xref: /freebsd/sys/net/if.c (revision f8b865d1d62d17626ab993212963277c06cc25b8)
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_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/refcount.h>
54 #include <sys/module.h>
55 #include <sys/rwlock.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <sys/priv.h>
64 
65 #include <machine/stdarg.h>
66 #include <vm/uma.h>
67 
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/if_media.h>
77 #include <net/if_vlan_var.h>
78 #include <net/radix.h>
79 #include <net/route.h>
80 #include <net/vnet.h>
81 
82 #if defined(INET) || defined(INET6)
83 #include <net/ethernet.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip.h>
87 #include <netinet/ip_carp.h>
88 #ifdef INET
89 #include <netinet/if_ether.h>
90 #endif /* INET */
91 #ifdef INET6
92 #include <netinet6/in6_var.h>
93 #include <netinet6/in6_ifattach.h>
94 #endif /* INET6 */
95 #endif /* INET || INET6 */
96 
97 #include <security/mac/mac_framework.h>
98 
99 /*
100  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
101  * and ifr_ifru when it is used in SIOCGIFCONF.
102  */
103 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
104     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
105 
106 #ifdef COMPAT_FREEBSD32
107 #include <sys/mount.h>
108 #include <compat/freebsd32/freebsd32.h>
109 
110 struct ifreq_buffer32 {
111 	uint32_t	length;		/* (size_t) */
112 	uint32_t	buffer;		/* (void *) */
113 };
114 
115 /*
116  * Interface request structure used for socket
117  * ioctl's.  All interface ioctl's must have parameter
118  * definitions which begin with ifr_name.  The
119  * remainder may be interface specific.
120  */
121 struct ifreq32 {
122 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
123 	union {
124 		struct sockaddr	ifru_addr;
125 		struct sockaddr	ifru_dstaddr;
126 		struct sockaddr	ifru_broadaddr;
127 		struct ifreq_buffer32 ifru_buffer;
128 		short		ifru_flags[2];
129 		short		ifru_index;
130 		int		ifru_jid;
131 		int		ifru_metric;
132 		int		ifru_mtu;
133 		int		ifru_phys;
134 		int		ifru_media;
135 		uint32_t	ifru_data;
136 		int		ifru_cap[2];
137 		u_int		ifru_fib;
138 		u_char		ifru_vlan_pcp;
139 	} ifr_ifru;
140 };
141 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
142 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
143     __offsetof(struct ifreq32, ifr_ifru));
144 
145 struct ifgroupreq32 {
146 	char	ifgr_name[IFNAMSIZ];
147 	u_int	ifgr_len;
148 	union {
149 		char		ifgru_group[IFNAMSIZ];
150 		uint32_t	ifgru_groups;
151 	} ifgr_ifgru;
152 };
153 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
154     case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
155 #else
156 #define _CASE_IOC_IFGROUPREQ_32(cmd)
157 #endif /* COMPAT_FREEBSD32 */
158 
159 #define CASE_IOC_IFGROUPREQ(cmd)	\
160     _CASE_IOC_IFGROUPREQ_32(cmd)	\
161     case (cmd)
162 
163 union ifreq_union {
164 	struct ifreq	ifr;
165 #ifdef COMPAT_FREEBSD32
166 	struct ifreq32	ifr32;
167 #endif
168 };
169 
170 union ifgroupreq_union {
171 	struct ifgroupreq ifgr;
172 #ifdef COMPAT_FREEBSD32
173 	struct ifgroupreq32 ifgr32;
174 #endif
175 };
176 
177 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
178 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
179 
180 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
181     &ifqmaxlen, 0, "max send queue size");
182 
183 /* Log link state change events */
184 static int log_link_state_change = 1;
185 
186 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
187 	&log_link_state_change, 0,
188 	"log interface link state change events");
189 
190 /* Log promiscuous mode change events */
191 static int log_promisc_mode_change = 1;
192 
193 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
194 	&log_promisc_mode_change, 1,
195 	"log promiscuous mode change events");
196 
197 /* Interface description */
198 static unsigned int ifdescr_maxlen = 1024;
199 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
200 	&ifdescr_maxlen, 0,
201 	"administrative maximum length for interface description");
202 
203 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
204 
205 /* global sx for non-critical path ifdescr */
206 static struct sx ifdescr_sx;
207 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
208 
209 void	(*bridge_linkstate_p)(struct ifnet *ifp);
210 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
211 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
212 /* These are external hooks for CARP. */
213 void	(*carp_linkstate_p)(struct ifnet *ifp);
214 void	(*carp_demote_adj_p)(int, char *);
215 int	(*carp_master_p)(struct ifaddr *);
216 #if defined(INET) || defined(INET6)
217 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
218 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
219     const struct sockaddr *sa);
220 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
221 int	(*carp_attach_p)(struct ifaddr *, int);
222 void	(*carp_detach_p)(struct ifaddr *, bool);
223 #endif
224 #ifdef INET
225 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
226 #endif
227 #ifdef INET6
228 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
229 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
230     const struct in6_addr *taddr);
231 #endif
232 
233 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
234 
235 /*
236  * XXX: Style; these should be sorted alphabetically, and unprototyped
237  * static functions should be prototyped. Currently they are sorted by
238  * declaration order.
239  */
240 static void	if_attachdomain(void *);
241 static void	if_attachdomain1(struct ifnet *);
242 static int	ifconf(u_long, caddr_t);
243 static void	if_freemulti(struct ifmultiaddr *);
244 static void	if_grow(void);
245 static void	if_input_default(struct ifnet *, struct mbuf *);
246 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
247 static void	if_route(struct ifnet *, int flag, int fam);
248 static int	if_setflag(struct ifnet *, int, int, int *, int);
249 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
250 static void	if_unroute(struct ifnet *, int flag, int fam);
251 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
252 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
253 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
254 static void	do_link_state_change(void *, int);
255 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
256 static int	if_getgroupmembers(struct ifgroupreq *);
257 static void	if_delgroups(struct ifnet *);
258 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
259 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
260 #ifdef VIMAGE
261 static void	if_vmove(struct ifnet *, struct vnet *);
262 #endif
263 
264 #ifdef INET6
265 /*
266  * XXX: declare here to avoid to include many inet6 related files..
267  * should be more generalized?
268  */
269 extern void	nd6_setmtu(struct ifnet *);
270 #endif
271 
272 /* ipsec helper hooks */
273 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
274 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
275 
276 VNET_DEFINE(int, if_index);
277 int	ifqmaxlen = IFQ_MAXLEN;
278 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
279 VNET_DEFINE(struct ifgrouphead, ifg_head);
280 
281 static VNET_DEFINE(int, if_indexlim) = 8;
282 
283 /* Table of ifnet by index. */
284 VNET_DEFINE(struct ifnet **, ifindex_table);
285 
286 #define	V_if_indexlim		VNET(if_indexlim)
287 #define	V_ifindex_table		VNET(ifindex_table)
288 
289 /*
290  * The global network interface list (V_ifnet) and related state (such as
291  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
292  * an rwlock.  Either may be acquired shared to stablize the list, but both
293  * must be acquired writable to modify the list.  This model allows us to
294  * both stablize the interface list during interrupt thread processing, but
295  * also to stablize it over long-running ioctls, without introducing priority
296  * inversions and deadlocks.
297  */
298 struct rwlock ifnet_rwlock;
299 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
300 struct sx ifnet_sxlock;
301 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
302 
303 /*
304  * The allocation of network interfaces is a rather non-atomic affair; we
305  * need to select an index before we are ready to expose the interface for
306  * use, so will use this pointer value to indicate reservation.
307  */
308 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
309 
310 static	if_com_alloc_t *if_com_alloc[256];
311 static	if_com_free_t *if_com_free[256];
312 
313 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
314 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
315 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
316 
317 struct ifnet *
318 ifnet_byindex_locked(u_short idx)
319 {
320 
321 	if (idx > V_if_index)
322 		return (NULL);
323 	if (V_ifindex_table[idx] == IFNET_HOLD)
324 		return (NULL);
325 	return (V_ifindex_table[idx]);
326 }
327 
328 struct ifnet *
329 ifnet_byindex(u_short idx)
330 {
331 	struct ifnet *ifp;
332 
333 	IFNET_RLOCK_NOSLEEP();
334 	ifp = ifnet_byindex_locked(idx);
335 	IFNET_RUNLOCK_NOSLEEP();
336 	return (ifp);
337 }
338 
339 struct ifnet *
340 ifnet_byindex_ref(u_short idx)
341 {
342 	struct ifnet *ifp;
343 
344 	IFNET_RLOCK_NOSLEEP();
345 	ifp = ifnet_byindex_locked(idx);
346 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
347 		IFNET_RUNLOCK_NOSLEEP();
348 		return (NULL);
349 	}
350 	if_ref(ifp);
351 	IFNET_RUNLOCK_NOSLEEP();
352 	return (ifp);
353 }
354 
355 /*
356  * Allocate an ifindex array entry; return 0 on success or an error on
357  * failure.
358  */
359 static u_short
360 ifindex_alloc(void)
361 {
362 	u_short idx;
363 
364 	IFNET_WLOCK_ASSERT();
365 retry:
366 	/*
367 	 * Try to find an empty slot below V_if_index.  If we fail, take the
368 	 * next slot.
369 	 */
370 	for (idx = 1; idx <= V_if_index; idx++) {
371 		if (V_ifindex_table[idx] == NULL)
372 			break;
373 	}
374 
375 	/* Catch if_index overflow. */
376 	if (idx >= V_if_indexlim) {
377 		if_grow();
378 		goto retry;
379 	}
380 	if (idx > V_if_index)
381 		V_if_index = idx;
382 	return (idx);
383 }
384 
385 static void
386 ifindex_free_locked(u_short idx)
387 {
388 
389 	IFNET_WLOCK_ASSERT();
390 
391 	V_ifindex_table[idx] = NULL;
392 	while (V_if_index > 0 &&
393 	    V_ifindex_table[V_if_index] == NULL)
394 		V_if_index--;
395 }
396 
397 static void
398 ifindex_free(u_short idx)
399 {
400 
401 	IFNET_WLOCK();
402 	ifindex_free_locked(idx);
403 	IFNET_WUNLOCK();
404 }
405 
406 static void
407 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
408 {
409 
410 	IFNET_WLOCK_ASSERT();
411 
412 	V_ifindex_table[idx] = ifp;
413 }
414 
415 static void
416 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
417 {
418 
419 	IFNET_WLOCK();
420 	ifnet_setbyindex_locked(idx, ifp);
421 	IFNET_WUNLOCK();
422 }
423 
424 struct ifaddr *
425 ifaddr_byindex(u_short idx)
426 {
427 	struct ifnet *ifp;
428 	struct ifaddr *ifa = NULL;
429 
430 	IFNET_RLOCK_NOSLEEP();
431 	ifp = ifnet_byindex_locked(idx);
432 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
433 		ifa_ref(ifa);
434 	IFNET_RUNLOCK_NOSLEEP();
435 	return (ifa);
436 }
437 
438 /*
439  * Network interface utility routines.
440  *
441  * Routines with ifa_ifwith* names take sockaddr *'s as
442  * parameters.
443  */
444 
445 static void
446 vnet_if_init(const void *unused __unused)
447 {
448 
449 	TAILQ_INIT(&V_ifnet);
450 	TAILQ_INIT(&V_ifg_head);
451 	IFNET_WLOCK();
452 	if_grow();				/* create initial table */
453 	IFNET_WUNLOCK();
454 	vnet_if_clone_init();
455 }
456 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
457     NULL);
458 
459 #ifdef VIMAGE
460 static void
461 vnet_if_uninit(const void *unused __unused)
462 {
463 
464 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
465 	    "not empty", __func__, __LINE__, &V_ifnet));
466 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
467 	    "not empty", __func__, __LINE__, &V_ifg_head));
468 
469 	free((caddr_t)V_ifindex_table, M_IFNET);
470 }
471 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
472     vnet_if_uninit, NULL);
473 
474 static void
475 vnet_if_return(const void *unused __unused)
476 {
477 	struct ifnet *ifp, *nifp;
478 
479 	/* Return all inherited interfaces to their parent vnets. */
480 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
481 		if (ifp->if_home_vnet != ifp->if_vnet)
482 			if_vmove(ifp, ifp->if_home_vnet);
483 	}
484 }
485 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
486     vnet_if_return, NULL);
487 #endif
488 
489 static void
490 if_grow(void)
491 {
492 	int oldlim;
493 	u_int n;
494 	struct ifnet **e;
495 
496 	IFNET_WLOCK_ASSERT();
497 	oldlim = V_if_indexlim;
498 	IFNET_WUNLOCK();
499 	n = (oldlim << 1) * sizeof(*e);
500 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
501 	IFNET_WLOCK();
502 	if (V_if_indexlim != oldlim) {
503 		free(e, M_IFNET);
504 		return;
505 	}
506 	if (V_ifindex_table != NULL) {
507 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
508 		free((caddr_t)V_ifindex_table, M_IFNET);
509 	}
510 	V_if_indexlim <<= 1;
511 	V_ifindex_table = e;
512 }
513 
514 /*
515  * Allocate a struct ifnet and an index for an interface.  A layer 2
516  * common structure will also be allocated if an allocation routine is
517  * registered for the passed type.
518  */
519 struct ifnet *
520 if_alloc(u_char type)
521 {
522 	struct ifnet *ifp;
523 	u_short idx;
524 
525 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
526 	IFNET_WLOCK();
527 	idx = ifindex_alloc();
528 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
529 	IFNET_WUNLOCK();
530 	ifp->if_index = idx;
531 	ifp->if_type = type;
532 	ifp->if_alloctype = type;
533 #ifdef VIMAGE
534 	ifp->if_vnet = curvnet;
535 #endif
536 	if (if_com_alloc[type] != NULL) {
537 		ifp->if_l2com = if_com_alloc[type](type, ifp);
538 		if (ifp->if_l2com == NULL) {
539 			free(ifp, M_IFNET);
540 			ifindex_free(idx);
541 			return (NULL);
542 		}
543 	}
544 
545 	IF_ADDR_LOCK_INIT(ifp);
546 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
547 	ifp->if_afdata_initialized = 0;
548 	IF_AFDATA_LOCK_INIT(ifp);
549 	TAILQ_INIT(&ifp->if_addrhead);
550 	TAILQ_INIT(&ifp->if_multiaddrs);
551 	TAILQ_INIT(&ifp->if_groups);
552 #ifdef MAC
553 	mac_ifnet_init(ifp);
554 #endif
555 	ifq_init(&ifp->if_snd, ifp);
556 
557 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
558 	for (int i = 0; i < IFCOUNTERS; i++)
559 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
560 	ifp->if_get_counter = if_get_counter_default;
561 	ifp->if_pcp = IFNET_PCP_NONE;
562 	ifnet_setbyindex(ifp->if_index, ifp);
563 	return (ifp);
564 }
565 
566 /*
567  * Do the actual work of freeing a struct ifnet, and layer 2 common
568  * structure.  This call is made when the last reference to an
569  * interface is released.
570  */
571 static void
572 if_free_internal(struct ifnet *ifp)
573 {
574 
575 	KASSERT((ifp->if_flags & IFF_DYING),
576 	    ("if_free_internal: interface not dying"));
577 
578 	if (if_com_free[ifp->if_alloctype] != NULL)
579 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
580 		    ifp->if_alloctype);
581 
582 #ifdef MAC
583 	mac_ifnet_destroy(ifp);
584 #endif /* MAC */
585 	if (ifp->if_description != NULL)
586 		free(ifp->if_description, M_IFDESCR);
587 	IF_AFDATA_DESTROY(ifp);
588 	IF_ADDR_LOCK_DESTROY(ifp);
589 	ifq_delete(&ifp->if_snd);
590 
591 	for (int i = 0; i < IFCOUNTERS; i++)
592 		counter_u64_free(ifp->if_counters[i]);
593 
594 	free(ifp, M_IFNET);
595 }
596 
597 /*
598  * Deregister an interface and free the associated storage.
