xref: /freebsd/sys/net/if.c (revision e9b148a3185f41e3a09e91ea75cae7828d908845)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/conf.h>
40 #include <sys/eventhandler.h>
41 #include <sys/malloc.h>
42 #include <sys/domainset.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/epoch.h>
46 #include <sys/mbuf.h>
47 #include <sys/systm.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/protosw.h>
53 #include <sys/kernel.h>
54 #include <sys/lock.h>
55 #include <sys/refcount.h>
56 #include <sys/module.h>
57 #include <sys/rwlock.h>
58 #include <sys/sockio.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/taskqueue.h>
63 #include <sys/domain.h>
64 #include <sys/jail.h>
65 #include <sys/priv.h>
66 
67 #include <machine/stdarg.h>
68 #include <vm/uma.h>
69 
70 #include <net/bpf.h>
71 #include <net/ethernet.h>
72 #include <net/if.h>
73 #include <net/if_arp.h>
74 #include <net/if_clone.h>
75 #include <net/if_dl.h>
76 #include <net/if_types.h>
77 #include <net/if_var.h>
78 #include <net/if_media.h>
79 #include <net/if_vlan_var.h>
80 #include <net/radix.h>
81 #include <net/route.h>
82 #include <net/vnet.h>
83 
84 #if defined(INET) || defined(INET6)
85 #include <net/ethernet.h>
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_carp.h>
90 #ifdef INET
91 #include <net/debugnet.h>
92 #include <netinet/if_ether.h>
93 #endif /* INET */
94 #ifdef INET6
95 #include <netinet6/in6_var.h>
96 #include <netinet6/in6_ifattach.h>
97 #endif /* INET6 */
98 #endif /* INET || INET6 */
99 
100 #include <security/mac/mac_framework.h>
101 
102 /*
103  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
104  * and ifr_ifru when it is used in SIOCGIFCONF.
105  */
106 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
107     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
108 
109 __read_mostly epoch_t net_epoch_preempt;
110 __read_mostly epoch_t net_epoch;
111 #ifdef COMPAT_FREEBSD32
112 #include <sys/mount.h>
113 #include <compat/freebsd32/freebsd32.h>
114 
115 struct ifreq_buffer32 {
116 	uint32_t	length;		/* (size_t) */
117 	uint32_t	buffer;		/* (void *) */
118 };
119 
120 /*
121  * Interface request structure used for socket
122  * ioctl's.  All interface ioctl's must have parameter
123  * definitions which begin with ifr_name.  The
124  * remainder may be interface specific.
125  */
126 struct ifreq32 {
127 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
128 	union {
129 		struct sockaddr	ifru_addr;
130 		struct sockaddr	ifru_dstaddr;
131 		struct sockaddr	ifru_broadaddr;
132 		struct ifreq_buffer32 ifru_buffer;
133 		short		ifru_flags[2];
134 		short		ifru_index;
135 		int		ifru_jid;
136 		int		ifru_metric;
137 		int		ifru_mtu;
138 		int		ifru_phys;
139 		int		ifru_media;
140 		uint32_t	ifru_data;
141 		int		ifru_cap[2];
142 		u_int		ifru_fib;
143 		u_char		ifru_vlan_pcp;
144 	} ifr_ifru;
145 };
146 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
147 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
148     __offsetof(struct ifreq32, ifr_ifru));
149 
150 struct ifgroupreq32 {
151 	char	ifgr_name[IFNAMSIZ];
152 	u_int	ifgr_len;
153 	union {
154 		char		ifgru_group[IFNAMSIZ];
155 		uint32_t	ifgru_groups;
156 	} ifgr_ifgru;
157 };
158 
159 struct ifmediareq32 {
160 	char		ifm_name[IFNAMSIZ];
161 	int		ifm_current;
162 	int		ifm_mask;
163 	int		ifm_status;
164 	int		ifm_active;
165 	int		ifm_count;
166 	uint32_t	ifm_ulist;	/* (int *) */
167 };
168 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
169 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
170 
171 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
172     _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
173 #else /* !COMPAT_FREEBSD32 */
174 #define _CASE_IOC_IFGROUPREQ_32(cmd)
175 #endif /* !COMPAT_FREEBSD32 */
176 
177 #define CASE_IOC_IFGROUPREQ(cmd)	\
178     _CASE_IOC_IFGROUPREQ_32(cmd)	\
179     (cmd)
180 
181 union ifreq_union {
182 	struct ifreq	ifr;
183 #ifdef COMPAT_FREEBSD32
184 	struct ifreq32	ifr32;
185 #endif
186 };
187 
188 union ifgroupreq_union {
189 	struct ifgroupreq ifgr;
190 #ifdef COMPAT_FREEBSD32
191 	struct ifgroupreq32 ifgr32;
192 #endif
193 };
194 
195 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
196 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
197 
198 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
199     &ifqmaxlen, 0, "max send queue size");
200 
201 /* Log link state change events */
202 static int log_link_state_change = 1;
203 
204 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
205 	&log_link_state_change, 0,
206 	"log interface link state change events");
207 
208 /* Log promiscuous mode change events */
209 static int log_promisc_mode_change = 1;
210 
211 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
212 	&log_promisc_mode_change, 1,
213 	"log promiscuous mode change events");
214 
215 /* Interface description */
216 static unsigned int ifdescr_maxlen = 1024;
217 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
218 	&ifdescr_maxlen, 0,
219 	"administrative maximum length for interface description");
220 
221 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
222 
223 /* global sx for non-critical path ifdescr */
224 static struct sx ifdescr_sx;
225 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
226 
227 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
228 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
229 /* These are external hooks for CARP. */
230 void	(*carp_linkstate_p)(struct ifnet *ifp);
231 void	(*carp_demote_adj_p)(int, char *);
232 int	(*carp_master_p)(struct ifaddr *);
233 #if defined(INET) || defined(INET6)
234 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
235 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
236     const struct sockaddr *sa);
237 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
238 int	(*carp_attach_p)(struct ifaddr *, int);
239 void	(*carp_detach_p)(struct ifaddr *, bool);
240 #endif
241 #ifdef INET
242 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
243 #endif
244 #ifdef INET6
245 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
246 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
247     const struct in6_addr *taddr);
248 #endif
249 
250 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
251 
252 /*
253  * XXX: Style; these should be sorted alphabetically, and unprototyped
254  * static functions should be prototyped. Currently they are sorted by
255  * declaration order.
256  */
257 static void	if_attachdomain(void *);
258 static void	if_attachdomain1(struct ifnet *);
259 static int	ifconf(u_long, caddr_t);
260 static void	*if_grow(void);
261 static void	if_input_default(struct ifnet *, struct mbuf *);
262 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
263 static void	if_route(struct ifnet *, int flag, int fam);
264 static int	if_setflag(struct ifnet *, int, int, int *, int);
265 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
266 static void	if_unroute(struct ifnet *, int flag, int fam);
267 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void	do_link_state_change(void *, int);
269 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int	if_getgroupmembers(struct ifgroupreq *);
271 static void	if_delgroups(struct ifnet *);
272 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
274 #ifdef VIMAGE
275 static void	if_vmove(struct ifnet *, struct vnet *);
276 #endif
277 
278 #ifdef INET6
279 /*
280  * XXX: declare here to avoid to include many inet6 related files..
281  * should be more generalized?
282  */
283 extern void	nd6_setmtu(struct ifnet *);
284 #endif
285 
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
289 
290 VNET_DEFINE(int, if_index);
291 int	ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
294 
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
296 
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
299 
300 #define	V_if_indexlim		VNET(if_indexlim)
301 #define	V_ifindex_table		VNET(ifindex_table)
302 
303 /*
304  * The global network interface list (V_ifnet) and related state (such as
305  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306  * an rwlock.  Either may be acquired shared to stablize the list, but both
307  * must be acquired writable to modify the list.  This model allows us to
308  * both stablize the interface list during interrupt thread processing, but
309  * also to stablize it over long-running ioctls, without introducing priority
310  * inversions and deadlocks.
311  */
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
316 
317 /*
318  * The allocation of network interfaces is a rather non-atomic affair; we
319  * need to select an index before we are ready to expose the interface for
320  * use, so will use this pointer value to indicate reservation.
321  */
322 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
323 
324 static	if_com_alloc_t *if_com_alloc[256];
325 static	if_com_free_t *if_com_free[256];
326 
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
330 
331 struct ifnet *
332 ifnet_byindex(u_short idx)
333 {
334 	struct ifnet *ifp;
335 
336 	if (__predict_false(idx > V_if_index))
337 		return (NULL);
338 
339 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
340 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
341 }
342 
343 struct ifnet *
344 ifnet_byindex_ref(u_short idx)
345 {
346 	struct ifnet *ifp;
347 
348 	NET_EPOCH_ASSERT();
349 
350 	ifp = ifnet_byindex(idx);
351 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
352 		return (NULL);
353 	if_ref(ifp);
354 	return (ifp);
355 }
356 
357 /*
358  * Allocate an ifindex array entry; return 0 on success or an error on
359  * failure.
360  */
361 static u_short
362 ifindex_alloc(void **old)
363 {
364 	u_short idx;
365 
366 	IFNET_WLOCK_ASSERT();
367 	/*
368 	 * Try to find an empty slot below V_if_index.  If we fail, take the
369 	 * next slot.
370 	 */
371 	for (idx = 1; idx <= V_if_index; idx++) {
372 		if (V_ifindex_table[idx] == NULL)
373 			break;
374 	}
375 
376 	/* Catch if_index overflow. */
377 	if (idx >= V_if_indexlim) {
378 		*old = if_grow();
379 		return (USHRT_MAX);
380 	}
381 	if (idx > V_if_index)
382 		V_if_index = idx;
383 	return (idx);
384 }
385 
386 static void
387 ifindex_free_locked(u_short idx)
388 {
389 
390 	IFNET_WLOCK_ASSERT();
391 
392 	V_ifindex_table[idx] = NULL;
393 	while (V_if_index > 0 &&
394 	    V_ifindex_table[V_if_index] == NULL)
395 		V_if_index--;
396 }
397 
398 static void
399 ifindex_free(u_short idx)
400 {
401 
402 	IFNET_WLOCK();
403 	ifindex_free_locked(idx);
404 	IFNET_WUNLOCK();
405 }
406 
407 static void
408 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
409 {
410 
411 	V_ifindex_table[idx] = ifp;
412 }
413 
414 struct ifaddr *
415 ifaddr_byindex(u_short idx)
416 {
417 	struct ifnet *ifp;
418 	struct ifaddr *ifa = NULL;
419 
420 	NET_EPOCH_ASSERT();
421 
422 	ifp = ifnet_byindex(idx);
423 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
424 		ifa_ref(ifa);
425 	return (ifa);
426 }
427 
428 /*
429  * Network interface utility routines.
430  *
431  * Routines with ifa_ifwith* names take sockaddr *'s as
432  * parameters.
433  */
434 
435 static void
436 vnet_if_init(const void *unused __unused)
437 {
438 	void *old;
439 
440 	CK_STAILQ_INIT(&V_ifnet);
441 	CK_STAILQ_INIT(&V_ifg_head);
442 	IFNET_WLOCK();
443 	old = if_grow();				/* create initial table */
444 	IFNET_WUNLOCK();
445 	epoch_wait_preempt(net_epoch_preempt);
446 	free(old, M_IFNET);
447 	vnet_if_clone_init();
448 }
449 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
450     NULL);
451 
452 #ifdef VIMAGE
453 static void
454 vnet_if_uninit(const void *unused __unused)
455 {
456 
457 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
458 	    "not empty", __func__, __LINE__, &V_ifnet));
459 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
460 	    "not empty", __func__, __LINE__, &V_ifg_head));
461 
462 	free((caddr_t)V_ifindex_table, M_IFNET);
463 }
464 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
465     vnet_if_uninit, NULL);
466 
467 static void
468 vnet_if_return(const void *unused __unused)
469 {
470 	struct ifnet *ifp, *nifp;
471 
472 	/* Return all inherited interfaces to their parent vnets. */
473 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
474 		if (ifp->if_home_vnet != ifp->if_vnet)
475 			if_vmove(ifp, ifp->if_home_vnet);
476 	}
477 }
478 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
479     vnet_if_return, NULL);
480 #endif
481 
482 
483 static void *
484 if_grow(void)
485 {
486 	int oldlim;
487 	u_int n;
488 	struct ifnet **e;
489 	void *old;
490 
491 	old = NULL;
492 	IFNET_WLOCK_ASSERT();
493 	oldlim = V_if_indexlim;
494 	IFNET_WUNLOCK();
495 	n = (oldlim << 1) * sizeof(*e);
496 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
497 	IFNET_WLOCK();
498 	if (V_if_indexlim != oldlim) {
499 		free(e, M_IFNET);
500 		return (NULL);
501 	}
502 	if (V_ifindex_table != NULL) {
503 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
504 		old = V_ifindex_table;
505 	}
506 	V_if_indexlim <<= 1;
507 	V_ifindex_table = e;
508 	return (old);
509 }
510 
511 /*
512  * Allocate a struct ifnet and an index for an interface.  A layer 2
513  * common structure will also be allocated if an allocation routine is
514  * registered for the passed type.
515  */
516 struct ifnet *
517 if_alloc_domain(u_char type, int numa_domain)
518 {
519 	struct ifnet *ifp;
520 	u_short idx;
521 	void *old;
522 
523 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
524 	if (numa_domain == IF_NODOM)
525 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
526 		    M_WAITOK | M_ZERO);
527 	else
528 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
529 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
530  restart:
531 	IFNET_WLOCK();
532 	idx = ifindex_alloc(&old);
533 	if (__predict_false(idx == USHRT_MAX)) {
534 		IFNET_WUNLOCK();
535 		epoch_wait_preempt(net_epoch_preempt);
536 		free(old, M_IFNET);
537 		goto restart;
538 	}
539 	ifnet_setbyindex(idx, IFNET_HOLD);
540 	IFNET_WUNLOCK();
541 	ifp->if_index = idx;
542 	ifp->if_type = type;
543 	ifp->if_alloctype = type;
544 	ifp->if_numa_domain = numa_domain;
545 #ifdef VIMAGE
546 	ifp->if_vnet = curvnet;
547 #endif
548 	if (if_com_alloc[type] != NULL) {
549 		ifp->if_l2com = if_com_alloc[type](type, ifp);
550 		if (ifp->if_l2com == NULL) {
551 			free(ifp, M_IFNET);
552 			ifindex_free(idx);
553 			return (NULL);
554 		}
555 	}
556 
557 	IF_ADDR_LOCK_INIT(ifp);
558 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
559 	ifp->if_afdata_initialized = 0;
560 	IF_AFDATA_LOCK_INIT(ifp);
561 	CK_STAILQ_INIT(&ifp->if_addrhead);
562 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
563 	CK_STAILQ_INIT(&ifp->if_groups);
564 #ifdef MAC
565 	mac_ifnet_init(ifp);
566 #endif
567 	ifq_init(&ifp->if_snd, ifp);
568 
569 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
570 	for (int i = 0; i < IFCOUNTERS; i++)
571 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
572 	ifp->if_get_counter = if_get_counter_default;
573 	ifp->if_pcp = IFNET_PCP_NONE;
574 	ifnet_setbyindex(ifp->if_index, ifp);
575 	return (ifp);
576 }
577 
578 struct ifnet *
579 if_alloc_dev(u_char type, device_t dev)
580 {
581 	int numa_domain;
582 
583 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
584 		return (if_alloc_domain(type, IF_NODOM));
585 	return (if_alloc_domain(type, numa_domain));
586 }
587 
588 struct ifnet *
589 if_alloc(u_char type)
590 {
591 
592 	return (if_alloc_domain(type, IF_NODOM));
593 }
594 /*
595  * Do the actual work of freeing a struct ifnet, and layer 2 common
596  * structure.  This call is made when the last reference to an
597  * interface is released.
598  */
599 static void
600 if_free_internal(struct ifnet *ifp)
601 {
602 
603 	KASSERT((ifp->if_flags & IFF_DYING),
604 	    ("if_free_internal: interface not dying"));
605 
606 	if (if_com_free[ifp->if_alloctype] != NULL)
607 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
608 		    ifp->if_alloctype);
609 
610 #ifdef MAC
611 	mac_ifnet_destroy(ifp);
612 #endif /* MAC */
613 	IF_AFDATA_DESTROY(ifp);
614 	IF_ADDR_LOCK_DESTROY(ifp);
615 	ifq_delete(&ifp->if_snd);
616 
617 	for (int i = 0; i < IFCOUNTERS; i++)
618 		counter_u64_free(ifp->if_counters[i]);
619 
620 	free(ifp->if_description, M_IFDESCR);
621 	free(ifp->if_hw_addr, M_IFADDR);
622 	if (ifp->if_numa_domain == IF_NODOM)
623 		free(ifp, M_IFNET);
624 	else
625 		free_domain(ifp, M_IFNET);
626 }
627 
628 static void
629 if_destroy(epoch_context_t ctx)
630 {
631 	struct ifnet *ifp;
632 
633 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
634 	if_free_internal(ifp);
635 }
636 
637 /*
638  * Deregister an interface and free the associated storage.
