xref: /freebsd/sys/net/if.c (revision 2c3632d14fe37fa35c262ee9fb66835be0a52621)
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_bpf.h"
36 #include "opt_inet6.h"
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/conf.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/domainset.h>
44 #include <sys/sbuf.h>
45 #include <sys/bus.h>
46 #include <sys/epoch.h>
47 #include <sys/mbuf.h>
48 #include <sys/systm.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/refcount.h>
57 #include <sys/module.h>
58 #include <sys/rwlock.h>
59 #include <sys/sockio.h>
60 #include <sys/syslog.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/taskqueue.h>
64 #include <sys/domain.h>
65 #include <sys/jail.h>
66 #include <sys/priv.h>
67 
68 #include <machine/stdarg.h>
69 #include <vm/uma.h>
70 
71 #include <net/bpf.h>
72 #include <net/ethernet.h>
73 #include <net/if.h>
74 #include <net/if_arp.h>
75 #include <net/if_clone.h>
76 #include <net/if_dl.h>
77 #include <net/if_types.h>
78 #include <net/if_var.h>
79 #include <net/if_media.h>
80 #include <net/if_vlan_var.h>
81 #include <net/radix.h>
82 #include <net/route.h>
83 #include <net/route/route_ctl.h>
84 #include <net/vnet.h>
85 
86 #if defined(INET) || defined(INET6)
87 #include <net/ethernet.h>
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip.h>
91 #include <netinet/ip_carp.h>
92 #ifdef INET
93 #include <net/debugnet.h>
94 #include <netinet/if_ether.h>
95 #endif /* INET */
96 #ifdef INET6
97 #include <netinet6/in6_var.h>
98 #include <netinet6/in6_ifattach.h>
99 #endif /* INET6 */
100 #endif /* INET || INET6 */
101 
102 #include <security/mac/mac_framework.h>
103 
104 /*
105  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
106  * and ifr_ifru when it is used in SIOCGIFCONF.
107  */
108 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
109     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
110 
111 __read_mostly epoch_t net_epoch_preempt;
112 #ifdef COMPAT_FREEBSD32
113 #include <sys/mount.h>
114 #include <compat/freebsd32/freebsd32.h>
115 
116 struct ifreq_buffer32 {
117 	uint32_t	length;		/* (size_t) */
118 	uint32_t	buffer;		/* (void *) */
119 };
120 
121 /*
122  * Interface request structure used for socket
123  * ioctl's.  All interface ioctl's must have parameter
124  * definitions which begin with ifr_name.  The
125  * remainder may be interface specific.
126  */
127 struct ifreq32 {
128 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
129 	union {
130 		struct sockaddr	ifru_addr;
131 		struct sockaddr	ifru_dstaddr;
132 		struct sockaddr	ifru_broadaddr;
133 		struct ifreq_buffer32 ifru_buffer;
134 		short		ifru_flags[2];
135 		short		ifru_index;
136 		int		ifru_jid;
137 		int		ifru_metric;
138 		int		ifru_mtu;
139 		int		ifru_phys;
140 		int		ifru_media;
141 		uint32_t	ifru_data;
142 		int		ifru_cap[2];
143 		u_int		ifru_fib;
144 		u_char		ifru_vlan_pcp;
145 	} ifr_ifru;
146 };
147 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
148 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
149     __offsetof(struct ifreq32, ifr_ifru));
150 
151 struct ifgroupreq32 {
152 	char	ifgr_name[IFNAMSIZ];
153 	u_int	ifgr_len;
154 	union {
155 		char		ifgru_group[IFNAMSIZ];
156 		uint32_t	ifgru_groups;
157 	} ifgr_ifgru;
158 };
159 
160 struct ifmediareq32 {
161 	char		ifm_name[IFNAMSIZ];
162 	int		ifm_current;
163 	int		ifm_mask;
164 	int		ifm_status;
165 	int		ifm_active;
166 	int		ifm_count;
167 	uint32_t	ifm_ulist;	/* (int *) */
168 };
169 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
170 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
171 
172 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
173     _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
174 #else /* !COMPAT_FREEBSD32 */
175 #define _CASE_IOC_IFGROUPREQ_32(cmd)
176 #endif /* !COMPAT_FREEBSD32 */
177 
178 #define CASE_IOC_IFGROUPREQ(cmd)	\
179     _CASE_IOC_IFGROUPREQ_32(cmd)	\
180     (cmd)
181 
182 union ifreq_union {
183 	struct ifreq	ifr;
184 #ifdef COMPAT_FREEBSD32
185 	struct ifreq32	ifr32;
186 #endif
187 };
188 
189 union ifgroupreq_union {
190 	struct ifgroupreq ifgr;
191 #ifdef COMPAT_FREEBSD32
192 	struct ifgroupreq32 ifgr32;
193 #endif
194 };
195 
196 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
197     "Link layers");
198 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
199     "Generic link-management");
200 
201 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
202     &ifqmaxlen, 0, "max send queue size");
203 
204 /* Log link state change events */
205 static int log_link_state_change = 1;
206 
207 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
208 	&log_link_state_change, 0,
209 	"log interface link state change events");
210 
211 /* Log promiscuous mode change events */
212 static int log_promisc_mode_change = 1;
213 
214 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
215 	&log_promisc_mode_change, 1,
216 	"log promiscuous mode change events");
217 
218 /* Interface description */
219 static unsigned int ifdescr_maxlen = 1024;
220 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
221 	&ifdescr_maxlen, 0,
222 	"administrative maximum length for interface description");
223 
224 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
225 
226 /* global sx for non-critical path ifdescr */
227 static struct sx ifdescr_sx;
228 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
229 
230 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
231 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
232 /* These are external hooks for CARP. */
233 void	(*carp_linkstate_p)(struct ifnet *ifp);
234 void	(*carp_demote_adj_p)(int, char *);
235 int	(*carp_master_p)(struct ifaddr *);
236 #if defined(INET) || defined(INET6)
237 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
238 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
239     const struct sockaddr *sa);
240 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
241 int	(*carp_attach_p)(struct ifaddr *, int);
242 void	(*carp_detach_p)(struct ifaddr *, bool);
243 #endif
244 #ifdef INET
245 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
246 #endif
247 #ifdef INET6
248 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
249 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
250     const struct in6_addr *taddr);
251 #endif
252 
253 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
254 
255 /*
256  * XXX: Style; these should be sorted alphabetically, and unprototyped
257  * static functions should be prototyped. Currently they are sorted by
258  * declaration order.
259  */
260 static void	if_attachdomain(void *);
261 static void	if_attachdomain1(struct ifnet *);
262 static int	ifconf(u_long, caddr_t);
263 static void	*if_grow(void);
264 static void	if_input_default(struct ifnet *, struct mbuf *);
265 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
266 static void	if_route(struct ifnet *, int flag, int fam);
267 static int	if_setflag(struct ifnet *, int, int, int *, int);
268 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
269 static void	if_unroute(struct ifnet *, int flag, int fam);
270 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
271 static void	do_link_state_change(void *, int);
272 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
273 static int	if_getgroupmembers(struct ifgroupreq *);
274 static void	if_delgroups(struct ifnet *);
275 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
276 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
277 static void	if_siocaddmulti(void *, int);
278 #ifdef VIMAGE
279 static int	if_vmove(struct ifnet *, struct vnet *);
280 #endif
281 
282 #ifdef INET6
283 /*
284  * XXX: declare here to avoid to include many inet6 related files..
285  * should be more generalized?
286  */
287 extern void	nd6_setmtu(struct ifnet *);
288 #endif
289 
290 /* ipsec helper hooks */
291 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
292 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
293 
294 VNET_DEFINE(int, if_index);
295 int	ifqmaxlen = IFQ_MAXLEN;
296 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
297 VNET_DEFINE(struct ifgrouphead, ifg_head);
298 
299 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
300 
301 /* Table of ifnet by index. */
302 VNET_DEFINE(struct ifnet **, ifindex_table);
303 
304 #define	V_if_indexlim		VNET(if_indexlim)
305 #define	V_ifindex_table		VNET(ifindex_table)
306 
307 /*
308  * The global network interface list (V_ifnet) and related state (such as
309  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
310  * an rwlock.  Either may be acquired shared to stablize the list, but both
311  * must be acquired writable to modify the list.  This model allows us to
312  * both stablize the interface list during interrupt thread processing, but
313  * also to stablize it over long-running ioctls, without introducing priority
314  * inversions and deadlocks.
315  */
316 struct rwlock ifnet_rwlock;
317 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
318 struct sx ifnet_sxlock;
319 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
320 
321 /*
322  * The allocation of network interfaces is a rather non-atomic affair; we
323  * need to select an index before we are ready to expose the interface for
324  * use, so will use this pointer value to indicate reservation.
325  */
326 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
327 
328 #ifdef VIMAGE
329 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
330     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
331 #endif
332 
333 static	if_com_alloc_t *if_com_alloc[256];
334 static	if_com_free_t *if_com_free[256];
335 
336 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
337 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
338 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
339 
340 struct ifnet *
341 ifnet_byindex(u_short idx)
342 {
343 	struct ifnet *ifp;
344 
345 	if (__predict_false(idx > V_if_index))
346 		return (NULL);
347 
348 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
349 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
350 }
351 
352 struct ifnet *
353 ifnet_byindex_ref(u_short idx)
354 {
355 	struct ifnet *ifp;
356 
357 	NET_EPOCH_ASSERT();
358 
359 	ifp = ifnet_byindex(idx);
360 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
361 		return (NULL);
362 	if_ref(ifp);
363 	return (ifp);
364 }
365 
366 /*
367  * Allocate an ifindex array entry; return 0 on success or an error on
368  * failure.
369  */
370 static u_short
371 ifindex_alloc(void **old)
372 {
373 	u_short idx;
374 
375 	IFNET_WLOCK_ASSERT();
376 	/*
377 	 * Try to find an empty slot below V_if_index.  If we fail, take the
378 	 * next slot.
379 	 */
380 	for (idx = 1; idx <= V_if_index; idx++) {
381 		if (V_ifindex_table[idx] == NULL)
382 			break;
383 	}
384 
385 	/* Catch if_index overflow. */
386 	if (idx >= V_if_indexlim) {
387 		*old = if_grow();
388 		return (USHRT_MAX);
389 	}
390 	if (idx > V_if_index)
391 		V_if_index = idx;
392 	return (idx);
393 }
394 
395 static void
396 ifindex_free_locked(u_short idx)
397 {
398 
399 	IFNET_WLOCK_ASSERT();
400 
401 	V_ifindex_table[idx] = NULL;
402 	while (V_if_index > 0 &&
403 	    V_ifindex_table[V_if_index] == NULL)
404 		V_if_index--;
405 }
406 
407 static void
408 ifindex_free(u_short idx)
409 {
410 
411 	IFNET_WLOCK();
412 	ifindex_free_locked(idx);
413 	IFNET_WUNLOCK();
414 }
415 
416 static void
417 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
418 {
419 
420 	V_ifindex_table[idx] = ifp;
421 }
422 
423 struct ifaddr *
424 ifaddr_byindex(u_short idx)
425 {
426 	struct ifnet *ifp;
427 	struct ifaddr *ifa = NULL;
428 
429 	NET_EPOCH_ASSERT();
430 
431 	ifp = ifnet_byindex(idx);
432 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
433 		ifa_ref(ifa);
434 	return (ifa);
435 }
436 
437 /*
438  * Network interface utility routines.
439  *
440  * Routines with ifa_ifwith* names take sockaddr *'s as
441  * parameters.
442  */
443 
444 static void
445 vnet_if_init(const void *unused __unused)
446 {
447 	void *old;
448 
449 	CK_STAILQ_INIT(&V_ifnet);
450 	CK_STAILQ_INIT(&V_ifg_head);
451 	IFNET_WLOCK();
452 	old = if_grow();				/* create initial table */
453 	IFNET_WUNLOCK();
454 	epoch_wait_preempt(net_epoch_preempt);
455 	free(old, M_IFNET);
456 	vnet_if_clone_init();
457 }
458 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
459     NULL);
460 
461 #ifdef VIMAGE
462 static void
463 vnet_if_uninit(const void *unused __unused)
464 {
465 
466 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
467 	    "not empty", __func__, __LINE__, &V_ifnet));
468 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
469 	    "not empty", __func__, __LINE__, &V_ifg_head));
470 
471 	free((caddr_t)V_ifindex_table, M_IFNET);
472 }
473 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
474     vnet_if_uninit, NULL);
475 
476 static void
477 vnet_if_return(const void *unused __unused)
478 {
479 	struct ifnet *ifp, *nifp;
480 
481 	/* Return all inherited interfaces to their parent vnets. */
482 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
483 		if (ifp->if_home_vnet != ifp->if_vnet)
484 			if_vmove(ifp, ifp->if_home_vnet);
485 	}
486 }
487 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
488     vnet_if_return, NULL);
489 #endif
490 
491 static void *
492 if_grow(void)
493 {
494 	int oldlim;
495 	u_int n;
496 	struct ifnet **e;
497 	void *old;
498 
499 	old = NULL;
500 	IFNET_WLOCK_ASSERT();
501 	oldlim = V_if_indexlim;
502 	IFNET_WUNLOCK();
503 	n = (oldlim << 1) * sizeof(*e);
504 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
505 	IFNET_WLOCK();
506 	if (V_if_indexlim != oldlim) {
507 		free(e, M_IFNET);
508 		return (NULL);
509 	}
510 	if (V_ifindex_table != NULL) {
511 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
512 		old = V_ifindex_table;
513 	}
514 	V_if_indexlim <<= 1;
515 	V_ifindex_table = e;
516 	return (old);
517 }
518 
519 /*
520  * Allocate a struct ifnet and an index for an interface.  A layer 2
521  * common structure will also be allocated if an allocation routine is
522  * registered for the passed type.
523  */
524 struct ifnet *
525 if_alloc_domain(u_char type, int numa_domain)
526 {
527 	struct ifnet *ifp;
528 	u_short idx;
529 	void *old;
530 
531 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
532 	if (numa_domain == IF_NODOM)
533 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
534 		    M_WAITOK | M_ZERO);
535 	else
536 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
537 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
538  restart:
539 	IFNET_WLOCK();
540 	idx = ifindex_alloc(&old);
541 	if (__predict_false(idx == USHRT_MAX)) {
542 		IFNET_WUNLOCK();
543 		epoch_wait_preempt(net_epoch_preempt);
544 		free(old, M_IFNET);
545 		goto restart;
546 	}
547 	ifnet_setbyindex(idx, IFNET_HOLD);
548 	IFNET_WUNLOCK();
549 	ifp->if_index = idx;
550 	ifp->if_type = type;
551 	ifp->if_alloctype = type;
552 	ifp->if_numa_domain = numa_domain;
553 #ifdef VIMAGE
554 	ifp->if_vnet = curvnet;
555 #endif
556 	if (if_com_alloc[type] != NULL) {
557 		ifp->if_l2com = if_com_alloc[type](type, ifp);
558 		if (ifp->if_l2com == NULL) {
559 			free(ifp, M_IFNET);
560 			ifindex_free(idx);
561 			return (NULL);
562 		}
563 	}
564 
565 	IF_ADDR_LOCK_INIT(ifp);
566 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
567 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
568 	ifp->if_afdata_initialized = 0;
569 	IF_AFDATA_LOCK_INIT(ifp);
570 	CK_STAILQ_INIT(&ifp->if_addrhead);
571 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
572 	CK_STAILQ_INIT(&ifp->if_groups);
573 #ifdef MAC
574 	mac_ifnet_init(ifp);
575 #endif
576 	ifq_init(&ifp->if_snd, ifp);
577 
578 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
579 	for (int i = 0; i < IFCOUNTERS; i++)
580 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
581 	ifp->if_get_counter = if_get_counter_default;
582 	ifp->if_pcp = IFNET_PCP_NONE;
583 	ifnet_setbyindex(ifp->if_index, ifp);
584 	return (ifp);
585 }
586 
587 struct ifnet *
588 if_alloc_dev(u_char type, device_t dev)
589 {
590 	int numa_domain;
591 
592 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
593 		return (if_alloc_domain(type, IF_NODOM));
594 	return (if_alloc_domain(type, numa_domain));
595 }
596 
597 struct ifnet *
598 if_alloc(u_char type)
599 {
600 
601 	return (if_alloc_domain(type, IF_NODOM));
602 }
603 /*
604  * Do the actual work of freeing a struct ifnet, and layer 2 common
605  * structure.  This call is made when the last reference to an
606  * interface is released.
