp;
268 
269 	IFNET_RLOCK_NOSLEEP();
270 	ifp = ifnet_byindex_locked(idx);
271 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
272 		IFNET_RUNLOCK_NOSLEEP();
273 		return (NULL);
274 	}
275 	if_ref(ifp);
276 	IFNET_RUNLOCK_NOSLEEP();
277 	return (ifp);
278 }
279 
280 /*
281  * Allocate an ifindex array entry; return 0 on success or an error on
282  * failure.
283  */
284 static u_short
285 ifindex_alloc(void)
286 {
287 	u_short idx;
288 
289 	IFNET_WLOCK_ASSERT();
290 retry:
291 	/*
292 	 * Try to find an empty slot below V_if_index.  If we fail, take the
293 	 * next slot.
294 	 */
295 	for (idx = 1; idx <= V_if_index; idx++) {
296 		if (V_ifindex_table[idx] == NULL)
297 			break;
298 	}
299 
300 	/* Catch if_index overflow. */
301 	if (idx >= V_if_indexlim) {
302 		if_grow();
303 		goto retry;
304 	}
305 	if (idx > V_if_index)
306 		V_if_index = idx;
307 	return (idx);
308 }
309 
310 static void
311 ifindex_free_locked(u_short idx)
312 {
313 
314 	IFNET_WLOCK_ASSERT();
315 
316 	V_ifindex_table[idx] = NULL;
317 	while (V_if_index > 0 &&
318 	    V_ifindex_table[V_if_index] == NULL)
319 		V_if_index--;
320 }
321 
322 static void
323 ifindex_free(u_short idx)
324 {
325 
326 	IFNET_WLOCK();
327 	ifindex_free_locked(idx);
328 	IFNET_WUNLOCK();
329 }
330 
331 static void
332 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
333 {
334 
335 	IFNET_WLOCK_ASSERT();
336 
337 	V_ifindex_table[idx] = ifp;
338 }
339 
340 static void
341 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
342 {
343 
344 	IFNET_WLOCK();
345 	ifnet_setbyindex_locked(idx, ifp);
346 	IFNET_WUNLOCK();
347 }
348 
349 struct ifaddr *
350 ifaddr_byindex(u_short idx)
351 {
352 	struct ifnet *ifp;
353 	struct ifaddr *ifa = NULL;
354 
355 	IFNET_RLOCK_NOSLEEP();
356 	ifp = ifnet_byindex_locked(idx);
357 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
358 		ifa_ref(ifa);
359 	IFNET_RUNLOCK_NOSLEEP();
360 	return (ifa);
361 }
362 
363 /*
364  * Network interface utility routines.
365  *
366  * Routines with ifa_ifwith* names take sockaddr *'s as
367  * parameters.
368  */
369 
370 static void
371 vnet_if_init(const void *unused __unused)
372 {
373 
374 	TAILQ_INIT(&V_ifnet);
375 	TAILQ_INIT(&V_ifg_head);
376 	IFNET_WLOCK();
377 	if_grow();				/* create initial table */
378 	IFNET_WUNLOCK();
379 	vnet_if_clone_init();
380 }
381 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
382     NULL);
383 
384 #ifdef VIMAGE
385 static void
386 vnet_if_uninit(const void *unused __unused)
387 {
388 
389 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
390 	    "not empty", __func__, __LINE__, &V_ifnet));
391 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
392 	    "not empty", __func__, __LINE__, &V_ifg_head));
393 
394 	free((caddr_t)V_ifindex_table, M_IFNET);
395 }
396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
397     vnet_if_uninit, NULL);
398 
399 static void
400 vnet_if_return(const void *unused __unused)
401 {
402 	struct ifnet *ifp, *nifp;
403 
404 	/* Return all inherited interfaces to their parent vnets. */
405 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
406 		if (ifp->if_home_vnet != ifp->if_vnet)
407 			if_vmove(ifp, ifp->if_home_vnet);
408 	}
409 }
410 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
411     vnet_if_return, NULL);
412 #endif
413 
414 static void
415 if_grow(void)
416 {
417 	int oldlim;
418 	u_int n;
419 	struct ifnet **e;
420 
421 	IFNET_WLOCK_ASSERT();
422 	oldlim = V_if_indexlim;
423 	IFNET_WUNLOCK();
424 	n = (oldlim << 1) * sizeof(*e);
425 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
426 	IFNET_WLOCK();
427 	if (V_if_indexlim != oldlim) {
428 		free(e, M_IFNET);
429 		return;
430 	}
431 	if (V_ifindex_table != NULL) {
432 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
433 		free((caddr_t)V_ifindex_table, M_IFNET);
434 	}
435 	V_if_indexlim <<= 1;
436 	V_ifindex_table = e;
437 }
438 
439 /*
440  * Allocate a struct ifnet and an index for an interface.  A layer 2
441  * common structure will also be allocated if an allocation routine is
442  * registered for the passed type.
443  */
444 struct ifnet *
445 if_alloc(u_char type)
446 {
447 	struct ifnet *ifp;
448 	u_short idx;
449 
450 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
451 	IFNET_WLOCK();
452 	idx = ifindex_alloc();
453 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
454 	IFNET_WUNLOCK();
455 	ifp->if_index = idx;
456 	ifp->if_type = type;
457 	ifp->if_alloctype = type;
458 #ifdef VIMAGE
459 	ifp->if_vnet = curvnet;
460 #endif
461 	if (if_com_alloc[type] != NULL) {
462 		ifp->if_l2com = if_com_alloc[type](type, ifp);
463 		if (ifp->if_l2com == NULL) {
464 			free(ifp, M_IFNET);
465 			ifindex_free(idx);
466 			return (NULL);
467 		}
468 	}
469 
470 	IF_ADDR_LOCK_INIT(ifp);
471 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
472 	ifp->if_afdata_initialized = 0;
473 	IF_AFDATA_LOCK_INIT(ifp);
474 	TAILQ_INIT(&ifp->if_addrhead);
475 	TAILQ_INIT(&ifp->if_multiaddrs);
476 	TAILQ_INIT(&ifp->if_groups);
477 #ifdef MAC
478 	mac_ifnet_init(ifp);
479 #endif
480 	ifq_init(&ifp->if_snd, ifp);
481 
482 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
483 	for (int i = 0; i < IFCOUNTERS; i++)
484 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
485 	ifp->if_get_counter = if_get_counter_default;
486 	ifnet_setbyindex(ifp->if_index, ifp);
487 	return (ifp);
488 }
489 
490 /*
491  * Do the actual work of freeing a struct ifnet, and layer 2 common
492  * structure.  This call is made when the last reference to an
493  * interface is released.
494  */
495 static void
496 if_free_internal(struct ifnet *ifp)
497 {
498 
499 	KASSERT((ifp->if_flags & IFF_DYING),
500 	    ("if_free_internal: interface not dying"));
501 
502 	if (if_com_free[ifp->if_alloctype] != NULL)
503 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
504 		    ifp->if_alloctype);
505 
506 #ifdef MAC
507 	mac_ifnet_destroy(ifp);
508 #endif /* MAC */
509 	if (ifp->if_description != NULL)
510 		free(ifp->if_description, M_IFDESCR);
511 	IF_AFDATA_DESTROY(ifp);
512 	IF_ADDR_LOCK_DESTROY(ifp);
513 	ifq_delete(&ifp->if_snd);
514 
515 	for (int i = 0; i < IFCOUNTERS; i++)
516 		counter_u64_free(ifp->if_counters[i]);
517 
518 	free(ifp, M_IFNET);
519 }
520 
521 /*
522  * Deregister an interface and free the associated storage.
523  */
524 void
525 if_free(struct ifnet *ifp)
526 {
527 
528 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
529 
530 	CURVNET_SET_QUIET(ifp->if_vnet);
531 	IFNET_WLOCK();
532 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
533 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
534 
535 	ifindex_free_locked(ifp->if_index);
536 	IFNET_WUNLOCK();
537 
538 	if (refcount_release(&ifp->if_refcount))
539 		if_free_internal(ifp);
540 	CURVNET_RESTORE();
541 }
542 
543 /*
544  * Interfaces to keep an ifnet type-stable despite the possibility of the
545  * driver calling if_free().  If there are additional references, we defer
546  * freeing the underlying data structure.
547  */
548 void
549 if_ref(struct ifnet *ifp)
550 {
551 
552 	/* We don't assert the ifnet list lock here, but arguably should. */
553 	refcount_acquire(&ifp->if_refcount);
554 }
555 
556 void
557 if_rele(struct ifnet *ifp)
558 {
559 
560 	if (!refcount_release(&ifp->if_refcount))
561 		return;
562 	if_free_internal(ifp);
563 }
564 
565 void
566 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
567 {
568 
569 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
570 
571 	if (ifq->ifq_maxlen == 0)
572 		ifq->ifq_maxlen = ifqmaxlen;
573 
574 	ifq->altq_type = 0;
575 	ifq->altq_disc = NULL;
576 	ifq->altq_flags &= ALTQF_CANTCHANGE;
577 	ifq->altq_tbr  = NULL;
578 	ifq->altq_ifp  = ifp;
579 }
580 
581 void
582 ifq_delete(struct ifaltq *ifq)
583 {
584 	mtx_destroy(&ifq->ifq_mtx);
585 }
586 
587 /*
588  * Perform generic interface initialization tasks and attach the interface
589  * to the list of "active" interfaces.  If vmove flag is set on entry
590  * to if_attach_internal(), perform only a limited subset of initialization
591  * tasks, given that we are moving from one vnet to another an ifnet which
592  * has already been fully initialized.
593  *
594  * Note that if_detach_internal() removes group membership unconditionally
595  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
596  * Thus, when if_vmove() is applied to a cloned interface, group membership
597  * is lost while a cloned one always joins a group whose name is
598  * ifc->ifc_name.  To recover this after if_detach_internal() and
599  * if_attach_internal(), the cloner should be specified to
600  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
601  * attempts to join a group whose name is ifc->ifc_name.
602  *
603  * XXX:
604  *  - The decision to return void and thus require this function to
605  *    succeed is questionable.
606  *  - We should probably do more sanity checking.  For instance we don't
607  *    do anything to insure if_xname is unique or non-empty.
608  */
609 void
610 if_attach(struct ifnet *ifp)
611 {
612 
613 	if_attach_internal(ifp, 0, NULL);
614 }
615 
616 /*
617  * Compute the least common TSO limit.
618  */
619 void
620 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
621 {
622 	/*
623 	 * 1) If there is no limit currently, take the limit from
624 	 * the network adapter.
625 	 *
626 	 * 2) If the network adapter has a limit below the current
627 	 * limit, apply it.
628 	 */
629 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
630 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
631 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
632 	}
633 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
634 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
635 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
636 	}
637 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
638 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
639 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
640 	}
641 }
642 
643 /*
644  * Update TSO limit of a network adapter.
645  *
646  * Returns zero if no change. Else non-zero.
647  */
648 int
649 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
650 {
651 	int retval = 0;
652 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
653 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
654 		retval++;
655 	}
656 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
657 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
658 		retval++;
659 	}
660 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
661 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
662 		retval++;
663 	}
664 	return (retval);
665 }
666 
667 static void
668 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
669 {
670 	unsigned socksize, ifasize;
671 	int namelen, masklen;
672 	struct sockaddr_dl *sdl;
673 	struct ifaddr *ifa;
674 
675 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
676 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
677 		    ifp->if_xname);
678 
679 #ifdef VIMAGE
680 	ifp->if_vnet = curvnet;
681 	if (ifp->if_home_vnet == NULL)
682 		ifp->if_home_vnet = curvnet;
683 #endif
684 
685 	if_addgroup(ifp, IFG_ALL);
686 
687 	/* Restore group membership for cloned interfaces. */
688 	if (vmove && ifc != NULL)
689 		if_clone_addgroup(ifp, ifc);
690 
691 	getmicrotime(&ifp->if_lastchange);
692 	ifp->if_epoch = time_uptime;
693 
694 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
695 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
696 	    ("transmit and qflush must both either be set or both be NULL"));
697 	if (ifp->if_transmit == NULL) {
698 		ifp->if_transmit = if_transmit;
699 		ifp->if_qflush = if_qflush;
700 	}
701 	if (ifp->if_input == NULL)
702 		ifp->if_input = if_input_default;
703 
704 	if (ifp->if_requestencap == NULL)
705 		ifp->if_requestencap = if_requestencap_default;
706 
707 	if (!vmove) {
708 #ifdef MAC
709 		mac_ifnet_create(ifp);
710 #endif
711 
712 		/*
713 		 * Create a Link Level name for this device.
714 		 */
715 		namelen = strlen(ifp->if_xname);
716 		/*
717 		 * Always save enough space for any possiable name so we
718 		 * can do a rename in place later.
719 		 */
720 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
721 		socksize = masklen + ifp->if_addrlen;
722 		if (socksize < sizeof(*sdl))
723 			socksize = sizeof(*sdl);
724 		socksize = roundup2(socksize, sizeof(long));
725 		ifasize = sizeof(*ifa) + 2 * socksize;
726 		ifa = ifa_alloc(ifasize, M_WAITOK);
727 		sdl = (struct sockaddr_dl *)(ifa + 1);
728 		sdl->sdl_len = socksize;
729 		sdl->sdl_family = AF_LINK;
730 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
731 		sdl->sdl_nlen = namelen;
732 		sdl->sdl_index = ifp->if_index;
733 		sdl->sdl_type = ifp->if_type;
734 		ifp->if_addr = ifa;
735 		ifa->ifa_ifp = ifp;
736 		ifa->ifa_rtrequest = link_rtrequest;
737 		ifa->ifa_addr = (struct sockaddr *)sdl;
738 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
739 		ifa->ifa_netmask = (struct sockaddr *)sdl;
740 		sdl->sdl_len = masklen;
741 		while (namelen != 0)
742 			sdl->sdl_data[--namelen] = 0xff;
743 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
744 		/* Reliably crash if used uninitialized. */
745 		ifp->if_broadcastaddr = NULL;
746 
747 		if (ifp->if_type == IFT_ETHER) {
748 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
749 			    M_WAITOK | M_ZERO);
750 		}
751 
752 #if defined(INET) || defined(INET6)
753 		/* Use defaults for TSO, if nothing is set */
754 		if (ifp->if_hw_tsomax == 0 &&
755 		    ifp->if_hw_tsomaxsegcount == 0 &&
756 		    ifp->if_hw_tsomaxsegsize == 0) {
757 			/*
758 			 * The TSO defaults needs to be such that an
759 			 * NFS mbuf list of 35 mbufs totalling just
760 			 * below 64K works and that a chain of mbufs
761 			 * can be defragged into at most 32 segments:
762 			 */
763 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
764 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
765 			ifp->if_hw_tsomaxsegcount = 35;
766 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
767 
768 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
769 			if (ifp->if_capabilities & IFCAP_TSO) {
770 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
771 				    ifp->if_hw_tsomax,
772 				    ifp->if_hw_tsomaxsegcount,
773 				    ifp->if_hw_tsomaxsegsize);
774 			}
775 		}
776 #endif
777 	}
778 #ifdef VIMAGE
779 	else {
780 		/*
781 		 * Update the interface index in the link layer address
782 		 * of the interface.
