xref: /freebsd/sys/net/if.c (revision 2f68cdb94454035989fa370bcb68db2f4e96b1ed)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_compat.h"
36 #include "opt_inet6.h"
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/conf.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysent.h>
61 #include <sys/taskqueue.h>
62 #include <sys/domain.h>
63 #include <sys/jail.h>
64 #include <sys/priv.h>
65 
66 #include <machine/stdarg.h>
67 #include <vm/uma.h>
68 
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if.h>
72 #include <net/if_arp.h>
73 #include <net/if_clone.h>
74 #include <net/if_dl.h>
75 #include <net/if_types.h>
76 #include <net/if_var.h>
77 #include <net/if_media.h>
78 #include <net/if_vlan_var.h>
79 #include <net/radix.h>
80 #include <net/route.h>
81 #include <net/vnet.h>
82 
83 #if defined(INET) || defined(INET6)
84 #include <net/ethernet.h>
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip_carp.h>
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #endif /* INET */
92 #ifdef INET6
93 #include <netinet6/in6_var.h>
94 #include <netinet6/in6_ifattach.h>
95 #endif /* INET6 */
96 #endif /* INET || INET6 */
97 
98 #include <security/mac/mac_framework.h>
99 
100 #ifdef COMPAT_FREEBSD32
101 #include <sys/mount.h>
102 #include <compat/freebsd32/freebsd32.h>
103 
104 struct ifreq_buffer32 {
105 	uint32_t	length;		/* (size_t) */
106 	uint32_t	buffer;		/* (void *) */
107 };
108 
109 /*
110  * Interface request structure used for socket
111  * ioctl's.  All interface ioctl's must have parameter
112  * definitions which begin with ifr_name.  The
113  * remainder may be interface specific.
114  */
115 struct ifreq32 {
116 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
117 	union {
118 		struct sockaddr	ifru_addr;
119 		struct sockaddr	ifru_dstaddr;
120 		struct sockaddr	ifru_broadaddr;
121 		struct ifreq_buffer32 ifru_buffer;
122 		short		ifru_flags[2];
123 		short		ifru_index;
124 		int		ifru_jid;
125 		int		ifru_metric;
126 		int		ifru_mtu;
127 		int		ifru_phys;
128 		int		ifru_media;
129 		uint32_t	ifru_data;
130 		int		ifru_cap[2];
131 		u_int		ifru_fib;
132 		u_char		ifru_vlan_pcp;
133 	} ifr_ifru;
134 };
135 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
136 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
137     __offsetof(struct ifreq32, ifr_ifru));
138 #endif
139 
140 union ifreq_union {
141 	struct ifreq	ifr;
142 #ifdef COMPAT_FREEBSD32
143 	struct ifreq32	ifr32;
144 #endif
145 };
146 
147 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
148 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
149 
150 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
151     &ifqmaxlen, 0, "max send queue size");
152 
153 /* Log link state change events */
154 static int log_link_state_change = 1;
155 
156 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
157 	&log_link_state_change, 0,
158 	"log interface link state change events");
159 
160 /* Log promiscuous mode change events */
161 static int log_promisc_mode_change = 1;
162 
163 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
164 	&log_promisc_mode_change, 1,
165 	"log promiscuous mode change events");
166 
167 /* Interface description */
168 static unsigned int ifdescr_maxlen = 1024;
169 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
170 	&ifdescr_maxlen, 0,
171 	"administrative maximum length for interface description");
172 
173 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
174 
175 /* global sx for non-critical path ifdescr */
176 static struct sx ifdescr_sx;
177 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
178 
179 void	(*bridge_linkstate_p)(struct ifnet *ifp);
180 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
181 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
182 /* These are external hooks for CARP. */
183 void	(*carp_linkstate_p)(struct ifnet *ifp);
184 void	(*carp_demote_adj_p)(int, char *);
185 int	(*carp_master_p)(struct ifaddr *);
186 #if defined(INET) || defined(INET6)
187 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
188 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
189     const struct sockaddr *sa);
190 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
191 int	(*carp_attach_p)(struct ifaddr *, int);
192 void	(*carp_detach_p)(struct ifaddr *, bool);
193 #endif
194 #ifdef INET
195 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
196 #endif
197 #ifdef INET6
198 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
199 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
200     const struct in6_addr *taddr);
201 #endif
202 
203 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
204 
205 /*
206  * XXX: Style; these should be sorted alphabetically, and unprototyped
207  * static functions should be prototyped. Currently they are sorted by
208  * declaration order.
209  */
210 static void	if_attachdomain(void *);
211 static void	if_attachdomain1(struct ifnet *);
212 static int	ifconf(u_long, caddr_t);
213 static void	if_freemulti(struct ifmultiaddr *);
214 static void	if_grow(void);
215 static void	if_input_default(struct ifnet *, struct mbuf *);
216 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
217 static void	if_route(struct ifnet *, int flag, int fam);
218 static int	if_setflag(struct ifnet *, int, int, int *, int);
219 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
220 static void	if_unroute(struct ifnet *, int flag, int fam);
221 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
222 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
223 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
224 static void	do_link_state_change(void *, int);
225 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
226 static int	if_getgroupmembers(struct ifgroupreq *);
227 static void	if_delgroups(struct ifnet *);
228 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
229 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
230 #ifdef VIMAGE
231 static void	if_vmove(struct ifnet *, struct vnet *);
232 #endif
233 
234 #ifdef INET6
235 /*
236  * XXX: declare here to avoid to include many inet6 related files..
237  * should be more generalized?
238  */
239 extern void	nd6_setmtu(struct ifnet *);
240 #endif
241 
242 /* ipsec helper hooks */
243 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
244 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
245 
246 VNET_DEFINE(int, if_index);
247 int	ifqmaxlen = IFQ_MAXLEN;
248 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
249 VNET_DEFINE(struct ifgrouphead, ifg_head);
250 
251 static VNET_DEFINE(int, if_indexlim) = 8;
252 
253 /* Table of ifnet by index. */
254 VNET_DEFINE(struct ifnet **, ifindex_table);
255 
256 #define	V_if_indexlim		VNET(if_indexlim)
257 #define	V_ifindex_table		VNET(ifindex_table)
258 
259 /*
260  * The global network interface list (V_ifnet) and related state (such as
261  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
262  * an rwlock.  Either may be acquired shared to stablize the list, but both
263  * must be acquired writable to modify the list.  This model allows us to
264  * both stablize the interface list during interrupt thread processing, but
265  * also to stablize it over long-running ioctls, without introducing priority
266  * inversions and deadlocks.
267  */
268 struct rwlock ifnet_rwlock;
269 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
270 struct sx ifnet_sxlock;
271 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
272 
273 /*
274  * The allocation of network interfaces is a rather non-atomic affair; we
275  * need to select an index before we are ready to expose the interface for
276  * use, so will use this pointer value to indicate reservation.
277  */
278 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
279 
280 static	if_com_alloc_t *if_com_alloc[256];
281 static	if_com_free_t *if_com_free[256];
282 
283 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
284 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
285 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
286 
287 struct ifnet *
288 ifnet_byindex_locked(u_short idx)
289 {
290 
291 	if (idx > V_if_index)
292 		return (NULL);
293 	if (V_ifindex_table[idx] == IFNET_HOLD)
294 		return (NULL);
295 	return (V_ifindex_table[idx]);
296 }
297 
298 struct ifnet *
299 ifnet_byindex(u_short idx)
300 {
301 	struct ifnet *ifp;
302 
303 	IFNET_RLOCK_NOSLEEP();
304 	ifp = ifnet_byindex_locked(idx);
305 	IFNET_RUNLOCK_NOSLEEP();
306 	return (ifp);
307 }
308 
309 struct ifnet *
310 ifnet_byindex_ref(u_short idx)
311 {
312 	struct ifnet *ifp;
313 
314 	IFNET_RLOCK_NOSLEEP();
315 	ifp = ifnet_byindex_locked(idx);
316 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
317 		IFNET_RUNLOCK_NOSLEEP();
318 		return (NULL);
319 	}
320 	if_ref(ifp);
321 	IFNET_RUNLOCK_NOSLEEP();
322 	return (ifp);
323 }
324 
325 /*
326  * Allocate an ifindex array entry; return 0 on success or an error on
327  * failure.
328  */
329 static u_short
330 ifindex_alloc(void)
331 {
332 	u_short idx;
333 
334 	IFNET_WLOCK_ASSERT();
335 retry:
336 	/*
337 	 * Try to find an empty slot below V_if_index.  If we fail, take the
338 	 * next slot.
339 	 */
340 	for (idx = 1; idx <= V_if_index; idx++) {
341 		if (V_ifindex_table[idx] == NULL)
342 			break;
343 	}
344 
345 	/* Catch if_index overflow. */
346 	if (idx >= V_if_indexlim) {
347 		if_grow();
348 		goto retry;
349 	}
350 	if (idx > V_if_index)
351 		V_if_index = idx;
352 	return (idx);
353 }
354 
355 static void
356 ifindex_free_locked(u_short idx)
357 {
358 
359 	IFNET_WLOCK_ASSERT();
360 
361 	V_ifindex_table[idx] = NULL;
362 	while (V_if_index > 0 &&
363 	    V_ifindex_table[V_if_index] == NULL)
364 		V_if_index--;
365 }
366 
367 static void
368 ifindex_free(u_short idx)
369 {
370 
371 	IFNET_WLOCK();
372 	ifindex_free_locked(idx);
373 	IFNET_WUNLOCK();
374 }
375 
376 static void
377 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
378 {
379 
380 	IFNET_WLOCK_ASSERT();
381 
382 	V_ifindex_table[idx] = ifp;
383 }
384 
385 static void
386 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
387 {
388 
389 	IFNET_WLOCK();
390 	ifnet_setbyindex_locked(idx, ifp);
391 	IFNET_WUNLOCK();
392 }
393 
394 struct ifaddr *
395 ifaddr_byindex(u_short idx)
396 {
397 	struct ifnet *ifp;
398 	struct ifaddr *ifa = NULL;
399 
400 	IFNET_RLOCK_NOSLEEP();
401 	ifp = ifnet_byindex_locked(idx);
402 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
403 		ifa_ref(ifa);
404 	IFNET_RUNLOCK_NOSLEEP();
405 	return (ifa);
406 }
407 
408 /*
409  * Network interface utility routines.
410  *
411  * Routines with ifa_ifwith* names take sockaddr *'s as
412  * parameters.
413  */
414 
415 static void
416 vnet_if_init(const void *unused __unused)
417 {
418 
419 	TAILQ_INIT(&V_ifnet);
420 	TAILQ_INIT(&V_ifg_head);
421 	IFNET_WLOCK();
422 	if_grow();				/* create initial table */
423 	IFNET_WUNLOCK();
424 	vnet_if_clone_init();
425 }
426 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
427     NULL);
428 
429 #ifdef VIMAGE
430 static void
431 vnet_if_uninit(const void *unused __unused)
432 {
433 
434 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
435 	    "not empty", __func__, __LINE__, &V_ifnet));
436 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
437 	    "not empty", __func__, __LINE__, &V_ifg_head));
438 
439 	free((caddr_t)V_ifindex_table, M_IFNET);
440 }
441 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
442     vnet_if_uninit, NULL);
443 
444 static void
445 vnet_if_return(const void *unused __unused)
446 {
447 	struct ifnet *ifp, *nifp;
448 
449 	/* Return all inherited interfaces to their parent vnets. */
450 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
451 		if (ifp->if_home_vnet != ifp->if_vnet)
452 			if_vmove(ifp, ifp->if_home_vnet);
453 	}
454 }
455 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
456     vnet_if_return, NULL);
457 #endif
458 
459 static void
460 if_grow(void)
461 {
462 	int oldlim;
463 	u_int n;
464 	struct ifnet **e;
465 
466 	IFNET_WLOCK_ASSERT();
467 	oldlim = V_if_indexlim;
468 	IFNET_WUNLOCK();
469 	n = (oldlim << 1) * sizeof(*e);
470 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
471 	IFNET_WLOCK();
472 	if (V_if_indexlim != oldlim) {
473 		free(e, M_IFNET);
474 		return;
475 	}
476 	if (V_ifindex_table != NULL) {
477 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
478 		free((caddr_t)V_ifindex_table, M_IFNET);
479 	}
480 	V_if_indexlim <<= 1;
481 	V_ifindex_table = e;
482 }
483 
484 /*
485  * Allocate a struct ifnet and an index for an interface.  A layer 2
486  * common structure will also be allocated if an allocation routine is
487  * registered for the passed type.
488  */
489 struct ifnet *
490 if_alloc(u_char type)
491 {
492 	struct ifnet *ifp;
493 	u_short idx;
494 
495 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
496 	IFNET_WLOCK();
497 	idx = ifindex_alloc();
498 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
499 	IFNET_WUNLOCK();
500 	ifp->if_index = idx;
501 	ifp->if_type = type;
502 	ifp->if_alloctype = type;
503 #ifdef VIMAGE
504 	ifp->if_vnet = curvnet;
505 #endif
506 	if (if_com_alloc[type] != NULL) {
507 		ifp->if_l2com = if_com_alloc[type](type, ifp);
508 		if (ifp->if_l2com == NULL) {
509 			free(ifp, M_IFNET);
510 			ifindex_free(idx);
511 			return (NULL);
512 		}
513 	}
514 
515 	IF_ADDR_LOCK_INIT(ifp);
516 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
517 	ifp->if_afdata_initialized = 0;
518 	IF_AFDATA_LOCK_INIT(ifp);
519 	TAILQ_INIT(&ifp->if_addrhead);
520 	TAILQ_INIT(&ifp->if_multiaddrs);
521 	TAILQ_INIT(&ifp->if_groups);
522 #ifdef MAC
523 	mac_ifnet_init(ifp);
524 #endif
525 	ifq_init(&ifp->if_snd, ifp);
526 
527 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
528 	for (int i = 0; i < IFCOUNTERS; i++)
529 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
530 	ifp->if_get_counter = if_get_counter_default;
531 	ifp->if_pcp = IFNET_PCP_NONE;
532 	ifnet_setbyindex(ifp->if_index, ifp);
533 	return (ifp);
534 }
535 
536 /*
537  * Do the actual work of freeing a struct ifnet, and layer 2 common
538  * structure.  This call is made when the last reference to an
539  * interface is released.
540  */
541 static void
542 if_free_internal(struct ifnet *ifp)
543 {
544 
545 	KASSERT((ifp->if_flags & IFF_DYING),
546 	    ("if_free_internal: interface not dying"));
547 
548 	if (if_com_free[ifp->if_alloctype] != NULL)
549 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
550 		    ifp->if_alloctype);
551 
552 #ifdef MAC
553 	mac_ifnet_destroy(ifp);
554 #endif /* MAC */
555 	if (ifp->if_description != NULL)
556 		free(ifp->if_description, M_IFDESCR);
557 	IF_AFDATA_DESTROY(ifp);
558 	IF_ADDR_LOCK_DESTROY(ifp);
559 	ifq_delete(&ifp->if_snd);
560 
561 	for (int i = 0; i < IFCOUNTERS; i++)
562 		counter_u64_free(ifp->if_counters[i]);
563 
564 	free(ifp, M_IFNET);
565 }
566 
567 /*
568  * Deregister an interface and free the associated storage.
569  */
570 void
571 if_free(struct ifnet *ifp)
572 {
573 
574 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
575 
576 	CURVNET_SET_QUIET(ifp->if_vnet);
577 	IFNET_WLOCK();
578 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
579 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
580 
581 	ifindex_free_locked(ifp->if_index);
582 	IFNET_WUNLOCK();
583 
584 	if (refcount_release(&ifp->if_refcount))
585 		if_free_internal(ifp);
586 	CURVNET_RESTORE();
587 }
588 
589 /*
590  * Interfaces to keep an ifnet type-stable despite the possibility of the
591  * driver calling if_free().  If there are additional references, we defer
592  * freeing the underlying data structure.