607  */
608 static void
609 if_free_internal(struct ifnet *ifp)
610 {
611 
612 	KASSERT((ifp->if_flags & IFF_DYING),
613 	    ("if_free_internal: interface not dying"));
614 
615 	if (if_com_free[ifp->if_alloctype] != NULL)
616 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
617 		    ifp->if_alloctype);
618 
619 #ifdef MAC
620 	mac_ifnet_destroy(ifp);
621 #endif /* MAC */
622 	IF_AFDATA_DESTROY(ifp);
623 	IF_ADDR_LOCK_DESTROY(ifp);
624 	ifq_delete(&ifp->if_snd);
625 
626 	for (int i = 0; i < IFCOUNTERS; i++)
627 		counter_u64_free(ifp->if_counters[i]);
628 
629 	free(ifp->if_description, M_IFDESCR);
630 	free(ifp->if_hw_addr, M_IFADDR);
631 	free(ifp, M_IFNET);
632 }
633 
634 static void
635 if_destroy(epoch_context_t ctx)
636 {
637 	struct ifnet *ifp;
638 
639 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
640 	if_free_internal(ifp);
641 }
642 
643 /*
644  * Deregister an interface and free the associated storage.
645  */
646 void
647 if_free(struct ifnet *ifp)
648 {
649 
650 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
651 
652 	CURVNET_SET_QUIET(ifp->if_vnet);
653 	IFNET_WLOCK();
654 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
655 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
656 
657 	ifindex_free_locked(ifp->if_index);
658 	IFNET_WUNLOCK();
659 
660 	if (refcount_release(&ifp->if_refcount))
661 		NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
662 	CURVNET_RESTORE();
663 }
664 
665 /*
666  * Interfaces to keep an ifnet type-stable despite the possibility of the
667  * driver calling if_free().  If there are additional references, we defer
668  * freeing the underlying data structure.
669  */
670 void
671 if_ref(struct ifnet *ifp)
672 {
673 
674 	/* We don't assert the ifnet list lock here, but arguably should. */
675 	refcount_acquire(&ifp->if_refcount);
676 }
677 
678 void
679 if_rele(struct ifnet *ifp)
680 {
681 
682 	if (!refcount_release(&ifp->if_refcount))
683 		return;
684 	NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
685 }
686 
687 void
688 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
689 {
690 
691 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
692 
693 	if (ifq->ifq_maxlen == 0)
694 		ifq->ifq_maxlen = ifqmaxlen;
695 
696 	ifq->altq_type = 0;
697 	ifq->altq_disc = NULL;
698 	ifq->altq_flags &= ALTQF_CANTCHANGE;
699 	ifq->altq_tbr  = NULL;
700 	ifq->altq_ifp  = ifp;
701 }
702 
703 void
704 ifq_delete(struct ifaltq *ifq)
705 {
706 	mtx_destroy(&ifq->ifq_mtx);
707 }
708 
709 /*
710  * Perform generic interface initialization tasks and attach the interface
711  * to the list of "active" interfaces.  If vmove flag is set on entry
712  * to if_attach_internal(), perform only a limited subset of initialization
713  * tasks, given that we are moving from one vnet to another an ifnet which
714  * has already been fully initialized.
715  *
716  * Note that if_detach_internal() removes group membership unconditionally
717  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
718  * Thus, when if_vmove() is applied to a cloned interface, group membership
719  * is lost while a cloned one always joins a group whose name is
720  * ifc->ifc_name.  To recover this after if_detach_internal() and
721  * if_attach_internal(), the cloner should be specified to
722  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
723  * attempts to join a group whose name is ifc->ifc_name.
724  *
725  * XXX:
726  *  - The decision to return void and thus require this function to
727  *    succeed is questionable.
728  *  - We should probably do more sanity checking.  For instance we don't
729  *    do anything to insure if_xname is unique or non-empty.
730  */
731 void
732 if_attach(struct ifnet *ifp)
733 {
734 
735 	if_attach_internal(ifp, 0, NULL);
736 }
737 
738 /*
739  * Compute the least common TSO limit.
740  */
741 void
742 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
743 {
744 	/*
745 	 * 1) If there is no limit currently, take the limit from
746 	 * the network adapter.
747 	 *
748 	 * 2) If the network adapter has a limit below the current
749 	 * limit, apply it.
750 	 */
751 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
752 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
753 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
754 	}
755 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
756 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
757 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
758 	}
759 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
760 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
761 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
762 	}
763 }
764 
765 /*
766  * Update TSO limit of a network adapter.
767  *
768  * Returns zero if no change. Else non-zero.
769  */
770 int
771 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
772 {
773 	int retval = 0;
774 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
775 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
776 		retval++;
777 	}
778 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
779 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
780 		retval++;
781 	}
782 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
783 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
784 		retval++;
785 	}
786 	return (retval);
787 }
788 
789 static void
790 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
791 {
792 	unsigned socksize, ifasize;
793 	int namelen, masklen;
794 	struct sockaddr_dl *sdl;
795 	struct ifaddr *ifa;
796 
797 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
798 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
799 		    ifp->if_xname);
800 
801 #ifdef VIMAGE
802 	ifp->if_vnet = curvnet;
803 	if (ifp->if_home_vnet == NULL)
804 		ifp->if_home_vnet = curvnet;
805 #endif
806 
807 	if_addgroup(ifp, IFG_ALL);
808 
809 	/* Restore group membership for cloned interfaces. */
810 	if (vmove && ifc != NULL)
811 		if_clone_addgroup(ifp, ifc);
812 
813 	getmicrotime(&ifp->if_lastchange);
814 	ifp->if_epoch = time_uptime;
815 
816 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
817 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
818 	    ("transmit and qflush must both either be set or both be NULL"));
819 	if (ifp->if_transmit == NULL) {
820 		ifp->if_transmit = if_transmit;
821 		ifp->if_qflush = if_qflush;
822 	}
823 	if (ifp->if_input == NULL)
824 		ifp->if_input = if_input_default;
825 
826 	if (ifp->if_requestencap == NULL)
827 		ifp->if_requestencap = if_requestencap_default;
828 
829 	if (!vmove) {
830 #ifdef MAC
831 		mac_ifnet_create(ifp);
832 #endif
833 
834 		/*
835 		 * Create a Link Level name for this device.
836 		 */
837 		namelen = strlen(ifp->if_xname);
838 		/*
839 		 * Always save enough space for any possiable name so we
840 		 * can do a rename in place later.
841 		 */
842 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
843 		socksize = masklen + ifp->if_addrlen;
844 		if (socksize < sizeof(*sdl))
845 			socksize = sizeof(*sdl);
846 		socksize = roundup2(socksize, sizeof(long));
847 		ifasize = sizeof(*ifa) + 2 * socksize;
848 		ifa = ifa_alloc(ifasize, M_WAITOK);
849 		sdl = (struct sockaddr_dl *)(ifa + 1);
850 		sdl->sdl_len = socksize;
851 		sdl->sdl_family = AF_LINK;
852 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
853 		sdl->sdl_nlen = namelen;
854 		sdl->sdl_index = ifp->if_index;
855 		sdl->sdl_type = ifp->if_type;
856 		ifp->if_addr = ifa;
857 		ifa->ifa_ifp = ifp;
858 		ifa->ifa_addr = (struct sockaddr *)sdl;
859 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
860 		ifa->ifa_netmask = (struct sockaddr *)sdl;
861 		sdl->sdl_len = masklen;
862 		while (namelen != 0)
863 			sdl->sdl_data[--namelen] = 0xff;
864 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
865 		/* Reliably crash if used uninitialized. */
866 		ifp->if_broadcastaddr = NULL;
867 
868 		if (ifp->if_type == IFT_ETHER) {
869 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
870 			    M_WAITOK | M_ZERO);
871 		}
872 
873 #if defined(INET) || defined(INET6)
874 		/* Use defaults for TSO, if nothing is set */
875 		if (ifp->if_hw_tsomax == 0 &&
876 		    ifp->if_hw_tsomaxsegcount == 0 &&
877 		    ifp->if_hw_tsomaxsegsize == 0) {
878 			/*
879 			 * The TSO defaults needs to be such that an
880 			 * NFS mbuf list of 35 mbufs totalling just
881 			 * below 64K works and that a chain of mbufs
882 			 * can be defragged into at most 32 segments:
883 			 */
884 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
885 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
886 			ifp->if_hw_tsomaxsegcount = 35;
887 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
888 
889 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
890 			if (ifp->if_capabilities & IFCAP_TSO) {
891 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
892 				    ifp->if_hw_tsomax,
893 				    ifp->if_hw_tsomaxsegcount,
894 				    ifp->if_hw_tsomaxsegsize);
895 			}
896 		}
897 #endif
898 	}
899 #ifdef VIMAGE
900 	else {
901 		/*
902 		 * Update the interface index in the link layer address
903 		 * of the interface.
904 		 */
905 		for (ifa = ifp->if_addr; ifa != NULL;
906 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
907 			if (ifa->ifa_addr->sa_family == AF_LINK) {
908 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
909 				sdl->sdl_index = ifp->if_index;
910 			}
911 		}
912 	}
913 #endif
914 
915 	IFNET_WLOCK();
916 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
917 #ifdef VIMAGE
918 	curvnet->vnet_ifcnt++;
919 #endif
920 	IFNET_WUNLOCK();
921 
922 	if (domain_init_status >= 2)
923 		if_attachdomain1(ifp);
924 
925 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
926 	if (IS_DEFAULT_VNET(curvnet))
927 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
928 
929 	/* Announce the interface. */
930 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
931 }
932 
933 static void
934 if_epochalloc(void *dummy __unused)
935 {
936 
937 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
938 }
939 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
940 
941 static void
942 if_attachdomain(void *dummy)
943 {
944 	struct ifnet *ifp;
945 
946 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
947 		if_attachdomain1(ifp);
948 }
949 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
950     if_attachdomain, NULL);
951 
952 static void
953 if_attachdomain1(struct ifnet *ifp)
954 {
955 	struct domain *dp;
956 
957 	/*
958 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
959 	 * cannot lock ifp->if_afdata initialization, entirely.
960 	 */
961 	IF_AFDATA_LOCK(ifp);
962 	if (ifp->if_afdata_initialized >= domain_init_status) {
963 		IF_AFDATA_UNLOCK(ifp);
964 		log(LOG_WARNING, "%s called more than once on %s\n",
965 		    __func__, ifp->if_xname);
966 		return;
967 	}
968 	ifp->if_afdata_initialized = domain_init_status;
969 	IF_AFDATA_UNLOCK(ifp);
970 
971 	/* address family dependent data region */
972 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
973 	for (dp = domains; dp; dp = dp->dom_next) {
974 		if (dp->dom_ifattach)
975 			ifp->if_afdata[dp->dom_family] =
976 			    (*dp->dom_ifattach)(ifp);
977 	}
978 }
979 
980 /*
981  * Remove any unicast or broadcast network addresses from an interface.
982  */
983 void
984 if_purgeaddrs(struct ifnet *ifp)
985 {
986 	struct ifaddr *ifa;
987 
988 	while (1) {
989 		struct epoch_tracker et;
990 
991 		NET_EPOCH_ENTER(et);
992 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
993 			if (ifa->ifa_addr->sa_family != AF_LINK)
994 				break;
995 		}
996 		NET_EPOCH_EXIT(et);
997 
998 		if (ifa == NULL)
999 			break;
1000 #ifdef INET
1001 		/* XXX: Ugly!! ad hoc just for INET */
1002 		if (ifa->ifa_addr->sa_family == AF_INET) {
1003 			struct ifaliasreq ifr;
1004 
1005 			bzero(&ifr, sizeof(ifr));
1006 			ifr.ifra_addr = *ifa->ifa_addr;
1007 			if (ifa->ifa_dstaddr)
1008 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1009 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1010 			    NULL) == 0)
1011 				continue;
1012 		}
1013 #endif /* INET */
1014 #ifdef INET6
1015 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1016 			in6_purgeaddr(ifa);
1017 			/* ifp_addrhead is already updated */
1018 			continue;
1019 		}
1020 #endif /* INET6 */
1021 		IF_ADDR_WLOCK(ifp);
1022 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1023 		IF_ADDR_WUNLOCK(ifp);
1024 		ifa_free(ifa);
1025 	}
1026 }
1027 
1028 /*
1029  * Remove any multicast network addresses from an interface when an ifnet
1030  * is going away.
1031  */
1032 static void
1033 if_purgemaddrs(struct ifnet *ifp)
1034 {
1035 	struct ifmultiaddr *ifma;
1036 
1037 	IF_ADDR_WLOCK(ifp);
1038 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1039 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1040 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1041 		if_delmulti_locked(ifp, ifma, 1);
1042 	}
1043 	IF_ADDR_WUNLOCK(ifp);
1044 }
1045 
1046 /*
1047  * Detach an interface, removing it from the list of "active" interfaces.
1048  * If vmove flag is set on entry to if_detach_internal(), perform only a
1049  * limited subset of cleanup tasks, given that we are moving an ifnet from
1050  * one vnet to another, where it must be fully operational.
1051  *
1052  * XXXRW: There are some significant questions about event ordering, and
1053  * how to prevent things from starting to use the interface during detach.
1054  */
1055 void
1056 if_detach(struct ifnet *ifp)
1057 {
1058 
1059 	CURVNET_SET_QUIET(ifp->if_vnet);
1060 	if_detach_internal(ifp, 0, NULL);
1061 	CURVNET_RESTORE();
1062 }
1063 
1064 /*
1065  * The vmove flag, if set, indicates that we are called from a callpath
1066  * that is moving an interface to a different vnet instance.
1067  *
1068  * The shutdown flag, if set, indicates that we are called in the
1069  * process of shutting down a vnet instance.  Currently only the
1070  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1071  * on a vnet instance shutdown without this flag being set, e.g., when
1072  * the cloned interfaces are destoyed as first thing of teardown.
1073  */
1074 static int
1075 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1076 {
1077 	struct ifaddr *ifa;
1078 	int i;
1079 	struct domain *dp;
1080  	struct ifnet *iter;
1081  	int found = 0;
1082 #ifdef VIMAGE
1083 	bool shutdown;
1084 
1085 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1086 #endif
1087 	IFNET_WLOCK();
1088 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1089 		if (iter == ifp) {
1090 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1091 			if (!vmove)
1092 				ifp->if_flags |= IFF_DYING;
1093 			found = 1;
1094 			break;
1095 		}
1096 	IFNET_WUNLOCK();
1097 	if (!found) {
1098 		/*
1099 		 * While we would want to panic here, we cannot
1100 		 * guarantee that the interface is indeed still on
1101 		 * the list given we don't hold locks all the way.
1102 		 */
1103 		return (ENOENT);
1104 #if 0
1105 		if (vmove)
1106 			panic("%s: ifp=%p not on the ifnet tailq %p",
1107 			    __func__, ifp, &V_ifnet);
1108 		else
1109 			return; /* XXX this should panic as well? */
1110 #endif
1111 	}
1112 
1113 	/*
1114 	 * At this point we know the interface still was on the ifnet list
1115 	 * and we removed it so we are in a stable state.
1116 	 */
1117 #ifdef VIMAGE
1118 	curvnet->vnet_ifcnt--;
1119 #endif
1120 	epoch_wait_preempt(net_epoch_preempt);
1121 
1122 	/*
1123 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1124 	 * fixes issues about late destruction of multicast options
1125 	 * which lead to leave group calls, which in turn access the
1126 	 * belonging ifnet structure:
1127 	 */
1128 	epoch_drain_callbacks(net_epoch_preempt);
1129 
1130 	/*
1131 	 * In any case (destroy or vmove) detach us from the groups
1132 	 * and remove/wait for pending events on the taskq.
1133 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1134 	 */
1135 	if_delgroups(ifp);
1136 
1137 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1138 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1139 
1140 	/*
1141 	 * Check if this is a cloned interface or not. Must do even if
1142 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1143 	 * the if_clone_addgroup() will have a corrupted string overwise
1144 	 * from a gibberish pointer.
1145 	 */
1146 	if (vmove && ifcp != NULL)
1147 		*ifcp = if_clone_findifc(ifp);
1148 
1149 	if_down(ifp);
1150 
1151 #ifdef VIMAGE
1152 	/*
1153 	 * On VNET shutdown abort here as the stack teardown will do all
1154 	 * the work top-down for us.
1155 	 */
1156 	if (shutdown) {
1157 		/* Give interface users the chance to clean up. */
1158 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1159 
1160 		/*
1161 		 * In case of a vmove we are done here without error.
1162 		 * If we would signal an error it would lead to the same
1163 		 * abort as if we did not find the ifnet anymore.
