xref: /freebsd/sys/net/if.c (revision 488ab515d6cc02f6f743f0badfc8e94eb553cd30)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/refcount.h>
54 #include <sys/module.h>
55 #include <sys/rwlock.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <sys/priv.h>
64 
65 #include <machine/stdarg.h>
66 #include <vm/uma.h>
67 
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/if_media.h>
77 #include <net/if_vlan_var.h>
78 #include <net/radix.h>
79 #include <net/route.h>
80 #include <net/vnet.h>
81 
82 #if defined(INET) || defined(INET6)
83 #include <net/ethernet.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip.h>
87 #include <netinet/ip_carp.h>
88 #ifdef INET
89 #include <netinet/if_ether.h>
90 #include <netinet/netdump/netdump.h>
91 #endif /* INET */
92 #ifdef INET6
93 #include <netinet6/in6_var.h>
94 #include <netinet6/in6_ifattach.h>
95 #endif /* INET6 */
96 #endif /* INET || INET6 */
97 
98 #include <security/mac/mac_framework.h>
99 
100 /*
101  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
102  * and ifr_ifru when it is used in SIOCGIFCONF.
103  */
104 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
105     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
106 
107 #ifdef COMPAT_FREEBSD32
108 #include <sys/mount.h>
109 #include <compat/freebsd32/freebsd32.h>
110 
111 struct ifreq_buffer32 {
112 	uint32_t	length;		/* (size_t) */
113 	uint32_t	buffer;		/* (void *) */
114 };
115 
116 /*
117  * Interface request structure used for socket
118  * ioctl's.  All interface ioctl's must have parameter
119  * definitions which begin with ifr_name.  The
120  * remainder may be interface specific.
121  */
122 struct ifreq32 {
123 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
124 	union {
125 		struct sockaddr	ifru_addr;
126 		struct sockaddr	ifru_dstaddr;
127 		struct sockaddr	ifru_broadaddr;
128 		struct ifreq_buffer32 ifru_buffer;
129 		short		ifru_flags[2];
130 		short		ifru_index;
131 		int		ifru_jid;
132 		int		ifru_metric;
133 		int		ifru_mtu;
134 		int		ifru_phys;
135 		int		ifru_media;
136 		uint32_t	ifru_data;
137 		int		ifru_cap[2];
138 		u_int		ifru_fib;
139 		u_char		ifru_vlan_pcp;
140 	} ifr_ifru;
141 };
142 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
143 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
144     __offsetof(struct ifreq32, ifr_ifru));
145 
146 struct ifgroupreq32 {
147 	char	ifgr_name[IFNAMSIZ];
148 	u_int	ifgr_len;
149 	union {
150 		char		ifgru_group[IFNAMSIZ];
151 		uint32_t	ifgru_groups;
152 	} ifgr_ifgru;
153 };
154 
155 struct ifmediareq32 {
156 	char		ifm_name[IFNAMSIZ];
157 	int		ifm_current;
158 	int		ifm_mask;
159 	int		ifm_status;
160 	int		ifm_active;
161 	int		ifm_count;
162 	uint32_t	ifm_ulist;	/* (int *) */
163 };
164 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
165 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
166 
167 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
168     case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
169 #else /* !COMPAT_FREEBSD32 */
170 #define _CASE_IOC_IFGROUPREQ_32(cmd)
171 #endif /* !COMPAT_FREEBSD32 */
172 
173 #define CASE_IOC_IFGROUPREQ(cmd)	\
174     _CASE_IOC_IFGROUPREQ_32(cmd)	\
175     case (cmd)
176 
177 union ifreq_union {
178 	struct ifreq	ifr;
179 #ifdef COMPAT_FREEBSD32
180 	struct ifreq32	ifr32;
181 #endif
182 };
183 
184 union ifgroupreq_union {
185 	struct ifgroupreq ifgr;
186 #ifdef COMPAT_FREEBSD32
187 	struct ifgroupreq32 ifgr32;
188 #endif
189 };
190 
191 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
192 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
193 
194 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
195     &ifqmaxlen, 0, "max send queue size");
196 
197 /* Log link state change events */
198 static int log_link_state_change = 1;
199 
200 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
201 	&log_link_state_change, 0,
202 	"log interface link state change events");
203 
204 /* Log promiscuous mode change events */
205 static int log_promisc_mode_change = 1;
206 
207 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
208 	&log_promisc_mode_change, 1,
209 	"log promiscuous mode change events");
210 
211 /* Interface description */
212 static unsigned int ifdescr_maxlen = 1024;
213 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