599  */
600 void
601 if_free(struct ifnet *ifp)
602 {
603 
604 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
605 
606 	CURVNET_SET_QUIET(ifp->if_vnet);
607 	IFNET_WLOCK();
608 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
609 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
610 
611 	ifindex_free_locked(ifp->if_index);
612 	IFNET_WUNLOCK();
613 
614 	if (refcount_release(&ifp->if_refcount))
615 		if_free_internal(ifp);
616 	CURVNET_RESTORE();
617 }
618 
619 /*
620  * Interfaces to keep an ifnet type-stable despite the possibility of the
621  * driver calling if_free().  If there are additional references, we defer
622  * freeing the underlying data structure.
623  */
624 void
625 if_ref(struct ifnet *ifp)
626 {
627 
628 	/* We don't assert the ifnet list lock here, but arguably should. */
629 	refcount_acquire(&ifp->if_refcount);
630 }
631 
632 void
633 if_rele(struct ifnet *ifp)
634 {
635 
636 	if (!refcount_release(&ifp->if_refcount))
637 		return;
638 	if_free_internal(ifp);
639 }
640 
641 void
642 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
643 {
644 
645 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
646 
647 	if (ifq->ifq_maxlen == 0)
648 		ifq->ifq_maxlen = ifqmaxlen;
649 
650 	ifq->altq_type = 0;
651 	ifq->altq_disc = NULL;
652 	ifq->altq_flags &= ALTQF_CANTCHANGE;
653 	ifq->altq_tbr  = NULL;
654 	ifq->altq_ifp  = ifp;
655 }
656 
657 void
658 ifq_delete(struct ifaltq *ifq)
659 {
660 	mtx_destroy(&ifq->ifq_mtx);
661 }
662 
663 /*
664  * Perform generic interface initialization tasks and attach the interface
665  * to the list of "active" interfaces.  If vmove flag is set on entry
666  * to if_attach_internal(), perform only a limited subset of initialization
667  * tasks, given that we are moving from one vnet to another an ifnet which
668  * has already been fully initialized.
669  *
670  * Note that if_detach_internal() removes group membership unconditionally
671  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
672  * Thus, when if_vmove() is applied to a cloned interface, group membership
673  * is lost while a cloned one always joins a group whose name is
674  * ifc->ifc_name.  To recover this after if_detach_internal() and
675  * if_attach_internal(), the cloner should be specified to
676  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
677  * attempts to join a group whose name is ifc->ifc_name.
678  *
679  * XXX:
680  *  - The decision to return void and thus require this function to
681  *    succeed is questionable.
682  *  - We should probably do more sanity checking.  For instance we don't
683  *    do anything to insure if_xname is unique or non-empty.
684  */
685 void
686 if_attach(struct ifnet *ifp)
687 {
688 
689 	if_attach_internal(ifp, 0, NULL);
690 }
691 
692 /*
693  * Compute the least common TSO limit.
694  */
695 void
696 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
697 {
698 	/*
699 	 * 1) If there is no limit currently, take the limit from
700 	 * the network adapter.
701 	 *
702 	 * 2) If the network adapter has a limit below the current
703 	 * limit, apply it.
704 	 */
705 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
706 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
707 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
708 	}
709 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
710 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
711 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
712 	}
713 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
714 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
715 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
716 	}
717 }
718 
719 /*
720  * Update TSO limit of a network adapter.
721  *
722  * Returns zero if no change. Else non-zero.
723  */
724 int
725 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
726 {
727 	int retval = 0;
728 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
729 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
730 		retval++;
731 	}
732 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
733 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
734 		retval++;
735 	}
736 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
737 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
738 		retval++;
739 	}
740 	return (retval);
741 }
742 
743 static void
744 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
745 {
746 	unsigned socksize, ifasize;
747 	int namelen, masklen;
748 	struct sockaddr_dl *sdl;
749 	struct ifaddr *ifa;
750 
751 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
752 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
753 		    ifp->if_xname);
754 
755 #ifdef VIMAGE
756 	ifp->if_vnet = curvnet;
757 	if (ifp->if_home_vnet == NULL)
758 		ifp->if_home_vnet = curvnet;
759 #endif
760 
761 	if_addgroup(ifp, IFG_ALL);
762 
763 	/* Restore group membership for cloned interfaces. */
764 	if (vmove && ifc != NULL)
765 		if_clone_addgroup(ifp, ifc);
766 
767 	getmicrotime(&ifp->if_lastchange);
768 	ifp->if_epoch = time_uptime;
769 
770 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
771 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
772 	    ("transmit and qflush must both either be set or both be NULL"));
773 	if (ifp->if_transmit == NULL) {
774 		ifp->if_transmit = if_transmit;
775 		ifp->if_qflush = if_qflush;
776 	}
777 	if (ifp->if_input == NULL)
778 		ifp->if_input = if_input_default;
779 
780 	if (ifp->if_requestencap == NULL)
781 		ifp->if_requestencap = if_requestencap_default;
782 
783 	if (!vmove) {
784 #ifdef MAC
785 		mac_ifnet_create(ifp);
786 #endif
787 
788 		/*
789 		 * Create a Link Level name for this device.
790 		 */
791 		namelen = strlen(ifp->if_xname);
792 		/*
793 		 * Always save enough space for any possiable name so we
794 		 * can do a rename in place later.
795 		 */
796 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
797 		socksize = masklen + ifp->if_addrlen;
798 		if (socksize < sizeof(*sdl))
799 			socksize = sizeof(*sdl);
800 		socksize = roundup2(socksize, sizeof(long));
801 		ifasize = sizeof(*ifa) + 2 * socksize;
802 		ifa = ifa_alloc(ifasize, M_WAITOK);
803 		sdl = (struct sockaddr_dl *)(ifa + 1);
804 		sdl->sdl_len = socksize;
805 		sdl->sdl_family = AF_LINK;
806 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
807 		sdl->sdl_nlen = namelen;
808 		sdl->sdl_index = ifp->if_index;
809 		sdl->sdl_type = ifp->if_type;
810 		ifp->if_addr = ifa;
811 		ifa->ifa_ifp = ifp;
812 		ifa->ifa_rtrequest = link_rtrequest;
813 		ifa->ifa_addr = (struct sockaddr *)sdl;
814 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
815 		ifa->ifa_netmask = (struct sockaddr *)sdl;
816 		sdl->sdl_len = masklen;
817 		while (namelen != 0)
818 			sdl->sdl_data[--namelen] = 0xff;
819 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
820 		/* Reliably crash if used uninitialized. */
821 		ifp->if_broadcastaddr = NULL;
822 
823 		if (ifp->if_type == IFT_ETHER) {
824 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
825 			    M_WAITOK | M_ZERO);
826 		}
827 
828 #if defined(INET) || defined(INET6)
829 		/* Use defaults for TSO, if nothing is set */
830 		if (ifp->if_hw_tsomax == 0 &&
831 		    ifp->if_hw_tsomaxsegcount == 0 &&
832 		    ifp->if_hw_tsomaxsegsize == 0) {
833 			/*
834 			 * The TSO defaults needs to be such that an
835 			 * NFS mbuf list of 35 mbufs totalling just
836 			 * below 64K works and that a chain of mbufs
837 			 * can be defragged into at most 32 segments:
838 			 */
839 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
840 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
841 			ifp->if_hw_tsomaxsegcount = 35;
842 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
843 
844 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
845 			if (ifp->if_capabilities & IFCAP_TSO) {
846 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
847 				    ifp->if_hw_tsomax,
848 				    ifp->if_hw_tsomaxsegcount,
849 				    ifp->if_hw_tsomaxsegsize);
850 			}
851 		}
852 #endif
853 	}
854 #ifdef VIMAGE
855 	else {
856 		/*
857 		 * Update the interface index in the link layer address
858 		 * of the interface.
859 		 */
860 		for (ifa = ifp->if_addr; ifa != NULL;
861 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
862 			if (ifa->ifa_addr->sa_family == AF_LINK) {
863 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
864 				sdl->sdl_index = ifp->if_index;
865 			}
866 		}
867 	}
868 #endif
869 
870 	IFNET_WLOCK();
871 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
872 #ifdef VIMAGE
873 	curvnet->vnet_ifcnt++;
874 #endif
875 	IFNET_WUNLOCK();
876 
877 	if (domain_init_status >= 2)
878 		if_attachdomain1(ifp);
879 
880 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
881 	if (IS_DEFAULT_VNET(curvnet))
882 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
883 
884 	/* Announce the interface. */
885 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
886 }
887 
888 static void
889 if_attachdomain(void *dummy)
890 {
891 	struct ifnet *ifp;
892 
893 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
894 		if_attachdomain1(ifp);
895 }
896 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
897     if_attachdomain, NULL);
898 
899 static void
900 if_attachdomain1(struct ifnet *ifp)
901 {
902 	struct domain *dp;
903 
904 	/*
905 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
906 	 * cannot lock ifp->if_afdata initialization, entirely.
907 	 */
908 	IF_AFDATA_LOCK(ifp);
909 	if (ifp->if_afdata_initialized >= domain_init_status) {
910 		IF_AFDATA_UNLOCK(ifp);
911 		log(LOG_WARNING, "%s called more than once on %s\n",
912 		    __func__, ifp->if_xname);
913 		return;
914 	}
915 	ifp->if_afdata_initialized = domain_init_status;
916 	IF_AFDATA_UNLOCK(ifp);
917 
918 	/* address family dependent data region */
919 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
920 	for (dp = domains; dp; dp = dp->dom_next) {
921 		if (dp->dom_ifattach)
922 			ifp->if_afdata[dp->dom_family] =
923 			    (*dp->dom_ifattach)(ifp);
924 	}
925 }
926 
927 /*
928  * Remove any unicast or broadcast network addresses from an interface.
929  */
930 void
931 if_purgeaddrs(struct ifnet *ifp)
932 {
933 	struct ifaddr *ifa, *next;
934 
935 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
936 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
937 		if (ifa->ifa_addr->sa_family == AF_LINK)
938 			continue;
939 #ifdef INET
940 		/* XXX: Ugly!! ad hoc just for INET */
941 		if (ifa->ifa_addr->sa_family == AF_INET) {
942 			struct ifaliasreq ifr;
943 
944 			bzero(&ifr, sizeof(ifr));
945 			ifr.ifra_addr = *ifa->ifa_addr;
946 			if (ifa->ifa_dstaddr)
947 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
948 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
949 			    NULL) == 0)
950 				continue;
951 		}
952 #endif /* INET */
953 #ifdef INET6
954 		if (ifa->ifa_addr->sa_family == AF_INET6) {
955 			in6_purgeaddr(ifa);
956 			/* ifp_addrhead is already updated */
957 			continue;
958 		}
959 #endif /* INET6 */
960 		IF_ADDR_WLOCK(ifp);
961 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
962 		IF_ADDR_WUNLOCK(ifp);
963 		ifa_free(ifa);
964 	}
965 }
966 
967 /*
968  * Remove any multicast network addresses from an interface when an ifnet
969  * is going away.
970  */
971 static void
972 if_purgemaddrs(struct ifnet *ifp)
973 {
974 	struct ifmultiaddr *ifma;
975 	struct ifmultiaddr *next;
976 
977 	IF_ADDR_WLOCK(ifp);
978 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
979 		if_delmulti_locked(ifp, ifma, 1);
980 	IF_ADDR_WUNLOCK(ifp);
981 }
982 
983 /*
984  * Detach an interface, removing it from the list of "active" interfaces.
985  * If vmove flag is set on entry to if_detach_internal(), perform only a
986  * limited subset of cleanup tasks, given that we are moving an ifnet from
987  * one vnet to another, where it must be fully operational.
988  *
989  * XXXRW: There are some significant questions about event ordering, and
990  * how to prevent things from starting to use the interface during detach.
991  */
992 void
993 if_detach(struct ifnet *ifp)
994 {
995 
996 	CURVNET_SET_QUIET(ifp->if_vnet);
997 	if_detach_internal(ifp, 0, NULL);
998 	CURVNET_RESTORE();
999 }
1000 
1001 /*
1002  * The vmove flag, if set, indicates that we are called from a callpath
1003  * that is moving an interface to a different vnet instance.
1004  *
1005  * The shutdown flag, if set, indicates that we are called in the
1006  * process of shutting down a vnet instance.  Currently only the
1007  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1008  * on a vnet instance shutdown without this flag being set, e.g., when
1009  * the cloned interfaces are destoyed as first thing of teardown.
1010  */
1011 static int
1012 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1013 {
1014 	struct ifaddr *ifa;
1015 	int i;
1016 	struct domain *dp;
1017  	struct ifnet *iter;
1018  	int found = 0;
1019 #ifdef VIMAGE
1020 	int shutdown;
1021 
1022 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1023 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1024 #endif
1025 	IFNET_WLOCK();
1026 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
1027 		if (iter == ifp) {
1028 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
1029 			found = 1;
1030 			break;
1031 		}
1032 	IFNET_WUNLOCK();
1033 	if (!found) {
1034 		/*
1035 		 * While we would want to panic here, we cannot
1036 		 * guarantee that the interface is indeed still on
1037 		 * the list given we don't hold locks all the way.
1038 		 */
1039 		return (ENOENT);
1040 #if 0
1041 		if (vmove)
1042 			panic("%s: ifp=%p not on the ifnet tailq %p",
1043 			    __func__, ifp, &V_ifnet);
1044 		else
1045 			return; /* XXX this should panic as well? */
1046 #endif
1047 	}
1048 
1049 	/*
1050 	 * At this point we know the interface still was on the ifnet list
1051 	 * and we removed it so we are in a stable state.
1052 	 */
1053 #ifdef VIMAGE
1054 	curvnet->vnet_ifcnt--;
1055 #endif
1056 
1057 	/*
1058 	 * In any case (destroy or vmove) detach us from the groups
1059 	 * and remove/wait for pending events on the taskq.
1060 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1061 	 */
1062 	if_delgroups(ifp);
1063 
1064 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1065 
1066 	/*
1067 	 * Check if this is a cloned interface or not. Must do even if
1068 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1069 	 * the if_clone_addgroup() will have a corrupted string overwise
1070 	 * from a gibberish pointer.
1071 	 */
1072 	if (vmove && ifcp != NULL)
1073 		*ifcp = if_clone_findifc(ifp);
1074 
1075 	if_down(ifp);
1076 
1077 #ifdef VIMAGE
1078 	/*
1079 	 * On VNET shutdown abort here as the stack teardown will do all
1080 	 * the work top-down for us.
1081 	 */
1082 	if (shutdown) {
1083 		/*
1084 		 * In case of a vmove we are done here without error.
1085 		 * If we would signal an error it would lead to the same
1086 		 * abort as if we did not find the ifnet anymore.
1087 		 * if_detach() calls us in void context and does not care
1088 		 * about an early abort notification, so life is splendid :)
1089 		 */
1090 		goto finish_vnet_shutdown;
1091 	}
1092 #endif
1093 
1094 	/*
1095 	 * At this point we are not tearing down a VNET and are either
1096 	 * going to destroy or vmove the interface and have to cleanup
1097 	 * accordingly.
1098 	 */
1099 
1100 	/*
1101 	 * Remove routes and flush queues.
1102 	 */
1103 #ifdef ALTQ
1104 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1105 		altq_disable(&ifp->if_snd);
1106 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1107 		altq_detach(&ifp->if_snd);
1108 #endif
1109 
1110 	if_purgeaddrs(ifp);
1111 
1112 #ifdef INET
1113 	in_ifdetach(ifp);
1114 #endif
1115 
1116 #ifdef INET6
1117 	/*
1118 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1119 	 * before removing routing entries below, since IPv6 interface direct
1120 	 * routes are expected to be removed by the IPv6-specific kernel API.
1121 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1122 	 */
1123 	in6_ifdetach(ifp);
1124 #endif
1125 	if_purgemaddrs(ifp);
1126 
1127 	/* Announce that the interface is gone. */
1128 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1129 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1130 	if (IS_DEFAULT_VNET(curvnet))
1131 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1132 
1133 	if (!vmove) {
1134 		/*
1135 		 * Prevent further calls into the device driver via ifnet.