1164 		 * if_detach() calls us in void context and does not care
1165 		 * about an early abort notification, so life is splendid :)
1166 		 */
1167 		goto finish_vnet_shutdown;
1168 	}
1169 #endif
1170 
1171 	/*
1172 	 * At this point we are not tearing down a VNET and are either
1173 	 * going to destroy or vmove the interface and have to cleanup
1174 	 * accordingly.
1175 	 */
1176 
1177 	/*
1178 	 * Remove routes and flush queues.
1179 	 */
1180 #ifdef ALTQ
1181 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1182 		altq_disable(&ifp->if_snd);
1183 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1184 		altq_detach(&ifp->if_snd);
1185 #endif
1186 
1187 	if_purgeaddrs(ifp);
1188 
1189 #ifdef INET
1190 	in_ifdetach(ifp);
1191 #endif
1192 
1193 #ifdef INET6
1194 	/*
1195 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1196 	 * before removing routing entries below, since IPv6 interface direct
1197 	 * routes are expected to be removed by the IPv6-specific kernel API.
1198 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1199 	 */
1200 	in6_ifdetach(ifp);
1201 #endif
1202 	if_purgemaddrs(ifp);
1203 
1204 	/* Announce that the interface is gone. */
1205 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1206 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1207 	if (IS_DEFAULT_VNET(curvnet))
1208 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1209 
1210 	if (!vmove) {
1211 		/*
1212 		 * Prevent further calls into the device driver via ifnet.
1213 		 */
1214 		if_dead(ifp);
1215 
1216 		/*
1217 		 * Clean up all addresses.
1218 		 */
1219 		IF_ADDR_WLOCK(ifp);
1220 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1221 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1222 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1223 			IF_ADDR_WUNLOCK(ifp);
1224 			ifa_free(ifa);
1225 		} else
1226 			IF_ADDR_WUNLOCK(ifp);
1227 	}
1228 
1229 	rt_flushifroutes(ifp);
1230 
1231 #ifdef VIMAGE
1232 finish_vnet_shutdown:
1233 #endif
1234 	/*
1235 	 * We cannot hold the lock over dom_ifdetach calls as they might
1236 	 * sleep, for example trying to drain a callout, thus open up the
1237 	 * theoretical race with re-attaching.
1238 	 */
1239 	IF_AFDATA_LOCK(ifp);
1240 	i = ifp->if_afdata_initialized;
1241 	ifp->if_afdata_initialized = 0;
1242 	IF_AFDATA_UNLOCK(ifp);
1243 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1244 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1245 			(*dp->dom_ifdetach)(ifp,
1246 			    ifp->if_afdata[dp->dom_family]);
1247 			ifp->if_afdata[dp->dom_family] = NULL;
1248 		}
1249 	}
1250 
1251 	return (0);
1252 }
1253 
1254 #ifdef VIMAGE
1255 /*
1256  * if_vmove() performs a limited version of if_detach() in current
1257  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1258  * An attempt is made to shrink if_index in current vnet, find an
1259  * unused if_index in target vnet and calls if_grow() if necessary,
1260  * and finally find an unused if_xname for the target vnet.
1261  */
1262 static int
1263 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1264 {
1265 	struct if_clone *ifc;
1266 #ifdef DEV_BPF
1267 	u_int bif_dlt, bif_hdrlen;
1268 #endif
1269 	void *old;
1270 	int rc;
1271 
1272 #ifdef DEV_BPF
1273  	/*
1274 	 * if_detach_internal() will call the eventhandler to notify
1275 	 * interface departure.  That will detach if_bpf.  We need to
1276 	 * safe the dlt and hdrlen so we can re-attach it later.
1277 	 */
1278 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1279 #endif
1280 
1281 	/*
1282 	 * Detach from current vnet, but preserve LLADDR info, do not
1283 	 * mark as dead etc. so that the ifnet can be reattached later.
1284 	 * If we cannot find it, we lost the race to someone else.
1285 	 */
1286 	rc = if_detach_internal(ifp, 1, &ifc);
1287 	if (rc != 0)
1288 		return (rc);
1289 
1290 	/*
1291 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1292 	 * the if_index for that vnet if possible.
1293 	 *
1294 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1295 	 * or we'd lock on one vnet and unlock on another.
1296 	 */
1297 	IFNET_WLOCK();
1298 	ifindex_free_locked(ifp->if_index);
1299 	IFNET_WUNLOCK();
1300 
1301 	/*
1302 	 * Perform interface-specific reassignment tasks, if provided by
1303 	 * the driver.
1304 	 */
1305 	if (ifp->if_reassign != NULL)
1306 		ifp->if_reassign(ifp, new_vnet, NULL);
1307 
1308 	/*
1309 	 * Switch to the context of the target vnet.
1310 	 */
1311 	CURVNET_SET_QUIET(new_vnet);
1312  restart:
1313 	IFNET_WLOCK();
1314 	ifp->if_index = ifindex_alloc(&old);
1315 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1316 		IFNET_WUNLOCK();
1317 		epoch_wait_preempt(net_epoch_preempt);
1318 		free(old, M_IFNET);
1319 		goto restart;
1320 	}
1321 	ifnet_setbyindex(ifp->if_index, ifp);
1322 	IFNET_WUNLOCK();
1323 
1324 	if_attach_internal(ifp, 1, ifc);
1325 
1326 #ifdef DEV_BPF
1327 	if (ifp->if_bpf == NULL)
1328 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1329 #endif
1330 
1331 	CURVNET_RESTORE();
1332 	return (0);
1333 }
1334 
1335 /*
1336  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1337  */
1338 static int
1339 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1340 {
1341 	struct prison *pr;
1342 	struct ifnet *difp;
1343 	int error;
1344 	bool shutdown;
1345 
1346 	/* Try to find the prison within our visibility. */
1347 	sx_slock(&allprison_lock);
1348 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1349 	sx_sunlock(&allprison_lock);
1350 	if (pr == NULL)
1351 		return (ENXIO);
1352 	prison_hold_locked(pr);
1353 	mtx_unlock(&pr->pr_mtx);
1354 
1355 	/* Do not try to move the iface from and to the same prison. */
1356 	if (pr->pr_vnet == ifp->if_vnet) {
1357 		prison_free(pr);
1358 		return (EEXIST);
1359 	}
1360 
1361 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1362 	/* XXX Lock interfaces to avoid races. */
1363 	CURVNET_SET_QUIET(pr->pr_vnet);
1364 	difp = ifunit(ifname);
1365 	if (difp != NULL) {
1366 		CURVNET_RESTORE();
1367 		prison_free(pr);
1368 		return (EEXIST);
1369 	}
1370 
1371 	/* Make sure the VNET is stable. */
1372 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1373 	if (shutdown) {
1374 		CURVNET_RESTORE();
1375 		prison_free(pr);
1376 		return (EBUSY);
1377 	}
1378 	CURVNET_RESTORE();
1379 
1380 	/* Move the interface into the child jail/vnet. */
1381 	error = if_vmove(ifp, pr->pr_vnet);
1382 
1383 	/* Report the new if_xname back to the userland on success. */
1384 	if (error == 0)
1385 		sprintf(ifname, "%s", ifp->if_xname);
1386 
1387 	prison_free(pr);
1388 	return (error);
1389 }
1390 
1391 static int
1392 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1393 {
1394 	struct prison *pr;
1395 	struct vnet *vnet_dst;
1396 	struct ifnet *ifp;
1397 	int error;
1398  	bool shutdown;
1399 
1400 	/* Try to find the prison within our visibility. */
1401 	sx_slock(&allprison_lock);
1402 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1403 	sx_sunlock(&allprison_lock);
1404 	if (pr == NULL)
1405 		return (ENXIO);
1406 	prison_hold_locked(pr);
1407 	mtx_unlock(&pr->pr_mtx);
1408 
1409 	/* Make sure the named iface exists in the source prison/vnet. */
1410 	CURVNET_SET(pr->pr_vnet);
1411 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1412 	if (ifp == NULL) {
1413 		CURVNET_RESTORE();
1414 		prison_free(pr);
1415 		return (ENXIO);
1416 	}
1417 
1418 	/* Do not try to move the iface from and to the same prison. */
1419 	vnet_dst = TD_TO_VNET(td);
1420 	if (vnet_dst == ifp->if_vnet) {
1421 		CURVNET_RESTORE();
1422 		prison_free(pr);
1423 		return (EEXIST);
1424 	}
1425 
1426 	/* Make sure the VNET is stable. */
1427 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1428 	if (shutdown) {
1429 		CURVNET_RESTORE();
1430 		prison_free(pr);
1431 		return (EBUSY);
1432 	}
1433 
1434 	/* Get interface back from child jail/vnet. */
1435 	error = if_vmove(ifp, vnet_dst);
1436 	CURVNET_RESTORE();
1437 
1438 	/* Report the new if_xname back to the userland on success. */
1439 	if (error == 0)
1440 		sprintf(ifname, "%s", ifp->if_xname);
1441 
1442 	prison_free(pr);
1443 	return (error);
1444 }
1445 #endif /* VIMAGE */
1446 
1447 /*
1448  * Add a group to an interface
1449  */
1450 int
1451 if_addgroup(struct ifnet *ifp, const char *groupname)
1452 {
1453 	struct ifg_list		*ifgl;
1454 	struct ifg_group	*ifg = NULL;
1455 	struct ifg_member	*ifgm;
1456 	int 			 new = 0;
1457 
1458 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1459 	    groupname[strlen(groupname) - 1] <= '9')
1460 		return (EINVAL);
1461 
1462 	IFNET_WLOCK();
1463 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1464 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1465 			IFNET_WUNLOCK();
1466 			return (EEXIST);
1467 		}
1468 
1469 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1470 	    	IFNET_WUNLOCK();
1471 		return (ENOMEM);
1472 	}
1473 
1474 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1475 		free(ifgl, M_TEMP);
1476 		IFNET_WUNLOCK();
1477 		return (ENOMEM);
1478 	}
1479 
1480 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1481 		if (!strcmp(ifg->ifg_group, groupname))
1482 			break;
1483 
1484 	if (ifg == NULL) {
1485 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1486 			free(ifgl, M_TEMP);
1487 			free(ifgm, M_TEMP);
1488 			IFNET_WUNLOCK();
1489 			return (ENOMEM);
1490 		}
1491 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1492 		ifg->ifg_refcnt = 0;
1493 		CK_STAILQ_INIT(&ifg->ifg_members);
1494 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1495 		new = 1;
1496 	}
1497 
1498 	ifg->ifg_refcnt++;
1499 	ifgl->ifgl_group = ifg;
1500 	ifgm->ifgm_ifp = ifp;
1501 
1502 	IF_ADDR_WLOCK(ifp);
1503 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1504 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1505 	IF_ADDR_WUNLOCK(ifp);
1506 
1507 	IFNET_WUNLOCK();
1508 
1509 	if (new)
1510 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1511 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1512 
1513 	return (0);
1514 }
1515 
1516 /*
1517  * Helper function to remove a group out of an interface.  Expects the global
1518  * ifnet lock to be write-locked, and drops it before returning.
1519  */
1520 static void
1521 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1522     const char *groupname)
1523 {
1524 	struct ifg_member *ifgm;
1525 	bool freeifgl;
1526 
1527 	IFNET_WLOCK_ASSERT();
1528 
1529 	IF_ADDR_WLOCK(ifp);
1530 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1531 	IF_ADDR_WUNLOCK(ifp);
1532 
1533 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1534 		if (ifgm->ifgm_ifp == ifp) {
1535 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1536 			    ifg_member, ifgm_next);
1537 			break;
1538 		}
1539 	}
1540 
1541 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1542 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1543 		    ifg_next);
1544 		freeifgl = true;
1545 	} else {
1546 		freeifgl = false;
1547 	}
1548 	IFNET_WUNLOCK();
1549 
1550 	epoch_wait_preempt(net_epoch_preempt);
1551 	if (freeifgl) {
1552 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1553 		free(ifgl->ifgl_group, M_TEMP);
1554 	}
1555 	free(ifgm, M_TEMP);
1556 	free(ifgl, M_TEMP);
1557 
1558 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1559 }
1560 
1561 /*
1562  * Remove a group from an interface
1563  */
1564 int
1565 if_delgroup(struct ifnet *ifp, const char *groupname)
1566 {
1567 	struct ifg_list *ifgl;
1568 
1569 	IFNET_WLOCK();
1570 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1571 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1572 			break;
1573 	if (ifgl == NULL) {
1574 		IFNET_WUNLOCK();
1575 		return (ENOENT);
1576 	}
1577 
1578 	_if_delgroup_locked(ifp, ifgl, groupname);
1579 
1580 	return (0);
1581 }
1582 
1583 /*
1584  * Remove an interface from all groups
1585  */
1586 static void
1587 if_delgroups(struct ifnet *ifp)
1588 {
1589 	struct ifg_list *ifgl;
1590 	char groupname[IFNAMSIZ];
1591 
1592 	IFNET_WLOCK();
1593 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1594 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1595 		_if_delgroup_locked(ifp, ifgl, groupname);
1596 		IFNET_WLOCK();
1597 	}
1598 	IFNET_WUNLOCK();
1599 }
1600 
1601 static char *
1602 ifgr_group_get(void *ifgrp)
1603 {
1604 	union ifgroupreq_union *ifgrup;
1605 
1606 	ifgrup = ifgrp;
1607 #ifdef COMPAT_FREEBSD32
1608 	if (SV_CURPROC_FLAG(SV_ILP32))
1609 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1610 #endif
1611 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1612 }
1613 
1614 static struct ifg_req *
1615 ifgr_groups_get(void *ifgrp)
1616 {
1617 	union ifgroupreq_union *ifgrup;
1618 
1619 	ifgrup = ifgrp;
1620 #ifdef COMPAT_FREEBSD32
1621 	if (SV_CURPROC_FLAG(SV_ILP32))
1622 		return ((struct ifg_req *)(uintptr_t)
1623 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1624 #endif
1625 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1626 }
1627 
1628 /*
1629  * Stores all groups from an interface in memory pointed to by ifgr.
1630  */
1631 static int
1632 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1633 {
1634 	int			 len, error;
1635 	struct ifg_list		*ifgl;
1636 	struct ifg_req		 ifgrq, *ifgp;
1637 
1638 	NET_EPOCH_ASSERT();
1639 
1640 	if (ifgr->ifgr_len == 0) {
1641 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1642 			ifgr->ifgr_len += sizeof(struct ifg_req);
1643 		return (0);
1644 	}
1645 
1646 	len = ifgr->ifgr_len;
1647 	ifgp = ifgr_groups_get(ifgr);
1648 	/* XXX: wire */
1649 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1650 		if (len < sizeof(ifgrq))
1651 			return (EINVAL);
1652 		bzero(&ifgrq, sizeof ifgrq);
1653 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1654 		    sizeof(ifgrq.ifgrq_group));
1655 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1656 			return (error);
1657 		len -= sizeof(ifgrq);
1658 		ifgp++;
1659 	}
1660 
1661 	return (0);
1662 }
1663 
1664 /*
1665  * Stores all members of a group in memory pointed to by igfr
1666  */
1667 static int
1668 if_getgroupmembers(struct ifgroupreq *ifgr)
1669 {
1670 	struct ifg_group	*ifg;
1671 	struct ifg_member	*ifgm;
1672 	struct ifg_req		 ifgrq, *ifgp;
1673 	int			 len, error;
1674 
1675 	IFNET_RLOCK();
1676 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1677 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1678 			break;
1679 	if (ifg == NULL) {
1680 		IFNET_RUNLOCK();
1681 		return (ENOENT);
1682 	}
1683 
1684 	if (ifgr->ifgr_len == 0) {
1685 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1686 			ifgr->ifgr_len += sizeof(ifgrq);
1687 		IFNET_RUNLOCK();
1688 		return (0);
1689 	}
1690 
1691 	len = ifgr->ifgr_len;
1692 	ifgp = ifgr_groups_get(ifgr);
1693 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1694 		if (len < sizeof(ifgrq)) {
1695 			IFNET_RUNLOCK();
1696 			return (EINVAL);
1697 		}
1698 		bzero(&ifgrq, sizeof ifgrq);
1699 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1700 		    sizeof(ifgrq.ifgrq_member));
1701 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1702 			IFNET_RUNLOCK();
1703 			return (error);
1704 		}
1705 		len -= sizeof(ifgrq);
1706 		ifgp++;
1707 	}
1708 	IFNET_RUNLOCK();
1709 
1710 	return (0);
1711 }
1712 
1713 /*
1714  * Return counter values from counter(9)s stored in ifnet.
1715  */
1716 uint64_t
1717 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1718 {
1719 
1720 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1721 
1722 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1723 }
1724 
1725 /*
1726  * Increase an ifnet counter. Usually used for counters shared
1727  * between the stack and a driver, but function supports them all.