593  */
594 void
595 if_ref(struct ifnet *ifp)
596 {
597 
598 	/* We don't assert the ifnet list lock here, but arguably should. */
599 	refcount_acquire(&ifp->if_refcount);
600 }
601 
602 void
603 if_rele(struct ifnet *ifp)
604 {
605 
606 	if (!refcount_release(&ifp->if_refcount))
607 		return;
608 	if_free_internal(ifp);
609 }
610 
611 void
612 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
613 {
614 
615 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
616 
617 	if (ifq->ifq_maxlen == 0)
618 		ifq->ifq_maxlen = ifqmaxlen;
619 
620 	ifq->altq_type = 0;
621 	ifq->altq_disc = NULL;
622 	ifq->altq_flags &= ALTQF_CANTCHANGE;
623 	ifq->altq_tbr  = NULL;
624 	ifq->altq_ifp  = ifp;
625 }
626 
627 void
628 ifq_delete(struct ifaltq *ifq)
629 {
630 	mtx_destroy(&ifq->ifq_mtx);
631 }
632 
633 /*
634  * Perform generic interface initialization tasks and attach the interface
635  * to the list of "active" interfaces.  If vmove flag is set on entry
636  * to if_attach_internal(), perform only a limited subset of initialization
637  * tasks, given that we are moving from one vnet to another an ifnet which
638  * has already been fully initialized.
639  *
640  * Note that if_detach_internal() removes group membership unconditionally
641  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
642  * Thus, when if_vmove() is applied to a cloned interface, group membership
643  * is lost while a cloned one always joins a group whose name is
644  * ifc->ifc_name.  To recover this after if_detach_internal() and
645  * if_attach_internal(), the cloner should be specified to
646  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
647  * attempts to join a group whose name is ifc->ifc_name.
648  *
649  * XXX:
650  *  - The decision to return void and thus require this function to
651  *    succeed is questionable.
652  *  - We should probably do more sanity checking.  For instance we don't
653  *    do anything to insure if_xname is unique or non-empty.
654  */
655 void
656 if_attach(struct ifnet *ifp)
657 {
658 
659 	if_attach_internal(ifp, 0, NULL);
660 }
661 
662 /*
663  * Compute the least common TSO limit.
664  */
665 void
666 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
667 {
668 	/*
669 	 * 1) If there is no limit currently, take the limit from
670 	 * the network adapter.
671 	 *
672 	 * 2) If the network adapter has a limit below the current
673 	 * limit, apply it.
674 	 */
675 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
676 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
677 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
678 	}
679 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
680 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
681 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
682 	}
683 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
684 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
685 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
686 	}
687 }
688 
689 /*
690  * Update TSO limit of a network adapter.
691  *
692  * Returns zero if no change. Else non-zero.
693  */
694 int
695 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
696 {
697 	int retval = 0;
698 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
699 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
700 		retval++;
701 	}
702 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
703 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
704 		retval++;
705 	}
706 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
707 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
708 		retval++;
709 	}
710 	return (retval);
711 }
712 
713 static void
714 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
715 {
716 	unsigned socksize, ifasize;
717 	int namelen, masklen;
718 	struct sockaddr_dl *sdl;
719 	struct ifaddr *ifa;
720 
721 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
722 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
723 		    ifp->if_xname);
724 
725 #ifdef VIMAGE
726 	ifp->if_vnet = curvnet;
727 	if (ifp->if_home_vnet == NULL)
728 		ifp->if_home_vnet = curvnet;
729 #endif
730 
731 	if_addgroup(ifp, IFG_ALL);
732 
733 	/* Restore group membership for cloned interfaces. */
734 	if (vmove && ifc != NULL)
735 		if_clone_addgroup(ifp, ifc);
736 
737 	getmicrotime(&ifp->if_lastchange);
738 	ifp->if_epoch = time_uptime;
739 
740 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
741 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
742 	    ("transmit and qflush must both either be set or both be NULL"));
743 	if (ifp->if_transmit == NULL) {
744 		ifp->if_transmit = if_transmit;
745 		ifp->if_qflush = if_qflush;
746 	}
747 	if (ifp->if_input == NULL)
748 		ifp->if_input = if_input_default;
749 
750 	if (ifp->if_requestencap == NULL)
751 		ifp->if_requestencap = if_requestencap_default;
752 
753 	if (!vmove) {
754 #ifdef MAC
755 		mac_ifnet_create(ifp);
756 #endif
757 
758 		/*
759 		 * Create a Link Level name for this device.
760 		 */
761 		namelen = strlen(ifp->if_xname);
762 		/*
763 		 * Always save enough space for any possiable name so we
764 		 * can do a rename in place later.
765 		 */
766 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
767 		socksize = masklen + ifp->if_addrlen;
768 		if (socksize < sizeof(*sdl))
769 			socksize = sizeof(*sdl);
770 		socksize = roundup2(socksize, sizeof(long));
771 		ifasize = sizeof(*ifa) + 2 * socksize;
772 		ifa = ifa_alloc(ifasize, M_WAITOK);
773 		sdl = (struct sockaddr_dl *)(ifa + 1);
774 		sdl->sdl_len = socksize;
775 		sdl->sdl_family = AF_LINK;
776 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
777 		sdl->sdl_nlen = namelen;
778 		sdl->sdl_index = ifp->if_index;
779 		sdl->sdl_type = ifp->if_type;
780 		ifp->if_addr = ifa;
781 		ifa->ifa_ifp = ifp;
782 		ifa->ifa_rtrequest = link_rtrequest;
783 		ifa->ifa_addr = (struct sockaddr *)sdl;
784 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
785 		ifa->ifa_netmask = (struct sockaddr *)sdl;
786 		sdl->sdl_len = masklen;
787 		while (namelen != 0)
788 			sdl->sdl_data[--namelen] = 0xff;
789 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
790 		/* Reliably crash if used uninitialized. */
791 		ifp->if_broadcastaddr = NULL;
792 
793 		if (ifp->if_type == IFT_ETHER) {
794 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
795 			    M_WAITOK | M_ZERO);
796 		}
797 
798 #if defined(INET) || defined(INET6)
799 		/* Use defaults for TSO, if nothing is set */
800 		if (ifp->if_hw_tsomax == 0 &&
801 		    ifp->if_hw_tsomaxsegcount == 0 &&
802 		    ifp->if_hw_tsomaxsegsize == 0) {
803 			/*
804 			 * The TSO defaults needs to be such that an
805 			 * NFS mbuf list of 35 mbufs totalling just
806 			 * below 64K works and that a chain of mbufs
807 			 * can be defragged into at most 32 segments:
808 			 */
809 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
810 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
811 			ifp->if_hw_tsomaxsegcount = 35;
812 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
813 
814 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
815 			if (ifp->if_capabilities & IFCAP_TSO) {
816 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
817 				    ifp->if_hw_tsomax,
818 				    ifp->if_hw_tsomaxsegcount,
819 				    ifp->if_hw_tsomaxsegsize);
820 			}
821 		}
822 #endif
823 	}
824 #ifdef VIMAGE
825 	else {
826 		/*
827 		 * Update the interface index in the link layer address
828 		 * of the interface.
829 		 */
830 		for (ifa = ifp->if_addr; ifa != NULL;
831 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
832 			if (ifa->ifa_addr->sa_family == AF_LINK) {
833 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
834 				sdl->sdl_index = ifp->if_index;
835 			}
836 		}
837 	}
838 #endif
839 
840 	IFNET_WLOCK();
841 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
842 #ifdef VIMAGE
843 	curvnet->vnet_ifcnt++;
844 #endif
845 	IFNET_WUNLOCK();
846 
847 	if (domain_init_status >= 2)
848 		if_attachdomain1(ifp);
849 
850 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
851 	if (IS_DEFAULT_VNET(curvnet))
852 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
853 
854 	/* Announce the interface. */
855 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
856 }
857 
858 static void
859 if_attachdomain(void *dummy)
860 {
861 	struct ifnet *ifp;
862 
863 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
864 		if_attachdomain1(ifp);
865 }
866 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
867     if_attachdomain, NULL);
868 
869 static void
870 if_attachdomain1(struct ifnet *ifp)
871 {
872 	struct domain *dp;
873 
874 	/*
875 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
876 	 * cannot lock ifp->if_afdata initialization, entirely.
877 	 */
878 	IF_AFDATA_LOCK(ifp);
879 	if (ifp->if_afdata_initialized >= domain_init_status) {
880 		IF_AFDATA_UNLOCK(ifp);
881 		log(LOG_WARNING, "%s called more than once on %s\n",
882 		    __func__, ifp->if_xname);
883 		return;
884 	}
885 	ifp->if_afdata_initialized = domain_init_status;
886 	IF_AFDATA_UNLOCK(ifp);
887 
888 	/* address family dependent data region */
889 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
890 	for (dp = domains; dp; dp = dp->dom_next) {
891 		if (dp->dom_ifattach)
892 			ifp->if_afdata[dp->dom_family] =
893 			    (*dp->dom_ifattach)(ifp);
894 	}
895 }
896 
897 /*
898  * Remove any unicast or broadcast network addresses from an interface.
899  */
900 void
901 if_purgeaddrs(struct ifnet *ifp)
902 {
903 	struct ifaddr *ifa, *next;
904 
905 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
906 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
907 		if (ifa->ifa_addr->sa_family == AF_LINK)
908 			continue;
909 #ifdef INET
910 		/* XXX: Ugly!! ad hoc just for INET */
911 		if (ifa->ifa_addr->sa_family == AF_INET) {
912 			struct ifaliasreq ifr;
913 
914 			bzero(&ifr, sizeof(ifr));
915 			ifr.ifra_addr = *ifa->ifa_addr;
916 			if (ifa->ifa_dstaddr)
917 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
918 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
919 			    NULL) == 0)
920 				continue;
921 		}
922 #endif /* INET */
923 #ifdef INET6
924 		if (ifa->ifa_addr->sa_family == AF_INET6) {
925 			in6_purgeaddr(ifa);
926 			/* ifp_addrhead is already updated */
927 			continue;
928 		}
929 #endif /* INET6 */
930 		IF_ADDR_WLOCK(ifp);
931 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
932 		IF_ADDR_WUNLOCK(ifp);
933 		ifa_free(ifa);
934 	}
935 }
936 
937 /*
938  * Remove any multicast network addresses from an interface when an ifnet
939  * is going away.
940  */
941 static void
942 if_purgemaddrs(struct ifnet *ifp)
943 {
944 	struct ifmultiaddr *ifma;
945 	struct ifmultiaddr *next;
946 
947 	IF_ADDR_WLOCK(ifp);
948 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
949 		if_delmulti_locked(ifp, ifma, 1);
950 	IF_ADDR_WUNLOCK(ifp);
951 }
952 
953 /*
954  * Detach an interface, removing it from the list of "active" interfaces.
955  * If vmove flag is set on entry to if_detach_internal(), perform only a
956  * limited subset of cleanup tasks, given that we are moving an ifnet from
957  * one vnet to another, where it must be fully operational.
958  *
959  * XXXRW: There are some significant questions about event ordering, and
960  * how to prevent things from starting to use the interface during detach.
961  */
962 void
963 if_detach(struct ifnet *ifp)
964 {
965 
966 	CURVNET_SET_QUIET(ifp->if_vnet);
967 	if_detach_internal(ifp, 0, NULL);
968 	CURVNET_RESTORE();
969 }
970 
971 /*
972  * The vmove flag, if set, indicates that we are called from a callpath
973  * that is moving an interface to a different vnet instance.
974  *
975  * The shutdown flag, if set, indicates that we are called in the
976  * process of shutting down a vnet instance.  Currently only the
977  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
978  * on a vnet instance shutdown without this flag being set, e.g., when
979  * the cloned interfaces are destoyed as first thing of teardown.
980  */
981 static int
982 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
983 {
984 	struct ifaddr *ifa;
985 	int i;
986 	struct domain *dp;
987  	struct ifnet *iter;
988  	int found = 0;
989 #ifdef VIMAGE
990 	int shutdown;
991 
992 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
993 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
994 #endif
995 	IFNET_WLOCK();
996 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
997 		if (iter == ifp) {
998 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
999 			found = 1;
1000 			break;
1001 		}
1002 	IFNET_WUNLOCK();
1003 	if (!found) {
1004 		/*
1005 		 * While we would want to panic here, we cannot
1006 		 * guarantee that the interface is indeed still on
1007 		 * the list given we don't hold locks all the way.
1008 		 */
1009 		return (ENOENT);
1010 #if 0
1011 		if (vmove)
1012 			panic("%s: ifp=%p not on the ifnet tailq %p",
1013 			    __func__, ifp, &V_ifnet);
1014 		else
1015 			return; /* XXX this should panic as well? */
1016 #endif
1017 	}
1018 
1019 	/*
1020 	 * At this point we know the interface still was on the ifnet list
1021 	 * and we removed it so we are in a stable state.
1022 	 */
1023 #ifdef VIMAGE
1024 	curvnet->vnet_ifcnt--;
1025 #endif
1026 
1027 	/*
1028 	 * In any case (destroy or vmove) detach us from the groups
1029 	 * and remove/wait for pending events on the taskq.
1030 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1031 	 */
1032 	if_delgroups(ifp);
1033 
1034 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1035 
1036 	/*
1037 	 * Check if this is a cloned interface or not. Must do even if
1038 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1039 	 * the if_clone_addgroup() will have a corrupted string overwise
1040 	 * from a gibberish pointer.
1041 	 */
1042 	if (vmove && ifcp != NULL)
1043 		*ifcp = if_clone_findifc(ifp);
1044 
1045 	if_down(ifp);
1046 
1047 #ifdef VIMAGE
1048 	/*
1049 	 * On VNET shutdown abort here as the stack teardown will do all
1050 	 * the work top-down for us.
1051 	 */
1052 	if (shutdown) {
1053 		/*
1054 		 * In case of a vmove we are done here without error.
1055 		 * If we would signal an error it would lead to the same
1056 		 * abort as if we did not find the ifnet anymore.
1057 		 * if_detach() calls us in void context and does not care
1058 		 * about an early abort notification, so life is splendid :)
1059 		 */
1060 		goto finish_vnet_shutdown;
1061 	}
1062 #endif
1063 
1064 	/*
1065 	 * At this point we are not tearing down a VNET and are either
1066 	 * going to destroy or vmove the interface and have to cleanup
1067 	 * accordingly.
1068 	 */
1069 
1070 	/*
1071 	 * Remove routes and flush queues.
1072 	 */
1073 #ifdef ALTQ
1074 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1075 		altq_disable(&ifp->if_snd);
1076 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1077 		altq_detach(&ifp->if_snd);
1078 #endif
1079 
1080 	if_purgeaddrs(ifp);
1081 
1082 #ifdef INET
1083 	in_ifdetach(ifp);
1084 #endif
1085 
1086 #ifdef INET6
1087 	/*
1088 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1089 	 * before removing routing entries below, since IPv6 interface direct
1090 	 * routes are expected to be removed by the IPv6-specific kernel API.
1091 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1092 	 */
1093 	in6_ifdetach(ifp);
1094 #endif
1095 	if_purgemaddrs(ifp);
1096 
1097 	/* Announce that the interface is gone. */
1098 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1099 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1100 	if (IS_DEFAULT_VNET(curvnet))
1101 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1102 
1103 	if (!vmove) {
1104 		/*
1105 		 * Prevent further calls into the device driver via ifnet.
1106 		 */
1107 		if_dead(ifp);
1108 
1109 		/*
1110 		 * Remove link ifaddr pointer and maybe decrement if_index.
1111 		 * Clean up all addresses.