1136 		 */
1137 		if_dead(ifp);
1138 
1139 		/*
1140 		 * Remove link ifaddr pointer and maybe decrement if_index.
1141 		 * Clean up all addresses.
1142 		 */
1143 		free(ifp->if_hw_addr, M_IFADDR);
1144 		ifp->if_hw_addr = NULL;
1145 		ifp->if_addr = NULL;
1146 
1147 		/* We can now free link ifaddr. */
1148 		IF_ADDR_WLOCK(ifp);
1149 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1150 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1151 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1152 			IF_ADDR_WUNLOCK(ifp);
1153 			ifa_free(ifa);
1154 		} else
1155 			IF_ADDR_WUNLOCK(ifp);
1156 	}
1157 
1158 	rt_flushifroutes(ifp);
1159 
1160 #ifdef VIMAGE
1161 finish_vnet_shutdown:
1162 #endif
1163 	/*
1164 	 * We cannot hold the lock over dom_ifdetach calls as they might
1165 	 * sleep, for example trying to drain a callout, thus open up the
1166 	 * theoretical race with re-attaching.
1167 	 */
1168 	IF_AFDATA_LOCK(ifp);
1169 	i = ifp->if_afdata_initialized;
1170 	ifp->if_afdata_initialized = 0;
1171 	IF_AFDATA_UNLOCK(ifp);
1172 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1173 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1174 			(*dp->dom_ifdetach)(ifp,
1175 			    ifp->if_afdata[dp->dom_family]);
1176 			ifp->if_afdata[dp->dom_family] = NULL;
1177 		}
1178 	}
1179 
1180 	return (0);
1181 }
1182 
1183 #ifdef VIMAGE
1184 /*
1185  * if_vmove() performs a limited version of if_detach() in current
1186  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1187  * An attempt is made to shrink if_index in current vnet, find an
1188  * unused if_index in target vnet and calls if_grow() if necessary,
1189  * and finally find an unused if_xname for the target vnet.
1190  */
1191 static void
1192 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1193 {
1194 	struct if_clone *ifc;
1195 	u_int bif_dlt, bif_hdrlen;
1196 	int rc;
1197 
1198  	/*
1199 	 * if_detach_internal() will call the eventhandler to notify
1200 	 * interface departure.  That will detach if_bpf.  We need to
1201 	 * safe the dlt and hdrlen so we can re-attach it later.
1202 	 */
1203 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1204 
1205 	/*
1206 	 * Detach from current vnet, but preserve LLADDR info, do not
1207 	 * mark as dead etc. so that the ifnet can be reattached later.
1208 	 * If we cannot find it, we lost the race to someone else.
1209 	 */
1210 	rc = if_detach_internal(ifp, 1, &ifc);
1211 	if (rc != 0)
1212 		return;
1213 
1214 	/*
1215 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1216 	 * the if_index for that vnet if possible.
1217 	 *
1218 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1219 	 * or we'd lock on one vnet and unlock on another.
1220 	 */
1221 	IFNET_WLOCK();
1222 	ifindex_free_locked(ifp->if_index);
1223 	IFNET_WUNLOCK();
1224 
1225 	/*
1226 	 * Perform interface-specific reassignment tasks, if provided by
1227 	 * the driver.
1228 	 */
1229 	if (ifp->if_reassign != NULL)
1230 		ifp->if_reassign(ifp, new_vnet, NULL);
1231 
1232 	/*
1233 	 * Switch to the context of the target vnet.
1234 	 */
1235 	CURVNET_SET_QUIET(new_vnet);
1236 
1237 	IFNET_WLOCK();
1238 	ifp->if_index = ifindex_alloc();
1239 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1240 	IFNET_WUNLOCK();
1241 
1242 	if_attach_internal(ifp, 1, ifc);
1243 
1244 	if (ifp->if_bpf == NULL)
1245 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1246 
1247 	CURVNET_RESTORE();
1248 }
1249 
1250 /*
1251  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1252  */
1253 static int
1254 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1255 {
1256 	struct prison *pr;
1257 	struct ifnet *difp;
1258 	int shutdown;
1259 
1260 	/* Try to find the prison within our visibility. */
1261 	sx_slock(&allprison_lock);
1262 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1263 	sx_sunlock(&allprison_lock);
1264 	if (pr == NULL)
1265 		return (ENXIO);
1266 	prison_hold_locked(pr);
1267 	mtx_unlock(&pr->pr_mtx);
1268 
1269 	/* Do not try to move the iface from and to the same prison. */
1270 	if (pr->pr_vnet == ifp->if_vnet) {
1271 		prison_free(pr);
1272 		return (EEXIST);
1273 	}
1274 
1275 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1276 	/* XXX Lock interfaces to avoid races. */
1277 	CURVNET_SET_QUIET(pr->pr_vnet);
1278 	difp = ifunit(ifname);
1279 	if (difp != NULL) {
1280 		CURVNET_RESTORE();
1281 		prison_free(pr);
1282 		return (EEXIST);
1283 	}
1284 
1285 	/* Make sure the VNET is stable. */
1286 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1287 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1288 	if (shutdown) {
1289 		CURVNET_RESTORE();
1290 		prison_free(pr);
1291 		return (EBUSY);
1292 	}
1293 	CURVNET_RESTORE();
1294 
1295 	/* Move the interface into the child jail/vnet. */
1296 	if_vmove(ifp, pr->pr_vnet);
1297 
1298 	/* Report the new if_xname back to the userland. */
1299 	sprintf(ifname, "%s", ifp->if_xname);
1300 
1301 	prison_free(pr);
1302 	return (0);
1303 }
1304 
1305 static int
1306 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1307 {
1308 	struct prison *pr;
1309 	struct vnet *vnet_dst;
1310 	struct ifnet *ifp;
1311  	int shutdown;
1312 
1313 	/* Try to find the prison within our visibility. */
1314 	sx_slock(&allprison_lock);
1315 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1316 	sx_sunlock(&allprison_lock);
1317 	if (pr == NULL)
1318 		return (ENXIO);
1319 	prison_hold_locked(pr);
1320 	mtx_unlock(&pr->pr_mtx);
1321 
1322 	/* Make sure the named iface exists in the source prison/vnet. */
1323 	CURVNET_SET(pr->pr_vnet);
1324 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1325 	if (ifp == NULL) {
1326 		CURVNET_RESTORE();
1327 		prison_free(pr);
1328 		return (ENXIO);
1329 	}
1330 
1331 	/* Do not try to move the iface from and to the same prison. */
1332 	vnet_dst = TD_TO_VNET(td);
1333 	if (vnet_dst == ifp->if_vnet) {
1334 		CURVNET_RESTORE();
1335 		prison_free(pr);
1336 		return (EEXIST);
1337 	}
1338 
1339 	/* Make sure the VNET is stable. */
1340 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1341 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1342 	if (shutdown) {
1343 		CURVNET_RESTORE();
1344 		prison_free(pr);
1345 		return (EBUSY);
1346 	}
1347 
1348 	/* Get interface back from child jail/vnet. */
1349 	if_vmove(ifp, vnet_dst);
1350 	CURVNET_RESTORE();
1351 
1352 	/* Report the new if_xname back to the userland. */
1353 	sprintf(ifname, "%s", ifp->if_xname);
1354 
1355 	prison_free(pr);
1356 	return (0);
1357 }
1358 #endif /* VIMAGE */
1359 
1360 /*
1361  * Add a group to an interface
1362  */
1363 int
1364 if_addgroup(struct ifnet *ifp, const char *groupname)
1365 {
1366 	struct ifg_list		*ifgl;
1367 	struct ifg_group	*ifg = NULL;
1368 	struct ifg_member	*ifgm;
1369 	int 			 new = 0;
1370 
1371 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1372 	    groupname[strlen(groupname) - 1] <= '9')
1373 		return (EINVAL);
1374 
1375 	IFNET_WLOCK();
1376 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1377 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1378 			IFNET_WUNLOCK();
1379 			return (EEXIST);
1380 		}
1381 
1382 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1383 	    M_NOWAIT)) == NULL) {
1384 	    	IFNET_WUNLOCK();
1385 		return (ENOMEM);
1386 	}
1387 
1388 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1389 	    M_TEMP, M_NOWAIT)) == NULL) {
1390 		free(ifgl, M_TEMP);
1391 		IFNET_WUNLOCK();
1392 		return (ENOMEM);
1393 	}
1394 
1395 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1396 		if (!strcmp(ifg->ifg_group, groupname))
1397 			break;
1398 
1399 	if (ifg == NULL) {
1400 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1401 		    M_TEMP, M_NOWAIT)) == NULL) {
1402 			free(ifgl, M_TEMP);
1403 			free(ifgm, M_TEMP);
1404 			IFNET_WUNLOCK();
1405 			return (ENOMEM);
1406 		}
1407 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1408 		ifg->ifg_refcnt = 0;
1409 		TAILQ_INIT(&ifg->ifg_members);
1410 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1411 		new = 1;
1412 	}
1413 
1414 	ifg->ifg_refcnt++;
1415 	ifgl->ifgl_group = ifg;
1416 	ifgm->ifgm_ifp = ifp;
1417 
1418 	IF_ADDR_WLOCK(ifp);
1419 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1420 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1421 	IF_ADDR_WUNLOCK(ifp);
1422 
1423 	IFNET_WUNLOCK();
1424 
1425 	if (new)
1426 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1427 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1428 
1429 	return (0);
1430 }
1431 
1432 /*
1433  * Remove a group from an interface
1434  */
1435 int
1436 if_delgroup(struct ifnet *ifp, const char *groupname)
1437 {
1438 	struct ifg_list		*ifgl;
1439 	struct ifg_member	*ifgm;
1440 
1441 	IFNET_WLOCK();
1442 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1443 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1444 			break;
1445 	if (ifgl == NULL) {
1446 		IFNET_WUNLOCK();
1447 		return (ENOENT);
1448 	}
1449 
1450 	IF_ADDR_WLOCK(ifp);
1451 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1452 	IF_ADDR_WUNLOCK(ifp);
1453 
1454 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1455 		if (ifgm->ifgm_ifp == ifp)
1456 			break;
1457 
1458 	if (ifgm != NULL) {
1459 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1460 		free(ifgm, M_TEMP);
1461 	}
1462 
1463 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1464 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1465 		IFNET_WUNLOCK();
1466 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1467 		free(ifgl->ifgl_group, M_TEMP);
1468 	} else
1469 		IFNET_WUNLOCK();
1470 
1471 	free(ifgl, M_TEMP);
1472 
1473 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1474 
1475 	return (0);
1476 }
1477 
1478 /*
1479  * Remove an interface from all groups
1480  */
1481 static void
1482 if_delgroups(struct ifnet *ifp)
1483 {
1484 	struct ifg_list		*ifgl;
1485 	struct ifg_member	*ifgm;
1486 	char groupname[IFNAMSIZ];
1487 
1488 	IFNET_WLOCK();
1489 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1490 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1491 
1492 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1493 
1494 		IF_ADDR_WLOCK(ifp);
1495 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1496 		IF_ADDR_WUNLOCK(ifp);
1497 
1498 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1499 			if (ifgm->ifgm_ifp == ifp)
1500 				break;
1501 
1502 		if (ifgm != NULL) {
1503 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1504 			    ifgm_next);
1505 			free(ifgm, M_TEMP);
1506 		}
1507 
1508 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1509 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1510 			IFNET_WUNLOCK();
1511 			EVENTHANDLER_INVOKE(group_detach_event,
1512 			    ifgl->ifgl_group);
1513 			free(ifgl->ifgl_group, M_TEMP);
1514 		} else
1515 			IFNET_WUNLOCK();
1516 
1517 		free(ifgl, M_TEMP);
1518 
1519 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1520 
1521 		IFNET_WLOCK();
1522 	}
1523 	IFNET_WUNLOCK();
1524 }
1525 
1526 static char *
1527 ifgr_group_get(void *ifgrp)
1528 {
1529 	union ifgroupreq_union *ifgrup;
1530 
1531 	ifgrup = ifgrp;
1532 #ifdef COMPAT_FREEBSD32
1533 	if (SV_CURPROC_FLAG(SV_ILP32))
1534 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1535 #endif
1536 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1537 }
1538 
1539 static struct ifg_req *
1540 ifgr_groups_get(void *ifgrp)
1541 {
1542 	union ifgroupreq_union *ifgrup;
1543 
1544 	ifgrup = ifgrp;
1545 #ifdef COMPAT_FREEBSD32
1546 	if (SV_CURPROC_FLAG(SV_ILP32))
1547 		return ((struct ifg_req *)(uintptr_t)
1548 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1549 #endif
1550 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1551 }
1552 
1553 /*
1554  * Stores all groups from an interface in memory pointed to by ifgr.
1555  */
1556 static int
1557 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1558 {
1559 	int			 len, error;
1560 	struct ifg_list		*ifgl;
1561 	struct ifg_req		 ifgrq, *ifgp;
1562 
1563 	if (ifgr->ifgr_len == 0) {
1564 		IF_ADDR_RLOCK(ifp);
1565 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1566 			ifgr->ifgr_len += sizeof(struct ifg_req);
1567 		IF_ADDR_RUNLOCK(ifp);
1568 		return (0);
1569 	}
1570 
1571 	len = ifgr->ifgr_len;
1572 	ifgp = ifgr_groups_get(ifgr);
1573 	/* XXX: wire */
1574 	IF_ADDR_RLOCK(ifp);
1575 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1576 		if (len < sizeof(ifgrq)) {
1577 			IF_ADDR_RUNLOCK(ifp);
1578 			return (EINVAL);
1579 		}
1580 		bzero(&ifgrq, sizeof ifgrq);
1581 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1582 		    sizeof(ifgrq.ifgrq_group));
1583 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1584 		    	IF_ADDR_RUNLOCK(ifp);
1585 			return (error);
1586 		}
1587 		len -= sizeof(ifgrq);
1588 		ifgp++;
1589 	}
1590 	IF_ADDR_RUNLOCK(ifp);
1591 
1592 	return (0);
1593 }
1594 
1595 /*
1596  * Stores all members of a group in memory pointed to by igfr
1597  */
1598 static int
1599 if_getgroupmembers(struct ifgroupreq *ifgr)
1600 {
1601 	struct ifg_group	*ifg;
1602 	struct ifg_member	*ifgm;
1603 	struct ifg_req		 ifgrq, *ifgp;
1604 	int			 len, error;
1605 
1606 	IFNET_RLOCK();
1607 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1608 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1609 			break;
1610 	if (ifg == NULL) {
1611 		IFNET_RUNLOCK();
1612 		return (ENOENT);
1613 	}
1614 
1615 	if (ifgr->ifgr_len == 0) {
1616 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1617 			ifgr->ifgr_len += sizeof(ifgrq);
1618 		IFNET_RUNLOCK();
1619 		return (0);
1620 	}
1621 
1622 	len = ifgr->ifgr_len;
1623 	ifgp = ifgr_groups_get(ifgr);
1624 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1625 		if (len < sizeof(ifgrq)) {
1626 			IFNET_RUNLOCK();
1627 			return (EINVAL);
1628 		}
1629 		bzero(&ifgrq, sizeof ifgrq);
1630 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1631 		    sizeof(ifgrq.ifgrq_member));
1632 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1633 			IFNET_RUNLOCK();
1634 			return (error);
1635 		}
1636 		len -= sizeof(ifgrq);
1637 		ifgp++;
1638 	}
1639 	IFNET_RUNLOCK();
1640 
1641 	return (0);
1642 }
1643 
1644 /*
1645  * Return counter values from counter(9)s stored in ifnet.
1646  */
1647 uint64_t
1648 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1649 {
1650 
1651 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1652 
1653 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1654 }
1655 
1656 /*
1657  * Increase an ifnet counter. Usually used for counters shared
1658  * between the stack and a driver, but function supports them all.
1659  */
1660 void
1661 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1662 {
1663 
1664 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1665 
1666 	counter_u64_add(ifp->if_counters[cnt], inc);
1667 }
1668 
1669 /*
1670  * Copy data from ifnet to userland API structure if_data.