214 	&ifdescr_maxlen, 0,
215 	"administrative maximum length for interface description");
216 
217 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
218 
219 /* global sx for non-critical path ifdescr */
220 static struct sx ifdescr_sx;
221 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
222 
223 void	(*bridge_linkstate_p)(struct ifnet *ifp);
224 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
225 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
226 /* These are external hooks for CARP. */
227 void	(*carp_linkstate_p)(struct ifnet *ifp);
228 void	(*carp_demote_adj_p)(int, char *);
229 int	(*carp_master_p)(struct ifaddr *);
230 #if defined(INET) || defined(INET6)
231 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
232 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
233     const struct sockaddr *sa);
234 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
235 int	(*carp_attach_p)(struct ifaddr *, int);
236 void	(*carp_detach_p)(struct ifaddr *, bool);
237 #endif
238 #ifdef INET
239 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
240 #endif
241 #ifdef INET6
242 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
243 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
244     const struct in6_addr *taddr);
245 #endif
246 
247 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
248 
249 /*
250  * XXX: Style; these should be sorted alphabetically, and unprototyped
251  * static functions should be prototyped. Currently they are sorted by
252  * declaration order.
253  */
254 static void	if_attachdomain(void *);
255 static void	if_attachdomain1(struct ifnet *);
256 static int	ifconf(u_long, caddr_t);
257 static void	if_grow(void);
258 static void	if_input_default(struct ifnet *, struct mbuf *);
259 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
260 static void	if_route(struct ifnet *, int flag, int fam);
261 static int	if_setflag(struct ifnet *, int, int, int *, int);
262 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
263 static void	if_unroute(struct ifnet *, int flag, int fam);
264 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
265 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
266 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
267 static void	do_link_state_change(void *, int);
268 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
269 static int	if_getgroupmembers(struct ifgroupreq *);
270 static void	if_delgroups(struct ifnet *);
271 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
272 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
273 #ifdef VIMAGE
274 static void	if_vmove(struct ifnet *, struct vnet *);
275 #endif
276 
277 #ifdef INET6
278 /*
279  * XXX: declare here to avoid to include many inet6 related files..
280  * should be more generalized?
281  */
282 extern void	nd6_setmtu(struct ifnet *);
283 #endif
284 
285 /* ipsec helper hooks */
286 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
288 
289 VNET_DEFINE(int, if_index);
290 int	ifqmaxlen = IFQ_MAXLEN;
291 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
292 VNET_DEFINE(struct ifgrouphead, ifg_head);
293 
294 static VNET_DEFINE(int, if_indexlim) = 8;
295 
296 /* Table of ifnet by index. */
297 VNET_DEFINE(struct ifnet **, ifindex_table);
298 
299 #define	V_if_indexlim		VNET(if_indexlim)
300 #define	V_ifindex_table		VNET(ifindex_table)
301 
302 /*
303  * The global network interface list (V_ifnet) and related state (such as
304  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
305  * an rwlock.  Either may be acquired shared to stablize the list, but both
306  * must be acquired writable to modify the list.  This model allows us to
307  * both stablize the interface list during interrupt thread processing, but
308  * also to stablize it over long-running ioctls, without introducing priority
309  * inversions and deadlocks.
310  */
311 struct rwlock ifnet_rwlock;
312 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
313 struct sx ifnet_sxlock;
314 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
315 
316 /*
317  * The allocation of network interfaces is a rather non-atomic affair; we
318  * need to select an index before we are ready to expose the interface for
319  * use, so will use this pointer value to indicate reservation.
