xref: /freebsd/sys/net/if.c (revision 4928135658a9d0eaee37003df6137ab363fcb0b4)
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/epoch.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysent.h>
61 #include <sys/taskqueue.h>
62 #include <sys/domain.h>
63 #include <sys/jail.h>
64 #include <sys/priv.h>
65 
66 #include <machine/stdarg.h>
67 #include <vm/uma.h>
68 
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if.h>
72 #include <net/if_arp.h>
73 #include <net/if_clone.h>
74 #include <net/if_dl.h>
75 #include <net/if_types.h>
76 #include <net/if_var.h>
77 #include <net/if_media.h>
78 #include <net/if_vlan_var.h>
79 #include <net/radix.h>
80 #include <net/route.h>
81 #include <net/vnet.h>
82 
83 #if defined(INET) || defined(INET6)
84 #include <net/ethernet.h>
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip_carp.h>
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #include <netinet/netdump/netdump.h>
92 #endif /* INET */
93 #ifdef INET6
94 #include <netinet6/in6_var.h>
95 #include <netinet6/in6_ifattach.h>
96 #endif /* INET6 */
97 #endif /* INET || INET6 */
98 
99 #include <security/mac/mac_framework.h>
100 
101 /*
102  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
103  * and ifr_ifru when it is used in SIOCGIFCONF.
104  */
105 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
106     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
107 
108 __read_mostly epoch_t net_epoch_preempt;
109 __read_mostly epoch_t net_epoch;
110 #ifdef COMPAT_FREEBSD32
111 #include <sys/mount.h>
112 #include <compat/freebsd32/freebsd32.h>
113 
114 struct ifreq_buffer32 {
115 	uint32_t	length;		/* (size_t) */
116 	uint32_t	buffer;		/* (void *) */
117 };
118 
119 /*
120  * Interface request structure used for socket
121  * ioctl's.  All interface ioctl's must have parameter
122  * definitions which begin with ifr_name.  The
123  * remainder may be interface specific.
124  */
125 struct ifreq32 {
126 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
127 	union {
128 		struct sockaddr	ifru_addr;
129 		struct sockaddr	ifru_dstaddr;
130 		struct sockaddr	ifru_broadaddr;
131 		struct ifreq_buffer32 ifru_buffer;
132 		short		ifru_flags[2];
133 		short		ifru_index;
134 		int		ifru_jid;
135 		int		ifru_metric;
136 		int		ifru_mtu;
137 		int		ifru_phys;
138 		int		ifru_media;
139 		uint32_t	ifru_data;
140 		int		ifru_cap[2];
141 		u_int		ifru_fib;
142 		u_char		ifru_vlan_pcp;
143 	} ifr_ifru;
144 };
145 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
146 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
147     __offsetof(struct ifreq32, ifr_ifru));
148 
149 struct ifgroupreq32 {
150 	char	ifgr_name[IFNAMSIZ];
151 	u_int	ifgr_len;
152 	union {
153 		char		ifgru_group[IFNAMSIZ];
154 		uint32_t	ifgru_groups;
155 	} ifgr_ifgru;
156 };
157 
158 struct ifmediareq32 {
159 	char		ifm_name[IFNAMSIZ];
160 	int		ifm_current;
161 	int		ifm_mask;
162 	int		ifm_status;
163 	int		ifm_active;
164 	int		ifm_count;
165 	uint32_t	ifm_ulist;	/* (int *) */
166 };
167 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
168 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
169 
170 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
171     case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
172 #else /* !COMPAT_FREEBSD32 */
173 #define _CASE_IOC_IFGROUPREQ_32(cmd)
174 #endif /* !COMPAT_FREEBSD32 */
175 
176 #define CASE_IOC_IFGROUPREQ(cmd)	\
177     _CASE_IOC_IFGROUPREQ_32(cmd)	\
178     case (cmd)
179 
180 union ifreq_union {
181 	struct ifreq	ifr;
182 #ifdef COMPAT_FREEBSD32
183 	struct ifreq32	ifr32;
184 #endif
185 };
186 
187 union ifgroupreq_union {
188 	struct ifgroupreq ifgr;
189 #ifdef COMPAT_FREEBSD32
190 	struct ifgroupreq32 ifgr32;
191 #endif
192 };
193 
194 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
195 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
196 
197 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
198     &ifqmaxlen, 0, "max send queue size");
199 
200 /* Log link state change events */
201 static int log_link_state_change = 1;
202 
203 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
204 	&log_link_state_change, 0,
205 	"log interface link state change events");
206 
207 /* Log promiscuous mode change events */
208 static int log_promisc_mode_change = 1;
209 
210 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
211 	&log_promisc_mode_change, 1,
212 	"log promiscuous mode change events");
213 
214 /* Interface description */
215 static unsigned int ifdescr_maxlen = 1024;
216 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
217 	&ifdescr_maxlen, 0,
218 	"administrative maximum length for interface description");
219 
220 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
221 
222 /* global sx for non-critical path ifdescr */
223 static struct sx ifdescr_sx;
224 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
225 
226 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
227 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
228 /* These are external hooks for CARP. */
229 void	(*carp_linkstate_p)(struct ifnet *ifp);
230 void	(*carp_demote_adj_p)(int, char *);
231 int	(*carp_master_p)(struct ifaddr *);
232 #if defined(INET) || defined(INET6)
233 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
234 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
235     const struct sockaddr *sa);
236 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
237 int	(*carp_attach_p)(struct ifaddr *, int);
238 void	(*carp_detach_p)(struct ifaddr *, bool);
239 #endif
240 #ifdef INET
241 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
242 #endif
243 #ifdef INET6
244 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
245 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
246     const struct in6_addr *taddr);
247 #endif
248 
249 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
250 
251 /*
252  * XXX: Style; these should be sorted alphabetically, and unprototyped
253  * static functions should be prototyped. Currently they are sorted by
254  * declaration order.
255  */
256 static void	if_attachdomain(void *);
257 static void	if_attachdomain1(struct ifnet *);
258 static int	ifconf(u_long, caddr_t);
259 static void	*if_grow(void);
260 static void	if_input_default(struct ifnet *, struct mbuf *);
261 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
262 static void	if_route(struct ifnet *, int flag, int fam);
263 static int	if_setflag(struct ifnet *, int, int, int *, int);
264 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
265 static void	if_unroute(struct ifnet *, int flag, int fam);
266 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
267 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
268 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
269 static void	do_link_state_change(void *, int);
270 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
271 static int	if_getgroupmembers(struct ifgroupreq *);
272 static void	if_delgroups(struct ifnet *);
273 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
274 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
275 #ifdef VIMAGE
276 static void	if_vmove(struct ifnet *, struct vnet *);
277 #endif
278 
279 #ifdef INET6
280 /*
281  * XXX: declare here to avoid to include many inet6 related files..
282  * should be more generalized?
283  */
284 extern void	nd6_setmtu(struct ifnet *);
285 #endif
286 
287 /* ipsec helper hooks */
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
289 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
290 
291 VNET_DEFINE(int, if_index);
292 int	ifqmaxlen = IFQ_MAXLEN;
293 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
294 VNET_DEFINE(struct ifgrouphead, ifg_head);
295 
296 static VNET_DEFINE(int, if_indexlim) = 8;
297 
298 /* Table of ifnet by index. */
299 VNET_DEFINE(struct ifnet **, ifindex_table);
300 
301 #define	V_if_indexlim		VNET(if_indexlim)
302 #define	V_ifindex_table		VNET(ifindex_table)
303 
304 /*
305  * The global network interface list (V_ifnet) and related state (such as
306  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
307  * an rwlock.  Either may be acquired shared to stablize the list, but both
308  * must be acquired writable to modify the list.  This model allows us to
309  * both stablize the interface list during interrupt thread processing, but
310  * also to stablize it over long-running ioctls, without introducing priority
311  * inversions and deadlocks.
312  */
313 struct rwlock ifnet_rwlock;
314 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
315 struct sx ifnet_sxlock;
316 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
317 
318 /*
319  * The allocation of network interfaces is a rather non-atomic affair; we
320  * need to select an index before we are ready to expose the interface for
321  * use, so will use this pointer value to indicate reservation.
322  */
323 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
324 
325 static	if_com_alloc_t *if_com_alloc[256];
326 static	if_com_free_t *if_com_free[256];
327 
328 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
329 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
330 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
331 
332 struct ifnet *
333 ifnet_byindex_locked(u_short idx)
334 {
335 
336 	if (idx > V_if_index)
337 		return (NULL);
338 	if (V_ifindex_table[idx] == IFNET_HOLD)
339 		return (NULL);
340 	return (V_ifindex_table[idx]);
341 }
342 
343 struct ifnet *
344 ifnet_byindex(u_short idx)
345 {
346 	struct ifnet *ifp;
347 
348 	ifp = ifnet_byindex_locked(idx);
349 	return (ifp);
350 }
351 
352 struct ifnet *
353 ifnet_byindex_ref(u_short idx)
354 {
355 	struct ifnet *ifp;
356 
357 	IFNET_RLOCK_NOSLEEP();
358 	ifp = ifnet_byindex_locked(idx);
359 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
360 		IFNET_RUNLOCK_NOSLEEP();
361 		return (NULL);
362 	}
363 	if_ref(ifp);
364 	IFNET_RUNLOCK_NOSLEEP();
365 	return (ifp);
366 }
367 
368 /*
369  * Allocate an ifindex array entry; return 0 on success or an error on
370  * failure.
371  */
372 static u_short
373 ifindex_alloc(void **old)
374 {
375 	u_short idx;
376 
377 	IFNET_WLOCK_ASSERT();
378 	/*
379 	 * Try to find an empty slot below V_if_index.  If we fail, take the
380 	 * next slot.
381 	 */
382 	for (idx = 1; idx <= V_if_index; idx++) {
383 		if (V_ifindex_table[idx] == NULL)
384 			break;
385 	}
386 
387 	/* Catch if_index overflow. */
388 	if (idx >= V_if_indexlim) {
389 		*old = if_grow();
390 		return (USHRT_MAX);
391 	}
392 	if (idx > V_if_index)
393 		V_if_index = idx;
394 	return (idx);
395 }
396 
397 static void
398 ifindex_free_locked(u_short idx)
399 {
400 
401 	IFNET_WLOCK_ASSERT();
402 
403 	V_ifindex_table[idx] = NULL;
404 	while (V_if_index > 0 &&
405 	    V_ifindex_table[V_if_index] == NULL)
406 		V_if_index--;
407 }
408 
409 static void
410 ifindex_free(u_short idx)
411 {
412 
413 	IFNET_WLOCK();
414 	ifindex_free_locked(idx);
415 	IFNET_WUNLOCK();
416 }
417 
418 static void
419 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
420 {
421 
422 	V_ifindex_table[idx] = ifp;
423 }
424 
425 struct ifaddr *
426 ifaddr_byindex(u_short idx)
427 {
428 	struct ifnet *ifp;
429 	struct ifaddr *ifa = NULL;
430 
431 	IFNET_RLOCK_NOSLEEP();
432 	ifp = ifnet_byindex_locked(idx);
433 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
434 		ifa_ref(ifa);
435 	IFNET_RUNLOCK_NOSLEEP();
436 	return (ifa);
437 }
438 
439 /*
440  * Network interface utility routines.
441  *
442  * Routines with ifa_ifwith* names take sockaddr *'s as
443  * parameters.
444  */
445 
446 static void
447 vnet_if_init(const void *unused __unused)
448 {
449 	void *old;
450 
451 	CK_STAILQ_INIT(&V_ifnet);
452 	CK_STAILQ_INIT(&V_ifg_head);
453 	IFNET_WLOCK();
454 	old = if_grow();				/* create initial table */
455 	IFNET_WUNLOCK();
456 	epoch_wait_preempt(net_epoch_preempt);
457 	free(old, M_IFNET);
458 	vnet_if_clone_init();
459 }
460 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
461     NULL);
462 
463 #ifdef VIMAGE
464 static void
465 vnet_if_uninit(const void *unused __unused)
466 {
467 
468 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
469 	    "not empty", __func__, __LINE__, &V_ifnet));
470 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
471 	    "not empty", __func__, __LINE__, &V_ifg_head));
472 
473 	free((caddr_t)V_ifindex_table, M_IFNET);
474 }
475 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
476     vnet_if_uninit, NULL);
477 
478 static void
479 vnet_if_return(const void *unused __unused)
480 {
481 	struct ifnet *ifp, *nifp;
482 
483 	/* Return all inherited interfaces to their parent vnets. */
484 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
485 		if (ifp->if_home_vnet != ifp->if_vnet)
486 			if_vmove(ifp, ifp->if_home_vnet);
487 	}
488 }
489 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
490     vnet_if_return, NULL);
491 #endif
492 
493 
494 static void *
495 if_grow(void)
496 {
497 	int oldlim;
498 	u_int n;
499 	struct ifnet **e;
500 	void *old;
501 
502 	old = NULL;
503 	IFNET_WLOCK_ASSERT();
504 	oldlim = V_if_indexlim;
505 	IFNET_WUNLOCK();
506 	n = (oldlim << 1) * sizeof(*e);
507 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
508 	IFNET_WLOCK();
509 	if (V_if_indexlim != oldlim) {
510 		free(e, M_IFNET);
511 		return (NULL);
512 	}
513 	if (V_ifindex_table != NULL) {
514 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
515 		old = V_ifindex_table;
516 	}
517 	V_if_indexlim <<= 1;
518 	V_ifindex_table = e;
519 	return (old);
520 }
521 
522 /*
523  * Allocate a struct ifnet and an index for an interface.  A layer 2
524  * common structure will also be allocated if an allocation routine is
525  * registered for the passed type.
526  */
527 struct ifnet *
528 if_alloc(u_char type)
529 {
530 	struct ifnet *ifp;
531 	u_short idx;
532 	void *old;
533 
534 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
535  restart:
536 	IFNET_WLOCK();
537 	idx = ifindex_alloc(&old);
538 	if (__predict_false(idx == USHRT_MAX)) {
539 		IFNET_WUNLOCK();
540 		epoch_wait_preempt(net_epoch_preempt);
541 		free(old, M_IFNET);
542 		goto restart;
543 	}
544 	ifnet_setbyindex(idx, IFNET_HOLD);
545 	IFNET_WUNLOCK();
546 	ifp->if_index = idx;
547 	ifp->if_type = type;
548 	ifp->if_alloctype = type;
549 #ifdef VIMAGE
550 	ifp->if_vnet = curvnet;
551 #endif
552 	if (if_com_alloc[type] != NULL) {
553 		ifp->if_l2com = if_com_alloc[type](type, ifp);
554 		if (ifp->if_l2com == NULL) {
555 			free(ifp, M_IFNET);
556 			ifindex_free(idx);
557 			return (NULL);
558 		}
559 	}
560 
561 	IF_ADDR_LOCK_INIT(ifp);
562 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
563 	ifp->if_afdata_initialized = 0;
564 	IF_AFDATA_LOCK_INIT(ifp);
565 	CK_STAILQ_INIT(&ifp->if_addrhead);
566 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
567 	CK_STAILQ_INIT(&ifp->if_groups);
568 #ifdef MAC
569 	mac_ifnet_init(ifp);
570 #endif
571 	ifq_init(&ifp->if_snd, ifp);
572 
573 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
574 	for (int i = 0; i < IFCOUNTERS; i++)
575 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
576 	ifp->if_get_counter = if_get_counter_default;
577 	ifp->if_pcp = IFNET_PCP_NONE;
578 	ifnet_setbyindex(ifp->if_index, ifp);
579 	return (ifp);
580 }
581 
582 /*
583  * Do the actual work of freeing a struct ifnet, and layer 2 common
584  * structure.  This call is made when the last reference to an
585  * interface is released.
586  */
587 static void
588 if_free_internal(struct ifnet *ifp)
589 {
590 
591 	KASSERT((ifp->if_flags & IFF_DYING),
592 	    ("if_free_internal: interface not dying"));
593 
594 	if (if_com_free[ifp->if_alloctype] != NULL)
595 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
596 		    ifp->if_alloctype);
597 
598 #ifdef MAC
599 	mac_ifnet_destroy(ifp);
600 #endif /* MAC */
601 	if (ifp->if_description != NULL)
602 		free(ifp->if_description, M_IFDESCR);
603 	IF_AFDATA_DESTROY(ifp);
604 	IF_ADDR_LOCK_DESTROY(ifp);
605 	ifq_delete(&ifp->if_snd);
606 
607 	for (int i = 0; i < IFCOUNTERS; i++)
608 		counter_u64_free(ifp->if_counters[i]);
609 
610 	free(ifp, M_IFNET);
611 }
612 
613 static void
614 if_destroy(epoch_context_t ctx)
615 {
616 	struct ifnet *ifp;
617 
618 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
619 	if_free_internal(ifp);
620 }
621 
622 /*
623  * Deregister an interface and free the associated storage.