1671  */
1672 void
1673 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1674 {
1675 
1676 	ifd->ifi_type = ifp->if_type;
1677 	ifd->ifi_physical = 0;
1678 	ifd->ifi_addrlen = ifp->if_addrlen;
1679 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1680 	ifd->ifi_link_state = ifp->if_link_state;
1681 	ifd->ifi_vhid = 0;
1682 	ifd->ifi_datalen = sizeof(struct if_data);
1683 	ifd->ifi_mtu = ifp->if_mtu;
1684 	ifd->ifi_metric = ifp->if_metric;
1685 	ifd->ifi_baudrate = ifp->if_baudrate;
1686 	ifd->ifi_hwassist = ifp->if_hwassist;
1687 	ifd->ifi_epoch = ifp->if_epoch;
1688 	ifd->ifi_lastchange = ifp->if_lastchange;
1689 
1690 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1691 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1692 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1693 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1694 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1695 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1696 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1697 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1698 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1699 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1700 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1701 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1702 }
1703 
1704 /*
1705  * Wrapper functions for struct ifnet address list locking macros.  These are
1706  * used by kernel modules to avoid encoding programming interface or binary
1707  * interface assumptions that may be violated when kernel-internal locking
1708  * approaches change.
1709  */
1710 void
1711 if_addr_rlock(struct ifnet *ifp)
1712 {
1713 
1714 	IF_ADDR_RLOCK(ifp);
1715 }
1716 
1717 void
1718 if_addr_runlock(struct ifnet *ifp)
1719 {
1720 
1721 	IF_ADDR_RUNLOCK(ifp);
1722 }
1723 
1724 void
1725 if_maddr_rlock(if_t ifp)
1726 {
1727 
1728 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1729 }
1730 
1731 void
1732 if_maddr_runlock(if_t ifp)
1733 {
1734 
1735 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1736 }
1737 
1738 /*
1739  * Initialization, destruction and refcounting functions for ifaddrs.
1740  */
1741 struct ifaddr *
1742 ifa_alloc(size_t size, int flags)
1743 {
1744 	struct ifaddr *ifa;
1745 
1746 	KASSERT(size >= sizeof(struct ifaddr),
1747 	    ("%s: invalid size %zu", __func__, size));
1748 
1749 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1750 	if (ifa == NULL)
1751 		return (NULL);
1752 
1753 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1754 		goto fail;
1755 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1756 		goto fail;
1757 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1758 		goto fail;
1759 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1760 		goto fail;
1761 
1762 	refcount_init(&ifa->ifa_refcnt, 1);
1763 
1764 	return (ifa);
1765 
1766 fail:
1767 	/* free(NULL) is okay */
1768 	counter_u64_free(ifa->ifa_opackets);
1769 	counter_u64_free(ifa->ifa_ipackets);
1770 	counter_u64_free(ifa->ifa_obytes);
1771 	counter_u64_free(ifa->ifa_ibytes);
1772 	free(ifa, M_IFADDR);
1773 
1774 	return (NULL);
1775 }
1776 
1777 void
1778 ifa_ref(struct ifaddr *ifa)
1779 {
1780 
1781 	refcount_acquire(&ifa->ifa_refcnt);
1782 }
1783 
1784 void
1785 ifa_free(struct ifaddr *ifa)
1786 {
1787 
1788 	if (refcount_release(&ifa->ifa_refcnt)) {
1789 		counter_u64_free(ifa->ifa_opackets);
1790 		counter_u64_free(ifa->ifa_ipackets);
1791 		counter_u64_free(ifa->ifa_obytes);
1792 		counter_u64_free(ifa->ifa_ibytes);
1793 		free(ifa, M_IFADDR);
1794 	}
1795 }
1796 
1797 static int
1798 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1799     struct sockaddr *ia)
1800 {
1801 	int error;
1802 	struct rt_addrinfo info;
1803 	struct sockaddr_dl null_sdl;
1804 	struct ifnet *ifp;
1805 
1806 	ifp = ifa->ifa_ifp;
1807 
1808 	bzero(&info, sizeof(info));
1809 	if (cmd != RTM_DELETE)
1810 		info.rti_ifp = V_loif;
1811 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1812 	info.rti_info[RTAX_DST] = ia;
1813 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1814 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1815 
1816 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1817 
1818 	if (error != 0)
1819 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1820 		    __func__, otype, if_name(ifp), error);
1821 
1822 	return (error);
1823 }
1824 
1825 int
1826 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1827 {
1828 
1829 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1830 }
1831 
1832 int
1833 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1834 {
1835 
1836 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1837 }
1838 
1839 int
1840 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1841 {
1842 
1843 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1844 }
1845 
1846 /*
1847  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1848  * structs used to represent other address families, it is necessary
1849  * to perform a different comparison.
1850  */
1851 
1852 #define	sa_dl_equal(a1, a2)	\
1853 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1854 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1855 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1856 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1857 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1858 
1859 /*
1860  * Locate an interface based on a complete address.
1861  */
1862 /*ARGSUSED*/
1863 static struct ifaddr *
1864 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1865 {
1866 	struct ifnet *ifp;
1867 	struct ifaddr *ifa;
1868 
1869 	IFNET_RLOCK_NOSLEEP();
1870 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1871 		IF_ADDR_RLOCK(ifp);
1872 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1873 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1874 				continue;
1875 			if (sa_equal(addr, ifa->ifa_addr)) {
1876 				if (getref)
1877 					ifa_ref(ifa);
1878 				IF_ADDR_RUNLOCK(ifp);
1879 				goto done;
1880 			}
1881 			/* IP6 doesn't have broadcast */
1882 			if ((ifp->if_flags & IFF_BROADCAST) &&
1883 			    ifa->ifa_broadaddr &&
1884 			    ifa->ifa_broadaddr->sa_len != 0 &&
1885 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1886 				if (getref)
1887 					ifa_ref(ifa);
1888 				IF_ADDR_RUNLOCK(ifp);
1889 				goto done;
1890 			}
1891 		}
1892 		IF_ADDR_RUNLOCK(ifp);
1893 	}
1894 	ifa = NULL;
1895 done:
1896 	IFNET_RUNLOCK_NOSLEEP();
1897 	return (ifa);
1898 }
1899 
1900 struct ifaddr *
1901 ifa_ifwithaddr(const struct sockaddr *addr)
1902 {
1903 
1904 	return (ifa_ifwithaddr_internal(addr, 1));
1905 }
1906 
1907 int
1908 ifa_ifwithaddr_check(const struct sockaddr *addr)
1909 {
1910 
1911 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1912 }
1913 
1914 /*
1915  * Locate an interface based on the broadcast address.
1916  */
1917 /* ARGSUSED */
1918 struct ifaddr *
1919 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1920 {
1921 	struct ifnet *ifp;
1922 	struct ifaddr *ifa;
1923 
1924 	IFNET_RLOCK_NOSLEEP();
1925 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1926 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1927 			continue;
1928 		IF_ADDR_RLOCK(ifp);
1929 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1930 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1931 				continue;
1932 			if ((ifp->if_flags & IFF_BROADCAST) &&
1933 			    ifa->ifa_broadaddr &&
1934 			    ifa->ifa_broadaddr->sa_len != 0 &&
1935 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1936 				ifa_ref(ifa);
1937 				IF_ADDR_RUNLOCK(ifp);
1938 				goto done;
1939 			}
1940 		}
1941 		IF_ADDR_RUNLOCK(ifp);
1942 	}
1943 	ifa = NULL;
1944 done:
1945 	IFNET_RUNLOCK_NOSLEEP();
1946 	return (ifa);
1947 }
1948 
1949 /*
1950  * Locate the point to point interface with a given destination address.
1951  */
1952 /*ARGSUSED*/
1953 struct ifaddr *
1954 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1955 {
1956 	struct ifnet *ifp;
1957 	struct ifaddr *ifa;
1958 
1959 	IFNET_RLOCK_NOSLEEP();
1960 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1961 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1962 			continue;
1963 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1964 			continue;
1965 		IF_ADDR_RLOCK(ifp);
1966 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1967 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1968 				continue;
1969 			if (ifa->ifa_dstaddr != NULL &&
1970 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1971 				ifa_ref(ifa);
1972 				IF_ADDR_RUNLOCK(ifp);
1973 				goto done;
1974 			}
1975 		}
1976 		IF_ADDR_RUNLOCK(ifp);
1977 	}
1978 	ifa = NULL;
1979 done:
1980 	IFNET_RUNLOCK_NOSLEEP();
1981 	return (ifa);
1982 }
1983 
1984 /*
1985  * Find an interface on a specific network.  If many, choice
1986  * is most specific found.
1987  */
1988 struct ifaddr *
1989 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1990 {
1991 	struct ifnet *ifp;
1992 	struct ifaddr *ifa;
1993 	struct ifaddr *ifa_maybe = NULL;
1994 	u_int af = addr->sa_family;
1995 	const char *addr_data = addr->sa_data, *cplim;
1996 
1997 	/*
1998 	 * AF_LINK addresses can be looked up directly by their index number,
1999 	 * so do that if we can.
2000 	 */
2001 	if (af == AF_LINK) {
2002 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2003 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2004 		return (ifaddr_byindex(sdl->sdl_index));
2005 	}
2006 
2007 	/*
2008 	 * Scan though each interface, looking for ones that have addresses
2009 	 * in this address family and the requested fib.  Maintain a reference
2010 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2011 	 * kept it stable when we move onto the next interface.
2012 	 */
2013 	IFNET_RLOCK_NOSLEEP();
2014 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2015 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2016 			continue;
2017 		IF_ADDR_RLOCK(ifp);
2018 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2019 			const char *cp, *cp2, *cp3;
2020 
2021 			if (ifa->ifa_addr->sa_family != af)
2022 next:				continue;
2023 			if (af == AF_INET &&
2024 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2025 				/*
2026 				 * This is a bit broken as it doesn't
2027 				 * take into account that the remote end may
2028 				 * be a single node in the network we are
2029 				 * looking for.
2030 				 * The trouble is that we don't know the
2031 				 * netmask for the remote end.
2032 				 */
2033 				if (ifa->ifa_dstaddr != NULL &&
2034 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2035 					ifa_ref(ifa);
2036 					IF_ADDR_RUNLOCK(ifp);
2037 					goto done;
2038 				}
2039 			} else {
2040 				/*
2041 				 * Scan all the bits in the ifa's address.
2042 				 * If a bit dissagrees with what we are
2043 				 * looking for, mask it with the netmask
2044 				 * to see if it really matters.
2045 				 * (A byte at a time)
2046 				 */
2047 				if (ifa->ifa_netmask == 0)
2048 					continue;
2049 				cp = addr_data;
2050 				cp2 = ifa->ifa_addr->sa_data;
2051 				cp3 = ifa->ifa_netmask->sa_data;
2052 				cplim = ifa->ifa_netmask->sa_len
2053 					+ (char *)ifa->ifa_netmask;
2054 				while (cp3 < cplim)
2055 					if ((*cp++ ^ *cp2++) & *cp3++)
2056 						goto next; /* next address! */
2057 				/*
2058 				 * If the netmask of what we just found
2059 				 * is more specific than what we had before
2060 				 * (if we had one), or if the virtual status
2061 				 * of new prefix is better than of the old one,
2062 				 * then remember the new one before continuing
2063 				 * to search for an even better one.
2064 				 */
2065 				if (ifa_maybe == NULL ||
2066 				    ifa_preferred(ifa_maybe, ifa) ||
2067 				    rn_refines((caddr_t)ifa->ifa_netmask,
2068 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2069 					if (ifa_maybe != NULL)
2070 						ifa_free(ifa_maybe);
2071 					ifa_maybe = ifa;
2072 					ifa_ref(ifa_maybe);
2073 				}
2074 			}
2075 		}
2076 		IF_ADDR_RUNLOCK(ifp);
2077 	}
2078 	ifa = ifa_maybe;
2079 	ifa_maybe = NULL;
2080 done:
2081 	IFNET_RUNLOCK_NOSLEEP();
2082 	if (ifa_maybe != NULL)
2083 		ifa_free(ifa_maybe);
2084 	return (ifa);
2085 }
2086 
2087 /*
2088  * Find an interface address specific to an interface best matching
2089  * a given address.
2090  */
2091 struct ifaddr *
2092 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2093 {
2094 	struct ifaddr *ifa;
2095 	const char *cp, *cp2, *cp3;
2096 	char *cplim;
2097 	struct ifaddr *ifa_maybe = NULL;
2098 	u_int af = addr->sa_family;
2099 
2100 	if (af >= AF_MAX)
2101 		return (NULL);
2102 	IF_ADDR_RLOCK(ifp);
2103 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2104 		if (ifa->ifa_addr->sa_family != af)
2105 			continue;
2106 		if (ifa_maybe == NULL)
2107 			ifa_maybe = ifa;
2108 		if (ifa->ifa_netmask == 0) {
2109 			if (sa_equal(addr, ifa->ifa_addr) ||
2110 			    (ifa->ifa_dstaddr &&
2111 			    sa_equal(addr, ifa->ifa_dstaddr)))
2112 				goto done;
2113 			continue;
2114 		}
2115 		if (ifp->if_flags & IFF_POINTOPOINT) {
2116 			if (sa_equal(addr, ifa->ifa_dstaddr))
2117 				goto done;
2118 		} else {
2119 			cp = addr->sa_data;
2120 			cp2 = ifa->ifa_addr->sa_data;
2121 			cp3 = ifa->ifa_netmask->sa_data;
2122 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2123 			for (; cp3 < cplim; cp3++)
2124 				if ((*cp++ ^ *cp2++) & *cp3)
2125 					break;
2126 			if (cp3 == cplim)
2127 				goto done;
2128 		}
2129 	}
2130 	ifa = ifa_maybe;
2131 done:
2132 	if (ifa != NULL)
2133 		ifa_ref(ifa);
2134 	IF_ADDR_RUNLOCK(ifp);
2135 	return (ifa);
2136 }
2137 
2138 /*
2139  * See whether new ifa is better than current one:
2140  * 1) A non-virtual one is preferred over virtual.
2141  * 2) A virtual in master state preferred over any other state.
2142  *
2143  * Used in several address selecting functions.
2144  */
2145 int
2146 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2147 {
2148 
2149 	return (cur->ifa_carp && (!next->ifa_carp ||
2150 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2151 }
2152 
2153 #include <net/if_llatbl.h>
2154 
2155 /*
2156  * Default action when installing a route with a Link Level gateway.
2157  * Lookup an appropriate real ifa to point to.
2158  * This should be moved to /sys/net/link.c eventually.
2159  */
2160 static void
2161 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2162 {
2163 	struct ifaddr *ifa, *oifa;
2164 	struct sockaddr *dst;
2165 	struct ifnet *ifp;
2166 
2167 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2168 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2169 		return;
2170 	ifa = ifaof_ifpforaddr(dst, ifp);
2171 	if (ifa) {
2172 		oifa = rt->rt_ifa;
2173 		rt->rt_ifa = ifa;
2174 		ifa_free(oifa);
2175 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2176 			ifa->ifa_rtrequest(cmd, rt, info);
2177 	}
2178 }
2179 
2180 struct sockaddr_dl *
2181 link_alloc_sdl(size_t size, int flags)
2182 {
2183 
2184 	return (malloc(size, M_TEMP, flags));
2185 }
2186 
2187 void
2188 link_free_sdl(struct sockaddr *sa)
2189 {
2190 	free(sa, M_TEMP);
2191 }
2192 
2193 /*
2194  * Fills in given sdl with interface basic info.
2195  * Returns pointer to filled sdl.
2196  */
2197 struct sockaddr_dl *
2198 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2199 {
2200 	struct sockaddr_dl *sdl;
2201 
2202 	sdl = (struct sockaddr_dl *)paddr;
2203 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2204 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2205 	sdl->sdl_family = AF_LINK;
2206 	sdl->sdl_index = ifp->if_index;
2207 	sdl->sdl_type = iftype;
2208 
2209 	return (sdl);
2210 }
2211 
2212 /*
2213  * Mark an interface down and notify protocols of
2214  * the transition.