1112 		 */
1113 		free(ifp->if_hw_addr, M_IFADDR);
1114 		ifp->if_hw_addr = NULL;
1115 		ifp->if_addr = NULL;
1116 
1117 		/* We can now free link ifaddr. */
1118 		IF_ADDR_WLOCK(ifp);
1119 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1120 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1121 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1122 			IF_ADDR_WUNLOCK(ifp);
1123 			ifa_free(ifa);
1124 		} else
1125 			IF_ADDR_WUNLOCK(ifp);
1126 	}
1127 
1128 	rt_flushifroutes(ifp);
1129 
1130 #ifdef VIMAGE
1131 finish_vnet_shutdown:
1132 #endif
1133 	/*
1134 	 * We cannot hold the lock over dom_ifdetach calls as they might
1135 	 * sleep, for example trying to drain a callout, thus open up the
1136 	 * theoretical race with re-attaching.
1137 	 */
1138 	IF_AFDATA_LOCK(ifp);
1139 	i = ifp->if_afdata_initialized;
1140 	ifp->if_afdata_initialized = 0;
1141 	IF_AFDATA_UNLOCK(ifp);
1142 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1143 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1144 			(*dp->dom_ifdetach)(ifp,
1145 			    ifp->if_afdata[dp->dom_family]);
1146 			ifp->if_afdata[dp->dom_family] = NULL;
1147 		}
1148 	}
1149 
1150 	return (0);
1151 }
1152 
1153 #ifdef VIMAGE
1154 /*
1155  * if_vmove() performs a limited version of if_detach() in current
1156  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1157  * An attempt is made to shrink if_index in current vnet, find an
1158  * unused if_index in target vnet and calls if_grow() if necessary,
1159  * and finally find an unused if_xname for the target vnet.
1160  */
1161 static void
1162 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1163 {
1164 	struct if_clone *ifc;
1165 	u_int bif_dlt, bif_hdrlen;
1166 	int rc;
1167 
1168  	/*
1169 	 * if_detach_internal() will call the eventhandler to notify
1170 	 * interface departure.  That will detach if_bpf.  We need to
1171 	 * safe the dlt and hdrlen so we can re-attach it later.
1172 	 */
1173 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1174 
1175 	/*
1176 	 * Detach from current vnet, but preserve LLADDR info, do not
1177 	 * mark as dead etc. so that the ifnet can be reattached later.
1178 	 * If we cannot find it, we lost the race to someone else.
1179 	 */
1180 	rc = if_detach_internal(ifp, 1, &ifc);
1181 	if (rc != 0)
1182 		return;
1183 
1184 	/*
1185 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1186 	 * the if_index for that vnet if possible.
1187 	 *
1188 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1189 	 * or we'd lock on one vnet and unlock on another.
1190 	 */
1191 	IFNET_WLOCK();
1192 	ifindex_free_locked(ifp->if_index);
1193 	IFNET_WUNLOCK();
1194 
1195 	/*
1196 	 * Perform interface-specific reassignment tasks, if provided by
1197 	 * the driver.
1198 	 */
1199 	if (ifp->if_reassign != NULL)
1200 		ifp->if_reassign(ifp, new_vnet, NULL);
1201 
1202 	/*
1203 	 * Switch to the context of the target vnet.
1204 	 */
1205 	CURVNET_SET_QUIET(new_vnet);
1206 
1207 	IFNET_WLOCK();
1208 	ifp->if_index = ifindex_alloc();
1209 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1210 	IFNET_WUNLOCK();
1211 
1212 	if_attach_internal(ifp, 1, ifc);
1213 
1214 	if (ifp->if_bpf == NULL)
1215 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1216 
1217 	CURVNET_RESTORE();
1218 }
1219 
1220 /*
1221  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1222  */
1223 static int
1224 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1225 {
1226 	struct prison *pr;
1227 	struct ifnet *difp;
1228 	int shutdown;
1229 
1230 	/* Try to find the prison within our visibility. */
1231 	sx_slock(&allprison_lock);
1232 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1233 	sx_sunlock(&allprison_lock);
1234 	if (pr == NULL)
1235 		return (ENXIO);
1236 	prison_hold_locked(pr);
1237 	mtx_unlock(&pr->pr_mtx);
1238 
1239 	/* Do not try to move the iface from and to the same prison. */
1240 	if (pr->pr_vnet == ifp->if_vnet) {
1241 		prison_free(pr);
1242 		return (EEXIST);
1243 	}
1244 
1245 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1246 	/* XXX Lock interfaces to avoid races. */
1247 	CURVNET_SET_QUIET(pr->pr_vnet);
1248 	difp = ifunit(ifname);
1249 	if (difp != NULL) {
1250 		CURVNET_RESTORE();
1251 		prison_free(pr);
1252 		return (EEXIST);
1253 	}
1254 
1255 	/* Make sure the VNET is stable. */
1256 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1257 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1258 	if (shutdown) {
1259 		CURVNET_RESTORE();
1260 		prison_free(pr);
1261 		return (EBUSY);
1262 	}
1263 	CURVNET_RESTORE();
1264 
1265 	/* Move the interface into the child jail/vnet. */
1266 	if_vmove(ifp, pr->pr_vnet);
1267 
1268 	/* Report the new if_xname back to the userland. */
1269 	sprintf(ifname, "%s", ifp->if_xname);
1270 
1271 	prison_free(pr);
1272 	return (0);
1273 }
1274 
1275 static int
1276 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1277 {
1278 	struct prison *pr;
1279 	struct vnet *vnet_dst;
1280 	struct ifnet *ifp;
1281  	int shutdown;
1282 
1283 	/* Try to find the prison within our visibility. */
1284 	sx_slock(&allprison_lock);
1285 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1286 	sx_sunlock(&allprison_lock);
1287 	if (pr == NULL)
1288 		return (ENXIO);
1289 	prison_hold_locked(pr);
1290 	mtx_unlock(&pr->pr_mtx);
1291 
1292 	/* Make sure the named iface exists in the source prison/vnet. */
1293 	CURVNET_SET(pr->pr_vnet);
1294 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1295 	if (ifp == NULL) {
1296 		CURVNET_RESTORE();
1297 		prison_free(pr);
1298 		return (ENXIO);
1299 	}
1300 
1301 	/* Do not try to move the iface from and to the same prison. */
1302 	vnet_dst = TD_TO_VNET(td);
1303 	if (vnet_dst == ifp->if_vnet) {
1304 		CURVNET_RESTORE();
1305 		prison_free(pr);
1306 		return (EEXIST);
1307 	}
1308 
1309 	/* Make sure the VNET is stable. */
1310 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1311 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1312 	if (shutdown) {
1313 		CURVNET_RESTORE();
1314 		prison_free(pr);
1315 		return (EBUSY);
1316 	}
1317 
1318 	/* Get interface back from child jail/vnet. */
1319 	if_vmove(ifp, vnet_dst);
1320 	CURVNET_RESTORE();
1321 
1322 	/* Report the new if_xname back to the userland. */
1323 	sprintf(ifname, "%s", ifp->if_xname);
1324 
1325 	prison_free(pr);
1326 	return (0);
1327 }
1328 #endif /* VIMAGE */
1329 
1330 /*
1331  * Add a group to an interface
1332  */
1333 int
1334 if_addgroup(struct ifnet *ifp, const char *groupname)
1335 {
1336 	struct ifg_list		*ifgl;
1337 	struct ifg_group	*ifg = NULL;
1338 	struct ifg_member	*ifgm;
1339 	int 			 new = 0;
1340 
1341 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1342 	    groupname[strlen(groupname) - 1] <= '9')
1343 		return (EINVAL);
1344 
1345 	IFNET_WLOCK();
1346 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1347 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1348 			IFNET_WUNLOCK();
1349 			return (EEXIST);
1350 		}
1351 
1352 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1353 	    M_NOWAIT)) == NULL) {
1354 	    	IFNET_WUNLOCK();
1355 		return (ENOMEM);
1356 	}
1357 
1358 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1359 	    M_TEMP, M_NOWAIT)) == NULL) {
1360 		free(ifgl, M_TEMP);
1361 		IFNET_WUNLOCK();
1362 		return (ENOMEM);
1363 	}
1364 
1365 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1366 		if (!strcmp(ifg->ifg_group, groupname))
1367 			break;
1368 
1369 	if (ifg == NULL) {
1370 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1371 		    M_TEMP, M_NOWAIT)) == NULL) {
1372 			free(ifgl, M_TEMP);
1373 			free(ifgm, M_TEMP);
1374 			IFNET_WUNLOCK();
1375 			return (ENOMEM);
1376 		}
1377 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1378 		ifg->ifg_refcnt = 0;
1379 		TAILQ_INIT(&ifg->ifg_members);
1380 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1381 		new = 1;
1382 	}
1383 
1384 	ifg->ifg_refcnt++;
1385 	ifgl->ifgl_group = ifg;
1386 	ifgm->ifgm_ifp = ifp;
1387 
1388 	IF_ADDR_WLOCK(ifp);
1389 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1390 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1391 	IF_ADDR_WUNLOCK(ifp);
1392 
1393 	IFNET_WUNLOCK();
1394 
1395 	if (new)
1396 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1397 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1398 
1399 	return (0);
1400 }
1401 
1402 /*
1403  * Remove a group from an interface
1404  */
1405 int
1406 if_delgroup(struct ifnet *ifp, const char *groupname)
1407 {
1408 	struct ifg_list		*ifgl;
1409 	struct ifg_member	*ifgm;
1410 
1411 	IFNET_WLOCK();
1412 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1413 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1414 			break;
1415 	if (ifgl == NULL) {
1416 		IFNET_WUNLOCK();
1417 		return (ENOENT);
1418 	}
1419 
1420 	IF_ADDR_WLOCK(ifp);
1421 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1422 	IF_ADDR_WUNLOCK(ifp);
1423 
1424 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1425 		if (ifgm->ifgm_ifp == ifp)
1426 			break;
1427 
1428 	if (ifgm != NULL) {
1429 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1430 		free(ifgm, M_TEMP);
1431 	}
1432 
1433 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1434 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1435 		IFNET_WUNLOCK();
1436 		EVENTHANDLER_INVOKE(group_detach_event, 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 	return (0);
1446 }
1447 
1448 /*
1449  * Remove an interface from all groups
1450  */
1451 static void
1452 if_delgroups(struct ifnet *ifp)
1453 {
1454 	struct ifg_list		*ifgl;
1455 	struct ifg_member	*ifgm;
1456 	char groupname[IFNAMSIZ];
1457 
1458 	IFNET_WLOCK();
1459 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1460 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1461 
1462 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1463 
1464 		IF_ADDR_WLOCK(ifp);
1465 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1466 		IF_ADDR_WUNLOCK(ifp);
1467 
1468 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1469 			if (ifgm->ifgm_ifp == ifp)
1470 				break;
1471 
1472 		if (ifgm != NULL) {
1473 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1474 			    ifgm_next);
1475 			free(ifgm, M_TEMP);
1476 		}
1477 
1478 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1479 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1480 			IFNET_WUNLOCK();
1481 			EVENTHANDLER_INVOKE(group_detach_event,
1482 			    ifgl->ifgl_group);
1483 			free(ifgl->ifgl_group, M_TEMP);
1484 		} else
1485 			IFNET_WUNLOCK();
1486 
1487 		free(ifgl, M_TEMP);
1488 
1489 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1490 
1491 		IFNET_WLOCK();
1492 	}
1493 	IFNET_WUNLOCK();
1494 }
1495 
1496 /*
1497  * Stores all groups from an interface in memory pointed
1498  * to by data
1499  */
1500 static int
1501 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1502 {
1503 	int			 len, error;
1504 	struct ifg_list		*ifgl;
1505 	struct ifg_req		 ifgrq, *ifgp;
1506 	struct ifgroupreq	*ifgr = data;
1507 
1508 	if (ifgr->ifgr_len == 0) {
1509 		IF_ADDR_RLOCK(ifp);
1510 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1511 			ifgr->ifgr_len += sizeof(struct ifg_req);
1512 		IF_ADDR_RUNLOCK(ifp);
1513 		return (0);
1514 	}
1515 
1516 	len = ifgr->ifgr_len;
1517 	ifgp = ifgr->ifgr_groups;
1518 	/* XXX: wire */
1519 	IF_ADDR_RLOCK(ifp);
1520 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1521 		if (len < sizeof(ifgrq)) {
1522 			IF_ADDR_RUNLOCK(ifp);
1523 			return (EINVAL);
1524 		}
1525 		bzero(&ifgrq, sizeof ifgrq);
1526 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1527 		    sizeof(ifgrq.ifgrq_group));
1528 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1529 		    	IF_ADDR_RUNLOCK(ifp);
1530 			return (error);
1531 		}
1532 		len -= sizeof(ifgrq);
1533 		ifgp++;
1534 	}
1535 	IF_ADDR_RUNLOCK(ifp);
1536 
1537 	return (0);
1538 }
1539 
1540 /*
1541  * Stores all members of a group in memory pointed to by data
1542  */
1543 static int
1544 if_getgroupmembers(struct ifgroupreq *data)
1545 {
1546 	struct ifgroupreq	*ifgr = data;
1547 	struct ifg_group	*ifg;
1548 	struct ifg_member	*ifgm;
1549 	struct ifg_req		 ifgrq, *ifgp;
1550 	int			 len, error;
1551 
1552 	IFNET_RLOCK();
1553 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1554 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1555 			break;
1556 	if (ifg == NULL) {
1557 		IFNET_RUNLOCK();
1558 		return (ENOENT);
1559 	}
1560 
1561 	if (ifgr->ifgr_len == 0) {
1562 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1563 			ifgr->ifgr_len += sizeof(ifgrq);
1564 		IFNET_RUNLOCK();
1565 		return (0);
1566 	}
1567 
1568 	len = ifgr->ifgr_len;
1569 	ifgp = ifgr->ifgr_groups;
1570 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1571 		if (len < sizeof(ifgrq)) {
1572 			IFNET_RUNLOCK();
1573 			return (EINVAL);
1574 		}
1575 		bzero(&ifgrq, sizeof ifgrq);
1576 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1577 		    sizeof(ifgrq.ifgrq_member));
1578 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1579 			IFNET_RUNLOCK();
1580 			return (error);
1581 		}
1582 		len -= sizeof(ifgrq);
1583 		ifgp++;
1584 	}
1585 	IFNET_RUNLOCK();
1586 
1587 	return (0);
1588 }
1589 
1590 /*
1591  * Return counter values from counter(9)s stored in ifnet.
1592  */
1593 uint64_t
1594 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1595 {
1596 
1597 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1598 
1599 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1600 }
1601 
1602 /*
1603  * Increase an ifnet counter. Usually used for counters shared
1604  * between the stack and a driver, but function supports them all.
1605  */
1606 void
1607 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1608 {
1609 
1610 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1611 
1612 	counter_u64_add(ifp->if_counters[cnt], inc);
1613 }
1614 
1615 /*
1616  * Copy data from ifnet to userland API structure if_data.
1617  */
1618 void
1619 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1620 {
1621 
1622 	ifd->ifi_type = ifp->if_type;
1623 	ifd->ifi_physical = 0;
1624 	ifd->ifi_addrlen = ifp->if_addrlen;
1625 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1626 	ifd->ifi_link_state = ifp->if_link_state;
1627 	ifd->ifi_vhid = 0;
1628 	ifd->ifi_datalen = sizeof(struct if_data);
1629 	ifd->ifi_mtu = ifp->if_mtu;
1630 	ifd->ifi_metric = ifp->if_metric;
1631 	ifd->ifi_baudrate = ifp->if_baudrate;
1632 	ifd->ifi_hwassist = ifp->if_hwassist;
1633 	ifd->ifi_epoch = ifp->if_epoch;
1634 	ifd->ifi_lastchange = ifp->if_lastchange;
1635 
1636 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1637 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1638 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1639 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1640 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1641 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1642 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1643 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1644 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1645 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1646 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1647 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1648 }
1649 
1650 /*
1651  * Wrapper functions for struct ifnet address list locking macros.  These are
1652  * used by kernel modules to avoid encoding programming interface or binary
1653  * interface assumptions that may be violated when kernel-internal locking
1654  * approaches change.