1728  */
1729 void
1730 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1731 {
1732 
1733 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1734 
1735 	counter_u64_add(ifp->if_counters[cnt], inc);
1736 }
1737 
1738 /*
1739  * Copy data from ifnet to userland API structure if_data.
1740  */
1741 void
1742 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1743 {
1744 
1745 	ifd->ifi_type = ifp->if_type;
1746 	ifd->ifi_physical = 0;
1747 	ifd->ifi_addrlen = ifp->if_addrlen;
1748 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1749 	ifd->ifi_link_state = ifp->if_link_state;
1750 	ifd->ifi_vhid = 0;
1751 	ifd->ifi_datalen = sizeof(struct if_data);
1752 	ifd->ifi_mtu = ifp->if_mtu;
1753 	ifd->ifi_metric = ifp->if_metric;
1754 	ifd->ifi_baudrate = ifp->if_baudrate;
1755 	ifd->ifi_hwassist = ifp->if_hwassist;
1756 	ifd->ifi_epoch = ifp->if_epoch;
1757 	ifd->ifi_lastchange = ifp->if_lastchange;
1758 
1759 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1760 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1761 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1762 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1763 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1764 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1765 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1766 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1767 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1768 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1769 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1770 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1771 }
1772 
1773 /*
1774  * Initialization, destruction and refcounting functions for ifaddrs.
1775  */
1776 struct ifaddr *
1777 ifa_alloc(size_t size, int flags)
1778 {
1779 	struct ifaddr *ifa;
1780 
1781 	KASSERT(size >= sizeof(struct ifaddr),
1782 	    ("%s: invalid size %zu", __func__, size));
1783 
1784 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1785 	if (ifa == NULL)
1786 		return (NULL);
1787 
1788 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1789 		goto fail;
1790 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1791 		goto fail;
1792 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1793 		goto fail;
1794 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1795 		goto fail;
1796 
1797 	refcount_init(&ifa->ifa_refcnt, 1);
1798 
1799 	return (ifa);
1800 
1801 fail:
1802 	/* free(NULL) is okay */
1803 	counter_u64_free(ifa->ifa_opackets);
1804 	counter_u64_free(ifa->ifa_ipackets);
1805 	counter_u64_free(ifa->ifa_obytes);
1806 	counter_u64_free(ifa->ifa_ibytes);
1807 	free(ifa, M_IFADDR);
1808 
1809 	return (NULL);
1810 }
1811 
1812 void
1813 ifa_ref(struct ifaddr *ifa)
1814 {
1815 
1816 	refcount_acquire(&ifa->ifa_refcnt);
1817 }
1818 
1819 static void
1820 ifa_destroy(epoch_context_t ctx)
1821 {
1822 	struct ifaddr *ifa;
1823 
1824 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1825 	counter_u64_free(ifa->ifa_opackets);
1826 	counter_u64_free(ifa->ifa_ipackets);
1827 	counter_u64_free(ifa->ifa_obytes);
1828 	counter_u64_free(ifa->ifa_ibytes);
1829 	free(ifa, M_IFADDR);
1830 }
1831 
1832 void
1833 ifa_free(struct ifaddr *ifa)
1834 {
1835 
1836 	if (refcount_release(&ifa->ifa_refcnt))
1837 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1838 }
1839 
1840 static int
1841 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1842     struct sockaddr *ia)
1843 {
1844 	struct rib_cmd_info rc;
1845 	struct epoch_tracker et;
1846 	int error;
1847 	struct rt_addrinfo info;
1848 	struct sockaddr_dl null_sdl;
1849 	struct ifnet *ifp;
1850 	struct ifaddr *rti_ifa = NULL;
1851 
1852 	ifp = ifa->ifa_ifp;
1853 
1854 	NET_EPOCH_ENTER(et);
1855 	bzero(&info, sizeof(info));
1856 	if (cmd != RTM_DELETE)
1857 		info.rti_ifp = V_loif;
1858 	if (cmd == RTM_ADD) {
1859 		/* explicitly specify (loopback) ifa */
1860 		if (info.rti_ifp != NULL) {
1861 			rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1862 			if (rti_ifa != NULL)
1863 				ifa_ref(rti_ifa);
1864 			info.rti_ifa = rti_ifa;
1865 		}
1866 	}
1867 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1868 	info.rti_info[RTAX_DST] = ia;
1869 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1870 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1871 
1872 	error = rib_action(ifp->if_fib, cmd, &info, &rc);
1873 	NET_EPOCH_EXIT(et);
1874 
1875 	if (rti_ifa != NULL)
1876 		ifa_free(rti_ifa);
1877 
1878 	if (error == 0 ||
1879 	    (cmd == RTM_ADD && error == EEXIST) ||
1880 	    (cmd == RTM_DELETE && (error == ENOENT || error == ESRCH)))
1881 		return (error);
1882 
1883 	log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1884 		__func__, otype, if_name(ifp), error);
1885 
1886 	return (error);
1887 }
1888 
1889 int
1890 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1891 {
1892 
1893 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1894 }
1895 
1896 int
1897 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1898 {
1899 
1900 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1901 }
1902 
1903 int
1904 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1905 {
1906 
1907 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1908 }
1909 
1910 /*
1911  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1912  * structs used to represent other address families, it is necessary
1913  * to perform a different comparison.
1914  */
1915 
1916 #define	sa_dl_equal(a1, a2)	\
1917 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1918 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1919 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1920 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1921 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1922 
1923 /*
1924  * Locate an interface based on a complete address.
1925  */
1926 /*ARGSUSED*/
1927 struct ifaddr *
1928 ifa_ifwithaddr(const struct sockaddr *addr)
1929 {
1930 	struct ifnet *ifp;
1931 	struct ifaddr *ifa;
1932 
1933 	NET_EPOCH_ASSERT();
1934 
1935 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1936 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1937 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1938 				continue;
1939 			if (sa_equal(addr, ifa->ifa_addr)) {
1940 				goto done;
1941 			}
1942 			/* IP6 doesn't have broadcast */
1943 			if ((ifp->if_flags & IFF_BROADCAST) &&
1944 			    ifa->ifa_broadaddr &&
1945 			    ifa->ifa_broadaddr->sa_len != 0 &&
1946 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1947 				goto done;
1948 			}
1949 		}
1950 	}
1951 	ifa = NULL;
1952 done:
1953 	return (ifa);
1954 }
1955 
1956 int
1957 ifa_ifwithaddr_check(const struct sockaddr *addr)
1958 {
1959 	struct epoch_tracker et;
1960 	int rc;
1961 
1962 	NET_EPOCH_ENTER(et);
1963 	rc = (ifa_ifwithaddr(addr) != NULL);
1964 	NET_EPOCH_EXIT(et);
1965 	return (rc);
1966 }
1967 
1968 /*
1969  * Locate an interface based on the broadcast address.
1970  */
1971 /* ARGSUSED */
1972 struct ifaddr *
1973 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1974 {
1975 	struct ifnet *ifp;
1976 	struct ifaddr *ifa;
1977 
1978 	NET_EPOCH_ASSERT();
1979 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1980 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1981 			continue;
1982 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1983 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1984 				continue;
1985 			if ((ifp->if_flags & IFF_BROADCAST) &&
1986 			    ifa->ifa_broadaddr &&
1987 			    ifa->ifa_broadaddr->sa_len != 0 &&
1988 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1989 				goto done;
1990 			}
1991 		}
1992 	}
1993 	ifa = NULL;
1994 done:
1995 	return (ifa);
1996 }
1997 
1998 /*
1999  * Locate the point to point interface with a given destination address.
2000  */
2001 /*ARGSUSED*/
2002 struct ifaddr *
2003 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2004 {
2005 	struct ifnet *ifp;
2006 	struct ifaddr *ifa;
2007 
2008 	NET_EPOCH_ASSERT();
2009 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2010 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2011 			continue;
2012 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2013 			continue;
2014 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2015 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2016 				continue;
2017 			if (ifa->ifa_dstaddr != NULL &&
2018 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2019 				goto done;
2020 			}
2021 		}
2022 	}
2023 	ifa = NULL;
2024 done:
2025 	return (ifa);
2026 }
2027 
2028 /*
2029  * Find an interface on a specific network.  If many, choice
2030  * is most specific found.
2031  */
2032 struct ifaddr *
2033 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2034 {
2035 	struct ifnet *ifp;
2036 	struct ifaddr *ifa;
2037 	struct ifaddr *ifa_maybe = NULL;
2038 	u_int af = addr->sa_family;
2039 	const char *addr_data = addr->sa_data, *cplim;
2040 
2041 	NET_EPOCH_ASSERT();
2042 	/*
2043 	 * AF_LINK addresses can be looked up directly by their index number,
2044 	 * so do that if we can.
2045 	 */
2046 	if (af == AF_LINK) {
2047 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2048 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2049 		return (ifaddr_byindex(sdl->sdl_index));
2050 	}
2051 
2052 	/*
2053 	 * Scan though each interface, looking for ones that have addresses
2054 	 * in this address family and the requested fib.
2055 	 */
2056 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2057 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2058 			continue;
2059 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2060 			const char *cp, *cp2, *cp3;
2061 
2062 			if (ifa->ifa_addr->sa_family != af)
2063 next:				continue;
2064 			if (af == AF_INET &&
2065 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2066 				/*
2067 				 * This is a bit broken as it doesn't
2068 				 * take into account that the remote end may
2069 				 * be a single node in the network we are
2070 				 * looking for.
2071 				 * The trouble is that we don't know the
2072 				 * netmask for the remote end.
2073 				 */
2074 				if (ifa->ifa_dstaddr != NULL &&
2075 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2076 					goto done;
2077 				}
2078 			} else {
2079 				/*
2080 				 * Scan all the bits in the ifa's address.
2081 				 * If a bit dissagrees with what we are
2082 				 * looking for, mask it with the netmask
2083 				 * to see if it really matters.
2084 				 * (A byte at a time)
2085 				 */
2086 				if (ifa->ifa_netmask == 0)
2087 					continue;
2088 				cp = addr_data;
2089 				cp2 = ifa->ifa_addr->sa_data;
2090 				cp3 = ifa->ifa_netmask->sa_data;
2091 				cplim = ifa->ifa_netmask->sa_len
2092 					+ (char *)ifa->ifa_netmask;
2093 				while (cp3 < cplim)
2094 					if ((*cp++ ^ *cp2++) & *cp3++)
2095 						goto next; /* next address! */
2096 				/*
2097 				 * If the netmask of what we just found
2098 				 * is more specific than what we had before
2099 				 * (if we had one), or if the virtual status
2100 				 * of new prefix is better than of the old one,
2101 				 * then remember the new one before continuing
2102 				 * to search for an even better one.
2103 				 */
2104 				if (ifa_maybe == NULL ||
2105 				    ifa_preferred(ifa_maybe, ifa) ||
2106 				    rn_refines((caddr_t)ifa->ifa_netmask,
2107 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2108 					ifa_maybe = ifa;
2109 				}
2110 			}
2111 		}
2112 	}
2113 	ifa = ifa_maybe;
2114 	ifa_maybe = NULL;
2115 done:
2116 	return (ifa);
2117 }
2118 
2119 /*
2120  * Find an interface address specific to an interface best matching
2121  * a given address.
2122  */
2123 struct ifaddr *
2124 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2125 {
2126 	struct ifaddr *ifa;
2127 	const char *cp, *cp2, *cp3;
2128 	char *cplim;
2129 	struct ifaddr *ifa_maybe = NULL;
2130 	u_int af = addr->sa_family;
2131 
2132 	if (af >= AF_MAX)
2133 		return (NULL);
2134 
2135 	NET_EPOCH_ASSERT();
2136 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2137 		if (ifa->ifa_addr->sa_family != af)
2138 			continue;
2139 		if (ifa_maybe == NULL)
2140 			ifa_maybe = ifa;
2141 		if (ifa->ifa_netmask == 0) {
2142 			if (sa_equal(addr, ifa->ifa_addr) ||
2143 			    (ifa->ifa_dstaddr &&
2144 			    sa_equal(addr, ifa->ifa_dstaddr)))
2145 				goto done;
2146 			continue;
2147 		}
2148 		if (ifp->if_flags & IFF_POINTOPOINT) {
2149 			if (sa_equal(addr, ifa->ifa_dstaddr))
2150 				goto done;
2151 		} else {
2152 			cp = addr->sa_data;
2153 			cp2 = ifa->ifa_addr->sa_data;
2154 			cp3 = ifa->ifa_netmask->sa_data;
2155 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2156 			for (; cp3 < cplim; cp3++)
2157 				if ((*cp++ ^ *cp2++) & *cp3)
2158 					break;
2159 			if (cp3 == cplim)
2160 				goto done;
2161 		}
2162 	}
2163 	ifa = ifa_maybe;
2164 done:
2165 	return (ifa);
2166 }
2167 
2168 /*
2169  * See whether new ifa is better than current one:
2170  * 1) A non-virtual one is preferred over virtual.
2171  * 2) A virtual in master state preferred over any other state.
2172  *
2173  * Used in several address selecting functions.
2174  */
2175 int
2176 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2177 {
2178 
2179 	return (cur->ifa_carp && (!next->ifa_carp ||
2180 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2181 }
2182 
2183 struct sockaddr_dl *
2184 link_alloc_sdl(size_t size, int flags)
2185 {
2186 
2187 	return (malloc(size, M_TEMP, flags));
2188 }
2189 
2190 void
2191 link_free_sdl(struct sockaddr *sa)
2192 {
2193 	free(sa, M_TEMP);
2194 }
2195 
2196 /*
2197  * Fills in given sdl with interface basic info.
2198  * Returns pointer to filled sdl.
2199  */
2200 struct sockaddr_dl *
2201 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2202 {
2203 	struct sockaddr_dl *sdl;
2204 
2205 	sdl = (struct sockaddr_dl *)paddr;
2206 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2207 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2208 	sdl->sdl_family = AF_LINK;
2209 	sdl->sdl_index = ifp->if_index;
2210 	sdl->sdl_type = iftype;
2211 
2212 	return (sdl);
2213 }
2214 
2215 /*
2216  * Mark an interface down and notify protocols of
2217  * the transition.
2218  */
2219 static void
2220 if_unroute(struct ifnet *ifp, int flag, int fam)
2221 {
2222 	struct ifaddr *ifa;
2223 
2224 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2225 
2226 	ifp->if_flags &= ~flag;
2227 	getmicrotime(&ifp->if_lastchange);
2228 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2229 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2230 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2231 	ifp->if_qflush(ifp);
2232 
2233 	if (ifp->if_carp)
2234 		(*carp_linkstate_p)(ifp);
2235 	rt_ifmsg(ifp);
2236 }
2237 
2238 /*
2239  * Mark an interface up and notify protocols of
2240  * the transition.
2241  */
2242 static void
2243 if_route(struct ifnet *ifp, int flag, int fam)
2244 {
2245 	struct ifaddr *ifa;
2246 
2247 	KASSERT(flag == IFF_UP, ("if_route: 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_IFUP, ifa->ifa_addr);
2254 	if (ifp->if_carp)
2255 		(*carp_linkstate_p)(ifp);
2256 	rt_ifmsg(ifp);
2257 #ifdef INET6
2258 	in6_if_up(ifp);
2259 #endif
2260 }
2261 
2262 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2263 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2264 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2265 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2266 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2267 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2268 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2269 void	*(*vlan_cookie_p)(struct ifnet *);
2270 
2271 /*
2272  * Handle a change in the interface link state. To avoid LORs
2273  * between driver lock and upper layer locks, as well as possible
2274  * recursions, we post event to taskqueue, and all job
2275  * is done in static do_link_state_change().
2276  */
2277 void
2278 if_link_state_change(struct ifnet *ifp, int link_state)
2279 {
2280 	/* Return if state hasn't changed. */
2281 	if (ifp->if_link_state == link_state)
2282 		return;
2283 
2284 	ifp->if_link_state = link_state;
2285 
2286 	/* XXXGL: reference ifp? */
2287 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2288 }
2289 
2290 static void
2291 do_link_state_change(void *arg, int pending)
2292 {
2293 	struct ifnet *ifp;
2294 	int link_state;
2295 
2296 	ifp = arg;
2297 	link_state = ifp->if_link_state;
2298 
2299 	CURVNET_SET(ifp->if_vnet);
2300 	rt_ifmsg(ifp);
2301 	if (ifp->if_vlantrunk != NULL)
2302 		(*vlan_link_state_p)(ifp);
2303 
2304 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2305 	    ifp->if_l2com != NULL)
2306 		(*ng_ether_link_state_p)(ifp, link_state);
2307 	if (ifp->if_carp)
2308 		(*carp_linkstate_p)(ifp);
2309 	if (ifp->if_bridge)
2310 		ifp->if_bridge_linkstate(ifp);
2311 	if (ifp->if_lagg)
2312 		(*lagg_linkstate_p)(ifp, link_state);
2313 
2314 	if (IS_DEFAULT_VNET(curvnet))
2315 		devctl_notify("IFNET", ifp->if_xname,
2316 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2317 		    NULL);
2318 	if (pending > 1)
2319 		if_printf(ifp, "%d link states coalesced\n", pending);
2320 	if (log_link_state_change)
2321 		if_printf(ifp, "link state changed to %s\n",
2322 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2323 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2324 	CURVNET_RESTORE();
2325 }
2326 
2327 /*
2328  * Mark an interface down and notify protocols of
2329  * the transition.