2215  */
2216 static void
2217 if_unroute(struct ifnet *ifp, int flag, int fam)
2218 {
2219 	struct ifaddr *ifa;
2220 
2221 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2222 
2223 	ifp->if_flags &= ~flag;
2224 	getmicrotime(&ifp->if_lastchange);
2225 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2226 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2227 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2228 	ifp->if_qflush(ifp);
2229 
2230 	if (ifp->if_carp)
2231 		(*carp_linkstate_p)(ifp);
2232 	rt_ifmsg(ifp);
2233 }
2234 
2235 /*
2236  * Mark an interface up and notify protocols of
2237  * the transition.
2238  */
2239 static void
2240 if_route(struct ifnet *ifp, int flag, int fam)
2241 {
2242 	struct ifaddr *ifa;
2243 
2244 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2245 
2246 	ifp->if_flags |= flag;
2247 	getmicrotime(&ifp->if_lastchange);
2248 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2249 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2250 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2251 	if (ifp->if_carp)
2252 		(*carp_linkstate_p)(ifp);
2253 	rt_ifmsg(ifp);
2254 #ifdef INET6
2255 	in6_if_up(ifp);
2256 #endif
2257 }
2258 
2259 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2260 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2261 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2262 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2263 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2264 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2265 void	*(*vlan_cookie_p)(struct ifnet *);
2266 
2267 /*
2268  * Handle a change in the interface link state. To avoid LORs
2269  * between driver lock and upper layer locks, as well as possible
2270  * recursions, we post event to taskqueue, and all job
2271  * is done in static do_link_state_change().
2272  */
2273 void
2274 if_link_state_change(struct ifnet *ifp, int link_state)
2275 {
2276 	/* Return if state hasn't changed. */
2277 	if (ifp->if_link_state == link_state)
2278 		return;
2279 
2280 	ifp->if_link_state = link_state;
2281 
2282 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2283 }
2284 
2285 static void
2286 do_link_state_change(void *arg, int pending)
2287 {
2288 	struct ifnet *ifp = (struct ifnet *)arg;
2289 	int link_state = ifp->if_link_state;
2290 	CURVNET_SET(ifp->if_vnet);
2291 
2292 	/* Notify that the link state has changed. */
2293 	rt_ifmsg(ifp);
2294 	if (ifp->if_vlantrunk != NULL)
2295 		(*vlan_link_state_p)(ifp);
2296 
2297 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2298 	    ifp->if_l2com != NULL)
2299 		(*ng_ether_link_state_p)(ifp, link_state);
2300 	if (ifp->if_carp)
2301 		(*carp_linkstate_p)(ifp);
2302 	if (ifp->if_bridge)
2303 		(*bridge_linkstate_p)(ifp);
2304 	if (ifp->if_lagg)
2305 		(*lagg_linkstate_p)(ifp, link_state);
2306 
2307 	if (IS_DEFAULT_VNET(curvnet))
2308 		devctl_notify("IFNET", ifp->if_xname,
2309 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2310 		    NULL);
2311 	if (pending > 1)
2312 		if_printf(ifp, "%d link states coalesced\n", pending);
2313 	if (log_link_state_change)
2314 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2315 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2316 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2317 	CURVNET_RESTORE();
2318 }
2319 
2320 /*
2321  * Mark an interface down and notify protocols of
2322  * the transition.
2323  */
2324 void
2325 if_down(struct ifnet *ifp)
2326 {
2327 
2328 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2329 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2330 }
2331 
2332 /*
2333  * Mark an interface up and notify protocols of
2334  * the transition.
2335  */
2336 void
2337 if_up(struct ifnet *ifp)
2338 {
2339 
2340 	if_route(ifp, IFF_UP, AF_UNSPEC);
2341 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2342 }
2343 
2344 /*
2345  * Flush an interface queue.
2346  */
2347 void
2348 if_qflush(struct ifnet *ifp)
2349 {
2350 	struct mbuf *m, *n;
2351 	struct ifaltq *ifq;
2352 
2353 	ifq = &ifp->if_snd;
2354 	IFQ_LOCK(ifq);
2355 #ifdef ALTQ
2356 	if (ALTQ_IS_ENABLED(ifq))
2357 		ALTQ_PURGE(ifq);
2358 #endif
2359 	n = ifq->ifq_head;
2360 	while ((m = n) != NULL) {
2361 		n = m->m_nextpkt;
2362 		m_freem(m);
2363 	}
2364 	ifq->ifq_head = 0;
2365 	ifq->ifq_tail = 0;
2366 	ifq->ifq_len = 0;
2367 	IFQ_UNLOCK(ifq);
2368 }
2369 
2370 /*
2371  * Map interface name to interface structure pointer, with or without
2372  * returning a reference.
2373  */
2374 struct ifnet *
2375 ifunit_ref(const char *name)
2376 {
2377 	struct ifnet *ifp;
2378 
2379 	IFNET_RLOCK_NOSLEEP();
2380 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2381 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2382 		    !(ifp->if_flags & IFF_DYING))
2383 			break;
2384 	}
2385 	if (ifp != NULL)
2386 		if_ref(ifp);
2387 	IFNET_RUNLOCK_NOSLEEP();
2388 	return (ifp);
2389 }
2390 
2391 struct ifnet *
2392 ifunit(const char *name)
2393 {
2394 	struct ifnet *ifp;
2395 
2396 	IFNET_RLOCK_NOSLEEP();
2397 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2398 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2399 			break;
2400 	}
2401 	IFNET_RUNLOCK_NOSLEEP();
2402 	return (ifp);
2403 }
2404 
2405 static void *
2406 ifr_buffer_get_buffer(void *data)
2407 {
2408 	union ifreq_union *ifrup;
2409 
2410 	ifrup = data;
2411 #ifdef COMPAT_FREEBSD32
2412 	if (SV_CURPROC_FLAG(SV_ILP32))
2413 		return ((void *)(uintptr_t)
2414 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2415 #endif
2416 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2417 }
2418 
2419 static void
2420 ifr_buffer_set_buffer_null(void *data)
2421 {
2422 	union ifreq_union *ifrup;
2423 
2424 	ifrup = data;
2425 #ifdef COMPAT_FREEBSD32
2426 	if (SV_CURPROC_FLAG(SV_ILP32))
2427 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2428 	else
2429 #endif
2430 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2431 }
2432 
2433 static size_t
2434 ifr_buffer_get_length(void *data)
2435 {
2436 	union ifreq_union *ifrup;
2437 
2438 	ifrup = data;
2439 #ifdef COMPAT_FREEBSD32
2440 	if (SV_CURPROC_FLAG(SV_ILP32))
2441 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2442 #endif
2443 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2444 }
2445 
2446 static void
2447 ifr_buffer_set_length(void *data, size_t len)
2448 {
2449 	union ifreq_union *ifrup;
2450 
2451 	ifrup = data;
2452 #ifdef COMPAT_FREEBSD32
2453 	if (SV_CURPROC_FLAG(SV_ILP32))
2454 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2455 	else
2456 #endif
2457 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2458 }
2459 
2460 void *
2461 ifr_data_get_ptr(void *ifrp)
2462 {
2463 	union ifreq_union *ifrup;
2464 
2465 	ifrup = ifrp;
2466 #ifdef COMPAT_FREEBSD32
2467 	if (SV_CURPROC_FLAG(SV_ILP32))
2468 		return ((void *)(uintptr_t)
2469 		    ifrup->ifr32.ifr_ifru.ifru_data);
2470 #endif
2471 		return (ifrup->ifr.ifr_ifru.ifru_data);
2472 }
2473 
2474 /*
2475  * Hardware specific interface ioctls.
2476  */
2477 static int
2478 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2479 {
2480 	struct ifreq *ifr;
2481 	int error = 0, do_ifup = 0;
2482 	int new_flags, temp_flags;
2483 	size_t namelen, onamelen;
2484 	size_t descrlen;
2485 	char *descrbuf, *odescrbuf;
2486 	char new_name[IFNAMSIZ];
2487 	struct ifaddr *ifa;
2488 	struct sockaddr_dl *sdl;
2489 
2490 	ifr = (struct ifreq *)data;
2491 	switch (cmd) {
2492 	case SIOCGIFINDEX:
2493 		ifr->ifr_index = ifp->if_index;
2494 		break;
2495 
2496 	case SIOCGIFFLAGS:
2497 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2498 		ifr->ifr_flags = temp_flags & 0xffff;
2499 		ifr->ifr_flagshigh = temp_flags >> 16;
2500 		break;
2501 
2502 	case SIOCGIFCAP:
2503 		ifr->ifr_reqcap = ifp->if_capabilities;
2504 		ifr->ifr_curcap = ifp->if_capenable;
2505 		break;
2506 
2507 #ifdef MAC
2508 	case SIOCGIFMAC:
2509 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2510 		break;
2511 #endif
2512 
2513 	case SIOCGIFMETRIC:
2514 		ifr->ifr_metric = ifp->if_metric;
2515 		break;
2516 
2517 	case SIOCGIFMTU:
2518 		ifr->ifr_mtu = ifp->if_mtu;
2519 		break;
2520 
2521 	case SIOCGIFPHYS:
2522 		/* XXXGL: did this ever worked? */
2523 		ifr->ifr_phys = 0;
2524 		break;
2525 
2526 	case SIOCGIFDESCR:
2527 		error = 0;
2528 		sx_slock(&ifdescr_sx);
2529 		if (ifp->if_description == NULL)
2530 			error = ENOMSG;
2531 		else {
2532 			/* space for terminating nul */
2533 			descrlen = strlen(ifp->if_description) + 1;
2534 			if (ifr_buffer_get_length(ifr) < descrlen)
2535 				ifr_buffer_set_buffer_null(ifr);
2536 			else
2537 				error = copyout(ifp->if_description,
2538 				    ifr_buffer_get_buffer(ifr), descrlen);
2539 			ifr_buffer_set_length(ifr, descrlen);
2540 		}
2541 		sx_sunlock(&ifdescr_sx);
2542 		break;
2543 
2544 	case SIOCSIFDESCR:
2545 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2546 		if (error)
2547 			return (error);
2548 
2549 		/*
2550 		 * Copy only (length-1) bytes to make sure that
2551 		 * if_description is always nul terminated.  The
2552 		 * length parameter is supposed to count the
2553 		 * terminating nul in.
2554 		 */
2555 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2556 			return (ENAMETOOLONG);
2557 		else if (ifr_buffer_get_length(ifr) == 0)
2558 			descrbuf = NULL;
2559 		else {
2560 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2561 			    M_IFDESCR, M_WAITOK | M_ZERO);
2562 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2563 			    ifr_buffer_get_length(ifr) - 1);
2564 			if (error) {
2565 				free(descrbuf, M_IFDESCR);
2566 				break;
2567 			}
2568 		}
2569 
2570 		sx_xlock(&ifdescr_sx);
2571 		odescrbuf = ifp->if_description;
2572 		ifp->if_description = descrbuf;
2573 		sx_xunlock(&ifdescr_sx);
2574 
2575 		getmicrotime(&ifp->if_lastchange);
2576 		free(odescrbuf, M_IFDESCR);
2577 		break;
2578 
2579 	case SIOCGIFFIB:
2580 		ifr->ifr_fib = ifp->if_fib;
2581 		break;
2582 
2583 	case SIOCSIFFIB:
2584 		error = priv_check(td, PRIV_NET_SETIFFIB);
2585 		if (error)
2586 			return (error);
2587 		if (ifr->ifr_fib >= rt_numfibs)
2588 			return (EINVAL);
2589 
2590 		ifp->if_fib = ifr->ifr_fib;
2591 		break;
2592 
2593 	case SIOCSIFFLAGS:
2594 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2595 		if (error)
2596 			return (error);
2597 		/*
2598 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2599 		 * check, so we don't need special handling here yet.
2600 		 */
2601 		new_flags = (ifr->ifr_flags & 0xffff) |
2602 		    (ifr->ifr_flagshigh << 16);
2603 		if (ifp->if_flags & IFF_UP &&
2604 		    (new_flags & IFF_UP) == 0) {
2605 			if_down(ifp);
2606 		} else if (new_flags & IFF_UP &&
2607 		    (ifp->if_flags & IFF_UP) == 0) {
2608 			do_ifup = 1;
2609 		}
2610 		/* See if permanently promiscuous mode bit is about to flip */
2611 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2612 			if (new_flags & IFF_PPROMISC)
2613 				ifp->if_flags |= IFF_PROMISC;
2614 			else if (ifp->if_pcount == 0)
2615 				ifp->if_flags &= ~IFF_PROMISC;
2616 			if (log_promisc_mode_change)
2617                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2618                                     ifp->if_xname,
2619                                     ((new_flags & IFF_PPROMISC) ?
2620                                      "enabled" : "disabled"));
2621 		}
2622 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2623 			(new_flags &~ IFF_CANTCHANGE);
2624 		if (ifp->if_ioctl) {
2625 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2626 		}
2627 		if (do_ifup)
2628 			if_up(ifp);
2629 		getmicrotime(&ifp->if_lastchange);
2630 		break;
2631 
2632 	case SIOCSIFCAP:
2633 		error = priv_check(td, PRIV_NET_SETIFCAP);
2634 		if (error)
2635 			return (error);
2636 		if (ifp->if_ioctl == NULL)
2637 			return (EOPNOTSUPP);
2638 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2639 			return (EINVAL);
2640 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2641 		if (error == 0)
2642 			getmicrotime(&ifp->if_lastchange);
2643 		break;
2644 
2645 #ifdef MAC
2646 	case SIOCSIFMAC:
2647 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2648 		break;
2649 #endif
2650 
2651 	case SIOCSIFNAME:
2652 		error = priv_check(td, PRIV_NET_SETIFNAME);
2653 		if (error)
2654 			return (error);
2655 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2656 		    NULL);
2657 		if (error != 0)
2658 			return (error);
2659 		if (new_name[0] == '\0')
2660 			return (EINVAL);
2661 		if (new_name[IFNAMSIZ-1] != '\0') {
2662 			new_name[IFNAMSIZ-1] = '\0';
2663 			if (strlen(new_name) == IFNAMSIZ-1)
2664 				return (EINVAL);
2665 		}
2666 		if (ifunit(new_name) != NULL)
2667 			return (EEXIST);
2668 
2669 		/*
2670 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2671 		 * and there are numerous other races with the
2672 		 * ifunit() checks not being atomic with namespace
2673 		 * changes (renames, vmoves, if_attach, etc).
2674 		 */
2675 		ifp->if_flags |= IFF_RENAMING;
2676 
2677 		/* Announce the departure of the interface. */
2678 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2679 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2680 
2681 		log(LOG_INFO, "%s: changing name to '%s'\n",
2682 		    ifp->if_xname, new_name);
2683 
2684 		IF_ADDR_WLOCK(ifp);
2685 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2686 		ifa = ifp->if_addr;
2687 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2688 		namelen = strlen(new_name);
2689 		onamelen = sdl->sdl_nlen;
2690 		/*
2691 		 * Move the address if needed.  This is safe because we
2692 		 * allocate space for a name of length IFNAMSIZ when we
2693 		 * create this in if_attach().
2694 		 */
2695 		if (namelen != onamelen) {
2696 			bcopy(sdl->sdl_data + onamelen,
2697 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2698 		}
2699 		bcopy(new_name, sdl->sdl_data, namelen);
2700 		sdl->sdl_nlen = namelen;
2701 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2702 		bzero(sdl->sdl_data, onamelen);
2703 		while (namelen != 0)
2704 			sdl->sdl_data[--namelen] = 0xff;
2705 		IF_ADDR_WUNLOCK(ifp);
2706 
2707 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2708 		/* Announce the return of the interface. */
2709 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2710 
2711 		ifp->if_flags &= ~IFF_RENAMING;
2712 		break;
2713 
2714 #ifdef VIMAGE
2715 	case SIOCSIFVNET:
2716 		error = priv_check(td, PRIV_NET_SETIFVNET);
2717 		if (error)
2718 			return (error);
2719 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2720 		break;
2721 #endif
2722 
2723 	case SIOCSIFMETRIC:
2724 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2725 		if (error)
2726 			return (error);
2727 		ifp->if_metric = ifr->ifr_metric;
2728 		getmicrotime(&ifp->if_lastchange);
2729 		break;
2730 
2731 	case SIOCSIFPHYS:
2732 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2733 		if (error)
2734 			return (error);
2735 		if (ifp->if_ioctl == NULL)
2736 			return (EOPNOTSUPP);
2737 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2738 		if (error == 0)
2739 			getmicrotime(&ifp->if_lastchange);
2740 		break;
2741 
2742 	case SIOCSIFMTU:
2743 	{
2744 		u_long oldmtu = ifp->if_mtu;
2745 
2746 		error = priv_check(td, PRIV_NET_SETIFMTU);
2747 		if (error)
2748 			return (error);
2749 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2750 			return (EINVAL);
2751 		if (ifp->if_ioctl == NULL)
2752 			return (EOPNOTSUPP);
2753 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2754 		if (error == 0) {
2755 			getmicrotime(&ifp->if_lastchange);
2756 			rt_ifmsg(ifp);
2757 		}
2758 		/*
2759 		 * If the link MTU changed, do network layer specific procedure.