1655  */
1656 void
1657 if_addr_rlock(struct ifnet *ifp)
1658 {
1659 
1660 	IF_ADDR_RLOCK(ifp);
1661 }
1662 
1663 void
1664 if_addr_runlock(struct ifnet *ifp)
1665 {
1666 
1667 	IF_ADDR_RUNLOCK(ifp);
1668 }
1669 
1670 void
1671 if_maddr_rlock(if_t ifp)
1672 {
1673 
1674 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1675 }
1676 
1677 void
1678 if_maddr_runlock(if_t ifp)
1679 {
1680 
1681 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1682 }
1683 
1684 /*
1685  * Initialization, destruction and refcounting functions for ifaddrs.
1686  */
1687 struct ifaddr *
1688 ifa_alloc(size_t size, int flags)
1689 {
1690 	struct ifaddr *ifa;
1691 
1692 	KASSERT(size >= sizeof(struct ifaddr),
1693 	    ("%s: invalid size %zu", __func__, size));
1694 
1695 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1696 	if (ifa == NULL)
1697 		return (NULL);
1698 
1699 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1700 		goto fail;
1701 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1702 		goto fail;
1703 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1704 		goto fail;
1705 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1706 		goto fail;
1707 
1708 	refcount_init(&ifa->ifa_refcnt, 1);
1709 
1710 	return (ifa);
1711 
1712 fail:
1713 	/* free(NULL) is okay */
1714 	counter_u64_free(ifa->ifa_opackets);
1715 	counter_u64_free(ifa->ifa_ipackets);
1716 	counter_u64_free(ifa->ifa_obytes);
1717 	counter_u64_free(ifa->ifa_ibytes);
1718 	free(ifa, M_IFADDR);
1719 
1720 	return (NULL);
1721 }
1722 
1723 void
1724 ifa_ref(struct ifaddr *ifa)
1725 {
1726 
1727 	refcount_acquire(&ifa->ifa_refcnt);
1728 }
1729 
1730 void
1731 ifa_free(struct ifaddr *ifa)
1732 {
1733 
1734 	if (refcount_release(&ifa->ifa_refcnt)) {
1735 		counter_u64_free(ifa->ifa_opackets);
1736 		counter_u64_free(ifa->ifa_ipackets);
1737 		counter_u64_free(ifa->ifa_obytes);
1738 		counter_u64_free(ifa->ifa_ibytes);
1739 		free(ifa, M_IFADDR);
1740 	}
1741 }
1742 
1743 static int
1744 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1745     struct sockaddr *ia)
1746 {
1747 	int error;
1748 	struct rt_addrinfo info;
1749 	struct sockaddr_dl null_sdl;
1750 	struct ifnet *ifp;
1751 
1752 	ifp = ifa->ifa_ifp;
1753 
1754 	bzero(&info, sizeof(info));
1755 	if (cmd != RTM_DELETE)
1756 		info.rti_ifp = V_loif;
1757 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1758 	info.rti_info[RTAX_DST] = ia;
1759 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1760 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1761 
1762 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1763 
1764 	if (error != 0)
1765 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1766 		    __func__, otype, if_name(ifp), error);
1767 
1768 	return (error);
1769 }
1770 
1771 int
1772 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1773 {
1774 
1775 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1776 }
1777 
1778 int
1779 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1780 {
1781 
1782 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1783 }
1784 
1785 int
1786 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1787 {
1788 
1789 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1790 }
1791 
1792 /*
1793  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1794  * structs used to represent other address families, it is necessary
1795  * to perform a different comparison.
1796  */
1797 
1798 #define	sa_dl_equal(a1, a2)	\
1799 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1800 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1801 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1802 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1803 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1804 
1805 /*
1806  * Locate an interface based on a complete address.
1807  */
1808 /*ARGSUSED*/
1809 static struct ifaddr *
1810 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1811 {
1812 	struct ifnet *ifp;
1813 	struct ifaddr *ifa;
1814 
1815 	IFNET_RLOCK_NOSLEEP();
1816 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1817 		IF_ADDR_RLOCK(ifp);
1818 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1819 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1820 				continue;
1821 			if (sa_equal(addr, ifa->ifa_addr)) {
1822 				if (getref)
1823 					ifa_ref(ifa);
1824 				IF_ADDR_RUNLOCK(ifp);
1825 				goto done;
1826 			}
1827 			/* IP6 doesn't have broadcast */
1828 			if ((ifp->if_flags & IFF_BROADCAST) &&
1829 			    ifa->ifa_broadaddr &&
1830 			    ifa->ifa_broadaddr->sa_len != 0 &&
1831 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1832 				if (getref)
1833 					ifa_ref(ifa);
1834 				IF_ADDR_RUNLOCK(ifp);
1835 				goto done;
1836 			}
1837 		}
1838 		IF_ADDR_RUNLOCK(ifp);
1839 	}
1840 	ifa = NULL;
1841 done:
1842 	IFNET_RUNLOCK_NOSLEEP();
1843 	return (ifa);
1844 }
1845 
1846 struct ifaddr *
1847 ifa_ifwithaddr(const struct sockaddr *addr)
1848 {
1849 
1850 	return (ifa_ifwithaddr_internal(addr, 1));
1851 }
1852 
1853 int
1854 ifa_ifwithaddr_check(const struct sockaddr *addr)
1855 {
1856 
1857 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1858 }
1859 
1860 /*
1861  * Locate an interface based on the broadcast address.
1862  */
1863 /* ARGSUSED */
1864 struct ifaddr *
1865 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1866 {
1867 	struct ifnet *ifp;
1868 	struct ifaddr *ifa;
1869 
1870 	IFNET_RLOCK_NOSLEEP();
1871 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1872 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1873 			continue;
1874 		IF_ADDR_RLOCK(ifp);
1875 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1876 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1877 				continue;
1878 			if ((ifp->if_flags & IFF_BROADCAST) &&
1879 			    ifa->ifa_broadaddr &&
1880 			    ifa->ifa_broadaddr->sa_len != 0 &&
1881 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1882 				ifa_ref(ifa);
1883 				IF_ADDR_RUNLOCK(ifp);
1884 				goto done;
1885 			}
1886 		}
1887 		IF_ADDR_RUNLOCK(ifp);
1888 	}
1889 	ifa = NULL;
1890 done:
1891 	IFNET_RUNLOCK_NOSLEEP();
1892 	return (ifa);
1893 }
1894 
1895 /*
1896  * Locate the point to point interface with a given destination address.
1897  */
1898 /*ARGSUSED*/
1899 struct ifaddr *
1900 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1901 {
1902 	struct ifnet *ifp;
1903 	struct ifaddr *ifa;
1904 
1905 	IFNET_RLOCK_NOSLEEP();
1906 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1907 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1908 			continue;
1909 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1910 			continue;
1911 		IF_ADDR_RLOCK(ifp);
1912 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1913 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1914 				continue;
1915 			if (ifa->ifa_dstaddr != NULL &&
1916 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1917 				ifa_ref(ifa);
1918 				IF_ADDR_RUNLOCK(ifp);
1919 				goto done;
1920 			}
1921 		}
1922 		IF_ADDR_RUNLOCK(ifp);
1923 	}
1924 	ifa = NULL;
1925 done:
1926 	IFNET_RUNLOCK_NOSLEEP();
1927 	return (ifa);
1928 }
1929 
1930 /*
1931  * Find an interface on a specific network.  If many, choice
1932  * is most specific found.
1933  */
1934 struct ifaddr *
1935 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1936 {
1937 	struct ifnet *ifp;
1938 	struct ifaddr *ifa;
1939 	struct ifaddr *ifa_maybe = NULL;
1940 	u_int af = addr->sa_family;
1941 	const char *addr_data = addr->sa_data, *cplim;
1942 
1943 	/*
1944 	 * AF_LINK addresses can be looked up directly by their index number,
1945 	 * so do that if we can.
1946 	 */
1947 	if (af == AF_LINK) {
1948 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1949 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1950 		return (ifaddr_byindex(sdl->sdl_index));
1951 	}
1952 
1953 	/*
1954 	 * Scan though each interface, looking for ones that have addresses
1955 	 * in this address family and the requested fib.  Maintain a reference
1956 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1957 	 * kept it stable when we move onto the next interface.
1958 	 */
1959 	IFNET_RLOCK_NOSLEEP();
1960 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1961 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1962 			continue;
1963 		IF_ADDR_RLOCK(ifp);
1964 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1965 			const char *cp, *cp2, *cp3;
1966 
1967 			if (ifa->ifa_addr->sa_family != af)
1968 next:				continue;
1969 			if (af == AF_INET &&
1970 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1971 				/*
1972 				 * This is a bit broken as it doesn't
1973 				 * take into account that the remote end may
1974 				 * be a single node in the network we are
1975 				 * looking for.
1976 				 * The trouble is that we don't know the
1977 				 * netmask for the remote end.
1978 				 */
1979 				if (ifa->ifa_dstaddr != NULL &&
1980 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1981 					ifa_ref(ifa);
1982 					IF_ADDR_RUNLOCK(ifp);
1983 					goto done;
1984 				}
1985 			} else {
1986 				/*
1987 				 * Scan all the bits in the ifa's address.
1988 				 * If a bit dissagrees with what we are
1989 				 * looking for, mask it with the netmask
1990 				 * to see if it really matters.
1991 				 * (A byte at a time)
1992 				 */
1993 				if (ifa->ifa_netmask == 0)
1994 					continue;
1995 				cp = addr_data;
1996 				cp2 = ifa->ifa_addr->sa_data;
1997 				cp3 = ifa->ifa_netmask->sa_data;
1998 				cplim = ifa->ifa_netmask->sa_len
1999 					+ (char *)ifa->ifa_netmask;
2000 				while (cp3 < cplim)
2001 					if ((*cp++ ^ *cp2++) & *cp3++)
2002 						goto next; /* next address! */
2003 				/*
2004 				 * If the netmask of what we just found
2005 				 * is more specific than what we had before
2006 				 * (if we had one), or if the virtual status
2007 				 * of new prefix is better than of the old one,
2008 				 * then remember the new one before continuing
2009 				 * to search for an even better one.
2010 				 */
2011 				if (ifa_maybe == NULL ||
2012 				    ifa_preferred(ifa_maybe, ifa) ||
2013 				    rn_refines((caddr_t)ifa->ifa_netmask,
2014 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2015 					if (ifa_maybe != NULL)
2016 						ifa_free(ifa_maybe);
2017 					ifa_maybe = ifa;
2018 					ifa_ref(ifa_maybe);
2019 				}
2020 			}
2021 		}
2022 		IF_ADDR_RUNLOCK(ifp);
2023 	}
2024 	ifa = ifa_maybe;
2025 	ifa_maybe = NULL;
2026 done:
2027 	IFNET_RUNLOCK_NOSLEEP();
2028 	if (ifa_maybe != NULL)
2029 		ifa_free(ifa_maybe);
2030 	return (ifa);
2031 }
2032 
2033 /*
2034  * Find an interface address specific to an interface best matching
2035  * a given address.
2036  */
2037 struct ifaddr *
2038 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2039 {
2040 	struct ifaddr *ifa;
2041 	const char *cp, *cp2, *cp3;
2042 	char *cplim;
2043 	struct ifaddr *ifa_maybe = NULL;
2044 	u_int af = addr->sa_family;
2045 
2046 	if (af >= AF_MAX)
2047 		return (NULL);
2048 	IF_ADDR_RLOCK(ifp);
2049 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2050 		if (ifa->ifa_addr->sa_family != af)
2051 			continue;
2052 		if (ifa_maybe == NULL)
2053 			ifa_maybe = ifa;
2054 		if (ifa->ifa_netmask == 0) {
2055 			if (sa_equal(addr, ifa->ifa_addr) ||
2056 			    (ifa->ifa_dstaddr &&
2057 			    sa_equal(addr, ifa->ifa_dstaddr)))
2058 				goto done;
2059 			continue;
2060 		}
2061 		if (ifp->if_flags & IFF_POINTOPOINT) {
2062 			if (sa_equal(addr, ifa->ifa_dstaddr))
2063 				goto done;
2064 		} else {
2065 			cp = addr->sa_data;
2066 			cp2 = ifa->ifa_addr->sa_data;
2067 			cp3 = ifa->ifa_netmask->sa_data;
2068 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2069 			for (; cp3 < cplim; cp3++)
2070 				if ((*cp++ ^ *cp2++) & *cp3)
2071 					break;
2072 			if (cp3 == cplim)
2073 				goto done;
2074 		}
2075 	}
2076 	ifa = ifa_maybe;
2077 done:
2078 	if (ifa != NULL)
2079 		ifa_ref(ifa);
2080 	IF_ADDR_RUNLOCK(ifp);
2081 	return (ifa);
2082 }
2083 
2084 /*
2085  * See whether new ifa is better than current one:
2086  * 1) A non-virtual one is preferred over virtual.
2087  * 2) A virtual in master state preferred over any other state.
2088  *
2089  * Used in several address selecting functions.
2090  */
2091 int
2092 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2093 {
2094 
2095 	return (cur->ifa_carp && (!next->ifa_carp ||
2096 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2097 }
2098 
2099 #include <net/if_llatbl.h>
2100 
2101 /*
2102  * Default action when installing a route with a Link Level gateway.
2103  * Lookup an appropriate real ifa to point to.
2104  * This should be moved to /sys/net/link.c eventually.
2105  */
2106 static void
2107 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2108 {
2109 	struct ifaddr *ifa, *oifa;
2110 	struct sockaddr *dst;
2111 	struct ifnet *ifp;
2112 
2113 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2114 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2115 		return;
2116 	ifa = ifaof_ifpforaddr(dst, ifp);
2117 	if (ifa) {
2118 		oifa = rt->rt_ifa;
2119 		rt->rt_ifa = ifa;
2120 		ifa_free(oifa);
2121 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2122 			ifa->ifa_rtrequest(cmd, rt, info);
2123 	}
2124 }
2125 
2126 struct sockaddr_dl *
2127 link_alloc_sdl(size_t size, int flags)
2128 {
2129 
2130 	return (malloc(size, M_TEMP, flags));
2131 }
2132 
2133 void
2134 link_free_sdl(struct sockaddr *sa)
2135 {
2136 	free(sa, M_TEMP);
2137 }
2138 
2139 /*
2140  * Fills in given sdl with interface basic info.
2141  * Returns pointer to filled sdl.
2142  */
2143 struct sockaddr_dl *
2144 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2145 {
2146 	struct sockaddr_dl *sdl;
2147 
2148 	sdl = (struct sockaddr_dl *)paddr;
2149 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2150 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2151 	sdl->sdl_family = AF_LINK;
2152 	sdl->sdl_index = ifp->if_index;
2153 	sdl->sdl_type = iftype;
2154 
2155 	return (sdl);
2156 }
2157 
2158 /*
2159  * Mark an interface down and notify protocols of
2160  * the transition.
2161  */
2162 static void
2163 if_unroute(struct ifnet *ifp, int flag, int fam)
2164 {
2165 	struct ifaddr *ifa;
2166 
2167 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2168 
2169 	ifp->if_flags &= ~flag;
2170 	getmicrotime(&ifp->if_lastchange);
2171 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2172 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2173 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2174 	ifp->if_qflush(ifp);
2175 
2176 	if (ifp->if_carp)
2177 		(*carp_linkstate_p)(ifp);
2178 	rt_ifmsg(ifp);
2179 }
2180 
2181 /*
2182  * Mark an interface up and notify protocols of
2183  * the transition.