783 		 */
784 		for (ifa = ifp->if_addr; ifa != NULL;
785 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
786 			if (ifa->ifa_addr->sa_family == AF_LINK) {
787 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
788 				sdl->sdl_index = ifp->if_index;
789 			}
790 		}
791 	}
792 #endif
793 
794 	IFNET_WLOCK();
795 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
796 #ifdef VIMAGE
797 	curvnet->vnet_ifcnt++;
798 #endif
799 	IFNET_WUNLOCK();
800 
801 	if (domain_init_status >= 2)
802 		if_attachdomain1(ifp);
803 
804 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
805 	if (IS_DEFAULT_VNET(curvnet))
806 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
807 
808 	/* Announce the interface. */
809 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
810 }
811 
812 static void
813 if_attachdomain(void *dummy)
814 {
815 	struct ifnet *ifp;
816 
817 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
818 		if_attachdomain1(ifp);
819 }
820 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
821     if_attachdomain, NULL);
822 
823 static void
824 if_attachdomain1(struct ifnet *ifp)
825 {
826 	struct domain *dp;
827 
828 	/*
829 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
830 	 * cannot lock ifp->if_afdata initialization, entirely.
831 	 */
832 	IF_AFDATA_LOCK(ifp);
833 	if (ifp->if_afdata_initialized >= domain_init_status) {
834 		IF_AFDATA_UNLOCK(ifp);
835 		log(LOG_WARNING, "%s called more than once on %s\n",
836 		    __func__, ifp->if_xname);
837 		return;
838 	}
839 	ifp->if_afdata_initialized = domain_init_status;
840 	IF_AFDATA_UNLOCK(ifp);
841 
842 	/* address family dependent data region */
843 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
844 	for (dp = domains; dp; dp = dp->dom_next) {
845 		if (dp->dom_ifattach)
846 			ifp->if_afdata[dp->dom_family] =
847 			    (*dp->dom_ifattach)(ifp);
848 	}
849 }
850 
851 /*
852  * Remove any unicast or broadcast network addresses from an interface.
853  */
854 void
855 if_purgeaddrs(struct ifnet *ifp)
856 {
857 	struct ifaddr *ifa, *next;
858 
859 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
860 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
861 		if (ifa->ifa_addr->sa_family == AF_LINK)
862 			continue;
863 #ifdef INET
864 		/* XXX: Ugly!! ad hoc just for INET */
865 		if (ifa->ifa_addr->sa_family == AF_INET) {
866 			struct ifaliasreq ifr;
867 
868 			bzero(&ifr, sizeof(ifr));
869 			ifr.ifra_addr = *ifa->ifa_addr;
870 			if (ifa->ifa_dstaddr)
871 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
872 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
873 			    NULL) == 0)
874 				continue;
875 		}
876 #endif /* INET */
877 #ifdef INET6
878 		if (ifa->ifa_addr->sa_family == AF_INET6) {
879 			in6_purgeaddr(ifa);
880 			/* ifp_addrhead is already updated */
881 			continue;
882 		}
883 #endif /* INET6 */
884 		IF_ADDR_WLOCK(ifp);
885 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
886 		IF_ADDR_WUNLOCK(ifp);
887 		ifa_free(ifa);
888 	}
889 }
890 
891 /*
892  * Remove any multicast network addresses from an interface when an ifnet
893  * is going away.
894  */
895 static void
896 if_purgemaddrs(struct ifnet *ifp)
897 {
898 	struct ifmultiaddr *ifma;
899 	struct ifmultiaddr *next;
900 
901 	IF_ADDR_WLOCK(ifp);
902 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
903 		if_delmulti_locked(ifp, ifma, 1);
904 	IF_ADDR_WUNLOCK(ifp);
905 }
906 
907 /*
908  * Detach an interface, removing it from the list of "active" interfaces.
909  * If vmove flag is set on entry to if_detach_internal(), perform only a
910  * limited subset of cleanup tasks, given that we are moving an ifnet from
911  * one vnet to another, where it must be fully operational.
912  *
913  * XXXRW: There are some significant questions about event ordering, and
914  * how to prevent things from starting to use the interface during detach.
915  */
916 void
917 if_detach(struct ifnet *ifp)
918 {
919 
920 	CURVNET_SET_QUIET(ifp->if_vnet);
921 	if_detach_internal(ifp, 0, NULL);
922 	CURVNET_RESTORE();
923 }
924 
925 /*
926  * The vmove flag, if set, indicates that we are called from a callpath
927  * that is moving an interface to a different vnet instance.
928  *
929  * The shutdown flag, if set, indicates that we are called in the
930  * process of shutting down a vnet instance.  Currently only the
931  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
932  * on a vnet instance shutdown without this flag being set, e.g., when
933  * the cloned interfaces are destoyed as first thing of teardown.
934  */
935 static int
936 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
937 {
938 	struct ifaddr *ifa;
939 	int i;
940 	struct domain *dp;
941  	struct ifnet *iter;
942  	int found = 0;
943 #ifdef VIMAGE
944 	int shutdown;
945 
946 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
947 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
948 #endif
949 	IFNET_WLOCK();
950 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
951 		if (iter == ifp) {
952 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
953 			found = 1;
954 			break;
955 		}
956 	IFNET_WUNLOCK();
957 	if (!found) {
958 		/*
959 		 * While we would want to panic here, we cannot
960 		 * guarantee that the interface is indeed still on
961 		 * the list given we don't hold locks all the way.
962 		 */
963 		return (ENOENT);
964 #if 0
965 		if (vmove)
966 			panic("%s: ifp=%p not on the ifnet tailq %p",
967 			    __func__, ifp, &V_ifnet);
968 		else
969 			return; /* XXX this should panic as well? */
970 #endif
971 	}
972 
973 	/*
974 	 * At this point we know the interface still was on the ifnet list
975 	 * and we removed it so we are in a stable state.
976 	 */
977 #ifdef VIMAGE
978 	curvnet->vnet_ifcnt--;
979 #endif
980 
981 	/*
982 	 * In any case (destroy or vmove) detach us from the groups
983 	 * and remove/wait for pending events on the taskq.
984 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
985 	 */
986 	if_delgroups(ifp);
987 
988 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
989 
990 	/*
991 	 * Check if this is a cloned interface or not. Must do even if
992 	 * shutting down as a if_vmove_reclaim() would move the ifp and
993 	 * the if_clone_addgroup() will have a corrupted string overwise
994 	 * from a gibberish pointer.
995 	 */
996 	if (vmove && ifcp != NULL)
997 		*ifcp = if_clone_findifc(ifp);
998 
999 	if_down(ifp);
1000 
1001 #ifdef VIMAGE
1002 	/*
1003 	 * On VNET shutdown abort here as the stack teardown will do all
1004 	 * the work top-down for us.
1005 	 */
1006 	if (shutdown) {
1007 		/*
1008 		 * In case of a vmove we are done here without error.
1009 		 * If we would signal an error it would lead to the same
1010 		 * abort as if we did not find the ifnet anymore.
1011 		 * if_detach() calls us in void context and does not care
1012 		 * about an early abort notification, so life is splendid :)
1013 		 */
1014 		goto finish_vnet_shutdown;
1015 	}
1016 #endif
1017 
1018 	/*
1019 	 * At this point we are not tearing down a VNET and are either
1020 	 * going to destroy or vmove the interface and have to cleanup
1021 	 * accordingly.
1022 	 */
1023 
1024 	/*
1025 	 * Remove routes and flush queues.
1026 	 */
1027 #ifdef ALTQ
1028 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1029 		altq_disable(&ifp->if_snd);
1030 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1031 		altq_detach(&ifp->if_snd);
1032 #endif
1033 
1034 	if_purgeaddrs(ifp);
1035 
1036 #ifdef INET
1037 	in_ifdetach(ifp);
1038 #endif
1039 
1040 #ifdef INET6
1041 	/*
1042 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1043 	 * before removing routing entries below, since IPv6 interface direct
1044 	 * routes are expected to be removed by the IPv6-specific kernel API.
1045 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1046 	 */
1047 	in6_ifdetach(ifp);
1048 #endif
1049 	if_purgemaddrs(ifp);
1050 
1051 	/* Announce that the interface is gone. */
1052 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1053 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1054 	if (IS_DEFAULT_VNET(curvnet))
1055 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1056 
1057 	if (!vmove) {
1058 		/*
1059 		 * Prevent further calls into the device driver via ifnet.
1060 		 */
1061 		if_dead(ifp);
1062 
1063 		/*
1064 		 * Remove link ifaddr pointer and maybe decrement if_index.
1065 		 * Clean up all addresses.
1066 		 */
1067 		free(ifp->if_hw_addr, M_IFADDR);
1068 		ifp->if_hw_addr = NULL;
1069 		ifp->if_addr = NULL;
1070 
1071 		/* We can now free link ifaddr. */
1072 		IF_ADDR_WLOCK(ifp);
1073 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1074 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1075 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1076 			IF_ADDR_WUNLOCK(ifp);
1077 			ifa_free(ifa);
1078 		} else
1079 			IF_ADDR_WUNLOCK(ifp);
1080 	}
1081 
1082 	rt_flushifroutes(ifp);
1083 
1084 #ifdef VIMAGE
1085 finish_vnet_shutdown:
1086 #endif
1087 	/*
1088 	 * We cannot hold the lock over dom_ifdetach calls as they might
1089 	 * sleep, for example trying to drain a callout, thus open up the
1090 	 * theoretical race with re-attaching.
1091 	 */
1092 	IF_AFDATA_LOCK(ifp);
1093 	i = ifp->if_afdata_initialized;
1094 	ifp->if_afdata_initialized = 0;
1095 	IF_AFDATA_UNLOCK(ifp);
1096 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1097 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1098 			(*dp->dom_ifdetach)(ifp,
1099 			    ifp->if_afdata[dp->dom_family]);
1100 			ifp->if_afdata[dp->dom_family] = NULL;
1101 		}
1102 	}
1103 
1104 	return (0);
1105 }
1106 
1107 #ifdef VIMAGE
1108 /*
1109  * if_vmove() performs a limited version of if_detach() in current
1110  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1111  * An attempt is made to shrink if_index in current vnet, find an
1112  * unused if_index in target vnet and calls if_grow() if necessary,
1113  * and finally find an unused if_xname for the target vnet.
1114  */
1115 static void
1116 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1117 {
1118 	struct if_clone *ifc;
1119 	u_int bif_dlt, bif_hdrlen;
1120 	int rc;
1121 
1122  	/*
1123 	 * if_detach_internal() will call the eventhandler to notify
1124 	 * interface departure.  That will detach if_bpf.  We need to
1125 	 * safe the dlt and hdrlen so we can re-attach it later.
1126 	 */
1127 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1128 
1129 	/*
1130 	 * Detach from current vnet, but preserve LLADDR info, do not
1131 	 * mark as dead etc. so that the ifnet can be reattached later.
1132 	 * If we cannot find it, we lost the race to someone else.
1133 	 */
1134 	rc = if_detach_internal(ifp, 1, &ifc);
1135 	if (rc != 0)
1136 		return;
1137 
1138 	/*
1139 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1140 	 * the if_index for that vnet if possible.
1141 	 *
1142 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1143 	 * or we'd lock on one vnet and unlock on another.
1144 	 */
1145 	IFNET_WLOCK();
1146 	ifindex_free_locked(ifp->if_index);
1147 	IFNET_WUNLOCK();
1148 
1149 	/*
1150 	 * Perform interface-specific reassignment tasks, if provided by
1151 	 * the driver.
1152 	 */
1153 	if (ifp->if_reassign != NULL)
1154 		ifp->if_reassign(ifp, new_vnet, NULL);
1155 
1156 	/*
1157 	 * Switch to the context of the target vnet.