2760 		 */
2761 		if (ifp->if_mtu != oldmtu) {
2762 #ifdef INET6
2763 			nd6_setmtu(ifp);
2764 #endif
2765 			rt_updatemtu(ifp);
2766 		}
2767 		break;
2768 	}
2769 
2770 	case SIOCADDMULTI:
2771 	case SIOCDELMULTI:
2772 		if (cmd == SIOCADDMULTI)
2773 			error = priv_check(td, PRIV_NET_ADDMULTI);
2774 		else
2775 			error = priv_check(td, PRIV_NET_DELMULTI);
2776 		if (error)
2777 			return (error);
2778 
2779 		/* Don't allow group membership on non-multicast interfaces. */
2780 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2781 			return (EOPNOTSUPP);
2782 
2783 		/* Don't let users screw up protocols' entries. */
2784 		if (ifr->ifr_addr.sa_family != AF_LINK)
2785 			return (EINVAL);
2786 
2787 		if (cmd == SIOCADDMULTI) {
2788 			struct ifmultiaddr *ifma;
2789 
2790 			/*
2791 			 * Userland is only permitted to join groups once
2792 			 * via the if_addmulti() KPI, because it cannot hold
2793 			 * struct ifmultiaddr * between calls. It may also
2794 			 * lose a race while we check if the membership
2795 			 * already exists.
2796 			 */
2797 			IF_ADDR_RLOCK(ifp);
2798 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2799 			IF_ADDR_RUNLOCK(ifp);
2800 			if (ifma != NULL)
2801 				error = EADDRINUSE;
2802 			else
2803 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2804 		} else {
2805 			error = if_delmulti(ifp, &ifr->ifr_addr);
2806 		}
2807 		if (error == 0)
2808 			getmicrotime(&ifp->if_lastchange);
2809 		break;
2810 
2811 	case SIOCSIFPHYADDR:
2812 	case SIOCDIFPHYADDR:
2813 #ifdef INET6
2814 	case SIOCSIFPHYADDR_IN6:
2815 #endif
2816 	case SIOCSIFMEDIA:
2817 	case SIOCSIFGENERIC:
2818 		error = priv_check(td, PRIV_NET_HWIOCTL);
2819 		if (error)
2820 			return (error);
2821 		if (ifp->if_ioctl == NULL)
2822 			return (EOPNOTSUPP);
2823 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2824 		if (error == 0)
2825 			getmicrotime(&ifp->if_lastchange);
2826 		break;
2827 
2828 	case SIOCGIFSTATUS:
2829 	case SIOCGIFPSRCADDR:
2830 	case SIOCGIFPDSTADDR:
2831 	case SIOCGIFMEDIA:
2832 	case SIOCGIFXMEDIA:
2833 	case SIOCGIFGENERIC:
2834 	case SIOCGIFRSSKEY:
2835 	case SIOCGIFRSSHASH:
2836 		if (ifp->if_ioctl == NULL)
2837 			return (EOPNOTSUPP);
2838 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2839 		break;
2840 
2841 	case SIOCSIFLLADDR:
2842 		error = priv_check(td, PRIV_NET_SETLLADDR);
2843 		if (error)
2844 			return (error);
2845 		error = if_setlladdr(ifp,
2846 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2847 		break;
2848 
2849 	case SIOCGHWADDR:
2850 		error = if_gethwaddr(ifp, ifr);
2851 		break;
2852 
2853 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2854 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2855 		if (error)
2856 			return (error);
2857 		if ((error = if_addgroup(ifp,
2858 		    ifgr_group_get((struct ifgroupreq *)data))))
2859 			return (error);
2860 		break;
2861 
2862 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2863 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2864 			return (error);
2865 		break;
2866 
2867 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2868 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2869 		if (error)
2870 			return (error);
2871 		if ((error = if_delgroup(ifp,
2872 		    ifgr_group_get((struct ifgroupreq *)data))))
2873 			return (error);
2874 		break;
2875 
2876 	default:
2877 		error = ENOIOCTL;
2878 		break;
2879 	}
2880 	return (error);
2881 }
2882 
2883 #ifdef COMPAT_FREEBSD32
2884 struct ifconf32 {
2885 	int32_t	ifc_len;
2886 	union {
2887 		uint32_t	ifcu_buf;
2888 		uint32_t	ifcu_req;
2889 	} ifc_ifcu;
2890 };
2891 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2892 #endif
2893 
2894 /*
2895  * Interface ioctls.
2896  */
2897 int
2898 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2899 {
2900 	struct ifnet *ifp;
2901 	struct ifreq *ifr;
2902 	int error;
2903 	int oif_flags;
2904 #ifdef VIMAGE
2905 	int shutdown;
2906 #endif
2907 
2908 	CURVNET_SET(so->so_vnet);
2909 #ifdef VIMAGE
2910 	/* Make sure the VNET is stable. */
2911 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2912 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2913 	if (shutdown) {
2914 		CURVNET_RESTORE();
2915 		return (EBUSY);
2916 	}
2917 #endif
2918 
2919 
2920 	switch (cmd) {
2921 	case SIOCGIFCONF:
2922 		error = ifconf(cmd, data);
2923 		CURVNET_RESTORE();
2924 		return (error);
2925 
2926 #ifdef COMPAT_FREEBSD32
2927 	case SIOCGIFCONF32:
2928 		{
2929 			struct ifconf32 *ifc32;
2930 			struct ifconf ifc;
2931 
2932 			ifc32 = (struct ifconf32 *)data;
2933 			ifc.ifc_len = ifc32->ifc_len;
2934 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2935 
2936 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2937 			CURVNET_RESTORE();
2938 			if (error == 0)
2939 				ifc32->ifc_len = ifc.ifc_len;
2940 			return (error);
2941 		}
2942 #endif
2943 	}
2944 	ifr = (struct ifreq *)data;
2945 
2946 	switch (cmd) {
2947 #ifdef VIMAGE
2948 	case SIOCSIFRVNET:
2949 		error = priv_check(td, PRIV_NET_SETIFVNET);
2950 		if (error == 0)
2951 			error = if_vmove_reclaim(td, ifr->ifr_name,
2952 			    ifr->ifr_jid);
2953 		CURVNET_RESTORE();
2954 		return (error);
2955 #endif
2956 	case SIOCIFCREATE:
2957 	case SIOCIFCREATE2:
2958 		error = priv_check(td, PRIV_NET_IFCREATE);
2959 		if (error == 0)
2960 			error = if_clone_create(ifr->ifr_name,
2961 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
2962 			    ifr_data_get_ptr(ifr) : NULL);
2963 		CURVNET_RESTORE();
2964 		return (error);
2965 	case SIOCIFDESTROY:
2966 		error = priv_check(td, PRIV_NET_IFDESTROY);
2967 		if (error == 0)
2968 			error = if_clone_destroy(ifr->ifr_name);
2969 		CURVNET_RESTORE();
2970 		return (error);
2971 
2972 	case SIOCIFGCLONERS:
2973 		error = if_clone_list((struct if_clonereq *)data);
2974 		CURVNET_RESTORE();
2975 		return (error);
2976 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
2977 		error = if_getgroupmembers((struct ifgroupreq *)data);
2978 		CURVNET_RESTORE();
2979 		return (error);
2980 #if defined(INET) || defined(INET6)
2981 	case SIOCSVH:
2982 	case SIOCGVH:
2983 		if (carp_ioctl_p == NULL)
2984 			error = EPROTONOSUPPORT;
2985 		else
2986 			error = (*carp_ioctl_p)(ifr, cmd, td);
2987 		CURVNET_RESTORE();
2988 		return (error);
2989 #endif
2990 	}
2991 
2992 	ifp = ifunit_ref(ifr->ifr_name);
2993 	if (ifp == NULL) {
2994 		CURVNET_RESTORE();
2995 		return (ENXIO);
2996 	}
2997 
2998 	error = ifhwioctl(cmd, ifp, data, td);
2999 	if (error != ENOIOCTL) {
3000 		if_rele(ifp);
3001 		CURVNET_RESTORE();
3002 		return (error);
3003 	}
3004 
3005 	oif_flags = ifp->if_flags;
3006 	if (so->so_proto == NULL) {
3007 		if_rele(ifp);
3008 		CURVNET_RESTORE();
3009 		return (EOPNOTSUPP);
3010 	}
3011 
3012 	/*
3013 	 * Pass the request on to the socket control method, and if the
3014 	 * latter returns EOPNOTSUPP, directly to the interface.
3015 	 *
3016 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3017 	 * trust SIOCSIFADDR et al to come from an already privileged
3018 	 * layer, and do not perform any credentials checks or input
3019 	 * validation.
3020 	 */
3021 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3022 	    ifp, td));
3023 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3024 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3025 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3026 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3027 
3028 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3029 #ifdef INET6
3030 		if (ifp->if_flags & IFF_UP)
3031 			in6_if_up(ifp);
3032 #endif
3033 	}
3034 	if_rele(ifp);
3035 	CURVNET_RESTORE();
3036 	return (error);
3037 }
3038 
3039 /*
3040  * The code common to handling reference counted flags,
3041  * e.g., in ifpromisc() and if_allmulti().
3042  * The "pflag" argument can specify a permanent mode flag to check,
3043  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3044  *
3045  * Only to be used on stack-owned flags, not driver-owned flags.
3046  */
3047 static int
3048 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3049 {
3050 	struct ifreq ifr;
3051 	int error;
3052 	int oldflags, oldcount;
3053 
3054 	/* Sanity checks to catch programming errors */
3055 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3056 	    ("%s: setting driver-owned flag %d", __func__, flag));
3057 
3058 	if (onswitch)
3059 		KASSERT(*refcount >= 0,
3060 		    ("%s: increment negative refcount %d for flag %d",
3061 		    __func__, *refcount, flag));
3062 	else
3063 		KASSERT(*refcount > 0,
3064 		    ("%s: decrement non-positive refcount %d for flag %d",
3065 		    __func__, *refcount, flag));
3066 
3067 	/* In case this mode is permanent, just touch refcount */
3068 	if (ifp->if_flags & pflag) {
3069 		*refcount += onswitch ? 1 : -1;
3070 		return (0);
3071 	}
3072 
3073 	/* Save ifnet parameters for if_ioctl() may fail */
3074 	oldcount = *refcount;
3075 	oldflags = ifp->if_flags;
3076 
3077 	/*
3078 	 * See if we aren't the only and touching refcount is enough.
3079 	 * Actually toggle interface flag if we are the first or last.
3080 	 */
3081 	if (onswitch) {
3082 		if ((*refcount)++)
3083 			return (0);
3084 		ifp->if_flags |= flag;
3085 	} else {
3086 		if (--(*refcount))
3087 			return (0);
3088 		ifp->if_flags &= ~flag;
3089 	}
3090 
3091 	/* Call down the driver since we've changed interface flags */
3092 	if (ifp->if_ioctl == NULL) {
3093 		error = EOPNOTSUPP;
3094 		goto recover;
3095 	}
3096 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3097 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3098 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3099 	if (error)
3100 		goto recover;
3101 	/* Notify userland that interface flags have changed */
3102 	rt_ifmsg(ifp);
3103 	return (0);
3104 
3105 recover:
3106 	/* Recover after driver error */
3107 	*refcount = oldcount;
3108 	ifp->if_flags = oldflags;
3109 	return (error);
3110 }
3111 
3112 /*
3113  * Set/clear promiscuous mode on interface ifp based on the truth value
3114  * of pswitch.  The calls are reference counted so that only the first
3115  * "on" request actually has an effect, as does the final "off" request.
3116  * Results are undefined if the "off" and "on" requests are not matched.
3117  */
3118 int
3119 ifpromisc(struct ifnet *ifp, int pswitch)
3120 {
3121 	int error;
3122 	int oldflags = ifp->if_flags;
3123 
3124 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3125 			   &ifp->if_pcount, pswitch);
3126 	/* If promiscuous mode status has changed, log a message */
3127 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3128             log_promisc_mode_change)
3129 		log(LOG_INFO, "%s: promiscuous mode %s\n",
3130 		    ifp->if_xname,
3131 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3132 	return (error);
3133 }
3134 
3135 /*
3136  * Return interface configuration
3137  * of system.  List may be used
3138  * in later ioctl's (above) to get
3139  * other information.
3140  */
3141 /*ARGSUSED*/
3142 static int
3143 ifconf(u_long cmd, caddr_t data)
3144 {
3145 	struct ifconf *ifc = (struct ifconf *)data;
3146 	struct ifnet *ifp;
3147 	struct ifaddr *ifa;
3148 	struct ifreq ifr;
3149 	struct sbuf *sb;
3150 	int error, full = 0, valid_len, max_len;
3151 
3152 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3153 	max_len = MAXPHYS - 1;
3154 
3155 	/* Prevent hostile input from being able to crash the system */
3156 	if (ifc->ifc_len <= 0)
3157 		return (EINVAL);
3158 
3159 again:
3160 	if (ifc->ifc_len <= max_len) {
3161 		max_len = ifc->ifc_len;
3162 		full = 1;
3163 	}
3164 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3165 	max_len = 0;
3166 	valid_len = 0;
3167 
3168 	IFNET_RLOCK();
3169 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3170 		int addrs;
3171 
3172 		/*
3173 		 * Zero the ifr to make sure we don't disclose the contents
3174 		 * of the stack.
3175 		 */
3176 		memset(&ifr, 0, sizeof(ifr));
3177 
3178 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3179 		    >= sizeof(ifr.ifr_name)) {
3180 			sbuf_delete(sb);
3181 			IFNET_RUNLOCK();
3182 			return (ENAMETOOLONG);
3183 		}
3184 
3185 		addrs = 0;
3186 		IF_ADDR_RLOCK(ifp);
3187 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3188 			struct sockaddr *sa = ifa->ifa_addr;
3189 
3190 			if (prison_if(curthread->td_ucred, sa) != 0)
3191 				continue;
3192 			addrs++;
3193 			if (sa->sa_len <= sizeof(*sa)) {
3194 				if (sa->sa_len < sizeof(*sa)) {
3195 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3196 					    sizeof(ifr.ifr_ifru.ifru_addr));
3197 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3198 					    sa->sa_len);
3199 				} else
3200 					ifr.ifr_ifru.ifru_addr = *sa;
3201 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3202 				max_len += sizeof(ifr);
3203 			} else {
3204 				sbuf_bcat(sb, &ifr,
3205 				    offsetof(struct ifreq, ifr_addr));
3206 				max_len += offsetof(struct ifreq, ifr_addr);
3207 				sbuf_bcat(sb, sa, sa->sa_len);
3208 				max_len += sa->sa_len;
3209 			}
3210 
3211 			if (sbuf_error(sb) == 0)
3212 				valid_len = sbuf_len(sb);
3213 		}
3214 		IF_ADDR_RUNLOCK(ifp);
3215 		if (addrs == 0) {
3216 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3217 			max_len += sizeof(ifr);
3218 
3219 			if (sbuf_error(sb) == 0)
3220 				valid_len = sbuf_len(sb);
3221 		}
3222 	}
3223 	IFNET_RUNLOCK();
3224 
3225 	/*
3226 	 * If we didn't allocate enough space (uncommon), try again.  If
3227 	 * we have already allocated as much space as we are allowed,
3228 	 * return what we've got.