2330  */
2331 void
2332 if_down(struct ifnet *ifp)
2333 {
2334 
2335 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2336 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2337 }
2338 
2339 /*
2340  * Mark an interface up and notify protocols of
2341  * the transition.
2342  */
2343 void
2344 if_up(struct ifnet *ifp)
2345 {
2346 
2347 	if_route(ifp, IFF_UP, AF_UNSPEC);
2348 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2349 }
2350 
2351 /*
2352  * Flush an interface queue.
2353  */
2354 void
2355 if_qflush(struct ifnet *ifp)
2356 {
2357 	struct mbuf *m, *n;
2358 	struct ifaltq *ifq;
2359 
2360 	ifq = &ifp->if_snd;
2361 	IFQ_LOCK(ifq);
2362 #ifdef ALTQ
2363 	if (ALTQ_IS_ENABLED(ifq))
2364 		ALTQ_PURGE(ifq);
2365 #endif
2366 	n = ifq->ifq_head;
2367 	while ((m = n) != NULL) {
2368 		n = m->m_nextpkt;
2369 		m_freem(m);
2370 	}
2371 	ifq->ifq_head = 0;
2372 	ifq->ifq_tail = 0;
2373 	ifq->ifq_len = 0;
2374 	IFQ_UNLOCK(ifq);
2375 }
2376 
2377 /*
2378  * Map interface name to interface structure pointer, with or without
2379  * returning a reference.
2380  */
2381 struct ifnet *
2382 ifunit_ref(const char *name)
2383 {
2384 	struct epoch_tracker et;
2385 	struct ifnet *ifp;
2386 
2387 	NET_EPOCH_ENTER(et);
2388 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2389 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2390 		    !(ifp->if_flags & IFF_DYING))
2391 			break;
2392 	}
2393 	if (ifp != NULL)
2394 		if_ref(ifp);
2395 	NET_EPOCH_EXIT(et);
2396 	return (ifp);
2397 }
2398 
2399 struct ifnet *
2400 ifunit(const char *name)
2401 {
2402 	struct epoch_tracker et;
2403 	struct ifnet *ifp;
2404 
2405 	NET_EPOCH_ENTER(et);
2406 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2407 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2408 			break;
2409 	}
2410 	NET_EPOCH_EXIT(et);
2411 	return (ifp);
2412 }
2413 
2414 void *
2415 ifr_buffer_get_buffer(void *data)
2416 {
2417 	union ifreq_union *ifrup;
2418 
2419 	ifrup = data;
2420 #ifdef COMPAT_FREEBSD32
2421 	if (SV_CURPROC_FLAG(SV_ILP32))
2422 		return ((void *)(uintptr_t)
2423 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2424 #endif
2425 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2426 }
2427 
2428 static void
2429 ifr_buffer_set_buffer_null(void *data)
2430 {
2431 	union ifreq_union *ifrup;
2432 
2433 	ifrup = data;
2434 #ifdef COMPAT_FREEBSD32
2435 	if (SV_CURPROC_FLAG(SV_ILP32))
2436 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2437 	else
2438 #endif
2439 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2440 }
2441 
2442 size_t
2443 ifr_buffer_get_length(void *data)
2444 {
2445 	union ifreq_union *ifrup;
2446 
2447 	ifrup = data;
2448 #ifdef COMPAT_FREEBSD32
2449 	if (SV_CURPROC_FLAG(SV_ILP32))
2450 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2451 #endif
2452 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2453 }
2454 
2455 static void
2456 ifr_buffer_set_length(void *data, size_t len)
2457 {
2458 	union ifreq_union *ifrup;
2459 
2460 	ifrup = data;
2461 #ifdef COMPAT_FREEBSD32
2462 	if (SV_CURPROC_FLAG(SV_ILP32))
2463 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2464 	else
2465 #endif
2466 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2467 }
2468 
2469 void *
2470 ifr_data_get_ptr(void *ifrp)
2471 {
2472 	union ifreq_union *ifrup;
2473 
2474 	ifrup = ifrp;
2475 #ifdef COMPAT_FREEBSD32
2476 	if (SV_CURPROC_FLAG(SV_ILP32))
2477 		return ((void *)(uintptr_t)
2478 		    ifrup->ifr32.ifr_ifru.ifru_data);
2479 #endif
2480 		return (ifrup->ifr.ifr_ifru.ifru_data);
2481 }
2482 
2483 /*
2484  * Hardware specific interface ioctls.
2485  */
2486 int
2487 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2488 {
2489 	struct ifreq *ifr;
2490 	int error = 0, do_ifup = 0;
2491 	int new_flags, temp_flags;
2492 	size_t namelen, onamelen;
2493 	size_t descrlen;
2494 	char *descrbuf, *odescrbuf;
2495 	char new_name[IFNAMSIZ];
2496 	struct ifaddr *ifa;
2497 	struct sockaddr_dl *sdl;
2498 
2499 	ifr = (struct ifreq *)data;
2500 	switch (cmd) {
2501 	case SIOCGIFINDEX:
2502 		ifr->ifr_index = ifp->if_index;
2503 		break;
2504 
2505 	case SIOCGIFFLAGS:
2506 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2507 		ifr->ifr_flags = temp_flags & 0xffff;
2508 		ifr->ifr_flagshigh = temp_flags >> 16;
2509 		break;
2510 
2511 	case SIOCGIFCAP:
2512 		ifr->ifr_reqcap = ifp->if_capabilities;
2513 		ifr->ifr_curcap = ifp->if_capenable;
2514 		break;
2515 
2516 #ifdef MAC
2517 	case SIOCGIFMAC:
2518 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2519 		break;
2520 #endif
2521 
2522 	case SIOCGIFMETRIC:
2523 		ifr->ifr_metric = ifp->if_metric;
2524 		break;
2525 
2526 	case SIOCGIFMTU:
2527 		ifr->ifr_mtu = ifp->if_mtu;
2528 		break;
2529 
2530 	case SIOCGIFPHYS:
2531 		/* XXXGL: did this ever worked? */
2532 		ifr->ifr_phys = 0;
2533 		break;
2534 
2535 	case SIOCGIFDESCR:
2536 		error = 0;
2537 		sx_slock(&ifdescr_sx);
2538 		if (ifp->if_description == NULL)
2539 			error = ENOMSG;
2540 		else {
2541 			/* space for terminating nul */
2542 			descrlen = strlen(ifp->if_description) + 1;
2543 			if (ifr_buffer_get_length(ifr) < descrlen)
2544 				ifr_buffer_set_buffer_null(ifr);
2545 			else
2546 				error = copyout(ifp->if_description,
2547 				    ifr_buffer_get_buffer(ifr), descrlen);
2548 			ifr_buffer_set_length(ifr, descrlen);
2549 		}
2550 		sx_sunlock(&ifdescr_sx);
2551 		break;
2552 
2553 	case SIOCSIFDESCR:
2554 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2555 		if (error)
2556 			return (error);
2557 
2558 		/*
2559 		 * Copy only (length-1) bytes to make sure that
2560 		 * if_description is always nul terminated.  The
2561 		 * length parameter is supposed to count the
2562 		 * terminating nul in.
2563 		 */
2564 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2565 			return (ENAMETOOLONG);
2566 		else if (ifr_buffer_get_length(ifr) == 0)
2567 			descrbuf = NULL;
2568 		else {
2569 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2570 			    M_IFDESCR, M_WAITOK | M_ZERO);
2571 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2572 			    ifr_buffer_get_length(ifr) - 1);
2573 			if (error) {
2574 				free(descrbuf, M_IFDESCR);
2575 				break;
2576 			}
2577 		}
2578 
2579 		sx_xlock(&ifdescr_sx);
2580 		odescrbuf = ifp->if_description;
2581 		ifp->if_description = descrbuf;
2582 		sx_xunlock(&ifdescr_sx);
2583 
2584 		getmicrotime(&ifp->if_lastchange);
2585 		free(odescrbuf, M_IFDESCR);
2586 		break;
2587 
2588 	case SIOCGIFFIB:
2589 		ifr->ifr_fib = ifp->if_fib;
2590 		break;
2591 
2592 	case SIOCSIFFIB:
2593 		error = priv_check(td, PRIV_NET_SETIFFIB);
2594 		if (error)
2595 			return (error);
2596 		if (ifr->ifr_fib >= rt_numfibs)
2597 			return (EINVAL);
2598 
2599 		ifp->if_fib = ifr->ifr_fib;
2600 		break;
2601 
2602 	case SIOCSIFFLAGS:
2603 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2604 		if (error)
2605 			return (error);
2606 		/*
2607 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2608 		 * check, so we don't need special handling here yet.
2609 		 */
2610 		new_flags = (ifr->ifr_flags & 0xffff) |
2611 		    (ifr->ifr_flagshigh << 16);
2612 		if (ifp->if_flags & IFF_UP &&
2613 		    (new_flags & IFF_UP) == 0) {
2614 			if_down(ifp);
2615 		} else if (new_flags & IFF_UP &&
2616 		    (ifp->if_flags & IFF_UP) == 0) {
2617 			do_ifup = 1;
2618 		}
2619 		/* See if permanently promiscuous mode bit is about to flip */
2620 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2621 			if (new_flags & IFF_PPROMISC)
2622 				ifp->if_flags |= IFF_PROMISC;
2623 			else if (ifp->if_pcount == 0)
2624 				ifp->if_flags &= ~IFF_PROMISC;
2625 			if (log_promisc_mode_change)
2626                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2627                                     ((new_flags & IFF_PPROMISC) ?
2628                                      "enabled" : "disabled"));
2629 		}
2630 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2631 			(new_flags &~ IFF_CANTCHANGE);
2632 		if (ifp->if_ioctl) {
2633 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2634 		}
2635 		if (do_ifup)
2636 			if_up(ifp);
2637 		getmicrotime(&ifp->if_lastchange);
2638 		break;
2639 
2640 	case SIOCSIFCAP:
2641 		error = priv_check(td, PRIV_NET_SETIFCAP);
2642 		if (error)
2643 			return (error);
2644 		if (ifp->if_ioctl == NULL)
2645 			return (EOPNOTSUPP);
2646 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2647 			return (EINVAL);
2648 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2649 		if (error == 0)
2650 			getmicrotime(&ifp->if_lastchange);
2651 		break;
2652 
2653 #ifdef MAC
2654 	case SIOCSIFMAC:
2655 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2656 		break;
2657 #endif
2658 
2659 	case SIOCSIFNAME:
2660 		error = priv_check(td, PRIV_NET_SETIFNAME);
2661 		if (error)
2662 			return (error);
2663 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2664 		    NULL);
2665 		if (error != 0)
2666 			return (error);
2667 		if (new_name[0] == '\0')
2668 			return (EINVAL);
2669 		if (new_name[IFNAMSIZ-1] != '\0') {
2670 			new_name[IFNAMSIZ-1] = '\0';
2671 			if (strlen(new_name) == IFNAMSIZ-1)
2672 				return (EINVAL);
2673 		}
2674 		if (strcmp(new_name, ifp->if_xname) == 0)
2675 			break;
2676 		if (ifunit(new_name) != NULL)
2677 			return (EEXIST);
2678 
2679 		/*
2680 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2681 		 * and there are numerous other races with the
2682 		 * ifunit() checks not being atomic with namespace
2683 		 * changes (renames, vmoves, if_attach, etc).
2684 		 */
2685 		ifp->if_flags |= IFF_RENAMING;
2686 
2687 		/* Announce the departure of the interface. */
2688 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2689 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2690 
2691 		if_printf(ifp, "changing name to '%s'\n", new_name);
2692 
2693 		IF_ADDR_WLOCK(ifp);
2694 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2695 		ifa = ifp->if_addr;
2696 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2697 		namelen = strlen(new_name);
2698 		onamelen = sdl->sdl_nlen;
2699 		/*
2700 		 * Move the address if needed.  This is safe because we
2701 		 * allocate space for a name of length IFNAMSIZ when we
2702 		 * create this in if_attach().
2703 		 */
2704 		if (namelen != onamelen) {
2705 			bcopy(sdl->sdl_data + onamelen,
2706 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2707 		}
2708 		bcopy(new_name, sdl->sdl_data, namelen);
2709 		sdl->sdl_nlen = namelen;
2710 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2711 		bzero(sdl->sdl_data, onamelen);
2712 		while (namelen != 0)
2713 			sdl->sdl_data[--namelen] = 0xff;
2714 		IF_ADDR_WUNLOCK(ifp);
2715 
2716 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2717 		/* Announce the return of the interface. */
2718 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2719 
2720 		ifp->if_flags &= ~IFF_RENAMING;
2721 		break;
2722 
2723 #ifdef VIMAGE
2724 	case SIOCSIFVNET:
2725 		error = priv_check(td, PRIV_NET_SETIFVNET);
2726 		if (error)
2727 			return (error);
2728 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2729 		break;
2730 #endif
2731 
2732 	case SIOCSIFMETRIC:
2733 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2734 		if (error)
2735 			return (error);
2736 		ifp->if_metric = ifr->ifr_metric;
2737 		getmicrotime(&ifp->if_lastchange);
2738 		break;
2739 
2740 	case SIOCSIFPHYS:
2741 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2742 		if (error)
2743 			return (error);
2744 		if (ifp->if_ioctl == NULL)
2745 			return (EOPNOTSUPP);
2746 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2747 		if (error == 0)
2748 			getmicrotime(&ifp->if_lastchange);
2749 		break;
2750 
2751 	case SIOCSIFMTU:
2752 	{
2753 		u_long oldmtu = ifp->if_mtu;
2754 
2755 		error = priv_check(td, PRIV_NET_SETIFMTU);
2756 		if (error)
2757 			return (error);
2758 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2759 			return (EINVAL);
2760 		if (ifp->if_ioctl == NULL)
2761 			return (EOPNOTSUPP);
2762 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2763 		if (error == 0) {
2764 			getmicrotime(&ifp->if_lastchange);
2765 			rt_ifmsg(ifp);
2766 #ifdef INET
2767 			DEBUGNET_NOTIFY_MTU(ifp);
2768 #endif
2769 		}
2770 		/*
2771 		 * If the link MTU changed, do network layer specific procedure.
2772 		 */
2773 		if (ifp->if_mtu != oldmtu) {
2774 #ifdef INET6
2775 			nd6_setmtu(ifp);
2776 #endif
2777 			rt_updatemtu(ifp);
2778 		}
2779 		break;
2780 	}
2781 
2782 	case SIOCADDMULTI:
2783 	case SIOCDELMULTI:
2784 		if (cmd == SIOCADDMULTI)
2785 			error = priv_check(td, PRIV_NET_ADDMULTI);
2786 		else
2787 			error = priv_check(td, PRIV_NET_DELMULTI);
2788 		if (error)
2789 			return (error);
2790 
2791 		/* Don't allow group membership on non-multicast interfaces. */
2792 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2793 			return (EOPNOTSUPP);
2794 
2795 		/* Don't let users screw up protocols' entries. */
2796 		if (ifr->ifr_addr.sa_family != AF_LINK)
2797 			return (EINVAL);
2798 
2799 		if (cmd == SIOCADDMULTI) {
2800 			struct epoch_tracker et;
2801 			struct ifmultiaddr *ifma;
2802 
2803 			/*
2804 			 * Userland is only permitted to join groups once
2805 			 * via the if_addmulti() KPI, because it cannot hold
2806 			 * struct ifmultiaddr * between calls. It may also
2807 			 * lose a race while we check if the membership
2808 			 * already exists.