639  */
640 void
641 if_free(struct ifnet *ifp)
642 {
643 
644 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
645 
646 	CURVNET_SET_QUIET(ifp->if_vnet);
647 	IFNET_WLOCK();
648 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
649 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
650 
651 	ifindex_free_locked(ifp->if_index);
652 	IFNET_WUNLOCK();
653 
654 	if (refcount_release(&ifp->if_refcount))
655 		epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
656 	CURVNET_RESTORE();
657 }
658 
659 /*
660  * Interfaces to keep an ifnet type-stable despite the possibility of the
661  * driver calling if_free().  If there are additional references, we defer
662  * freeing the underlying data structure.
663  */
664 void
665 if_ref(struct ifnet *ifp)
666 {
667 
668 	/* We don't assert the ifnet list lock here, but arguably should. */
669 	refcount_acquire(&ifp->if_refcount);
670 }
671 
672 void
673 if_rele(struct ifnet *ifp)
674 {
675 
676 	if (!refcount_release(&ifp->if_refcount))
677 		return;
678 	epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
679 }
680 
681 void
682 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
683 {
684 
685 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
686 
687 	if (ifq->ifq_maxlen == 0)
688 		ifq->ifq_maxlen = ifqmaxlen;
689 
690 	ifq->altq_type = 0;
691 	ifq->altq_disc = NULL;
692 	ifq->altq_flags &= ALTQF_CANTCHANGE;
693 	ifq->altq_tbr  = NULL;
694 	ifq->altq_ifp  = ifp;
695 }
696 
697 void
698 ifq_delete(struct ifaltq *ifq)
699 {
700 	mtx_destroy(&ifq->ifq_mtx);
701 }
702 
703 /*
704  * Perform generic interface initialization tasks and attach the interface
705  * to the list of "active" interfaces.  If vmove flag is set on entry
706  * to if_attach_internal(), perform only a limited subset of initialization
707  * tasks, given that we are moving from one vnet to another an ifnet which
708  * has already been fully initialized.
709  *
710  * Note that if_detach_internal() removes group membership unconditionally
711  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
712  * Thus, when if_vmove() is applied to a cloned interface, group membership
713  * is lost while a cloned one always joins a group whose name is
714  * ifc->ifc_name.  To recover this after if_detach_internal() and
715  * if_attach_internal(), the cloner should be specified to
716  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
717  * attempts to join a group whose name is ifc->ifc_name.
718  *
719  * XXX:
720  *  - The decision to return void and thus require this function to
721  *    succeed is questionable.
722  *  - We should probably do more sanity checking.  For instance we don't
723  *    do anything to insure if_xname is unique or non-empty.
724  */
725 void
726 if_attach(struct ifnet *ifp)
727 {
728 
729 	if_attach_internal(ifp, 0, NULL);
730 }
731 
732 /*
733  * Compute the least common TSO limit.
734  */
735 void
736 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
737 {
738 	/*
739 	 * 1) If there is no limit currently, take the limit from
740 	 * the network adapter.
741 	 *
742 	 * 2) If the network adapter has a limit below the current
743 	 * limit, apply it.
744 	 */
745 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
746 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
747 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
748 	}
749 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
750 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
751 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
752 	}
753 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
754 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
755 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
756 	}
757 }
758 
759 /*
760  * Update TSO limit of a network adapter.
761  *
762  * Returns zero if no change. Else non-zero.
763  */
764 int
765 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
766 {
767 	int retval = 0;
768 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
769 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
770 		retval++;
771 	}
772 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
773 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
774 		retval++;
775 	}
776 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
777 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
778 		retval++;
779 	}
780 	return (retval);
781 }
782 
783 static void
784 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
785 {
786 	unsigned socksize, ifasize;
787 	int namelen, masklen;
788 	struct sockaddr_dl *sdl;
789 	struct ifaddr *ifa;
790 
791 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
792 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
793 		    ifp->if_xname);
794 
795 #ifdef VIMAGE
796 	ifp->if_vnet = curvnet;
797 	if (ifp->if_home_vnet == NULL)
798 		ifp->if_home_vnet = curvnet;
799 #endif
800 
801 	if_addgroup(ifp, IFG_ALL);
802 
803 	/* Restore group membership for cloned interfaces. */
804 	if (vmove && ifc != NULL)
805 		if_clone_addgroup(ifp, ifc);
806 
807 	getmicrotime(&ifp->if_lastchange);
808 	ifp->if_epoch = time_uptime;
809 
810 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
811 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
812 	    ("transmit and qflush must both either be set or both be NULL"));
813 	if (ifp->if_transmit == NULL) {
814 		ifp->if_transmit = if_transmit;
815 		ifp->if_qflush = if_qflush;
816 	}
817 	if (ifp->if_input == NULL)
818 		ifp->if_input = if_input_default;
819 
820 	if (ifp->if_requestencap == NULL)
821 		ifp->if_requestencap = if_requestencap_default;
822 
823 	if (!vmove) {
824 #ifdef MAC
825 		mac_ifnet_create(ifp);
826 #endif
827 
828 		/*
829 		 * Create a Link Level name for this device.
830 		 */
831 		namelen = strlen(ifp->if_xname);
832 		/*
833 		 * Always save enough space for any possiable name so we
834 		 * can do a rename in place later.
835 		 */
836 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
837 		socksize = masklen + ifp->if_addrlen;
838 		if (socksize < sizeof(*sdl))
839 			socksize = sizeof(*sdl);
840 		socksize = roundup2(socksize, sizeof(long));
841 		ifasize = sizeof(*ifa) + 2 * socksize;
842 		ifa = ifa_alloc(ifasize, M_WAITOK);
843 		sdl = (struct sockaddr_dl *)(ifa + 1);
844 		sdl->sdl_len = socksize;
845 		sdl->sdl_family = AF_LINK;
846 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
847 		sdl->sdl_nlen = namelen;
848 		sdl->sdl_index = ifp->if_index;
849 		sdl->sdl_type = ifp->if_type;
850 		ifp->if_addr = ifa;
851 		ifa->ifa_ifp = ifp;
852 		ifa->ifa_addr = (struct sockaddr *)sdl;
853 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
854 		ifa->ifa_netmask = (struct sockaddr *)sdl;
855 		sdl->sdl_len = masklen;
856 		while (namelen != 0)
857 			sdl->sdl_data[--namelen] = 0xff;
858 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
859 		/* Reliably crash if used uninitialized. */
860 		ifp->if_broadcastaddr = NULL;
861 
862 		if (ifp->if_type == IFT_ETHER) {
863 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
864 			    M_WAITOK | M_ZERO);
865 		}
866 
867 #if defined(INET) || defined(INET6)
868 		/* Use defaults for TSO, if nothing is set */
869 		if (ifp->if_hw_tsomax == 0 &&
870 		    ifp->if_hw_tsomaxsegcount == 0 &&
871 		    ifp->if_hw_tsomaxsegsize == 0) {
872 			/*
873 			 * The TSO defaults needs to be such that an
874 			 * NFS mbuf list of 35 mbufs totalling just
875 			 * below 64K works and that a chain of mbufs
876 			 * can be defragged into at most 32 segments:
877 			 */
878 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
879 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
880 			ifp->if_hw_tsomaxsegcount = 35;
881 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
882 
883 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
884 			if (ifp->if_capabilities & IFCAP_TSO) {
885 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
886 				    ifp->if_hw_tsomax,
887 				    ifp->if_hw_tsomaxsegcount,
888 				    ifp->if_hw_tsomaxsegsize);
889 			}
890 		}
891 #endif
892 	}
893 #ifdef VIMAGE
894 	else {
895 		/*
896 		 * Update the interface index in the link layer address
897 		 * of the interface.
898 		 */
899 		for (ifa = ifp->if_addr; ifa != NULL;
900 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
901 			if (ifa->ifa_addr->sa_family == AF_LINK) {
902 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
903 				sdl->sdl_index = ifp->if_index;
904 			}
905 		}
906 	}
907 #endif
908 
909 	IFNET_WLOCK();
910 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
911 #ifdef VIMAGE
912 	curvnet->vnet_ifcnt++;
913 #endif
914 	IFNET_WUNLOCK();
915 
916 	if (domain_init_status >= 2)
917 		if_attachdomain1(ifp);
918 
919 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
920 	if (IS_DEFAULT_VNET(curvnet))
921 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
922 
923 	/* Announce the interface. */
924 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
925 }
926 
927 static void
928 if_epochalloc(void *dummy __unused)
929 {
930 
931 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
932 	net_epoch = epoch_alloc("Net", 0);
933 }
934 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
935     if_epochalloc, NULL);
936 
937 static void
938 if_attachdomain(void *dummy)
939 {
940 	struct ifnet *ifp;
941 
942 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
943 		if_attachdomain1(ifp);
944 }
945 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
946     if_attachdomain, NULL);
947 
948 static void
949 if_attachdomain1(struct ifnet *ifp)
950 {
951 	struct domain *dp;
952 
953 	/*
954 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
955 	 * cannot lock ifp->if_afdata initialization, entirely.
956 	 */
957 	IF_AFDATA_LOCK(ifp);
958 	if (ifp->if_afdata_initialized >= domain_init_status) {
959 		IF_AFDATA_UNLOCK(ifp);
960 		log(LOG_WARNING, "%s called more than once on %s\n",
961 		    __func__, ifp->if_xname);
962 		return;
963 	}
964 	ifp->if_afdata_initialized = domain_init_status;
965 	IF_AFDATA_UNLOCK(ifp);
966 
967 	/* address family dependent data region */
968 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
969 	for (dp = domains; dp; dp = dp->dom_next) {
970 		if (dp->dom_ifattach)
971 			ifp->if_afdata[dp->dom_family] =
972 			    (*dp->dom_ifattach)(ifp);
973 	}
974 }
975 
976 /*
977  * Remove any unicast or broadcast network addresses from an interface.
978  */
979 void
980 if_purgeaddrs(struct ifnet *ifp)
981 {
982 	struct ifaddr *ifa;
983 
984 	while (1) {
985 		struct epoch_tracker et;
986 
987 		NET_EPOCH_ENTER(et);
988 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
989 			if (ifa->ifa_addr->sa_family != AF_LINK)
990 				break;
991 		}
992 		NET_EPOCH_EXIT(et);
993 
994 		if (ifa == NULL)
995 			break;
996 #ifdef INET
997 		/* XXX: Ugly!! ad hoc just for INET */
998 		if (ifa->ifa_addr->sa_family == AF_INET) {
999 			struct ifaliasreq ifr;
1000 
1001 			bzero(&ifr, sizeof(ifr));
1002 			ifr.ifra_addr = *ifa->ifa_addr;
1003 			if (ifa->ifa_dstaddr)
1004 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1005 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1006 			    NULL) == 0)
1007 				continue;
1008 		}
1009 #endif /* INET */
1010 #ifdef INET6
1011 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1012 			in6_purgeaddr(ifa);
1013 			/* ifp_addrhead is already updated */
1014 			continue;
1015 		}
1016 #endif /* INET6 */
1017 		IF_ADDR_WLOCK(ifp);
1018 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1019 		IF_ADDR_WUNLOCK(ifp);
1020 		ifa_free(ifa);
1021 	}
1022 }
1023 
1024 /*
1025  * Remove any multicast network addresses from an interface when an ifnet
1026  * is going away.
1027  */
1028 static void
1029 if_purgemaddrs(struct ifnet *ifp)
1030 {
1031 	struct ifmultiaddr *ifma;
1032 
1033 	IF_ADDR_WLOCK(ifp);
1034 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1035 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1036 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1037 		if_delmulti_locked(ifp, ifma, 1);
1038 	}
1039 	IF_ADDR_WUNLOCK(ifp);
1040 }
1041 
1042 /*
1043  * Detach an interface, removing it from the list of "active" interfaces.
1044  * If vmove flag is set on entry to if_detach_internal(), perform only a
1045  * limited subset of cleanup tasks, given that we are moving an ifnet from
1046  * one vnet to another, where it must be fully operational.
1047  *
1048  * XXXRW: There are some significant questions about event ordering, and
1049  * how to prevent things from starting to use the interface during detach.
1050  */
1051 void
1052 if_detach(struct ifnet *ifp)
1053 {
1054 
1055 	CURVNET_SET_QUIET(ifp->if_vnet);
1056 	if_detach_internal(ifp, 0, NULL);
1057 	CURVNET_RESTORE();
1058 }
1059 
1060 /*
1061  * The vmove flag, if set, indicates that we are called from a callpath
1062  * that is moving an interface to a different vnet instance.
1063  *
1064  * The shutdown flag, if set, indicates that we are called in the
1065  * process of shutting down a vnet instance.  Currently only the
1066  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1067  * on a vnet instance shutdown without this flag being set, e.g., when
1068  * the cloned interfaces are destoyed as first thing of teardown.
1069  */
1070 static int
1071 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1072 {
1073 	struct ifaddr *ifa;
1074 	int i;
1075 	struct domain *dp;
1076  	struct ifnet *iter;
1077  	int found = 0;
1078 #ifdef VIMAGE
1079 	bool shutdown;
1080 
1081 	shutdown = ifp->if_vnet->vnet_shutdown;
1082 #endif
1083 	IFNET_WLOCK();
1084 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1085 		if (iter == ifp) {
1086 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1087 			if (!vmove)
1088 				ifp->if_flags |= IFF_DYING;
1089 			found = 1;
1090 			break;
1091 		}
1092 	IFNET_WUNLOCK();
1093 	if (!found) {
1094 		/*
1095 		 * While we would want to panic here, we cannot
1096 		 * guarantee that the interface is indeed still on
1097 		 * the list given we don't hold locks all the way.
1098 		 */
1099 		return (ENOENT);
1100 #if 0
1101 		if (vmove)
1102 			panic("%s: ifp=%p not on the ifnet tailq %p",
1103 			    __func__, ifp, &V_ifnet);
1104 		else
1105 			return; /* XXX this should panic as well? */
1106 #endif
1107 	}
1108 
1109 	/*
1110 	 * At this point we know the interface still was on the ifnet list
1111 	 * and we removed it so we are in a stable state.
1112 	 */
1113 #ifdef VIMAGE
1114 	curvnet->vnet_ifcnt--;
1115 #endif
1116 	epoch_wait_preempt(net_epoch_preempt);
1117 
1118 	/*
1119 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1120 	 * fixes issues about late destruction of multicast options
1121 	 * which lead to leave group calls, which in turn access the
1122 	 * belonging ifnet structure:
1123 	 */
1124 	epoch_drain_callbacks(net_epoch_preempt);
1125 
1126 	/*
1127 	 * In any case (destroy or vmove) detach us from the groups
1128 	 * and remove/wait for pending events on the taskq.
1129 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1130 	 */
1131 	if_delgroups(ifp);
1132 
1133 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1134 
1135 	/*
1136 	 * Check if this is a cloned interface or not. Must do even if
1137 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1138 	 * the if_clone_addgroup() will have a corrupted string overwise
1139 	 * from a gibberish pointer.
1140 	 */
1141 	if (vmove && ifcp != NULL)
1142 		*ifcp = if_clone_findifc(ifp);
1143 
1144 	if_down(ifp);
1145 
1146 #ifdef VIMAGE
1147 	/*
1148 	 * On VNET shutdown abort here as the stack teardown will do all
1149 	 * the work top-down for us.
1150 	 */
1151 	if (shutdown) {
1152 		/* Give interface users the chance to clean up. */
1153 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1154 
1155 		/*
1156 		 * In case of a vmove we are done here without error.
1157 		 * If we would signal an error it would lead to the same
1158 		 * abort as if we did not find the ifnet anymore.
1159 		 * if_detach() calls us in void context and does not care
1160 		 * about an early abort notification, so life is splendid :)
1161 		 */
1162 		goto finish_vnet_shutdown;
1163 	}
1164 #endif
1165 
1166 	/*
1167 	 * At this point we are not tearing down a VNET and are either
1168 	 * going to destroy or vmove the interface and have to cleanup
1169 	 * accordingly.