2184  */
2185 static void
2186 if_route(struct ifnet *ifp, int flag, int fam)
2187 {
2188 	struct ifaddr *ifa;
2189 
2190 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2191 
2192 	ifp->if_flags |= flag;
2193 	getmicrotime(&ifp->if_lastchange);
2194 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2195 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2196 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2197 	if (ifp->if_carp)
2198 		(*carp_linkstate_p)(ifp);
2199 	rt_ifmsg(ifp);
2200 #ifdef INET6
2201 	in6_if_up(ifp);
2202 #endif
2203 }
2204 
2205 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2206 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2207 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2208 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2209 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2210 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2211 void	*(*vlan_cookie_p)(struct ifnet *);
2212 
2213 /*
2214  * Handle a change in the interface link state. To avoid LORs
2215  * between driver lock and upper layer locks, as well as possible
2216  * recursions, we post event to taskqueue, and all job
2217  * is done in static do_link_state_change().
2218  */
2219 void
2220 if_link_state_change(struct ifnet *ifp, int link_state)
2221 {
2222 	/* Return if state hasn't changed. */
2223 	if (ifp->if_link_state == link_state)
2224 		return;
2225 
2226 	ifp->if_link_state = link_state;
2227 
2228 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2229 }
2230 
2231 static void
2232 do_link_state_change(void *arg, int pending)
2233 {
2234 	struct ifnet *ifp = (struct ifnet *)arg;
2235 	int link_state = ifp->if_link_state;
2236 	CURVNET_SET(ifp->if_vnet);
2237 
2238 	/* Notify that the link state has changed. */
2239 	rt_ifmsg(ifp);
2240 	if (ifp->if_vlantrunk != NULL)
2241 		(*vlan_link_state_p)(ifp);
2242 
2243 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2244 	    ifp->if_l2com != NULL)
2245 		(*ng_ether_link_state_p)(ifp, link_state);
2246 	if (ifp->if_carp)
2247 		(*carp_linkstate_p)(ifp);
2248 	if (ifp->if_bridge)
2249 		(*bridge_linkstate_p)(ifp);
2250 	if (ifp->if_lagg)
2251 		(*lagg_linkstate_p)(ifp, link_state);
2252 
2253 	if (IS_DEFAULT_VNET(curvnet))
2254 		devctl_notify("IFNET", ifp->if_xname,
2255 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2256 		    NULL);
2257 	if (pending > 1)
2258 		if_printf(ifp, "%d link states coalesced\n", pending);
2259 	if (log_link_state_change)
2260 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2261 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2262 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2263 	CURVNET_RESTORE();
2264 }
2265 
2266 /*
2267  * Mark an interface down and notify protocols of
2268  * the transition.
2269  */
2270 void
2271 if_down(struct ifnet *ifp)
2272 {
2273 
2274 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2275 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2276 }
2277 
2278 /*
2279  * Mark an interface up and notify protocols of
2280  * the transition.
2281  */
2282 void
2283 if_up(struct ifnet *ifp)
2284 {
2285 
2286 	if_route(ifp, IFF_UP, AF_UNSPEC);
2287 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2288 }
2289 
2290 /*
2291  * Flush an interface queue.
2292  */
2293 void
2294 if_qflush(struct ifnet *ifp)
2295 {
2296 	struct mbuf *m, *n;
2297 	struct ifaltq *ifq;
2298 
2299 	ifq = &ifp->if_snd;
2300 	IFQ_LOCK(ifq);
2301 #ifdef ALTQ
2302 	if (ALTQ_IS_ENABLED(ifq))
2303 		ALTQ_PURGE(ifq);
2304 #endif
2305 	n = ifq->ifq_head;
2306 	while ((m = n) != NULL) {
2307 		n = m->m_nextpkt;
2308 		m_freem(m);
2309 	}
2310 	ifq->ifq_head = 0;
2311 	ifq->ifq_tail = 0;
2312 	ifq->ifq_len = 0;
2313 	IFQ_UNLOCK(ifq);
2314 }
2315 
2316 /*
2317  * Map interface name to interface structure pointer, with or without
2318  * returning a reference.
2319  */
2320 struct ifnet *
2321 ifunit_ref(const char *name)
2322 {
2323 	struct ifnet *ifp;
2324 
2325 	IFNET_RLOCK_NOSLEEP();
2326 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2327 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2328 		    !(ifp->if_flags & IFF_DYING))
2329 			break;
2330 	}
2331 	if (ifp != NULL)
2332 		if_ref(ifp);
2333 	IFNET_RUNLOCK_NOSLEEP();
2334 	return (ifp);
2335 }
2336 
2337 struct ifnet *
2338 ifunit(const char *name)
2339 {
2340 	struct ifnet *ifp;
2341 
2342 	IFNET_RLOCK_NOSLEEP();
2343 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2344 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2345 			break;
2346 	}
2347 	IFNET_RUNLOCK_NOSLEEP();
2348 	return (ifp);
2349 }
2350 
2351 static void *
2352 ifr_buffer_get_buffer(struct thread *td, void *data)
2353 {
2354 	union ifreq_union *ifrup;
2355 
2356 	ifrup = data;
2357 #ifdef COMPAT_FREEBSD32
2358 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32))
2359 		return ((void *)(uintptr_t)
2360 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2361 #endif
2362 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2363 }
2364 
2365 static void
2366 ifr_buffer_set_buffer_null(struct thread *td, void *data)
2367 {
2368 	union ifreq_union *ifrup;
2369 
2370 	ifrup = data;
2371 #ifdef COMPAT_FREEBSD32
2372 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32))
2373 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2374 	else
2375 #endif
2376 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2377 }
2378 
2379 static size_t
2380 ifr_buffer_get_length(struct thread *td, void *data)
2381 {
2382 	union ifreq_union *ifrup;
2383 
2384 	ifrup = data;
2385 #ifdef COMPAT_FREEBSD32
2386 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32))
2387 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2388 #endif
2389 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2390 }
2391 
2392 static void
2393 ifr_buffer_set_length(struct thread *td, void *data, size_t len)
2394 {
2395 	union ifreq_union *ifrup;
2396 
2397 	ifrup = data;
2398 #ifdef COMPAT_FREEBSD32
2399 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32))
2400 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2401 	else
2402 #endif
2403 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2404 }
2405 
2406 /*
2407  * Hardware specific interface ioctls.
2408  */
2409 static int
2410 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2411 {
2412 	struct ifreq *ifr;
2413 	int error = 0, do_ifup = 0;
2414 	int new_flags, temp_flags;
2415 	size_t namelen, onamelen;
2416 	size_t descrlen;
2417 	char *descrbuf, *odescrbuf;
2418 	char new_name[IFNAMSIZ];
2419 	struct ifaddr *ifa;
2420 	struct sockaddr_dl *sdl;
2421 
2422 	ifr = (struct ifreq *)data;
2423 	switch (cmd) {
2424 	case SIOCGIFINDEX:
2425 		ifr->ifr_index = ifp->if_index;
2426 		break;
2427 
2428 	case SIOCGIFFLAGS:
2429 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2430 		ifr->ifr_flags = temp_flags & 0xffff;
2431 		ifr->ifr_flagshigh = temp_flags >> 16;
2432 		break;
2433 
2434 	case SIOCGIFCAP:
2435 		ifr->ifr_reqcap = ifp->if_capabilities;
2436 		ifr->ifr_curcap = ifp->if_capenable;
2437 		break;
2438 
2439 #ifdef MAC
2440 	case SIOCGIFMAC:
2441 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2442 		break;
2443 #endif
2444 
2445 	case SIOCGIFMETRIC:
2446 		ifr->ifr_metric = ifp->if_metric;
2447 		break;
2448 
2449 	case SIOCGIFMTU:
2450 		ifr->ifr_mtu = ifp->if_mtu;
2451 		break;
2452 
2453 	case SIOCGIFPHYS:
2454 		/* XXXGL: did this ever worked? */
2455 		ifr->ifr_phys = 0;
2456 		break;
2457 
2458 	case SIOCGIFDESCR:
2459 		error = 0;
2460 		sx_slock(&ifdescr_sx);
2461 		if (ifp->if_description == NULL)
2462 			error = ENOMSG;
2463 		else {
2464 			/* space for terminating nul */
2465 			descrlen = strlen(ifp->if_description) + 1;
2466 			if (ifr_buffer_get_length(td, ifr) < descrlen)
2467 				ifr_buffer_set_buffer_null(td, ifr);
2468 			else
2469 				error = copyout(ifp->if_description,
2470 				    ifr_buffer_get_buffer(td, ifr), descrlen);
2471 			ifr_buffer_set_length(td, ifr, descrlen);
2472 		}
2473 		sx_sunlock(&ifdescr_sx);
2474 		break;
2475 
2476 	case SIOCSIFDESCR:
2477 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2478 		if (error)
2479 			return (error);
2480 
2481 		/*
2482 		 * Copy only (length-1) bytes to make sure that
2483 		 * if_description is always nul terminated.  The
2484 		 * length parameter is supposed to count the
2485 		 * terminating nul in.
2486 		 */
2487 		if (ifr_buffer_get_length(td, ifr) > ifdescr_maxlen)
2488 			return (ENAMETOOLONG);
2489 		else if (ifr_buffer_get_length(td, ifr) == 0)
2490 			descrbuf = NULL;
2491 		else {
2492 			descrbuf = malloc(ifr_buffer_get_length(td, ifr),
2493 			    M_IFDESCR, M_WAITOK | M_ZERO);
2494 			error = copyin(ifr_buffer_get_buffer(td, ifr), descrbuf,
2495 			    ifr_buffer_get_length(td, ifr) - 1);
2496 			if (error) {
2497 				free(descrbuf, M_IFDESCR);
2498 				break;
2499 			}
2500 		}
2501 
2502 		sx_xlock(&ifdescr_sx);
2503 		odescrbuf = ifp->if_description;
2504 		ifp->if_description = descrbuf;
2505 		sx_xunlock(&ifdescr_sx);
2506 
2507 		getmicrotime(&ifp->if_lastchange);
2508 		free(odescrbuf, M_IFDESCR);
2509 		break;
2510 
2511 	case SIOCGIFFIB:
2512 		ifr->ifr_fib = ifp->if_fib;
2513 		break;
2514 
2515 	case SIOCSIFFIB:
2516 		error = priv_check(td, PRIV_NET_SETIFFIB);
2517 		if (error)
2518 			return (error);
2519 		if (ifr->ifr_fib >= rt_numfibs)
2520 			return (EINVAL);
2521 
2522 		ifp->if_fib = ifr->ifr_fib;
2523 		break;
2524 
2525 	case SIOCSIFFLAGS:
2526 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2527 		if (error)
2528 			return (error);
2529 		/*
2530 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2531 		 * check, so we don't need special handling here yet.
2532 		 */
2533 		new_flags = (ifr->ifr_flags & 0xffff) |
2534 		    (ifr->ifr_flagshigh << 16);
2535 		if (ifp->if_flags & IFF_UP &&
2536 		    (new_flags & IFF_UP) == 0) {
2537 			if_down(ifp);
2538 		} else if (new_flags & IFF_UP &&
2539 		    (ifp->if_flags & IFF_UP) == 0) {
2540 			do_ifup = 1;
2541 		}
2542 		/* See if permanently promiscuous mode bit is about to flip */
2543 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2544 			if (new_flags & IFF_PPROMISC)
2545 				ifp->if_flags |= IFF_PROMISC;
2546 			else if (ifp->if_pcount == 0)
2547 				ifp->if_flags &= ~IFF_PROMISC;
2548 			if (log_promisc_mode_change)
2549                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2550                                     ifp->if_xname,
2551                                     ((new_flags & IFF_PPROMISC) ?
2552                                      "enabled" : "disabled"));
2553 		}
2554 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2555 			(new_flags &~ IFF_CANTCHANGE);
2556 		if (ifp->if_ioctl) {
2557 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2558 		}
2559 		if (do_ifup)
2560 			if_up(ifp);
2561 		getmicrotime(&ifp->if_lastchange);
2562 		break;
2563 
2564 	case SIOCSIFCAP:
2565 		error = priv_check(td, PRIV_NET_SETIFCAP);
2566 		if (error)
2567 			return (error);
2568 		if (ifp->if_ioctl == NULL)
2569 			return (EOPNOTSUPP);
2570 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2571 			return (EINVAL);
2572 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2573 		if (error == 0)
2574 			getmicrotime(&ifp->if_lastchange);
2575 		break;
2576 
2577 #ifdef MAC
2578 	case SIOCSIFMAC:
2579 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2580 		break;
2581 #endif
2582 
2583 	case SIOCSIFNAME:
2584 		error = priv_check(td, PRIV_NET_SETIFNAME);
2585 		if (error)
2586 			return (error);
2587 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2588 		if (error != 0)
2589 			return (error);
2590 		if (new_name[0] == '\0')
2591 			return (EINVAL);
2592 		if (new_name[IFNAMSIZ-1] != '\0') {
2593 			new_name[IFNAMSIZ-1] = '\0';
2594 			if (strlen(new_name) == IFNAMSIZ-1)
2595 				return (EINVAL);
2596 		}
2597 		if (ifunit(new_name) != NULL)
2598 			return (EEXIST);
2599 
2600 		/*
2601 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2602 		 * and there are numerous other races with the
2603 		 * ifunit() checks not being atomic with namespace
2604 		 * changes (renames, vmoves, if_attach, etc).
2605 		 */
2606 		ifp->if_flags |= IFF_RENAMING;
2607 
2608 		/* Announce the departure of the interface. */
2609 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2610 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2611 
2612 		log(LOG_INFO, "%s: changing name to '%s'\n",
2613 		    ifp->if_xname, new_name);
2614 
2615 		IF_ADDR_WLOCK(ifp);
2616 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2617 		ifa = ifp->if_addr;
2618 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2619 		namelen = strlen(new_name);
2620 		onamelen = sdl->sdl_nlen;
2621 		/*
2622 		 * Move the address if needed.  This is safe because we
2623 		 * allocate space for a name of length IFNAMSIZ when we
2624 		 * create this in if_attach().
2625 		 */
2626 		if (namelen != onamelen) {
2627 			bcopy(sdl->sdl_data + onamelen,
2628 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2629 		}
2630 		bcopy(new_name, sdl->sdl_data, namelen);
2631 		sdl->sdl_nlen = namelen;
2632 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2633 		bzero(sdl->sdl_data, onamelen);
2634 		while (namelen != 0)
2635 			sdl->sdl_data[--namelen] = 0xff;
2636 		IF_ADDR_WUNLOCK(ifp);
2637 
2638 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2639 		/* Announce the return of the interface. */
2640 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2641 
2642 		ifp->if_flags &= ~IFF_RENAMING;
2643 		break;
2644 
2645 #ifdef VIMAGE
2646 	case SIOCSIFVNET:
2647 		error = priv_check(td, PRIV_NET_SETIFVNET);
2648 		if (error)
2649 			return (error);
2650 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2651 		break;
2652 #endif
2653 
2654 	case SIOCSIFMETRIC:
2655 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2656 		if (error)
2657 			return (error);
2658 		ifp->if_metric = ifr->ifr_metric;
2659 		getmicrotime(&ifp->if_lastchange);
2660 		break;
2661 
2662 	case SIOCSIFPHYS:
2663 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2664 		if (error)
2665 			return (error);
2666 		if (ifp->if_ioctl == NULL)
2667 			return (EOPNOTSUPP);
2668 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2669 		if (error == 0)
2670 			getmicrotime(&ifp->if_lastchange);
2671 		break;
2672 
2673 	case SIOCSIFMTU:
2674 	{
2675 		u_long oldmtu = ifp->if_mtu;
2676 
2677 		error = priv_check(td, PRIV_NET_SETIFMTU);
2678 		if (error)
2679 			return (error);
2680 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2681 			return (EINVAL);
2682 		if (ifp->if_ioctl == NULL)
2683 			return (EOPNOTSUPP);
2684 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2685 		if (error == 0) {
2686 			getmicrotime(&ifp->if_lastchange);
2687 			rt_ifmsg(ifp);
2688 		}
2689 		/*
2690 		 * If the link MTU changed, do network layer specific procedure.