624  */
625 void
626 if_free(struct ifnet *ifp)
627 {
628 
629 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
630 
631 	CURVNET_SET_QUIET(ifp->if_vnet);
632 	IFNET_WLOCK();
633 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
634 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
635 
636 	ifindex_free_locked(ifp->if_index);
637 	IFNET_WUNLOCK();
638 
639 	if (refcount_release(&ifp->if_refcount))
640 		epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
641 	CURVNET_RESTORE();
642 }
643 
644 /*
645  * Interfaces to keep an ifnet type-stable despite the possibility of the
646  * driver calling if_free().  If there are additional references, we defer
647  * freeing the underlying data structure.
648  */
649 void
650 if_ref(struct ifnet *ifp)
651 {
652 
653 	/* We don't assert the ifnet list lock here, but arguably should. */
654 	refcount_acquire(&ifp->if_refcount);
655 }
656 
657 void
658 if_rele(struct ifnet *ifp)
659 {
660 
661 	if (!refcount_release(&ifp->if_refcount))
662 		return;
663 	epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
664 }
665 
666 void
667 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
668 {
669 
670 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
671 
672 	if (ifq->ifq_maxlen == 0)
673 		ifq->ifq_maxlen = ifqmaxlen;
674 
675 	ifq->altq_type = 0;
676 	ifq->altq_disc = NULL;
677 	ifq->altq_flags &= ALTQF_CANTCHANGE;
678 	ifq->altq_tbr  = NULL;
679 	ifq->altq_ifp  = ifp;
680 }
681 
682 void
683 ifq_delete(struct ifaltq *ifq)
684 {
685 	mtx_destroy(&ifq->ifq_mtx);
686 }
687 
688 /*
689  * Perform generic interface initialization tasks and attach the interface
690  * to the list of "active" interfaces.  If vmove flag is set on entry
691  * to if_attach_internal(), perform only a limited subset of initialization
692  * tasks, given that we are moving from one vnet to another an ifnet which
693  * has already been fully initialized.
694  *
695  * Note that if_detach_internal() removes group membership unconditionally
696  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
697  * Thus, when if_vmove() is applied to a cloned interface, group membership
698  * is lost while a cloned one always joins a group whose name is
699  * ifc->ifc_name.  To recover this after if_detach_internal() and
700  * if_attach_internal(), the cloner should be specified to
701  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
702  * attempts to join a group whose name is ifc->ifc_name.
703  *
704  * XXX:
705  *  - The decision to return void and thus require this function to
706  *    succeed is questionable.
707  *  - We should probably do more sanity checking.  For instance we don't
708  *    do anything to insure if_xname is unique or non-empty.
709  */
710 void
711 if_attach(struct ifnet *ifp)
712 {
713 
714 	if_attach_internal(ifp, 0, NULL);
715 }
716 
717 /*
718  * Compute the least common TSO limit.
719  */
720 void
721 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
722 {
723 	/*
724 	 * 1) If there is no limit currently, take the limit from
725 	 * the network adapter.
726 	 *
727 	 * 2) If the network adapter has a limit below the current
728 	 * limit, apply it.
729 	 */
730 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
731 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
732 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
733 	}
734 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
735 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
736 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
737 	}
738 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
739 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
740 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
741 	}
742 }
743 
744 /*
745  * Update TSO limit of a network adapter.
746  *
747  * Returns zero if no change. Else non-zero.
748  */
749 int
750 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
751 {
752 	int retval = 0;
753 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
754 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
755 		retval++;
756 	}
757 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
758 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
759 		retval++;
760 	}
761 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
762 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
763 		retval++;
764 	}
765 	return (retval);
766 }
767 
768 static void
769 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
770 {
771 	unsigned socksize, ifasize;
772 	int namelen, masklen;
773 	struct sockaddr_dl *sdl;
774 	struct ifaddr *ifa;
775 
776 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
777 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
778 		    ifp->if_xname);
779 
780 #ifdef VIMAGE
781 	ifp->if_vnet = curvnet;
782 	if (ifp->if_home_vnet == NULL)
783 		ifp->if_home_vnet = curvnet;
784 #endif
785 
786 	if_addgroup(ifp, IFG_ALL);
787 
788 	/* Restore group membership for cloned interfaces. */
789 	if (vmove && ifc != NULL)
790 		if_clone_addgroup(ifp, ifc);
791 
792 	getmicrotime(&ifp->if_lastchange);
793 	ifp->if_epoch = time_uptime;
794 
795 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
796 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
797 	    ("transmit and qflush must both either be set or both be NULL"));
798 	if (ifp->if_transmit == NULL) {
799 		ifp->if_transmit = if_transmit;
800 		ifp->if_qflush = if_qflush;
801 	}
802 	if (ifp->if_input == NULL)
803 		ifp->if_input = if_input_default;
804 
805 	if (ifp->if_requestencap == NULL)
806 		ifp->if_requestencap = if_requestencap_default;
807 
808 	if (!vmove) {
809 #ifdef MAC
810 		mac_ifnet_create(ifp);
811 #endif
812 
813 		/*
814 		 * Create a Link Level name for this device.
815 		 */
816 		namelen = strlen(ifp->if_xname);
817 		/*
818 		 * Always save enough space for any possiable name so we
819 		 * can do a rename in place later.
820 		 */
821 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
822 		socksize = masklen + ifp->if_addrlen;
823 		if (socksize < sizeof(*sdl))
824 			socksize = sizeof(*sdl);
825 		socksize = roundup2(socksize, sizeof(long));
826 		ifasize = sizeof(*ifa) + 2 * socksize;
827 		ifa = ifa_alloc(ifasize, M_WAITOK);
828 		sdl = (struct sockaddr_dl *)(ifa + 1);
829 		sdl->sdl_len = socksize;
830 		sdl->sdl_family = AF_LINK;
831 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
832 		sdl->sdl_nlen = namelen;
833 		sdl->sdl_index = ifp->if_index;
834 		sdl->sdl_type = ifp->if_type;
835 		ifp->if_addr = ifa;
836 		ifa->ifa_ifp = ifp;
837 		ifa->ifa_rtrequest = link_rtrequest;
838 		ifa->ifa_addr = (struct sockaddr *)sdl;
839 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
840 		ifa->ifa_netmask = (struct sockaddr *)sdl;
841 		sdl->sdl_len = masklen;
842 		while (namelen != 0)
843 			sdl->sdl_data[--namelen] = 0xff;
844 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
845 		/* Reliably crash if used uninitialized. */
846 		ifp->if_broadcastaddr = NULL;
847 
848 		if (ifp->if_type == IFT_ETHER) {
849 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
850 			    M_WAITOK | M_ZERO);
851 		}
852 
853 #if defined(INET) || defined(INET6)
854 		/* Use defaults for TSO, if nothing is set */
855 		if (ifp->if_hw_tsomax == 0 &&
856 		    ifp->if_hw_tsomaxsegcount == 0 &&
857 		    ifp->if_hw_tsomaxsegsize == 0) {
858 			/*
859 			 * The TSO defaults needs to be such that an
860 			 * NFS mbuf list of 35 mbufs totalling just
861 			 * below 64K works and that a chain of mbufs
862 			 * can be defragged into at most 32 segments:
863 			 */
864 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
865 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
866 			ifp->if_hw_tsomaxsegcount = 35;
867 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
868 
869 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
870 			if (ifp->if_capabilities & IFCAP_TSO) {
871 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
872 				    ifp->if_hw_tsomax,
873 				    ifp->if_hw_tsomaxsegcount,
874 				    ifp->if_hw_tsomaxsegsize);
875 			}
876 		}
877 #endif
878 	}
879 #ifdef VIMAGE
880 	else {
881 		/*
882 		 * Update the interface index in the link layer address
883 		 * of the interface.
884 		 */
885 		for (ifa = ifp->if_addr; ifa != NULL;
886 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
887 			if (ifa->ifa_addr->sa_family == AF_LINK) {
888 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
889 				sdl->sdl_index = ifp->if_index;
890 			}
891 		}
892 	}
893 #endif
894 
895 	IFNET_WLOCK();
896 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
897 #ifdef VIMAGE
898 	curvnet->vnet_ifcnt++;
899 #endif
900 	IFNET_WUNLOCK();
901 
902 	if (domain_init_status >= 2)
903 		if_attachdomain1(ifp);
904 
905 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
906 	if (IS_DEFAULT_VNET(curvnet))
907 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
908 
909 	/* Announce the interface. */
910 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
911 }
912 
913 static void
914 if_epochalloc(void *dummy __unused)
915 {
916 
917 	net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT);
918 	net_epoch = epoch_alloc(0);
919 }
920 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
921     if_epochalloc, NULL);
922 
923 static void
924 if_attachdomain(void *dummy)
925 {
926 	struct ifnet *ifp;
927 
928 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
929 		if_attachdomain1(ifp);
930 }
931 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
932     if_attachdomain, NULL);
933 
934 static void
935 if_attachdomain1(struct ifnet *ifp)
936 {
937 	struct domain *dp;
938 
939 	/*
940 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
941 	 * cannot lock ifp->if_afdata initialization, entirely.
942 	 */
943 	IF_AFDATA_LOCK(ifp);
944 	if (ifp->if_afdata_initialized >= domain_init_status) {
945 		IF_AFDATA_UNLOCK(ifp);
946 		log(LOG_WARNING, "%s called more than once on %s\n",
947 		    __func__, ifp->if_xname);
948 		return;
949 	}
950 	ifp->if_afdata_initialized = domain_init_status;
951 	IF_AFDATA_UNLOCK(ifp);
952 
953 	/* address family dependent data region */
954 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
955 	for (dp = domains; dp; dp = dp->dom_next) {
956 		if (dp->dom_ifattach)
957 			ifp->if_afdata[dp->dom_family] =
958 			    (*dp->dom_ifattach)(ifp);
959 	}
960 }
961 
962 /*
963  * Remove any unicast or broadcast network addresses from an interface.
964  */
965 void
966 if_purgeaddrs(struct ifnet *ifp)
967 {
968 	struct ifaddr *ifa, *next;
969 
970 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
971 	CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
972 		if (ifa->ifa_addr->sa_family == AF_LINK)
973 			continue;
974 #ifdef INET
975 		/* XXX: Ugly!! ad hoc just for INET */
976 		if (ifa->ifa_addr->sa_family == AF_INET) {
977 			struct ifaliasreq ifr;
978 
979 			bzero(&ifr, sizeof(ifr));
980 			ifr.ifra_addr = *ifa->ifa_addr;
981 			if (ifa->ifa_dstaddr)
982 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
983 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
984 			    NULL) == 0)
985 				continue;
986 		}
987 #endif /* INET */
988 #ifdef INET6
989 		if (ifa->ifa_addr->sa_family == AF_INET6) {
990 			in6_purgeaddr(ifa);
991 			/* ifp_addrhead is already updated */
992 			continue;
993 		}
994 #endif /* INET6 */
995 		IF_ADDR_WLOCK(ifp);
996 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
997 		IF_ADDR_WUNLOCK(ifp);
998 		ifa_free(ifa);
999 	}
1000 }
1001 
1002 /*
1003  * Remove any multicast network addresses from an interface when an ifnet
1004  * is going away.
1005  */
1006 static void
1007 if_purgemaddrs(struct ifnet *ifp)
1008 {
1009 	struct ifmultiaddr *ifma;
1010 
1011 	IF_ADDR_WLOCK(ifp);
1012 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1013 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1014 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1015 		if_delmulti_locked(ifp, ifma, 1);
1016 	}
1017 	IF_ADDR_WUNLOCK(ifp);
1018 }
1019 
1020 /*
1021  * Detach an interface, removing it from the list of "active" interfaces.
1022  * If vmove flag is set on entry to if_detach_internal(), perform only a
1023  * limited subset of cleanup tasks, given that we are moving an ifnet from
1024  * one vnet to another, where it must be fully operational.
1025  *
1026  * XXXRW: There are some significant questions about event ordering, and
1027  * how to prevent things from starting to use the interface during detach.
1028  */
1029 void
1030 if_detach(struct ifnet *ifp)
1031 {
1032 
1033 	CURVNET_SET_QUIET(ifp->if_vnet);
1034 	if_detach_internal(ifp, 0, NULL);
1035 	CURVNET_RESTORE();
1036 }
1037 
1038 /*
1039  * The vmove flag, if set, indicates that we are called from a callpath
1040  * that is moving an interface to a different vnet instance.
1041  *
1042  * The shutdown flag, if set, indicates that we are called in the
1043  * process of shutting down a vnet instance.  Currently only the
1044  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1045  * on a vnet instance shutdown without this flag being set, e.g., when
1046  * the cloned interfaces are destoyed as first thing of teardown.
1047  */
1048 static int
1049 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1050 {
1051 	struct ifaddr *ifa;
1052 	int i;
1053 	struct domain *dp;
1054  	struct ifnet *iter;
1055  	int found = 0;
1056 #ifdef VIMAGE
1057 	int shutdown;
1058 
1059 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1060 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1061 #endif
1062 	IFNET_WLOCK();
1063 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1064 		if (iter == ifp) {
1065 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1066 			found = 1;
1067 			break;
1068 		}
1069 	IFNET_WUNLOCK();
1070 	if (!found) {
1071 		/*
1072 		 * While we would want to panic here, we cannot
1073 		 * guarantee that the interface is indeed still on
1074 		 * the list given we don't hold locks all the way.
1075 		 */
1076 		return (ENOENT);
1077 #if 0
1078 		if (vmove)
1079 			panic("%s: ifp=%p not on the ifnet tailq %p",
1080 			    __func__, ifp, &V_ifnet);
1081 		else
1082 			return; /* XXX this should panic as well? */
1083 #endif
1084 	}
1085 
1086 	/*
1087 	 * At this point we know the interface still was on the ifnet list
1088 	 * and we removed it so we are in a stable state.
1089 	 */
1090 #ifdef VIMAGE
1091 	curvnet->vnet_ifcnt--;
1092 #endif
1093 	epoch_wait_preempt(net_epoch_preempt);
1094 	/*
1095 	 * In any case (destroy or vmove) detach us from the groups
1096 	 * and remove/wait for pending events on the taskq.
1097 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1098 	 */
1099 	if_delgroups(ifp);
1100 
1101 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1102 
1103 	/*
1104 	 * Check if this is a cloned interface or not. Must do even if
1105 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1106 	 * the if_clone_addgroup() will have a corrupted string overwise
1107 	 * from a gibberish pointer.
1108 	 */
1109 	if (vmove && ifcp != NULL)
1110 		*ifcp = if_clone_findifc(ifp);
1111 
1112 	if_down(ifp);
1113 
1114 #ifdef VIMAGE
1115 	/*
1116 	 * On VNET shutdown abort here as the stack teardown will do all
1117 	 * the work top-down for us.
1118 	 */
1119 	if (shutdown) {
1120 		/*
1121 		 * In case of a vmove we are done here without error.
1122 		 * If we would signal an error it would lead to the same
1123 		 * abort as if we did not find the ifnet anymore.
1124 		 * if_detach() calls us in void context and does not care
1125 		 * about an early abort notification, so life is splendid :)
1126 		 */
1127 		goto finish_vnet_shutdown;
1128 	}
1129 #endif
1130 
1131 	/*
1132 	 * At this point we are not tearing down a VNET and are either
1133 	 * going to destroy or vmove the interface and have to cleanup
1134 	 * accordingly.
1135 	 */
1136 
1137 	/*
1138 	 * Remove routes and flush queues.
1139 	 */
1140 #ifdef ALTQ
1141 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1142 		altq_disable(&ifp->if_snd);
1143 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1144 		altq_detach(&ifp->if_snd);
1145 #endif
1146 
1147 	if_purgeaddrs(ifp);
1148 
1149 #ifdef INET
1150 	in_ifdetach(ifp);
1151 #endif
1152 
1153 #ifdef INET6
1154 	/*
1155 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1156 	 * before removing routing entries below, since IPv6 interface direct
1157 	 * routes are expected to be removed by the IPv6-specific kernel API.