320  */
321 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
322 
323 static	if_com_alloc_t *if_com_alloc[256];
324 static	if_com_free_t *if_com_free[256];
325 
326 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
327 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
328 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
329 
330 struct ifnet *
331 ifnet_byindex_locked(u_short idx)
332 {
333 
334 	if (idx > V_if_index)
335 		return (NULL);
336 	if (V_ifindex_table[idx] == IFNET_HOLD)
337 		return (NULL);
338 	return (V_ifindex_table[idx]);
339 }
340 
341 struct ifnet *
342 ifnet_byindex(u_short idx)
343 {
344 	struct ifnet *ifp;
345 
346 	IFNET_RLOCK_NOSLEEP();
347 	ifp = ifnet_byindex_locked(idx);
348 	IFNET_RUNLOCK_NOSLEEP();
349 	return (ifp);
350 }
351 
352 struct ifnet *
353 ifnet_byindex_ref(u_short idx)
354 {
355 	struct ifnet *ifp;
356 
357 	IFNET_RLOCK_NOSLEEP();
358 	ifp = ifnet_byindex_locked(idx);
359 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
360 		IFNET_RUNLOCK_NOSLEEP();
361 		return (NULL);
362 	}
363 	if_ref(ifp);
364 	IFNET_RUNLOCK_NOSLEEP();
365 	return (ifp);
366 }
367 
368 /*
369  * Allocate an ifindex array entry; return 0 on success or an error on
370  * failure.
371  */
372 static u_short
373 ifindex_alloc(void)
374 {
375 	u_short idx;
376 
377 	IFNET_WLOCK_ASSERT();
378 retry:
379 	/*
380 	 * Try to find an empty slot below V_if_index.  If we fail, take the
381 	 * next slot.
382 	 */
383 	for (idx = 1; idx <= V_if_index; idx++) {
384 		if (V_ifindex_table[idx] == NULL)
385 			break;
386 	}
387 
388 	/* Catch if_index overflow. */
389 	if (idx >= V_if_indexlim) {
390 		if_grow();
391 		goto retry;
392 	}
393 	if (idx > V_if_index)
394 		V_if_index = idx;
395 	return (idx);
396 }
397 
398 static void
399 ifindex_free_locked(u_short idx)
400 {
401 
402 	IFNET_WLOCK_ASSERT();
403 
404 	V_ifindex_table[idx] = NULL;
405 	while (V_if_index > 0 &&
406 	    V_ifindex_table[V_if_index] == NULL)
407 		V_if_index--;
408 }
409 
410 static void
411 ifindex_free(u_short idx)
412 {
413 
414 	IFNET_WLOCK();
415 	ifindex_free_locked(idx);
416 	IFNET_WUNLOCK();
417 }
418 
419 static void
420 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
421 {
422 
423 	IFNET_WLOCK_ASSERT();
424 
425 	V_ifindex_table[idx] = ifp;
426 }
427 
428 static void
429 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
430 {
431 
432 	IFNET_WLOCK();
433 	ifnet_setbyindex_locked(idx, ifp);
434 	IFNET_WUNLOCK();
435 }
436 
437 struct ifaddr *
438 ifaddr_byindex(u_short idx)
439 {
440 	struct ifnet *ifp;
441 	struct ifaddr *ifa = NULL;
442 
443 	IFNET_RLOCK_NOSLEEP();
444 	ifp = ifnet_byindex_locked(idx);
445 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
446 		ifa_ref(ifa);
447 	IFNET_RUNLOCK_NOSLEEP();
448 	return (ifa);
449 }
450 
451 /*
452  * Network interface utility routines.
453  *
454  * Routines with ifa_ifwith* names take sockaddr *'s as
455  * parameters.