1170 	 */
1171 
1172 	/*
1173 	 * Remove routes and flush queues.
1174 	 */
1175 #ifdef ALTQ
1176 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1177 		altq_disable(&ifp->if_snd);
1178 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1179 		altq_detach(&ifp->if_snd);
1180 #endif
1181 
1182 	if_purgeaddrs(ifp);
1183 
1184 #ifdef INET
1185 	in_ifdetach(ifp);
1186 #endif
1187 
1188 #ifdef INET6
1189 	/*
1190 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1191 	 * before removing routing entries below, since IPv6 interface direct
1192 	 * routes are expected to be removed by the IPv6-specific kernel API.
1193 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1194 	 */
1195 	in6_ifdetach(ifp);
1196 #endif
1197 	if_purgemaddrs(ifp);
1198 
1199 	/* Announce that the interface is gone. */
1200 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1201 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1202 	if (IS_DEFAULT_VNET(curvnet))
1203 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1204 
1205 	if (!vmove) {
1206 		/*
1207 		 * Prevent further calls into the device driver via ifnet.
1208 		 */
1209 		if_dead(ifp);
1210 
1211 		/*
1212 		 * Clean up all addresses.
1213 		 */
1214 		IF_ADDR_WLOCK(ifp);
1215 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1216 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1217 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1218 			IF_ADDR_WUNLOCK(ifp);
1219 			ifa_free(ifa);
1220 		} else
1221 			IF_ADDR_WUNLOCK(ifp);
1222 	}
1223 
1224 	rt_flushifroutes(ifp);
1225 
1226 #ifdef VIMAGE
1227 finish_vnet_shutdown:
1228 #endif
1229 	/*
1230 	 * We cannot hold the lock over dom_ifdetach calls as they might
1231 	 * sleep, for example trying to drain a callout, thus open up the
1232 	 * theoretical race with re-attaching.
1233 	 */
1234 	IF_AFDATA_LOCK(ifp);
1235 	i = ifp->if_afdata_initialized;
1236 	ifp->if_afdata_initialized = 0;
1237 	IF_AFDATA_UNLOCK(ifp);
1238 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1239 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1240 			(*dp->dom_ifdetach)(ifp,
1241 			    ifp->if_afdata[dp->dom_family]);
1242 			ifp->if_afdata[dp->dom_family] = NULL;
1243 		}
1244 	}
1245 
1246 	return (0);
1247 }
1248 
1249 #ifdef VIMAGE
1250 /*
1251  * if_vmove() performs a limited version of if_detach() in current
1252  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1253  * An attempt is made to shrink if_index in current vnet, find an
1254  * unused if_index in target vnet and calls if_grow() if necessary,
1255  * and finally find an unused if_xname for the target vnet.
1256  */
1257 static void
1258 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1259 {
1260 	struct if_clone *ifc;
1261 	u_int bif_dlt, bif_hdrlen;
1262 	void *old;
1263 	int rc;
1264 
1265  	/*
1266 	 * if_detach_internal() will call the eventhandler to notify
1267 	 * interface departure.  That will detach if_bpf.  We need to
1268 	 * safe the dlt and hdrlen so we can re-attach it later.
1269 	 */
1270 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1271 
1272 	/*
1273 	 * Detach from current vnet, but preserve LLADDR info, do not
1274 	 * mark as dead etc. so that the ifnet can be reattached later.
1275 	 * If we cannot find it, we lost the race to someone else.
1276 	 */
1277 	rc = if_detach_internal(ifp, 1, &ifc);
1278 	if (rc != 0)
1279 		return;
1280 
1281 	/*
1282 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1283 	 * the if_index for that vnet if possible.
1284 	 *
1285 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1286 	 * or we'd lock on one vnet and unlock on another.
1287 	 */
1288 	IFNET_WLOCK();
1289 	ifindex_free_locked(ifp->if_index);
1290 	IFNET_WUNLOCK();
1291 
1292 	/*
1293 	 * Perform interface-specific reassignment tasks, if provided by
1294 	 * the driver.
1295 	 */
1296 	if (ifp->if_reassign != NULL)
1297 		ifp->if_reassign(ifp, new_vnet, NULL);
1298 
1299 	/*
1300 	 * Switch to the context of the target vnet.
1301 	 */
1302 	CURVNET_SET_QUIET(new_vnet);
1303  restart:
1304 	IFNET_WLOCK();
1305 	ifp->if_index = ifindex_alloc(&old);
1306 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1307 		IFNET_WUNLOCK();
1308 		epoch_wait_preempt(net_epoch_preempt);
1309 		free(old, M_IFNET);
1310 		goto restart;
1311 	}
1312 	ifnet_setbyindex(ifp->if_index, ifp);
1313 	IFNET_WUNLOCK();
1314 
1315 	if_attach_internal(ifp, 1, ifc);
1316 
1317 	if (ifp->if_bpf == NULL)
1318 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1319 
1320 	CURVNET_RESTORE();
1321 }
1322 
1323 /*
1324  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1325  */
1326 static int
1327 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1328 {
1329 	struct prison *pr;
1330 	struct ifnet *difp;
1331 
1332 	/* Try to find the prison within our visibility. */
1333 	sx_slock(&allprison_lock);
1334 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1335 	sx_sunlock(&allprison_lock);
1336 	if (pr == NULL)
1337 		return (ENXIO);
1338 	prison_hold_locked(pr);
1339 	mtx_unlock(&pr->pr_mtx);
1340 
1341 	/* Do not try to move the iface from and to the same prison. */
1342 	if (pr->pr_vnet == ifp->if_vnet) {
1343 		prison_free(pr);
1344 		return (EEXIST);
1345 	}
1346 
1347 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1348 	/* XXX Lock interfaces to avoid races. */
1349 	CURVNET_SET_QUIET(pr->pr_vnet);
1350 	difp = ifunit(ifname);
1351 	if (difp != NULL) {
1352 		CURVNET_RESTORE();
1353 		prison_free(pr);
1354 		return (EEXIST);
1355 	}
1356 
1357 	/* Make sure the VNET is stable. */
1358 	if (ifp->if_vnet->vnet_shutdown) {
1359 		CURVNET_RESTORE();
1360 		prison_free(pr);
1361 		return (EBUSY);
1362 	}
1363 	CURVNET_RESTORE();
1364 
1365 	/* Move the interface into the child jail/vnet. */
1366 	if_vmove(ifp, pr->pr_vnet);
1367 
1368 	/* Report the new if_xname back to the userland. */
1369 	sprintf(ifname, "%s", ifp->if_xname);
1370 
1371 	prison_free(pr);
1372 	return (0);
1373 }
1374 
1375 static int
1376 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1377 {
1378 	struct prison *pr;
1379 	struct vnet *vnet_dst;
1380 	struct ifnet *ifp;
1381 
1382 	/* Try to find the prison within our visibility. */
1383 	sx_slock(&allprison_lock);
1384 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1385 	sx_sunlock(&allprison_lock);
1386 	if (pr == NULL)
1387 		return (ENXIO);
1388 	prison_hold_locked(pr);
1389 	mtx_unlock(&pr->pr_mtx);
1390 
1391 	/* Make sure the named iface exists in the source prison/vnet. */
1392 	CURVNET_SET(pr->pr_vnet);
1393 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1394 	if (ifp == NULL) {
1395 		CURVNET_RESTORE();
1396 		prison_free(pr);
1397 		return (ENXIO);
1398 	}
1399 
1400 	/* Do not try to move the iface from and to the same prison. */
1401 	vnet_dst = TD_TO_VNET(td);
1402 	if (vnet_dst == ifp->if_vnet) {
1403 		CURVNET_RESTORE();
1404 		prison_free(pr);
1405 		return (EEXIST);
1406 	}
1407 
1408 	/* Make sure the VNET is stable. */
1409 	if (ifp->if_vnet->vnet_shutdown) {
1410 		CURVNET_RESTORE();
1411 		prison_free(pr);
1412 		return (EBUSY);
1413 	}
1414 
1415 	/* Get interface back from child jail/vnet. */
1416 	if_vmove(ifp, vnet_dst);
1417 	CURVNET_RESTORE();
1418 
1419 	/* Report the new if_xname back to the userland. */
1420 	sprintf(ifname, "%s", ifp->if_xname);
1421 
1422 	prison_free(pr);
1423 	return (0);
1424 }
1425 #endif /* VIMAGE */
1426 
1427 /*
1428  * Add a group to an interface
1429  */
1430 int
1431 if_addgroup(struct ifnet *ifp, const char *groupname)
1432 {
1433 	struct ifg_list		*ifgl;
1434 	struct ifg_group	*ifg = NULL;
1435 	struct ifg_member	*ifgm;
1436 	int 			 new = 0;
1437 
1438 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1439 	    groupname[strlen(groupname) - 1] <= '9')
1440 		return (EINVAL);
1441 
1442 	IFNET_WLOCK();
1443 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1444 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1445 			IFNET_WUNLOCK();
1446 			return (EEXIST);
1447 		}
1448 
1449 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1450 	    M_NOWAIT)) == NULL) {
1451 	    	IFNET_WUNLOCK();
1452 		return (ENOMEM);
1453 	}
1454 
1455 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1456 	    M_TEMP, M_NOWAIT)) == NULL) {
1457 		free(ifgl, M_TEMP);
1458 		IFNET_WUNLOCK();
1459 		return (ENOMEM);
1460 	}
1461 
1462 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1463 		if (!strcmp(ifg->ifg_group, groupname))
1464 			break;
1465 
1466 	if (ifg == NULL) {
1467 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1468 		    M_TEMP, M_NOWAIT)) == NULL) {
1469 			free(ifgl, M_TEMP);
1470 			free(ifgm, M_TEMP);
1471 			IFNET_WUNLOCK();
1472 			return (ENOMEM);
1473 		}
1474 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1475 		ifg->ifg_refcnt = 0;
1476 		CK_STAILQ_INIT(&ifg->ifg_members);
1477 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1478 		new = 1;
1479 	}
1480 
1481 	ifg->ifg_refcnt++;
1482 	ifgl->ifgl_group = ifg;
1483 	ifgm->ifgm_ifp = ifp;
1484 
1485 	IF_ADDR_WLOCK(ifp);
1486 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1487 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1488 	IF_ADDR_WUNLOCK(ifp);
1489 
1490 	IFNET_WUNLOCK();
1491 
1492 	if (new)
1493 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1494 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1495 
1496 	return (0);
1497 }
1498 
1499 /*
1500  * Remove a group from an interface
1501  */
1502 int
1503 if_delgroup(struct ifnet *ifp, const char *groupname)
1504 {
1505 	struct ifg_list		*ifgl;
1506 	struct ifg_member	*ifgm;
1507 	int freeifgl;
1508 
1509 	IFNET_WLOCK();
1510 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1511 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1512 			break;
1513 	if (ifgl == NULL) {
1514 		IFNET_WUNLOCK();
1515 		return (ENOENT);
1516 	}
1517 
1518 	freeifgl = 0;
1519 	IF_ADDR_WLOCK(ifp);
1520 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1521 	IF_ADDR_WUNLOCK(ifp);
1522 
1523 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1524 		if (ifgm->ifgm_ifp == ifp)
1525 			break;
1526 
1527 	if (ifgm != NULL)
1528 		CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1529 
1530 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1531 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1532 		freeifgl = 1;
1533 	}
1534 	IFNET_WUNLOCK();
1535 
1536 	epoch_wait_preempt(net_epoch_preempt);
1537 	if (freeifgl) {
1538 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1539 		free(ifgl->ifgl_group, M_TEMP);
1540 	}
1541 	free(ifgm, M_TEMP);
1542 	free(ifgl, M_TEMP);
1543 
1544 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1545 
1546 	return (0);
1547 }
1548 
1549 /*
1550  * Remove an interface from all groups
1551  */
1552 static void
1553 if_delgroups(struct ifnet *ifp)
1554 {
1555 	struct ifg_list		*ifgl;
1556 	struct ifg_member	*ifgm;
1557 	char groupname[IFNAMSIZ];
1558 	int ifglfree;
1559 
1560 	IFNET_WLOCK();
1561 	while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1562 		ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1563 
1564 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1565 
1566 		IF_ADDR_WLOCK(ifp);
1567 		CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1568 		IF_ADDR_WUNLOCK(ifp);
1569 
1570 		CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1571 			if (ifgm->ifgm_ifp == ifp)
1572 				break;
1573 
1574 		if (ifgm != NULL)
1575 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1576 			    ifgm_next);
1577 		ifglfree = 0;
1578 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1579 			CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1580 			ifglfree = 1;
1581 		}
1582 
1583 		IFNET_WUNLOCK();
1584 		epoch_wait_preempt(net_epoch_preempt);
1585 		free(ifgm, M_TEMP);
1586 		if (ifglfree) {
1587 			EVENTHANDLER_INVOKE(group_detach_event,
1588 								ifgl->ifgl_group);
1589 			free(ifgl->ifgl_group, M_TEMP);
1590 		}
1591 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1592 
1593 		IFNET_WLOCK();
1594 	}
1595 	IFNET_WUNLOCK();
1596 }
1597 
1598 static char *
1599 ifgr_group_get(void *ifgrp)
1600 {
1601 	union ifgroupreq_union *ifgrup;
1602 
1603 	ifgrup = ifgrp;
1604 #ifdef COMPAT_FREEBSD32
1605 	if (SV_CURPROC_FLAG(SV_ILP32))
1606 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1607 #endif
1608 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1609 }
1610 
1611 static struct ifg_req *
1612 ifgr_groups_get(void *ifgrp)
1613 {
1614 	union ifgroupreq_union *ifgrup;
1615 
1616 	ifgrup = ifgrp;
1617 #ifdef COMPAT_FREEBSD32
1618 	if (SV_CURPROC_FLAG(SV_ILP32))
1619 		return ((struct ifg_req *)(uintptr_t)
1620 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1621 #endif
1622 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1623 }
1624 
1625 /*
1626  * Stores all groups from an interface in memory pointed to by ifgr.
1627  */
1628 static int
1629 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1630 {
1631 	int			 len, error;
1632 	struct ifg_list		*ifgl;
1633 	struct ifg_req		 ifgrq, *ifgp;
1634 
1635 	NET_EPOCH_ASSERT();
1636 
1637 	if (ifgr->ifgr_len == 0) {
1638 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1639 			ifgr->ifgr_len += sizeof(struct ifg_req);
1640 		return (0);
1641 	}
1642 
1643 	len = ifgr->ifgr_len;
1644 	ifgp = ifgr_groups_get(ifgr);
1645 	/* XXX: wire */
1646 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1647 		if (len < sizeof(ifgrq))
1648 			return (EINVAL);
1649 		bzero(&ifgrq, sizeof ifgrq);
1650 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1651 		    sizeof(ifgrq.ifgrq_group));
1652 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1653 			return (error);
1654 		len -= sizeof(ifgrq);
1655 		ifgp++;
1656 	}
1657 
1658 	return (0);
1659 }
1660 
1661 /*
1662  * Stores all members of a group in memory pointed to by igfr
1663  */
1664 static int
1665 if_getgroupmembers(struct ifgroupreq *ifgr)
1666 {
1667 	struct ifg_group	*ifg;
1668 	struct ifg_member	*ifgm;
1669 	struct ifg_req		 ifgrq, *ifgp;
1670 	int			 len, error;
1671 
1672 	IFNET_RLOCK();
1673 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1674 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1675 			break;
1676 	if (ifg == NULL) {
1677 		IFNET_RUNLOCK();
1678 		return (ENOENT);
1679 	}
1680 
1681 	if (ifgr->ifgr_len == 0) {
1682 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1683 			ifgr->ifgr_len += sizeof(ifgrq);
1684 		IFNET_RUNLOCK();
1685 		return (0);
1686 	}
1687 
1688 	len = ifgr->ifgr_len;
1689 	ifgp = ifgr_groups_get(ifgr);
1690 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1691 		if (len < sizeof(ifgrq)) {
1692 			IFNET_RUNLOCK();
1693 			return (EINVAL);
1694 		}
1695 		bzero(&ifgrq, sizeof ifgrq);
1696 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1697 		    sizeof(ifgrq.ifgrq_member));
1698 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1699 			IFNET_RUNLOCK();
1700 			return (error);
1701 		}
1702 		len -= sizeof(ifgrq);
1703 		ifgp++;
1704 	}
1705 	IFNET_RUNLOCK();
1706 
1707 	return (0);
1708 }
1709 
1710 /*
1711  * Return counter values from counter(9)s stored in ifnet.