2809 			 */
2810 			NET_EPOCH_ENTER(et);
2811 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2812 			NET_EPOCH_EXIT(et);
2813 			if (ifma != NULL)
2814 				error = EADDRINUSE;
2815 			else
2816 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2817 		} else {
2818 			error = if_delmulti(ifp, &ifr->ifr_addr);
2819 		}
2820 		if (error == 0)
2821 			getmicrotime(&ifp->if_lastchange);
2822 		break;
2823 
2824 	case SIOCSIFPHYADDR:
2825 	case SIOCDIFPHYADDR:
2826 #ifdef INET6
2827 	case SIOCSIFPHYADDR_IN6:
2828 #endif
2829 	case SIOCSIFMEDIA:
2830 	case SIOCSIFGENERIC:
2831 		error = priv_check(td, PRIV_NET_HWIOCTL);
2832 		if (error)
2833 			return (error);
2834 		if (ifp->if_ioctl == NULL)
2835 			return (EOPNOTSUPP);
2836 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2837 		if (error == 0)
2838 			getmicrotime(&ifp->if_lastchange);
2839 		break;
2840 
2841 	case SIOCGIFSTATUS:
2842 	case SIOCGIFPSRCADDR:
2843 	case SIOCGIFPDSTADDR:
2844 	case SIOCGIFMEDIA:
2845 	case SIOCGIFXMEDIA:
2846 	case SIOCGIFGENERIC:
2847 	case SIOCGIFRSSKEY:
2848 	case SIOCGIFRSSHASH:
2849 	case SIOCGIFDOWNREASON:
2850 		if (ifp->if_ioctl == NULL)
2851 			return (EOPNOTSUPP);
2852 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2853 		break;
2854 
2855 	case SIOCSIFLLADDR:
2856 		error = priv_check(td, PRIV_NET_SETLLADDR);
2857 		if (error)
2858 			return (error);
2859 		error = if_setlladdr(ifp,
2860 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2861 		break;
2862 
2863 	case SIOCGHWADDR:
2864 		error = if_gethwaddr(ifp, ifr);
2865 		break;
2866 
2867 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2868 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2869 		if (error)
2870 			return (error);
2871 		if ((error = if_addgroup(ifp,
2872 		    ifgr_group_get((struct ifgroupreq *)data))))
2873 			return (error);
2874 		break;
2875 
2876 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2877 	{
2878 		struct epoch_tracker et;
2879 
2880 		NET_EPOCH_ENTER(et);
2881 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2882 		NET_EPOCH_EXIT(et);
2883 		break;
2884 	}
2885 
2886 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2887 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2888 		if (error)
2889 			return (error);
2890 		if ((error = if_delgroup(ifp,
2891 		    ifgr_group_get((struct ifgroupreq *)data))))
2892 			return (error);
2893 		break;
2894 
2895 	default:
2896 		error = ENOIOCTL;
2897 		break;
2898 	}
2899 	return (error);
2900 }
2901 
2902 #ifdef COMPAT_FREEBSD32
2903 struct ifconf32 {
2904 	int32_t	ifc_len;
2905 	union {
2906 		uint32_t	ifcu_buf;
2907 		uint32_t	ifcu_req;
2908 	} ifc_ifcu;
2909 };
2910 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2911 #endif
2912 
2913 #ifdef COMPAT_FREEBSD32
2914 static void
2915 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2916 {
2917 	struct ifmediareq32 *ifmr32;
2918 
2919 	ifmr32 = (struct ifmediareq32 *)data;
2920 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2921 	    sizeof(ifmr->ifm_name));
2922 	ifmr->ifm_current = ifmr32->ifm_current;
2923 	ifmr->ifm_mask = ifmr32->ifm_mask;
2924 	ifmr->ifm_status = ifmr32->ifm_status;
2925 	ifmr->ifm_active = ifmr32->ifm_active;
2926 	ifmr->ifm_count = ifmr32->ifm_count;
2927 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2928 }
2929 
2930 static void
2931 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2932 {
2933 	struct ifmediareq32 *ifmr32;
2934 
2935 	ifmr32 = (struct ifmediareq32 *)data;
2936 	ifmr32->ifm_current = ifmr->ifm_current;
2937 	ifmr32->ifm_mask = ifmr->ifm_mask;
2938 	ifmr32->ifm_status = ifmr->ifm_status;
2939 	ifmr32->ifm_active = ifmr->ifm_active;
2940 	ifmr32->ifm_count = ifmr->ifm_count;
2941 }
2942 #endif
2943 
2944 /*
2945  * Interface ioctls.
2946  */
2947 int
2948 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2949 {
2950 #ifdef COMPAT_FREEBSD32
2951 	caddr_t saved_data = NULL;
2952 	struct ifmediareq ifmr;
2953 	struct ifmediareq *ifmrp = NULL;
2954 #endif
2955 	struct ifnet *ifp;
2956 	struct ifreq *ifr;
2957 	int error;
2958 	int oif_flags;
2959 #ifdef VIMAGE
2960 	bool shutdown;
2961 #endif
2962 
2963 	CURVNET_SET(so->so_vnet);
2964 #ifdef VIMAGE
2965 	/* Make sure the VNET is stable. */
2966 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2967 	if (shutdown) {
2968 		CURVNET_RESTORE();
2969 		return (EBUSY);
2970 	}
2971 #endif
2972 
2973 	switch (cmd) {
2974 	case SIOCGIFCONF:
2975 		error = ifconf(cmd, data);
2976 		goto out_noref;
2977 
2978 #ifdef COMPAT_FREEBSD32
2979 	case SIOCGIFCONF32:
2980 		{
2981 			struct ifconf32 *ifc32;
2982 			struct ifconf ifc;
2983 
2984 			ifc32 = (struct ifconf32 *)data;
2985 			ifc.ifc_len = ifc32->ifc_len;
2986 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2987 
2988 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2989 			if (error == 0)
2990 				ifc32->ifc_len = ifc.ifc_len;
2991 			goto out_noref;
2992 		}
2993 #endif
2994 	}
2995 
2996 #ifdef COMPAT_FREEBSD32
2997 	switch (cmd) {
2998 	case SIOCGIFMEDIA32:
2999 	case SIOCGIFXMEDIA32:
3000 		ifmrp = &ifmr;
3001 		ifmr_init(ifmrp, data);
3002 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3003 		saved_data = data;
3004 		data = (caddr_t)ifmrp;
3005 	}
3006 #endif
3007 
3008 	ifr = (struct ifreq *)data;
3009 	switch (cmd) {
3010 #ifdef VIMAGE
3011 	case SIOCSIFRVNET:
3012 		error = priv_check(td, PRIV_NET_SETIFVNET);
3013 		if (error == 0)
3014 			error = if_vmove_reclaim(td, ifr->ifr_name,
3015 			    ifr->ifr_jid);
3016 		goto out_noref;
3017 #endif
3018 	case SIOCIFCREATE:
3019 	case SIOCIFCREATE2:
3020 		error = priv_check(td, PRIV_NET_IFCREATE);
3021 		if (error == 0)
3022 			error = if_clone_create(ifr->ifr_name,
3023 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3024 			    ifr_data_get_ptr(ifr) : NULL);
3025 		goto out_noref;
3026 	case SIOCIFDESTROY:
3027 		error = priv_check(td, PRIV_NET_IFDESTROY);
3028 		if (error == 0)
3029 			error = if_clone_destroy(ifr->ifr_name);
3030 		goto out_noref;
3031 
3032 	case SIOCIFGCLONERS:
3033 		error = if_clone_list((struct if_clonereq *)data);
3034 		goto out_noref;
3035 
3036 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3037 		error = if_getgroupmembers((struct ifgroupreq *)data);
3038 		goto out_noref;
3039 
3040 #if defined(INET) || defined(INET6)
3041 	case SIOCSVH:
3042 	case SIOCGVH:
3043 		if (carp_ioctl_p == NULL)
3044 			error = EPROTONOSUPPORT;
3045 		else
3046 			error = (*carp_ioctl_p)(ifr, cmd, td);
3047 		goto out_noref;
3048 #endif
3049 	}
3050 
3051 	ifp = ifunit_ref(ifr->ifr_name);
3052 	if (ifp == NULL) {
3053 		error = ENXIO;
3054 		goto out_noref;
3055 	}
3056 
3057 	error = ifhwioctl(cmd, ifp, data, td);
3058 	if (error != ENOIOCTL)
3059 		goto out_ref;
3060 
3061 	oif_flags = ifp->if_flags;
3062 	if (so->so_proto == NULL) {
3063 		error = EOPNOTSUPP;
3064 		goto out_ref;
3065 	}
3066 
3067 	/*
3068 	 * Pass the request on to the socket control method, and if the
3069 	 * latter returns EOPNOTSUPP, directly to the interface.
3070 	 *
3071 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3072 	 * trust SIOCSIFADDR et al to come from an already privileged
3073 	 * layer, and do not perform any credentials checks or input
3074 	 * validation.
3075 	 */
3076 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3077 	    ifp, td));
3078 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3079 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3080 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3081 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3082 
3083 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3084 #ifdef INET6
3085 		if (ifp->if_flags & IFF_UP)
3086 			in6_if_up(ifp);
3087 #endif
3088 	}
3089 
3090 out_ref:
3091 	if_rele(ifp);
3092 out_noref:
3093 #ifdef COMPAT_FREEBSD32
3094 	if (ifmrp != NULL) {
3095 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3096 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3097 		     cmd));
3098 		data = saved_data;
3099 		ifmr_update(ifmrp, data);
3100 	}
3101 #endif
3102 	CURVNET_RESTORE();
3103 	return (error);
3104 }
3105 
3106 /*
3107  * The code common to handling reference counted flags,
3108  * e.g., in ifpromisc() and if_allmulti().
3109  * The "pflag" argument can specify a permanent mode flag to check,
3110  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3111  *
3112  * Only to be used on stack-owned flags, not driver-owned flags.
3113  */
3114 static int
3115 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3116 {
3117 	struct ifreq ifr;
3118 	int error;
3119 	int oldflags, oldcount;
3120 
3121 	/* Sanity checks to catch programming errors */
3122 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3123 	    ("%s: setting driver-owned flag %d", __func__, flag));
3124 
3125 	if (onswitch)
3126 		KASSERT(*refcount >= 0,
3127 		    ("%s: increment negative refcount %d for flag %d",
3128 		    __func__, *refcount, flag));
3129 	else
3130 		KASSERT(*refcount > 0,
3131 		    ("%s: decrement non-positive refcount %d for flag %d",
3132 		    __func__, *refcount, flag));
3133 
3134 	/* In case this mode is permanent, just touch refcount */
3135 	if (ifp->if_flags & pflag) {
3136 		*refcount += onswitch ? 1 : -1;
3137 		return (0);
3138 	}
3139 
3140 	/* Save ifnet parameters for if_ioctl() may fail */
3141 	oldcount = *refcount;
3142 	oldflags = ifp->if_flags;
3143 
3144 	/*
3145 	 * See if we aren't the only and touching refcount is enough.
3146 	 * Actually toggle interface flag if we are the first or last.
3147 	 */
3148 	if (onswitch) {
3149 		if ((*refcount)++)
3150 			return (0);
3151 		ifp->if_flags |= flag;
3152 	} else {
3153 		if (--(*refcount))
3154 			return (0);
3155 		ifp->if_flags &= ~flag;
3156 	}
3157 
3158 	/* Call down the driver since we've changed interface flags */
3159 	if (ifp->if_ioctl == NULL) {
3160 		error = EOPNOTSUPP;
3161 		goto recover;
3162 	}
3163 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3164 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3165 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3166 	if (error)
3167 		goto recover;
3168 	/* Notify userland that interface flags have changed */
3169 	rt_ifmsg(ifp);
3170 	return (0);
3171 
3172 recover:
3173 	/* Recover after driver error */
3174 	*refcount = oldcount;
3175 	ifp->if_flags = oldflags;
3176 	return (error);
3177 }
3178 
3179 /*
3180  * Set/clear promiscuous mode on interface ifp based on the truth value
3181  * of pswitch.  The calls are reference counted so that only the first
3182  * "on" request actually has an effect, as does the final "off" request.
3183  * Results are undefined if the "off" and "on" requests are not matched.
3184  */
3185 int
3186 ifpromisc(struct ifnet *ifp, int pswitch)
3187 {
3188 	int error;
3189 	int oldflags = ifp->if_flags;
3190 
3191 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3192 			   &ifp->if_pcount, pswitch);
3193 	/* If promiscuous mode status has changed, log a message */
3194 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3195             log_promisc_mode_change)
3196 		if_printf(ifp, "promiscuous mode %s\n",
3197 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3198 	return (error);
3199 }
3200 
3201 /*
3202  * Return interface configuration
3203  * of system.  List may be used
3204  * in later ioctl's (above) to get
3205  * other information.
3206  */
3207 /*ARGSUSED*/
3208 static int
3209 ifconf(u_long cmd, caddr_t data)
3210 {
3211 	struct ifconf *ifc = (struct ifconf *)data;
3212 	struct ifnet *ifp;
3213 	struct ifaddr *ifa;
3214 	struct ifreq ifr;
3215 	struct sbuf *sb;
3216 	int error, full = 0, valid_len, max_len;
3217 
3218 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3219 	max_len = MAXPHYS - 1;
3220 
3221 	/* Prevent hostile input from being able to crash the system */
3222 	if (ifc->ifc_len <= 0)
3223 		return (EINVAL);
3224 
3225 again:
3226 	if (ifc->ifc_len <= max_len) {
3227 		max_len = ifc->ifc_len;
3228 		full = 1;
3229 	}
3230 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3231 	max_len = 0;
3232 	valid_len = 0;
3233 
3234 	IFNET_RLOCK();
3235 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3236 		struct epoch_tracker et;
3237 		int addrs;
3238 
3239 		/*
3240 		 * Zero the ifr to make sure we don't disclose the contents
3241 		 * of the stack.
3242 		 */
3243 		memset(&ifr, 0, sizeof(ifr));
3244 
3245 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3246 		    >= sizeof(ifr.ifr_name)) {
3247 			sbuf_delete(sb);
3248 			IFNET_RUNLOCK();
3249 			return (ENAMETOOLONG);
3250 		}
3251 
3252 		addrs = 0;
3253 		NET_EPOCH_ENTER(et);
3254 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3255 			struct sockaddr *sa = ifa->ifa_addr;
3256 
3257 			if (prison_if(curthread->td_ucred, sa) != 0)
3258 				continue;
3259 			addrs++;
3260 			if (sa->sa_len <= sizeof(*sa)) {
3261 				if (sa->sa_len < sizeof(*sa)) {
3262 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3263 					    sizeof(ifr.ifr_ifru.ifru_addr));
3264 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3265 					    sa->sa_len);
3266 				} else
3267 					ifr.ifr_ifru.ifru_addr = *sa;
3268 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3269 				max_len += sizeof(ifr);
3270 			} else {
3271 				sbuf_bcat(sb, &ifr,
3272 				    offsetof(struct ifreq, ifr_addr));
3273 				max_len += offsetof(struct ifreq, ifr_addr);
3274 				sbuf_bcat(sb, sa, sa->sa_len);
3275 				max_len += sa->sa_len;
3276 			}
3277 
3278 			if (sbuf_error(sb) == 0)
3279 				valid_len = sbuf_len(sb);
3280 		}
3281 		NET_EPOCH_EXIT(et);
3282 		if (addrs == 0) {
3283 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3284 			max_len += sizeof(ifr);
3285 
3286 			if (sbuf_error(sb) == 0)
3287 				valid_len = sbuf_len(sb);
3288 		}
3289 	}
3290 	IFNET_RUNLOCK();
3291 
3292 	/*
3293 	 * If we didn't allocate enough space (uncommon), try again.  If
3294 	 * we have already allocated as much space as we are allowed,
3295 	 * return what we've got.
3296 	 */
3297 	if (valid_len != max_len && !full) {
3298 		sbuf_delete(sb);
3299 		goto again;
3300 	}
3301 
3302 	ifc->ifc_len = valid_len;
3303 	sbuf_finish(sb);
3304 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3305 	sbuf_delete(sb);
3306 	return (error);
3307 }
3308 
3309 /*
3310  * Just like ifpromisc(), but for all-multicast-reception mode.
3311  */
3312 int
3313 if_allmulti(struct ifnet *ifp, int onswitch)
3314 {
3315 
3316 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3317 }
3318 
3319 struct ifmultiaddr *
3320 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3321 {
3322 	struct ifmultiaddr *ifma;
3323 
3324 	IF_ADDR_LOCK_ASSERT(ifp);
3325 
3326 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3327 		if (sa->sa_family == AF_LINK) {
3328 			if (sa_dl_equal(ifma->ifma_addr, sa))
3329 				break;
3330 		} else {
3331 			if (sa_equal(ifma->ifma_addr, sa))
3332 				break;
3333 		}
3334 	}
3335 
3336 	return ifma;
3337 }
3338 
3339 /*
3340  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3341  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3342  * the ifnet multicast address list here, so the caller must do that and
3343  * other setup work (such as notifying the device driver).  The reference
3344  * count is initialized to 1.