456  */
457 
458 static void
459 vnet_if_init(const void *unused __unused)
460 {
461 
462 	TAILQ_INIT(&V_ifnet);
463 	TAILQ_INIT(&V_ifg_head);
464 	IFNET_WLOCK();
465 	if_grow();				/* create initial table */
466 	IFNET_WUNLOCK();
467 	vnet_if_clone_init();
468 }
469 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
470     NULL);
471 
472 #ifdef VIMAGE
473 static void
474 vnet_if_uninit(const void *unused __unused)
475 {
476 
477 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
478 	    "not empty", __func__, __LINE__, &V_ifnet));
479 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
480 	    "not empty", __func__, __LINE__, &V_ifg_head));
481 
482 	free((caddr_t)V_ifindex_table, M_IFNET);
483 }
484 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
485     vnet_if_uninit, NULL);
486 
487 static void
488 vnet_if_return(const void *unused __unused)
489 {
490 	struct ifnet *ifp, *nifp;
491 
492 	/* Return all inherited interfaces to their parent vnets. */
493 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
494 		if (ifp->if_home_vnet != ifp->if_vnet)
495 			if_vmove(ifp, ifp->if_home_vnet);
496 	}
497 }
498 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
499     vnet_if_return, NULL);
500 #endif
501 
502 static void
503 if_grow(void)
504 {
505 	int oldlim;
506 	u_int n;
507 	struct ifnet **e;
508 
509 	IFNET_WLOCK_ASSERT();
510 	oldlim = V_if_indexlim;
511 	IFNET_WUNLOCK();
512 	n = (oldlim << 1) * sizeof(*e);
513 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
514 	IFNET_WLOCK();
515 	if (V_if_indexlim != oldlim) {
516 		free(e, M_IFNET);
517 		return;
518 	}
519 	if (V_ifindex_table != NULL) {
520 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
521 		free((caddr_t)V_ifindex_table, M_IFNET);
522 	}
523 	V_if_indexlim <<= 1;
524 	V_ifindex_table = e;
525 }
526 
527 /*
528  * Allocate a struct ifnet and an index for an interface.  A layer 2
529  * common structure will also be allocated if an allocation routine is
530  * registered for the passed type.
531  */
532 struct ifnet *
533 if_alloc(u_char type)
534 {
535 	struct ifnet *ifp;
536 	u_short idx;
537 
538 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
539 	IFNET_WLOCK();
540 	idx = ifindex_alloc();
541 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
542 	IFNET_WUNLOCK();
543 	ifp->if_index = idx;
544 	ifp->if_type = type;
545 	ifp->if_alloctype = type;
546 #ifdef VIMAGE
547 	ifp->if_vnet = curvnet;
548 #endif
549 	if (if_com_alloc[type] != NULL) {
550 		ifp->if_l2com = if_com_alloc[type](type, ifp);
551 		if (ifp->if_l2com == NULL) {
552 			free(ifp, M_IFNET);
553 			ifindex_free(idx);
554 			return (NULL);
555 		}
556 	}
557 
558 	IF_ADDR_LOCK_INIT(ifp);
559 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
560 	ifp->if_afdata_initialized = 0;
561 	IF_AFDATA_LOCK_INIT(ifp);
562 	TAILQ_INIT(&ifp->if_addrhead);
563 	TAILQ_INIT(&ifp->if_multiaddrs);
564 	TAILQ_INIT(&ifp->if_groups);
565 #ifdef MAC
566 	mac_ifnet_init(ifp);
567 #endif
568 	ifq_init(&ifp->if_snd, ifp);
569 
570 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
571 	for (int i = 0; i < IFCOUNTERS; i++)
572 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
573 	ifp->if_get_counter = if_get_counter_default;
574 	ifp->if_pcp = IFNET_PCP_NONE;
575 	ifnet_setbyindex(ifp->if_index, ifp);
576 	return (ifp);
577 }
578 
579 /*
580  * Do the actual work of freeing a struct ifnet, and layer 2 common
581  * structure.  This call is made when the last reference to an
582  * interface is released.