1158 	 */
1159 	CURVNET_SET_QUIET(new_vnet);
1160 
1161 	IFNET_WLOCK();
1162 	ifp->if_index = ifindex_alloc();
1163 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1164 	IFNET_WUNLOCK();
1165 
1166 	if_attach_internal(ifp, 1, ifc);
1167 
1168 	if (ifp->if_bpf == NULL)
1169 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1170 
1171 	CURVNET_RESTORE();
1172 }
1173 
1174 /*
1175  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1176  */
1177 static int
1178 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1179 {
1180 	struct prison *pr;
1181 	struct ifnet *difp;
1182 	int shutdown;
1183 
1184 	/* Try to find the prison within our visibility. */
1185 	sx_slock(&allprison_lock);
1186 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1187 	sx_sunlock(&allprison_lock);
1188 	if (pr == NULL)
1189 		return (ENXIO);
1190 	prison_hold_locked(pr);
1191 	mtx_unlock(&pr->pr_mtx);
1192 
1193 	/* Do not try to move the iface from and to the same prison. */
1194 	if (pr->pr_vnet == ifp->if_vnet) {
1195 		prison_free(pr);
1196 		return (EEXIST);
1197 	}
1198 
1199 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1200 	/* XXX Lock interfaces to avoid races. */
1201 	CURVNET_SET_QUIET(pr->pr_vnet);
1202 	difp = ifunit(ifname);
1203 	if (difp != NULL) {
1204 		CURVNET_RESTORE();
1205 		prison_free(pr);
1206 		return (EEXIST);
1207 	}
1208 
1209 	/* Make sure the VNET is stable. */
1210 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1211 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1212 	if (shutdown) {
1213 		CURVNET_RESTORE();
1214 		prison_free(pr);
1215 		return (EBUSY);
1216 	}
1217 	CURVNET_RESTORE();
1218 
1219 	/* Move the interface into the child jail/vnet. */
1220 	if_vmove(ifp, pr->pr_vnet);
1221 
1222 	/* Report the new if_xname back to the userland. */
1223 	sprintf(ifname, "%s", ifp->if_xname);
1224 
1225 	prison_free(pr);
1226 	return (0);
1227 }
1228 
1229 static int
1230 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1231 {
1232 	struct prison *pr;
1233 	struct vnet *vnet_dst;
1234 	struct ifnet *ifp;
1235  	int shutdown;
1236 
1237 	/* Try to find the prison within our visibility. */
1238 	sx_slock(&allprison_lock);
1239 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1240 	sx_sunlock(&allprison_lock);
1241 	if (pr == NULL)
1242 		return (ENXIO);
1243 	prison_hold_locked(pr);
1244 	mtx_unlock(&pr->pr_mtx);
1245 
1246 	/* Make sure the named iface exists in the source prison/vnet. */
1247 	CURVNET_SET(pr->pr_vnet);
1248 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1249 	if (ifp == NULL) {
1250 		CURVNET_RESTORE();
1251 		prison_free(pr);
1252 		return (ENXIO);
1253 	}
1254 
1255 	/* Do not try to move the iface from and to the same prison. */
1256 	vnet_dst = TD_TO_VNET(td);
1257 	if (vnet_dst == ifp->if_vnet) {
1258 		CURVNET_RESTORE();
1259 		prison_free(pr);
1260 		return (EEXIST);
1261 	}
1262 
1263 	/* Make sure the VNET is stable. */
1264 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1265 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1266 	if (shutdown) {
1267 		CURVNET_RESTORE();
1268 		prison_free(pr);
1269 		return (EBUSY);
1270 	}
1271 
1272 	/* Get interface back from child jail/vnet. */
1273 	if_vmove(ifp, vnet_dst);
1274 	CURVNET_RESTORE();
1275 
1276 	/* Report the new if_xname back to the userland. */
1277 	sprintf(ifname, "%s", ifp->if_xname);
1278 
1279 	prison_free(pr);
1280 	return (0);
1281 }
1282 #endif /* VIMAGE */
1283 
1284 /*
1285  * Add a group to an interface
1286  */
1287 int
1288 if_addgroup(struct ifnet *ifp, const char *groupname)
1289 {
1290 	struct ifg_list		*ifgl;
1291 	struct ifg_group	*ifg = NULL;
1292 	struct ifg_member	*ifgm;
1293 	int 			 new = 0;
1294 
1295 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1296 	    groupname[strlen(groupname) - 1] <= '9')
1297 		return (EINVAL);
1298 
1299 	IFNET_WLOCK();
1300 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1301 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1302 			IFNET_WUNLOCK();
1303 			return (EEXIST);
1304 		}
1305 
1306 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1307 	    M_NOWAIT)) == NULL) {
1308 	    	IFNET_WUNLOCK();
1309 		return (ENOMEM);
1310 	}
1311 
1312 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1313 	    M_TEMP, M_NOWAIT)) == NULL) {
1314 		free(ifgl, M_TEMP);
1315 		IFNET_WUNLOCK();
1316 		return (ENOMEM);
1317 	}
1318 
1319 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1320 		if (!strcmp(ifg->ifg_group, groupname))
1321 			break;
1322 
1323 	if (ifg == NULL) {
1324 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1325 		    M_TEMP, M_NOWAIT)) == NULL) {
1326 			free(ifgl, M_TEMP);
1327 			free(ifgm, M_TEMP);
1328 			IFNET_WUNLOCK();
1329 			return (ENOMEM);
1330 		}
1331 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1332 		ifg->ifg_refcnt = 0;
1333 		TAILQ_INIT(&ifg->ifg_members);
1334 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1335 		new = 1;
1336 	}
1337 
1338 	ifg->ifg_refcnt++;
1339 	ifgl->ifgl_group = ifg;
1340 	ifgm->ifgm_ifp = ifp;
1341 
1342 	IF_ADDR_WLOCK(ifp);
1343 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1344 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1345 	IF_ADDR_WUNLOCK(ifp);
1346 
1347 	IFNET_WUNLOCK();
1348 
1349 	if (new)
1350 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1351 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1352 
1353 	return (0);
1354 }
1355 
1356 /*
1357  * Remove a group from an interface
1358  */
1359 int
1360 if_delgroup(struct ifnet *ifp, const char *groupname)
1361 {
1362 	struct ifg_list		*ifgl;
1363 	struct ifg_member	*ifgm;
1364 
1365 	IFNET_WLOCK();
1366 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1367 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1368 			break;
1369 	if (ifgl == NULL) {
1370 		IFNET_WUNLOCK();
1371 		return (ENOENT);
1372 	}
1373 
1374 	IF_ADDR_WLOCK(ifp);
1375 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1376 	IF_ADDR_WUNLOCK(ifp);
1377 
1378 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1379 		if (ifgm->ifgm_ifp == ifp)
1380 			break;
1381 
1382 	if (ifgm != NULL) {
1383 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1384 		free(ifgm, M_TEMP);
1385 	}
1386 
1387 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1388 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1389 		IFNET_WUNLOCK();
1390 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1391 		free(ifgl->ifgl_group, M_TEMP);
1392 	} else
1393 		IFNET_WUNLOCK();
1394 
1395 	free(ifgl, M_TEMP);
1396 
1397 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1398 
1399 	return (0);
1400 }
1401 
1402 /*
1403  * Remove an interface from all groups
1404  */
1405 static void
1406 if_delgroups(struct ifnet *ifp)
1407 {
1408 	struct ifg_list		*ifgl;
1409 	struct ifg_member	*ifgm;
1410 	char groupname[IFNAMSIZ];
1411 
1412 	IFNET_WLOCK();
1413 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1414 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1415 
1416 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1417 
1418 		IF_ADDR_WLOCK(ifp);
1419 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1420 		IF_ADDR_WUNLOCK(ifp);
1421 
1422 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1423 			if (ifgm->ifgm_ifp == ifp)
1424 				break;
1425 
1426 		if (ifgm != NULL) {
1427 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1428 			    ifgm_next);
1429 			free(ifgm, M_TEMP);
1430 		}
1431 
1432 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1433 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1434 			IFNET_WUNLOCK();
1435 			EVENTHANDLER_INVOKE(group_detach_event,
1436 			    ifgl->ifgl_group);
1437 			free(ifgl->ifgl_group, M_TEMP);
1438 		} else
1439 			IFNET_WUNLOCK();
1440 
1441 		free(ifgl, M_TEMP);
1442 
1443 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1444 
1445 		IFNET_WLOCK();
1446 	}
1447 	IFNET_WUNLOCK();
1448 }
1449 
1450 /*
1451  * Stores all groups from an interface in memory pointed
1452  * to by data
1453  */
1454 static int
1455 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1456 {
1457 	int			 len, error;
1458 	struct ifg_list		*ifgl;
1459 	struct ifg_req		 ifgrq, *ifgp;
1460 	struct ifgroupreq	*ifgr = data;
1461 
1462 	if (ifgr->ifgr_len == 0) {
1463 		IF_ADDR_RLOCK(ifp);
1464 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1465 			ifgr->ifgr_len += sizeof(struct ifg_req);
1466 		IF_ADDR_RUNLOCK(ifp);
1467 		return (0);
1468 	}
1469 
1470 	len = ifgr->ifgr_len;
1471 	ifgp = ifgr->ifgr_groups;
1472 	/* XXX: wire */
1473 	IF_ADDR_RLOCK(ifp);
1474 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1475 		if (len < sizeof(ifgrq)) {
1476 			IF_ADDR_RUNLOCK(ifp);
1477 			return (EINVAL);
1478 		}
1479 		bzero(&ifgrq, sizeof ifgrq);
1480 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1481 		    sizeof(ifgrq.ifgrq_group));
1482 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1483 		    	IF_ADDR_RUNLOCK(ifp);
1484 			return (error);
1485 		}
1486 		len -= sizeof(ifgrq);
1487 		ifgp++;
1488 	}
1489 	IF_ADDR_RUNLOCK(ifp);
1490 
1491 	return (0);
1492 }
1493 
1494 /*
1495  * Stores all members of a group in memory pointed to by data
1496  */
1497 static int
1498 if_getgroupmembers(struct ifgroupreq *data)
1499 {
1500 	struct ifgroupreq	*ifgr = data;
1501 	struct ifg_group	*ifg;
1502 	struct ifg_member	*ifgm;
1503 	struct ifg_req		 ifgrq, *ifgp;
1504 	int			 len, error;
1505 
1506 	IFNET_RLOCK();
1507 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1508 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1509 			break;
1510 	if (ifg == NULL) {
1511 		IFNET_RUNLOCK();
1512 		return (ENOENT);
1513 	}
1514 
1515 	if (ifgr->ifgr_len == 0) {
1516 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1517 			ifgr->ifgr_len += sizeof(ifgrq);
1518 		IFNET_RUNLOCK();
1519 		return (0);
1520 	}
1521 
1522 	len = ifgr->ifgr_len;
1523 	ifgp = ifgr->ifgr_groups;
1524 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1525 		if (len < sizeof(ifgrq)) {
1526 			IFNET_RUNLOCK();
1527 			return (EINVAL);
1528 		}
1529 		bzero(&ifgrq, sizeof ifgrq);
1530 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1531 		    sizeof(ifgrq.ifgrq_member));
1532 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1533 			IFNET_RUNLOCK();
1534 			return (error);
1535 		}
1536 		len -= sizeof(ifgrq);
1537 		ifgp++;
1538 	}
1539 	IFNET_RUNLOCK();
1540 
1541 	return (0);
1542 }
1543 
1544 /*
1545  * Return counter values from counter(9)s stored in ifnet.
1546  */
1547 uint64_t
1548 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1549 {
1550 
1551 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1552 
1553 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1554 }
1555 
1556 /*
1557  * Increase an ifnet counter. Usually used for counters shared
1558  * between the stack and a driver, but function supports them all.
1559  */
1560 void
1561 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1562 {
1563 
1564 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1565 
1566 	counter_u64_add(ifp->if_counters[cnt], inc);
1567 }
1568 
1569 /*
1570  * Copy data from ifnet to userland API structure if_data.
1571  */
1572 void
1573 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1574 {
1575 
1576 	ifd->ifi_type = ifp->if_type;
1577 	ifd->ifi_physical = 0;
1578 	ifd->ifi_addrlen = ifp->if_addrlen;
1579 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1580 	ifd->ifi_link_state = ifp->if_link_state;
1581 	ifd->ifi_vhid = 0;
1582 	ifd->ifi_datalen = sizeof(struct if_data);
1583 	ifd->ifi_mtu = ifp->if_mtu;
1584 	ifd->ifi_metric = ifp->if_metric;
1585 	ifd->ifi_baudrate = ifp->if_baudrate;
1586 	ifd->ifi_hwassist = ifp->if_hwassist;
1587 	ifd->ifi_epoch = ifp->if_epoch;
1588 	ifd->ifi_lastchange = ifp->if_lastchange;
1589 
1590 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1591 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1592 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1593 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1594 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1595 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1596 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1597 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1598 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1599 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1600 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1601 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1602 }
1603 
1604 /*
1605  * Wrapper functions for struct ifnet address list locking macros.  These are
1606  * used by kernel modules to avoid encoding programming interface or binary
1607  * interface assumptions that may be violated when kernel-internal locking
1608  * approaches change.
1609  */
1610 void
1611 if_addr_rlock(struct ifnet *ifp)
1612 {
1613 
1614 	IF_ADDR_RLOCK(ifp);
1615 }
1616 
1617 void
1618 if_addr_runlock(struct ifnet *ifp)
1619 {
1620 
1621 	IF_ADDR_RUNLOCK(ifp);
1622 }
1623 
1624 void
1625 if_maddr_rlock(if_t ifp)
1626 {
1627 
1628 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1629 }
1630 
1631 void
1632 if_maddr_runlock(if_t ifp)
1633 {
1634 
1635 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1636 }
1637 
1638 /*
1639  * Initialization, destruction and refcounting functions for ifaddrs.
1640  */
1641 struct ifaddr *
1642 ifa_alloc(size_t size, int flags)
1643 {
1644 	struct ifaddr *ifa;
1645 
1646 	KASSERT(size >= sizeof(struct ifaddr),
1647 	    ("%s: invalid size %zu", __func__, size));
1648 
1649 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1650 	if (ifa == NULL)
1651 		return (NULL);
1652 
1653 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1654 		goto fail;
1655 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1656 		goto fail;
1657 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1658 		goto fail;
1659 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1660 		goto fail;
1661 
1662 	refcount_init(&ifa->ifa_refcnt, 1);
1663 
1664 	return (ifa);
1665 
1666 fail:
1667 	/* free(NULL) is okay */
1668 	counter_u64_free(ifa->ifa_opackets);
1669 	counter_u64_free(ifa->ifa_ipackets);
1670 	counter_u64_free(ifa->ifa_obytes);
1671 	counter_u64_free(ifa->ifa_ibytes);
1672 	free(ifa, M_IFADDR);
1673 
1674 	return (NULL);
1675 }
1676 
1677 void
1678 ifa_ref(struct ifaddr *ifa)
1679 {
1680 
1681 	refcount_acquire(&ifa->ifa_refcnt);
1682 }
1683 
1684 void
1685 ifa_free(struct ifaddr *ifa)
1686 {
1687 
1688 	if (refcount_release(&ifa->ifa_refcnt)) {
1689 		counter_u64_free(ifa->ifa_opackets);
1690 		counter_u64_free(ifa->ifa_ipackets);
1691 		counter_u64_free(ifa->ifa_obytes);
1692 		counter_u64_free(ifa->ifa_ibytes);
1693 		free(ifa, M_IFADDR);
1694 	}
1695 }
1696 
1697 static int
1698 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1699     struct sockaddr *ia)
1700 {
1701 	int error;
1702 	struct rt_addrinfo info;
1703 	struct sockaddr_dl null_sdl;
1704 	struct ifnet *ifp;
1705 
1706 	ifp = ifa->ifa_ifp;
1707 
1708 	bzero(&info, sizeof(info));
1709 	if (cmd != RTM_DELETE)
1710 		info.rti_ifp = V_loif;
1711 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1712 	info.rti_info[RTAX_DST] = ia;
1713 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1714 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1715 
1716 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1717 
1718 	if (error != 0)
1719 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1720 		    __func__, otype, if_name(ifp), error);
1721 
1722 	return (error);
1723 }
1724 
1725 int
1726 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1727 {
1728 
1729 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1730 }
1731 
1732 int
1733 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1734 {
1735 
1736 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1737 }
1738 
1739 int
1740 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1741 {
1742 
1743 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1744 }
1745 
1746 /*
1747  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1748  * structs used to represent other address families, it is necessary
1749  * to perform a different comparison.
1750  */
1751 
1752 #define	sa_dl_equal(a1, a2)	\
1753 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1754 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1755 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1756 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1757 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1758 
1759 /*
1760  * Locate an interface based on a complete address.