3345  */
3346 static struct ifmultiaddr *
3347 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3348     int mflags)
3349 {
3350 	struct ifmultiaddr *ifma;
3351 	struct sockaddr *dupsa;
3352 
3353 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3354 	    M_ZERO);
3355 	if (ifma == NULL)
3356 		return (NULL);
3357 
3358 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3359 	if (dupsa == NULL) {
3360 		free(ifma, M_IFMADDR);
3361 		return (NULL);
3362 	}
3363 	bcopy(sa, dupsa, sa->sa_len);
3364 	ifma->ifma_addr = dupsa;
3365 
3366 	ifma->ifma_ifp = ifp;
3367 	ifma->ifma_refcount = 1;
3368 	ifma->ifma_protospec = NULL;
3369 
3370 	if (llsa == NULL) {
3371 		ifma->ifma_lladdr = NULL;
3372 		return (ifma);
3373 	}
3374 
3375 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3376 	if (dupsa == NULL) {
3377 		free(ifma->ifma_addr, M_IFMADDR);
3378 		free(ifma, M_IFMADDR);
3379 		return (NULL);
3380 	}
3381 	bcopy(llsa, dupsa, llsa->sa_len);
3382 	ifma->ifma_lladdr = dupsa;
3383 
3384 	return (ifma);
3385 }
3386 
3387 /*
3388  * if_freemulti: free ifmultiaddr structure and possibly attached related
3389  * addresses.  The caller is responsible for implementing reference
3390  * counting, notifying the driver, handling routing messages, and releasing
3391  * any dependent link layer state.
3392  */
3393 #ifdef MCAST_VERBOSE
3394 extern void kdb_backtrace(void);
3395 #endif
3396 static void
3397 if_freemulti_internal(struct ifmultiaddr *ifma)
3398 {
3399 
3400 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3401 	    ifma->ifma_refcount));
3402 
3403 	if (ifma->ifma_lladdr != NULL)
3404 		free(ifma->ifma_lladdr, M_IFMADDR);
3405 #ifdef MCAST_VERBOSE
3406 	kdb_backtrace();
3407 	printf("%s freeing ifma: %p\n", __func__, ifma);
3408 #endif
3409 	free(ifma->ifma_addr, M_IFMADDR);
3410 	free(ifma, M_IFMADDR);
3411 }
3412 
3413 static void
3414 if_destroymulti(epoch_context_t ctx)
3415 {
3416 	struct ifmultiaddr *ifma;
3417 
3418 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3419 	if_freemulti_internal(ifma);
3420 }
3421 
3422 void
3423 if_freemulti(struct ifmultiaddr *ifma)
3424 {
3425 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3426 	    ifma->ifma_refcount));
3427 
3428 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3429 }
3430 
3431 /*
3432  * Register an additional multicast address with a network interface.
3433  *
3434  * - If the address is already present, bump the reference count on the
3435  *   address and return.
3436  * - If the address is not link-layer, look up a link layer address.
3437  * - Allocate address structures for one or both addresses, and attach to the
3438  *   multicast address list on the interface.  If automatically adding a link
3439  *   layer address, the protocol address will own a reference to the link
3440  *   layer address, to be freed when it is freed.
3441  * - Notify the network device driver of an addition to the multicast address
3442  *   list.
3443  *
3444  * 'sa' points to caller-owned memory with the desired multicast address.
3445  *
3446  * 'retifma' will be used to return a pointer to the resulting multicast
3447  * address reference, if desired.
3448  */
3449 int
3450 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3451     struct ifmultiaddr **retifma)
3452 {
3453 	struct ifmultiaddr *ifma, *ll_ifma;
3454 	struct sockaddr *llsa;
3455 	struct sockaddr_dl sdl;
3456 	int error;
3457 
3458 #ifdef INET
3459 	IN_MULTI_LIST_UNLOCK_ASSERT();
3460 #endif
3461 #ifdef INET6
3462 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3463 #endif
3464 	/*
3465 	 * If the address is already present, return a new reference to it;
3466 	 * otherwise, allocate storage and set up a new address.
3467 	 */
3468 	IF_ADDR_WLOCK(ifp);
3469 	ifma = if_findmulti(ifp, sa);
3470 	if (ifma != NULL) {
3471 		ifma->ifma_refcount++;
3472 		if (retifma != NULL)
3473 			*retifma = ifma;
3474 		IF_ADDR_WUNLOCK(ifp);
3475 		return (0);
3476 	}
3477 
3478 	/*
3479 	 * The address isn't already present; resolve the protocol address
3480 	 * into a link layer address, and then look that up, bump its
3481 	 * refcount or allocate an ifma for that also.
3482 	 * Most link layer resolving functions returns address data which
3483 	 * fits inside default sockaddr_dl structure. However callback
3484 	 * can allocate another sockaddr structure, in that case we need to
3485 	 * free it later.
3486 	 */
3487 	llsa = NULL;
3488 	ll_ifma = NULL;
3489 	if (ifp->if_resolvemulti != NULL) {
3490 		/* Provide called function with buffer size information */
3491 		sdl.sdl_len = sizeof(sdl);
3492 		llsa = (struct sockaddr *)&sdl;
3493 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3494 		if (error)
3495 			goto unlock_out;
3496 	}
3497 
3498 	/*
3499 	 * Allocate the new address.  Don't hook it up yet, as we may also
3500 	 * need to allocate a link layer multicast address.
3501 	 */
3502 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3503 	if (ifma == NULL) {
3504 		error = ENOMEM;
3505 		goto free_llsa_out;
3506 	}
3507 
3508 	/*
3509 	 * If a link layer address is found, we'll need to see if it's
3510 	 * already present in the address list, or allocate is as well.
3511 	 * When this block finishes, the link layer address will be on the
3512 	 * list.
3513 	 */
3514 	if (llsa != NULL) {
3515 		ll_ifma = if_findmulti(ifp, llsa);
3516 		if (ll_ifma == NULL) {
3517 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3518 			if (ll_ifma == NULL) {
3519 				--ifma->ifma_refcount;
3520 				if_freemulti(ifma);
3521 				error = ENOMEM;
3522 				goto free_llsa_out;
3523 			}
3524 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3525 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3526 			    ifma_link);
3527 		} else
3528 			ll_ifma->ifma_refcount++;
3529 		ifma->ifma_llifma = ll_ifma;
3530 	}
3531 
3532 	/*
3533 	 * We now have a new multicast address, ifma, and possibly a new or
3534 	 * referenced link layer address.  Add the primary address to the
3535 	 * ifnet address list.
3536 	 */
3537 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3538 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3539 
3540 	if (retifma != NULL)
3541 		*retifma = ifma;
3542 
3543 	/*
3544 	 * Must generate the message while holding the lock so that 'ifma'
3545 	 * pointer is still valid.
3546 	 */
3547 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3548 	IF_ADDR_WUNLOCK(ifp);
3549 
3550 	/*
3551 	 * We are certain we have added something, so call down to the
3552 	 * interface to let them know about it.
3553 	 */
3554 	if (ifp->if_ioctl != NULL) {
3555 		if (THREAD_CAN_SLEEP())
3556 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3557 		else
3558 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3559 	}
3560 
3561 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3562 		link_free_sdl(llsa);
3563 
3564 	return (0);
3565 
3566 free_llsa_out:
3567 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3568 		link_free_sdl(llsa);
3569 
3570 unlock_out:
3571 	IF_ADDR_WUNLOCK(ifp);
3572 	return (error);
3573 }
3574 
3575 static void
3576 if_siocaddmulti(void *arg, int pending)
3577 {
3578 	struct ifnet *ifp;
3579 
3580 	ifp = arg;
3581 #ifdef DIAGNOSTIC
3582 	if (pending > 1)
3583 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3584 #endif
3585 	CURVNET_SET(ifp->if_vnet);
3586 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3587 	CURVNET_RESTORE();
3588 }
3589 
3590 /*
3591  * Delete a multicast group membership by network-layer group address.
3592  *
3593  * Returns ENOENT if the entry could not be found. If ifp no longer
3594  * exists, results are undefined. This entry point should only be used
3595  * from subsystems which do appropriate locking to hold ifp for the
3596  * duration of the call.
3597  * Network-layer protocol domains must use if_delmulti_ifma().
3598  */
3599 int
3600 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3601 {
3602 	struct ifmultiaddr *ifma;
3603 	int lastref;
3604 
3605 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3606 
3607 	IF_ADDR_WLOCK(ifp);
3608 	lastref = 0;
3609 	ifma = if_findmulti(ifp, sa);
3610 	if (ifma != NULL)
3611 		lastref = if_delmulti_locked(ifp, ifma, 0);
3612 	IF_ADDR_WUNLOCK(ifp);
3613 
3614 	if (ifma == NULL)
3615 		return (ENOENT);
3616 
3617 	if (lastref && ifp->if_ioctl != NULL) {
3618 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3619 	}
3620 
3621 	return (0);
3622 }
3623 
3624 /*
3625  * Delete all multicast group membership for an interface.
3626  * Should be used to quickly flush all multicast filters.
3627  */
3628 void
3629 if_delallmulti(struct ifnet *ifp)
3630 {
3631 	struct ifmultiaddr *ifma;
3632 	struct ifmultiaddr *next;
3633 
3634 	IF_ADDR_WLOCK(ifp);
3635 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3636 		if_delmulti_locked(ifp, ifma, 0);
3637 	IF_ADDR_WUNLOCK(ifp);
3638 }
3639 
3640 void
3641 if_delmulti_ifma(struct ifmultiaddr *ifma)
3642 {
3643 	if_delmulti_ifma_flags(ifma, 0);
3644 }
3645 
3646 /*
3647  * Delete a multicast group membership by group membership pointer.
3648  * Network-layer protocol domains must use this routine.
3649  *
3650  * It is safe to call this routine if the ifp disappeared.
3651  */
3652 void
3653 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3654 {
3655 	struct ifnet *ifp;
3656 	int lastref;
3657 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3658 #ifdef INET
3659 	IN_MULTI_LIST_UNLOCK_ASSERT();
3660 #endif
3661 	ifp = ifma->ifma_ifp;
3662 #ifdef DIAGNOSTIC
3663 	if (ifp == NULL) {
3664 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3665 	} else {
3666 		struct epoch_tracker et;
3667 		struct ifnet *oifp;
3668 
3669 		NET_EPOCH_ENTER(et);
3670 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3671 			if (ifp == oifp)
3672 				break;
3673 		NET_EPOCH_EXIT(et);
3674 		if (ifp != oifp)
3675 			ifp = NULL;
3676 	}
3677 #endif
3678 	/*
3679 	 * If and only if the ifnet instance exists: Acquire the address lock.
3680 	 */
3681 	if (ifp != NULL)
3682 		IF_ADDR_WLOCK(ifp);
3683 
3684 	lastref = if_delmulti_locked(ifp, ifma, flags);
3685 
3686 	if (ifp != NULL) {
3687 		/*
3688 		 * If and only if the ifnet instance exists:
3689 		 *  Release the address lock.
3690 		 *  If the group was left: update the hardware hash filter.
3691 		 */
3692 		IF_ADDR_WUNLOCK(ifp);
3693 		if (lastref && ifp->if_ioctl != NULL) {
3694 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3695 		}
3696 	}
3697 }
3698 
3699 /*
3700  * Perform deletion of network-layer and/or link-layer multicast address.
3701  *
3702  * Return 0 if the reference count was decremented.
3703  * Return 1 if the final reference was released, indicating that the
3704  * hardware hash filter should be reprogrammed.
3705  */
3706 static int
3707 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3708 {
3709 	struct ifmultiaddr *ll_ifma;
3710 
3711 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3712 		KASSERT(ifma->ifma_ifp == ifp,
3713 		    ("%s: inconsistent ifp %p", __func__, ifp));
3714 		IF_ADDR_WLOCK_ASSERT(ifp);
3715 	}
3716 
3717 	ifp = ifma->ifma_ifp;
3718 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3719 
3720 	/*
3721 	 * If the ifnet is detaching, null out references to ifnet,
3722 	 * so that upper protocol layers will notice, and not attempt
3723 	 * to obtain locks for an ifnet which no longer exists. The
3724 	 * routing socket announcement must happen before the ifnet
3725 	 * instance is detached from the system.
3726 	 */
3727 	if (detaching) {
3728 #ifdef DIAGNOSTIC
3729 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3730 #endif
3731 		/*
3732 		 * ifp may already be nulled out if we are being reentered
3733 		 * to delete the ll_ifma.
3734 		 */
3735 		if (ifp != NULL) {
3736 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3737 			ifma->ifma_ifp = NULL;
3738 		}
3739 	}
3740 
3741 	if (--ifma->ifma_refcount > 0)
3742 		return 0;
3743 
3744 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3745 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3746 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3747 	}
3748 	/*
3749 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3750 	 * have been associated with it. Release it first if so.
3751 	 */
3752 	ll_ifma = ifma->ifma_llifma;
3753 	if (ll_ifma != NULL) {
3754 		KASSERT(ifma->ifma_lladdr != NULL,
3755 		    ("%s: llifma w/o lladdr", __func__));
3756 		if (detaching)
3757 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3758 		if (--ll_ifma->ifma_refcount == 0) {
3759 			if (ifp != NULL) {
3760 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3761 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3762 						ifma_link);
3763 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3764 				}
3765 			}
3766 			if_freemulti(ll_ifma);
3767 		}
3768 	}
3769 #ifdef INVARIANTS
3770 	if (ifp) {
3771 		struct ifmultiaddr *ifmatmp;
3772 
3773 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3774 			MPASS(ifma != ifmatmp);
3775 	}
3776 #endif
3777 	if_freemulti(ifma);
3778 	/*
3779 	 * The last reference to this instance of struct ifmultiaddr
3780 	 * was released; the hardware should be notified of this change.
3781 	 */
3782 	return 1;
3783 }
3784 
3785 /*
3786  * Set the link layer address on an interface.
3787  *
3788  * At this time we only support certain types of interfaces,
3789  * and we don't allow the length of the address to change.
3790  *
3791  * Set noinline to be dtrace-friendly
3792  */
3793 __noinline int
3794 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3795 {
3796 	struct sockaddr_dl *sdl;
3797 	struct ifaddr *ifa;
3798 	struct ifreq ifr;
3799 
3800 	ifa = ifp->if_addr;
3801 	if (ifa == NULL)
3802 		return (EINVAL);
3803 
3804 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3805 	if (sdl == NULL)
3806 		return (EINVAL);
3807 
3808 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3809 		return (EINVAL);
3810 
3811 	switch (ifp->if_type) {
3812 	case IFT_ETHER:
3813 	case IFT_XETHER:
3814 	case IFT_L2VLAN:
3815 	case IFT_BRIDGE:
3816 	case IFT_IEEE8023ADLAG:
3817 		bcopy(lladdr, LLADDR(sdl), len);
3818 		break;
3819 	default:
3820 		return (ENODEV);
3821 	}
3822 
3823 	/*
3824 	 * If the interface is already up, we need
3825 	 * to re-init it in order to reprogram its
3826 	 * address filter.
3827 	 */
3828 	if ((ifp->if_flags & IFF_UP) != 0) {
3829 		if (ifp->if_ioctl) {
3830 			ifp->if_flags &= ~IFF_UP;
3831 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3832 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3833 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3834 			ifp->if_flags |= IFF_UP;
3835 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3836 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3837 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3838 		}
3839 	}
3840 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3841 
3842 	return (0);
3843 }
3844 
3845 /*
3846  * Compat function for handling basic encapsulation requests.
3847  * Not converted stacks (FDDI, IB, ..) supports traditional
3848  * output model: ARP (and other similar L2 protocols) are handled
3849  * inside output routine, arpresolve/nd6_resolve() returns MAC
3850  * address instead of full prepend.
3851  *
3852  * This function creates calculated header==MAC for IPv4/IPv6 and
3853  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3854  * address families.
3855  */
3856 static int
3857 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3858 {
3859 
3860 	if (req->rtype != IFENCAP_LL)
3861 		return (EOPNOTSUPP);
3862 
3863 	if (req->bufsize < req->lladdr_len)
3864 		return (ENOMEM);
3865 
3866 	switch (req->family) {
3867 	case AF_INET:
3868 	case AF_INET6:
3869 		break;
3870 	default:
3871 		return (EAFNOSUPPORT);
3872 	}
3873 
3874 	/* Copy lladdr to storage as is */
3875 	memmove(req->buf, req->lladdr, req->lladdr_len);
3876 	req->bufsize = req->lladdr_len;
3877 	req->lladdr_off = 0;
3878 
3879 	return (0);
3880 }
3881 
3882 /*
3883  * Tunnel interfaces can nest, also they may cause infinite recursion
3884  * calls when misconfigured. We'll prevent this by detecting loops.