583  */
584 static void
585 if_free_internal(struct ifnet *ifp)
586 {
587 
588 	KASSERT((ifp->if_flags & IFF_DYING),
589 	    ("if_free_internal: interface not dying"));
590 
591 	if (if_com_free[ifp->if_alloctype] != NULL)
592 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
593 		    ifp->if_alloctype);
594 
595 #ifdef MAC
596 	mac_ifnet_destroy(ifp);
597 #endif /* MAC */
598 	if (ifp->if_description != NULL)
599 		free(ifp->if_description, M_IFDESCR);
600 	IF_AFDATA_DESTROY(ifp);
601 	IF_ADDR_LOCK_DESTROY(ifp);
602 	ifq_delete(&ifp->if_snd);
603 
604 	for (int i = 0; i < IFCOUNTERS; i++)
605 		counter_u64_free(ifp->if_counters[i]);
606 
607 	free(ifp, M_IFNET);
608 }
609 
610 /*
611  * Deregister an interface and free the associated storage.
612  */
613 void
614 if_free(struct ifnet *ifp)
615 {
616 
617 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
618 
619 	CURVNET_SET_QUIET(ifp->if_vnet);
620 	IFNET_WLOCK();
621 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
622 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
623 
624 	ifindex_free_locked(ifp->if_index);
625 	IFNET_WUNLOCK();
626 
627 	if (refcount_release(&ifp->if_refcount))
628 		if_free_internal(ifp);
629 	CURVNET_RESTORE();
630 }
631 
632 /*
633  * Interfaces to keep an ifnet type-stable despite the possibility of the
634  * driver calling if_free().  If there are additional references, we defer
635  * freeing the underlying data structure.
636  */
637 void
638 if_ref(struct ifnet *ifp)
639 {
640 
641 	/* We don't assert the ifnet list lock here, but arguably should. */
642 	refcount_acquire(&ifp->if_refcount);
643 }
644 
645 void
646 if_rele(struct ifnet *ifp)
647 {
648 
649 	if (!refcount_release(&ifp->if_refcount))
650 		return;
651 	if_free_internal(ifp);
652 }
653 
654 void
655 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
656 {
657 
658 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
659 
660 	if (ifq->ifq_maxlen == 0)
661 		ifq->ifq_maxlen = ifqmaxlen;
662 
663 	ifq->altq_type = 0;
664 	ifq->altq_disc = NULL;
665 	ifq->altq_flags &= ALTQF_CANTCHANGE;
666 	ifq->altq_tbr  = NULL;
667 	ifq->altq_ifp  = ifp;
668 }
669 
670 void
671 ifq_delete(struct ifaltq *ifq)
672 {
673 	mtx_destroy(&ifq->ifq_mtx);
674 }
675 
676 /*
677  * Perform generic interface initialization tasks and attach the interface
678  * to the list of "active" interfaces.  If vmove flag is set on entry
679  * to if_attach_internal(), perform only a limited subset of initialization
680  * tasks, given that we are moving from one vnet to another an ifnet which
681  * has already been fully initialized.
682  *
683  * Note that if_detach_internal() removes group membership unconditionally
684  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
685  * Thus, when if_vmove() is applied to a cloned interface, group membership
686  * is lost while a cloned one always joins a group whose name is
687  * ifc->ifc_name.  To recover this after if_detach_internal() and
688  * if_attach_internal(), the cloner should be specified to
689  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
690  * attempts to join a group whose name is ifc->ifc_name.
691  *
692  * XXX:
693  *  - The decision to return void and thus require this function to
694  *    succeed is questionable.
695  *  - We should probably do more sanity checking.  For instance we don't
696  *    do anything to insure if_xname is unique or non-empty.
697  */
698 void
699 if_attach(struct ifnet *ifp)
700 {
701 
702 	if_attach_internal(ifp, 0, NULL);
703 }
704 
705 /*
706  * Compute the least common TSO limit.
707  */
708 void
709 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
710 {
711 	/*
712 	 * 1) If there is no limit currently, take the limit from
713 	 * the network adapter.
714 	 *
715 	 * 2) If the network adapter has a limit below the current
716 	 * limit, apply it.