1158 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1159 	 */
1160 	in6_ifdetach(ifp);
1161 #endif
1162 	if_purgemaddrs(ifp);
1163 
1164 	/* Announce that the interface is gone. */
1165 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1166 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1167 	if (IS_DEFAULT_VNET(curvnet))
1168 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1169 
1170 	if (!vmove) {
1171 		/*
1172 		 * Prevent further calls into the device driver via ifnet.
1173 		 */
1174 		if_dead(ifp);
1175 
1176 		/*
1177 		 * Remove link ifaddr pointer and maybe decrement if_index.
1178 		 * Clean up all addresses.
1179 		 */
1180 		free(ifp->if_hw_addr, M_IFADDR);
1181 		ifp->if_hw_addr = NULL;
1182 		ifp->if_addr = NULL;
1183 
1184 		/* We can now free link ifaddr. */
1185 		IF_ADDR_WLOCK(ifp);
1186 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1187 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1188 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1189 			IF_ADDR_WUNLOCK(ifp);
1190 			ifa_free(ifa);
1191 		} else
1192 			IF_ADDR_WUNLOCK(ifp);
1193 	}
1194 
1195 	rt_flushifroutes(ifp);
1196 
1197 #ifdef VIMAGE
1198 finish_vnet_shutdown:
1199 #endif
1200 	/*
1201 	 * We cannot hold the lock over dom_ifdetach calls as they might
1202 	 * sleep, for example trying to drain a callout, thus open up the
1203 	 * theoretical race with re-attaching.
1204 	 */
1205 	IF_AFDATA_LOCK(ifp);
1206 	i = ifp->if_afdata_initialized;
1207 	ifp->if_afdata_initialized = 0;
1208 	IF_AFDATA_UNLOCK(ifp);
1209 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1210 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1211 			(*dp->dom_ifdetach)(ifp,
1212 			    ifp->if_afdata[dp->dom_family]);
1213 			ifp->if_afdata[dp->dom_family] = NULL;
1214 		}
1215 	}
1216 
1217 	return (0);
1218 }
1219 
1220 #ifdef VIMAGE
1221 /*
1222  * if_vmove() performs a limited version of if_detach() in current
1223  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1224  * An attempt is made to shrink if_index in current vnet, find an
1225  * unused if_index in target vnet and calls if_grow() if necessary,
1226  * and finally find an unused if_xname for the target vnet.
1227  */
1228 static void
1229 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1230 {
1231 	struct if_clone *ifc;
1232 	u_int bif_dlt, bif_hdrlen;
1233 	void *old;
1234 	int rc;
1235 
1236  	/*
1237 	 * if_detach_internal() will call the eventhandler to notify
1238 	 * interface departure.  That will detach if_bpf.  We need to
1239 	 * safe the dlt and hdrlen so we can re-attach it later.
1240 	 */
1241 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1242 
1243 	/*
1244 	 * Detach from current vnet, but preserve LLADDR info, do not
1245 	 * mark as dead etc. so that the ifnet can be reattached later.
1246 	 * If we cannot find it, we lost the race to someone else.
1247 	 */
1248 	rc = if_detach_internal(ifp, 1, &ifc);
1249 	if (rc != 0)
1250 		return;
1251 
1252 	/*
1253 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1254 	 * the if_index for that vnet if possible.
1255 	 *
1256 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1257 	 * or we'd lock on one vnet and unlock on another.
1258 	 */
1259 	IFNET_WLOCK();
1260 	ifindex_free_locked(ifp->if_index);
1261 	IFNET_WUNLOCK();
1262 
1263 	/*
1264 	 * Perform interface-specific reassignment tasks, if provided by
1265 	 * the driver.
1266 	 */
1267 	if (ifp->if_reassign != NULL)
1268 		ifp->if_reassign(ifp, new_vnet, NULL);
1269 
1270 	/*
1271 	 * Switch to the context of the target vnet.
1272 	 */
1273 	CURVNET_SET_QUIET(new_vnet);
1274  restart:
1275 	IFNET_WLOCK();
1276 	ifp->if_index = ifindex_alloc(&old);
1277 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1278 		IFNET_WUNLOCK();
1279 		epoch_wait_preempt(net_epoch_preempt);
1280 		free(old, M_IFNET);
1281 		goto restart;
1282 	}
1283 	ifnet_setbyindex(ifp->if_index, ifp);
1284 	IFNET_WUNLOCK();
1285 
1286 	if_attach_internal(ifp, 1, ifc);
1287 
1288 	if (ifp->if_bpf == NULL)
1289 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1290 
1291 	CURVNET_RESTORE();
1292 }
1293 
1294 /*
1295  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1296  */
1297 static int
1298 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1299 {
1300 	struct prison *pr;
1301 	struct ifnet *difp;
1302 	int shutdown;
1303 
1304 	/* Try to find the prison within our visibility. */
1305 	sx_slock(&allprison_lock);
1306 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1307 	sx_sunlock(&allprison_lock);
1308 	if (pr == NULL)
1309 		return (ENXIO);
1310 	prison_hold_locked(pr);
1311 	mtx_unlock(&pr->pr_mtx);
1312 
1313 	/* Do not try to move the iface from and to the same prison. */
1314 	if (pr->pr_vnet == ifp->if_vnet) {
1315 		prison_free(pr);
1316 		return (EEXIST);
1317 	}
1318 
1319 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1320 	/* XXX Lock interfaces to avoid races. */
1321 	CURVNET_SET_QUIET(pr->pr_vnet);
1322 	difp = ifunit(ifname);
1323 	if (difp != NULL) {
1324 		CURVNET_RESTORE();
1325 		prison_free(pr);
1326 		return (EEXIST);
1327 	}
1328 
1329 	/* Make sure the VNET is stable. */
1330 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1331 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1332 	if (shutdown) {
1333 		CURVNET_RESTORE();
1334 		prison_free(pr);
1335 		return (EBUSY);
1336 	}
1337 	CURVNET_RESTORE();
1338 
1339 	/* Move the interface into the child jail/vnet. */
1340 	if_vmove(ifp, pr->pr_vnet);
1341 
1342 	/* Report the new if_xname back to the userland. */
1343 	sprintf(ifname, "%s", ifp->if_xname);
1344 
1345 	prison_free(pr);
1346 	return (0);
1347 }
1348 
1349 static int
1350 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1351 {
1352 	struct prison *pr;
1353 	struct vnet *vnet_dst;
1354 	struct ifnet *ifp;
1355  	int shutdown;
1356 
1357 	/* Try to find the prison within our visibility. */
1358 	sx_slock(&allprison_lock);
1359 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1360 	sx_sunlock(&allprison_lock);
1361 	if (pr == NULL)
1362 		return (ENXIO);
1363 	prison_hold_locked(pr);
1364 	mtx_unlock(&pr->pr_mtx);
1365 
1366 	/* Make sure the named iface exists in the source prison/vnet. */
1367 	CURVNET_SET(pr->pr_vnet);
1368 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1369 	if (ifp == NULL) {
1370 		CURVNET_RESTORE();
1371 		prison_free(pr);
1372 		return (ENXIO);
1373 	}
1374 
1375 	/* Do not try to move the iface from and to the same prison. */
1376 	vnet_dst = TD_TO_VNET(td);
1377 	if (vnet_dst == ifp->if_vnet) {
1378 		CURVNET_RESTORE();
1379 		prison_free(pr);
1380 		return (EEXIST);
1381 	}
1382 
1383 	/* Make sure the VNET is stable. */
1384 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1385 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1386 	if (shutdown) {
1387 		CURVNET_RESTORE();
1388 		prison_free(pr);
1389 		return (EBUSY);
1390 	}
1391 
1392 	/* Get interface back from child jail/vnet. */
1393 	if_vmove(ifp, vnet_dst);
1394 	CURVNET_RESTORE();
1395 
1396 	/* Report the new if_xname back to the userland. */
1397 	sprintf(ifname, "%s", ifp->if_xname);
1398 
1399 	prison_free(pr);
1400 	return (0);
1401 }
1402 #endif /* VIMAGE */
1403 
1404 /*
1405  * Add a group to an interface
1406  */
1407 int
1408 if_addgroup(struct ifnet *ifp, const char *groupname)
1409 {
1410 	struct ifg_list		*ifgl;
1411 	struct ifg_group	*ifg = NULL;
1412 	struct ifg_member	*ifgm;
1413 	int 			 new = 0;
1414 
1415 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1416 	    groupname[strlen(groupname) - 1] <= '9')
1417 		return (EINVAL);
1418 
1419 	IFNET_WLOCK();
1420 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1421 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1422 			IFNET_WUNLOCK();
1423 			return (EEXIST);
1424 		}
1425 
1426 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1427 	    M_NOWAIT)) == NULL) {
1428 	    	IFNET_WUNLOCK();
1429 		return (ENOMEM);
1430 	}
1431 
1432 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1433 	    M_TEMP, M_NOWAIT)) == NULL) {
1434 		free(ifgl, M_TEMP);
1435 		IFNET_WUNLOCK();
1436 		return (ENOMEM);
1437 	}
1438 
1439 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1440 		if (!strcmp(ifg->ifg_group, groupname))
1441 			break;
1442 
1443 	if (ifg == NULL) {
1444 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1445 		    M_TEMP, M_NOWAIT)) == NULL) {
1446 			free(ifgl, M_TEMP);
1447 			free(ifgm, M_TEMP);
1448 			IFNET_WUNLOCK();
1449 			return (ENOMEM);
1450 		}
1451 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1452 		ifg->ifg_refcnt = 0;
1453 		CK_STAILQ_INIT(&ifg->ifg_members);
1454 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1455 		new = 1;
1456 	}
1457 
1458 	ifg->ifg_refcnt++;
1459 	ifgl->ifgl_group = ifg;
1460 	ifgm->ifgm_ifp = ifp;
1461 
1462 	IF_ADDR_WLOCK(ifp);
1463 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1464 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1465 	IF_ADDR_WUNLOCK(ifp);
1466 
1467 	IFNET_WUNLOCK();
1468 
1469 	if (new)
1470 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1471 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1472 
1473 	return (0);
1474 }
1475 
1476 /*
1477  * Remove a group from an interface
1478  */
1479 int
1480 if_delgroup(struct ifnet *ifp, const char *groupname)
1481 {
1482 	struct ifg_list		*ifgl;
1483 	struct ifg_member	*ifgm;
1484 	int freeifgl;
1485 
1486 	IFNET_WLOCK();
1487 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1488 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1489 			break;
1490 	if (ifgl == NULL) {
1491 		IFNET_WUNLOCK();
1492 		return (ENOENT);
1493 	}
1494 
1495 	freeifgl = 0;
1496 	IF_ADDR_WLOCK(ifp);
1497 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1498 	IF_ADDR_WUNLOCK(ifp);
1499 
1500 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1501 		if (ifgm->ifgm_ifp == ifp)
1502 			break;
1503 
1504 	if (ifgm != NULL)
1505 		CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1506 
1507 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1508 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1509 		freeifgl = 1;
1510 	}
1511 	IFNET_WUNLOCK();
1512 
1513 	epoch_wait_preempt(net_epoch_preempt);
1514 	if (freeifgl) {
1515 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1516 		free(ifgl->ifgl_group, M_TEMP);
1517 	}
1518 	free(ifgm, M_TEMP);
1519 	free(ifgl, M_TEMP);
1520 
1521 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1522 
1523 	return (0);
1524 }
1525 
1526 /*
1527  * Remove an interface from all groups
1528  */
1529 static void
1530 if_delgroups(struct ifnet *ifp)
1531 {
1532 	struct ifg_list		*ifgl;
1533 	struct ifg_member	*ifgm;
1534 	char groupname[IFNAMSIZ];
1535 	int ifglfree;
1536 
1537 	IFNET_WLOCK();
1538 	while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1539 		ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1540 
1541 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1542 
1543 		IF_ADDR_WLOCK(ifp);
1544 		CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1545 		IF_ADDR_WUNLOCK(ifp);
1546 
1547 		CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1548 			if (ifgm->ifgm_ifp == ifp)
1549 				break;
1550 
1551 		if (ifgm != NULL)
1552 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1553 			    ifgm_next);
1554 		ifglfree = 0;
1555 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1556 			CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1557 			ifglfree = 1;
1558 		}
1559 
1560 		IFNET_WUNLOCK();
1561 		epoch_wait_preempt(net_epoch_preempt);
1562 		free(ifgm, M_TEMP);
1563 		if (ifglfree) {
1564 			EVENTHANDLER_INVOKE(group_detach_event,
1565 								ifgl->ifgl_group);
1566 			free(ifgl->ifgl_group, M_TEMP);
1567 		}
1568 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1569 
1570 		IFNET_WLOCK();
1571 	}
1572 	IFNET_WUNLOCK();
1573 }
1574 
1575 static char *
1576 ifgr_group_get(void *ifgrp)
1577 {
1578 	union ifgroupreq_union *ifgrup;
1579 
1580 	ifgrup = ifgrp;
1581 #ifdef COMPAT_FREEBSD32
1582 	if (SV_CURPROC_FLAG(SV_ILP32))
1583 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1584 #endif
1585 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1586 }
1587 
1588 static struct ifg_req *
1589 ifgr_groups_get(void *ifgrp)
1590 {
1591 	union ifgroupreq_union *ifgrup;
1592 
1593 	ifgrup = ifgrp;
1594 #ifdef COMPAT_FREEBSD32
1595 	if (SV_CURPROC_FLAG(SV_ILP32))
1596 		return ((struct ifg_req *)(uintptr_t)
1597 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1598 #endif
1599 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1600 }
1601 
1602 /*
1603  * Stores all groups from an interface in memory pointed to by ifgr.
1604  */
1605 static int
1606 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1607 {
1608 	int			 len, error;
1609 	struct ifg_list		*ifgl;
1610 	struct ifg_req		 ifgrq, *ifgp;
1611 
1612 	if (ifgr->ifgr_len == 0) {
1613 		IF_ADDR_RLOCK(ifp);
1614 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1615 			ifgr->ifgr_len += sizeof(struct ifg_req);
1616 		IF_ADDR_RUNLOCK(ifp);
1617 		return (0);
1618 	}
1619 
1620 	len = ifgr->ifgr_len;
1621 	ifgp = ifgr_groups_get(ifgr);
1622 	/* XXX: wire */
1623 	IF_ADDR_RLOCK(ifp);
1624 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1625 		if (len < sizeof(ifgrq)) {
1626 			IF_ADDR_RUNLOCK(ifp);
1627 			return (EINVAL);
1628 		}
1629 		bzero(&ifgrq, sizeof ifgrq);
1630 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1631 		    sizeof(ifgrq.ifgrq_group));
1632 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1633 		    	IF_ADDR_RUNLOCK(ifp);
1634 			return (error);
1635 		}
1636 		len -= sizeof(ifgrq);
1637 		ifgp++;
1638 	}
1639 	IF_ADDR_RUNLOCK(ifp);
1640 
1641 	return (0);
1642 }
1643 
1644 /*
1645  * Stores all members of a group in memory pointed to by igfr
1646  */
1647 static int
1648 if_getgroupmembers(struct ifgroupreq *ifgr)
1649 {
1650 	struct ifg_group	*ifg;
1651 	struct ifg_member	*ifgm;
1652 	struct ifg_req		 ifgrq, *ifgp;
1653 	int			 len, error;
1654 
1655 	IFNET_RLOCK();
1656 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1657 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1658 			break;
1659 	if (ifg == NULL) {
1660 		IFNET_RUNLOCK();
1661 		return (ENOENT);
1662 	}
1663 
1664 	if (ifgr->ifgr_len == 0) {
1665 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1666 			ifgr->ifgr_len += sizeof(ifgrq);
1667 		IFNET_RUNLOCK();
1668 		return (0);
1669 	}
1670 
1671 	len = ifgr->ifgr_len;
1672 	ifgp = ifgr_groups_get(ifgr);
1673 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1674 		if (len < sizeof(ifgrq)) {
1675 			IFNET_RUNLOCK();
1676 			return (EINVAL);
1677 		}
1678 		bzero(&ifgrq, sizeof ifgrq);
1679 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1680 		    sizeof(ifgrq.ifgrq_member));
1681 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1682 			IFNET_RUNLOCK();
1683 			return (error);
1684 		}
1685 		len -= sizeof(ifgrq);
1686 		ifgp++;
1687 	}
1688 	IFNET_RUNLOCK();
1689 
1690 	return (0);
1691 }
1692 
1693 /*
1694  * Return counter values from counter(9)s stored in ifnet.
1695  */
1696 uint64_t
1697 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1698 {
1699 
1700 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1701 
1702 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1703 }
1704 
1705 /*
1706  * Increase an ifnet counter. Usually used for counters shared
1707  * between the stack and a driver, but function supports them all.