3229 	 */
3230 	if (valid_len != max_len && !full) {
3231 		sbuf_delete(sb);
3232 		goto again;
3233 	}
3234 
3235 	ifc->ifc_len = valid_len;
3236 	sbuf_finish(sb);
3237 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3238 	sbuf_delete(sb);
3239 	return (error);
3240 }
3241 
3242 /*
3243  * Just like ifpromisc(), but for all-multicast-reception mode.
3244  */
3245 int
3246 if_allmulti(struct ifnet *ifp, int onswitch)
3247 {
3248 
3249 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3250 }
3251 
3252 struct ifmultiaddr *
3253 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3254 {
3255 	struct ifmultiaddr *ifma;
3256 
3257 	IF_ADDR_LOCK_ASSERT(ifp);
3258 
3259 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3260 		if (sa->sa_family == AF_LINK) {
3261 			if (sa_dl_equal(ifma->ifma_addr, sa))
3262 				break;
3263 		} else {
3264 			if (sa_equal(ifma->ifma_addr, sa))
3265 				break;
3266 		}
3267 	}
3268 
3269 	return ifma;
3270 }
3271 
3272 /*
3273  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3274  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3275  * the ifnet multicast address list here, so the caller must do that and
3276  * other setup work (such as notifying the device driver).  The reference
3277  * count is initialized to 1.
3278  */
3279 static struct ifmultiaddr *
3280 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3281     int mflags)
3282 {
3283 	struct ifmultiaddr *ifma;
3284 	struct sockaddr *dupsa;
3285 
3286 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3287 	    M_ZERO);
3288 	if (ifma == NULL)
3289 		return (NULL);
3290 
3291 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3292 	if (dupsa == NULL) {
3293 		free(ifma, M_IFMADDR);
3294 		return (NULL);
3295 	}
3296 	bcopy(sa, dupsa, sa->sa_len);
3297 	ifma->ifma_addr = dupsa;
3298 
3299 	ifma->ifma_ifp = ifp;
3300 	ifma->ifma_refcount = 1;
3301 	ifma->ifma_protospec = NULL;
3302 
3303 	if (llsa == NULL) {
3304 		ifma->ifma_lladdr = NULL;
3305 		return (ifma);
3306 	}
3307 
3308 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3309 	if (dupsa == NULL) {
3310 		free(ifma->ifma_addr, M_IFMADDR);
3311 		free(ifma, M_IFMADDR);
3312 		return (NULL);
3313 	}
3314 	bcopy(llsa, dupsa, llsa->sa_len);
3315 	ifma->ifma_lladdr = dupsa;
3316 
3317 	return (ifma);
3318 }
3319 
3320 /*
3321  * if_freemulti: free ifmultiaddr structure and possibly attached related
3322  * addresses.  The caller is responsible for implementing reference
3323  * counting, notifying the driver, handling routing messages, and releasing
3324  * any dependent link layer state.
3325  */
3326 static void
3327 if_freemulti(struct ifmultiaddr *ifma)
3328 {
3329 
3330 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3331 	    ifma->ifma_refcount));
3332 
3333 	if (ifma->ifma_lladdr != NULL)
3334 		free(ifma->ifma_lladdr, M_IFMADDR);
3335 	free(ifma->ifma_addr, M_IFMADDR);
3336 	free(ifma, M_IFMADDR);
3337 }
3338 
3339 /*
3340  * Register an additional multicast address with a network interface.
3341  *
3342  * - If the address is already present, bump the reference count on the
3343  *   address and return.
3344  * - If the address is not link-layer, look up a link layer address.
3345  * - Allocate address structures for one or both addresses, and attach to the
3346  *   multicast address list on the interface.  If automatically adding a link
3347  *   layer address, the protocol address will own a reference to the link
3348  *   layer address, to be freed when it is freed.
3349  * - Notify the network device driver of an addition to the multicast address
3350  *   list.
3351  *
3352  * 'sa' points to caller-owned memory with the desired multicast address.
3353  *
3354  * 'retifma' will be used to return a pointer to the resulting multicast
3355  * address reference, if desired.
3356  */
3357 int
3358 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3359     struct ifmultiaddr **retifma)
3360 {
3361 	struct ifmultiaddr *ifma, *ll_ifma;
3362 	struct sockaddr *llsa;
3363 	struct sockaddr_dl sdl;
3364 	int error;
3365 
3366 	/*
3367 	 * If the address is already present, return a new reference to it;
3368 	 * otherwise, allocate storage and set up a new address.
3369 	 */
3370 	IF_ADDR_WLOCK(ifp);
3371 	ifma = if_findmulti(ifp, sa);
3372 	if (ifma != NULL) {
3373 		ifma->ifma_refcount++;
3374 		if (retifma != NULL)
3375 			*retifma = ifma;
3376 		IF_ADDR_WUNLOCK(ifp);
3377 		return (0);
3378 	}
3379 
3380 	/*
3381 	 * The address isn't already present; resolve the protocol address
3382 	 * into a link layer address, and then look that up, bump its
3383 	 * refcount or allocate an ifma for that also.
3384 	 * Most link layer resolving functions returns address data which
3385 	 * fits inside default sockaddr_dl structure. However callback
3386 	 * can allocate another sockaddr structure, in that case we need to
3387 	 * free it later.
3388 	 */
3389 	llsa = NULL;
3390 	ll_ifma = NULL;
3391 	if (ifp->if_resolvemulti != NULL) {
3392 		/* Provide called function with buffer size information */
3393 		sdl.sdl_len = sizeof(sdl);
3394 		llsa = (struct sockaddr *)&sdl;
3395 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3396 		if (error)
3397 			goto unlock_out;
3398 	}
3399 
3400 	/*
3401 	 * Allocate the new address.  Don't hook it up yet, as we may also
3402 	 * need to allocate a link layer multicast address.
3403 	 */
3404 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3405 	if (ifma == NULL) {
3406 		error = ENOMEM;
3407 		goto free_llsa_out;
3408 	}
3409 
3410 	/*
3411 	 * If a link layer address is found, we'll need to see if it's
3412 	 * already present in the address list, or allocate is as well.
3413 	 * When this block finishes, the link layer address will be on the
3414 	 * list.
3415 	 */
3416 	if (llsa != NULL) {
3417 		ll_ifma = if_findmulti(ifp, llsa);
3418 		if (ll_ifma == NULL) {
3419 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3420 			if (ll_ifma == NULL) {
3421 				--ifma->ifma_refcount;
3422 				if_freemulti(ifma);
3423 				error = ENOMEM;
3424 				goto free_llsa_out;
3425 			}
3426 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3427 			    ifma_link);
3428 		} else
3429 			ll_ifma->ifma_refcount++;
3430 		ifma->ifma_llifma = ll_ifma;
3431 	}
3432 
3433 	/*
3434 	 * We now have a new multicast address, ifma, and possibly a new or
3435 	 * referenced link layer address.  Add the primary address to the
3436 	 * ifnet address list.
3437 	 */
3438 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3439 
3440 	if (retifma != NULL)
3441 		*retifma = ifma;
3442 
3443 	/*
3444 	 * Must generate the message while holding the lock so that 'ifma'
3445 	 * pointer is still valid.
3446 	 */
3447 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3448 	IF_ADDR_WUNLOCK(ifp);
3449 
3450 	/*
3451 	 * We are certain we have added something, so call down to the
3452 	 * interface to let them know about it.
3453 	 */
3454 	if (ifp->if_ioctl != NULL) {
3455 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3456 	}
3457 
3458 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3459 		link_free_sdl(llsa);
3460 
3461 	return (0);
3462 
3463 free_llsa_out:
3464 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3465 		link_free_sdl(llsa);
3466 
3467 unlock_out:
3468 	IF_ADDR_WUNLOCK(ifp);
3469 	return (error);
3470 }
3471 
3472 /*
3473  * Delete a multicast group membership by network-layer group address.
3474  *
3475  * Returns ENOENT if the entry could not be found. If ifp no longer
3476  * exists, results are undefined. This entry point should only be used
3477  * from subsystems which do appropriate locking to hold ifp for the
3478  * duration of the call.
3479  * Network-layer protocol domains must use if_delmulti_ifma().
3480  */
3481 int
3482 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3483 {
3484 	struct ifmultiaddr *ifma;
3485 	int lastref;
3486 #ifdef INVARIANTS
3487 	struct ifnet *oifp;
3488 
3489 	IFNET_RLOCK_NOSLEEP();
3490 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3491 		if (ifp == oifp)
3492 			break;
3493 	if (ifp != oifp)
3494 		ifp = NULL;
3495 	IFNET_RUNLOCK_NOSLEEP();
3496 
3497 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3498 #endif
3499 	if (ifp == NULL)
3500 		return (ENOENT);
3501 
3502 	IF_ADDR_WLOCK(ifp);
3503 	lastref = 0;
3504 	ifma = if_findmulti(ifp, sa);
3505 	if (ifma != NULL)
3506 		lastref = if_delmulti_locked(ifp, ifma, 0);
3507 	IF_ADDR_WUNLOCK(ifp);
3508 
3509 	if (ifma == NULL)
3510 		return (ENOENT);
3511 
3512 	if (lastref && ifp->if_ioctl != NULL) {
3513 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3514 	}
3515 
3516 	return (0);
3517 }
3518 
3519 /*
3520  * Delete all multicast group membership for an interface.
3521  * Should be used to quickly flush all multicast filters.
3522  */
3523 void
3524 if_delallmulti(struct ifnet *ifp)
3525 {
3526 	struct ifmultiaddr *ifma;
3527 	struct ifmultiaddr *next;
3528 
3529 	IF_ADDR_WLOCK(ifp);
3530 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3531 		if_delmulti_locked(ifp, ifma, 0);
3532 	IF_ADDR_WUNLOCK(ifp);
3533 }
3534 
3535 /*
3536  * Delete a multicast group membership by group membership pointer.
3537  * Network-layer protocol domains must use this routine.
3538  *
3539  * It is safe to call this routine if the ifp disappeared.
3540  */
3541 void
3542 if_delmulti_ifma(struct ifmultiaddr *ifma)
3543 {
3544 	struct ifnet *ifp;
3545 	int lastref;
3546 
3547 	ifp = ifma->ifma_ifp;
3548 #ifdef DIAGNOSTIC
3549 	if (ifp == NULL) {
3550 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3551 	} else {
3552 		struct ifnet *oifp;
3553 
3554 		IFNET_RLOCK_NOSLEEP();
3555 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3556 			if (ifp == oifp)
3557 				break;
3558 		if (ifp != oifp) {
3559 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3560 			ifp = NULL;
3561 		}
3562 		IFNET_RUNLOCK_NOSLEEP();
3563 	}
3564 #endif
3565 	/*
3566 	 * If and only if the ifnet instance exists: Acquire the address lock.
3567 	 */
3568 	if (ifp != NULL)
3569 		IF_ADDR_WLOCK(ifp);
3570 
3571 	lastref = if_delmulti_locked(ifp, ifma, 0);
3572 
3573 	if (ifp != NULL) {
3574 		/*
3575 		 * If and only if the ifnet instance exists:
3576 		 *  Release the address lock.
3577 		 *  If the group was left: update the hardware hash filter.
3578 		 */
3579 		IF_ADDR_WUNLOCK(ifp);
3580 		if (lastref && ifp->if_ioctl != NULL) {
3581 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3582 		}
3583 	}
3584 }
3585 
3586 /*
3587  * Perform deletion of network-layer and/or link-layer multicast address.
3588  *
3589  * Return 0 if the reference count was decremented.
3590  * Return 1 if the final reference was released, indicating that the
3591  * hardware hash filter should be reprogrammed.
3592  */
3593 static int
3594 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3595 {
3596 	struct ifmultiaddr *ll_ifma;
3597 
3598 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3599 		KASSERT(ifma->ifma_ifp == ifp,
3600 		    ("%s: inconsistent ifp %p", __func__, ifp));
3601 		IF_ADDR_WLOCK_ASSERT(ifp);
3602 	}
3603 
3604 	ifp = ifma->ifma_ifp;
3605 
3606 	/*
3607 	 * If the ifnet is detaching, null out references to ifnet,
3608 	 * so that upper protocol layers will notice, and not attempt
3609 	 * to obtain locks for an ifnet which no longer exists. The
3610 	 * routing socket announcement must happen before the ifnet
3611 	 * instance is detached from the system.
3612 	 */
3613 	if (detaching) {
3614 #ifdef DIAGNOSTIC
3615 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3616 #endif
3617 		/*
3618 		 * ifp may already be nulled out if we are being reentered
3619 		 * to delete the ll_ifma.
3620 		 */
3621 		if (ifp != NULL) {
3622 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3623 			ifma->ifma_ifp = NULL;
3624 		}
3625 	}
3626 
3627 	if (--ifma->ifma_refcount > 0)
3628 		return 0;
3629 
3630 	/*
3631 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3632 	 * have been associated with it. Release it first if so.
3633 	 */
3634 	ll_ifma = ifma->ifma_llifma;
3635 	if (ll_ifma != NULL) {
3636 		KASSERT(ifma->ifma_lladdr != NULL,
3637 		    ("%s: llifma w/o lladdr", __func__));
3638 		if (detaching)
3639 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3640 		if (--ll_ifma->ifma_refcount == 0) {
3641 			if (ifp != NULL) {
3642 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3643 				    ifma_link);
3644 			}
3645 			if_freemulti(ll_ifma);
3646 		}
3647 	}
3648 
3649 	if (ifp != NULL)
3650 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3651 
3652 	if_freemulti(ifma);
3653 
3654 	/*
3655 	 * The last reference to this instance of struct ifmultiaddr
3656 	 * was released; the hardware should be notified of this change.
3657 	 */
3658 	return 1;
3659 }
3660 
3661 /*
3662  * Set the link layer address on an interface.
3663  *
3664  * At this time we only support certain types of interfaces,
3665  * and we don't allow the length of the address to change.
3666  *
3667  * Set noinline to be dtrace-friendly
3668  */
3669 __noinline int
3670 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3671 {
3672 	struct sockaddr_dl *sdl;
3673 	struct ifaddr *ifa;
3674 	struct ifreq ifr;
3675 
3676 	IF_ADDR_RLOCK(ifp);
3677 	ifa = ifp->if_addr;
3678 	if (ifa == NULL) {
3679 		IF_ADDR_RUNLOCK(ifp);
3680 		return (EINVAL);
3681 	}
3682 	ifa_ref(ifa);
3683 	IF_ADDR_RUNLOCK(ifp);
3684 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3685 	if (sdl == NULL) {
3686 		ifa_free(ifa);
3687 		return (EINVAL);
3688 	}
3689 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3690 		ifa_free(ifa);
3691 		return (EINVAL);
3692 	}
3693 	switch (ifp->if_type) {
3694 	case IFT_ETHER:
3695 	case IFT_XETHER:
3696 	case IFT_L2VLAN:
3697 	case IFT_BRIDGE:
3698 	case IFT_IEEE8023ADLAG:
3699 		bcopy(lladdr, LLADDR(sdl), len);
3700 		ifa_free(ifa);
3701 		break;
3702 	default:
3703 		ifa_free(ifa);
3704 		return (ENODEV);
3705 	}
3706 
3707 	/*
3708 	 * If the interface is already up, we need
3709 	 * to re-init it in order to reprogram its
3710 	 * address filter.
3711 	 */
3712 	if ((ifp->if_flags & IFF_UP) != 0) {
3713 		if (ifp->if_ioctl) {
3714 			ifp->if_flags &= ~IFF_UP;
3715 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3716 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3717 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3718 			ifp->if_flags |= IFF_UP;
3719 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3720 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3721 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3722 		}
3723 	}
3724 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3725 	return (0);
3726 }
3727 
3728 /*
3729  * Compat function for handling basic encapsulation requests.
3730  * Not converted stacks (FDDI, IB, ..) supports traditional
3731  * output model: ARP (and other similar L2 protocols) are handled
3732  * inside output routine, arpresolve/nd6_resolve() returns MAC
3733  * address instead of full prepend.