1712  */
1713 uint64_t
1714 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1715 {
1716 
1717 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1718 
1719 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1720 }
1721 
1722 /*
1723  * Increase an ifnet counter. Usually used for counters shared
1724  * between the stack and a driver, but function supports them all.
1725  */
1726 void
1727 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1728 {
1729 
1730 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1731 
1732 	counter_u64_add(ifp->if_counters[cnt], inc);
1733 }
1734 
1735 /*
1736  * Copy data from ifnet to userland API structure if_data.
1737  */
1738 void
1739 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1740 {
1741 
1742 	ifd->ifi_type = ifp->if_type;
1743 	ifd->ifi_physical = 0;
1744 	ifd->ifi_addrlen = ifp->if_addrlen;
1745 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1746 	ifd->ifi_link_state = ifp->if_link_state;
1747 	ifd->ifi_vhid = 0;
1748 	ifd->ifi_datalen = sizeof(struct if_data);
1749 	ifd->ifi_mtu = ifp->if_mtu;
1750 	ifd->ifi_metric = ifp->if_metric;
1751 	ifd->ifi_baudrate = ifp->if_baudrate;
1752 	ifd->ifi_hwassist = ifp->if_hwassist;
1753 	ifd->ifi_epoch = ifp->if_epoch;
1754 	ifd->ifi_lastchange = ifp->if_lastchange;
1755 
1756 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1757 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1758 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1759 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1760 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1761 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1762 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1763 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1764 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1765 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1766 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1767 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1768 }
1769 
1770 /*
1771  * Initialization, destruction and refcounting functions for ifaddrs.
1772  */
1773 struct ifaddr *
1774 ifa_alloc(size_t size, int flags)
1775 {
1776 	struct ifaddr *ifa;
1777 
1778 	KASSERT(size >= sizeof(struct ifaddr),
1779 	    ("%s: invalid size %zu", __func__, size));
1780 
1781 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1782 	if (ifa == NULL)
1783 		return (NULL);
1784 
1785 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1786 		goto fail;
1787 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1788 		goto fail;
1789 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1790 		goto fail;
1791 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1792 		goto fail;
1793 
1794 	refcount_init(&ifa->ifa_refcnt, 1);
1795 
1796 	return (ifa);
1797 
1798 fail:
1799 	/* free(NULL) is okay */
1800 	counter_u64_free(ifa->ifa_opackets);
1801 	counter_u64_free(ifa->ifa_ipackets);
1802 	counter_u64_free(ifa->ifa_obytes);
1803 	counter_u64_free(ifa->ifa_ibytes);
1804 	free(ifa, M_IFADDR);
1805 
1806 	return (NULL);
1807 }
1808 
1809 void
1810 ifa_ref(struct ifaddr *ifa)
1811 {
1812 
1813 	refcount_acquire(&ifa->ifa_refcnt);
1814 }
1815 
1816 static void
1817 ifa_destroy(epoch_context_t ctx)
1818 {
1819 	struct ifaddr *ifa;
1820 
1821 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1822 	counter_u64_free(ifa->ifa_opackets);
1823 	counter_u64_free(ifa->ifa_ipackets);
1824 	counter_u64_free(ifa->ifa_obytes);
1825 	counter_u64_free(ifa->ifa_ibytes);
1826 	free(ifa, M_IFADDR);
1827 }
1828 
1829 void
1830 ifa_free(struct ifaddr *ifa)
1831 {
1832 
1833 	if (refcount_release(&ifa->ifa_refcnt))
1834 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1835 }
1836 
1837 
1838 static int
1839 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1840     struct sockaddr *ia)
1841 {
1842 	struct epoch_tracker et;
1843 	int error;
1844 	struct rt_addrinfo info;
1845 	struct sockaddr_dl null_sdl;
1846 	struct ifnet *ifp;
1847 
1848 	ifp = ifa->ifa_ifp;
1849 
1850 	bzero(&info, sizeof(info));
1851 	if (cmd != RTM_DELETE)
1852 		info.rti_ifp = V_loif;
1853 	if (cmd == RTM_ADD) {
1854 		/* explicitly specify (loopback) ifa */
1855 		if (info.rti_ifp != NULL) {
1856 			NET_EPOCH_ENTER(et);
1857 			info.rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1858 			if (info.rti_ifa != NULL)
1859 				ifa_ref(info.rti_ifa);
1860 			NET_EPOCH_EXIT(et);
1861 		}
1862 	}
1863 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1864 	info.rti_info[RTAX_DST] = ia;
1865 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1866 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1867 
1868 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1869 
1870 	if (error != 0 &&
1871 	    !(cmd == RTM_ADD && error == EEXIST) &&
1872 	    !(cmd == RTM_DELETE && error == ENOENT))
1873 		if_printf(ifp, "%s failed: %d\n", otype, error);
1874 
1875 	return (error);
1876 }
1877 
1878 int
1879 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1880 {
1881 
1882 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1883 }
1884 
1885 int
1886 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1887 {
1888 
1889 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1890 }
1891 
1892 int
1893 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1894 {
1895 
1896 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1897 }
1898 
1899 /*
1900  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1901  * structs used to represent other address families, it is necessary
1902  * to perform a different comparison.
1903  */
1904 
1905 #define	sa_dl_equal(a1, a2)	\
1906 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1907 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1908 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1909 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1910 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1911 
1912 /*
1913  * Locate an interface based on a complete address.
1914  */
1915 /*ARGSUSED*/
1916 struct ifaddr *
1917 ifa_ifwithaddr(const struct sockaddr *addr)
1918 {
1919 	struct ifnet *ifp;
1920 	struct ifaddr *ifa;
1921 
1922 	NET_EPOCH_ASSERT();
1923 
1924 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1925 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1926 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1927 				continue;
1928 			if (sa_equal(addr, ifa->ifa_addr)) {
1929 				goto done;
1930 			}
1931 			/* IP6 doesn't have broadcast */
1932 			if ((ifp->if_flags & IFF_BROADCAST) &&
1933 			    ifa->ifa_broadaddr &&
1934 			    ifa->ifa_broadaddr->sa_len != 0 &&
1935 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1936 				goto done;
1937 			}
1938 		}
1939 	}
1940 	ifa = NULL;
1941 done:
1942 	return (ifa);
1943 }
1944 
1945 int
1946 ifa_ifwithaddr_check(const struct sockaddr *addr)
1947 {
1948 	struct epoch_tracker et;
1949 	int rc;
1950 
1951 	NET_EPOCH_ENTER(et);
1952 	rc = (ifa_ifwithaddr(addr) != NULL);
1953 	NET_EPOCH_EXIT(et);
1954 	return (rc);
1955 }
1956 
1957 /*
1958  * Locate an interface based on the broadcast address.
1959  */
1960 /* ARGSUSED */
1961 struct ifaddr *
1962 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1963 {
1964 	struct ifnet *ifp;
1965 	struct ifaddr *ifa;
1966 
1967 	MPASS(in_epoch(net_epoch_preempt));
1968 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1969 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1970 			continue;
1971 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1972 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1973 				continue;
1974 			if ((ifp->if_flags & IFF_BROADCAST) &&
1975 			    ifa->ifa_broadaddr &&
1976 			    ifa->ifa_broadaddr->sa_len != 0 &&
1977 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1978 				goto done;
1979 			}
1980 		}
1981 	}
1982 	ifa = NULL;
1983 done:
1984 	return (ifa);
1985 }
1986 
1987 /*
1988  * Locate the point to point interface with a given destination address.
1989  */
1990 /*ARGSUSED*/
1991 struct ifaddr *
1992 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1993 {
1994 	struct ifnet *ifp;
1995 	struct ifaddr *ifa;
1996 
1997 	MPASS(in_epoch(net_epoch_preempt));
1998 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1999 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2000 			continue;
2001 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2002 			continue;
2003 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2004 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2005 				continue;
2006 			if (ifa->ifa_dstaddr != NULL &&
2007 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2008 				goto done;
2009 			}
2010 		}
2011 	}
2012 	ifa = NULL;
2013 done:
2014 	return (ifa);
2015 }
2016 
2017 /*
2018  * Find an interface on a specific network.  If many, choice
2019  * is most specific found.
2020  */
2021 struct ifaddr *
2022 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2023 {
2024 	struct ifnet *ifp;
2025 	struct ifaddr *ifa;
2026 	struct ifaddr *ifa_maybe = NULL;
2027 	u_int af = addr->sa_family;
2028 	const char *addr_data = addr->sa_data, *cplim;
2029 
2030 	MPASS(in_epoch(net_epoch_preempt));
2031 	/*
2032 	 * AF_LINK addresses can be looked up directly by their index number,
2033 	 * so do that if we can.
2034 	 */
2035 	if (af == AF_LINK) {
2036 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2037 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2038 		return (ifaddr_byindex(sdl->sdl_index));
2039 	}
2040 
2041 	/*
2042 	 * Scan though each interface, looking for ones that have addresses
2043 	 * in this address family and the requested fib.
2044 	 */
2045 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2046 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2047 			continue;
2048 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2049 			const char *cp, *cp2, *cp3;
2050 
2051 			if (ifa->ifa_addr->sa_family != af)
2052 next:				continue;
2053 			if (af == AF_INET &&
2054 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2055 				/*
2056 				 * This is a bit broken as it doesn't
2057 				 * take into account that the remote end may
2058 				 * be a single node in the network we are
2059 				 * looking for.
2060 				 * The trouble is that we don't know the
2061 				 * netmask for the remote end.
2062 				 */
2063 				if (ifa->ifa_dstaddr != NULL &&
2064 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2065 					goto done;
2066 				}
2067 			} else {
2068 				/*
2069 				 * Scan all the bits in the ifa's address.
2070 				 * If a bit dissagrees with what we are
2071 				 * looking for, mask it with the netmask
2072 				 * to see if it really matters.
2073 				 * (A byte at a time)
2074 				 */
2075 				if (ifa->ifa_netmask == 0)
2076 					continue;
2077 				cp = addr_data;
2078 				cp2 = ifa->ifa_addr->sa_data;
2079 				cp3 = ifa->ifa_netmask->sa_data;
2080 				cplim = ifa->ifa_netmask->sa_len
2081 					+ (char *)ifa->ifa_netmask;
2082 				while (cp3 < cplim)
2083 					if ((*cp++ ^ *cp2++) & *cp3++)
2084 						goto next; /* next address! */
2085 				/*
2086 				 * If the netmask of what we just found
2087 				 * is more specific than what we had before
2088 				 * (if we had one), or if the virtual status
2089 				 * of new prefix is better than of the old one,
2090 				 * then remember the new one before continuing
2091 				 * to search for an even better one.
2092 				 */
2093 				if (ifa_maybe == NULL ||
2094 				    ifa_preferred(ifa_maybe, ifa) ||
2095 				    rn_refines((caddr_t)ifa->ifa_netmask,
2096 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2097 					ifa_maybe = ifa;
2098 				}
2099 			}
2100 		}
2101 	}
2102 	ifa = ifa_maybe;
2103 	ifa_maybe = NULL;
2104 done:
2105 	return (ifa);
2106 }
2107 
2108 /*
2109  * Find an interface address specific to an interface best matching
2110  * a given address.
2111  */
2112 struct ifaddr *
2113 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2114 {
2115 	struct ifaddr *ifa;
2116 	const char *cp, *cp2, *cp3;
2117 	char *cplim;
2118 	struct ifaddr *ifa_maybe = NULL;
2119 	u_int af = addr->sa_family;
2120 
2121 	if (af >= AF_MAX)
2122 		return (NULL);
2123 
2124 	MPASS(in_epoch(net_epoch_preempt));
2125 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2126 		if (ifa->ifa_addr->sa_family != af)
2127 			continue;
2128 		if (ifa_maybe == NULL)
2129 			ifa_maybe = ifa;
2130 		if (ifa->ifa_netmask == 0) {
2131 			if (sa_equal(addr, ifa->ifa_addr) ||
2132 			    (ifa->ifa_dstaddr &&
2133 			    sa_equal(addr, ifa->ifa_dstaddr)))
2134 				goto done;
2135 			continue;
2136 		}
2137 		if (ifp->if_flags & IFF_POINTOPOINT) {
2138 			if (sa_equal(addr, ifa->ifa_dstaddr))
2139 				goto done;
2140 		} else {
2141 			cp = addr->sa_data;
2142 			cp2 = ifa->ifa_addr->sa_data;
2143 			cp3 = ifa->ifa_netmask->sa_data;
2144 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2145 			for (; cp3 < cplim; cp3++)
2146 				if ((*cp++ ^ *cp2++) & *cp3)
2147 					break;
2148 			if (cp3 == cplim)
2149 				goto done;
2150 		}
2151 	}
2152 	ifa = ifa_maybe;
2153 done:
2154 	return (ifa);
2155 }
2156 
2157 /*
2158  * See whether new ifa is better than current one:
2159  * 1) A non-virtual one is preferred over virtual.
2160  * 2) A virtual in master state preferred over any other state.
2161  *
2162  * Used in several address selecting functions.
2163  */
2164 int
2165 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2166 {
2167 
2168 	return (cur->ifa_carp && (!next->ifa_carp ||
2169 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2170 }
2171 
2172 struct sockaddr_dl *
2173 link_alloc_sdl(size_t size, int flags)
2174 {
2175 
2176 	return (malloc(size, M_TEMP, flags));
2177 }
2178 
2179 void
2180 link_free_sdl(struct sockaddr *sa)
2181 {
2182 	free(sa, M_TEMP);
2183 }
2184 
2185 /*
2186  * Fills in given sdl with interface basic info.
2187  * Returns pointer to filled sdl.
2188  */
2189 struct sockaddr_dl *
2190 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2191 {
2192 	struct sockaddr_dl *sdl;
2193 
2194 	sdl = (struct sockaddr_dl *)paddr;
2195 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2196 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2197 	sdl->sdl_family = AF_LINK;
2198 	sdl->sdl_index = ifp->if_index;
2199 	sdl->sdl_type = iftype;
2200 
2201 	return (sdl);
2202 }
2203 
2204 /*
2205  * Mark an interface down and notify protocols of
2206  * the transition.
2207  */
2208 static void
2209 if_unroute(struct ifnet *ifp, int flag, int fam)
2210 {
2211 	struct ifaddr *ifa;
2212 
2213 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2214 
2215 	ifp->if_flags &= ~flag;
2216 	getmicrotime(&ifp->if_lastchange);
2217 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2218 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2219 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2220 	ifp->if_qflush(ifp);
2221 
2222 	if (ifp->if_carp)
2223 		(*carp_linkstate_p)(ifp);
2224 	rt_ifmsg(ifp);
2225 }
2226 
2227 /*
2228  * Mark an interface up and notify protocols of
2229  * the transition.
2230  */
2231 static void
2232 if_route(struct ifnet *ifp, int flag, int fam)
2233 {
2234 	struct ifaddr *ifa;
2235 
2236 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2237 
2238 	ifp->if_flags |= flag;
2239 	getmicrotime(&ifp->if_lastchange);
2240 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2241 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2242 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2243 	if (ifp->if_carp)
2244 		(*carp_linkstate_p)(ifp);
2245 	rt_ifmsg(ifp);
2246 #ifdef INET6
2247 	in6_if_up(ifp);
2248 #endif
2249 }
2250 
2251 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2252 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2253 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2254 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2255 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2256 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2257 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2258 void	*(*vlan_cookie_p)(struct ifnet *);
2259 
2260 /*
2261  * Handle a change in the interface link state. To avoid LORs
2262  * between driver lock and upper layer locks, as well as possible
2263  * recursions, we post event to taskqueue, and all job
2264  * is done in static do_link_state_change().
2265  */
2266 void
2267 if_link_state_change(struct ifnet *ifp, int link_state)
2268 {
2269 	/* Return if state hasn't changed. */
2270 	if (ifp->if_link_state == link_state)
2271 		return;
2272 
2273 	ifp->if_link_state = link_state;
2274 
2275 	/* XXXGL: reference ifp? */
2276 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2277 }
2278 
2279 static void
2280 do_link_state_change(void *arg, int pending)
2281 {
2282 	struct ifnet *ifp;
2283 	int link_state;
2284 
2285 	ifp = arg;
2286 	link_state = ifp->if_link_state;
2287 
2288 	CURVNET_SET(ifp->if_vnet);
2289 	rt_ifmsg(ifp);
2290 	if (ifp->if_vlantrunk != NULL)
2291 		(*vlan_link_state_p)(ifp);
2292 
2293 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2294 	    ifp->if_l2com != NULL)
2295 		(*ng_ether_link_state_p)(ifp, link_state);
2296 	if (ifp->if_carp)
2297 		(*carp_linkstate_p)(ifp);
2298 	if (ifp->if_bridge)
2299 		ifp->if_bridge_linkstate(ifp);
2300 	if (ifp->if_lagg)
2301 		(*lagg_linkstate_p)(ifp, link_state);
2302 
2303 	if (IS_DEFAULT_VNET(curvnet))
2304 		devctl_notify("IFNET", ifp->if_xname,
2305 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2306 		    NULL);
2307 	if (pending > 1)
2308 		if_printf(ifp, "%d link states coalesced\n", pending);
2309 	if (log_link_state_change)
2310 		if_printf(ifp, "link state changed to %s\n",
2311 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2312 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2313 	CURVNET_RESTORE();
2314 }
2315 
2316 /*
2317  * Mark an interface down and notify protocols of
2318  * the transition.