1708  */
1709 void
1710 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1711 {
1712 
1713 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1714 
1715 	counter_u64_add(ifp->if_counters[cnt], inc);
1716 }
1717 
1718 /*
1719  * Copy data from ifnet to userland API structure if_data.
1720  */
1721 void
1722 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1723 {
1724 
1725 	ifd->ifi_type = ifp->if_type;
1726 	ifd->ifi_physical = 0;
1727 	ifd->ifi_addrlen = ifp->if_addrlen;
1728 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1729 	ifd->ifi_link_state = ifp->if_link_state;
1730 	ifd->ifi_vhid = 0;
1731 	ifd->ifi_datalen = sizeof(struct if_data);
1732 	ifd->ifi_mtu = ifp->if_mtu;
1733 	ifd->ifi_metric = ifp->if_metric;
1734 	ifd->ifi_baudrate = ifp->if_baudrate;
1735 	ifd->ifi_hwassist = ifp->if_hwassist;
1736 	ifd->ifi_epoch = ifp->if_epoch;
1737 	ifd->ifi_lastchange = ifp->if_lastchange;
1738 
1739 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1740 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1741 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1742 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1743 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1744 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1745 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1746 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1747 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1748 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1749 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1750 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1751 }
1752 
1753 /*
1754  * Wrapper functions for struct ifnet address list locking macros.  These are
1755  * used by kernel modules to avoid encoding programming interface or binary
1756  * interface assumptions that may be violated when kernel-internal locking
1757  * approaches change.
1758  */
1759 void
1760 if_addr_rlock(struct ifnet *ifp)
1761 {
1762 
1763 	IF_ADDR_RLOCK(ifp);
1764 }
1765 
1766 void
1767 if_addr_runlock(struct ifnet *ifp)
1768 {
1769 
1770 	IF_ADDR_RUNLOCK(ifp);
1771 }
1772 
1773 void
1774 if_maddr_rlock(if_t ifp)
1775 {
1776 
1777 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1778 }
1779 
1780 void
1781 if_maddr_runlock(if_t ifp)
1782 {
1783 
1784 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1785 }
1786 
1787 /*
1788  * Initialization, destruction and refcounting functions for ifaddrs.
1789  */
1790 struct ifaddr *
1791 ifa_alloc(size_t size, int flags)
1792 {
1793 	struct ifaddr *ifa;
1794 
1795 	KASSERT(size >= sizeof(struct ifaddr),
1796 	    ("%s: invalid size %zu", __func__, size));
1797 
1798 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1799 	if (ifa == NULL)
1800 		return (NULL);
1801 
1802 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1803 		goto fail;
1804 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1805 		goto fail;
1806 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1807 		goto fail;
1808 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1809 		goto fail;
1810 
1811 	refcount_init(&ifa->ifa_refcnt, 1);
1812 
1813 	return (ifa);
1814 
1815 fail:
1816 	/* free(NULL) is okay */
1817 	counter_u64_free(ifa->ifa_opackets);
1818 	counter_u64_free(ifa->ifa_ipackets);
1819 	counter_u64_free(ifa->ifa_obytes);
1820 	counter_u64_free(ifa->ifa_ibytes);
1821 	free(ifa, M_IFADDR);
1822 
1823 	return (NULL);
1824 }
1825 
1826 void
1827 ifa_ref(struct ifaddr *ifa)
1828 {
1829 
1830 	refcount_acquire(&ifa->ifa_refcnt);
1831 }
1832 
1833 static void
1834 ifa_destroy(epoch_context_t ctx)
1835 {
1836 	struct ifaddr *ifa;
1837 
1838 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1839 	counter_u64_free(ifa->ifa_opackets);
1840 	counter_u64_free(ifa->ifa_ipackets);
1841 	counter_u64_free(ifa->ifa_obytes);
1842 	counter_u64_free(ifa->ifa_ibytes);
1843 	free(ifa, M_IFADDR);
1844 }
1845 
1846 void
1847 ifa_free(struct ifaddr *ifa)
1848 {
1849 
1850 	if (refcount_release(&ifa->ifa_refcnt))
1851 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1852 }
1853 
1854 
1855 static int
1856 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1857     struct sockaddr *ia)
1858 {
1859 	int error;
1860 	struct rt_addrinfo info;
1861 	struct sockaddr_dl null_sdl;
1862 	struct ifnet *ifp;
1863 
1864 	ifp = ifa->ifa_ifp;
1865 
1866 	bzero(&info, sizeof(info));
1867 	if (cmd != RTM_DELETE)
1868 		info.rti_ifp = V_loif;
1869 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1870 	info.rti_info[RTAX_DST] = ia;
1871 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1872 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1873 
1874 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1875 
1876 	if (error != 0 &&
1877 	    !(cmd == RTM_ADD && error == EEXIST) &&
1878 	    !(cmd == RTM_DELETE && error == ENOENT))
1879 		if_printf(ifp, "%s failed: %d\n", otype, error);
1880 
1881 	return (error);
1882 }
1883 
1884 int
1885 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1886 {
1887 
1888 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1889 }
1890 
1891 int
1892 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1893 {
1894 
1895 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1896 }
1897 
1898 int
1899 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1900 {
1901 
1902 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1903 }
1904 
1905 /*
1906  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1907  * structs used to represent other address families, it is necessary
1908  * to perform a different comparison.
1909  */
1910 
1911 #define	sa_dl_equal(a1, a2)	\
1912 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1913 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1914 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1915 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1916 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1917 
1918 /*
1919  * Locate an interface based on a complete address.
1920  */
1921 /*ARGSUSED*/
1922 struct ifaddr *
1923 ifa_ifwithaddr(const struct sockaddr *addr)
1924 {
1925 	struct ifnet *ifp;
1926 	struct ifaddr *ifa;
1927 
1928 	MPASS(in_epoch());
1929 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1930 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1931 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1932 				continue;
1933 			if (sa_equal(addr, ifa->ifa_addr)) {
1934 				goto done;
1935 			}
1936 			/* IP6 doesn't have broadcast */
1937 			if ((ifp->if_flags & IFF_BROADCAST) &&
1938 			    ifa->ifa_broadaddr &&
1939 			    ifa->ifa_broadaddr->sa_len != 0 &&
1940 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1941 				goto done;
1942 			}
1943 		}
1944 	}
1945 	ifa = NULL;
1946 done:
1947 	return (ifa);
1948 }
1949 
1950 int
1951 ifa_ifwithaddr_check(const struct sockaddr *addr)
1952 {
1953 	int rc;
1954 
1955 	NET_EPOCH_ENTER();
1956 	rc = (ifa_ifwithaddr(addr) != NULL);
1957 	NET_EPOCH_EXIT();
1958 	return (rc);
1959 }
1960 
1961 /*
1962  * Locate an interface based on the broadcast address.
1963  */
1964 /* ARGSUSED */
1965 struct ifaddr *
1966 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1967 {
1968 	struct ifnet *ifp;
1969 	struct ifaddr *ifa;
1970 
1971 	MPASS(in_epoch());
1972 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1973 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1974 			continue;
1975 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1976 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1977 				continue;
1978 			if ((ifp->if_flags & IFF_BROADCAST) &&
1979 			    ifa->ifa_broadaddr &&
1980 			    ifa->ifa_broadaddr->sa_len != 0 &&
1981 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1982 				goto done;
1983 			}
1984 		}
1985 	}
1986 	ifa = NULL;
1987 done:
1988 	return (ifa);
1989 }
1990 
1991 /*
1992  * Locate the point to point interface with a given destination address.
1993  */
1994 /*ARGSUSED*/
1995 struct ifaddr *
1996 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1997 {
1998 	struct ifnet *ifp;
1999 	struct ifaddr *ifa;
2000 
2001 	MPASS(in_epoch());
2002 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2003 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2004 			continue;
2005 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2006 			continue;
2007 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2008 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2009 				continue;
2010 			if (ifa->ifa_dstaddr != NULL &&
2011 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2012 				goto done;
2013 			}
2014 		}
2015 	}
2016 	ifa = NULL;
2017 done:
2018 	return (ifa);
2019 }
2020 
2021 /*
2022  * Find an interface on a specific network.  If many, choice
2023  * is most specific found.
2024  */
2025 struct ifaddr *
2026 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2027 {
2028 	struct ifnet *ifp;
2029 	struct ifaddr *ifa;
2030 	struct ifaddr *ifa_maybe = NULL;
2031 	u_int af = addr->sa_family;
2032 	const char *addr_data = addr->sa_data, *cplim;
2033 
2034 	MPASS(in_epoch());
2035 	/*
2036 	 * AF_LINK addresses can be looked up directly by their index number,
2037 	 * so do that if we can.
2038 	 */
2039 	if (af == AF_LINK) {
2040 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2041 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2042 		return (ifaddr_byindex(sdl->sdl_index));
2043 	}
2044 
2045 	/*
2046 	 * Scan though each interface, looking for ones that have addresses
2047 	 * in this address family and the requested fib.  Maintain a reference
2048 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2049 	 * kept it stable when we move onto the next interface.
2050 	 */
2051 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2052 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2053 			continue;
2054 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2055 			const char *cp, *cp2, *cp3;
2056 
2057 			if (ifa->ifa_addr->sa_family != af)
2058 next:				continue;
2059 			if (af == AF_INET &&
2060 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2061 				/*
2062 				 * This is a bit broken as it doesn't
2063 				 * take into account that the remote end may
2064 				 * be a single node in the network we are
2065 				 * looking for.
2066 				 * The trouble is that we don't know the
2067 				 * netmask for the remote end.
2068 				 */
2069 				if (ifa->ifa_dstaddr != NULL &&
2070 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2071 					IF_ADDR_RUNLOCK(ifp);
2072 					goto done;
2073 				}
2074 			} else {
2075 				/*
2076 				 * Scan all the bits in the ifa's address.
2077 				 * If a bit dissagrees with what we are
2078 				 * looking for, mask it with the netmask
2079 				 * to see if it really matters.
2080 				 * (A byte at a time)
2081 				 */
2082 				if (ifa->ifa_netmask == 0)
2083 					continue;
2084 				cp = addr_data;
2085 				cp2 = ifa->ifa_addr->sa_data;
2086 				cp3 = ifa->ifa_netmask->sa_data;
2087 				cplim = ifa->ifa_netmask->sa_len
2088 					+ (char *)ifa->ifa_netmask;
2089 				while (cp3 < cplim)
2090 					if ((*cp++ ^ *cp2++) & *cp3++)
2091 						goto next; /* next address! */
2092 				/*
2093 				 * If the netmask of what we just found
2094 				 * is more specific than what we had before
2095 				 * (if we had one), or if the virtual status
2096 				 * of new prefix is better than of the old one,
2097 				 * then remember the new one before continuing
2098 				 * to search for an even better one.
2099 				 */
2100 				if (ifa_maybe == NULL ||
2101 				    ifa_preferred(ifa_maybe, ifa) ||
2102 				    rn_refines((caddr_t)ifa->ifa_netmask,
2103 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2104 					ifa_maybe = ifa;
2105 				}
2106 			}
2107 		}
2108 	}
2109 	ifa = ifa_maybe;
2110 	ifa_maybe = NULL;
2111 done:
2112 	return (ifa);
2113 }
2114 
2115 /*
2116  * Find an interface address specific to an interface best matching
2117  * a given address.
2118  */
2119 struct ifaddr *
2120 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2121 {
2122 	struct ifaddr *ifa;
2123 	const char *cp, *cp2, *cp3;
2124 	char *cplim;
2125 	struct ifaddr *ifa_maybe = NULL;
2126 	u_int af = addr->sa_family;
2127 
2128 	if (af >= AF_MAX)
2129 		return (NULL);
2130 	MPASS(in_epoch());
2131 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2132 		if (ifa->ifa_addr->sa_family != af)
2133 			continue;
2134 		if (ifa_maybe == NULL)
2135 			ifa_maybe = ifa;
2136 		if (ifa->ifa_netmask == 0) {
2137 			if (sa_equal(addr, ifa->ifa_addr) ||
2138 			    (ifa->ifa_dstaddr &&
2139 			    sa_equal(addr, ifa->ifa_dstaddr)))
2140 				goto done;
2141 			continue;
2142 		}
2143 		if (ifp->if_flags & IFF_POINTOPOINT) {
2144 			if (sa_equal(addr, ifa->ifa_dstaddr))
2145 				goto done;
2146 		} else {
2147 			cp = addr->sa_data;
2148 			cp2 = ifa->ifa_addr->sa_data;
2149 			cp3 = ifa->ifa_netmask->sa_data;
2150 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2151 			for (; cp3 < cplim; cp3++)
2152 				if ((*cp++ ^ *cp2++) & *cp3)
2153 					break;
2154 			if (cp3 == cplim)
2155 				goto done;
2156 		}
2157 	}
2158 	ifa = ifa_maybe;
2159 done:
2160 	return (ifa);
2161 }
2162 
2163 /*
2164  * See whether new ifa is better than current one:
2165  * 1) A non-virtual one is preferred over virtual.
2166  * 2) A virtual in master state preferred over any other state.
2167  *
2168  * Used in several address selecting functions.
2169  */
2170 int
2171 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2172 {
2173 
2174 	return (cur->ifa_carp && (!next->ifa_carp ||
2175 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2176 }
2177 
2178 #include <net/if_llatbl.h>
2179 
2180 /*
2181  * Default action when installing a route with a Link Level gateway.
2182  * Lookup an appropriate real ifa to point to.
2183  * This should be moved to /sys/net/link.c eventually.
2184  */
2185 static void
2186 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2187 {
2188 	struct ifaddr *ifa, *oifa;
2189 	struct sockaddr *dst;
2190 	struct ifnet *ifp;
2191 
2192 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2193 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2194 		return;
2195 	NET_EPOCH_ENTER();
2196 	ifa = ifaof_ifpforaddr(dst, ifp);
2197 	if (ifa) {
2198 		oifa = rt->rt_ifa;
2199 		rt->rt_ifa = ifa;
2200 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2201 			ifa->ifa_rtrequest(cmd, rt, info);
2202 	}
2203 	NET_EPOCH_EXIT();
2204 }
2205 
2206 struct sockaddr_dl *
2207 link_alloc_sdl(size_t size, int flags)
2208 {
2209 
2210 	return (malloc(size, M_TEMP, flags));
2211 }
2212 
2213 void
2214 link_free_sdl(struct sockaddr *sa)
2215 {
2216 	free(sa, M_TEMP);
2217 }
2218 
2219 /*
2220  * Fills in given sdl with interface basic info.
2221  * Returns pointer to filled sdl.
2222  */
2223 struct sockaddr_dl *
2224 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2225 {
2226 	struct sockaddr_dl *sdl;
2227 
2228 	sdl = (struct sockaddr_dl *)paddr;
2229 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2230 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2231 	sdl->sdl_family = AF_LINK;
2232 	sdl->sdl_index = ifp->if_index;
2233 	sdl->sdl_type = iftype;
2234 
2235 	return (sdl);
2236 }
2237 
2238 /*
2239  * Mark an interface down and notify protocols of
2240  * the transition.
2241  */
2242 static void
2243 if_unroute(struct ifnet *ifp, int flag, int fam)
2244 {
2245 	struct ifaddr *ifa;
2246 
2247 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2248 
2249 	ifp->if_flags &= ~flag;
2250 	getmicrotime(&ifp->if_lastchange);
2251 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2252 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2253 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2254 	ifp->if_qflush(ifp);
2255 
2256 	if (ifp->if_carp)
2257 		(*carp_linkstate_p)(ifp);
2258 	rt_ifmsg(ifp);
2259 }
2260 
2261 /*
2262  * Mark an interface up and notify protocols of
2263  * the transition.
2264  */
2265 static void
2266 if_route(struct ifnet *ifp, int flag, int fam)
2267 {
2268 	struct ifaddr *ifa;
2269 
2270 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2271 
2272 	ifp->if_flags |= flag;
2273 	getmicrotime(&ifp->if_lastchange);
2274 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2275 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2276 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2277 	if (ifp->if_carp)
2278 		(*carp_linkstate_p)(ifp);
2279 	rt_ifmsg(ifp);
2280 #ifdef INET6
2281 	in6_if_up(ifp);
2282 #endif
2283 }
2284 
2285 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2286 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2287 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2288 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2289 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2290 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2291 void	*(*vlan_cookie_p)(struct ifnet *);
2292 
2293 /*
2294  * Handle a change in the interface link state. To avoid LORs
2295  * between driver lock and upper layer locks, as well as possible
2296  * recursions, we post event to taskqueue, and all job
2297  * is done in static do_link_state_change().