717 	 */
718 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
719 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
720 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
721 	}
722 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
723 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
724 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
725 	}
726 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
727 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
728 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
729 	}
730 }
731 
732 /*
733  * Update TSO limit of a network adapter.
734  *
735  * Returns zero if no change. Else non-zero.
736  */
737 int
738 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
739 {
740 	int retval = 0;
741 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
742 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
743 		retval++;
744 	}
745 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
746 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
747 		retval++;
748 	}
749 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
750 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
751 		retval++;
752 	}
753 	return (retval);
754 }
755 
756 static void
757 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
758 {
759 	unsigned socksize, ifasize;
760 	int namelen, masklen;
761 	struct sockaddr_dl *sdl;
762 	struct ifaddr *ifa;
763 
764 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
765 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
766 		    ifp->if_xname);
767 
768 #ifdef VIMAGE
769 	ifp->if_vnet = curvnet;
770 	if (ifp->if_home_vnet == NULL)
771 		ifp->if_home_vnet = curvnet;
772 #endif
773 
774 	if_addgroup(ifp, IFG_ALL);
775 
776 	/* Restore group membership for cloned interfaces. */
777 	if (vmove && ifc != NULL)
778 		if_clone_addgroup(ifp, ifc);
779 
780 	getmicrotime(&ifp->if_lastchange);
781 	ifp->if_epoch = time_uptime;
782 
783 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
784 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
785 	    ("transmit and qflush must both either be set or both be NULL"));
786 	if (ifp->if_transmit == NULL) {
787 		ifp->if_transmit = if_transmit;
788 		ifp->if_qflush = if_qflush;
789 	}
790 	if (ifp->if_input == NULL)
791 		ifp->if_input = if_input_default;
792 
793 	if (ifp->if_requestencap == NULL)
794 		ifp->if_requestencap = if_requestencap_default;
795 
796 	if (!vmove) {
797 #ifdef MAC
798 		mac_ifnet_create(ifp);
799 #endif
800 
801 		/*
802 		 * Create a Link Level name for this device.
803 		 */
804 		namelen = strlen(ifp->if_xname);
805 		/*
806 		 * Always save enough space for any possiable name so we
807 		 * can do a rename in place later.
808 		 */
809 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
810 		socksize = masklen + ifp->if_addrlen;
811 		if (socksize < sizeof(*sdl))
812 			socksize = sizeof(*sdl);
813 		socksize = roundup2(socksize, sizeof(long));
814 		ifasize = sizeof(*ifa) + 2 * socksize;
815 		ifa = ifa_alloc(ifasize, M_WAITOK);
816 		sdl = (struct sockaddr_dl *)(ifa + 1);
817 		sdl->sdl_len = socksize;
818 		sdl->sdl_family = AF_LINK;
819 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
820 		sdl->sdl_nlen = namelen;
821 		sdl->sdl_index = ifp->if_index;
822 		sdl->sdl_type = ifp->if_type;
823 		ifp->if_addr = ifa;
824 		ifa->ifa_ifp = ifp;
825 		ifa->ifa_rtrequest = link_rtrequest;
826 		ifa->ifa_addr = (struct sockaddr *)sdl;
827 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
828 		ifa->ifa_netmask = (struct sockaddr *)sdl;
829 		sdl->sdl_len = masklen;
830 		while (namelen != 0)
831 			sdl->sdl_data[--namelen] = 0xff;
832 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
833 		/* Reliably crash if used uninitialized. */
834 		ifp->if_broadcastaddr = NULL;
835 
836 		if (ifp->if_type == IFT_ETHER) {
837 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
838 			    M_WAITOK | M_ZERO);
839 		}
840 
841 #if defined(INET) || defined(INET6)
842 		/* Use defaults for TSO, if nothing is set */
843 		if (ifp->if_hw_tsomax == 0 &&
844 		    ifp->if_hw_tsomaxsegcount == 0 &&
845 		    ifp->if_hw_tsomaxsegsize == 0) {
846 			/*
847 			 * The TSO defaults needs to be such that an
848 			 * NFS mbuf list of 35 mbufs totalling just
849 			 * below 64K works and that a chain of mbufs
850 			 * can be defragged into at most 32 segments:
851 			 */
852 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
853 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
854 			ifp->if_hw_tsomaxsegcount = 35;
855 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
856 
857 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
858 			if (ifp->if_capabilities & IFCAP_TSO) {
859 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
860 				    ifp->if_hw_tsomax,
861 				    ifp->if_hw_tsomaxsegcount,
862 				    ifp->if_hw_tsomaxsegsize);
863 			}
864 		}
865 #endif
866 	}
867 #ifdef VIMAGE
868 	else {
869 		/*
870 		 * Update the interface index in the link layer address
871 		 * of the interface.