2319  */
2320 void
2321 if_down(struct ifnet *ifp)
2322 {
2323 
2324 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2325 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2326 }
2327 
2328 /*
2329  * Mark an interface up and notify protocols of
2330  * the transition.
2331  */
2332 void
2333 if_up(struct ifnet *ifp)
2334 {
2335 
2336 	if_route(ifp, IFF_UP, AF_UNSPEC);
2337 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2338 }
2339 
2340 /*
2341  * Flush an interface queue.
2342  */
2343 void
2344 if_qflush(struct ifnet *ifp)
2345 {
2346 	struct mbuf *m, *n;
2347 	struct ifaltq *ifq;
2348 
2349 	ifq = &ifp->if_snd;
2350 	IFQ_LOCK(ifq);
2351 #ifdef ALTQ
2352 	if (ALTQ_IS_ENABLED(ifq))
2353 		ALTQ_PURGE(ifq);
2354 #endif
2355 	n = ifq->ifq_head;
2356 	while ((m = n) != NULL) {
2357 		n = m->m_nextpkt;
2358 		m_freem(m);
2359 	}
2360 	ifq->ifq_head = 0;
2361 	ifq->ifq_tail = 0;
2362 	ifq->ifq_len = 0;
2363 	IFQ_UNLOCK(ifq);
2364 }
2365 
2366 /*
2367  * Map interface name to interface structure pointer, with or without
2368  * returning a reference.
2369  */
2370 struct ifnet *
2371 ifunit_ref(const char *name)
2372 {
2373 	struct epoch_tracker et;
2374 	struct ifnet *ifp;
2375 
2376 	NET_EPOCH_ENTER(et);
2377 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2378 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2379 		    !(ifp->if_flags & IFF_DYING))
2380 			break;
2381 	}
2382 	if (ifp != NULL)
2383 		if_ref(ifp);
2384 	NET_EPOCH_EXIT(et);
2385 	return (ifp);
2386 }
2387 
2388 struct ifnet *
2389 ifunit(const char *name)
2390 {
2391 	struct epoch_tracker et;
2392 	struct ifnet *ifp;
2393 
2394 	NET_EPOCH_ENTER(et);
2395 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2396 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2397 			break;
2398 	}
2399 	NET_EPOCH_EXIT(et);
2400 	return (ifp);
2401 }
2402 
2403 static void *
2404 ifr_buffer_get_buffer(void *data)
2405 {
2406 	union ifreq_union *ifrup;
2407 
2408 	ifrup = data;
2409 #ifdef COMPAT_FREEBSD32
2410 	if (SV_CURPROC_FLAG(SV_ILP32))
2411 		return ((void *)(uintptr_t)
2412 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2413 #endif
2414 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2415 }
2416 
2417 static void
2418 ifr_buffer_set_buffer_null(void *data)
2419 {
2420 	union ifreq_union *ifrup;
2421 
2422 	ifrup = data;
2423 #ifdef COMPAT_FREEBSD32
2424 	if (SV_CURPROC_FLAG(SV_ILP32))
2425 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2426 	else
2427 #endif
2428 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2429 }
2430 
2431 static size_t
2432 ifr_buffer_get_length(void *data)
2433 {
2434 	union ifreq_union *ifrup;
2435 
2436 	ifrup = data;
2437 #ifdef COMPAT_FREEBSD32
2438 	if (SV_CURPROC_FLAG(SV_ILP32))
2439 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2440 #endif
2441 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2442 }
2443 
2444 static void
2445 ifr_buffer_set_length(void *data, size_t len)
2446 {
2447 	union ifreq_union *ifrup;
2448 
2449 	ifrup = data;
2450 #ifdef COMPAT_FREEBSD32
2451 	if (SV_CURPROC_FLAG(SV_ILP32))
2452 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2453 	else
2454 #endif
2455 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2456 }
2457 
2458 void *
2459 ifr_data_get_ptr(void *ifrp)
2460 {
2461 	union ifreq_union *ifrup;
2462 
2463 	ifrup = ifrp;
2464 #ifdef COMPAT_FREEBSD32
2465 	if (SV_CURPROC_FLAG(SV_ILP32))
2466 		return ((void *)(uintptr_t)
2467 		    ifrup->ifr32.ifr_ifru.ifru_data);
2468 #endif
2469 		return (ifrup->ifr.ifr_ifru.ifru_data);
2470 }
2471 
2472 /*
2473  * Hardware specific interface ioctls.
2474  */
2475 int
2476 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2477 {
2478 	struct ifreq *ifr;
2479 	int error = 0, do_ifup = 0;
2480 	int new_flags, temp_flags;
2481 	size_t namelen, onamelen;
2482 	size_t descrlen;
2483 	char *descrbuf, *odescrbuf;
2484 	char new_name[IFNAMSIZ];
2485 	struct ifaddr *ifa;
2486 	struct sockaddr_dl *sdl;
2487 
2488 	ifr = (struct ifreq *)data;
2489 	switch (cmd) {
2490 	case SIOCGIFINDEX:
2491 		ifr->ifr_index = ifp->if_index;
2492 		break;
2493 
2494 	case SIOCGIFFLAGS:
2495 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2496 		ifr->ifr_flags = temp_flags & 0xffff;
2497 		ifr->ifr_flagshigh = temp_flags >> 16;
2498 		break;
2499 
2500 	case SIOCGIFCAP:
2501 		ifr->ifr_reqcap = ifp->if_capabilities;
2502 		ifr->ifr_curcap = ifp->if_capenable;
2503 		break;
2504 
2505 #ifdef MAC
2506 	case SIOCGIFMAC:
2507 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2508 		break;
2509 #endif
2510 
2511 	case SIOCGIFMETRIC:
2512 		ifr->ifr_metric = ifp->if_metric;
2513 		break;
2514 
2515 	case SIOCGIFMTU:
2516 		ifr->ifr_mtu = ifp->if_mtu;
2517 		break;
2518 
2519 	case SIOCGIFPHYS:
2520 		/* XXXGL: did this ever worked? */
2521 		ifr->ifr_phys = 0;
2522 		break;
2523 
2524 	case SIOCGIFDESCR:
2525 		error = 0;
2526 		sx_slock(&ifdescr_sx);
2527 		if (ifp->if_description == NULL)
2528 			error = ENOMSG;
2529 		else {
2530 			/* space for terminating nul */
2531 			descrlen = strlen(ifp->if_description) + 1;
2532 			if (ifr_buffer_get_length(ifr) < descrlen)
2533 				ifr_buffer_set_buffer_null(ifr);
2534 			else
2535 				error = copyout(ifp->if_description,
2536 				    ifr_buffer_get_buffer(ifr), descrlen);
2537 			ifr_buffer_set_length(ifr, descrlen);
2538 		}
2539 		sx_sunlock(&ifdescr_sx);
2540 		break;
2541 
2542 	case SIOCSIFDESCR:
2543 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2544 		if (error)
2545 			return (error);
2546 
2547 		/*
2548 		 * Copy only (length-1) bytes to make sure that
2549 		 * if_description is always nul terminated.  The
2550 		 * length parameter is supposed to count the
2551 		 * terminating nul in.
2552 		 */
2553 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2554 			return (ENAMETOOLONG);
2555 		else if (ifr_buffer_get_length(ifr) == 0)
2556 			descrbuf = NULL;
2557 		else {
2558 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2559 			    M_IFDESCR, M_WAITOK | M_ZERO);
2560 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2561 			    ifr_buffer_get_length(ifr) - 1);
2562 			if (error) {
2563 				free(descrbuf, M_IFDESCR);
2564 				break;
2565 			}
2566 		}
2567 
2568 		sx_xlock(&ifdescr_sx);
2569 		odescrbuf = ifp->if_description;
2570 		ifp->if_description = descrbuf;
2571 		sx_xunlock(&ifdescr_sx);
2572 
2573 		getmicrotime(&ifp->if_lastchange);
2574 		free(odescrbuf, M_IFDESCR);
2575 		break;
2576 
2577 	case SIOCGIFFIB:
2578 		ifr->ifr_fib = ifp->if_fib;
2579 		break;
2580 
2581 	case SIOCSIFFIB:
2582 		error = priv_check(td, PRIV_NET_SETIFFIB);
2583 		if (error)
2584 			return (error);
2585 		if (ifr->ifr_fib >= rt_numfibs)
2586 			return (EINVAL);
2587 
2588 		ifp->if_fib = ifr->ifr_fib;
2589 		break;
2590 
2591 	case SIOCSIFFLAGS:
2592 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2593 		if (error)
2594 			return (error);
2595 		/*
2596 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2597 		 * check, so we don't need special handling here yet.
2598 		 */
2599 		new_flags = (ifr->ifr_flags & 0xffff) |
2600 		    (ifr->ifr_flagshigh << 16);
2601 		if (ifp->if_flags & IFF_UP &&
2602 		    (new_flags & IFF_UP) == 0) {
2603 			if_down(ifp);
2604 		} else if (new_flags & IFF_UP &&
2605 		    (ifp->if_flags & IFF_UP) == 0) {
2606 			do_ifup = 1;
2607 		}
2608 		/* See if permanently promiscuous mode bit is about to flip */
2609 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2610 			if (new_flags & IFF_PPROMISC)
2611 				ifp->if_flags |= IFF_PROMISC;
2612 			else if (ifp->if_pcount == 0)
2613 				ifp->if_flags &= ~IFF_PROMISC;
2614 			if (log_promisc_mode_change)
2615                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2616                                     ((new_flags & IFF_PPROMISC) ?
2617                                      "enabled" : "disabled"));
2618 		}
2619 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2620 			(new_flags &~ IFF_CANTCHANGE);
2621 		if (ifp->if_ioctl) {
2622 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2623 		}
2624 		if (do_ifup)
2625 			if_up(ifp);
2626 		getmicrotime(&ifp->if_lastchange);
2627 		break;
2628 
2629 	case SIOCSIFCAP:
2630 		error = priv_check(td, PRIV_NET_SETIFCAP);
2631 		if (error)
2632 			return (error);
2633 		if (ifp->if_ioctl == NULL)
2634 			return (EOPNOTSUPP);
2635 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2636 			return (EINVAL);
2637 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2638 		if (error == 0)
2639 			getmicrotime(&ifp->if_lastchange);
2640 		break;
2641 
2642 #ifdef MAC
2643 	case SIOCSIFMAC:
2644 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2645 		break;
2646 #endif
2647 
2648 	case SIOCSIFNAME:
2649 		error = priv_check(td, PRIV_NET_SETIFNAME);
2650 		if (error)
2651 			return (error);
2652 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2653 		    NULL);
2654 		if (error != 0)
2655 			return (error);
2656 		if (new_name[0] == '\0')
2657 			return (EINVAL);
2658 		if (new_name[IFNAMSIZ-1] != '\0') {
2659 			new_name[IFNAMSIZ-1] = '\0';
2660 			if (strlen(new_name) == IFNAMSIZ-1)
2661 				return (EINVAL);
2662 		}
2663 		if (strcmp(new_name, ifp->if_xname) == 0)
2664 			break;
2665 		if (ifunit(new_name) != NULL)
2666 			return (EEXIST);
2667 
2668 		/*
2669 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2670 		 * and there are numerous other races with the
2671 		 * ifunit() checks not being atomic with namespace
2672 		 * changes (renames, vmoves, if_attach, etc).
2673 		 */
2674 		ifp->if_flags |= IFF_RENAMING;
2675 
2676 		/* Announce the departure of the interface. */
2677 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2678 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2679 
2680 		if_printf(ifp, "changing name to '%s'\n", new_name);
2681 
2682 		IF_ADDR_WLOCK(ifp);
2683 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2684 		ifa = ifp->if_addr;
2685 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2686 		namelen = strlen(new_name);
2687 		onamelen = sdl->sdl_nlen;
2688 		/*
2689 		 * Move the address if needed.  This is safe because we
2690 		 * allocate space for a name of length IFNAMSIZ when we
2691 		 * create this in if_attach().
2692 		 */
2693 		if (namelen != onamelen) {
2694 			bcopy(sdl->sdl_data + onamelen,
2695 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2696 		}
2697 		bcopy(new_name, sdl->sdl_data, namelen);
2698 		sdl->sdl_nlen = namelen;
2699 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2700 		bzero(sdl->sdl_data, onamelen);
2701 		while (namelen != 0)
2702 			sdl->sdl_data[--namelen] = 0xff;
2703 		IF_ADDR_WUNLOCK(ifp);
2704 
2705 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2706 		/* Announce the return of the interface. */
2707 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2708 
2709 		ifp->if_flags &= ~IFF_RENAMING;
2710 		break;
2711 
2712 #ifdef VIMAGE
2713 	case SIOCSIFVNET:
2714 		error = priv_check(td, PRIV_NET_SETIFVNET);
2715 		if (error)
2716 			return (error);
2717 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2718 		break;
2719 #endif
2720 
2721 	case SIOCSIFMETRIC:
2722 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2723 		if (error)
2724 			return (error);
2725 		ifp->if_metric = ifr->ifr_metric;
2726 		getmicrotime(&ifp->if_lastchange);
2727 		break;
2728 
2729 	case SIOCSIFPHYS:
2730 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2731 		if (error)
2732 			return (error);
2733 		if (ifp->if_ioctl == NULL)
2734 			return (EOPNOTSUPP);
2735 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2736 		if (error == 0)
2737 			getmicrotime(&ifp->if_lastchange);
2738 		break;
2739 
2740 	case SIOCSIFMTU:
2741 	{
2742 		u_long oldmtu = ifp->if_mtu;
2743 
2744 		error = priv_check(td, PRIV_NET_SETIFMTU);
2745 		if (error)
2746 			return (error);
2747 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2748 			return (EINVAL);
2749 		if (ifp->if_ioctl == NULL)
2750 			return (EOPNOTSUPP);
2751 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2752 		if (error == 0) {
2753 			getmicrotime(&ifp->if_lastchange);
2754 			rt_ifmsg(ifp);
2755 #ifdef INET
2756 			DEBUGNET_NOTIFY_MTU(ifp);
2757 #endif
2758 		}
2759 		/*
2760 		 * If the link MTU changed, do network layer specific procedure.
2761 		 */
2762 		if (ifp->if_mtu != oldmtu) {
2763 #ifdef INET6
2764 			nd6_setmtu(ifp);
2765 #endif
2766 			rt_updatemtu(ifp);
2767 		}
2768 		break;
2769 	}
2770 
2771 	case SIOCADDMULTI:
2772 	case SIOCDELMULTI:
2773 		if (cmd == SIOCADDMULTI)
2774 			error = priv_check(td, PRIV_NET_ADDMULTI);
2775 		else
2776 			error = priv_check(td, PRIV_NET_DELMULTI);
2777 		if (error)
2778 			return (error);
2779 
2780 		/* Don't allow group membership on non-multicast interfaces. */
2781 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2782 			return (EOPNOTSUPP);
2783 
2784 		/* Don't let users screw up protocols' entries. */
2785 		if (ifr->ifr_addr.sa_family != AF_LINK)
2786 			return (EINVAL);
2787 
2788 		if (cmd == SIOCADDMULTI) {
2789 			struct epoch_tracker et;
2790 			struct ifmultiaddr *ifma;
2791 
2792 			/*
2793 			 * Userland is only permitted to join groups once
2794 			 * via the if_addmulti() KPI, because it cannot hold
2795 			 * struct ifmultiaddr * between calls. It may also
2796 			 * lose a race while we check if the membership
2797 			 * already exists.