1761  */
1762 /*ARGSUSED*/
1763 static struct ifaddr *
1764 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1765 {
1766 	struct ifnet *ifp;
1767 	struct ifaddr *ifa;
1768 
1769 	IFNET_RLOCK_NOSLEEP();
1770 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1771 		IF_ADDR_RLOCK(ifp);
1772 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1773 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1774 				continue;
1775 			if (sa_equal(addr, ifa->ifa_addr)) {
1776 				if (getref)
1777 					ifa_ref(ifa);
1778 				IF_ADDR_RUNLOCK(ifp);
1779 				goto done;
1780 			}
1781 			/* IP6 doesn't have broadcast */
1782 			if ((ifp->if_flags & IFF_BROADCAST) &&
1783 			    ifa->ifa_broadaddr &&
1784 			    ifa->ifa_broadaddr->sa_len != 0 &&
1785 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1786 				if (getref)
1787 					ifa_ref(ifa);
1788 				IF_ADDR_RUNLOCK(ifp);
1789 				goto done;
1790 			}
1791 		}
1792 		IF_ADDR_RUNLOCK(ifp);
1793 	}
1794 	ifa = NULL;
1795 done:
1796 	IFNET_RUNLOCK_NOSLEEP();
1797 	return (ifa);
1798 }
1799 
1800 struct ifaddr *
1801 ifa_ifwithaddr(const struct sockaddr *addr)
1802 {
1803 
1804 	return (ifa_ifwithaddr_internal(addr, 1));
1805 }
1806 
1807 int
1808 ifa_ifwithaddr_check(const struct sockaddr *addr)
1809 {
1810 
1811 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1812 }
1813 
1814 /*
1815  * Locate an interface based on the broadcast address.
1816  */
1817 /* ARGSUSED */
1818 struct ifaddr *
1819 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1820 {
1821 	struct ifnet *ifp;
1822 	struct ifaddr *ifa;
1823 
1824 	IFNET_RLOCK_NOSLEEP();
1825 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1826 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1827 			continue;
1828 		IF_ADDR_RLOCK(ifp);
1829 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1830 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1831 				continue;
1832 			if ((ifp->if_flags & IFF_BROADCAST) &&
1833 			    ifa->ifa_broadaddr &&
1834 			    ifa->ifa_broadaddr->sa_len != 0 &&
1835 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1836 				ifa_ref(ifa);
1837 				IF_ADDR_RUNLOCK(ifp);
1838 				goto done;
1839 			}
1840 		}
1841 		IF_ADDR_RUNLOCK(ifp);
1842 	}
1843 	ifa = NULL;
1844 done:
1845 	IFNET_RUNLOCK_NOSLEEP();
1846 	return (ifa);
1847 }
1848 
1849 /*
1850  * Locate the point to point interface with a given destination address.
1851  */
1852 /*ARGSUSED*/
1853 struct ifaddr *
1854 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1855 {
1856 	struct ifnet *ifp;
1857 	struct ifaddr *ifa;
1858 
1859 	IFNET_RLOCK_NOSLEEP();
1860 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1861 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1862 			continue;
1863 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1864 			continue;
1865 		IF_ADDR_RLOCK(ifp);
1866 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1867 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1868 				continue;
1869 			if (ifa->ifa_dstaddr != NULL &&
1870 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1871 				ifa_ref(ifa);
1872 				IF_ADDR_RUNLOCK(ifp);
1873 				goto done;
1874 			}
1875 		}
1876 		IF_ADDR_RUNLOCK(ifp);
1877 	}
1878 	ifa = NULL;
1879 done:
1880 	IFNET_RUNLOCK_NOSLEEP();
1881 	return (ifa);
1882 }
1883 
1884 /*
1885  * Find an interface on a specific network.  If many, choice
1886  * is most specific found.
1887  */
1888 struct ifaddr *
1889 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1890 {
1891 	struct ifnet *ifp;
1892 	struct ifaddr *ifa;
1893 	struct ifaddr *ifa_maybe = NULL;
1894 	u_int af = addr->sa_family;
1895 	const char *addr_data = addr->sa_data, *cplim;
1896 
1897 	/*
1898 	 * AF_LINK addresses can be looked up directly by their index number,
1899 	 * so do that if we can.
1900 	 */
1901 	if (af == AF_LINK) {
1902 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1903 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1904 		return (ifaddr_byindex(sdl->sdl_index));
1905 	}
1906 
1907 	/*
1908 	 * Scan though each interface, looking for ones that have addresses
1909 	 * in this address family and the requested fib.  Maintain a reference
1910 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1911 	 * kept it stable when we move onto the next interface.
1912 	 */
1913 	IFNET_RLOCK_NOSLEEP();
1914 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1915 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1916 			continue;
1917 		IF_ADDR_RLOCK(ifp);
1918 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1919 			const char *cp, *cp2, *cp3;
1920 
1921 			if (ifa->ifa_addr->sa_family != af)
1922 next:				continue;
1923 			if (af == AF_INET &&
1924 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1925 				/*
1926 				 * This is a bit broken as it doesn't
1927 				 * take into account that the remote end may
1928 				 * be a single node in the network we are
1929 				 * looking for.
1930 				 * The trouble is that we don't know the
1931 				 * netmask for the remote end.
1932 				 */
1933 				if (ifa->ifa_dstaddr != NULL &&
1934 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1935 					ifa_ref(ifa);
1936 					IF_ADDR_RUNLOCK(ifp);
1937 					goto done;
1938 				}
1939 			} else {
1940 				/*
1941 				 * Scan all the bits in the ifa's address.
1942 				 * If a bit dissagrees with what we are
1943 				 * looking for, mask it with the netmask
1944 				 * to see if it really matters.
1945 				 * (A byte at a time)
1946 				 */
1947 				if (ifa->ifa_netmask == 0)
1948 					continue;
1949 				cp = addr_data;
1950 				cp2 = ifa->ifa_addr->sa_data;
1951 				cp3 = ifa->ifa_netmask->sa_data;
1952 				cplim = ifa->ifa_netmask->sa_len
1953 					+ (char *)ifa->ifa_netmask;
1954 				while (cp3 < cplim)
1955 					if ((*cp++ ^ *cp2++) & *cp3++)
1956 						goto next; /* next address! */
1957 				/*
1958 				 * If the netmask of what we just found
1959 				 * is more specific than what we had before
1960 				 * (if we had one), or if the virtual status
1961 				 * of new prefix is better than of the old one,
1962 				 * then remember the new one before continuing
1963 				 * to search for an even better one.
1964 				 */
1965 				if (ifa_maybe == NULL ||
1966 				    ifa_preferred(ifa_maybe, ifa) ||
1967 				    rn_refines((caddr_t)ifa->ifa_netmask,
1968 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1969 					if (ifa_maybe != NULL)
1970 						ifa_free(ifa_maybe);
1971 					ifa_maybe = ifa;
1972 					ifa_ref(ifa_maybe);
1973 				}
1974 			}
1975 		}
1976 		IF_ADDR_RUNLOCK(ifp);
1977 	}
1978 	ifa = ifa_maybe;
1979 	ifa_maybe = NULL;
1980 done:
1981 	IFNET_RUNLOCK_NOSLEEP();
1982 	if (ifa_maybe != NULL)
1983 		ifa_free(ifa_maybe);
1984 	return (ifa);
1985 }
1986 
1987 /*
1988  * Find an interface address specific to an interface best matching
1989  * a given address.
1990  */
1991 struct ifaddr *
1992 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1993 {
1994 	struct ifaddr *ifa;
1995 	const char *cp, *cp2, *cp3;
1996 	char *cplim;
1997 	struct ifaddr *ifa_maybe = NULL;
1998 	u_int af = addr->sa_family;
1999 
2000 	if (af >= AF_MAX)
2001 		return (NULL);
2002 	IF_ADDR_RLOCK(ifp);
2003 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2004 		if (ifa->ifa_addr->sa_family != af)
2005 			continue;
2006 		if (ifa_maybe == NULL)
2007 			ifa_maybe = ifa;
2008 		if (ifa->ifa_netmask == 0) {
2009 			if (sa_equal(addr, ifa->ifa_addr) ||
2010 			    (ifa->ifa_dstaddr &&
2011 			    sa_equal(addr, ifa->ifa_dstaddr)))
2012 				goto done;
2013 			continue;
2014 		}
2015 		if (ifp->if_flags & IFF_POINTOPOINT) {
2016 			if (sa_equal(addr, ifa->ifa_dstaddr))
2017 				goto done;
2018 		} else {
2019 			cp = addr->sa_data;
2020 			cp2 = ifa->ifa_addr->sa_data;
2021 			cp3 = ifa->ifa_netmask->sa_data;
2022 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2023 			for (; cp3 < cplim; cp3++)
2024 				if ((*cp++ ^ *cp2++) & *cp3)
2025 					break;
2026 			if (cp3 == cplim)
2027 				goto done;
2028 		}
2029 	}
2030 	ifa = ifa_maybe;
2031 done:
2032 	if (ifa != NULL)
2033 		ifa_ref(ifa);
2034 	IF_ADDR_RUNLOCK(ifp);
2035 	return (ifa);
2036 }
2037 
2038 /*
2039  * See whether new ifa is better than current one:
2040  * 1) A non-virtual one is preferred over virtual.
2041  * 2) A virtual in master state preferred over any other state.
2042  *
2043  * Used in several address selecting functions.
2044  */
2045 int
2046 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2047 {
2048 
2049 	return (cur->ifa_carp && (!next->ifa_carp ||
2050 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2051 }
2052 
2053 #include <net/if_llatbl.h>
2054 
2055 /*
2056  * Default action when installing a route with a Link Level gateway.
2057  * Lookup an appropriate real ifa to point to.
2058  * This should be moved to /sys/net/link.c eventually.
2059  */
2060 static void
2061 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2062 {
2063 	struct ifaddr *ifa, *oifa;
2064 	struct sockaddr *dst;
2065 	struct ifnet *ifp;
2066 
2067 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2068 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2069 		return;
2070 	ifa = ifaof_ifpforaddr(dst, ifp);
2071 	if (ifa) {
2072 		oifa = rt->rt_ifa;
2073 		rt->rt_ifa = ifa;
2074 		ifa_free(oifa);
2075 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2076 			ifa->ifa_rtrequest(cmd, rt, info);
2077 	}
2078 }
2079 
2080 struct sockaddr_dl *
2081 link_alloc_sdl(size_t size, int flags)
2082 {
2083 
2084 	return (malloc(size, M_TEMP, flags));
2085 }
2086 
2087 void
2088 link_free_sdl(struct sockaddr *sa)
2089 {
2090 	free(sa, M_TEMP);
2091 }
2092 
2093 /*
2094  * Fills in given sdl with interface basic info.
2095  * Returns pointer to filled sdl.
2096  */
2097 struct sockaddr_dl *
2098 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2099 {
2100 	struct sockaddr_dl *sdl;
2101 
2102 	sdl = (struct sockaddr_dl *)paddr;
2103 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2104 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2105 	sdl->sdl_family = AF_LINK;
2106 	sdl->sdl_index = ifp->if_index;
2107 	sdl->sdl_type = iftype;
2108 
2109 	return (sdl);
2110 }
2111 
2112 /*
2113  * Mark an interface down and notify protocols of
2114  * the transition.
2115  */
2116 static void
2117 if_unroute(struct ifnet *ifp, int flag, int fam)
2118 {
2119 	struct ifaddr *ifa;
2120 
2121 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2122 
2123 	ifp->if_flags &= ~flag;
2124 	getmicrotime(&ifp->if_lastchange);
2125 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2126 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2127 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2128 	ifp->if_qflush(ifp);
2129 
2130 	if (ifp->if_carp)
2131 		(*carp_linkstate_p)(ifp);
2132 	rt_ifmsg(ifp);
2133 }
2134 
2135 /*
2136  * Mark an interface up and notify protocols of
2137  * the transition.
2138  */
2139 static void
2140 if_route(struct ifnet *ifp, int flag, int fam)
2141 {
2142 	struct ifaddr *ifa;
2143 
2144 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2145 
2146 	ifp->if_flags |= flag;
2147 	getmicrotime(&ifp->if_lastchange);
2148 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2149 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2150 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2151 	if (ifp->if_carp)
2152 		(*carp_linkstate_p)(ifp);
2153 	rt_ifmsg(ifp);
2154 #ifdef INET6
2155 	in6_if_up(ifp);
2156 #endif
2157 }
2158 
2159 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2160 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2161 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2162 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2163 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2164 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2165 void	*(*vlan_cookie_p)(struct ifnet *);
2166 
2167 /*
2168  * Handle a change in the interface link state. To avoid LORs
2169  * between driver lock and upper layer locks, as well as possible
2170  * recursions, we post event to taskqueue, and all job
2171  * is done in static do_link_state_change().
2172  */
2173 void
2174 if_link_state_change(struct ifnet *ifp, int link_state)
2175 {
2176 	/* Return if state hasn't changed. */
2177 	if (ifp->if_link_state == link_state)
2178 		return;
2179 
2180 	ifp->if_link_state = link_state;
2181 
2182 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2183 }
2184 
2185 static void
2186 do_link_state_change(void *arg, int pending)
2187 {
2188 	struct ifnet *ifp = (struct ifnet *)arg;
2189 	int link_state = ifp->if_link_state;
2190 	CURVNET_SET(ifp->if_vnet);
2191 
2192 	/* Notify that the link state has changed. */
2193 	rt_ifmsg(ifp);
2194 	if (ifp->if_vlantrunk != NULL)
2195 		(*vlan_link_state_p)(ifp);
2196 
2197 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2198 	    ifp->if_l2com != NULL)
2199 		(*ng_ether_link_state_p)(ifp, link_state);
2200 	if (ifp->if_carp)
2201 		(*carp_linkstate_p)(ifp);
2202 	if (ifp->if_bridge)
2203 		(*bridge_linkstate_p)(ifp);
2204 	if (ifp->if_lagg)
2205 		(*lagg_linkstate_p)(ifp, link_state);
2206 
2207 	if (IS_DEFAULT_VNET(curvnet))
2208 		devctl_notify("IFNET", ifp->if_xname,
2209 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2210 		    NULL);
2211 	if (pending > 1)
2212 		if_printf(ifp, "%d link states coalesced\n", pending);
2213 	if (log_link_state_change)
2214 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2215 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2216 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2217 	CURVNET_RESTORE();
2218 }
2219 
2220 /*
2221  * Mark an interface down and notify protocols of
2222  * the transition.
2223  */
2224 void
2225 if_down(struct ifnet *ifp)
2226 {
2227 
2228 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2229 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2230 }
2231 
2232 /*
2233  * Mark an interface up and notify protocols of
2234  * the transition.
2235  */
2236 void
2237 if_up(struct ifnet *ifp)
2238 {
2239 
2240 	if_route(ifp, IFF_UP, AF_UNSPEC);
2241 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2242 }
2243 
2244 /*
2245  * Flush an interface queue.