2298  */
2299 void
2300 if_link_state_change(struct ifnet *ifp, int link_state)
2301 {
2302 	/* Return if state hasn't changed. */
2303 	if (ifp->if_link_state == link_state)
2304 		return;
2305 
2306 	ifp->if_link_state = link_state;
2307 
2308 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2309 }
2310 
2311 static void
2312 do_link_state_change(void *arg, int pending)
2313 {
2314 	struct ifnet *ifp = (struct ifnet *)arg;
2315 	int link_state = ifp->if_link_state;
2316 	CURVNET_SET(ifp->if_vnet);
2317 
2318 	/* Notify that the link state has changed. */
2319 	rt_ifmsg(ifp);
2320 	if (ifp->if_vlantrunk != NULL)
2321 		(*vlan_link_state_p)(ifp);
2322 
2323 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2324 	    ifp->if_l2com != NULL)
2325 		(*ng_ether_link_state_p)(ifp, link_state);
2326 	if (ifp->if_carp)
2327 		(*carp_linkstate_p)(ifp);
2328 	if (ifp->if_bridge)
2329 		ifp->if_bridge_linkstate(ifp);
2330 	if (ifp->if_lagg)
2331 		(*lagg_linkstate_p)(ifp, link_state);
2332 
2333 	if (IS_DEFAULT_VNET(curvnet))
2334 		devctl_notify("IFNET", ifp->if_xname,
2335 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2336 		    NULL);
2337 	if (pending > 1)
2338 		if_printf(ifp, "%d link states coalesced\n", pending);
2339 	if (log_link_state_change)
2340 		if_printf(ifp, "link state changed to %s\n",
2341 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2342 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2343 	CURVNET_RESTORE();
2344 }
2345 
2346 /*
2347  * Mark an interface down and notify protocols of
2348  * the transition.
2349  */
2350 void
2351 if_down(struct ifnet *ifp)
2352 {
2353 
2354 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2355 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2356 }
2357 
2358 /*
2359  * Mark an interface up and notify protocols of
2360  * the transition.
2361  */
2362 void
2363 if_up(struct ifnet *ifp)
2364 {
2365 
2366 	if_route(ifp, IFF_UP, AF_UNSPEC);
2367 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2368 }
2369 
2370 /*
2371  * Flush an interface queue.
2372  */
2373 void
2374 if_qflush(struct ifnet *ifp)
2375 {
2376 	struct mbuf *m, *n;
2377 	struct ifaltq *ifq;
2378 
2379 	ifq = &ifp->if_snd;
2380 	IFQ_LOCK(ifq);
2381 #ifdef ALTQ
2382 	if (ALTQ_IS_ENABLED(ifq))
2383 		ALTQ_PURGE(ifq);
2384 #endif
2385 	n = ifq->ifq_head;
2386 	while ((m = n) != NULL) {
2387 		n = m->m_nextpkt;
2388 		m_freem(m);
2389 	}
2390 	ifq->ifq_head = 0;
2391 	ifq->ifq_tail = 0;
2392 	ifq->ifq_len = 0;
2393 	IFQ_UNLOCK(ifq);
2394 }
2395 
2396 /*
2397  * Map interface name to interface structure pointer, with or without
2398  * returning a reference.
2399  */
2400 struct ifnet *
2401 ifunit_ref(const char *name)
2402 {
2403 	struct ifnet *ifp;
2404 
2405 	IFNET_RLOCK_NOSLEEP();
2406 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2407 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2408 		    !(ifp->if_flags & IFF_DYING))
2409 			break;
2410 	}
2411 	if (ifp != NULL)
2412 		if_ref(ifp);
2413 	IFNET_RUNLOCK_NOSLEEP();
2414 	return (ifp);
2415 }
2416 
2417 struct ifnet *
2418 ifunit(const char *name)
2419 {
2420 	struct ifnet *ifp;
2421 
2422 	IFNET_RLOCK_NOSLEEP();
2423 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2424 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2425 			break;
2426 	}
2427 	IFNET_RUNLOCK_NOSLEEP();
2428 	return (ifp);
2429 }
2430 
2431 static void *
2432 ifr_buffer_get_buffer(void *data)
2433 {
2434 	union ifreq_union *ifrup;
2435 
2436 	ifrup = data;
2437 #ifdef COMPAT_FREEBSD32
2438 	if (SV_CURPROC_FLAG(SV_ILP32))
2439 		return ((void *)(uintptr_t)
2440 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2441 #endif
2442 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2443 }
2444 
2445 static void
2446 ifr_buffer_set_buffer_null(void *data)
2447 {
2448 	union ifreq_union *ifrup;
2449 
2450 	ifrup = data;
2451 #ifdef COMPAT_FREEBSD32
2452 	if (SV_CURPROC_FLAG(SV_ILP32))
2453 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2454 	else
2455 #endif
2456 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2457 }
2458 
2459 static size_t
2460 ifr_buffer_get_length(void *data)
2461 {
2462 	union ifreq_union *ifrup;
2463 
2464 	ifrup = data;
2465 #ifdef COMPAT_FREEBSD32
2466 	if (SV_CURPROC_FLAG(SV_ILP32))
2467 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2468 #endif
2469 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2470 }
2471 
2472 static void
2473 ifr_buffer_set_length(void *data, size_t len)
2474 {
2475 	union ifreq_union *ifrup;
2476 
2477 	ifrup = data;
2478 #ifdef COMPAT_FREEBSD32
2479 	if (SV_CURPROC_FLAG(SV_ILP32))
2480 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2481 	else
2482 #endif
2483 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2484 }
2485 
2486 void *
2487 ifr_data_get_ptr(void *ifrp)
2488 {
2489 	union ifreq_union *ifrup;
2490 
2491 	ifrup = ifrp;
2492 #ifdef COMPAT_FREEBSD32
2493 	if (SV_CURPROC_FLAG(SV_ILP32))
2494 		return ((void *)(uintptr_t)
2495 		    ifrup->ifr32.ifr_ifru.ifru_data);
2496 #endif
2497 		return (ifrup->ifr.ifr_ifru.ifru_data);
2498 }
2499 
2500 /*
2501  * Hardware specific interface ioctls.
2502  */
2503 static int
2504 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2505 {
2506 	struct ifreq *ifr;
2507 	int error = 0, do_ifup = 0;
2508 	int new_flags, temp_flags;
2509 	size_t namelen, onamelen;
2510 	size_t descrlen;
2511 	char *descrbuf, *odescrbuf;
2512 	char new_name[IFNAMSIZ];
2513 	struct ifaddr *ifa;
2514 	struct sockaddr_dl *sdl;
2515 
2516 	ifr = (struct ifreq *)data;
2517 	switch (cmd) {
2518 	case SIOCGIFINDEX:
2519 		ifr->ifr_index = ifp->if_index;
2520 		break;
2521 
2522 	case SIOCGIFFLAGS:
2523 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2524 		ifr->ifr_flags = temp_flags & 0xffff;
2525 		ifr->ifr_flagshigh = temp_flags >> 16;
2526 		break;
2527 
2528 	case SIOCGIFCAP:
2529 		ifr->ifr_reqcap = ifp->if_capabilities;
2530 		ifr->ifr_curcap = ifp->if_capenable;
2531 		break;
2532 
2533 #ifdef MAC
2534 	case SIOCGIFMAC:
2535 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2536 		break;
2537 #endif
2538 
2539 	case SIOCGIFMETRIC:
2540 		ifr->ifr_metric = ifp->if_metric;
2541 		break;
2542 
2543 	case SIOCGIFMTU:
2544 		ifr->ifr_mtu = ifp->if_mtu;
2545 		break;
2546 
2547 	case SIOCGIFPHYS:
2548 		/* XXXGL: did this ever worked? */
2549 		ifr->ifr_phys = 0;
2550 		break;
2551 
2552 	case SIOCGIFDESCR:
2553 		error = 0;
2554 		sx_slock(&ifdescr_sx);
2555 		if (ifp->if_description == NULL)
2556 			error = ENOMSG;
2557 		else {
2558 			/* space for terminating nul */
2559 			descrlen = strlen(ifp->if_description) + 1;
2560 			if (ifr_buffer_get_length(ifr) < descrlen)
2561 				ifr_buffer_set_buffer_null(ifr);
2562 			else
2563 				error = copyout(ifp->if_description,
2564 				    ifr_buffer_get_buffer(ifr), descrlen);
2565 			ifr_buffer_set_length(ifr, descrlen);
2566 		}
2567 		sx_sunlock(&ifdescr_sx);
2568 		break;
2569 
2570 	case SIOCSIFDESCR:
2571 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2572 		if (error)
2573 			return (error);
2574 
2575 		/*
2576 		 * Copy only (length-1) bytes to make sure that
2577 		 * if_description is always nul terminated.  The
2578 		 * length parameter is supposed to count the
2579 		 * terminating nul in.
2580 		 */
2581 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2582 			return (ENAMETOOLONG);
2583 		else if (ifr_buffer_get_length(ifr) == 0)
2584 			descrbuf = NULL;
2585 		else {
2586 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2587 			    M_IFDESCR, M_WAITOK | M_ZERO);
2588 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2589 			    ifr_buffer_get_length(ifr) - 1);
2590 			if (error) {
2591 				free(descrbuf, M_IFDESCR);
2592 				break;
2593 			}
2594 		}
2595 
2596 		sx_xlock(&ifdescr_sx);
2597 		odescrbuf = ifp->if_description;
2598 		ifp->if_description = descrbuf;
2599 		sx_xunlock(&ifdescr_sx);
2600 
2601 		getmicrotime(&ifp->if_lastchange);
2602 		free(odescrbuf, M_IFDESCR);
2603 		break;
2604 
2605 	case SIOCGIFFIB:
2606 		ifr->ifr_fib = ifp->if_fib;
2607 		break;
2608 
2609 	case SIOCSIFFIB:
2610 		error = priv_check(td, PRIV_NET_SETIFFIB);
2611 		if (error)
2612 			return (error);
2613 		if (ifr->ifr_fib >= rt_numfibs)
2614 			return (EINVAL);
2615 
2616 		ifp->if_fib = ifr->ifr_fib;
2617 		break;
2618 
2619 	case SIOCSIFFLAGS:
2620 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2621 		if (error)
2622 			return (error);
2623 		/*
2624 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2625 		 * check, so we don't need special handling here yet.
2626 		 */
2627 		new_flags = (ifr->ifr_flags & 0xffff) |
2628 		    (ifr->ifr_flagshigh << 16);
2629 		if (ifp->if_flags & IFF_UP &&
2630 		    (new_flags & IFF_UP) == 0) {
2631 			if_down(ifp);
2632 		} else if (new_flags & IFF_UP &&
2633 		    (ifp->if_flags & IFF_UP) == 0) {
2634 			do_ifup = 1;
2635 		}
2636 		/* See if permanently promiscuous mode bit is about to flip */
2637 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2638 			if (new_flags & IFF_PPROMISC)
2639 				ifp->if_flags |= IFF_PROMISC;
2640 			else if (ifp->if_pcount == 0)
2641 				ifp->if_flags &= ~IFF_PROMISC;
2642 			if (log_promisc_mode_change)
2643                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2644                                     ((new_flags & IFF_PPROMISC) ?
2645                                      "enabled" : "disabled"));
2646 		}
2647 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2648 			(new_flags &~ IFF_CANTCHANGE);
2649 		if (ifp->if_ioctl) {
2650 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2651 		}
2652 		if (do_ifup)
2653 			if_up(ifp);
2654 		getmicrotime(&ifp->if_lastchange);
2655 		break;
2656 
2657 	case SIOCSIFCAP:
2658 		error = priv_check(td, PRIV_NET_SETIFCAP);
2659 		if (error)
2660 			return (error);
2661 		if (ifp->if_ioctl == NULL)
2662 			return (EOPNOTSUPP);
2663 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2664 			return (EINVAL);
2665 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2666 		if (error == 0)
2667 			getmicrotime(&ifp->if_lastchange);
2668 		break;
2669 
2670 #ifdef MAC
2671 	case SIOCSIFMAC:
2672 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2673 		break;
2674 #endif
2675 
2676 	case SIOCSIFNAME:
2677 		error = priv_check(td, PRIV_NET_SETIFNAME);
2678 		if (error)
2679 			return (error);
2680 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2681 		    NULL);
2682 		if (error != 0)
2683 			return (error);
2684 		if (new_name[0] == '\0')
2685 			return (EINVAL);
2686 		if (new_name[IFNAMSIZ-1] != '\0') {
2687 			new_name[IFNAMSIZ-1] = '\0';
2688 			if (strlen(new_name) == IFNAMSIZ-1)
2689 				return (EINVAL);
2690 		}
2691 		if (ifunit(new_name) != NULL)
2692 			return (EEXIST);
2693 
2694 		/*
2695 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2696 		 * and there are numerous other races with the
2697 		 * ifunit() checks not being atomic with namespace
2698 		 * changes (renames, vmoves, if_attach, etc).
2699 		 */
2700 		ifp->if_flags |= IFF_RENAMING;
2701 
2702 		/* Announce the departure of the interface. */
2703 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2704 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2705 
2706 		if_printf(ifp, "changing name to '%s'\n", new_name);
2707 
2708 		IF_ADDR_WLOCK(ifp);
2709 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2710 		ifa = ifp->if_addr;
2711 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2712 		namelen = strlen(new_name);
2713 		onamelen = sdl->sdl_nlen;
2714 		/*
2715 		 * Move the address if needed.  This is safe because we
2716 		 * allocate space for a name of length IFNAMSIZ when we
2717 		 * create this in if_attach().
2718 		 */
2719 		if (namelen != onamelen) {
2720 			bcopy(sdl->sdl_data + onamelen,
2721 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2722 		}
2723 		bcopy(new_name, sdl->sdl_data, namelen);
2724 		sdl->sdl_nlen = namelen;
2725 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2726 		bzero(sdl->sdl_data, onamelen);
2727 		while (namelen != 0)
2728 			sdl->sdl_data[--namelen] = 0xff;
2729 		IF_ADDR_WUNLOCK(ifp);
2730 
2731 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2732 		/* Announce the return of the interface. */
2733 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2734 
2735 		ifp->if_flags &= ~IFF_RENAMING;
2736 		break;
2737 
2738 #ifdef VIMAGE
2739 	case SIOCSIFVNET:
2740 		error = priv_check(td, PRIV_NET_SETIFVNET);
2741 		if (error)
2742 			return (error);
2743 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2744 		break;
2745 #endif
2746 
2747 	case SIOCSIFMETRIC:
2748 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2749 		if (error)
2750 			return (error);
2751 		ifp->if_metric = ifr->ifr_metric;
2752 		getmicrotime(&ifp->if_lastchange);
2753 		break;
2754 
2755 	case SIOCSIFPHYS:
2756 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2757 		if (error)
2758 			return (error);
2759 		if (ifp->if_ioctl == NULL)
2760 			return (EOPNOTSUPP);
2761 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2762 		if (error == 0)
2763 			getmicrotime(&ifp->if_lastchange);
2764 		break;
2765 
2766 	case SIOCSIFMTU:
2767 	{
2768 		u_long oldmtu = ifp->if_mtu;
2769 
2770 		error = priv_check(td, PRIV_NET_SETIFMTU);
2771 		if (error)
2772 			return (error);
2773 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2774 			return (EINVAL);
2775 		if (ifp->if_ioctl == NULL)
2776 			return (EOPNOTSUPP);
2777 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2778 		if (error == 0) {
2779 			getmicrotime(&ifp->if_lastchange);
2780 			rt_ifmsg(ifp);
2781 #ifdef INET
2782 			NETDUMP_REINIT(ifp);
2783 #endif
2784 		}
2785 		/*
2786 		 * If the link MTU changed, do network layer specific procedure.
2787 		 */
2788 		if (ifp->if_mtu != oldmtu) {
2789 #ifdef INET6
2790 			nd6_setmtu(ifp);
2791 #endif
2792 			rt_updatemtu(ifp);
2793 		}
2794 		break;
2795 	}
2796 
2797 	case SIOCADDMULTI:
2798 	case SIOCDELMULTI:
2799 		if (cmd == SIOCADDMULTI)
2800 			error = priv_check(td, PRIV_NET_ADDMULTI);
2801 		else
2802 			error = priv_check(td, PRIV_NET_DELMULTI);
2803 		if (error)
2804 			return (error);
2805 
2806 		/* Don't allow group membership on non-multicast interfaces. */
2807 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2808 			return (EOPNOTSUPP);
2809 
2810 		/* Don't let users screw up protocols' entries. */
2811 		if (ifr->ifr_addr.sa_family != AF_LINK)
2812 			return (EINVAL);
2813 
2814 		if (cmd == SIOCADDMULTI) {
2815 			struct ifmultiaddr *ifma;
2816 
2817 			/*
2818 			 * Userland is only permitted to join groups once
2819 			 * via the if_addmulti() KPI, because it cannot hold
2820 			 * struct ifmultiaddr * between calls. It may also
2821 			 * lose a race while we check if the membership
2822 			 * already exists.