2691 		 */
2692 		if (ifp->if_mtu != oldmtu) {
2693 #ifdef INET6
2694 			nd6_setmtu(ifp);
2695 #endif
2696 			rt_updatemtu(ifp);
2697 		}
2698 		break;
2699 	}
2700 
2701 	case SIOCADDMULTI:
2702 	case SIOCDELMULTI:
2703 		if (cmd == SIOCADDMULTI)
2704 			error = priv_check(td, PRIV_NET_ADDMULTI);
2705 		else
2706 			error = priv_check(td, PRIV_NET_DELMULTI);
2707 		if (error)
2708 			return (error);
2709 
2710 		/* Don't allow group membership on non-multicast interfaces. */
2711 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2712 			return (EOPNOTSUPP);
2713 
2714 		/* Don't let users screw up protocols' entries. */
2715 		if (ifr->ifr_addr.sa_family != AF_LINK)
2716 			return (EINVAL);
2717 
2718 		if (cmd == SIOCADDMULTI) {
2719 			struct ifmultiaddr *ifma;
2720 
2721 			/*
2722 			 * Userland is only permitted to join groups once
2723 			 * via the if_addmulti() KPI, because it cannot hold
2724 			 * struct ifmultiaddr * between calls. It may also
2725 			 * lose a race while we check if the membership
2726 			 * already exists.
2727 			 */
2728 			IF_ADDR_RLOCK(ifp);
2729 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2730 			IF_ADDR_RUNLOCK(ifp);
2731 			if (ifma != NULL)
2732 				error = EADDRINUSE;
2733 			else
2734 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2735 		} else {
2736 			error = if_delmulti(ifp, &ifr->ifr_addr);
2737 		}
2738 		if (error == 0)
2739 			getmicrotime(&ifp->if_lastchange);
2740 		break;
2741 
2742 	case SIOCSIFPHYADDR:
2743 	case SIOCDIFPHYADDR:
2744 #ifdef INET6
2745 	case SIOCSIFPHYADDR_IN6:
2746 #endif
2747 	case SIOCSIFMEDIA:
2748 	case SIOCSIFGENERIC:
2749 		error = priv_check(td, PRIV_NET_HWIOCTL);
2750 		if (error)
2751 			return (error);
2752 		if (ifp->if_ioctl == NULL)
2753 			return (EOPNOTSUPP);
2754 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2755 		if (error == 0)
2756 			getmicrotime(&ifp->if_lastchange);
2757 		break;
2758 
2759 	case SIOCGIFSTATUS:
2760 	case SIOCGIFPSRCADDR:
2761 	case SIOCGIFPDSTADDR:
2762 	case SIOCGIFMEDIA:
2763 	case SIOCGIFXMEDIA:
2764 	case SIOCGIFGENERIC:
2765 	case SIOCGIFRSSKEY:
2766 	case SIOCGIFRSSHASH:
2767 		if (ifp->if_ioctl == NULL)
2768 			return (EOPNOTSUPP);
2769 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2770 		break;
2771 
2772 	case SIOCSIFLLADDR:
2773 		error = priv_check(td, PRIV_NET_SETLLADDR);
2774 		if (error)
2775 			return (error);
2776 		error = if_setlladdr(ifp,
2777 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2778 		break;
2779 
2780 	case SIOCGHWADDR:
2781 		error = if_gethwaddr(ifp, ifr);
2782 		break;
2783 
2784 	case SIOCAIFGROUP:
2785 	{
2786 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2787 
2788 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2789 		if (error)
2790 			return (error);
2791 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2792 			return (error);
2793 		break;
2794 	}
2795 
2796 	case SIOCGIFGROUP:
2797 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2798 			return (error);
2799 		break;
2800 
2801 	case SIOCDIFGROUP:
2802 	{
2803 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2804 
2805 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2806 		if (error)
2807 			return (error);
2808 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2809 			return (error);
2810 		break;
2811 	}
2812 
2813 	default:
2814 		error = ENOIOCTL;
2815 		break;
2816 	}
2817 	return (error);
2818 }
2819 
2820 #ifdef COMPAT_FREEBSD32
2821 struct ifconf32 {
2822 	int32_t	ifc_len;
2823 	union {
2824 		uint32_t	ifcu_buf;
2825 		uint32_t	ifcu_req;
2826 	} ifc_ifcu;
2827 };
2828 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2829 #endif
2830 
2831 /*
2832  * Interface ioctls.
2833  */
2834 int
2835 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2836 {
2837 	struct ifnet *ifp;
2838 	struct ifreq *ifr;
2839 	int error;
2840 	int oif_flags;
2841 #ifdef VIMAGE
2842 	int shutdown;
2843 #endif
2844 
2845 	CURVNET_SET(so->so_vnet);
2846 #ifdef VIMAGE
2847 	/* Make sure the VNET is stable. */
2848 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2849 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2850 	if (shutdown) {
2851 		CURVNET_RESTORE();
2852 		return (EBUSY);
2853 	}
2854 #endif
2855 
2856 
2857 	switch (cmd) {
2858 	case SIOCGIFCONF:
2859 		error = ifconf(cmd, data);
2860 		CURVNET_RESTORE();
2861 		return (error);
2862 
2863 #ifdef COMPAT_FREEBSD32
2864 	case SIOCGIFCONF32:
2865 		{
2866 			struct ifconf32 *ifc32;
2867 			struct ifconf ifc;
2868 
2869 			ifc32 = (struct ifconf32 *)data;
2870 			ifc.ifc_len = ifc32->ifc_len;
2871 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2872 
2873 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2874 			CURVNET_RESTORE();
2875 			if (error == 0)
2876 				ifc32->ifc_len = ifc.ifc_len;
2877 			return (error);
2878 		}
2879 #endif
2880 	}
2881 	ifr = (struct ifreq *)data;
2882 
2883 	switch (cmd) {
2884 #ifdef VIMAGE
2885 	case SIOCSIFRVNET:
2886 		error = priv_check(td, PRIV_NET_SETIFVNET);
2887 		if (error == 0)
2888 			error = if_vmove_reclaim(td, ifr->ifr_name,
2889 			    ifr->ifr_jid);
2890 		CURVNET_RESTORE();
2891 		return (error);
2892 #endif
2893 	case SIOCIFCREATE:
2894 	case SIOCIFCREATE2:
2895 		error = priv_check(td, PRIV_NET_IFCREATE);
2896 		if (error == 0)
2897 			error = if_clone_create(ifr->ifr_name,
2898 			    sizeof(ifr->ifr_name),
2899 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2900 		CURVNET_RESTORE();
2901 		return (error);
2902 	case SIOCIFDESTROY:
2903 		error = priv_check(td, PRIV_NET_IFDESTROY);
2904 		if (error == 0)
2905 			error = if_clone_destroy(ifr->ifr_name);
2906 		CURVNET_RESTORE();
2907 		return (error);
2908 
2909 	case SIOCIFGCLONERS:
2910 		error = if_clone_list((struct if_clonereq *)data);
2911 		CURVNET_RESTORE();
2912 		return (error);
2913 	case SIOCGIFGMEMB:
2914 		error = if_getgroupmembers((struct ifgroupreq *)data);
2915 		CURVNET_RESTORE();
2916 		return (error);
2917 #if defined(INET) || defined(INET6)
2918 	case SIOCSVH:
2919 	case SIOCGVH:
2920 		if (carp_ioctl_p == NULL)
2921 			error = EPROTONOSUPPORT;
2922 		else
2923 			error = (*carp_ioctl_p)(ifr, cmd, td);
2924 		CURVNET_RESTORE();
2925 		return (error);
2926 #endif
2927 	}
2928 
2929 	ifp = ifunit_ref(ifr->ifr_name);
2930 	if (ifp == NULL) {
2931 		CURVNET_RESTORE();
2932 		return (ENXIO);
2933 	}
2934 
2935 	error = ifhwioctl(cmd, ifp, data, td);
2936 	if (error != ENOIOCTL) {
2937 		if_rele(ifp);
2938 		CURVNET_RESTORE();
2939 		return (error);
2940 	}
2941 
2942 	oif_flags = ifp->if_flags;
2943 	if (so->so_proto == NULL) {
2944 		if_rele(ifp);
2945 		CURVNET_RESTORE();
2946 		return (EOPNOTSUPP);
2947 	}
2948 
2949 	/*
2950 	 * Pass the request on to the socket control method, and if the
2951 	 * latter returns EOPNOTSUPP, directly to the interface.
2952 	 *
2953 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2954 	 * trust SIOCSIFADDR et al to come from an already privileged
2955 	 * layer, and do not perform any credentials checks or input
2956 	 * validation.
2957 	 */
2958 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2959 	    ifp, td));
2960 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2961 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2962 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2963 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2964 
2965 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2966 #ifdef INET6
2967 		if (ifp->if_flags & IFF_UP)
2968 			in6_if_up(ifp);
2969 #endif
2970 	}
2971 	if_rele(ifp);
2972 	CURVNET_RESTORE();
2973 	return (error);
2974 }
2975 
2976 /*
2977  * The code common to handling reference counted flags,
2978  * e.g., in ifpromisc() and if_allmulti().
2979  * The "pflag" argument can specify a permanent mode flag to check,
2980  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2981  *
2982  * Only to be used on stack-owned flags, not driver-owned flags.
2983  */
2984 static int
2985 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2986 {
2987 	struct ifreq ifr;
2988 	int error;
2989 	int oldflags, oldcount;
2990 
2991 	/* Sanity checks to catch programming errors */
2992 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2993 	    ("%s: setting driver-owned flag %d", __func__, flag));
2994 
2995 	if (onswitch)
2996 		KASSERT(*refcount >= 0,
2997 		    ("%s: increment negative refcount %d for flag %d",
2998 		    __func__, *refcount, flag));
2999 	else
3000 		KASSERT(*refcount > 0,
3001 		    ("%s: decrement non-positive refcount %d for flag %d",
3002 		    __func__, *refcount, flag));
3003 
3004 	/* In case this mode is permanent, just touch refcount */
3005 	if (ifp->if_flags & pflag) {
3006 		*refcount += onswitch ? 1 : -1;
3007 		return (0);
3008 	}
3009 
3010 	/* Save ifnet parameters for if_ioctl() may fail */
3011 	oldcount = *refcount;
3012 	oldflags = ifp->if_flags;
3013 
3014 	/*
3015 	 * See if we aren't the only and touching refcount is enough.
3016 	 * Actually toggle interface flag if we are the first or last.
3017 	 */
3018 	if (onswitch) {
3019 		if ((*refcount)++)
3020 			return (0);
3021 		ifp->if_flags |= flag;
3022 	} else {
3023 		if (--(*refcount))
3024 			return (0);
3025 		ifp->if_flags &= ~flag;
3026 	}
3027 
3028 	/* Call down the driver since we've changed interface flags */
3029 	if (ifp->if_ioctl == NULL) {
3030 		error = EOPNOTSUPP;
3031 		goto recover;
3032 	}
3033 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3034 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3035 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3036 	if (error)
3037 		goto recover;
3038 	/* Notify userland that interface flags have changed */
3039 	rt_ifmsg(ifp);
3040 	return (0);
3041 
3042 recover:
3043 	/* Recover after driver error */
3044 	*refcount = oldcount;
3045 	ifp->if_flags = oldflags;
3046 	return (error);
3047 }
3048 
3049 /*
3050  * Set/clear promiscuous mode on interface ifp based on the truth value
3051  * of pswitch.  The calls are reference counted so that only the first
3052  * "on" request actually has an effect, as does the final "off" request.
3053  * Results are undefined if the "off" and "on" requests are not matched.
3054  */
3055 int
3056 ifpromisc(struct ifnet *ifp, int pswitch)
3057 {
3058 	int error;
3059 	int oldflags = ifp->if_flags;
3060 
3061 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3062 			   &ifp->if_pcount, pswitch);
3063 	/* If promiscuous mode status has changed, log a message */
3064 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3065             log_promisc_mode_change)
3066 		log(LOG_INFO, "%s: promiscuous mode %s\n",
3067 		    ifp->if_xname,
3068 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3069 	return (error);
3070 }
3071 
3072 /*
3073  * Return interface configuration
3074  * of system.  List may be used
3075  * in later ioctl's (above) to get
3076  * other information.
3077  */
3078 /*ARGSUSED*/
3079 static int
3080 ifconf(u_long cmd, caddr_t data)
3081 {
3082 	struct ifconf *ifc = (struct ifconf *)data;
3083 	struct ifnet *ifp;
3084 	struct ifaddr *ifa;
3085 	struct ifreq ifr;
3086 	struct sbuf *sb;
3087 	int error, full = 0, valid_len, max_len;
3088 
3089 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3090 	max_len = MAXPHYS - 1;
3091 
3092 	/* Prevent hostile input from being able to crash the system */
3093 	if (ifc->ifc_len <= 0)
3094 		return (EINVAL);
3095 
3096 again:
3097 	if (ifc->ifc_len <= max_len) {
3098 		max_len = ifc->ifc_len;
3099 		full = 1;
3100 	}
3101 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3102 	max_len = 0;
3103 	valid_len = 0;
3104 
3105 	IFNET_RLOCK();
3106 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3107 		int addrs;
3108 
3109 		/*
3110 		 * Zero the ifr_name buffer to make sure we don't
3111 		 * disclose the contents of the stack.
3112 		 */
3113 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
3114 
3115 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3116 		    >= sizeof(ifr.ifr_name)) {
3117 			sbuf_delete(sb);
3118 			IFNET_RUNLOCK();
3119 			return (ENAMETOOLONG);
3120 		}
3121 
3122 		addrs = 0;
3123 		IF_ADDR_RLOCK(ifp);
3124 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3125 			struct sockaddr *sa = ifa->ifa_addr;
3126 
3127 			if (prison_if(curthread->td_ucred, sa) != 0)
3128 				continue;
3129 			addrs++;
3130 			if (sa->sa_len <= sizeof(*sa)) {
3131 				ifr.ifr_addr = *sa;
3132 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3133 				max_len += sizeof(ifr);
3134 			} else {
3135 				sbuf_bcat(sb, &ifr,
3136 				    offsetof(struct ifreq, ifr_addr));
3137 				max_len += offsetof(struct ifreq, ifr_addr);
3138 				sbuf_bcat(sb, sa, sa->sa_len);
3139 				max_len += sa->sa_len;
3140 			}
3141 
3142 			if (sbuf_error(sb) == 0)
3143 				valid_len = sbuf_len(sb);
3144 		}
3145 		IF_ADDR_RUNLOCK(ifp);
3146 		if (addrs == 0) {
3147 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3148 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3149 			max_len += sizeof(ifr);
3150 
3151 			if (sbuf_error(sb) == 0)
3152 				valid_len = sbuf_len(sb);
3153 		}
3154 	}
3155 	IFNET_RUNLOCK();
3156 
3157 	/*
3158 	 * If we didn't allocate enough space (uncommon), try again.  If
3159 	 * we have already allocated as much space as we are allowed,
3160 	 * return what we've got.
3161 	 */
3162 	if (valid_len != max_len && !full) {
3163 		sbuf_delete(sb);
3164 		goto again;
3165 	}
3166 
3167 	ifc->ifc_len = valid_len;
3168 	sbuf_finish(sb);
3169 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3170 	sbuf_delete(sb);
3171 	return (error);
3172 }
3173 
3174 /*
3175  * Just like ifpromisc(), but for all-multicast-reception mode.