2246  */
2247 void
2248 if_qflush(struct ifnet *ifp)
2249 {
2250 	struct mbuf *m, *n;
2251 	struct ifaltq *ifq;
2252 
2253 	ifq = &ifp->if_snd;
2254 	IFQ_LOCK(ifq);
2255 #ifdef ALTQ
2256 	if (ALTQ_IS_ENABLED(ifq))
2257 		ALTQ_PURGE(ifq);
2258 #endif
2259 	n = ifq->ifq_head;
2260 	while ((m = n) != NULL) {
2261 		n = m->m_nextpkt;
2262 		m_freem(m);
2263 	}
2264 	ifq->ifq_head = 0;
2265 	ifq->ifq_tail = 0;
2266 	ifq->ifq_len = 0;
2267 	IFQ_UNLOCK(ifq);
2268 }
2269 
2270 /*
2271  * Map interface name to interface structure pointer, with or without
2272  * returning a reference.
2273  */
2274 struct ifnet *
2275 ifunit_ref(const char *name)
2276 {
2277 	struct ifnet *ifp;
2278 
2279 	IFNET_RLOCK_NOSLEEP();
2280 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2281 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2282 		    !(ifp->if_flags & IFF_DYING))
2283 			break;
2284 	}
2285 	if (ifp != NULL)
2286 		if_ref(ifp);
2287 	IFNET_RUNLOCK_NOSLEEP();
2288 	return (ifp);
2289 }
2290 
2291 struct ifnet *
2292 ifunit(const char *name)
2293 {
2294 	struct ifnet *ifp;
2295 
2296 	IFNET_RLOCK_NOSLEEP();
2297 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2298 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2299 			break;
2300 	}
2301 	IFNET_RUNLOCK_NOSLEEP();
2302 	return (ifp);
2303 }
2304 
2305 /*
2306  * Hardware specific interface ioctls.
2307  */
2308 static int
2309 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2310 {
2311 	struct ifreq *ifr;
2312 	int error = 0, do_ifup = 0;
2313 	int new_flags, temp_flags;
2314 	size_t namelen, onamelen;
2315 	size_t descrlen;
2316 	char *descrbuf, *odescrbuf;
2317 	char new_name[IFNAMSIZ];
2318 	struct ifaddr *ifa;
2319 	struct sockaddr_dl *sdl;
2320 
2321 	ifr = (struct ifreq *)data;
2322 	switch (cmd) {
2323 	case SIOCGIFINDEX:
2324 		ifr->ifr_index = ifp->if_index;
2325 		break;
2326 
2327 	case SIOCGIFFLAGS:
2328 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2329 		ifr->ifr_flags = temp_flags & 0xffff;
2330 		ifr->ifr_flagshigh = temp_flags >> 16;
2331 		break;
2332 
2333 	case SIOCGIFCAP:
2334 		ifr->ifr_reqcap = ifp->if_capabilities;
2335 		ifr->ifr_curcap = ifp->if_capenable;
2336 		break;
2337 
2338 #ifdef MAC
2339 	case SIOCGIFMAC:
2340 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2341 		break;
2342 #endif
2343 
2344 	case SIOCGIFMETRIC:
2345 		ifr->ifr_metric = ifp->if_metric;
2346 		break;
2347 
2348 	case SIOCGIFMTU:
2349 		ifr->ifr_mtu = ifp->if_mtu;
2350 		break;
2351 
2352 	case SIOCGIFPHYS:
2353 		/* XXXGL: did this ever worked? */
2354 		ifr->ifr_phys = 0;
2355 		break;
2356 
2357 	case SIOCGIFDESCR:
2358 		error = 0;
2359 		sx_slock(&ifdescr_sx);
2360 		if (ifp->if_description == NULL)
2361 			error = ENOMSG;
2362 		else {
2363 			/* space for terminating nul */
2364 			descrlen = strlen(ifp->if_description) + 1;
2365 			if (ifr->ifr_buffer.length < descrlen)
2366 				ifr->ifr_buffer.buffer = NULL;
2367 			else
2368 				error = copyout(ifp->if_description,
2369 				    ifr->ifr_buffer.buffer, descrlen);
2370 			ifr->ifr_buffer.length = descrlen;
2371 		}
2372 		sx_sunlock(&ifdescr_sx);
2373 		break;
2374 
2375 	case SIOCSIFDESCR:
2376 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2377 		if (error)
2378 			return (error);
2379 
2380 		/*
2381 		 * Copy only (length-1) bytes to make sure that
2382 		 * if_description is always nul terminated.  The
2383 		 * length parameter is supposed to count the
2384 		 * terminating nul in.
2385 		 */
2386 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2387 			return (ENAMETOOLONG);
2388 		else if (ifr->ifr_buffer.length == 0)
2389 			descrbuf = NULL;
2390 		else {
2391 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2392 			    M_WAITOK | M_ZERO);
2393 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2394 			    ifr->ifr_buffer.length - 1);
2395 			if (error) {
2396 				free(descrbuf, M_IFDESCR);
2397 				break;
2398 			}
2399 		}
2400 
2401 		sx_xlock(&ifdescr_sx);
2402 		odescrbuf = ifp->if_description;
2403 		ifp->if_description = descrbuf;
2404 		sx_xunlock(&ifdescr_sx);
2405 
2406 		getmicrotime(&ifp->if_lastchange);
2407 		free(odescrbuf, M_IFDESCR);
2408 		break;
2409 
2410 	case SIOCGIFFIB:
2411 		ifr->ifr_fib = ifp->if_fib;
2412 		break;
2413 
2414 	case SIOCSIFFIB:
2415 		error = priv_check(td, PRIV_NET_SETIFFIB);
2416 		if (error)
2417 			return (error);
2418 		if (ifr->ifr_fib >= rt_numfibs)
2419 			return (EINVAL);
2420 
2421 		ifp->if_fib = ifr->ifr_fib;
2422 		break;
2423 
2424 	case SIOCSIFFLAGS:
2425 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2426 		if (error)
2427 			return (error);
2428 		/*
2429 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2430 		 * check, so we don't need special handling here yet.
2431 		 */
2432 		new_flags = (ifr->ifr_flags & 0xffff) |
2433 		    (ifr->ifr_flagshigh << 16);
2434 		if (ifp->if_flags & IFF_UP &&
2435 		    (new_flags & IFF_UP) == 0) {
2436 			if_down(ifp);
2437 		} else if (new_flags & IFF_UP &&
2438 		    (ifp->if_flags & IFF_UP) == 0) {
2439 			do_ifup = 1;
2440 		}
2441 		/* See if permanently promiscuous mode bit is about to flip */
2442 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2443 			if (new_flags & IFF_PPROMISC)
2444 				ifp->if_flags |= IFF_PROMISC;
2445 			else if (ifp->if_pcount == 0)
2446 				ifp->if_flags &= ~IFF_PROMISC;
2447 			if (log_promisc_mode_change)
2448                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2449                                     ifp->if_xname,
2450                                     ((new_flags & IFF_PPROMISC) ?
2451                                      "enabled" : "disabled"));
2452 		}
2453 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2454 			(new_flags &~ IFF_CANTCHANGE);
2455 		if (ifp->if_ioctl) {
2456 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2457 		}
2458 		if (do_ifup)
2459 			if_up(ifp);
2460 		getmicrotime(&ifp->if_lastchange);
2461 		break;
2462 
2463 	case SIOCSIFCAP:
2464 		error = priv_check(td, PRIV_NET_SETIFCAP);
2465 		if (error)
2466 			return (error);
2467 		if (ifp->if_ioctl == NULL)
2468 			return (EOPNOTSUPP);
2469 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2470 			return (EINVAL);
2471 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2472 		if (error == 0)
2473 			getmicrotime(&ifp->if_lastchange);
2474 		break;
2475 
2476 #ifdef MAC
2477 	case SIOCSIFMAC:
2478 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2479 		break;
2480 #endif
2481 
2482 	case SIOCSIFNAME:
2483 		error = priv_check(td, PRIV_NET_SETIFNAME);
2484 		if (error)
2485 			return (error);
2486 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2487 		if (error != 0)
2488 			return (error);
2489 		if (new_name[0] == '\0')
2490 			return (EINVAL);
2491 		if (new_name[IFNAMSIZ-1] != '\0') {
2492 			new_name[IFNAMSIZ-1] = '\0';
2493 			if (strlen(new_name) == IFNAMSIZ-1)
2494 				return (EINVAL);
2495 		}
2496 		if (ifunit(new_name) != NULL)
2497 			return (EEXIST);
2498 
2499 		/*
2500 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2501 		 * and there are numerous other races with the
2502 		 * ifunit() checks not being atomic with namespace
2503 		 * changes (renames, vmoves, if_attach, etc).
2504 		 */
2505 		ifp->if_flags |= IFF_RENAMING;
2506 
2507 		/* Announce the departure of the interface. */
2508 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2509 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2510 
2511 		log(LOG_INFO, "%s: changing name to '%s'\n",
2512 		    ifp->if_xname, new_name);
2513 
2514 		IF_ADDR_WLOCK(ifp);
2515 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2516 		ifa = ifp->if_addr;
2517 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2518 		namelen = strlen(new_name);
2519 		onamelen = sdl->sdl_nlen;
2520 		/*
2521 		 * Move the address if needed.  This is safe because we
2522 		 * allocate space for a name of length IFNAMSIZ when we
2523 		 * create this in if_attach().
2524 		 */
2525 		if (namelen != onamelen) {
2526 			bcopy(sdl->sdl_data + onamelen,
2527 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2528 		}
2529 		bcopy(new_name, sdl->sdl_data, namelen);
2530 		sdl->sdl_nlen = namelen;
2531 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2532 		bzero(sdl->sdl_data, onamelen);
2533 		while (namelen != 0)
2534 			sdl->sdl_data[--namelen] = 0xff;
2535 		IF_ADDR_WUNLOCK(ifp);
2536 
2537 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2538 		/* Announce the return of the interface. */
2539 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2540 
2541 		ifp->if_flags &= ~IFF_RENAMING;
2542 		break;
2543 
2544 #ifdef VIMAGE
2545 	case SIOCSIFVNET:
2546 		error = priv_check(td, PRIV_NET_SETIFVNET);
2547 		if (error)
2548 			return (error);
2549 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2550 		break;
2551 #endif
2552 
2553 	case SIOCSIFMETRIC:
2554 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2555 		if (error)
2556 			return (error);
2557 		ifp->if_metric = ifr->ifr_metric;
2558 		getmicrotime(&ifp->if_lastchange);
2559 		break;
2560 
2561 	case SIOCSIFPHYS:
2562 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2563 		if (error)
2564 			return (error);
2565 		if (ifp->if_ioctl == NULL)
2566 			return (EOPNOTSUPP);
2567 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2568 		if (error == 0)
2569 			getmicrotime(&ifp->if_lastchange);
2570 		break;
2571 
2572 	case SIOCSIFMTU:
2573 	{
2574 		u_long oldmtu = ifp->if_mtu;
2575 
2576 		error = priv_check(td, PRIV_NET_SETIFMTU);
2577 		if (error)
2578 			return (error);
2579 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2580 			return (EINVAL);
2581 		if (ifp->if_ioctl == NULL)
2582 			return (EOPNOTSUPP);
2583 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2584 		if (error == 0) {
2585 			getmicrotime(&ifp->if_lastchange);
2586 			rt_ifmsg(ifp);
2587 		}
2588 		/*
2589 		 * If the link MTU changed, do network layer specific procedure.
2590 		 */
2591 		if (ifp->if_mtu != oldmtu) {
2592 #ifdef INET6
2593 			nd6_setmtu(ifp);
2594 #endif
2595 			rt_updatemtu(ifp);
2596 		}
2597 		break;
2598 	}
2599 
2600 	case SIOCADDMULTI:
2601 	case SIOCDELMULTI:
2602 		if (cmd == SIOCADDMULTI)
2603 			error = priv_check(td, PRIV_NET_ADDMULTI);
2604 		else
2605 			error = priv_check(td, PRIV_NET_DELMULTI);
2606 		if (error)
2607 			return (error);
2608 
2609 		/* Don't allow group membership on non-multicast interfaces. */
2610 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2611 			return (EOPNOTSUPP);
2612 
2613 		/* Don't let users screw up protocols' entries. */
2614 		if (ifr->ifr_addr.sa_family != AF_LINK)
2615 			return (EINVAL);
2616 
2617 		if (cmd == SIOCADDMULTI) {
2618 			struct ifmultiaddr *ifma;
2619 
2620 			/*
2621 			 * Userland is only permitted to join groups once
2622 			 * via the if_addmulti() KPI, because it cannot hold
2623 			 * struct ifmultiaddr * between calls. It may also
2624 			 * lose a race while we check if the membership
2625 			 * already exists.
2626 			 */
2627 			IF_ADDR_RLOCK(ifp);
2628 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2629 			IF_ADDR_RUNLOCK(ifp);
2630 			if (ifma != NULL)
2631 				error = EADDRINUSE;
2632 			else
2633 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2634 		} else {
2635 			error = if_delmulti(ifp, &ifr->ifr_addr);
2636 		}
2637 		if (error == 0)
2638 			getmicrotime(&ifp->if_lastchange);
2639 		break;
2640 
2641 	case SIOCSIFPHYADDR:
2642 	case SIOCDIFPHYADDR:
2643 #ifdef INET6
2644 	case SIOCSIFPHYADDR_IN6:
2645 #endif
2646 	case SIOCSIFMEDIA:
2647 	case SIOCSIFGENERIC:
2648 		error = priv_check(td, PRIV_NET_HWIOCTL);
2649 		if (error)
2650 			return (error);
2651 		if (ifp->if_ioctl == NULL)
2652 			return (EOPNOTSUPP);
2653 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2654 		if (error == 0)
2655 			getmicrotime(&ifp->if_lastchange);
2656 		break;
2657 
2658 	case SIOCGIFSTATUS:
2659 	case SIOCGIFPSRCADDR:
2660 	case SIOCGIFPDSTADDR:
2661 	case SIOCGIFMEDIA:
2662 	case SIOCGIFXMEDIA:
2663 	case SIOCGIFGENERIC:
2664 		if (ifp->if_ioctl == NULL)
2665 			return (EOPNOTSUPP);
2666 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2667 		break;
2668 
2669 	case SIOCSIFLLADDR:
2670 		error = priv_check(td, PRIV_NET_SETLLADDR);
2671 		if (error)
2672 			return (error);
2673 		error = if_setlladdr(ifp,
2674 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2675 		break;
2676 
2677 	case SIOCGHWADDR:
2678 		error = if_gethwaddr(ifp, ifr);
2679 		break;
2680 
2681 	case SIOCAIFGROUP:
2682 	{
2683 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2684 
2685 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2686 		if (error)
2687 			return (error);
2688 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2689 			return (error);
2690 		break;
2691 	}
2692 
2693 	case SIOCGIFGROUP:
2694 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2695 			return (error);
2696 		break;
2697 
2698 	case SIOCDIFGROUP:
2699 	{
2700 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2701 
2702 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2703 		if (error)
2704 			return (error);
2705 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2706 			return (error);
2707 		break;
2708 	}
2709 
2710 	default:
2711 		error = ENOIOCTL;
2712 		break;
2713 	}
2714 	return (error);
2715 }
2716 
2717 #ifdef COMPAT_FREEBSD32
2718 struct ifconf32 {
2719 	int32_t	ifc_len;
2720 	union {
2721 		uint32_t	ifcu_buf;
2722 		uint32_t	ifcu_req;
2723 	} ifc_ifcu;
2724 };
2725 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2726 #endif
2727 
2728 /*
2729  * Interface ioctls.