2798 			 */
2799 			NET_EPOCH_ENTER(et);
2800 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2801 			NET_EPOCH_EXIT(et);
2802 			if (ifma != NULL)
2803 				error = EADDRINUSE;
2804 			else
2805 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2806 		} else {
2807 			error = if_delmulti(ifp, &ifr->ifr_addr);
2808 		}
2809 		if (error == 0)
2810 			getmicrotime(&ifp->if_lastchange);
2811 		break;
2812 
2813 	case SIOCSIFPHYADDR:
2814 	case SIOCDIFPHYADDR:
2815 #ifdef INET6
2816 	case SIOCSIFPHYADDR_IN6:
2817 #endif
2818 	case SIOCSIFMEDIA:
2819 	case SIOCSIFGENERIC:
2820 		error = priv_check(td, PRIV_NET_HWIOCTL);
2821 		if (error)
2822 			return (error);
2823 		if (ifp->if_ioctl == NULL)
2824 			return (EOPNOTSUPP);
2825 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2826 		if (error == 0)
2827 			getmicrotime(&ifp->if_lastchange);
2828 		break;
2829 
2830 	case SIOCGIFSTATUS:
2831 	case SIOCGIFPSRCADDR:
2832 	case SIOCGIFPDSTADDR:
2833 	case SIOCGIFMEDIA:
2834 	case SIOCGIFXMEDIA:
2835 	case SIOCGIFGENERIC:
2836 	case SIOCGIFRSSKEY:
2837 	case SIOCGIFRSSHASH:
2838 	case SIOCGIFDOWNREASON:
2839 		if (ifp->if_ioctl == NULL)
2840 			return (EOPNOTSUPP);
2841 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2842 		break;
2843 
2844 	case SIOCSIFLLADDR:
2845 		error = priv_check(td, PRIV_NET_SETLLADDR);
2846 		if (error)
2847 			return (error);
2848 		error = if_setlladdr(ifp,
2849 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2850 		break;
2851 
2852 	case SIOCGHWADDR:
2853 		error = if_gethwaddr(ifp, ifr);
2854 		break;
2855 
2856 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2857 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2858 		if (error)
2859 			return (error);
2860 		if ((error = if_addgroup(ifp,
2861 		    ifgr_group_get((struct ifgroupreq *)data))))
2862 			return (error);
2863 		break;
2864 
2865 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2866 	{
2867 		struct epoch_tracker et;
2868 
2869 		NET_EPOCH_ENTER(et);
2870 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2871 		NET_EPOCH_EXIT(et);
2872 		break;
2873 	}
2874 
2875 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2876 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2877 		if (error)
2878 			return (error);
2879 		if ((error = if_delgroup(ifp,
2880 		    ifgr_group_get((struct ifgroupreq *)data))))
2881 			return (error);
2882 		break;
2883 
2884 	default:
2885 		error = ENOIOCTL;
2886 		break;
2887 	}
2888 	return (error);
2889 }
2890 
2891 #ifdef COMPAT_FREEBSD32
2892 struct ifconf32 {
2893 	int32_t	ifc_len;
2894 	union {
2895 		uint32_t	ifcu_buf;
2896 		uint32_t	ifcu_req;
2897 	} ifc_ifcu;
2898 };
2899 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2900 #endif
2901 
2902 #ifdef COMPAT_FREEBSD32
2903 static void
2904 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2905 {
2906 	struct ifmediareq32 *ifmr32;
2907 
2908 	ifmr32 = (struct ifmediareq32 *)data;
2909 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2910 	    sizeof(ifmr->ifm_name));
2911 	ifmr->ifm_current = ifmr32->ifm_current;
2912 	ifmr->ifm_mask = ifmr32->ifm_mask;
2913 	ifmr->ifm_status = ifmr32->ifm_status;
2914 	ifmr->ifm_active = ifmr32->ifm_active;
2915 	ifmr->ifm_count = ifmr32->ifm_count;
2916 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2917 }
2918 
2919 static void
2920 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2921 {
2922 	struct ifmediareq32 *ifmr32;
2923 
2924 	ifmr32 = (struct ifmediareq32 *)data;
2925 	ifmr32->ifm_current = ifmr->ifm_current;
2926 	ifmr32->ifm_mask = ifmr->ifm_mask;
2927 	ifmr32->ifm_status = ifmr->ifm_status;
2928 	ifmr32->ifm_active = ifmr->ifm_active;
2929 	ifmr32->ifm_count = ifmr->ifm_count;
2930 }
2931 #endif
2932 
2933 /*
2934  * Interface ioctls.
2935  */
2936 int
2937 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2938 {
2939 #ifdef COMPAT_FREEBSD32
2940 	caddr_t saved_data = NULL;
2941 	struct ifmediareq ifmr;
2942 	struct ifmediareq *ifmrp = NULL;
2943 #endif
2944 	struct ifnet *ifp;
2945 	struct ifreq *ifr;
2946 	int error;
2947 	int oif_flags;
2948 
2949 	CURVNET_SET(so->so_vnet);
2950 #ifdef VIMAGE
2951 	/* Make sure the VNET is stable. */
2952 	if (so->so_vnet->vnet_shutdown) {
2953 		CURVNET_RESTORE();
2954 		return (EBUSY);
2955 	}
2956 #endif
2957 
2958 	switch (cmd) {
2959 	case SIOCGIFCONF:
2960 		error = ifconf(cmd, data);
2961 		goto out_noref;
2962 
2963 #ifdef COMPAT_FREEBSD32
2964 	case SIOCGIFCONF32:
2965 		{
2966 			struct ifconf32 *ifc32;
2967 			struct ifconf ifc;
2968 
2969 			ifc32 = (struct ifconf32 *)data;
2970 			ifc.ifc_len = ifc32->ifc_len;
2971 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2972 
2973 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2974 			if (error == 0)
2975 				ifc32->ifc_len = ifc.ifc_len;
2976 			goto out_noref;
2977 		}
2978 #endif
2979 	}
2980 
2981 #ifdef COMPAT_FREEBSD32
2982 	switch (cmd) {
2983 	case SIOCGIFMEDIA32:
2984 	case SIOCGIFXMEDIA32:
2985 		ifmrp = &ifmr;
2986 		ifmr_init(ifmrp, data);
2987 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2988 		saved_data = data;
2989 		data = (caddr_t)ifmrp;
2990 	}
2991 #endif
2992 
2993 	ifr = (struct ifreq *)data;
2994 	switch (cmd) {
2995 #ifdef VIMAGE
2996 	case SIOCSIFRVNET:
2997 		error = priv_check(td, PRIV_NET_SETIFVNET);
2998 		if (error == 0)
2999 			error = if_vmove_reclaim(td, ifr->ifr_name,
3000 			    ifr->ifr_jid);
3001 		goto out_noref;
3002 #endif
3003 	case SIOCIFCREATE:
3004 	case SIOCIFCREATE2:
3005 		error = priv_check(td, PRIV_NET_IFCREATE);
3006 		if (error == 0)
3007 			error = if_clone_create(ifr->ifr_name,
3008 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3009 			    ifr_data_get_ptr(ifr) : NULL);
3010 		goto out_noref;
3011 	case SIOCIFDESTROY:
3012 		error = priv_check(td, PRIV_NET_IFDESTROY);
3013 		if (error == 0)
3014 			error = if_clone_destroy(ifr->ifr_name);
3015 		goto out_noref;
3016 
3017 	case SIOCIFGCLONERS:
3018 		error = if_clone_list((struct if_clonereq *)data);
3019 		goto out_noref;
3020 
3021 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3022 		error = if_getgroupmembers((struct ifgroupreq *)data);
3023 		goto out_noref;
3024 
3025 #if defined(INET) || defined(INET6)
3026 	case SIOCSVH:
3027 	case SIOCGVH:
3028 		if (carp_ioctl_p == NULL)
3029 			error = EPROTONOSUPPORT;
3030 		else
3031 			error = (*carp_ioctl_p)(ifr, cmd, td);
3032 		goto out_noref;
3033 #endif
3034 	}
3035 
3036 	ifp = ifunit_ref(ifr->ifr_name);
3037 	if (ifp == NULL) {
3038 		error = ENXIO;
3039 		goto out_noref;
3040 	}
3041 
3042 	error = ifhwioctl(cmd, ifp, data, td);
3043 	if (error != ENOIOCTL)
3044 		goto out_ref;
3045 
3046 	oif_flags = ifp->if_flags;
3047 	if (so->so_proto == NULL) {
3048 		error = EOPNOTSUPP;
3049 		goto out_ref;
3050 	}
3051 
3052 	/*
3053 	 * Pass the request on to the socket control method, and if the
3054 	 * latter returns EOPNOTSUPP, directly to the interface.
3055 	 *
3056 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3057 	 * trust SIOCSIFADDR et al to come from an already privileged
3058 	 * layer, and do not perform any credentials checks or input
3059 	 * validation.
3060 	 */
3061 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3062 	    ifp, td));
3063 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3064 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3065 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3066 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3067 
3068 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3069 #ifdef INET6
3070 		if (ifp->if_flags & IFF_UP)
3071 			in6_if_up(ifp);
3072 #endif
3073 	}
3074 
3075 out_ref:
3076 	if_rele(ifp);
3077 out_noref:
3078 #ifdef COMPAT_FREEBSD32
3079 	if (ifmrp != NULL) {
3080 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3081 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3082 		     cmd));
3083 		data = saved_data;
3084 		ifmr_update(ifmrp, data);
3085 	}
3086 #endif
3087 	CURVNET_RESTORE();
3088 	return (error);
3089 }
3090 
3091 /*
3092  * The code common to handling reference counted flags,
3093  * e.g., in ifpromisc() and if_allmulti().
3094  * The "pflag" argument can specify a permanent mode flag to check,
3095  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3096  *
3097  * Only to be used on stack-owned flags, not driver-owned flags.
3098  */
3099 static int
3100 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3101 {
3102 	struct ifreq ifr;
3103 	int error;
3104 	int oldflags, oldcount;
3105 
3106 	/* Sanity checks to catch programming errors */
3107 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3108 	    ("%s: setting driver-owned flag %d", __func__, flag));
3109 
3110 	if (onswitch)
3111 		KASSERT(*refcount >= 0,
3112 		    ("%s: increment negative refcount %d for flag %d",
3113 		    __func__, *refcount, flag));
3114 	else
3115 		KASSERT(*refcount > 0,
3116 		    ("%s: decrement non-positive refcount %d for flag %d",
3117 		    __func__, *refcount, flag));
3118 
3119 	/* In case this mode is permanent, just touch refcount */
3120 	if (ifp->if_flags & pflag) {
3121 		*refcount += onswitch ? 1 : -1;
3122 		return (0);
3123 	}
3124 
3125 	/* Save ifnet parameters for if_ioctl() may fail */
3126 	oldcount = *refcount;
3127 	oldflags = ifp->if_flags;
3128 
3129 	/*
3130 	 * See if we aren't the only and touching refcount is enough.
3131 	 * Actually toggle interface flag if we are the first or last.
3132 	 */
3133 	if (onswitch) {
3134 		if ((*refcount)++)
3135 			return (0);
3136 		ifp->if_flags |= flag;
3137 	} else {
3138 		if (--(*refcount))
3139 			return (0);
3140 		ifp->if_flags &= ~flag;
3141 	}
3142 
3143 	/* Call down the driver since we've changed interface flags */
3144 	if (ifp->if_ioctl == NULL) {
3145 		error = EOPNOTSUPP;
3146 		goto recover;
3147 	}
3148 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3149 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3150 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3151 	if (error)
3152 		goto recover;
3153 	/* Notify userland that interface flags have changed */
3154 	rt_ifmsg(ifp);
3155 	return (0);
3156 
3157 recover:
3158 	/* Recover after driver error */
3159 	*refcount = oldcount;
3160 	ifp->if_flags = oldflags;
3161 	return (error);
3162 }
3163 
3164 /*
3165  * Set/clear promiscuous mode on interface ifp based on the truth value
3166  * of pswitch.  The calls are reference counted so that only the first
3167  * "on" request actually has an effect, as does the final "off" request.
3168  * Results are undefined if the "off" and "on" requests are not matched.
3169  */
3170 int
3171 ifpromisc(struct ifnet *ifp, int pswitch)
3172 {
3173 	int error;
3174 	int oldflags = ifp->if_flags;
3175 
3176 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3177 			   &ifp->if_pcount, pswitch);
3178 	/* If promiscuous mode status has changed, log a message */
3179 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3180             log_promisc_mode_change)
3181 		if_printf(ifp, "promiscuous mode %s\n",
3182 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3183 	return (error);
3184 }
3185 
3186 /*
3187  * Return interface configuration
3188  * of system.  List may be used
3189  * in later ioctl's (above) to get
3190  * other information.
3191  */
3192 /*ARGSUSED*/
3193 static int
3194 ifconf(u_long cmd, caddr_t data)
3195 {
3196 	struct ifconf *ifc = (struct ifconf *)data;
3197 	struct ifnet *ifp;
3198 	struct ifaddr *ifa;
3199 	struct ifreq ifr;
3200 	struct sbuf *sb;
3201 	int error, full = 0, valid_len, max_len;
3202 
3203 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3204 	max_len = MAXPHYS - 1;
3205 
3206 	/* Prevent hostile input from being able to crash the system */
3207 	if (ifc->ifc_len <= 0)
3208 		return (EINVAL);
3209 
3210 again:
3211 	if (ifc->ifc_len <= max_len) {
3212 		max_len = ifc->ifc_len;
3213 		full = 1;
3214 	}
3215 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3216 	max_len = 0;
3217 	valid_len = 0;
3218 
3219 	IFNET_RLOCK();
3220 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3221 		struct epoch_tracker et;
3222 		int addrs;
3223 
3224 		/*
3225 		 * Zero the ifr to make sure we don't disclose the contents
3226 		 * of the stack.
3227 		 */
3228 		memset(&ifr, 0, sizeof(ifr));
3229 
3230 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3231 		    >= sizeof(ifr.ifr_name)) {
3232 			sbuf_delete(sb);
3233 			IFNET_RUNLOCK();
3234 			return (ENAMETOOLONG);
3235 		}
3236 
3237 		addrs = 0;
3238 		NET_EPOCH_ENTER(et);
3239 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3240 			struct sockaddr *sa = ifa->ifa_addr;
3241 
3242 			if (prison_if(curthread->td_ucred, sa) != 0)
3243 				continue;
3244 			addrs++;
3245 			if (sa->sa_len <= sizeof(*sa)) {
3246 				if (sa->sa_len < sizeof(*sa)) {
3247 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3248 					    sizeof(ifr.ifr_ifru.ifru_addr));
3249 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3250 					    sa->sa_len);
3251 				} else
3252 					ifr.ifr_ifru.ifru_addr = *sa;
3253 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3254 				max_len += sizeof(ifr);
3255 			} else {
3256 				sbuf_bcat(sb, &ifr,
3257 				    offsetof(struct ifreq, ifr_addr));
3258 				max_len += offsetof(struct ifreq, ifr_addr);
3259 				sbuf_bcat(sb, sa, sa->sa_len);
3260 				max_len += sa->sa_len;
3261 			}
3262 
3263 			if (sbuf_error(sb) == 0)
3264 				valid_len = sbuf_len(sb);
3265 		}
3266 		NET_EPOCH_EXIT(et);
3267 		if (addrs == 0) {
3268 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3269 			max_len += sizeof(ifr);
3270 
3271 			if (sbuf_error(sb) == 0)
3272 				valid_len = sbuf_len(sb);
3273 		}
3274 	}
3275 	IFNET_RUNLOCK();
3276 
3277 	/*
3278 	 * If we didn't allocate enough space (uncommon), try again.  If
3279 	 * we have already allocated as much space as we are allowed,
3280 	 * return what we've got.
3281 	 */
3282 	if (valid_len != max_len && !full) {
3283 		sbuf_delete(sb);
3284 		goto again;
3285 	}
3286 
3287 	ifc->ifc_len = valid_len;
3288 	sbuf_finish(sb);
3289 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3290 	sbuf_delete(sb);
3291 	return (error);
3292 }
3293 
3294 /*
3295  * Just like ifpromisc(), but for all-multicast-reception mode.
3296  */
3297 int
3298 if_allmulti(struct ifnet *ifp, int onswitch)
3299 {
3300 
3301 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3302 }
3303 
3304 struct ifmultiaddr *
3305 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3306 {
3307 	struct ifmultiaddr *ifma;
3308 
3309 	IF_ADDR_LOCK_ASSERT(ifp);
3310 
3311 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3312 		if (sa->sa_family == AF_LINK) {
3313 			if (sa_dl_equal(ifma->ifma_addr, sa))
3314 				break;
3315 		} else {
3316 			if (sa_equal(ifma->ifma_addr, sa))
3317 				break;
3318 		}
3319 	}
3320 
3321 	return ifma;
3322 }
3323 
3324 /*
3325  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3326  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3327  * the ifnet multicast address list here, so the caller must do that and
3328  * other setup work (such as notifying the device driver).  The reference
3329  * count is initialized to 1.