3176  */
3177 int
3178 if_allmulti(struct ifnet *ifp, int onswitch)
3179 {
3180 
3181 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3182 }
3183 
3184 struct ifmultiaddr *
3185 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3186 {
3187 	struct ifmultiaddr *ifma;
3188 
3189 	IF_ADDR_LOCK_ASSERT(ifp);
3190 
3191 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3192 		if (sa->sa_family == AF_LINK) {
3193 			if (sa_dl_equal(ifma->ifma_addr, sa))
3194 				break;
3195 		} else {
3196 			if (sa_equal(ifma->ifma_addr, sa))
3197 				break;
3198 		}
3199 	}
3200 
3201 	return ifma;
3202 }
3203 
3204 /*
3205  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3206  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3207  * the ifnet multicast address list here, so the caller must do that and
3208  * other setup work (such as notifying the device driver).  The reference
3209  * count is initialized to 1.
3210  */
3211 static struct ifmultiaddr *
3212 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3213     int mflags)
3214 {
3215 	struct ifmultiaddr *ifma;
3216 	struct sockaddr *dupsa;
3217 
3218 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3219 	    M_ZERO);
3220 	if (ifma == NULL)
3221 		return (NULL);
3222 
3223 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3224 	if (dupsa == NULL) {
3225 		free(ifma, M_IFMADDR);
3226 		return (NULL);
3227 	}
3228 	bcopy(sa, dupsa, sa->sa_len);
3229 	ifma->ifma_addr = dupsa;
3230 
3231 	ifma->ifma_ifp = ifp;
3232 	ifma->ifma_refcount = 1;
3233 	ifma->ifma_protospec = NULL;
3234 
3235 	if (llsa == NULL) {
3236 		ifma->ifma_lladdr = NULL;
3237 		return (ifma);
3238 	}
3239 
3240 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3241 	if (dupsa == NULL) {
3242 		free(ifma->ifma_addr, M_IFMADDR);
3243 		free(ifma, M_IFMADDR);
3244 		return (NULL);
3245 	}
3246 	bcopy(llsa, dupsa, llsa->sa_len);
3247 	ifma->ifma_lladdr = dupsa;
3248 
3249 	return (ifma);
3250 }
3251 
3252 /*
3253  * if_freemulti: free ifmultiaddr structure and possibly attached related
3254  * addresses.  The caller is responsible for implementing reference
3255  * counting, notifying the driver, handling routing messages, and releasing
3256  * any dependent link layer state.
3257  */
3258 static void
3259 if_freemulti(struct ifmultiaddr *ifma)
3260 {
3261 
3262 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3263 	    ifma->ifma_refcount));
3264 
3265 	if (ifma->ifma_lladdr != NULL)
3266 		free(ifma->ifma_lladdr, M_IFMADDR);
3267 	free(ifma->ifma_addr, M_IFMADDR);
3268 	free(ifma, M_IFMADDR);
3269 }
3270 
3271 /*
3272  * Register an additional multicast address with a network interface.
3273  *
3274  * - If the address is already present, bump the reference count on the
3275  *   address and return.
3276  * - If the address is not link-layer, look up a link layer address.
3277  * - Allocate address structures for one or both addresses, and attach to the
3278  *   multicast address list on the interface.  If automatically adding a link
3279  *   layer address, the protocol address will own a reference to the link
3280  *   layer address, to be freed when it is freed.
3281  * - Notify the network device driver of an addition to the multicast address
3282  *   list.
3283  *
3284  * 'sa' points to caller-owned memory with the desired multicast address.
3285  *
3286  * 'retifma' will be used to return a pointer to the resulting multicast
3287  * address reference, if desired.
3288  */
3289 int
3290 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3291     struct ifmultiaddr **retifma)
3292 {
3293 	struct ifmultiaddr *ifma, *ll_ifma;
3294 	struct sockaddr *llsa;
3295 	struct sockaddr_dl sdl;
3296 	int error;
3297 
3298 	/*
3299 	 * If the address is already present, return a new reference to it;
3300 	 * otherwise, allocate storage and set up a new address.
3301 	 */
3302 	IF_ADDR_WLOCK(ifp);
3303 	ifma = if_findmulti(ifp, sa);
3304 	if (ifma != NULL) {
3305 		ifma->ifma_refcount++;
3306 		if (retifma != NULL)
3307 			*retifma = ifma;
3308 		IF_ADDR_WUNLOCK(ifp);
3309 		return (0);
3310 	}
3311 
3312 	/*
3313 	 * The address isn't already present; resolve the protocol address
3314 	 * into a link layer address, and then look that up, bump its
3315 	 * refcount or allocate an ifma for that also.
3316 	 * Most link layer resolving functions returns address data which
3317 	 * fits inside default sockaddr_dl structure. However callback
3318 	 * can allocate another sockaddr structure, in that case we need to
3319 	 * free it later.
3320 	 */
3321 	llsa = NULL;
3322 	ll_ifma = NULL;
3323 	if (ifp->if_resolvemulti != NULL) {
3324 		/* Provide called function with buffer size information */
3325 		sdl.sdl_len = sizeof(sdl);
3326 		llsa = (struct sockaddr *)&sdl;
3327 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3328 		if (error)
3329 			goto unlock_out;
3330 	}
3331 
3332 	/*
3333 	 * Allocate the new address.  Don't hook it up yet, as we may also
3334 	 * need to allocate a link layer multicast address.
3335 	 */
3336 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3337 	if (ifma == NULL) {
3338 		error = ENOMEM;
3339 		goto free_llsa_out;
3340 	}
3341 
3342 	/*
3343 	 * If a link layer address is found, we'll need to see if it's
3344 	 * already present in the address list, or allocate is as well.
3345 	 * When this block finishes, the link layer address will be on the
3346 	 * list.
3347 	 */
3348 	if (llsa != NULL) {
3349 		ll_ifma = if_findmulti(ifp, llsa);
3350 		if (ll_ifma == NULL) {
3351 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3352 			if (ll_ifma == NULL) {
3353 				--ifma->ifma_refcount;
3354 				if_freemulti(ifma);
3355 				error = ENOMEM;
3356 				goto free_llsa_out;
3357 			}
3358 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3359 			    ifma_link);
3360 		} else
3361 			ll_ifma->ifma_refcount++;
3362 		ifma->ifma_llifma = ll_ifma;
3363 	}
3364 
3365 	/*
3366 	 * We now have a new multicast address, ifma, and possibly a new or
3367 	 * referenced link layer address.  Add the primary address to the
3368 	 * ifnet address list.
3369 	 */
3370 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3371 
3372 	if (retifma != NULL)
3373 		*retifma = ifma;
3374 
3375 	/*
3376 	 * Must generate the message while holding the lock so that 'ifma'
3377 	 * pointer is still valid.
3378 	 */
3379 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3380 	IF_ADDR_WUNLOCK(ifp);
3381 
3382 	/*
3383 	 * We are certain we have added something, so call down to the
3384 	 * interface to let them know about it.
3385 	 */
3386 	if (ifp->if_ioctl != NULL) {
3387 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3388 	}
3389 
3390 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3391 		link_free_sdl(llsa);
3392 
3393 	return (0);
3394 
3395 free_llsa_out:
3396 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3397 		link_free_sdl(llsa);
3398 
3399 unlock_out:
3400 	IF_ADDR_WUNLOCK(ifp);
3401 	return (error);
3402 }
3403 
3404 /*
3405  * Delete a multicast group membership by network-layer group address.
3406  *
3407  * Returns ENOENT if the entry could not be found. If ifp no longer
3408  * exists, results are undefined. This entry point should only be used
3409  * from subsystems which do appropriate locking to hold ifp for the
3410  * duration of the call.
3411  * Network-layer protocol domains must use if_delmulti_ifma().
3412  */
3413 int
3414 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3415 {
3416 	struct ifmultiaddr *ifma;
3417 	int lastref;
3418 #ifdef INVARIANTS
3419 	struct ifnet *oifp;
3420 
3421 	IFNET_RLOCK_NOSLEEP();
3422 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3423 		if (ifp == oifp)
3424 			break;
3425 	if (ifp != oifp)
3426 		ifp = NULL;
3427 	IFNET_RUNLOCK_NOSLEEP();
3428 
3429 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3430 #endif
3431 	if (ifp == NULL)
3432 		return (ENOENT);
3433 
3434 	IF_ADDR_WLOCK(ifp);
3435 	lastref = 0;
3436 	ifma = if_findmulti(ifp, sa);
3437 	if (ifma != NULL)
3438 		lastref = if_delmulti_locked(ifp, ifma, 0);
3439 	IF_ADDR_WUNLOCK(ifp);
3440 
3441 	if (ifma == NULL)
3442 		return (ENOENT);
3443 
3444 	if (lastref && ifp->if_ioctl != NULL) {
3445 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3446 	}
3447 
3448 	return (0);
3449 }
3450 
3451 /*
3452  * Delete all multicast group membership for an interface.
3453  * Should be used to quickly flush all multicast filters.
3454  */
3455 void
3456 if_delallmulti(struct ifnet *ifp)
3457 {
3458 	struct ifmultiaddr *ifma;
3459 	struct ifmultiaddr *next;
3460 
3461 	IF_ADDR_WLOCK(ifp);
3462 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3463 		if_delmulti_locked(ifp, ifma, 0);
3464 	IF_ADDR_WUNLOCK(ifp);
3465 }
3466 
3467 /*
3468  * Delete a multicast group membership by group membership pointer.
3469  * Network-layer protocol domains must use this routine.
3470  *
3471  * It is safe to call this routine if the ifp disappeared.
3472  */
3473 void
3474 if_delmulti_ifma(struct ifmultiaddr *ifma)
3475 {
3476 	struct ifnet *ifp;
3477 	int lastref;
3478 
3479 	ifp = ifma->ifma_ifp;
3480 #ifdef DIAGNOSTIC
3481 	if (ifp == NULL) {
3482 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3483 	} else {
3484 		struct ifnet *oifp;
3485 
3486 		IFNET_RLOCK_NOSLEEP();
3487 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3488 			if (ifp == oifp)
3489 				break;
3490 		if (ifp != oifp) {
3491 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3492 			ifp = NULL;
3493 		}
3494 		IFNET_RUNLOCK_NOSLEEP();
3495 	}
3496 #endif
3497 	/*
3498 	 * If and only if the ifnet instance exists: Acquire the address lock.
3499 	 */
3500 	if (ifp != NULL)
3501 		IF_ADDR_WLOCK(ifp);
3502 
3503 	lastref = if_delmulti_locked(ifp, ifma, 0);
3504 
3505 	if (ifp != NULL) {
3506 		/*
3507 		 * If and only if the ifnet instance exists:
3508 		 *  Release the address lock.
3509 		 *  If the group was left: update the hardware hash filter.
3510 		 */
3511 		IF_ADDR_WUNLOCK(ifp);
3512 		if (lastref && ifp->if_ioctl != NULL) {
3513 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3514 		}
3515 	}
3516 }
3517 
3518 /*
3519  * Perform deletion of network-layer and/or link-layer multicast address.
3520  *
3521  * Return 0 if the reference count was decremented.
3522  * Return 1 if the final reference was released, indicating that the
3523  * hardware hash filter should be reprogrammed.
3524  */
3525 static int
3526 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3527 {
3528 	struct ifmultiaddr *ll_ifma;
3529 
3530 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3531 		KASSERT(ifma->ifma_ifp == ifp,
3532 		    ("%s: inconsistent ifp %p", __func__, ifp));
3533 		IF_ADDR_WLOCK_ASSERT(ifp);
3534 	}
3535 
3536 	ifp = ifma->ifma_ifp;
3537 
3538 	/*
3539 	 * If the ifnet is detaching, null out references to ifnet,
3540 	 * so that upper protocol layers will notice, and not attempt
3541 	 * to obtain locks for an ifnet which no longer exists. The
3542 	 * routing socket announcement must happen before the ifnet
3543 	 * instance is detached from the system.
3544 	 */
3545 	if (detaching) {
3546 #ifdef DIAGNOSTIC
3547 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3548 #endif
3549 		/*
3550 		 * ifp may already be nulled out if we are being reentered
3551 		 * to delete the ll_ifma.
3552 		 */
3553 		if (ifp != NULL) {
3554 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3555 			ifma->ifma_ifp = NULL;
3556 		}
3557 	}
3558 
3559 	if (--ifma->ifma_refcount > 0)
3560 		return 0;
3561 
3562 	/*
3563 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3564 	 * have been associated with it. Release it first if so.
3565 	 */
3566 	ll_ifma = ifma->ifma_llifma;
3567 	if (ll_ifma != NULL) {
3568 		KASSERT(ifma->ifma_lladdr != NULL,
3569 		    ("%s: llifma w/o lladdr", __func__));
3570 		if (detaching)
3571 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3572 		if (--ll_ifma->ifma_refcount == 0) {
3573 			if (ifp != NULL) {
3574 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3575 				    ifma_link);
3576 			}
3577 			if_freemulti(ll_ifma);
3578 		}
3579 	}
3580 
3581 	if (ifp != NULL)
3582 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3583 
3584 	if_freemulti(ifma);
3585 
3586 	/*
3587 	 * The last reference to this instance of struct ifmultiaddr
3588 	 * was released; the hardware should be notified of this change.
3589 	 */
3590 	return 1;
3591 }
3592 
3593 /*
3594  * Set the link layer address on an interface.
3595  *
3596  * At this time we only support certain types of interfaces,
3597  * and we don't allow the length of the address to change.
3598  *
3599  * Set noinline to be dtrace-friendly
3600  */
3601 __noinline int
3602 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3603 {
3604 	struct sockaddr_dl *sdl;
3605 	struct ifaddr *ifa;
3606 	struct ifreq ifr;
3607 
3608 	IF_ADDR_RLOCK(ifp);
3609 	ifa = ifp->if_addr;
3610 	if (ifa == NULL) {
3611 		IF_ADDR_RUNLOCK(ifp);
3612 		return (EINVAL);
3613 	}
3614 	ifa_ref(ifa);
3615 	IF_ADDR_RUNLOCK(ifp);
3616 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3617 	if (sdl == NULL) {
3618 		ifa_free(ifa);
3619 		return (EINVAL);
3620 	}
3621 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3622 		ifa_free(ifa);
3623 		return (EINVAL);
3624 	}
3625 	switch (ifp->if_type) {
3626 	case IFT_ETHER:
3627 	case IFT_FDDI:
3628 	case IFT_XETHER:
3629 	case IFT_ISO88025:
3630 	case IFT_L2VLAN:
3631 	case IFT_BRIDGE:
3632 	case IFT_ARCNET:
3633 	case IFT_IEEE8023ADLAG:
3634 		bcopy(lladdr, LLADDR(sdl), len);
3635 		ifa_free(ifa);
3636 		break;
3637 	default:
3638 		ifa_free(ifa);
3639 		return (ENODEV);
3640 	}
3641 
3642 	/*
3643 	 * If the interface is already up, we need
3644 	 * to re-init it in order to reprogram its
3645 	 * address filter.
3646 	 */
3647 	if ((ifp->if_flags & IFF_UP) != 0) {
3648 		if (ifp->if_ioctl) {
3649 			ifp->if_flags &= ~IFF_UP;
3650 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3651 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3652 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3653 			ifp->if_flags |= IFF_UP;
3654 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3655 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3656 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3657 		}
3658 	}
3659 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3660 	return (0);
3661 }
3662 
3663 /*
3664  * Compat function for handling basic encapsulation requests.
3665  * Not converted stacks (FDDI, IB, ..) supports traditional
3666  * output model: ARP (and other similar L2 protocols) are handled
3667  * inside output routine, arpresolve/nd6_resolve() returns MAC
3668  * address instead of full prepend.
3669  *
3670  * This function creates calculated header==MAC for IPv4/IPv6 and
3671  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3672  * address families.