2730  */
2731 int
2732 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2733 {
2734 	struct ifnet *ifp;
2735 	struct ifreq *ifr;
2736 	int error;
2737 	int oif_flags;
2738 #ifdef VIMAGE
2739 	int shutdown;
2740 #endif
2741 
2742 	CURVNET_SET(so->so_vnet);
2743 #ifdef VIMAGE
2744 	/* Make sure the VNET is stable. */
2745 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2746 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2747 	if (shutdown) {
2748 		CURVNET_RESTORE();
2749 		return (EBUSY);
2750 	}
2751 #endif
2752 
2753 
2754 	switch (cmd) {
2755 	case SIOCGIFCONF:
2756 		error = ifconf(cmd, data);
2757 		CURVNET_RESTORE();
2758 		return (error);
2759 
2760 #ifdef COMPAT_FREEBSD32
2761 	case SIOCGIFCONF32:
2762 		{
2763 			struct ifconf32 *ifc32;
2764 			struct ifconf ifc;
2765 
2766 			ifc32 = (struct ifconf32 *)data;
2767 			ifc.ifc_len = ifc32->ifc_len;
2768 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2769 
2770 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2771 			CURVNET_RESTORE();
2772 			if (error == 0)
2773 				ifc32->ifc_len = ifc.ifc_len;
2774 			return (error);
2775 		}
2776 #endif
2777 	}
2778 	ifr = (struct ifreq *)data;
2779 
2780 	switch (cmd) {
2781 #ifdef VIMAGE
2782 	case SIOCSIFRVNET:
2783 		error = priv_check(td, PRIV_NET_SETIFVNET);
2784 		if (error == 0)
2785 			error = if_vmove_reclaim(td, ifr->ifr_name,
2786 			    ifr->ifr_jid);
2787 		CURVNET_RESTORE();
2788 		return (error);
2789 #endif
2790 	case SIOCIFCREATE:
2791 	case SIOCIFCREATE2:
2792 		error = priv_check(td, PRIV_NET_IFCREATE);
2793 		if (error == 0)
2794 			error = if_clone_create(ifr->ifr_name,
2795 			    sizeof(ifr->ifr_name),
2796 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2797 		CURVNET_RESTORE();
2798 		return (error);
2799 	case SIOCIFDESTROY:
2800 		error = priv_check(td, PRIV_NET_IFDESTROY);
2801 		if (error == 0)
2802 			error = if_clone_destroy(ifr->ifr_name);
2803 		CURVNET_RESTORE();
2804 		return (error);
2805 
2806 	case SIOCIFGCLONERS:
2807 		error = if_clone_list((struct if_clonereq *)data);
2808 		CURVNET_RESTORE();
2809 		return (error);
2810 	case SIOCGIFGMEMB:
2811 		error = if_getgroupmembers((struct ifgroupreq *)data);
2812 		CURVNET_RESTORE();
2813 		return (error);
2814 #if defined(INET) || defined(INET6)
2815 	case SIOCSVH:
2816 	case SIOCGVH:
2817 		if (carp_ioctl_p == NULL)
2818 			error = EPROTONOSUPPORT;
2819 		else
2820 			error = (*carp_ioctl_p)(ifr, cmd, td);
2821 		CURVNET_RESTORE();
2822 		return (error);
2823 #endif
2824 	}
2825 
2826 	ifp = ifunit_ref(ifr->ifr_name);
2827 	if (ifp == NULL) {
2828 		CURVNET_RESTORE();
2829 		return (ENXIO);
2830 	}
2831 
2832 	error = ifhwioctl(cmd, ifp, data, td);
2833 	if (error != ENOIOCTL) {
2834 		if_rele(ifp);
2835 		CURVNET_RESTORE();
2836 		return (error);
2837 	}
2838 
2839 	oif_flags = ifp->if_flags;
2840 	if (so->so_proto == NULL) {
2841 		if_rele(ifp);
2842 		CURVNET_RESTORE();
2843 		return (EOPNOTSUPP);
2844 	}
2845 
2846 	/*
2847 	 * Pass the request on to the socket control method, and if the
2848 	 * latter returns EOPNOTSUPP, directly to the interface.
2849 	 *
2850 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2851 	 * trust SIOCSIFADDR et al to come from an already privileged
2852 	 * layer, and do not perform any credentials checks or input
2853 	 * validation.
2854 	 */
2855 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2856 	    ifp, td));
2857 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2858 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2859 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2860 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2861 
2862 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2863 #ifdef INET6
2864 		if (ifp->if_flags & IFF_UP)
2865 			in6_if_up(ifp);
2866 #endif
2867 	}
2868 	if_rele(ifp);
2869 	CURVNET_RESTORE();
2870 	return (error);
2871 }
2872 
2873 /*
2874  * The code common to handling reference counted flags,
2875  * e.g., in ifpromisc() and if_allmulti().
2876  * The "pflag" argument can specify a permanent mode flag to check,
2877  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2878  *
2879  * Only to be used on stack-owned flags, not driver-owned flags.
2880  */
2881 static int
2882 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2883 {
2884 	struct ifreq ifr;
2885 	int error;
2886 	int oldflags, oldcount;
2887 
2888 	/* Sanity checks to catch programming errors */
2889 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2890 	    ("%s: setting driver-owned flag %d", __func__, flag));
2891 
2892 	if (onswitch)
2893 		KASSERT(*refcount >= 0,
2894 		    ("%s: increment negative refcount %d for flag %d",
2895 		    __func__, *refcount, flag));
2896 	else
2897 		KASSERT(*refcount > 0,
2898 		    ("%s: decrement non-positive refcount %d for flag %d",
2899 		    __func__, *refcount, flag));
2900 
2901 	/* In case this mode is permanent, just touch refcount */
2902 	if (ifp->if_flags & pflag) {
2903 		*refcount += onswitch ? 1 : -1;
2904 		return (0);
2905 	}
2906 
2907 	/* Save ifnet parameters for if_ioctl() may fail */
2908 	oldcount = *refcount;
2909 	oldflags = ifp->if_flags;
2910 
2911 	/*
2912 	 * See if we aren't the only and touching refcount is enough.
2913 	 * Actually toggle interface flag if we are the first or last.
2914 	 */
2915 	if (onswitch) {
2916 		if ((*refcount)++)
2917 			return (0);
2918 		ifp->if_flags |= flag;
2919 	} else {
2920 		if (--(*refcount))
2921 			return (0);
2922 		ifp->if_flags &= ~flag;
2923 	}
2924 
2925 	/* Call down the driver since we've changed interface flags */
2926 	if (ifp->if_ioctl == NULL) {
2927 		error = EOPNOTSUPP;
2928 		goto recover;
2929 	}
2930 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2931 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2932 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2933 	if (error)
2934 		goto recover;
2935 	/* Notify userland that interface flags have changed */
2936 	rt_ifmsg(ifp);
2937 	return (0);
2938 
2939 recover:
2940 	/* Recover after driver error */
2941 	*refcount = oldcount;
2942 	ifp->if_flags = oldflags;
2943 	return (error);
2944 }
2945 
2946 /*
2947  * Set/clear promiscuous mode on interface ifp based on the truth value
2948  * of pswitch.  The calls are reference counted so that only the first
2949  * "on" request actually has an effect, as does the final "off" request.
2950  * Results are undefined if the "off" and "on" requests are not matched.
2951  */
2952 int
2953 ifpromisc(struct ifnet *ifp, int pswitch)
2954 {
2955 	int error;
2956 	int oldflags = ifp->if_flags;
2957 
2958 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2959 			   &ifp->if_pcount, pswitch);
2960 	/* If promiscuous mode status has changed, log a message */
2961 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2962             log_promisc_mode_change)
2963 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2964 		    ifp->if_xname,
2965 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2966 	return (error);
2967 }
2968 
2969 /*
2970  * Return interface configuration
2971  * of system.  List may be used
2972  * in later ioctl's (above) to get
2973  * other information.
2974  */
2975 /*ARGSUSED*/
2976 static int
2977 ifconf(u_long cmd, caddr_t data)
2978 {
2979 	struct ifconf *ifc = (struct ifconf *)data;
2980 	struct ifnet *ifp;
2981 	struct ifaddr *ifa;
2982 	struct ifreq ifr;
2983 	struct sbuf *sb;
2984 	int error, full = 0, valid_len, max_len;
2985 
2986 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2987 	max_len = MAXPHYS - 1;
2988 
2989 	/* Prevent hostile input from being able to crash the system */
2990 	if (ifc->ifc_len <= 0)
2991 		return (EINVAL);
2992 
2993 again:
2994 	if (ifc->ifc_len <= max_len) {
2995 		max_len = ifc->ifc_len;
2996 		full = 1;
2997 	}
2998 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2999 	max_len = 0;
3000 	valid_len = 0;
3001 
3002 	IFNET_RLOCK();
3003 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3004 		int addrs;
3005 
3006 		/*
3007 		 * Zero the ifr_name buffer to make sure we don't
3008 		 * disclose the contents of the stack.
3009 		 */
3010 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
3011 
3012 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3013 		    >= sizeof(ifr.ifr_name)) {
3014 			sbuf_delete(sb);
3015 			IFNET_RUNLOCK();
3016 			return (ENAMETOOLONG);
3017 		}
3018 
3019 		addrs = 0;
3020 		IF_ADDR_RLOCK(ifp);
3021 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3022 			struct sockaddr *sa = ifa->ifa_addr;
3023 
3024 			if (prison_if(curthread->td_ucred, sa) != 0)
3025 				continue;
3026 			addrs++;
3027 			if (sa->sa_len <= sizeof(*sa)) {
3028 				ifr.ifr_addr = *sa;
3029 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3030 				max_len += sizeof(ifr);
3031 			} else {
3032 				sbuf_bcat(sb, &ifr,
3033 				    offsetof(struct ifreq, ifr_addr));
3034 				max_len += offsetof(struct ifreq, ifr_addr);
3035 				sbuf_bcat(sb, sa, sa->sa_len);
3036 				max_len += sa->sa_len;
3037 			}
3038 
3039 			if (sbuf_error(sb) == 0)
3040 				valid_len = sbuf_len(sb);
3041 		}
3042 		IF_ADDR_RUNLOCK(ifp);
3043 		if (addrs == 0) {
3044 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3045 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3046 			max_len += sizeof(ifr);
3047 
3048 			if (sbuf_error(sb) == 0)
3049 				valid_len = sbuf_len(sb);
3050 		}
3051 	}
3052 	IFNET_RUNLOCK();
3053 
3054 	/*
3055 	 * If we didn't allocate enough space (uncommon), try again.  If
3056 	 * we have already allocated as much space as we are allowed,
3057 	 * return what we've got.
3058 	 */
3059 	if (valid_len != max_len && !full) {
3060 		sbuf_delete(sb);
3061 		goto again;
3062 	}
3063 
3064 	ifc->ifc_len = valid_len;
3065 	sbuf_finish(sb);
3066 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3067 	sbuf_delete(sb);
3068 	return (error);
3069 }
3070 
3071 /*
3072  * Just like ifpromisc(), but for all-multicast-reception mode.
3073  */
3074 int
3075 if_allmulti(struct ifnet *ifp, int onswitch)
3076 {
3077 
3078 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3079 }
3080 
3081 struct ifmultiaddr *
3082 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3083 {
3084 	struct ifmultiaddr *ifma;
3085 
3086 	IF_ADDR_LOCK_ASSERT(ifp);
3087 
3088 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3089 		if (sa->sa_family == AF_LINK) {
3090 			if (sa_dl_equal(ifma->ifma_addr, sa))
3091 				break;
3092 		} else {
3093 			if (sa_equal(ifma->ifma_addr, sa))
3094 				break;
3095 		}
3096 	}
3097 
3098 	return ifma;
3099 }
3100 
3101 /*
3102  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3103  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3104  * the ifnet multicast address list here, so the caller must do that and
3105  * other setup work (such as notifying the device driver).  The reference
3106  * count is initialized to 1.
3107  */
3108 static struct ifmultiaddr *
3109 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3110     int mflags)
3111 {
3112 	struct ifmultiaddr *ifma;
3113 	struct sockaddr *dupsa;
3114 
3115 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3116 	    M_ZERO);
3117 	if (ifma == NULL)
3118 		return (NULL);
3119 
3120 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3121 	if (dupsa == NULL) {
3122 		free(ifma, M_IFMADDR);
3123 		return (NULL);
3124 	}
3125 	bcopy(sa, dupsa, sa->sa_len);
3126 	ifma->ifma_addr = dupsa;
3127 
3128 	ifma->ifma_ifp = ifp;
3129 	ifma->ifma_refcount = 1;
3130 	ifma->ifma_protospec = NULL;
3131 
3132 	if (llsa == NULL) {
3133 		ifma->ifma_lladdr = NULL;
3134 		return (ifma);
3135 	}
3136 
3137 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3138 	if (dupsa == NULL) {
3139 		free(ifma->ifma_addr, M_IFMADDR);
3140 		free(ifma, M_IFMADDR);
3141 		return (NULL);
3142 	}
3143 	bcopy(llsa, dupsa, llsa->sa_len);
3144 	ifma->ifma_lladdr = dupsa;
3145 
3146 	return (ifma);
3147 }
3148 
3149 /*
3150  * if_freemulti: free ifmultiaddr structure and possibly attached related
3151  * addresses.  The caller is responsible for implementing reference
3152  * counting, notifying the driver, handling routing messages, and releasing
3153  * any dependent link layer state.