2823 			 */
2824 			IF_ADDR_RLOCK(ifp);
2825 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2826 			IF_ADDR_RUNLOCK(ifp);
2827 			if (ifma != NULL)
2828 				error = EADDRINUSE;
2829 			else
2830 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2831 		} else {
2832 			error = if_delmulti(ifp, &ifr->ifr_addr);
2833 		}
2834 		if (error == 0)
2835 			getmicrotime(&ifp->if_lastchange);
2836 		break;
2837 
2838 	case SIOCSIFPHYADDR:
2839 	case SIOCDIFPHYADDR:
2840 #ifdef INET6
2841 	case SIOCSIFPHYADDR_IN6:
2842 #endif
2843 	case SIOCSIFMEDIA:
2844 	case SIOCSIFGENERIC:
2845 		error = priv_check(td, PRIV_NET_HWIOCTL);
2846 		if (error)
2847 			return (error);
2848 		if (ifp->if_ioctl == NULL)
2849 			return (EOPNOTSUPP);
2850 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2851 		if (error == 0)
2852 			getmicrotime(&ifp->if_lastchange);
2853 		break;
2854 
2855 	case SIOCGIFSTATUS:
2856 	case SIOCGIFPSRCADDR:
2857 	case SIOCGIFPDSTADDR:
2858 	case SIOCGIFMEDIA:
2859 	case SIOCGIFXMEDIA:
2860 	case SIOCGIFGENERIC:
2861 	case SIOCGIFRSSKEY:
2862 	case SIOCGIFRSSHASH:
2863 		if (ifp->if_ioctl == NULL)
2864 			return (EOPNOTSUPP);
2865 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2866 		break;
2867 
2868 	case SIOCSIFLLADDR:
2869 		error = priv_check(td, PRIV_NET_SETLLADDR);
2870 		if (error)
2871 			return (error);
2872 		error = if_setlladdr(ifp,
2873 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2874 		break;
2875 
2876 	case SIOCGHWADDR:
2877 		error = if_gethwaddr(ifp, ifr);
2878 		break;
2879 
2880 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2881 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2882 		if (error)
2883 			return (error);
2884 		if ((error = if_addgroup(ifp,
2885 		    ifgr_group_get((struct ifgroupreq *)data))))
2886 			return (error);
2887 		break;
2888 
2889 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2890 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2891 			return (error);
2892 		break;
2893 
2894 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2895 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2896 		if (error)
2897 			return (error);
2898 		if ((error = if_delgroup(ifp,
2899 		    ifgr_group_get((struct ifgroupreq *)data))))
2900 			return (error);
2901 		break;
2902 
2903 	default:
2904 		error = ENOIOCTL;
2905 		break;
2906 	}
2907 	return (error);
2908 }
2909 
2910 #ifdef COMPAT_FREEBSD32
2911 struct ifconf32 {
2912 	int32_t	ifc_len;
2913 	union {
2914 		uint32_t	ifcu_buf;
2915 		uint32_t	ifcu_req;
2916 	} ifc_ifcu;
2917 };
2918 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2919 #endif
2920 
2921 #ifdef COMPAT_FREEBSD32
2922 static void
2923 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2924 {
2925 	struct ifmediareq32 *ifmr32;
2926 
2927 	ifmr32 = (struct ifmediareq32 *)data;
2928 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2929 	    sizeof(ifmr->ifm_name));
2930 	ifmr->ifm_current = ifmr32->ifm_current;
2931 	ifmr->ifm_mask = ifmr32->ifm_mask;
2932 	ifmr->ifm_status = ifmr32->ifm_status;
2933 	ifmr->ifm_active = ifmr32->ifm_active;
2934 	ifmr->ifm_count = ifmr32->ifm_count;
2935 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2936 }
2937 
2938 static void
2939 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2940 {
2941 	struct ifmediareq32 *ifmr32;
2942 
2943 	ifmr32 = (struct ifmediareq32 *)data;
2944 	ifmr32->ifm_current = ifmr->ifm_current;
2945 	ifmr32->ifm_mask = ifmr->ifm_mask;
2946 	ifmr32->ifm_status = ifmr->ifm_status;
2947 	ifmr32->ifm_active = ifmr->ifm_active;
2948 	ifmr32->ifm_count = ifmr->ifm_count;
2949 }
2950 #endif
2951 
2952 /*
2953  * Interface ioctls.
2954  */
2955 int
2956 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2957 {
2958 #ifdef COMPAT_FREEBSD32
2959 	caddr_t saved_data = NULL;
2960 	struct ifmediareq ifmr;
2961 #endif
2962 	struct ifmediareq *ifmrp;
2963 	struct ifnet *ifp;
2964 	struct ifreq *ifr;
2965 	int error;
2966 	int oif_flags;
2967 #ifdef VIMAGE
2968 	int shutdown;
2969 #endif
2970 
2971 	CURVNET_SET(so->so_vnet);
2972 #ifdef VIMAGE
2973 	/* Make sure the VNET is stable. */
2974 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2975 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2976 	if (shutdown) {
2977 		CURVNET_RESTORE();
2978 		return (EBUSY);
2979 	}
2980 #endif
2981 
2982 
2983 	switch (cmd) {
2984 	case SIOCGIFCONF:
2985 		error = ifconf(cmd, data);
2986 		CURVNET_RESTORE();
2987 		return (error);
2988 
2989 #ifdef COMPAT_FREEBSD32
2990 	case SIOCGIFCONF32:
2991 		{
2992 			struct ifconf32 *ifc32;
2993 			struct ifconf ifc;
2994 
2995 			ifc32 = (struct ifconf32 *)data;
2996 			ifc.ifc_len = ifc32->ifc_len;
2997 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2998 
2999 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
3000 			CURVNET_RESTORE();
3001 			if (error == 0)
3002 				ifc32->ifc_len = ifc.ifc_len;
3003 			return (error);
3004 		}
3005 #endif
3006 	}
3007 
3008 	ifmrp = NULL;
3009 #ifdef COMPAT_FREEBSD32
3010 	switch (cmd) {
3011 	case SIOCGIFMEDIA32:
3012 	case SIOCGIFXMEDIA32:
3013 		ifmrp = &ifmr;
3014 		ifmr_init(ifmrp, data);
3015 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3016 		saved_data = data;
3017 		data = (caddr_t)ifmrp;
3018 	}
3019 #endif
3020 
3021 	ifr = (struct ifreq *)data;
3022 	switch (cmd) {
3023 #ifdef VIMAGE
3024 	case SIOCSIFRVNET:
3025 		error = priv_check(td, PRIV_NET_SETIFVNET);
3026 		if (error == 0)
3027 			error = if_vmove_reclaim(td, ifr->ifr_name,
3028 			    ifr->ifr_jid);
3029 		goto out_noref;
3030 #endif
3031 	case SIOCIFCREATE:
3032 	case SIOCIFCREATE2:
3033 		error = priv_check(td, PRIV_NET_IFCREATE);
3034 		if (error == 0)
3035 			error = if_clone_create(ifr->ifr_name,
3036 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3037 			    ifr_data_get_ptr(ifr) : NULL);
3038 		goto out_noref;
3039 	case SIOCIFDESTROY:
3040 		error = priv_check(td, PRIV_NET_IFDESTROY);
3041 		if (error == 0)
3042 			error = if_clone_destroy(ifr->ifr_name);
3043 		goto out_noref;
3044 
3045 	case SIOCIFGCLONERS:
3046 		error = if_clone_list((struct if_clonereq *)data);
3047 		goto out_noref;
3048 
3049 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3050 		error = if_getgroupmembers((struct ifgroupreq *)data);
3051 		goto out_noref;
3052 
3053 #if defined(INET) || defined(INET6)
3054 	case SIOCSVH:
3055 	case SIOCGVH:
3056 		if (carp_ioctl_p == NULL)
3057 			error = EPROTONOSUPPORT;
3058 		else
3059 			error = (*carp_ioctl_p)(ifr, cmd, td);
3060 		goto out_noref;
3061 #endif
3062 	}
3063 
3064 	ifp = ifunit_ref(ifr->ifr_name);
3065 	if (ifp == NULL) {
3066 		error = ENXIO;
3067 		goto out_noref;
3068 	}
3069 
3070 	error = ifhwioctl(cmd, ifp, data, td);
3071 	if (error != ENOIOCTL)
3072 		goto out_ref;
3073 
3074 	oif_flags = ifp->if_flags;
3075 	if (so->so_proto == NULL) {
3076 		error = EOPNOTSUPP;
3077 		goto out_ref;
3078 	}
3079 
3080 	/*
3081 	 * Pass the request on to the socket control method, and if the
3082 	 * latter returns EOPNOTSUPP, directly to the interface.
3083 	 *
3084 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3085 	 * trust SIOCSIFADDR et al to come from an already privileged
3086 	 * layer, and do not perform any credentials checks or input
3087 	 * validation.
3088 	 */
3089 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3090 	    ifp, td));
3091 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3092 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3093 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3094 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3095 
3096 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3097 #ifdef INET6
3098 		if (ifp->if_flags & IFF_UP)
3099 			in6_if_up(ifp);
3100 #endif
3101 	}
3102 
3103 out_ref:
3104 	if_rele(ifp);
3105 out_noref:
3106 #ifdef COMPAT_FREEBSD32
3107 	if (ifmrp != NULL) {
3108 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3109 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3110 		     cmd));
3111 		data = saved_data;
3112 		ifmr_update(ifmrp, data);
3113 	}
3114 #endif
3115 	CURVNET_RESTORE();
3116 	return (error);
3117 }
3118 
3119 /*
3120  * The code common to handling reference counted flags,
3121  * e.g., in ifpromisc() and if_allmulti().
3122  * The "pflag" argument can specify a permanent mode flag to check,
3123  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3124  *
3125  * Only to be used on stack-owned flags, not driver-owned flags.
3126  */
3127 static int
3128 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3129 {
3130 	struct ifreq ifr;
3131 	int error;
3132 	int oldflags, oldcount;
3133 
3134 	/* Sanity checks to catch programming errors */
3135 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3136 	    ("%s: setting driver-owned flag %d", __func__, flag));
3137 
3138 	if (onswitch)
3139 		KASSERT(*refcount >= 0,
3140 		    ("%s: increment negative refcount %d for flag %d",
3141 		    __func__, *refcount, flag));
3142 	else
3143 		KASSERT(*refcount > 0,
3144 		    ("%s: decrement non-positive refcount %d for flag %d",
3145 		    __func__, *refcount, flag));
3146 
3147 	/* In case this mode is permanent, just touch refcount */
3148 	if (ifp->if_flags & pflag) {
3149 		*refcount += onswitch ? 1 : -1;
3150 		return (0);
3151 	}
3152 
3153 	/* Save ifnet parameters for if_ioctl() may fail */
3154 	oldcount = *refcount;
3155 	oldflags = ifp->if_flags;
3156 
3157 	/*
3158 	 * See if we aren't the only and touching refcount is enough.
3159 	 * Actually toggle interface flag if we are the first or last.
3160 	 */
3161 	if (onswitch) {
3162 		if ((*refcount)++)
3163 			return (0);
3164 		ifp->if_flags |= flag;
3165 	} else {
3166 		if (--(*refcount))
3167 			return (0);
3168 		ifp->if_flags &= ~flag;
3169 	}
3170 
3171 	/* Call down the driver since we've changed interface flags */
3172 	if (ifp->if_ioctl == NULL) {
3173 		error = EOPNOTSUPP;
3174 		goto recover;
3175 	}
3176 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3177 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3178 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3179 	if (error)
3180 		goto recover;
3181 	/* Notify userland that interface flags have changed */
3182 	rt_ifmsg(ifp);
3183 	return (0);
3184 
3185 recover:
3186 	/* Recover after driver error */
3187 	*refcount = oldcount;
3188 	ifp->if_flags = oldflags;
3189 	return (error);
3190 }
3191 
3192 /*
3193  * Set/clear promiscuous mode on interface ifp based on the truth value
3194  * of pswitch.  The calls are reference counted so that only the first
3195  * "on" request actually has an effect, as does the final "off" request.
3196  * Results are undefined if the "off" and "on" requests are not matched.
3197  */
3198 int
3199 ifpromisc(struct ifnet *ifp, int pswitch)
3200 {
3201 	int error;
3202 	int oldflags = ifp->if_flags;
3203 
3204 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3205 			   &ifp->if_pcount, pswitch);
3206 	/* If promiscuous mode status has changed, log a message */
3207 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3208             log_promisc_mode_change)
3209 		if_printf(ifp, "promiscuous mode %s\n",
3210 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3211 	return (error);
3212 }
3213 
3214 /*
3215  * Return interface configuration
3216  * of system.  List may be used
3217  * in later ioctl's (above) to get
3218  * other information.
3219  */
3220 /*ARGSUSED*/
3221 static int
3222 ifconf(u_long cmd, caddr_t data)
3223 {
3224 	struct ifconf *ifc = (struct ifconf *)data;
3225 	struct ifnet *ifp;
3226 	struct ifaddr *ifa;
3227 	struct ifreq ifr;
3228 	struct sbuf *sb;
3229 	int error, full = 0, valid_len, max_len;
3230 
3231 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3232 	max_len = MAXPHYS - 1;
3233 
3234 	/* Prevent hostile input from being able to crash the system */
3235 	if (ifc->ifc_len <= 0)
3236 		return (EINVAL);
3237 
3238 again:
3239 	if (ifc->ifc_len <= max_len) {
3240 		max_len = ifc->ifc_len;
3241 		full = 1;
3242 	}
3243 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3244 	max_len = 0;
3245 	valid_len = 0;
3246 
3247 	IFNET_RLOCK();
3248 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3249 		int addrs;
3250 
3251 		/*
3252 		 * Zero the ifr to make sure we don't disclose the contents
3253 		 * of the stack.
3254 		 */
3255 		memset(&ifr, 0, sizeof(ifr));
3256 
3257 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3258 		    >= sizeof(ifr.ifr_name)) {
3259 			sbuf_delete(sb);
3260 			IFNET_RUNLOCK();
3261 			return (ENAMETOOLONG);
3262 		}
3263 
3264 		addrs = 0;
3265 		IF_ADDR_RLOCK(ifp);
3266 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3267 			struct sockaddr *sa = ifa->ifa_addr;
3268 
3269 			if (prison_if(curthread->td_ucred, sa) != 0)
3270 				continue;
3271 			addrs++;
3272 			if (sa->sa_len <= sizeof(*sa)) {
3273 				if (sa->sa_len < sizeof(*sa)) {
3274 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3275 					    sizeof(ifr.ifr_ifru.ifru_addr));
3276 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3277 					    sa->sa_len);
3278 				} else
3279 					ifr.ifr_ifru.ifru_addr = *sa;
3280 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3281 				max_len += sizeof(ifr);
3282 			} else {
3283 				sbuf_bcat(sb, &ifr,
3284 				    offsetof(struct ifreq, ifr_addr));
3285 				max_len += offsetof(struct ifreq, ifr_addr);
3286 				sbuf_bcat(sb, sa, sa->sa_len);
3287 				max_len += sa->sa_len;
3288 			}
3289 
3290 			if (sbuf_error(sb) == 0)
3291 				valid_len = sbuf_len(sb);
3292 		}
3293 		IF_ADDR_RUNLOCK(ifp);
3294 		if (addrs == 0) {
3295 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3296 			max_len += sizeof(ifr);
3297 
3298 			if (sbuf_error(sb) == 0)
3299 				valid_len = sbuf_len(sb);
3300 		}
3301 	}
3302 	IFNET_RUNLOCK();
3303 
3304 	/*
3305 	 * If we didn't allocate enough space (uncommon), try again.  If
3306 	 * we have already allocated as much space as we are allowed,
3307 	 * return what we've got.