3734  *
3735  * This function creates calculated header==MAC for IPv4/IPv6 and
3736  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3737  * address families.
3738  */
3739 static int
3740 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3741 {
3742 
3743 	if (req->rtype != IFENCAP_LL)
3744 		return (EOPNOTSUPP);
3745 
3746 	if (req->bufsize < req->lladdr_len)
3747 		return (ENOMEM);
3748 
3749 	switch (req->family) {
3750 	case AF_INET:
3751 	case AF_INET6:
3752 		break;
3753 	default:
3754 		return (EAFNOSUPPORT);
3755 	}
3756 
3757 	/* Copy lladdr to storage as is */
3758 	memmove(req->buf, req->lladdr, req->lladdr_len);
3759 	req->bufsize = req->lladdr_len;
3760 	req->lladdr_off = 0;
3761 
3762 	return (0);
3763 }
3764 
3765 /*
3766  * Get the link layer address that was read from the hardware at attach.
3767  *
3768  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3769  * their component interfaces as IFT_IEEE8023ADLAG.
3770  */
3771 int
3772 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3773 {
3774 
3775 	if (ifp->if_hw_addr == NULL)
3776 		return (ENODEV);
3777 
3778 	switch (ifp->if_type) {
3779 	case IFT_ETHER:
3780 	case IFT_IEEE8023ADLAG:
3781 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3782 		return (0);
3783 	default:
3784 		return (ENODEV);
3785 	}
3786 }
3787 
3788 /*
3789  * The name argument must be a pointer to storage which will last as
3790  * long as the interface does.  For physical devices, the result of
3791  * device_get_name(dev) is a good choice and for pseudo-devices a
3792  * static string works well.
3793  */
3794 void
3795 if_initname(struct ifnet *ifp, const char *name, int unit)
3796 {
3797 	ifp->if_dname = name;
3798 	ifp->if_dunit = unit;
3799 	if (unit != IF_DUNIT_NONE)
3800 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3801 	else
3802 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3803 }
3804 
3805 int
3806 if_printf(struct ifnet *ifp, const char * fmt, ...)
3807 {
3808 	va_list ap;
3809 	int retval;
3810 
3811 	retval = printf("%s: ", ifp->if_xname);
3812 	va_start(ap, fmt);
3813 	retval += vprintf(fmt, ap);
3814 	va_end(ap);
3815 	return (retval);
3816 }
3817 
3818 void
3819 if_start(struct ifnet *ifp)
3820 {
3821 
3822 	(*(ifp)->if_start)(ifp);
3823 }
3824 
3825 /*
3826  * Backwards compatibility interface for drivers
3827  * that have not implemented it
3828  */
3829 static int
3830 if_transmit(struct ifnet *ifp, struct mbuf *m)
3831 {
3832 	int error;
3833 
3834 	IFQ_HANDOFF(ifp, m, error);
3835 	return (error);
3836 }
3837 
3838 static void
3839 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3840 {
3841 
3842 	m_freem(m);
3843 }
3844 
3845 int
3846 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3847 {
3848 	int active = 0;
3849 
3850 	IF_LOCK(ifq);
3851 	if (_IF_QFULL(ifq)) {
3852 		IF_UNLOCK(ifq);
3853 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3854 		m_freem(m);
3855 		return (0);
3856 	}
3857 	if (ifp != NULL) {
3858 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3859 		if (m->m_flags & (M_BCAST|M_MCAST))
3860 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3861 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3862 	}
3863 	_IF_ENQUEUE(ifq, m);
3864 	IF_UNLOCK(ifq);
3865 	if (ifp != NULL && !active)
3866 		(*(ifp)->if_start)(ifp);
3867 	return (1);
3868 }
3869 
3870 void
3871 if_register_com_alloc(u_char type,
3872     if_com_alloc_t *a, if_com_free_t *f)
3873 {
3874 
3875 	KASSERT(if_com_alloc[type] == NULL,
3876 	    ("if_register_com_alloc: %d already registered", type));
3877 	KASSERT(if_com_free[type] == NULL,
3878 	    ("if_register_com_alloc: %d free already registered", type));
3879 
3880 	if_com_alloc[type] = a;
3881 	if_com_free[type] = f;
3882 }
3883 
3884 void
3885 if_deregister_com_alloc(u_char type)
3886 {
3887 
3888 	KASSERT(if_com_alloc[type] != NULL,
3889 	    ("if_deregister_com_alloc: %d not registered", type));
3890 	KASSERT(if_com_free[type] != NULL,
3891 	    ("if_deregister_com_alloc: %d free not registered", type));
3892 	if_com_alloc[type] = NULL;
3893 	if_com_free[type] = NULL;
3894 }
3895 
3896 /* API for driver access to network stack owned ifnet.*/
3897 uint64_t
3898 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3899 {
3900 	uint64_t oldbrate;
3901 
3902 	oldbrate = ifp->if_baudrate;
3903 	ifp->if_baudrate = baudrate;
3904 	return (oldbrate);
3905 }
3906 
3907 uint64_t
3908 if_getbaudrate(if_t ifp)
3909 {
3910 
3911 	return (((struct ifnet *)ifp)->if_baudrate);
3912 }
3913 
3914 int
3915 if_setcapabilities(if_t ifp, int capabilities)
3916 {
3917 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3918 	return (0);
3919 }
3920 
3921 int
3922 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3923 {
3924 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3925 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3926 
3927 	return (0);
3928 }
3929 
3930 int
3931 if_getcapabilities(if_t ifp)
3932 {
3933 	return ((struct ifnet *)ifp)->if_capabilities;
3934 }
3935 
3936 int
3937 if_setcapenable(if_t ifp, int capabilities)
3938 {
3939 	((struct ifnet *)ifp)->if_capenable = capabilities;
3940 	return (0);
3941 }
3942 
3943 int
3944 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3945 {
3946 	if(setcap)
3947 		((struct ifnet *)ifp)->if_capenable |= setcap;
3948 	if(clearcap)
3949 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3950 
3951 	return (0);
3952 }
3953 
3954 const char *
3955 if_getdname(if_t ifp)
3956 {
3957 	return ((struct ifnet *)ifp)->if_dname;
3958 }
3959 
3960 int
3961 if_togglecapenable(if_t ifp, int togglecap)
3962 {
3963 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3964 	return (0);
3965 }
3966 
3967 int
3968 if_getcapenable(if_t ifp)
3969 {
3970 	return ((struct ifnet *)ifp)->if_capenable;
3971 }
3972 
3973 /*
3974  * This is largely undesirable because it ties ifnet to a device, but does
3975  * provide flexiblity for an embedded product vendor. Should be used with
3976  * the understanding that it violates the interface boundaries, and should be
3977  * a last resort only.
3978  */
3979 int
3980 if_setdev(if_t ifp, void *dev)
3981 {
3982 	return (0);
3983 }
3984 
3985 int
3986 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3987 {
3988 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3989 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3990 
3991 	return (0);
3992 }
3993 
3994 int
3995 if_getdrvflags(if_t ifp)
3996 {
3997 	return ((struct ifnet *)ifp)->if_drv_flags;
3998 }
3999 
4000 int
4001 if_setdrvflags(if_t ifp, int flags)
4002 {
4003 	((struct ifnet *)ifp)->if_drv_flags = flags;
4004 	return (0);
4005 }
4006 
4007 
4008 int
4009 if_setflags(if_t ifp, int flags)
4010 {
4011 	((struct ifnet *)ifp)->if_flags = flags;
4012 	return (0);
4013 }
4014 
4015 int
4016 if_setflagbits(if_t ifp, int set, int clear)
4017 {
4018 	((struct ifnet *)ifp)->if_flags |= set;
4019 	((struct ifnet *)ifp)->if_flags &= ~clear;
4020 
4021 	return (0);
4022 }
4023 
4024 int
4025 if_getflags(if_t ifp)
4026 {
4027 	return ((struct ifnet *)ifp)->if_flags;
4028 }
4029 
4030 int
4031 if_clearhwassist(if_t ifp)
4032 {
4033 	((struct ifnet *)ifp)->if_hwassist = 0;
4034 	return (0);
4035 }
4036 
4037 int
4038 if_sethwassistbits(if_t ifp, int toset, int toclear)
4039 {
4040 	((struct ifnet *)ifp)->if_hwassist |= toset;
4041 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4042 
4043 	return (0);
4044 }
4045 
4046 int
4047 if_sethwassist(if_t ifp, int hwassist_bit)
4048 {
4049 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4050 	return (0);
4051 }
4052 
4053 int
4054 if_gethwassist(if_t ifp)
4055 {
4056 	return ((struct ifnet *)ifp)->if_hwassist;
4057 }
4058 
4059 int
4060 if_setmtu(if_t ifp, int mtu)
4061 {
4062 	((struct ifnet *)ifp)->if_mtu = mtu;
4063 	return (0);
4064 }
4065 
4066 int
4067 if_getmtu(if_t ifp)
4068 {
4069 	return ((struct ifnet *)ifp)->if_mtu;
4070 }
4071 
4072 int
4073 if_getmtu_family(if_t ifp, int family)
4074 {
4075 	struct domain *dp;
4076 
4077 	for (dp = domains; dp; dp = dp->dom_next) {
4078 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4079 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4080 	}
4081 
4082 	return (((struct ifnet *)ifp)->if_mtu);
4083 }
4084 
4085 int
4086 if_setsoftc(if_t ifp, void *softc)
4087 {
4088 	((struct ifnet *)ifp)->if_softc = softc;
4089 	return (0);
4090 }
4091 
4092 void *
4093 if_getsoftc(if_t ifp)
4094 {
4095 	return ((struct ifnet *)ifp)->if_softc;
4096 }
4097 
4098 void
4099 if_setrcvif(struct mbuf *m, if_t ifp)
4100 {
4101 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4102 }
4103 
4104 void
4105 if_setvtag(struct mbuf *m, uint16_t tag)
4106 {
4107 	m->m_pkthdr.ether_vtag = tag;
4108 }
4109 
4110 uint16_t
4111 if_getvtag(struct mbuf *m)
4112 {
4113 
4114 	return (m->m_pkthdr.ether_vtag);
4115 }
4116 
4117 int
4118 if_sendq_empty(if_t ifp)
4119 {
4120 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4121 }
4122 
4123 struct ifaddr *
4124 if_getifaddr(if_t ifp)
4125 {
4126 	return ((struct ifnet *)ifp)->if_addr;
4127 }
4128 
4129 int
4130 if_getamcount(if_t ifp)
4131 {
4132 	return ((struct ifnet *)ifp)->if_amcount;
4133 }
4134 
4135 
4136 int
4137 if_setsendqready(if_t ifp)
4138 {
4139 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4140 	return (0);
4141 }
4142 
4143 int
4144 if_setsendqlen(if_t ifp, int tx_desc_count)
4145 {
4146 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4147 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4148 
4149 	return (0);
4150 }
4151 
4152 int
4153 if_vlantrunkinuse(if_t ifp)
4154 {
4155 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4156 }
4157 
4158 int
4159 if_input(if_t ifp, struct mbuf* sendmp)
4160 {
4161 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4162 	return (0);
4163 
4164 }
4165 
4166 /* XXX */
4167 #ifndef ETH_ADDR_LEN
4168 #define ETH_ADDR_LEN 6
4169 #endif
4170 
4171 int
4172 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4173 {
4174 	struct ifmultiaddr *ifma;
4175 	uint8_t *lmta = (uint8_t *)mta;
4176 	int mcnt = 0;
4177 
4178 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4179 		if (ifma->ifma_addr->sa_family != AF_LINK)
4180 			continue;
4181 
4182 		if (mcnt == max)
4183 			break;
4184 
4185 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4186 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4187 		mcnt++;
4188 	}
4189 	*cnt = mcnt;
4190 
4191 	return (0);
4192 }
4193 
4194 int
4195 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4196 {
4197 	int error;
4198 
4199 	if_maddr_rlock(ifp);
4200 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4201 	if_maddr_runlock(ifp);
4202 	return (error);
4203 }
4204 
4205 int
4206 if_multiaddr_count(if_t ifp, int max)
4207 {
4208 	struct ifmultiaddr *ifma;
4209 	int count;
4210 
4211 	count = 0;
4212 	if_maddr_rlock(ifp);
4213 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4214 		if (ifma->ifma_addr->sa_family != AF_LINK)
4215 			continue;
4216 		count++;
4217 		if (count == max)
4218 			break;
4219 	}
4220 	if_maddr_runlock(ifp);
4221 	return (count);
4222 }
4223 
4224 int
4225 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4226 {
4227 	struct ifmultiaddr *ifma;
4228 	int cnt = 0;
4229 
4230 	if_maddr_rlock(ifp);
4231 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4232 		cnt += filter(arg, ifma, cnt);
4233 	if_maddr_runlock(ifp);
4234 	return (cnt);
4235 }
4236 
4237 struct mbuf *
4238 if_dequeue(if_t ifp)
4239 {
4240 	struct mbuf *m;
4241 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4242 
4243 	return (m);
4244 }
4245 
4246 int
4247 if_sendq_prepend(if_t ifp, struct mbuf *m)
4248 {
4249 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4250 	return (0);
4251 }
4252 
4253 int
4254 if_setifheaderlen(if_t ifp, int len)
4255 {
4256 	((struct ifnet *)ifp)->if_hdrlen = len;
4257 	return (0);
4258 }
4259 
4260 caddr_t
4261 if_getlladdr(if_t ifp)
4262 {
4263 	return (IF_LLADDR((struct ifnet *)ifp));
4264 }
4265 
4266 void *
4267 if_gethandle(u_char type)
4268 {
4269 	return (if_alloc(type));
4270 }
4271 
4272 void
4273 if_bpfmtap(if_t ifh, struct mbuf *m)
4274 {
4275 	struct ifnet *ifp = (struct ifnet *)ifh;
4276 
4277 	BPF_MTAP(ifp, m);
4278 }
4279 
4280 void
4281 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4282 {
4283 	struct ifnet *ifp = (struct ifnet *)ifh;
4284 
4285 	ETHER_BPF_MTAP(ifp, m);
4286 }
4287 
4288 void
4289 if_vlancap(if_t ifh)
4290 {
4291 	struct ifnet *ifp = (struct ifnet *)ifh;
4292 	VLAN_CAPABILITIES(ifp);
4293 }
4294 
4295 int
4296 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4297 {
4298 
4299 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4300         return (0);
4301 }
4302 
4303 int
4304 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4305 {
4306 
4307 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4308         return (0);
4309 }
4310 
4311 int
4312 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4313 {
4314 
4315 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4316         return (0);
4317 }
4318 
4319 u_int
4320 if_gethwtsomax(if_t ifp)
4321 {
4322 
4323 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4324 }
4325 
4326 u_int
4327 if_gethwtsomaxsegcount(if_t ifp)
4328 {
4329 
4330 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4331 }
4332 
4333 u_int
4334 if_gethwtsomaxsegsize(if_t ifp)
4335 {
4336 
4337 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4338 }
4339 
4340 void
4341 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4342 {
4343 	((struct ifnet *)ifp)->if_init = init_fn;
4344 }
4345 
4346 void
4347 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4348 {
4349 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4350 }
4351 
4352 void
4353 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4354 {
4355 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4356 }
4357 
4358 void
4359 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4360 {
4361 	((struct ifnet *)ifp)->if_transmit = start_fn;
4362 }
4363 
4364 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4365 {
4366 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4367 
4368 }
4369 
4370 void
4371 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4372 {
4373 
4374 	ifp->if_get_counter = fn;
4375 }
4376 
4377 /* Revisit these - These are inline functions originally. */
4378 int
4379 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4380 {
4381 	return drbr_inuse(ifh, br);
4382 }
4383 
4384 struct mbuf*
4385 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4386 {
4387 	return drbr_dequeue(ifh, br);
4388 }
4389 
4390 int
4391 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4392 {
4393 	return drbr_needs_enqueue(ifh, br);
4394 }
4395 
4396 int
4397 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4398 {
4399 	return drbr_enqueue(ifh, br, m);
4400 
4401 }
4402