3154  */
3155 static void
3156 if_freemulti(struct ifmultiaddr *ifma)
3157 {
3158 
3159 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3160 	    ifma->ifma_refcount));
3161 
3162 	if (ifma->ifma_lladdr != NULL)
3163 		free(ifma->ifma_lladdr, M_IFMADDR);
3164 	free(ifma->ifma_addr, M_IFMADDR);
3165 	free(ifma, M_IFMADDR);
3166 }
3167 
3168 /*
3169  * Register an additional multicast address with a network interface.
3170  *
3171  * - If the address is already present, bump the reference count on the
3172  *   address and return.
3173  * - If the address is not link-layer, look up a link layer address.
3174  * - Allocate address structures for one or both addresses, and attach to the
3175  *   multicast address list on the interface.  If automatically adding a link
3176  *   layer address, the protocol address will own a reference to the link
3177  *   layer address, to be freed when it is freed.
3178  * - Notify the network device driver of an addition to the multicast address
3179  *   list.
3180  *
3181  * 'sa' points to caller-owned memory with the desired multicast address.
3182  *
3183  * 'retifma' will be used to return a pointer to the resulting multicast
3184  * address reference, if desired.
3185  */
3186 int
3187 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3188     struct ifmultiaddr **retifma)
3189 {
3190 	struct ifmultiaddr *ifma, *ll_ifma;
3191 	struct sockaddr *llsa;
3192 	struct sockaddr_dl sdl;
3193 	int error;
3194 
3195 	/*
3196 	 * If the address is already present, return a new reference to it;
3197 	 * otherwise, allocate storage and set up a new address.
3198 	 */
3199 	IF_ADDR_WLOCK(ifp);
3200 	ifma = if_findmulti(ifp, sa);
3201 	if (ifma != NULL) {
3202 		ifma->ifma_refcount++;
3203 		if (retifma != NULL)
3204 			*retifma = ifma;
3205 		IF_ADDR_WUNLOCK(ifp);
3206 		return (0);
3207 	}
3208 
3209 	/*
3210 	 * The address isn't already present; resolve the protocol address
3211 	 * into a link layer address, and then look that up, bump its
3212 	 * refcount or allocate an ifma for that also.
3213 	 * Most link layer resolving functions returns address data which
3214 	 * fits inside default sockaddr_dl structure. However callback
3215 	 * can allocate another sockaddr structure, in that case we need to
3216 	 * free it later.
3217 	 */
3218 	llsa = NULL;
3219 	ll_ifma = NULL;
3220 	if (ifp->if_resolvemulti != NULL) {
3221 		/* Provide called function with buffer size information */
3222 		sdl.sdl_len = sizeof(sdl);
3223 		llsa = (struct sockaddr *)&sdl;
3224 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3225 		if (error)
3226 			goto unlock_out;
3227 	}
3228 
3229 	/*
3230 	 * Allocate the new address.  Don't hook it up yet, as we may also
3231 	 * need to allocate a link layer multicast address.
3232 	 */
3233 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3234 	if (ifma == NULL) {
3235 		error = ENOMEM;
3236 		goto free_llsa_out;
3237 	}
3238 
3239 	/*
3240 	 * If a link layer address is found, we'll need to see if it's
3241 	 * already present in the address list, or allocate is as well.
3242 	 * When this block finishes, the link layer address will be on the
3243 	 * list.
3244 	 */
3245 	if (llsa != NULL) {
3246 		ll_ifma = if_findmulti(ifp, llsa);
3247 		if (ll_ifma == NULL) {
3248 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3249 			if (ll_ifma == NULL) {
3250 				--ifma->ifma_refcount;
3251 				if_freemulti(ifma);
3252 				error = ENOMEM;
3253 				goto free_llsa_out;
3254 			}
3255 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3256 			    ifma_link);
3257 		} else
3258 			ll_ifma->ifma_refcount++;
3259 		ifma->ifma_llifma = ll_ifma;
3260 	}
3261 
3262 	/*
3263 	 * We now have a new multicast address, ifma, and possibly a new or
3264 	 * referenced link layer address.  Add the primary address to the
3265 	 * ifnet address list.
3266 	 */
3267 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3268 
3269 	if (retifma != NULL)
3270 		*retifma = ifma;
3271 
3272 	/*
3273 	 * Must generate the message while holding the lock so that 'ifma'
3274 	 * pointer is still valid.
3275 	 */
3276 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3277 	IF_ADDR_WUNLOCK(ifp);
3278 
3279 	/*
3280 	 * We are certain we have added something, so call down to the
3281 	 * interface to let them know about it.
3282 	 */
3283 	if (ifp->if_ioctl != NULL) {
3284 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3285 	}
3286 
3287 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3288 		link_free_sdl(llsa);
3289 
3290 	return (0);
3291 
3292 free_llsa_out:
3293 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3294 		link_free_sdl(llsa);
3295 
3296 unlock_out:
3297 	IF_ADDR_WUNLOCK(ifp);
3298 	return (error);
3299 }
3300 
3301 /*
3302  * Delete a multicast group membership by network-layer group address.
3303  *
3304  * Returns ENOENT if the entry could not be found. If ifp no longer
3305  * exists, results are undefined. This entry point should only be used
3306  * from subsystems which do appropriate locking to hold ifp for the
3307  * duration of the call.
3308  * Network-layer protocol domains must use if_delmulti_ifma().
3309  */
3310 int
3311 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3312 {
3313 	struct ifmultiaddr *ifma;
3314 	int lastref;
3315 #ifdef INVARIANTS
3316 	struct ifnet *oifp;
3317 
3318 	IFNET_RLOCK_NOSLEEP();
3319 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3320 		if (ifp == oifp)
3321 			break;
3322 	if (ifp != oifp)
3323 		ifp = NULL;
3324 	IFNET_RUNLOCK_NOSLEEP();
3325 
3326 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3327 #endif
3328 	if (ifp == NULL)
3329 		return (ENOENT);
3330 
3331 	IF_ADDR_WLOCK(ifp);
3332 	lastref = 0;
3333 	ifma = if_findmulti(ifp, sa);
3334 	if (ifma != NULL)
3335 		lastref = if_delmulti_locked(ifp, ifma, 0);
3336 	IF_ADDR_WUNLOCK(ifp);
3337 
3338 	if (ifma == NULL)
3339 		return (ENOENT);
3340 
3341 	if (lastref && ifp->if_ioctl != NULL) {
3342 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3343 	}
3344 
3345 	return (0);
3346 }
3347 
3348 /*
3349  * Delete all multicast group membership for an interface.
3350  * Should be used to quickly flush all multicast filters.
3351  */
3352 void
3353 if_delallmulti(struct ifnet *ifp)
3354 {
3355 	struct ifmultiaddr *ifma;
3356 	struct ifmultiaddr *next;
3357 
3358 	IF_ADDR_WLOCK(ifp);
3359 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3360 		if_delmulti_locked(ifp, ifma, 0);
3361 	IF_ADDR_WUNLOCK(ifp);
3362 }
3363 
3364 /*
3365  * Delete a multicast group membership by group membership pointer.
3366  * Network-layer protocol domains must use this routine.
3367  *
3368  * It is safe to call this routine if the ifp disappeared.
3369  */
3370 void
3371 if_delmulti_ifma(struct ifmultiaddr *ifma)
3372 {
3373 	struct ifnet *ifp;
3374 	int lastref;
3375 
3376 	ifp = ifma->ifma_ifp;
3377 #ifdef DIAGNOSTIC
3378 	if (ifp == NULL) {
3379 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3380 	} else {
3381 		struct ifnet *oifp;
3382 
3383 		IFNET_RLOCK_NOSLEEP();
3384 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3385 			if (ifp == oifp)
3386 				break;
3387 		if (ifp != oifp) {
3388 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3389 			ifp = NULL;
3390 		}
3391 		IFNET_RUNLOCK_NOSLEEP();
3392 	}
3393 #endif
3394 	/*
3395 	 * If and only if the ifnet instance exists: Acquire the address lock.
3396 	 */
3397 	if (ifp != NULL)
3398 		IF_ADDR_WLOCK(ifp);
3399 
3400 	lastref = if_delmulti_locked(ifp, ifma, 0);
3401 
3402 	if (ifp != NULL) {
3403 		/*
3404 		 * If and only if the ifnet instance exists:
3405 		 *  Release the address lock.
3406 		 *  If the group was left: update the hardware hash filter.
3407 		 */
3408 		IF_ADDR_WUNLOCK(ifp);
3409 		if (lastref && ifp->if_ioctl != NULL) {
3410 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3411 		}
3412 	}
3413 }
3414 
3415 /*
3416  * Perform deletion of network-layer and/or link-layer multicast address.
3417  *
3418  * Return 0 if the reference count was decremented.
3419  * Return 1 if the final reference was released, indicating that the
3420  * hardware hash filter should be reprogrammed.
3421  */
3422 static int
3423 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3424 {
3425 	struct ifmultiaddr *ll_ifma;
3426 
3427 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3428 		KASSERT(ifma->ifma_ifp == ifp,
3429 		    ("%s: inconsistent ifp %p", __func__, ifp));
3430 		IF_ADDR_WLOCK_ASSERT(ifp);
3431 	}
3432 
3433 	ifp = ifma->ifma_ifp;
3434 
3435 	/*
3436 	 * If the ifnet is detaching, null out references to ifnet,
3437 	 * so that upper protocol layers will notice, and not attempt
3438 	 * to obtain locks for an ifnet which no longer exists. The
3439 	 * routing socket announcement must happen before the ifnet
3440 	 * instance is detached from the system.
3441 	 */
3442 	if (detaching) {
3443 #ifdef DIAGNOSTIC
3444 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3445 #endif
3446 		/*
3447 		 * ifp may already be nulled out if we are being reentered
3448 		 * to delete the ll_ifma.
3449 		 */
3450 		if (ifp != NULL) {
3451 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3452 			ifma->ifma_ifp = NULL;
3453 		}
3454 	}
3455 
3456 	if (--ifma->ifma_refcount > 0)
3457 		return 0;
3458 
3459 	/*
3460 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3461 	 * have been associated with it. Release it first if so.
3462 	 */
3463 	ll_ifma = ifma->ifma_llifma;
3464 	if (ll_ifma != NULL) {
3465 		KASSERT(ifma->ifma_lladdr != NULL,
3466 		    ("%s: llifma w/o lladdr", __func__));
3467 		if (detaching)
3468 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3469 		if (--ll_ifma->ifma_refcount == 0) {
3470 			if (ifp != NULL) {
3471 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3472 				    ifma_link);
3473 			}
3474 			if_freemulti(ll_ifma);
3475 		}
3476 	}
3477 
3478 	if (ifp != NULL)
3479 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3480 
3481 	if_freemulti(ifma);
3482 
3483 	/*
3484 	 * The last reference to this instance of struct ifmultiaddr
3485 	 * was released; the hardware should be notified of this change.
3486 	 */
3487 	return 1;
3488 }
3489 
3490 /*
3491  * Set the link layer address on an interface.
3492  *
3493  * At this time we only support certain types of interfaces,
3494  * and we don't allow the length of the address to change.
3495  *
3496  * Set noinline to be dtrace-friendly
3497  */
3498 __noinline int
3499 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3500 {
3501 	struct sockaddr_dl *sdl;
3502 	struct ifaddr *ifa;
3503 	struct ifreq ifr;
3504 
3505 	IF_ADDR_RLOCK(ifp);
3506 	ifa = ifp->if_addr;
3507 	if (ifa == NULL) {
3508 		IF_ADDR_RUNLOCK(ifp);
3509 		return (EINVAL);
3510 	}
3511 	ifa_ref(ifa);
3512 	IF_ADDR_RUNLOCK(ifp);
3513 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3514 	if (sdl == NULL) {
3515 		ifa_free(ifa);
3516 		return (EINVAL);
3517 	}
3518 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3519 		ifa_free(ifa);
3520 		return (EINVAL);
3521 	}
3522 	switch (ifp->if_type) {
3523 	case IFT_ETHER:
3524 	case IFT_FDDI:
3525 	case IFT_XETHER:
3526 	case IFT_ISO88025:
3527 	case IFT_L2VLAN:
3528 	case IFT_BRIDGE:
3529 	case IFT_ARCNET:
3530 	case IFT_IEEE8023ADLAG:
3531 		bcopy(lladdr, LLADDR(sdl), len);
3532 		ifa_free(ifa);
3533 		break;
3534 	default:
3535 		ifa_free(ifa);
3536 		return (ENODEV);
3537 	}
3538 
3539 	/*
3540 	 * If the interface is already up, we need
3541 	 * to re-init it in order to reprogram its
3542 	 * address filter.
3543 	 */
3544 	if ((ifp->if_flags & IFF_UP) != 0) {
3545 		if (ifp->if_ioctl) {
3546 			ifp->if_flags &= ~IFF_UP;
3547 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3548 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3549 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3550 			ifp->if_flags |= IFF_UP;
3551 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3552 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3553 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3554 		}
3555 	}
3556 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3557 	return (0);
3558 }
3559 
3560 /*
3561  * Compat function for handling basic encapsulation requests.
3562  * Not converted stacks (FDDI, IB, ..) supports traditional
3563  * output model: ARP (and other similar L2 protocols) are handled
3564  * inside output routine, arpresolve/nd6_resolve() returns MAC
3565  * address instead of full prepend.
3566  *
3567  * This function creates calculated header==MAC for IPv4/IPv6 and
3568  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3569  * address families.
3570  */
3571 static int
3572 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3573 {
3574 
3575 	if (req->rtype != IFENCAP_LL)
3576 		return (EOPNOTSUPP);
3577 
3578 	if (req->bufsize < req->lladdr_len)
3579 		return (ENOMEM);
3580 
3581 	switch (req->family) {
3582 	case AF_INET:
3583 	case AF_INET6:
3584 		break;
3585 	default:
3586 		return (EAFNOSUPPORT);
3587 	}
3588 
3589 	/* Copy lladdr to storage as is */
3590 	memmove(req->buf, req->lladdr, req->lladdr_len);
3591 	req->bufsize = req->lladdr_len;
3592 	req->lladdr_off = 0;
3593 
3594 	return (0);
3595 }
3596 
3597 /*
3598  * Get the link layer address that was read from the hardware at attach.
3599  *
3600  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3601  * their component interfaces as IFT_IEEE8023ADLAG.
3602  */
3603 int
3604 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3605 {
3606 
3607 	if (ifp->if_hw_addr == NULL)
3608 		return (ENODEV);
3609 
3610 	switch (ifp->if_type) {
3611 	case IFT_ETHER:
3612 	case IFT_IEEE8023ADLAG:
3613 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3614 		return (0);
3615 	default:
3616 		return (ENODEV);
3617 	}
3618 }
3619 
3620 /*
3621  * The name argument must be a pointer to storage which will last as
3622  * long as the interface does.  For physical devices, the result of
3623  * device_get_name(dev) is a good choice and for pseudo-devices a
3624  * static string works well.