3330  */
3331 static struct ifmultiaddr *
3332 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3333     int mflags)
3334 {
3335 	struct ifmultiaddr *ifma;
3336 	struct sockaddr *dupsa;
3337 
3338 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3339 	    M_ZERO);
3340 	if (ifma == NULL)
3341 		return (NULL);
3342 
3343 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3344 	if (dupsa == NULL) {
3345 		free(ifma, M_IFMADDR);
3346 		return (NULL);
3347 	}
3348 	bcopy(sa, dupsa, sa->sa_len);
3349 	ifma->ifma_addr = dupsa;
3350 
3351 	ifma->ifma_ifp = ifp;
3352 	ifma->ifma_refcount = 1;
3353 	ifma->ifma_protospec = NULL;
3354 
3355 	if (llsa == NULL) {
3356 		ifma->ifma_lladdr = NULL;
3357 		return (ifma);
3358 	}
3359 
3360 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3361 	if (dupsa == NULL) {
3362 		free(ifma->ifma_addr, M_IFMADDR);
3363 		free(ifma, M_IFMADDR);
3364 		return (NULL);
3365 	}
3366 	bcopy(llsa, dupsa, llsa->sa_len);
3367 	ifma->ifma_lladdr = dupsa;
3368 
3369 	return (ifma);
3370 }
3371 
3372 /*
3373  * if_freemulti: free ifmultiaddr structure and possibly attached related
3374  * addresses.  The caller is responsible for implementing reference
3375  * counting, notifying the driver, handling routing messages, and releasing
3376  * any dependent link layer state.
3377  */
3378 #ifdef MCAST_VERBOSE
3379 extern void kdb_backtrace(void);
3380 #endif
3381 static void
3382 if_freemulti_internal(struct ifmultiaddr *ifma)
3383 {
3384 
3385 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3386 	    ifma->ifma_refcount));
3387 
3388 	if (ifma->ifma_lladdr != NULL)
3389 		free(ifma->ifma_lladdr, M_IFMADDR);
3390 #ifdef MCAST_VERBOSE
3391 	kdb_backtrace();
3392 	printf("%s freeing ifma: %p\n", __func__, ifma);
3393 #endif
3394 	free(ifma->ifma_addr, M_IFMADDR);
3395 	free(ifma, M_IFMADDR);
3396 }
3397 
3398 static void
3399 if_destroymulti(epoch_context_t ctx)
3400 {
3401 	struct ifmultiaddr *ifma;
3402 
3403 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3404 	if_freemulti_internal(ifma);
3405 }
3406 
3407 void
3408 if_freemulti(struct ifmultiaddr *ifma)
3409 {
3410 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3411 	    ifma->ifma_refcount));
3412 
3413 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3414 }
3415 
3416 
3417 /*
3418  * Register an additional multicast address with a network interface.
3419  *
3420  * - If the address is already present, bump the reference count on the
3421  *   address and return.
3422  * - If the address is not link-layer, look up a link layer address.
3423  * - Allocate address structures for one or both addresses, and attach to the
3424  *   multicast address list on the interface.  If automatically adding a link
3425  *   layer address, the protocol address will own a reference to the link
3426  *   layer address, to be freed when it is freed.
3427  * - Notify the network device driver of an addition to the multicast address
3428  *   list.
3429  *
3430  * 'sa' points to caller-owned memory with the desired multicast address.
3431  *
3432  * 'retifma' will be used to return a pointer to the resulting multicast
3433  * address reference, if desired.
3434  */
3435 int
3436 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3437     struct ifmultiaddr **retifma)
3438 {
3439 	struct ifmultiaddr *ifma, *ll_ifma;
3440 	struct sockaddr *llsa;
3441 	struct sockaddr_dl sdl;
3442 	int error;
3443 
3444 #ifdef INET
3445 	IN_MULTI_LIST_UNLOCK_ASSERT();
3446 #endif
3447 #ifdef INET6
3448 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3449 #endif
3450 	/*
3451 	 * If the address is already present, return a new reference to it;
3452 	 * otherwise, allocate storage and set up a new address.
3453 	 */
3454 	IF_ADDR_WLOCK(ifp);
3455 	ifma = if_findmulti(ifp, sa);
3456 	if (ifma != NULL) {
3457 		ifma->ifma_refcount++;
3458 		if (retifma != NULL)
3459 			*retifma = ifma;
3460 		IF_ADDR_WUNLOCK(ifp);
3461 		return (0);
3462 	}
3463 
3464 	/*
3465 	 * The address isn't already present; resolve the protocol address
3466 	 * into a link layer address, and then look that up, bump its
3467 	 * refcount or allocate an ifma for that also.
3468 	 * Most link layer resolving functions returns address data which
3469 	 * fits inside default sockaddr_dl structure. However callback
3470 	 * can allocate another sockaddr structure, in that case we need to
3471 	 * free it later.
3472 	 */
3473 	llsa = NULL;
3474 	ll_ifma = NULL;
3475 	if (ifp->if_resolvemulti != NULL) {
3476 		/* Provide called function with buffer size information */
3477 		sdl.sdl_len = sizeof(sdl);
3478 		llsa = (struct sockaddr *)&sdl;
3479 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3480 		if (error)
3481 			goto unlock_out;
3482 	}
3483 
3484 	/*
3485 	 * Allocate the new address.  Don't hook it up yet, as we may also
3486 	 * need to allocate a link layer multicast address.
3487 	 */
3488 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3489 	if (ifma == NULL) {
3490 		error = ENOMEM;
3491 		goto free_llsa_out;
3492 	}
3493 
3494 	/*
3495 	 * If a link layer address is found, we'll need to see if it's
3496 	 * already present in the address list, or allocate is as well.
3497 	 * When this block finishes, the link layer address will be on the
3498 	 * list.
3499 	 */
3500 	if (llsa != NULL) {
3501 		ll_ifma = if_findmulti(ifp, llsa);
3502 		if (ll_ifma == NULL) {
3503 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3504 			if (ll_ifma == NULL) {
3505 				--ifma->ifma_refcount;
3506 				if_freemulti(ifma);
3507 				error = ENOMEM;
3508 				goto free_llsa_out;
3509 			}
3510 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3511 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3512 			    ifma_link);
3513 		} else
3514 			ll_ifma->ifma_refcount++;
3515 		ifma->ifma_llifma = ll_ifma;
3516 	}
3517 
3518 	/*
3519 	 * We now have a new multicast address, ifma, and possibly a new or
3520 	 * referenced link layer address.  Add the primary address to the
3521 	 * ifnet address list.
3522 	 */
3523 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3524 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3525 
3526 	if (retifma != NULL)
3527 		*retifma = ifma;
3528 
3529 	/*
3530 	 * Must generate the message while holding the lock so that 'ifma'
3531 	 * pointer is still valid.
3532 	 */
3533 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3534 	IF_ADDR_WUNLOCK(ifp);
3535 
3536 	/*
3537 	 * We are certain we have added something, so call down to the
3538 	 * interface to let them know about it.
3539 	 */
3540 	if (ifp->if_ioctl != NULL) {
3541 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3542 	}
3543 
3544 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3545 		link_free_sdl(llsa);
3546 
3547 	return (0);
3548 
3549 free_llsa_out:
3550 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3551 		link_free_sdl(llsa);
3552 
3553 unlock_out:
3554 	IF_ADDR_WUNLOCK(ifp);
3555 	return (error);
3556 }
3557 
3558 /*
3559  * Delete a multicast group membership by network-layer group address.
3560  *
3561  * Returns ENOENT if the entry could not be found. If ifp no longer
3562  * exists, results are undefined. This entry point should only be used
3563  * from subsystems which do appropriate locking to hold ifp for the
3564  * duration of the call.
3565  * Network-layer protocol domains must use if_delmulti_ifma().
3566  */
3567 int
3568 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3569 {
3570 	struct ifmultiaddr *ifma;
3571 	int lastref;
3572 
3573 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3574 
3575 	IF_ADDR_WLOCK(ifp);
3576 	lastref = 0;
3577 	ifma = if_findmulti(ifp, sa);
3578 	if (ifma != NULL)
3579 		lastref = if_delmulti_locked(ifp, ifma, 0);
3580 	IF_ADDR_WUNLOCK(ifp);
3581 
3582 	if (ifma == NULL)
3583 		return (ENOENT);
3584 
3585 	if (lastref && ifp->if_ioctl != NULL) {
3586 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3587 	}
3588 
3589 	return (0);
3590 }
3591 
3592 /*
3593  * Delete all multicast group membership for an interface.
3594  * Should be used to quickly flush all multicast filters.
3595  */
3596 void
3597 if_delallmulti(struct ifnet *ifp)
3598 {
3599 	struct ifmultiaddr *ifma;
3600 	struct ifmultiaddr *next;
3601 
3602 	IF_ADDR_WLOCK(ifp);
3603 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3604 		if_delmulti_locked(ifp, ifma, 0);
3605 	IF_ADDR_WUNLOCK(ifp);
3606 }
3607 
3608 void
3609 if_delmulti_ifma(struct ifmultiaddr *ifma)
3610 {
3611 	if_delmulti_ifma_flags(ifma, 0);
3612 }
3613 
3614 /*
3615  * Delete a multicast group membership by group membership pointer.
3616  * Network-layer protocol domains must use this routine.
3617  *
3618  * It is safe to call this routine if the ifp disappeared.
3619  */
3620 void
3621 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3622 {
3623 	struct ifnet *ifp;
3624 	int lastref;
3625 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3626 #ifdef INET
3627 	IN_MULTI_LIST_UNLOCK_ASSERT();
3628 #endif
3629 	ifp = ifma->ifma_ifp;
3630 #ifdef DIAGNOSTIC
3631 	if (ifp == NULL) {
3632 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3633 	} else {
3634 		struct epoch_tracker et;
3635 		struct ifnet *oifp;
3636 
3637 		NET_EPOCH_ENTER(et);
3638 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3639 			if (ifp == oifp)
3640 				break;
3641 		NET_EPOCH_EXIT(et);
3642 		if (ifp != oifp)
3643 			ifp = NULL;
3644 	}
3645 #endif
3646 	/*
3647 	 * If and only if the ifnet instance exists: Acquire the address lock.
3648 	 */
3649 	if (ifp != NULL)
3650 		IF_ADDR_WLOCK(ifp);
3651 
3652 	lastref = if_delmulti_locked(ifp, ifma, flags);
3653 
3654 	if (ifp != NULL) {
3655 		/*
3656 		 * If and only if the ifnet instance exists:
3657 		 *  Release the address lock.
3658 		 *  If the group was left: update the hardware hash filter.
3659 		 */
3660 		IF_ADDR_WUNLOCK(ifp);
3661 		if (lastref && ifp->if_ioctl != NULL) {
3662 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3663 		}
3664 	}
3665 }
3666 
3667 /*
3668  * Perform deletion of network-layer and/or link-layer multicast address.
3669  *
3670  * Return 0 if the reference count was decremented.
3671  * Return 1 if the final reference was released, indicating that the
3672  * hardware hash filter should be reprogrammed.
3673  */
3674 static int
3675 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3676 {
3677 	struct ifmultiaddr *ll_ifma;
3678 
3679 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3680 		KASSERT(ifma->ifma_ifp == ifp,
3681 		    ("%s: inconsistent ifp %p", __func__, ifp));
3682 		IF_ADDR_WLOCK_ASSERT(ifp);
3683 	}
3684 
3685 	ifp = ifma->ifma_ifp;
3686 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3687 
3688 	/*
3689 	 * If the ifnet is detaching, null out references to ifnet,
3690 	 * so that upper protocol layers will notice, and not attempt
3691 	 * to obtain locks for an ifnet which no longer exists. The
3692 	 * routing socket announcement must happen before the ifnet
3693 	 * instance is detached from the system.
3694 	 */
3695 	if (detaching) {
3696 #ifdef DIAGNOSTIC
3697 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3698 #endif
3699 		/*
3700 		 * ifp may already be nulled out if we are being reentered
3701 		 * to delete the ll_ifma.
3702 		 */
3703 		if (ifp != NULL) {
3704 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3705 			ifma->ifma_ifp = NULL;
3706 		}
3707 	}
3708 
3709 	if (--ifma->ifma_refcount > 0)
3710 		return 0;
3711 
3712 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3713 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3714 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3715 	}
3716 	/*
3717 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3718 	 * have been associated with it. Release it first if so.
3719 	 */
3720 	ll_ifma = ifma->ifma_llifma;
3721 	if (ll_ifma != NULL) {
3722 		KASSERT(ifma->ifma_lladdr != NULL,
3723 		    ("%s: llifma w/o lladdr", __func__));
3724 		if (detaching)
3725 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3726 		if (--ll_ifma->ifma_refcount == 0) {
3727 			if (ifp != NULL) {
3728 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3729 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3730 						ifma_link);
3731 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3732 				}
3733 			}
3734 			if_freemulti(ll_ifma);
3735 		}
3736 	}
3737 #ifdef INVARIANTS
3738 	if (ifp) {
3739 		struct ifmultiaddr *ifmatmp;
3740 
3741 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3742 			MPASS(ifma != ifmatmp);
3743 	}
3744 #endif
3745 	if_freemulti(ifma);
3746 	/*
3747 	 * The last reference to this instance of struct ifmultiaddr
3748 	 * was released; the hardware should be notified of this change.
3749 	 */
3750 	return 1;
3751 }
3752 
3753 /*
3754  * Set the link layer address on an interface.
3755  *
3756  * At this time we only support certain types of interfaces,
3757  * and we don't allow the length of the address to change.
3758  *
3759  * Set noinline to be dtrace-friendly
3760  */
3761 __noinline int
3762 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3763 {
3764 	struct sockaddr_dl *sdl;
3765 	struct ifaddr *ifa;
3766 	struct ifreq ifr;
3767 
3768 	ifa = ifp->if_addr;
3769 	if (ifa == NULL)
3770 		return (EINVAL);
3771 
3772 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3773 	if (sdl == NULL)
3774 		return (EINVAL);
3775 
3776 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3777 		return (EINVAL);
3778 
3779 	switch (ifp->if_type) {
3780 	case IFT_ETHER:
3781 	case IFT_XETHER:
3782 	case IFT_L2VLAN:
3783 	case IFT_BRIDGE:
3784 	case IFT_IEEE8023ADLAG:
3785 		bcopy(lladdr, LLADDR(sdl), len);
3786 		break;
3787 	default:
3788 		return (ENODEV);
3789 	}
3790 
3791 	/*
3792 	 * If the interface is already up, we need
3793 	 * to re-init it in order to reprogram its
3794 	 * address filter.
3795 	 */
3796 	if ((ifp->if_flags & IFF_UP) != 0) {
3797 		if (ifp->if_ioctl) {
3798 			ifp->if_flags &= ~IFF_UP;
3799 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3800 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3801 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3802 			ifp->if_flags |= IFF_UP;
3803 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3804 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3805 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3806 		}
3807 	}
3808 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3809 
3810 	return (0);
3811 }
3812 
3813 /*
3814  * Compat function for handling basic encapsulation requests.
3815  * Not converted stacks (FDDI, IB, ..) supports traditional
3816  * output model: ARP (and other similar L2 protocols) are handled
3817  * inside output routine, arpresolve/nd6_resolve() returns MAC
3818  * address instead of full prepend.
3819  *
3820  * This function creates calculated header==MAC for IPv4/IPv6 and
3821  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3822  * address families.
3823  */
3824 static int
3825 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3826 {
3827 
3828 	if (req->rtype != IFENCAP_LL)
3829 		return (EOPNOTSUPP);
3830 
3831 	if (req->bufsize < req->lladdr_len)
3832 		return (ENOMEM);
3833 
3834 	switch (req->family) {
3835 	case AF_INET:
3836 	case AF_INET6:
3837 		break;
3838 	default:
3839 		return (EAFNOSUPPORT);
3840 	}
3841 
3842 	/* Copy lladdr to storage as is */
3843 	memmove(req->buf, req->lladdr, req->lladdr_len);
3844 	req->bufsize = req->lladdr_len;
3845 	req->lladdr_off = 0;
3846 
3847 	return (0);
3848 }
3849 
3850 /*
3851  * Tunnel interfaces can nest, also they may cause infinite recursion
3852  * calls when misconfigured. We'll prevent this by detecting loops.