3673  */
3674 static int
3675 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3676 {
3677 
3678 	if (req->rtype != IFENCAP_LL)
3679 		return (EOPNOTSUPP);
3680 
3681 	if (req->bufsize < req->lladdr_len)
3682 		return (ENOMEM);
3683 
3684 	switch (req->family) {
3685 	case AF_INET:
3686 	case AF_INET6:
3687 		break;
3688 	default:
3689 		return (EAFNOSUPPORT);
3690 	}
3691 
3692 	/* Copy lladdr to storage as is */
3693 	memmove(req->buf, req->lladdr, req->lladdr_len);
3694 	req->bufsize = req->lladdr_len;
3695 	req->lladdr_off = 0;
3696 
3697 	return (0);
3698 }
3699 
3700 /*
3701  * Get the link layer address that was read from the hardware at attach.
3702  *
3703  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3704  * their component interfaces as IFT_IEEE8023ADLAG.
3705  */
3706 int
3707 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3708 {
3709 
3710 	if (ifp->if_hw_addr == NULL)
3711 		return (ENODEV);
3712 
3713 	switch (ifp->if_type) {
3714 	case IFT_ETHER:
3715 	case IFT_IEEE8023ADLAG:
3716 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3717 		return (0);
3718 	default:
3719 		return (ENODEV);
3720 	}
3721 }
3722 
3723 /*
3724  * The name argument must be a pointer to storage which will last as
3725  * long as the interface does.  For physical devices, the result of
3726  * device_get_name(dev) is a good choice and for pseudo-devices a
3727  * static string works well.
3728  */
3729 void
3730 if_initname(struct ifnet *ifp, const char *name, int unit)
3731 {
3732 	ifp->if_dname = name;
3733 	ifp->if_dunit = unit;
3734 	if (unit != IF_DUNIT_NONE)
3735 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3736 	else
3737 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3738 }
3739 
3740 int
3741 if_printf(struct ifnet *ifp, const char * fmt, ...)
3742 {
3743 	va_list ap;
3744 	int retval;
3745 
3746 	retval = printf("%s: ", ifp->if_xname);
3747 	va_start(ap, fmt);
3748 	retval += vprintf(fmt, ap);
3749 	va_end(ap);
3750 	return (retval);
3751 }
3752 
3753 void
3754 if_start(struct ifnet *ifp)
3755 {
3756 
3757 	(*(ifp)->if_start)(ifp);
3758 }
3759 
3760 /*
3761  * Backwards compatibility interface for drivers
3762  * that have not implemented it
3763  */
3764 static int
3765 if_transmit(struct ifnet *ifp, struct mbuf *m)
3766 {
3767 	int error;
3768 
3769 	IFQ_HANDOFF(ifp, m, error);
3770 	return (error);
3771 }
3772 
3773 static void
3774 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3775 {
3776 
3777 	m_freem(m);
3778 }
3779 
3780 int
3781 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3782 {
3783 	int active = 0;
3784 
3785 	IF_LOCK(ifq);
3786 	if (_IF_QFULL(ifq)) {
3787 		IF_UNLOCK(ifq);
3788 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3789 		m_freem(m);
3790 		return (0);
3791 	}
3792 	if (ifp != NULL) {
3793 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3794 		if (m->m_flags & (M_BCAST|M_MCAST))
3795 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3796 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3797 	}
3798 	_IF_ENQUEUE(ifq, m);
3799 	IF_UNLOCK(ifq);
3800 	if (ifp != NULL && !active)
3801 		(*(ifp)->if_start)(ifp);
3802 	return (1);
3803 }
3804 
3805 void
3806 if_register_com_alloc(u_char type,
3807     if_com_alloc_t *a, if_com_free_t *f)
3808 {
3809 
3810 	KASSERT(if_com_alloc[type] == NULL,
3811 	    ("if_register_com_alloc: %d already registered", type));
3812 	KASSERT(if_com_free[type] == NULL,
3813 	    ("if_register_com_alloc: %d free already registered", type));
3814 
3815 	if_com_alloc[type] = a;
3816 	if_com_free[type] = f;
3817 }
3818 
3819 void
3820 if_deregister_com_alloc(u_char type)
3821 {
3822 
3823 	KASSERT(if_com_alloc[type] != NULL,
3824 	    ("if_deregister_com_alloc: %d not registered", type));
3825 	KASSERT(if_com_free[type] != NULL,
3826 	    ("if_deregister_com_alloc: %d free not registered", type));
3827 	if_com_alloc[type] = NULL;
3828 	if_com_free[type] = NULL;
3829 }
3830 
3831 /* API for driver access to network stack owned ifnet.*/
3832 uint64_t
3833 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3834 {
3835 	uint64_t oldbrate;
3836 
3837 	oldbrate = ifp->if_baudrate;
3838 	ifp->if_baudrate = baudrate;
3839 	return (oldbrate);
3840 }
3841 
3842 uint64_t
3843 if_getbaudrate(if_t ifp)
3844 {
3845 
3846 	return (((struct ifnet *)ifp)->if_baudrate);
3847 }
3848 
3849 int
3850 if_setcapabilities(if_t ifp, int capabilities)
3851 {
3852 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3853 	return (0);
3854 }
3855 
3856 int
3857 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3858 {
3859 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3860 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3861 
3862 	return (0);
3863 }
3864 
3865 int
3866 if_getcapabilities(if_t ifp)
3867 {
3868 	return ((struct ifnet *)ifp)->if_capabilities;
3869 }
3870 
3871 int
3872 if_setcapenable(if_t ifp, int capabilities)
3873 {
3874 	((struct ifnet *)ifp)->if_capenable = capabilities;
3875 	return (0);
3876 }
3877 
3878 int
3879 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3880 {
3881 	if(setcap)
3882 		((struct ifnet *)ifp)->if_capenable |= setcap;
3883 	if(clearcap)
3884 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3885 
3886 	return (0);
3887 }
3888 
3889 const char *
3890 if_getdname(if_t ifp)
3891 {
3892 	return ((struct ifnet *)ifp)->if_dname;
3893 }
3894 
3895 int
3896 if_togglecapenable(if_t ifp, int togglecap)
3897 {
3898 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3899 	return (0);
3900 }
3901 
3902 int
3903 if_getcapenable(if_t ifp)
3904 {
3905 	return ((struct ifnet *)ifp)->if_capenable;
3906 }
3907 
3908 /*
3909  * This is largely undesirable because it ties ifnet to a device, but does
3910  * provide flexiblity for an embedded product vendor. Should be used with
3911  * the understanding that it violates the interface boundaries, and should be
3912  * a last resort only.
3913  */
3914 int
3915 if_setdev(if_t ifp, void *dev)
3916 {
3917 	return (0);
3918 }
3919 
3920 int
3921 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3922 {
3923 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3924 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3925 
3926 	return (0);
3927 }
3928 
3929 int
3930 if_getdrvflags(if_t ifp)
3931 {
3932 	return ((struct ifnet *)ifp)->if_drv_flags;
3933 }
3934 
3935 int
3936 if_setdrvflags(if_t ifp, int flags)
3937 {
3938 	((struct ifnet *)ifp)->if_drv_flags = flags;
3939 	return (0);
3940 }
3941 
3942 
3943 int
3944 if_setflags(if_t ifp, int flags)
3945 {
3946 	((struct ifnet *)ifp)->if_flags = flags;
3947 	return (0);
3948 }
3949 
3950 int
3951 if_setflagbits(if_t ifp, int set, int clear)
3952 {
3953 	((struct ifnet *)ifp)->if_flags |= set;
3954 	((struct ifnet *)ifp)->if_flags &= ~clear;
3955 
3956 	return (0);
3957 }
3958 
3959 int
3960 if_getflags(if_t ifp)
3961 {
3962 	return ((struct ifnet *)ifp)->if_flags;
3963 }
3964 
3965 int
3966 if_clearhwassist(if_t ifp)
3967 {
3968 	((struct ifnet *)ifp)->if_hwassist = 0;
3969 	return (0);
3970 }
3971 
3972 int
3973 if_sethwassistbits(if_t ifp, int toset, int toclear)
3974 {
3975 	((struct ifnet *)ifp)->if_hwassist |= toset;
3976 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3977 
3978 	return (0);
3979 }
3980 
3981 int
3982 if_sethwassist(if_t ifp, int hwassist_bit)
3983 {
3984 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3985 	return (0);
3986 }
3987 
3988 int
3989 if_gethwassist(if_t ifp)
3990 {
3991 	return ((struct ifnet *)ifp)->if_hwassist;
3992 }
3993 
3994 int
3995 if_setmtu(if_t ifp, int mtu)
3996 {
3997 	((struct ifnet *)ifp)->if_mtu = mtu;
3998 	return (0);
3999 }
4000 
4001 int
4002 if_getmtu(if_t ifp)
4003 {
4004 	return ((struct ifnet *)ifp)->if_mtu;
4005 }
4006 
4007 int
4008 if_getmtu_family(if_t ifp, int family)
4009 {
4010 	struct domain *dp;
4011 
4012 	for (dp = domains; dp; dp = dp->dom_next) {
4013 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4014 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4015 	}
4016 
4017 	return (((struct ifnet *)ifp)->if_mtu);
4018 }
4019 
4020 int
4021 if_setsoftc(if_t ifp, void *softc)
4022 {
4023 	((struct ifnet *)ifp)->if_softc = softc;
4024 	return (0);
4025 }
4026 
4027 void *
4028 if_getsoftc(if_t ifp)
4029 {
4030 	return ((struct ifnet *)ifp)->if_softc;
4031 }
4032 
4033 void
4034 if_setrcvif(struct mbuf *m, if_t ifp)
4035 {
4036 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4037 }
4038 
4039 void
4040 if_setvtag(struct mbuf *m, uint16_t tag)
4041 {
4042 	m->m_pkthdr.ether_vtag = tag;
4043 }
4044 
4045 uint16_t
4046 if_getvtag(struct mbuf *m)
4047 {
4048 
4049 	return (m->m_pkthdr.ether_vtag);
4050 }
4051 
4052 int
4053 if_sendq_empty(if_t ifp)
4054 {
4055 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4056 }
4057 
4058 struct ifaddr *
4059 if_getifaddr(if_t ifp)
4060 {
4061 	return ((struct ifnet *)ifp)->if_addr;
4062 }
4063 
4064 int
4065 if_getamcount(if_t ifp)
4066 {
4067 	return ((struct ifnet *)ifp)->if_amcount;
4068 }
4069 
4070 
4071 int
4072 if_setsendqready(if_t ifp)
4073 {
4074 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4075 	return (0);
4076 }
4077 
4078 int
4079 if_setsendqlen(if_t ifp, int tx_desc_count)
4080 {
4081 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4082 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4083 
4084 	return (0);
4085 }
4086 
4087 int
4088 if_vlantrunkinuse(if_t ifp)
4089 {
4090 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4091 }
4092 
4093 int
4094 if_input(if_t ifp, struct mbuf* sendmp)
4095 {
4096 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4097 	return (0);
4098 
4099 }
4100 
4101 /* XXX */
4102 #ifndef ETH_ADDR_LEN
4103 #define ETH_ADDR_LEN 6
4104 #endif
4105 
4106 int
4107 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4108 {
4109 	struct ifmultiaddr *ifma;
4110 	uint8_t *lmta = (uint8_t *)mta;
4111 	int mcnt = 0;
4112 
4113 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4114 		if (ifma->ifma_addr->sa_family != AF_LINK)
4115 			continue;
4116 
4117 		if (mcnt == max)
4118 			break;
4119 
4120 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4121 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4122 		mcnt++;
4123 	}
4124 	*cnt = mcnt;
4125 
4126 	return (0);
4127 }
4128 
4129 int
4130 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4131 {
4132 	int error;
4133 
4134 	if_maddr_rlock(ifp);
4135 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4136 	if_maddr_runlock(ifp);
4137 	return (error);
4138 }
4139 
4140 int
4141 if_multiaddr_count(if_t ifp, int max)
4142 {
4143 	struct ifmultiaddr *ifma;
4144 	int count;
4145 
4146 	count = 0;
4147 	if_maddr_rlock(ifp);
4148 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4149 		if (ifma->ifma_addr->sa_family != AF_LINK)
4150 			continue;
4151 		count++;
4152 		if (count == max)
4153 			break;
4154 	}
4155 	if_maddr_runlock(ifp);
4156 	return (count);
4157 }
4158 
4159 int
4160 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4161 {
4162 	struct ifmultiaddr *ifma;
4163 	int cnt = 0;
4164 
4165 	if_maddr_rlock(ifp);
4166 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4167 		cnt += filter(arg, ifma, cnt);
4168 	if_maddr_runlock(ifp);
4169 	return (cnt);
4170 }
4171 
4172 struct mbuf *
4173 if_dequeue(if_t ifp)
4174 {
4175 	struct mbuf *m;
4176 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4177 
4178 	return (m);
4179 }
4180 
4181 int
4182 if_sendq_prepend(if_t ifp, struct mbuf *m)
4183 {
4184 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4185 	return (0);
4186 }
4187 
4188 int
4189 if_setifheaderlen(if_t ifp, int len)
4190 {
4191 	((struct ifnet *)ifp)->if_hdrlen = len;
4192 	return (0);
4193 }
4194 
4195 caddr_t
4196 if_getlladdr(if_t ifp)
4197 {
4198 	return (IF_LLADDR((struct ifnet *)ifp));
4199 }
4200 
4201 void *
4202 if_gethandle(u_char type)
4203 {
4204 	return (if_alloc(type));
4205 }
4206 
4207 void
4208 if_bpfmtap(if_t ifh, struct mbuf *m)
4209 {
4210 	struct ifnet *ifp = (struct ifnet *)ifh;
4211 
4212 	BPF_MTAP(ifp, m);
4213 }
4214 
4215 void
4216 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4217 {
4218 	struct ifnet *ifp = (struct ifnet *)ifh;
4219 
4220 	ETHER_BPF_MTAP(ifp, m);
4221 }
4222 
4223 void
4224 if_vlancap(if_t ifh)
4225 {
4226 	struct ifnet *ifp = (struct ifnet *)ifh;
4227 	VLAN_CAPABILITIES(ifp);
4228 }
4229 
4230 int
4231 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4232 {
4233 
4234 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4235         return (0);
4236 }
4237 
4238 int
4239 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4240 {
4241 
4242 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4243         return (0);
4244 }
4245 
4246 int
4247 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4248 {
4249 
4250 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4251         return (0);
4252 }
4253 
4254 u_int
4255 if_gethwtsomax(if_t ifp)
4256 {
4257 
4258 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4259 }
4260 
4261 u_int
4262 if_gethwtsomaxsegcount(if_t ifp)
4263 {
4264 
4265 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4266 }
4267 
4268 u_int
4269 if_gethwtsomaxsegsize(if_t ifp)
4270 {
4271 
4272 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4273 }
4274 
4275 void
4276 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4277 {
4278 	((struct ifnet *)ifp)->if_init = init_fn;
4279 }
4280 
4281 void
4282 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4283 {
4284 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4285 }
4286 
4287 void
4288 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4289 {
4290 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4291 }
4292 
4293 void
4294 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4295 {
4296 	((struct ifnet *)ifp)->if_transmit = start_fn;
4297 }
4298 
4299 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4300 {
4301 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4302 
4303 }
4304 
4305 void
4306 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4307 {
4308 
4309 	ifp->if_get_counter = fn;
4310 }
4311 
4312 /* Revisit these - These are inline functions originally. */
4313 int
4314 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4315 {
4316 	return drbr_inuse(ifh, br);
4317 }
4318 
4319 struct mbuf*
4320 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4321 {
4322 	return drbr_dequeue(ifh, br);
4323 }
4324 
4325 int
4326 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4327 {
4328 	return drbr_needs_enqueue(ifh, br);
4329 }
4330 
4331 int
4332 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4333 {
4334 	return drbr_enqueue(ifh, br, m);
4335 
4336 }
4337