3625  */
3626 void
3627 if_initname(struct ifnet *ifp, const char *name, int unit)
3628 {
3629 	ifp->if_dname = name;
3630 	ifp->if_dunit = unit;
3631 	if (unit != IF_DUNIT_NONE)
3632 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3633 	else
3634 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3635 }
3636 
3637 int
3638 if_printf(struct ifnet *ifp, const char * fmt, ...)
3639 {
3640 	va_list ap;
3641 	int retval;
3642 
3643 	retval = printf("%s: ", ifp->if_xname);
3644 	va_start(ap, fmt);
3645 	retval += vprintf(fmt, ap);
3646 	va_end(ap);
3647 	return (retval);
3648 }
3649 
3650 void
3651 if_start(struct ifnet *ifp)
3652 {
3653 
3654 	(*(ifp)->if_start)(ifp);
3655 }
3656 
3657 /*
3658  * Backwards compatibility interface for drivers
3659  * that have not implemented it
3660  */
3661 static int
3662 if_transmit(struct ifnet *ifp, struct mbuf *m)
3663 {
3664 	int error;
3665 
3666 	IFQ_HANDOFF(ifp, m, error);
3667 	return (error);
3668 }
3669 
3670 static void
3671 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3672 {
3673 
3674 	m_freem(m);
3675 }
3676 
3677 int
3678 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3679 {
3680 	int active = 0;
3681 
3682 	IF_LOCK(ifq);
3683 	if (_IF_QFULL(ifq)) {
3684 		IF_UNLOCK(ifq);
3685 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3686 		m_freem(m);
3687 		return (0);
3688 	}
3689 	if (ifp != NULL) {
3690 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3691 		if (m->m_flags & (M_BCAST|M_MCAST))
3692 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3693 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3694 	}
3695 	_IF_ENQUEUE(ifq, m);
3696 	IF_UNLOCK(ifq);
3697 	if (ifp != NULL && !active)
3698 		(*(ifp)->if_start)(ifp);
3699 	return (1);
3700 }
3701 
3702 void
3703 if_register_com_alloc(u_char type,
3704     if_com_alloc_t *a, if_com_free_t *f)
3705 {
3706 
3707 	KASSERT(if_com_alloc[type] == NULL,
3708 	    ("if_register_com_alloc: %d already registered", type));
3709 	KASSERT(if_com_free[type] == NULL,
3710 	    ("if_register_com_alloc: %d free already registered", type));
3711 
3712 	if_com_alloc[type] = a;
3713 	if_com_free[type] = f;
3714 }
3715 
3716 void
3717 if_deregister_com_alloc(u_char type)
3718 {
3719 
3720 	KASSERT(if_com_alloc[type] != NULL,
3721 	    ("if_deregister_com_alloc: %d not registered", type));
3722 	KASSERT(if_com_free[type] != NULL,
3723 	    ("if_deregister_com_alloc: %d free not registered", type));
3724 	if_com_alloc[type] = NULL;
3725 	if_com_free[type] = NULL;
3726 }
3727 
3728 /* API for driver access to network stack owned ifnet.*/
3729 uint64_t
3730 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3731 {
3732 	uint64_t oldbrate;
3733 
3734 	oldbrate = ifp->if_baudrate;
3735 	ifp->if_baudrate = baudrate;
3736 	return (oldbrate);
3737 }
3738 
3739 uint64_t
3740 if_getbaudrate(if_t ifp)
3741 {
3742 
3743 	return (((struct ifnet *)ifp)->if_baudrate);
3744 }
3745 
3746 int
3747 if_setcapabilities(if_t ifp, int capabilities)
3748 {
3749 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3750 	return (0);
3751 }
3752 
3753 int
3754 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3755 {
3756 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3757 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3758 
3759 	return (0);
3760 }
3761 
3762 int
3763 if_getcapabilities(if_t ifp)
3764 {
3765 	return ((struct ifnet *)ifp)->if_capabilities;
3766 }
3767 
3768 int
3769 if_setcapenable(if_t ifp, int capabilities)
3770 {
3771 	((struct ifnet *)ifp)->if_capenable = capabilities;
3772 	return (0);
3773 }
3774 
3775 int
3776 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3777 {
3778 	if(setcap)
3779 		((struct ifnet *)ifp)->if_capenable |= setcap;
3780 	if(clearcap)
3781 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3782 
3783 	return (0);
3784 }
3785 
3786 const char *
3787 if_getdname(if_t ifp)
3788 {
3789 	return ((struct ifnet *)ifp)->if_dname;
3790 }
3791 
3792 int
3793 if_togglecapenable(if_t ifp, int togglecap)
3794 {
3795 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3796 	return (0);
3797 }
3798 
3799 int
3800 if_getcapenable(if_t ifp)
3801 {
3802 	return ((struct ifnet *)ifp)->if_capenable;
3803 }
3804 
3805 /*
3806  * This is largely undesirable because it ties ifnet to a device, but does
3807  * provide flexiblity for an embedded product vendor. Should be used with
3808  * the understanding that it violates the interface boundaries, and should be
3809  * a last resort only.
3810  */
3811 int
3812 if_setdev(if_t ifp, void *dev)
3813 {
3814 	return (0);
3815 }
3816 
3817 int
3818 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3819 {
3820 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3821 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3822 
3823 	return (0);
3824 }
3825 
3826 int
3827 if_getdrvflags(if_t ifp)
3828 {
3829 	return ((struct ifnet *)ifp)->if_drv_flags;
3830 }
3831 
3832 int
3833 if_setdrvflags(if_t ifp, int flags)
3834 {
3835 	((struct ifnet *)ifp)->if_drv_flags = flags;
3836 	return (0);
3837 }
3838 
3839 
3840 int
3841 if_setflags(if_t ifp, int flags)
3842 {
3843 	((struct ifnet *)ifp)->if_flags = flags;
3844 	return (0);
3845 }
3846 
3847 int
3848 if_setflagbits(if_t ifp, int set, int clear)
3849 {
3850 	((struct ifnet *)ifp)->if_flags |= set;
3851 	((struct ifnet *)ifp)->if_flags &= ~clear;
3852 
3853 	return (0);
3854 }
3855 
3856 int
3857 if_getflags(if_t ifp)
3858 {
3859 	return ((struct ifnet *)ifp)->if_flags;
3860 }
3861 
3862 int
3863 if_clearhwassist(if_t ifp)
3864 {
3865 	((struct ifnet *)ifp)->if_hwassist = 0;
3866 	return (0);
3867 }
3868 
3869 int
3870 if_sethwassistbits(if_t ifp, int toset, int toclear)
3871 {
3872 	((struct ifnet *)ifp)->if_hwassist |= toset;
3873 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3874 
3875 	return (0);
3876 }
3877 
3878 int
3879 if_sethwassist(if_t ifp, int hwassist_bit)
3880 {
3881 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3882 	return (0);
3883 }
3884 
3885 int
3886 if_gethwassist(if_t ifp)
3887 {
3888 	return ((struct ifnet *)ifp)->if_hwassist;
3889 }
3890 
3891 int
3892 if_setmtu(if_t ifp, int mtu)
3893 {
3894 	((struct ifnet *)ifp)->if_mtu = mtu;
3895 	return (0);
3896 }
3897 
3898 int
3899 if_getmtu(if_t ifp)
3900 {
3901 	return ((struct ifnet *)ifp)->if_mtu;
3902 }
3903 
3904 int
3905 if_getmtu_family(if_t ifp, int family)
3906 {
3907 	struct domain *dp;
3908 
3909 	for (dp = domains; dp; dp = dp->dom_next) {
3910 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3911 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3912 	}
3913 
3914 	return (((struct ifnet *)ifp)->if_mtu);
3915 }
3916 
3917 int
3918 if_setsoftc(if_t ifp, void *softc)
3919 {
3920 	((struct ifnet *)ifp)->if_softc = softc;
3921 	return (0);
3922 }
3923 
3924 void *
3925 if_getsoftc(if_t ifp)
3926 {
3927 	return ((struct ifnet *)ifp)->if_softc;
3928 }
3929 
3930 void
3931 if_setrcvif(struct mbuf *m, if_t ifp)
3932 {
3933 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3934 }
3935 
3936 void
3937 if_setvtag(struct mbuf *m, uint16_t tag)
3938 {
3939 	m->m_pkthdr.ether_vtag = tag;
3940 }
3941 
3942 uint16_t
3943 if_getvtag(struct mbuf *m)
3944 {
3945 
3946 	return (m->m_pkthdr.ether_vtag);
3947 }
3948 
3949 int
3950 if_sendq_empty(if_t ifp)
3951 {
3952 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3953 }
3954 
3955 struct ifaddr *
3956 if_getifaddr(if_t ifp)
3957 {
3958 	return ((struct ifnet *)ifp)->if_addr;
3959 }
3960 
3961 int
3962 if_getamcount(if_t ifp)
3963 {
3964 	return ((struct ifnet *)ifp)->if_amcount;
3965 }
3966 
3967 
3968 int
3969 if_setsendqready(if_t ifp)
3970 {
3971 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3972 	return (0);
3973 }
3974 
3975 int
3976 if_setsendqlen(if_t ifp, int tx_desc_count)
3977 {
3978 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3979 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3980 
3981 	return (0);
3982 }
3983 
3984 int
3985 if_vlantrunkinuse(if_t ifp)
3986 {
3987 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3988 }
3989 
3990 int
3991 if_input(if_t ifp, struct mbuf* sendmp)
3992 {
3993 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3994 	return (0);
3995 
3996 }
3997 
3998 /* XXX */
3999 #ifndef ETH_ADDR_LEN
4000 #define ETH_ADDR_LEN 6
4001 #endif
4002 
4003 int
4004 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4005 {
4006 	struct ifmultiaddr *ifma;
4007 	uint8_t *lmta = (uint8_t *)mta;
4008 	int mcnt = 0;
4009 
4010 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4011 		if (ifma->ifma_addr->sa_family != AF_LINK)
4012 			continue;
4013 
4014 		if (mcnt == max)
4015 			break;
4016 
4017 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4018 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4019 		mcnt++;
4020 	}
4021 	*cnt = mcnt;
4022 
4023 	return (0);
4024 }
4025 
4026 int
4027 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4028 {
4029 	int error;
4030 
4031 	if_maddr_rlock(ifp);
4032 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4033 	if_maddr_runlock(ifp);
4034 	return (error);
4035 }
4036 
4037 int
4038 if_multiaddr_count(if_t ifp, int max)
4039 {
4040 	struct ifmultiaddr *ifma;
4041 	int count;
4042 
4043 	count = 0;
4044 	if_maddr_rlock(ifp);
4045 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4046 		if (ifma->ifma_addr->sa_family != AF_LINK)
4047 			continue;
4048 		count++;
4049 		if (count == max)
4050 			break;
4051 	}
4052 	if_maddr_runlock(ifp);
4053 	return (count);
4054 }
4055 
4056 int
4057 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4058 {
4059 	struct ifmultiaddr *ifma;
4060 	int cnt = 0;
4061 
4062 	if_maddr_rlock(ifp);
4063 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4064 		cnt += filter(arg, ifma, cnt);
4065 	if_maddr_runlock(ifp);
4066 	return (cnt);
4067 }
4068 
4069 struct mbuf *
4070 if_dequeue(if_t ifp)
4071 {
4072 	struct mbuf *m;
4073 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4074 
4075 	return (m);
4076 }
4077 
4078 int
4079 if_sendq_prepend(if_t ifp, struct mbuf *m)
4080 {
4081 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4082 	return (0);
4083 }
4084 
4085 int
4086 if_setifheaderlen(if_t ifp, int len)
4087 {
4088 	((struct ifnet *)ifp)->if_hdrlen = len;
4089 	return (0);
4090 }
4091 
4092 caddr_t
4093 if_getlladdr(if_t ifp)
4094 {
4095 	return (IF_LLADDR((struct ifnet *)ifp));
4096 }
4097 
4098 void *
4099 if_gethandle(u_char type)
4100 {
4101 	return (if_alloc(type));
4102 }
4103 
4104 void
4105 if_bpfmtap(if_t ifh, struct mbuf *m)
4106 {
4107 	struct ifnet *ifp = (struct ifnet *)ifh;
4108 
4109 	BPF_MTAP(ifp, m);
4110 }
4111 
4112 void
4113 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4114 {
4115 	struct ifnet *ifp = (struct ifnet *)ifh;
4116 
4117 	ETHER_BPF_MTAP(ifp, m);
4118 }
4119 
4120 void
4121 if_vlancap(if_t ifh)
4122 {
4123 	struct ifnet *ifp = (struct ifnet *)ifh;
4124 	VLAN_CAPABILITIES(ifp);
4125 }
4126 
4127 int
4128 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4129 {
4130 
4131 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4132         return (0);
4133 }
4134 
4135 int
4136 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4137 {
4138 
4139 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4140         return (0);
4141 }
4142 
4143 int
4144 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4145 {
4146 
4147 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4148         return (0);
4149 }
4150 
4151 u_int
4152 if_gethwtsomax(if_t ifp)
4153 {
4154 
4155 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4156 }
4157 
4158 u_int
4159 if_gethwtsomaxsegcount(if_t ifp)
4160 {
4161 
4162 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4163 }
4164 
4165 u_int
4166 if_gethwtsomaxsegsize(if_t ifp)
4167 {
4168 
4169 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4170 }
4171 
4172 void
4173 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4174 {
4175 	((struct ifnet *)ifp)->if_init = init_fn;
4176 }
4177 
4178 void
4179 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4180 {
4181 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4182 }
4183 
4184 void
4185 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4186 {
4187 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4188 }
4189 
4190 void
4191 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4192 {
4193 	((struct ifnet *)ifp)->if_transmit = start_fn;
4194 }
4195 
4196 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4197 {
4198 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4199 
4200 }
4201 
4202 void
4203 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4204 {
4205 
4206 	ifp->if_get_counter = fn;
4207 }
4208 
4209 /* Revisit these - These are inline functions originally. */
4210 int
4211 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4212 {
4213 	return drbr_inuse(ifh, br);
4214 }
4215 
4216 struct mbuf*
4217 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4218 {
4219 	return drbr_dequeue(ifh, br);
4220 }
4221 
4222 int
4223 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4224 {
4225 	return drbr_needs_enqueue(ifh, br);
4226 }
4227 
4228 int
4229 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4230 {
4231 	return drbr_enqueue(ifh, br, m);
4232 
4233 }
4234