3853  * High nesting level may cause stack exhaustion. We'll prevent this
3854  * by introducing upper limit.
3855  *
3856  * Return 0, if tunnel nesting count is equal or less than limit.
3857  */
3858 int
3859 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3860     int limit)
3861 {
3862 	struct m_tag *mtag;
3863 	int count;
3864 
3865 	count = 1;
3866 	mtag = NULL;
3867 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3868 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3869 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3870 			return (EIO);
3871 		}
3872 		count++;
3873 	}
3874 	if (count > limit) {
3875 		log(LOG_NOTICE,
3876 		    "%s: if_output recursively called too many times(%d)\n",
3877 		    if_name(ifp), count);
3878 		return (EIO);
3879 	}
3880 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3881 	if (mtag == NULL)
3882 		return (ENOMEM);
3883 	*(struct ifnet **)(mtag + 1) = ifp;
3884 	m_tag_prepend(m, mtag);
3885 	return (0);
3886 }
3887 
3888 /*
3889  * Get the link layer address that was read from the hardware at attach.
3890  *
3891  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3892  * their component interfaces as IFT_IEEE8023ADLAG.
3893  */
3894 int
3895 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3896 {
3897 
3898 	if (ifp->if_hw_addr == NULL)
3899 		return (ENODEV);
3900 
3901 	switch (ifp->if_type) {
3902 	case IFT_ETHER:
3903 	case IFT_IEEE8023ADLAG:
3904 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3905 		return (0);
3906 	default:
3907 		return (ENODEV);
3908 	}
3909 }
3910 
3911 /*
3912  * The name argument must be a pointer to storage which will last as
3913  * long as the interface does.  For physical devices, the result of
3914  * device_get_name(dev) is a good choice and for pseudo-devices a
3915  * static string works well.
3916  */
3917 void
3918 if_initname(struct ifnet *ifp, const char *name, int unit)
3919 {
3920 	ifp->if_dname = name;
3921 	ifp->if_dunit = unit;
3922 	if (unit != IF_DUNIT_NONE)
3923 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3924 	else
3925 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3926 }
3927 
3928 int
3929 if_printf(struct ifnet *ifp, const char *fmt, ...)
3930 {
3931 	char if_fmt[256];
3932 	va_list ap;
3933 
3934 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3935 	va_start(ap, fmt);
3936 	vlog(LOG_INFO, if_fmt, ap);
3937 	va_end(ap);
3938 	return (0);
3939 }
3940 
3941 void
3942 if_start(struct ifnet *ifp)
3943 {
3944 
3945 	(*(ifp)->if_start)(ifp);
3946 }
3947 
3948 /*
3949  * Backwards compatibility interface for drivers
3950  * that have not implemented it
3951  */
3952 static int
3953 if_transmit(struct ifnet *ifp, struct mbuf *m)
3954 {
3955 	int error;
3956 
3957 	IFQ_HANDOFF(ifp, m, error);
3958 	return (error);
3959 }
3960 
3961 static void
3962 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3963 {
3964 
3965 	m_freem(m);
3966 }
3967 
3968 int
3969 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3970 {
3971 	int active = 0;
3972 
3973 	IF_LOCK(ifq);
3974 	if (_IF_QFULL(ifq)) {
3975 		IF_UNLOCK(ifq);
3976 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3977 		m_freem(m);
3978 		return (0);
3979 	}
3980 	if (ifp != NULL) {
3981 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3982 		if (m->m_flags & (M_BCAST|M_MCAST))
3983 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3984 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3985 	}
3986 	_IF_ENQUEUE(ifq, m);
3987 	IF_UNLOCK(ifq);
3988 	if (ifp != NULL && !active)
3989 		(*(ifp)->if_start)(ifp);
3990 	return (1);
3991 }
3992 
3993 void
3994 if_register_com_alloc(u_char type,
3995     if_com_alloc_t *a, if_com_free_t *f)
3996 {
3997 
3998 	KASSERT(if_com_alloc[type] == NULL,
3999 	    ("if_register_com_alloc: %d already registered", type));
4000 	KASSERT(if_com_free[type] == NULL,
4001 	    ("if_register_com_alloc: %d free already registered", type));
4002 
4003 	if_com_alloc[type] = a;
4004 	if_com_free[type] = f;
4005 }
4006 
4007 void
4008 if_deregister_com_alloc(u_char type)
4009 {
4010 
4011 	KASSERT(if_com_alloc[type] != NULL,
4012 	    ("if_deregister_com_alloc: %d not registered", type));
4013 	KASSERT(if_com_free[type] != NULL,
4014 	    ("if_deregister_com_alloc: %d free not registered", type));
4015 	if_com_alloc[type] = NULL;
4016 	if_com_free[type] = NULL;
4017 }
4018 
4019 /* API for driver access to network stack owned ifnet.*/
4020 uint64_t
4021 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4022 {
4023 	uint64_t oldbrate;
4024 
4025 	oldbrate = ifp->if_baudrate;
4026 	ifp->if_baudrate = baudrate;
4027 	return (oldbrate);
4028 }
4029 
4030 uint64_t
4031 if_getbaudrate(if_t ifp)
4032 {
4033 
4034 	return (((struct ifnet *)ifp)->if_baudrate);
4035 }
4036 
4037 int
4038 if_setcapabilities(if_t ifp, int capabilities)
4039 {
4040 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4041 	return (0);
4042 }
4043 
4044 int
4045 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4046 {
4047 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4048 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4049 
4050 	return (0);
4051 }
4052 
4053 int
4054 if_getcapabilities(if_t ifp)
4055 {
4056 	return ((struct ifnet *)ifp)->if_capabilities;
4057 }
4058 
4059 int
4060 if_setcapenable(if_t ifp, int capabilities)
4061 {
4062 	((struct ifnet *)ifp)->if_capenable = capabilities;
4063 	return (0);
4064 }
4065 
4066 int
4067 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4068 {
4069 	if(setcap)
4070 		((struct ifnet *)ifp)->if_capenable |= setcap;
4071 	if(clearcap)
4072 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4073 
4074 	return (0);
4075 }
4076 
4077 const char *
4078 if_getdname(if_t ifp)
4079 {
4080 	return ((struct ifnet *)ifp)->if_dname;
4081 }
4082 
4083 int
4084 if_togglecapenable(if_t ifp, int togglecap)
4085 {
4086 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4087 	return (0);
4088 }
4089 
4090 int
4091 if_getcapenable(if_t ifp)
4092 {
4093 	return ((struct ifnet *)ifp)->if_capenable;
4094 }
4095 
4096 /*
4097  * This is largely undesirable because it ties ifnet to a device, but does
4098  * provide flexiblity for an embedded product vendor. Should be used with
4099  * the understanding that it violates the interface boundaries, and should be
4100  * a last resort only.
4101  */
4102 int
4103 if_setdev(if_t ifp, void *dev)
4104 {
4105 	return (0);
4106 }
4107 
4108 int
4109 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4110 {
4111 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4112 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4113 
4114 	return (0);
4115 }
4116 
4117 int
4118 if_getdrvflags(if_t ifp)
4119 {
4120 	return ((struct ifnet *)ifp)->if_drv_flags;
4121 }
4122 
4123 int
4124 if_setdrvflags(if_t ifp, int flags)
4125 {
4126 	((struct ifnet *)ifp)->if_drv_flags = flags;
4127 	return (0);
4128 }
4129 
4130 
4131 int
4132 if_setflags(if_t ifp, int flags)
4133 {
4134 	((struct ifnet *)ifp)->if_flags = flags;
4135 	return (0);
4136 }
4137 
4138 int
4139 if_setflagbits(if_t ifp, int set, int clear)
4140 {
4141 	((struct ifnet *)ifp)->if_flags |= set;
4142 	((struct ifnet *)ifp)->if_flags &= ~clear;
4143 
4144 	return (0);
4145 }
4146 
4147 int
4148 if_getflags(if_t ifp)
4149 {
4150 	return ((struct ifnet *)ifp)->if_flags;
4151 }
4152 
4153 int
4154 if_clearhwassist(if_t ifp)
4155 {
4156 	((struct ifnet *)ifp)->if_hwassist = 0;
4157 	return (0);
4158 }
4159 
4160 int
4161 if_sethwassistbits(if_t ifp, int toset, int toclear)
4162 {
4163 	((struct ifnet *)ifp)->if_hwassist |= toset;
4164 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4165 
4166 	return (0);
4167 }
4168 
4169 int
4170 if_sethwassist(if_t ifp, int hwassist_bit)
4171 {
4172 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4173 	return (0);
4174 }
4175 
4176 int
4177 if_gethwassist(if_t ifp)
4178 {
4179 	return ((struct ifnet *)ifp)->if_hwassist;
4180 }
4181 
4182 int
4183 if_setmtu(if_t ifp, int mtu)
4184 {
4185 	((struct ifnet *)ifp)->if_mtu = mtu;
4186 	return (0);
4187 }
4188 
4189 int
4190 if_getmtu(if_t ifp)
4191 {
4192 	return ((struct ifnet *)ifp)->if_mtu;
4193 }
4194 
4195 int
4196 if_getmtu_family(if_t ifp, int family)
4197 {
4198 	struct domain *dp;
4199 
4200 	for (dp = domains; dp; dp = dp->dom_next) {
4201 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4202 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4203 	}
4204 
4205 	return (((struct ifnet *)ifp)->if_mtu);
4206 }
4207 
4208 /*
4209  * Methods for drivers to access interface unicast and multicast
4210  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4211  * 'struct ifmultiaddr'.
4212  */
4213 u_int
4214 if_lladdr_count(if_t ifp)
4215 {
4216 	struct epoch_tracker et;
4217 	struct ifaddr *ifa;
4218 	u_int count;
4219 
4220 	count = 0;
4221 	NET_EPOCH_ENTER(et);
4222 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4223 		if (ifa->ifa_addr->sa_family == AF_LINK)
4224 			count++;
4225 	NET_EPOCH_EXIT(et);
4226 
4227 	return (count);
4228 }
4229 
4230 u_int
4231 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4232 {
4233 	struct epoch_tracker et;
4234 	struct ifaddr *ifa;
4235 	u_int count;
4236 
4237 	MPASS(cb);
4238 
4239 	count = 0;
4240 	NET_EPOCH_ENTER(et);
4241 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4242 		if (ifa->ifa_addr->sa_family != AF_LINK)
4243 			continue;
4244 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4245 		    count);
4246 	}
4247 	NET_EPOCH_EXIT(et);
4248 
4249 	return (count);
4250 }
4251 
4252 u_int
4253 if_llmaddr_count(if_t ifp)
4254 {
4255 	struct epoch_tracker et;
4256 	struct ifmultiaddr *ifma;
4257 	int count;
4258 
4259 	count = 0;
4260 	NET_EPOCH_ENTER(et);
4261 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4262 		if (ifma->ifma_addr->sa_family == AF_LINK)
4263 			count++;
4264 	NET_EPOCH_EXIT(et);
4265 
4266 	return (count);
4267 }
4268 
4269 u_int
4270 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4271 {
4272 	struct epoch_tracker et;
4273 	struct ifmultiaddr *ifma;
4274 	u_int count;
4275 
4276 	MPASS(cb);
4277 
4278 	count = 0;
4279 	NET_EPOCH_ENTER(et);
4280 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4281 		if (ifma->ifma_addr->sa_family != AF_LINK)
4282 			continue;
4283 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4284 		    count);
4285 	}
4286 	NET_EPOCH_EXIT(et);
4287 
4288 	return (count);
4289 }
4290 
4291 int
4292 if_setsoftc(if_t ifp, void *softc)
4293 {
4294 	((struct ifnet *)ifp)->if_softc = softc;
4295 	return (0);
4296 }
4297 
4298 void *
4299 if_getsoftc(if_t ifp)
4300 {
4301 	return ((struct ifnet *)ifp)->if_softc;
4302 }
4303 
4304 void
4305 if_setrcvif(struct mbuf *m, if_t ifp)
4306 {
4307 
4308 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4309 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4310 }
4311 
4312 void
4313 if_setvtag(struct mbuf *m, uint16_t tag)
4314 {
4315 	m->m_pkthdr.ether_vtag = tag;
4316 }
4317 
4318 uint16_t
4319 if_getvtag(struct mbuf *m)
4320 {
4321 
4322 	return (m->m_pkthdr.ether_vtag);
4323 }
4324 
4325 int
4326 if_sendq_empty(if_t ifp)
4327 {
4328 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4329 }
4330 
4331 struct ifaddr *
4332 if_getifaddr(if_t ifp)
4333 {
4334 	return ((struct ifnet *)ifp)->if_addr;
4335 }
4336 
4337 int
4338 if_getamcount(if_t ifp)
4339 {
4340 	return ((struct ifnet *)ifp)->if_amcount;
4341 }
4342 
4343 
4344 int
4345 if_setsendqready(if_t ifp)
4346 {
4347 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4348 	return (0);
4349 }
4350 
4351 int
4352 if_setsendqlen(if_t ifp, int tx_desc_count)
4353 {
4354 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4355 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4356 
4357 	return (0);
4358 }
4359 
4360 int
4361 if_vlantrunkinuse(if_t ifp)
4362 {
4363 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4364 }
4365 
4366 int
4367 if_input(if_t ifp, struct mbuf* sendmp)
4368 {
4369 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4370 	return (0);
4371 
4372 }
4373 
4374 struct mbuf *
4375 if_dequeue(if_t ifp)
4376 {
4377 	struct mbuf *m;
4378 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4379 
4380 	return (m);
4381 }
4382 
4383 int
4384 if_sendq_prepend(if_t ifp, struct mbuf *m)
4385 {
4386 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4387 	return (0);
4388 }
4389 
4390 int
4391 if_setifheaderlen(if_t ifp, int len)
4392 {
4393 	((struct ifnet *)ifp)->if_hdrlen = len;
4394 	return (0);
4395 }
4396 
4397 caddr_t
4398 if_getlladdr(if_t ifp)
4399 {
4400 	return (IF_LLADDR((struct ifnet *)ifp));
4401 }
4402 
4403 void *
4404 if_gethandle(u_char type)
4405 {
4406 	return (if_alloc(type));
4407 }
4408 
4409 void
4410 if_bpfmtap(if_t ifh, struct mbuf *m)
4411 {
4412 	struct ifnet *ifp = (struct ifnet *)ifh;
4413 
4414 	BPF_MTAP(ifp, m);
4415 }
4416 
4417 void
4418 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4419 {
4420 	struct ifnet *ifp = (struct ifnet *)ifh;
4421 
4422 	ETHER_BPF_MTAP(ifp, m);
4423 }
4424 
4425 void
4426 if_vlancap(if_t ifh)
4427 {
4428 	struct ifnet *ifp = (struct ifnet *)ifh;
4429 	VLAN_CAPABILITIES(ifp);
4430 }
4431 
4432 int
4433 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4434 {
4435 
4436 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4437         return (0);
4438 }
4439 
4440 int
4441 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4442 {
4443 
4444 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4445         return (0);
4446 }
4447 
4448 int
4449 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4450 {
4451 
4452 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4453         return (0);
4454 }
4455 
4456 u_int
4457 if_gethwtsomax(if_t ifp)
4458 {
4459 
4460 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4461 }
4462 
4463 u_int
4464 if_gethwtsomaxsegcount(if_t ifp)
4465 {
4466 
4467 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4468 }
4469 
4470 u_int
4471 if_gethwtsomaxsegsize(if_t ifp)
4472 {
4473 
4474 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4475 }
4476 
4477 void
4478 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4479 {
4480 	((struct ifnet *)ifp)->if_init = init_fn;
4481 }
4482 
4483 void
4484 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4485 {
4486 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4487 }
4488 
4489 void
4490 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4491 {
4492 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4493 }
4494 
4495 void
4496 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4497 {
4498 	((struct ifnet *)ifp)->if_transmit = start_fn;
4499 }
4500 
4501 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4502 {
4503 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4504 
4505 }
4506 
4507 void
4508 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4509 {
4510 
4511 	ifp->if_get_counter = fn;
4512 }
4513 
4514 /* Revisit these - These are inline functions originally. */
4515 int
4516 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4517 {
4518 	return drbr_inuse(ifh, br);
4519 }
4520 
4521 struct mbuf*
4522 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4523 {
4524 	return drbr_dequeue(ifh, br);
4525 }
4526 
4527 int
4528 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4529 {
4530 	return drbr_needs_enqueue(ifh, br);
4531 }
4532 
4533 int
4534 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4535 {
4536 	return drbr_enqueue(ifh, br, m);
4537 
4538 }
4539