3308 	 */
3309 	if (valid_len != max_len && !full) {
3310 		sbuf_delete(sb);
3311 		goto again;
3312 	}
3313 
3314 	ifc->ifc_len = valid_len;
3315 	sbuf_finish(sb);
3316 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3317 	sbuf_delete(sb);
3318 	return (error);
3319 }
3320 
3321 /*
3322  * Just like ifpromisc(), but for all-multicast-reception mode.
3323  */
3324 int
3325 if_allmulti(struct ifnet *ifp, int onswitch)
3326 {
3327 
3328 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3329 }
3330 
3331 struct ifmultiaddr *
3332 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3333 {
3334 	struct ifmultiaddr *ifma;
3335 
3336 	IF_ADDR_LOCK_ASSERT(ifp);
3337 
3338 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3339 		if (sa->sa_family == AF_LINK) {
3340 			if (sa_dl_equal(ifma->ifma_addr, sa))
3341 				break;
3342 		} else {
3343 			if (sa_equal(ifma->ifma_addr, sa))
3344 				break;
3345 		}
3346 	}
3347 
3348 	return ifma;
3349 }
3350 
3351 /*
3352  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3353  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3354  * the ifnet multicast address list here, so the caller must do that and
3355  * other setup work (such as notifying the device driver).  The reference
3356  * count is initialized to 1.
3357  */
3358 static struct ifmultiaddr *
3359 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3360     int mflags)
3361 {
3362 	struct ifmultiaddr *ifma;
3363 	struct sockaddr *dupsa;
3364 
3365 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3366 	    M_ZERO);
3367 	if (ifma == NULL)
3368 		return (NULL);
3369 
3370 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3371 	if (dupsa == NULL) {
3372 		free(ifma, M_IFMADDR);
3373 		return (NULL);
3374 	}
3375 	bcopy(sa, dupsa, sa->sa_len);
3376 	ifma->ifma_addr = dupsa;
3377 
3378 	ifma->ifma_ifp = ifp;
3379 	ifma->ifma_refcount = 1;
3380 	ifma->ifma_protospec = NULL;
3381 
3382 	if (llsa == NULL) {
3383 		ifma->ifma_lladdr = NULL;
3384 		return (ifma);
3385 	}
3386 
3387 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3388 	if (dupsa == NULL) {
3389 		free(ifma->ifma_addr, M_IFMADDR);
3390 		free(ifma, M_IFMADDR);
3391 		return (NULL);
3392 	}
3393 	bcopy(llsa, dupsa, llsa->sa_len);
3394 	ifma->ifma_lladdr = dupsa;
3395 
3396 	return (ifma);
3397 }
3398 
3399 /*
3400  * if_freemulti: free ifmultiaddr structure and possibly attached related
3401  * addresses.  The caller is responsible for implementing reference
3402  * counting, notifying the driver, handling routing messages, and releasing
3403  * any dependent link layer state.
3404  */
3405 #ifdef MCAST_VERBOSE
3406 extern void kdb_backtrace(void);
3407 #endif
3408 static void
3409 if_freemulti_internal(struct ifmultiaddr *ifma)
3410 {
3411 
3412 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3413 	    ifma->ifma_refcount));
3414 
3415 	if (ifma->ifma_lladdr != NULL)
3416 		free(ifma->ifma_lladdr, M_IFMADDR);
3417 #ifdef MCAST_VERBOSE
3418 	kdb_backtrace();
3419 	printf("%s freeing ifma: %p\n", __func__, ifma);
3420 #endif
3421 	free(ifma->ifma_addr, M_IFMADDR);
3422 	free(ifma, M_IFMADDR);
3423 }
3424 
3425 static void
3426 if_destroymulti(epoch_context_t ctx)
3427 {
3428 	struct ifmultiaddr *ifma;
3429 
3430 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3431 	if_freemulti_internal(ifma);
3432 }
3433 
3434 void
3435 if_freemulti(struct ifmultiaddr *ifma)
3436 {
3437 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3438 	    ifma->ifma_refcount));
3439 
3440 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3441 }
3442 
3443 
3444 /*
3445  * Register an additional multicast address with a network interface.
3446  *
3447  * - If the address is already present, bump the reference count on the
3448  *   address and return.
3449  * - If the address is not link-layer, look up a link layer address.
3450  * - Allocate address structures for one or both addresses, and attach to the
3451  *   multicast address list on the interface.  If automatically adding a link
3452  *   layer address, the protocol address will own a reference to the link
3453  *   layer address, to be freed when it is freed.
3454  * - Notify the network device driver of an addition to the multicast address
3455  *   list.
3456  *
3457  * 'sa' points to caller-owned memory with the desired multicast address.
3458  *
3459  * 'retifma' will be used to return a pointer to the resulting multicast
3460  * address reference, if desired.
3461  */
3462 int
3463 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3464     struct ifmultiaddr **retifma)
3465 {
3466 	struct ifmultiaddr *ifma, *ll_ifma;
3467 	struct sockaddr *llsa;
3468 	struct sockaddr_dl sdl;
3469 	int error;
3470 
3471 #ifdef INET
3472 	IN_MULTI_LIST_UNLOCK_ASSERT();
3473 #endif
3474 #ifdef INET6
3475 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3476 #endif
3477 	/*
3478 	 * If the address is already present, return a new reference to it;
3479 	 * otherwise, allocate storage and set up a new address.
3480 	 */
3481 	IF_ADDR_WLOCK(ifp);
3482 	ifma = if_findmulti(ifp, sa);
3483 	if (ifma != NULL) {
3484 		ifma->ifma_refcount++;
3485 		if (retifma != NULL)
3486 			*retifma = ifma;
3487 		IF_ADDR_WUNLOCK(ifp);
3488 		return (0);
3489 	}
3490 
3491 	/*
3492 	 * The address isn't already present; resolve the protocol address
3493 	 * into a link layer address, and then look that up, bump its
3494 	 * refcount or allocate an ifma for that also.
3495 	 * Most link layer resolving functions returns address data which
3496 	 * fits inside default sockaddr_dl structure. However callback
3497 	 * can allocate another sockaddr structure, in that case we need to
3498 	 * free it later.
3499 	 */
3500 	llsa = NULL;
3501 	ll_ifma = NULL;
3502 	if (ifp->if_resolvemulti != NULL) {
3503 		/* Provide called function with buffer size information */
3504 		sdl.sdl_len = sizeof(sdl);
3505 		llsa = (struct sockaddr *)&sdl;
3506 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3507 		if (error)
3508 			goto unlock_out;
3509 	}
3510 
3511 	/*
3512 	 * Allocate the new address.  Don't hook it up yet, as we may also
3513 	 * need to allocate a link layer multicast address.
3514 	 */
3515 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3516 	if (ifma == NULL) {
3517 		error = ENOMEM;
3518 		goto free_llsa_out;
3519 	}
3520 
3521 	/*
3522 	 * If a link layer address is found, we'll need to see if it's
3523 	 * already present in the address list, or allocate is as well.
3524 	 * When this block finishes, the link layer address will be on the
3525 	 * list.
3526 	 */
3527 	if (llsa != NULL) {
3528 		ll_ifma = if_findmulti(ifp, llsa);
3529 		if (ll_ifma == NULL) {
3530 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3531 			if (ll_ifma == NULL) {
3532 				--ifma->ifma_refcount;
3533 				if_freemulti(ifma);
3534 				error = ENOMEM;
3535 				goto free_llsa_out;
3536 			}
3537 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3538 			    ifma_link);
3539 		} else
3540 			ll_ifma->ifma_refcount++;
3541 		ifma->ifma_llifma = ll_ifma;
3542 	}
3543 
3544 	/*
3545 	 * We now have a new multicast address, ifma, and possibly a new or
3546 	 * referenced link layer address.  Add the primary address to the
3547 	 * ifnet address list.
3548 	 */
3549 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3550 
3551 	if (retifma != NULL)
3552 		*retifma = ifma;
3553 
3554 	/*
3555 	 * Must generate the message while holding the lock so that 'ifma'
3556 	 * pointer is still valid.
3557 	 */
3558 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3559 	IF_ADDR_WUNLOCK(ifp);
3560 
3561 	/*
3562 	 * We are certain we have added something, so call down to the
3563 	 * interface to let them know about it.
3564 	 */
3565 	if (ifp->if_ioctl != NULL) {
3566 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3567 	}
3568 
3569 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3570 		link_free_sdl(llsa);
3571 
3572 	return (0);
3573 
3574 free_llsa_out:
3575 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3576 		link_free_sdl(llsa);
3577 
3578 unlock_out:
3579 	IF_ADDR_WUNLOCK(ifp);
3580 	return (error);
3581 }
3582 
3583 /*
3584  * Delete a multicast group membership by network-layer group address.
3585  *
3586  * Returns ENOENT if the entry could not be found. If ifp no longer
3587  * exists, results are undefined. This entry point should only be used
3588  * from subsystems which do appropriate locking to hold ifp for the
3589  * duration of the call.
3590  * Network-layer protocol domains must use if_delmulti_ifma().
3591  */
3592 int
3593 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3594 {
3595 	struct ifmultiaddr *ifma;
3596 	int lastref;
3597 #ifdef INVARIANTS
3598 	struct ifnet *oifp;
3599 
3600 	IFNET_RLOCK_NOSLEEP();
3601 	CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3602 		if (ifp == oifp)
3603 			break;
3604 	if (ifp != oifp)
3605 		ifp = NULL;
3606 	IFNET_RUNLOCK_NOSLEEP();
3607 
3608 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3609 #endif
3610 	if (ifp == NULL)
3611 		return (ENOENT);
3612 
3613 	IF_ADDR_WLOCK(ifp);
3614 	lastref = 0;
3615 	ifma = if_findmulti(ifp, sa);
3616 	if (ifma != NULL)
3617 		lastref = if_delmulti_locked(ifp, ifma, 0);
3618 	IF_ADDR_WUNLOCK(ifp);
3619 
3620 	if (ifma == NULL)
3621 		return (ENOENT);
3622 
3623 	if (lastref && ifp->if_ioctl != NULL) {
3624 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3625 	}
3626 
3627 	return (0);
3628 }
3629 
3630 /*
3631  * Delete all multicast group membership for an interface.
3632  * Should be used to quickly flush all multicast filters.
3633  */
3634 void
3635 if_delallmulti(struct ifnet *ifp)
3636 {
3637 	struct ifmultiaddr *ifma;
3638 	struct ifmultiaddr *next;
3639 
3640 	IF_ADDR_WLOCK(ifp);
3641 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3642 		if_delmulti_locked(ifp, ifma, 0);
3643 	IF_ADDR_WUNLOCK(ifp);
3644 }
3645 
3646 void
3647 if_delmulti_ifma(struct ifmultiaddr *ifma)
3648 {
3649 	if_delmulti_ifma_flags(ifma, 0);
3650 }
3651 
3652 /*
3653  * Delete a multicast group membership by group membership pointer.
3654  * Network-layer protocol domains must use this routine.
3655  *
3656  * It is safe to call this routine if the ifp disappeared.
3657  */
3658 void
3659 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3660 {
3661 	struct ifnet *ifp;
3662 	int lastref;
3663 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3664 #ifdef INET
3665 	IN_MULTI_LIST_UNLOCK_ASSERT();
3666 #endif
3667 	ifp = ifma->ifma_ifp;
3668 #ifdef DIAGNOSTIC
3669 	if (ifp == NULL) {
3670 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3671 	} else {
3672 		struct ifnet *oifp;
3673 
3674 		IFNET_RLOCK_NOSLEEP();
3675 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3676 			if (ifp == oifp)
3677 				break;
3678 		if (ifp != oifp) {
3679 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3680 			ifp = NULL;
3681 		}
3682 		IFNET_RUNLOCK_NOSLEEP();
3683 	}
3684 #endif
3685 	/*
3686 	 * If and only if the ifnet instance exists: Acquire the address lock.
3687 	 */
3688 	if (ifp != NULL)
3689 		IF_ADDR_WLOCK(ifp);
3690 
3691 	lastref = if_delmulti_locked(ifp, ifma, flags);
3692 
3693 	if (ifp != NULL) {
3694 		/*
3695 		 * If and only if the ifnet instance exists:
3696 		 *  Release the address lock.
3697 		 *  If the group was left: update the hardware hash filter.
3698 		 */
3699 		IF_ADDR_WUNLOCK(ifp);
3700 		if (lastref && ifp->if_ioctl != NULL) {
3701 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3702 		}
3703 	}
3704 }
3705 
3706 /*
3707  * Perform deletion of network-layer and/or link-layer multicast address.
3708  *
3709  * Return 0 if the reference count was decremented.
3710  * Return 1 if the final reference was released, indicating that the
3711  * hardware hash filter should be reprogrammed.
3712  */
3713 static int
3714 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3715 {
3716 	struct ifmultiaddr *ll_ifma;
3717 
3718 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3719 		KASSERT(ifma->ifma_ifp == ifp,
3720 		    ("%s: inconsistent ifp %p", __func__, ifp));
3721 		IF_ADDR_WLOCK_ASSERT(ifp);
3722 	}
3723 
3724 	ifp = ifma->ifma_ifp;
3725 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3726 
3727 	/*
3728 	 * If the ifnet is detaching, null out references to ifnet,
3729 	 * so that upper protocol layers will notice, and not attempt
3730 	 * to obtain locks for an ifnet which no longer exists. The
3731 	 * routing socket announcement must happen before the ifnet
3732 	 * instance is detached from the system.
3733 	 */
3734 	if (detaching) {
3735 #ifdef DIAGNOSTIC
3736 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3737 #endif
3738 		/*
3739 		 * ifp may already be nulled out if we are being reentered
3740 		 * to delete the ll_ifma.
3741 		 */
3742 		if (ifp != NULL) {
3743 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3744 			ifma->ifma_ifp = NULL;
3745 		}
3746 	}
3747 
3748 	if (--ifma->ifma_refcount > 0)
3749 		return 0;
3750 
3751 	if (ifp != NULL && detaching == 0)
3752 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3753 
3754 	/*
3755 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3756 	 * have been associated with it. Release it first if so.
3757 	 */
3758 	ll_ifma = ifma->ifma_llifma;
3759 	if (ll_ifma != NULL) {
3760 		KASSERT(ifma->ifma_lladdr != NULL,
3761 		    ("%s: llifma w/o lladdr", __func__));
3762 		if (detaching)
3763 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3764 		if (--ll_ifma->ifma_refcount == 0) {
3765 			if (ifp != NULL) {
3766 				CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3767 				    ifma_link);
3768 			}
3769 			if_freemulti(ll_ifma);
3770 		}
3771 	}
3772 #ifdef INVARIANTS
3773 	if (ifp) {
3774 		struct ifmultiaddr *ifmatmp;
3775 
3776 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3777 			MPASS(ifma != ifmatmp);
3778 	}
3779 #endif
3780 	if_freemulti(ifma);
3781 	/*
3782 	 * The last reference to this instance of struct ifmultiaddr
3783 	 * was released; the hardware should be notified of this change.
3784 	 */
3785 	return 1;
3786 }
3787 
3788 /*
3789  * Set the link layer address on an interface.
3790  *
3791  * At this time we only support certain types of interfaces,
3792  * and we don't allow the length of the address to change.
3793  *
3794  * Set noinline to be dtrace-friendly
3795  */
3796 __noinline int
3797 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3798 {
3799 	struct sockaddr_dl *sdl;
3800 	struct ifaddr *ifa;
3801 	struct ifreq ifr;
3802 	int rc;
3803 
3804 	rc = 0;
3805 	NET_EPOCH_ENTER();
3806 	ifa = ifp->if_addr;
3807 	if (ifa == NULL) {
3808 		rc = EINVAL;
3809 		goto out;
3810 	}
3811 
3812 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3813 	if (sdl == NULL) {
3814 		rc = EINVAL;
3815 		goto out;
3816 	}
3817 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3818 		rc = EINVAL;
3819 		goto out;
3820 	}
3821 	switch (ifp->if_type) {
3822 	case IFT_ETHER:
3823 	case IFT_XETHER:
3824 	case IFT_L2VLAN:
3825 	case IFT_BRIDGE:
3826 	case IFT_IEEE8023ADLAG:
3827 		bcopy(lladdr, LLADDR(sdl), len);
3828 		break;
3829 	default:
3830 		rc = ENODEV;
3831 		goto out;
3832 	}
3833 
3834 	/*
3835 	 * If the interface is already up, we need
3836 	 * to re-init it in order to reprogram its
3837 	 * address filter.