3885  * High nesting level may cause stack exhaustion. We'll prevent this
3886  * by introducing upper limit.
3887  *
3888  * Return 0, if tunnel nesting count is equal or less than limit.
3889  */
3890 int
3891 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3892     int limit)
3893 {
3894 	struct m_tag *mtag;
3895 	int count;
3896 
3897 	count = 1;
3898 	mtag = NULL;
3899 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3900 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3901 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3902 			return (EIO);
3903 		}
3904 		count++;
3905 	}
3906 	if (count > limit) {
3907 		log(LOG_NOTICE,
3908 		    "%s: if_output recursively called too many times(%d)\n",
3909 		    if_name(ifp), count);
3910 		return (EIO);
3911 	}
3912 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3913 	if (mtag == NULL)
3914 		return (ENOMEM);
3915 	*(struct ifnet **)(mtag + 1) = ifp;
3916 	m_tag_prepend(m, mtag);
3917 	return (0);
3918 }
3919 
3920 /*
3921  * Get the link layer address that was read from the hardware at attach.
3922  *
3923  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3924  * their component interfaces as IFT_IEEE8023ADLAG.
3925  */
3926 int
3927 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3928 {
3929 
3930 	if (ifp->if_hw_addr == NULL)
3931 		return (ENODEV);
3932 
3933 	switch (ifp->if_type) {
3934 	case IFT_ETHER:
3935 	case IFT_IEEE8023ADLAG:
3936 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3937 		return (0);
3938 	default:
3939 		return (ENODEV);
3940 	}
3941 }
3942 
3943 /*
3944  * The name argument must be a pointer to storage which will last as
3945  * long as the interface does.  For physical devices, the result of
3946  * device_get_name(dev) is a good choice and for pseudo-devices a
3947  * static string works well.
3948  */
3949 void
3950 if_initname(struct ifnet *ifp, const char *name, int unit)
3951 {
3952 	ifp->if_dname = name;
3953 	ifp->if_dunit = unit;
3954 	if (unit != IF_DUNIT_NONE)
3955 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3956 	else
3957 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3958 }
3959 
3960 int
3961 if_printf(struct ifnet *ifp, const char *fmt, ...)
3962 {
3963 	char if_fmt[256];
3964 	va_list ap;
3965 
3966 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3967 	va_start(ap, fmt);
3968 	vlog(LOG_INFO, if_fmt, ap);
3969 	va_end(ap);
3970 	return (0);
3971 }
3972 
3973 void
3974 if_start(struct ifnet *ifp)
3975 {
3976 
3977 	(*(ifp)->if_start)(ifp);
3978 }
3979 
3980 /*
3981  * Backwards compatibility interface for drivers
3982  * that have not implemented it
3983  */
3984 static int
3985 if_transmit(struct ifnet *ifp, struct mbuf *m)
3986 {
3987 	int error;
3988 
3989 	IFQ_HANDOFF(ifp, m, error);
3990 	return (error);
3991 }
3992 
3993 static void
3994 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3995 {
3996 
3997 	m_freem(m);
3998 }
3999 
4000 int
4001 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4002 {
4003 	int active = 0;
4004 
4005 	IF_LOCK(ifq);
4006 	if (_IF_QFULL(ifq)) {
4007 		IF_UNLOCK(ifq);
4008 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4009 		m_freem(m);
4010 		return (0);
4011 	}
4012 	if (ifp != NULL) {
4013 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4014 		if (m->m_flags & (M_BCAST|M_MCAST))
4015 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4016 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4017 	}
4018 	_IF_ENQUEUE(ifq, m);
4019 	IF_UNLOCK(ifq);
4020 	if (ifp != NULL && !active)
4021 		(*(ifp)->if_start)(ifp);
4022 	return (1);
4023 }
4024 
4025 void
4026 if_register_com_alloc(u_char type,
4027     if_com_alloc_t *a, if_com_free_t *f)
4028 {
4029 
4030 	KASSERT(if_com_alloc[type] == NULL,
4031 	    ("if_register_com_alloc: %d already registered", type));
4032 	KASSERT(if_com_free[type] == NULL,
4033 	    ("if_register_com_alloc: %d free already registered", type));
4034 
4035 	if_com_alloc[type] = a;
4036 	if_com_free[type] = f;
4037 }
4038 
4039 void
4040 if_deregister_com_alloc(u_char type)
4041 {
4042 
4043 	KASSERT(if_com_alloc[type] != NULL,
4044 	    ("if_deregister_com_alloc: %d not registered", type));
4045 	KASSERT(if_com_free[type] != NULL,
4046 	    ("if_deregister_com_alloc: %d free not registered", type));
4047 	if_com_alloc[type] = NULL;
4048 	if_com_free[type] = NULL;
4049 }
4050 
4051 /* API for driver access to network stack owned ifnet.*/
4052 uint64_t
4053 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4054 {
4055 	uint64_t oldbrate;
4056 
4057 	oldbrate = ifp->if_baudrate;
4058 	ifp->if_baudrate = baudrate;
4059 	return (oldbrate);
4060 }
4061 
4062 uint64_t
4063 if_getbaudrate(if_t ifp)
4064 {
4065 
4066 	return (((struct ifnet *)ifp)->if_baudrate);
4067 }
4068 
4069 int
4070 if_setcapabilities(if_t ifp, int capabilities)
4071 {
4072 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4073 	return (0);
4074 }
4075 
4076 int
4077 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4078 {
4079 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4080 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4081 
4082 	return (0);
4083 }
4084 
4085 int
4086 if_getcapabilities(if_t ifp)
4087 {
4088 	return ((struct ifnet *)ifp)->if_capabilities;
4089 }
4090 
4091 int
4092 if_setcapenable(if_t ifp, int capabilities)
4093 {
4094 	((struct ifnet *)ifp)->if_capenable = capabilities;
4095 	return (0);
4096 }
4097 
4098 int
4099 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4100 {
4101 	if(setcap)
4102 		((struct ifnet *)ifp)->if_capenable |= setcap;
4103 	if(clearcap)
4104 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4105 
4106 	return (0);
4107 }
4108 
4109 const char *
4110 if_getdname(if_t ifp)
4111 {
4112 	return ((struct ifnet *)ifp)->if_dname;
4113 }
4114 
4115 int
4116 if_togglecapenable(if_t ifp, int togglecap)
4117 {
4118 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4119 	return (0);
4120 }
4121 
4122 int
4123 if_getcapenable(if_t ifp)
4124 {
4125 	return ((struct ifnet *)ifp)->if_capenable;
4126 }
4127 
4128 /*
4129  * This is largely undesirable because it ties ifnet to a device, but does
4130  * provide flexiblity for an embedded product vendor. Should be used with
4131  * the understanding that it violates the interface boundaries, and should be
4132  * a last resort only.
4133  */
4134 int
4135 if_setdev(if_t ifp, void *dev)
4136 {
4137 	return (0);
4138 }
4139 
4140 int
4141 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4142 {
4143 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4144 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4145 
4146 	return (0);
4147 }
4148 
4149 int
4150 if_getdrvflags(if_t ifp)
4151 {
4152 	return ((struct ifnet *)ifp)->if_drv_flags;
4153 }
4154 
4155 int
4156 if_setdrvflags(if_t ifp, int flags)
4157 {
4158 	((struct ifnet *)ifp)->if_drv_flags = flags;
4159 	return (0);
4160 }
4161 
4162 int
4163 if_setflags(if_t ifp, int flags)
4164 {
4165 
4166 	ifp->if_flags = flags;
4167 	return (0);
4168 }
4169 
4170 int
4171 if_setflagbits(if_t ifp, int set, int clear)
4172 {
4173 	((struct ifnet *)ifp)->if_flags |= set;
4174 	((struct ifnet *)ifp)->if_flags &= ~clear;
4175 
4176 	return (0);
4177 }
4178 
4179 int
4180 if_getflags(if_t ifp)
4181 {
4182 	return ((struct ifnet *)ifp)->if_flags;
4183 }
4184 
4185 int
4186 if_clearhwassist(if_t ifp)
4187 {
4188 	((struct ifnet *)ifp)->if_hwassist = 0;
4189 	return (0);
4190 }
4191 
4192 int
4193 if_sethwassistbits(if_t ifp, int toset, int toclear)
4194 {
4195 	((struct ifnet *)ifp)->if_hwassist |= toset;
4196 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4197 
4198 	return (0);
4199 }
4200 
4201 int
4202 if_sethwassist(if_t ifp, int hwassist_bit)
4203 {
4204 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4205 	return (0);
4206 }
4207 
4208 int
4209 if_gethwassist(if_t ifp)
4210 {
4211 	return ((struct ifnet *)ifp)->if_hwassist;
4212 }
4213 
4214 int
4215 if_setmtu(if_t ifp, int mtu)
4216 {
4217 	((struct ifnet *)ifp)->if_mtu = mtu;
4218 	return (0);
4219 }
4220 
4221 int
4222 if_getmtu(if_t ifp)
4223 {
4224 	return ((struct ifnet *)ifp)->if_mtu;
4225 }
4226 
4227 int
4228 if_getmtu_family(if_t ifp, int family)
4229 {
4230 	struct domain *dp;
4231 
4232 	for (dp = domains; dp; dp = dp->dom_next) {
4233 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4234 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4235 	}
4236 
4237 	return (((struct ifnet *)ifp)->if_mtu);
4238 }
4239 
4240 /*
4241  * Methods for drivers to access interface unicast and multicast
4242  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4243  * 'struct ifmultiaddr'.
4244  */
4245 u_int
4246 if_lladdr_count(if_t ifp)
4247 {
4248 	struct epoch_tracker et;
4249 	struct ifaddr *ifa;
4250 	u_int count;
4251 
4252 	count = 0;
4253 	NET_EPOCH_ENTER(et);
4254 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4255 		if (ifa->ifa_addr->sa_family == AF_LINK)
4256 			count++;
4257 	NET_EPOCH_EXIT(et);
4258 
4259 	return (count);
4260 }
4261 
4262 u_int
4263 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4264 {
4265 	struct epoch_tracker et;
4266 	struct ifaddr *ifa;
4267 	u_int count;
4268 
4269 	MPASS(cb);
4270 
4271 	count = 0;
4272 	NET_EPOCH_ENTER(et);
4273 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4274 		if (ifa->ifa_addr->sa_family != AF_LINK)
4275 			continue;
4276 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4277 		    count);
4278 	}
4279 	NET_EPOCH_EXIT(et);
4280 
4281 	return (count);
4282 }
4283 
4284 u_int
4285 if_llmaddr_count(if_t ifp)
4286 {
4287 	struct epoch_tracker et;
4288 	struct ifmultiaddr *ifma;
4289 	int count;
4290 
4291 	count = 0;
4292 	NET_EPOCH_ENTER(et);
4293 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4294 		if (ifma->ifma_addr->sa_family == AF_LINK)
4295 			count++;
4296 	NET_EPOCH_EXIT(et);
4297 
4298 	return (count);
4299 }
4300 
4301 u_int
4302 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4303 {
4304 	struct epoch_tracker et;
4305 	struct ifmultiaddr *ifma;
4306 	u_int count;
4307 
4308 	MPASS(cb);
4309 
4310 	count = 0;
4311 	NET_EPOCH_ENTER(et);
4312 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4313 		if (ifma->ifma_addr->sa_family != AF_LINK)
4314 			continue;
4315 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4316 		    count);
4317 	}
4318 	NET_EPOCH_EXIT(et);
4319 
4320 	return (count);
4321 }
4322 
4323 int
4324 if_setsoftc(if_t ifp, void *softc)
4325 {
4326 	((struct ifnet *)ifp)->if_softc = softc;
4327 	return (0);
4328 }
4329 
4330 void *
4331 if_getsoftc(if_t ifp)
4332 {
4333 	return ((struct ifnet *)ifp)->if_softc;
4334 }
4335 
4336 void
4337 if_setrcvif(struct mbuf *m, if_t ifp)
4338 {
4339 
4340 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4341 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4342 }
4343 
4344 void
4345 if_setvtag(struct mbuf *m, uint16_t tag)
4346 {
4347 	m->m_pkthdr.ether_vtag = tag;
4348 }
4349 
4350 uint16_t
4351 if_getvtag(struct mbuf *m)
4352 {
4353 
4354 	return (m->m_pkthdr.ether_vtag);
4355 }
4356 
4357 int
4358 if_sendq_empty(if_t ifp)
4359 {
4360 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4361 }
4362 
4363 struct ifaddr *
4364 if_getifaddr(if_t ifp)
4365 {
4366 	return ((struct ifnet *)ifp)->if_addr;
4367 }
4368 
4369 int
4370 if_getamcount(if_t ifp)
4371 {
4372 	return ((struct ifnet *)ifp)->if_amcount;
4373 }
4374 
4375 int
4376 if_setsendqready(if_t ifp)
4377 {
4378 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4379 	return (0);
4380 }
4381 
4382 int
4383 if_setsendqlen(if_t ifp, int tx_desc_count)
4384 {
4385 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4386 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4387 
4388 	return (0);
4389 }
4390 
4391 int
4392 if_vlantrunkinuse(if_t ifp)
4393 {
4394 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4395 }
4396 
4397 int
4398 if_input(if_t ifp, struct mbuf* sendmp)
4399 {
4400 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4401 	return (0);
4402 
4403 }
4404 
4405 struct mbuf *
4406 if_dequeue(if_t ifp)
4407 {
4408 	struct mbuf *m;
4409 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4410 
4411 	return (m);
4412 }
4413 
4414 int
4415 if_sendq_prepend(if_t ifp, struct mbuf *m)
4416 {
4417 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4418 	return (0);
4419 }
4420 
4421 int
4422 if_setifheaderlen(if_t ifp, int len)
4423 {
4424 	((struct ifnet *)ifp)->if_hdrlen = len;
4425 	return (0);
4426 }
4427 
4428 caddr_t
4429 if_getlladdr(if_t ifp)
4430 {
4431 	return (IF_LLADDR((struct ifnet *)ifp));
4432 }
4433 
4434 void *
4435 if_gethandle(u_char type)
4436 {
4437 	return (if_alloc(type));
4438 }
4439 
4440 void
4441 if_bpfmtap(if_t ifh, struct mbuf *m)
4442 {
4443 	struct ifnet *ifp = (struct ifnet *)ifh;
4444 
4445 	BPF_MTAP(ifp, m);
4446 }
4447 
4448 void
4449 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4450 {
4451 	struct ifnet *ifp = (struct ifnet *)ifh;
4452 
4453 	ETHER_BPF_MTAP(ifp, m);
4454 }
4455 
4456 void
4457 if_vlancap(if_t ifh)
4458 {
4459 	struct ifnet *ifp = (struct ifnet *)ifh;
4460 	VLAN_CAPABILITIES(ifp);
4461 }
4462 
4463 int
4464 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4465 {
4466 
4467 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4468         return (0);
4469 }
4470 
4471 int
4472 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4473 {
4474 
4475 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4476         return (0);
4477 }
4478 
4479 int
4480 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4481 {
4482 
4483 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4484         return (0);
4485 }
4486 
4487 u_int
4488 if_gethwtsomax(if_t ifp)
4489 {
4490 
4491 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4492 }
4493 
4494 u_int
4495 if_gethwtsomaxsegcount(if_t ifp)
4496 {
4497 
4498 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4499 }
4500 
4501 u_int
4502 if_gethwtsomaxsegsize(if_t ifp)
4503 {
4504 
4505 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4506 }
4507 
4508 void
4509 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4510 {
4511 	((struct ifnet *)ifp)->if_init = init_fn;
4512 }
4513 
4514 void
4515 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4516 {
4517 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4518 }
4519 
4520 void
4521 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4522 {
4523 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4524 }
4525 
4526 void
4527 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4528 {
4529 	((struct ifnet *)ifp)->if_transmit = start_fn;
4530 }
4531 
4532 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4533 {
4534 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4535 
4536 }
4537 
4538 void
4539 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4540 {
4541 
4542 	ifp->if_get_counter = fn;
4543 }
4544 
4545 /* Revisit these - These are inline functions originally. */
4546 int
4547 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4548 {
4549 	return drbr_inuse(ifh, br);
4550 }
4551 
4552 struct mbuf*
4553 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4554 {
4555 	return drbr_dequeue(ifh, br);
4556 }
4557 
4558 int
4559 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4560 {
4561 	return drbr_needs_enqueue(ifh, br);
4562 }
4563 
4564 int
4565 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4566 {
4567 	return drbr_enqueue(ifh, br, m);
4568 
4569 }
4570