3838 	 */
3839 	if ((ifp->if_flags & IFF_UP) != 0) {
3840 		if (ifp->if_ioctl) {
3841 			ifp->if_flags &= ~IFF_UP;
3842 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3843 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3844 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3845 			ifp->if_flags |= IFF_UP;
3846 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3847 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3848 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3849 		}
3850 	}
3851 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3852  out:
3853 	NET_EPOCH_EXIT();
3854 	return (rc);
3855 }
3856 
3857 /*
3858  * Compat function for handling basic encapsulation requests.
3859  * Not converted stacks (FDDI, IB, ..) supports traditional
3860  * output model: ARP (and other similar L2 protocols) are handled
3861  * inside output routine, arpresolve/nd6_resolve() returns MAC
3862  * address instead of full prepend.
3863  *
3864  * This function creates calculated header==MAC for IPv4/IPv6 and
3865  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3866  * address families.
3867  */
3868 static int
3869 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3870 {
3871 
3872 	if (req->rtype != IFENCAP_LL)
3873 		return (EOPNOTSUPP);
3874 
3875 	if (req->bufsize < req->lladdr_len)
3876 		return (ENOMEM);
3877 
3878 	switch (req->family) {
3879 	case AF_INET:
3880 	case AF_INET6:
3881 		break;
3882 	default:
3883 		return (EAFNOSUPPORT);
3884 	}
3885 
3886 	/* Copy lladdr to storage as is */
3887 	memmove(req->buf, req->lladdr, req->lladdr_len);
3888 	req->bufsize = req->lladdr_len;
3889 	req->lladdr_off = 0;
3890 
3891 	return (0);
3892 }
3893 
3894 /*
3895  * Get the link layer address that was read from the hardware at attach.
3896  *
3897  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3898  * their component interfaces as IFT_IEEE8023ADLAG.
3899  */
3900 int
3901 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3902 {
3903 
3904 	if (ifp->if_hw_addr == NULL)
3905 		return (ENODEV);
3906 
3907 	switch (ifp->if_type) {
3908 	case IFT_ETHER:
3909 	case IFT_IEEE8023ADLAG:
3910 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3911 		return (0);
3912 	default:
3913 		return (ENODEV);
3914 	}
3915 }
3916 
3917 /*
3918  * The name argument must be a pointer to storage which will last as
3919  * long as the interface does.  For physical devices, the result of
3920  * device_get_name(dev) is a good choice and for pseudo-devices a
3921  * static string works well.
3922  */
3923 void
3924 if_initname(struct ifnet *ifp, const char *name, int unit)
3925 {
3926 	ifp->if_dname = name;
3927 	ifp->if_dunit = unit;
3928 	if (unit != IF_DUNIT_NONE)
3929 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3930 	else
3931 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3932 }
3933 
3934 int
3935 if_printf(struct ifnet *ifp, const char *fmt, ...)
3936 {
3937 	char if_fmt[256];
3938 	va_list ap;
3939 
3940 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3941 	va_start(ap, fmt);
3942 	vlog(LOG_INFO, if_fmt, ap);
3943 	va_end(ap);
3944 	return (0);
3945 }
3946 
3947 void
3948 if_start(struct ifnet *ifp)
3949 {
3950 
3951 	(*(ifp)->if_start)(ifp);
3952 }
3953 
3954 /*
3955  * Backwards compatibility interface for drivers
3956  * that have not implemented it
3957  */
3958 static int
3959 if_transmit(struct ifnet *ifp, struct mbuf *m)
3960 {
3961 	int error;
3962 
3963 	IFQ_HANDOFF(ifp, m, error);
3964 	return (error);
3965 }
3966 
3967 static void
3968 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3969 {
3970 
3971 	m_freem(m);
3972 }
3973 
3974 int
3975 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3976 {
3977 	int active = 0;
3978 
3979 	IF_LOCK(ifq);
3980 	if (_IF_QFULL(ifq)) {
3981 		IF_UNLOCK(ifq);
3982 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3983 		m_freem(m);
3984 		return (0);
3985 	}
3986 	if (ifp != NULL) {
3987 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3988 		if (m->m_flags & (M_BCAST|M_MCAST))
3989 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3990 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3991 	}
3992 	_IF_ENQUEUE(ifq, m);
3993 	IF_UNLOCK(ifq);
3994 	if (ifp != NULL && !active)
3995 		(*(ifp)->if_start)(ifp);
3996 	return (1);
3997 }
3998 
3999 void
4000 if_register_com_alloc(u_char type,
4001     if_com_alloc_t *a, if_com_free_t *f)
4002 {
4003 
4004 	KASSERT(if_com_alloc[type] == NULL,
4005 	    ("if_register_com_alloc: %d already registered", type));
4006 	KASSERT(if_com_free[type] == NULL,
4007 	    ("if_register_com_alloc: %d free already registered", type));
4008 
4009 	if_com_alloc[type] = a;
4010 	if_com_free[type] = f;
4011 }
4012 
4013 void
4014 if_deregister_com_alloc(u_char type)
4015 {
4016 
4017 	KASSERT(if_com_alloc[type] != NULL,
4018 	    ("if_deregister_com_alloc: %d not registered", type));
4019 	KASSERT(if_com_free[type] != NULL,
4020 	    ("if_deregister_com_alloc: %d free not registered", type));
4021 	if_com_alloc[type] = NULL;
4022 	if_com_free[type] = NULL;
4023 }
4024 
4025 /* API for driver access to network stack owned ifnet.*/
4026 uint64_t
4027 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4028 {
4029 	uint64_t oldbrate;
4030 
4031 	oldbrate = ifp->if_baudrate;
4032 	ifp->if_baudrate = baudrate;
4033 	return (oldbrate);
4034 }
4035 
4036 uint64_t
4037 if_getbaudrate(if_t ifp)
4038 {
4039 
4040 	return (((struct ifnet *)ifp)->if_baudrate);
4041 }
4042 
4043 int
4044 if_setcapabilities(if_t ifp, int capabilities)
4045 {
4046 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4047 	return (0);
4048 }
4049 
4050 int
4051 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4052 {
4053 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4054 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4055 
4056 	return (0);
4057 }
4058 
4059 int
4060 if_getcapabilities(if_t ifp)
4061 {
4062 	return ((struct ifnet *)ifp)->if_capabilities;
4063 }
4064 
4065 int
4066 if_setcapenable(if_t ifp, int capabilities)
4067 {
4068 	((struct ifnet *)ifp)->if_capenable = capabilities;
4069 	return (0);
4070 }
4071 
4072 int
4073 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4074 {
4075 	if(setcap)
4076 		((struct ifnet *)ifp)->if_capenable |= setcap;
4077 	if(clearcap)
4078 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4079 
4080 	return (0);
4081 }
4082 
4083 const char *
4084 if_getdname(if_t ifp)
4085 {
4086 	return ((struct ifnet *)ifp)->if_dname;
4087 }
4088 
4089 int
4090 if_togglecapenable(if_t ifp, int togglecap)
4091 {
4092 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4093 	return (0);
4094 }
4095 
4096 int
4097 if_getcapenable(if_t ifp)
4098 {
4099 	return ((struct ifnet *)ifp)->if_capenable;
4100 }
4101 
4102 /*
4103  * This is largely undesirable because it ties ifnet to a device, but does
4104  * provide flexiblity for an embedded product vendor. Should be used with
4105  * the understanding that it violates the interface boundaries, and should be
4106  * a last resort only.
4107  */
4108 int
4109 if_setdev(if_t ifp, void *dev)
4110 {
4111 	return (0);
4112 }
4113 
4114 int
4115 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4116 {
4117 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4118 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4119 
4120 	return (0);
4121 }
4122 
4123 int
4124 if_getdrvflags(if_t ifp)
4125 {
4126 	return ((struct ifnet *)ifp)->if_drv_flags;
4127 }
4128 
4129 int
4130 if_setdrvflags(if_t ifp, int flags)
4131 {
4132 	((struct ifnet *)ifp)->if_drv_flags = flags;
4133 	return (0);
4134 }
4135 
4136 
4137 int
4138 if_setflags(if_t ifp, int flags)
4139 {
4140 	((struct ifnet *)ifp)->if_flags = flags;
4141 	return (0);
4142 }
4143 
4144 int
4145 if_setflagbits(if_t ifp, int set, int clear)
4146 {
4147 	((struct ifnet *)ifp)->if_flags |= set;
4148 	((struct ifnet *)ifp)->if_flags &= ~clear;
4149 
4150 	return (0);
4151 }
4152 
4153 int
4154 if_getflags(if_t ifp)
4155 {
4156 	return ((struct ifnet *)ifp)->if_flags;
4157 }
4158 
4159 int
4160 if_clearhwassist(if_t ifp)
4161 {
4162 	((struct ifnet *)ifp)->if_hwassist = 0;
4163 	return (0);
4164 }
4165 
4166 int
4167 if_sethwassistbits(if_t ifp, int toset, int toclear)
4168 {
4169 	((struct ifnet *)ifp)->if_hwassist |= toset;
4170 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4171 
4172 	return (0);
4173 }
4174 
4175 int
4176 if_sethwassist(if_t ifp, int hwassist_bit)
4177 {
4178 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4179 	return (0);
4180 }
4181 
4182 int
4183 if_gethwassist(if_t ifp)
4184 {
4185 	return ((struct ifnet *)ifp)->if_hwassist;
4186 }
4187 
4188 int
4189 if_setmtu(if_t ifp, int mtu)
4190 {
4191 	((struct ifnet *)ifp)->if_mtu = mtu;
4192 	return (0);
4193 }
4194 
4195 int
4196 if_getmtu(if_t ifp)
4197 {
4198 	return ((struct ifnet *)ifp)->if_mtu;
4199 }
4200 
4201 int
4202 if_getmtu_family(if_t ifp, int family)
4203 {
4204 	struct domain *dp;
4205 
4206 	for (dp = domains; dp; dp = dp->dom_next) {
4207 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4208 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4209 	}
4210 
4211 	return (((struct ifnet *)ifp)->if_mtu);
4212 }
4213 
4214 int
4215 if_setsoftc(if_t ifp, void *softc)
4216 {
4217 	((struct ifnet *)ifp)->if_softc = softc;
4218 	return (0);
4219 }
4220 
4221 void *
4222 if_getsoftc(if_t ifp)
4223 {
4224 	return ((struct ifnet *)ifp)->if_softc;
4225 }
4226 
4227 void
4228 if_setrcvif(struct mbuf *m, if_t ifp)
4229 {
4230 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4231 }
4232 
4233 void
4234 if_setvtag(struct mbuf *m, uint16_t tag)
4235 {
4236 	m->m_pkthdr.ether_vtag = tag;
4237 }
4238 
4239 uint16_t
4240 if_getvtag(struct mbuf *m)
4241 {
4242 
4243 	return (m->m_pkthdr.ether_vtag);
4244 }
4245 
4246 int
4247 if_sendq_empty(if_t ifp)
4248 {
4249 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4250 }
4251 
4252 struct ifaddr *
4253 if_getifaddr(if_t ifp)
4254 {
4255 	return ((struct ifnet *)ifp)->if_addr;
4256 }
4257 
4258 int
4259 if_getamcount(if_t ifp)
4260 {
4261 	return ((struct ifnet *)ifp)->if_amcount;
4262 }
4263 
4264 
4265 int
4266 if_setsendqready(if_t ifp)
4267 {
4268 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4269 	return (0);
4270 }
4271 
4272 int
4273 if_setsendqlen(if_t ifp, int tx_desc_count)
4274 {
4275 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4276 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4277 
4278 	return (0);
4279 }
4280 
4281 int
4282 if_vlantrunkinuse(if_t ifp)
4283 {
4284 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4285 }
4286 
4287 int
4288 if_input(if_t ifp, struct mbuf* sendmp)
4289 {
4290 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4291 	return (0);
4292 
4293 }
4294 
4295 /* XXX */
4296 #ifndef ETH_ADDR_LEN
4297 #define ETH_ADDR_LEN 6
4298 #endif
4299 
4300 int
4301 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4302 {
4303 	struct ifmultiaddr *ifma;
4304 	uint8_t *lmta = (uint8_t *)mta;
4305 	int mcnt = 0;
4306 
4307 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4308 		if (ifma->ifma_addr->sa_family != AF_LINK)
4309 			continue;
4310 
4311 		if (mcnt == max)
4312 			break;
4313 
4314 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4315 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4316 		mcnt++;
4317 	}
4318 	*cnt = mcnt;
4319 
4320 	return (0);
4321 }
4322 
4323 int
4324 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4325 {
4326 	int error;
4327 
4328 	if_maddr_rlock(ifp);
4329 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4330 	if_maddr_runlock(ifp);
4331 	return (error);
4332 }
4333 
4334 int
4335 if_multiaddr_count(if_t ifp, int max)
4336 {
4337 	struct ifmultiaddr *ifma;
4338 	int count;
4339 
4340 	count = 0;
4341 	if_maddr_rlock(ifp);
4342 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4343 		if (ifma->ifma_addr->sa_family != AF_LINK)
4344 			continue;
4345 		count++;
4346 		if (count == max)
4347 			break;
4348 	}
4349 	if_maddr_runlock(ifp);
4350 	return (count);
4351 }
4352 
4353 int
4354 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4355 {
4356 	struct ifmultiaddr *ifma;
4357 	int cnt = 0;
4358 
4359 	if_maddr_rlock(ifp);
4360 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4361 		cnt += filter(arg, ifma, cnt);
4362 	if_maddr_runlock(ifp);
4363 	return (cnt);
4364 }
4365 
4366 struct mbuf *
4367 if_dequeue(if_t ifp)
4368 {
4369 	struct mbuf *m;
4370 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4371 
4372 	return (m);
4373 }
4374 
4375 int
4376 if_sendq_prepend(if_t ifp, struct mbuf *m)
4377 {
4378 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4379 	return (0);
4380 }
4381 
4382 int
4383 if_setifheaderlen(if_t ifp, int len)
4384 {
4385 	((struct ifnet *)ifp)->if_hdrlen = len;
4386 	return (0);
4387 }
4388 
4389 caddr_t
4390 if_getlladdr(if_t ifp)
4391 {
4392 	return (IF_LLADDR((struct ifnet *)ifp));
4393 }
4394 
4395 void *
4396 if_gethandle(u_char type)
4397 {
4398 	return (if_alloc(type));
4399 }
4400 
4401 void
4402 if_bpfmtap(if_t ifh, struct mbuf *m)
4403 {
4404 	struct ifnet *ifp = (struct ifnet *)ifh;
4405 
4406 	BPF_MTAP(ifp, m);
4407 }
4408 
4409 void
4410 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4411 {
4412 	struct ifnet *ifp = (struct ifnet *)ifh;
4413 
4414 	ETHER_BPF_MTAP(ifp, m);
4415 }
4416 
4417 void
4418 if_vlancap(if_t ifh)
4419 {
4420 	struct ifnet *ifp = (struct ifnet *)ifh;
4421 	VLAN_CAPABILITIES(ifp);
4422 }
4423 
4424 int
4425 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4426 {
4427 
4428 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4429         return (0);
4430 }
4431 
4432 int
4433 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4434 {
4435 
4436 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4437         return (0);
4438 }
4439 
4440 int
4441 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4442 {
4443 
4444 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4445         return (0);
4446 }
4447 
4448 u_int
4449 if_gethwtsomax(if_t ifp)
4450 {
4451 
4452 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4453 }
4454 
4455 u_int
4456 if_gethwtsomaxsegcount(if_t ifp)
4457 {
4458 
4459 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4460 }
4461 
4462 u_int
4463 if_gethwtsomaxsegsize(if_t ifp)
4464 {
4465 
4466 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4467 }
4468 
4469 void
4470 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4471 {
4472 	((struct ifnet *)ifp)->if_init = init_fn;
4473 }
4474 
4475 void
4476 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4477 {
4478 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4479 }
4480 
4481 void
4482 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4483 {
4484 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4485 }
4486 
4487 void
4488 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4489 {
4490 	((struct ifnet *)ifp)->if_transmit = start_fn;
4491 }
4492 
4493 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4494 {
4495 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4496 
4497 }
4498 
4499 void
4500 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4501 {
4502 
4503 	ifp->if_get_counter = fn;
4504 }
4505 
4506 /* Revisit these - These are inline functions originally. */
4507 int
4508 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4509 {
4510 	return drbr_inuse(ifh, br);
4511 }
4512 
4513 struct mbuf*
4514 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4515 {
4516 	return drbr_dequeue(ifh, br);
4517 }
4518 
4519 int
4520 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4521 {
4522 	return drbr_needs_enqueue(ifh, br);
4523 }
4524 
4525 int
4526 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4527 {
4528 	return drbr_enqueue(ifh, br, m);
4529 
4530 }
4531