872 		 */
873 		for (ifa = ifp->if_addr; ifa != NULL;
874 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
875 			if (ifa->ifa_addr->sa_family == AF_LINK) {
876 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
877 				sdl->sdl_index = ifp->if_index;
878 			}
879 		}
880 	}
881 #endif
882 
883 	IFNET_WLOCK();
884 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
885 #ifdef VIMAGE
886 	curvnet->vnet_ifcnt++;
887 #endif
888 	IFNET_WUNLOCK();
889 
890 	if (domain_init_status >= 2)
891 		if_attachdomain1(ifp);
892 
893 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
894 	if (IS_DEFAULT_VNET(curvnet))
895 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
896 
897 	/* Announce the interface. */
898 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
899 }
900 
901 static void
902 if_attachdomain(void *dummy)
903 {
904 	struct ifnet *ifp;
905 
906 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
907 		if_attachdomain1(ifp);
908 }
909 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
910     if_attachdomain, NULL);
911 
912 static void
913 if_attachdomain1(struct ifnet *ifp)
914 {
915 	struct domain *dp;
916 
917 	/*
918 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
919 	 * cannot lock ifp->if_afdata initialization, entirely.
920 	 */
921 	IF_AFDATA_LOCK(ifp);
922 	if (ifp->if_afdata_initialized >= domain_init_status) {
923 		IF_AFDATA_UNLOCK(ifp);
924 		log(LOG_WARNING, "%s called more than once on %s\n",
925 		    __func__, ifp->if_xname);
926 		return;
927 	}
928 	ifp->if_afdata_initialized = domain_init_status;
929 	IF_AFDATA_UNLOCK(ifp);
930 
931 	/* address family dependent data region */
932 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
933 	for (dp = domains; dp; dp = dp->dom_next) {
934 		if (dp->dom_ifattach)
935 			ifp->if_afdata[dp->dom_family] =
936 			    (*dp->dom_ifattach)(ifp);
937 	}
938 }
939 
940 /*
941  * Remove any unicast or broadcast network addresses from an interface.
942  */
943 void
944 if_purgeaddrs(struct ifnet *ifp)
945 {
946 	struct ifaddr *ifa, *next;
947 
948 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
949 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
950 		if (ifa->ifa_addr->sa_family == AF_LINK)
951 			continue;
952 #ifdef INET
953 		/* XXX: Ugly!! ad hoc just for INET */
954 		if (ifa->ifa_addr->sa_family == AF_INET) {
955 			struct ifaliasreq ifr;
956 
957 			bzero(&ifr, sizeof(ifr));
958 			ifr.ifra_addr = *ifa->ifa_addr;
959 			if (ifa->ifa_dstaddr)
960 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
961 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
962 			    NULL) == 0)
963 				continue;
964 		}
965 #endif /* INET */
966 #ifdef INET6
967 		if (ifa->ifa_addr->sa_family == AF_INET6) {
968 			in6_purgeaddr(ifa);
969 			/* ifp_addrhead is already updated */
970 			continue;
971 		}
972 #endif /* INET6 */
973 		IF_ADDR_WLOCK(ifp);
974 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
975 		IF_ADDR_WUNLOCK(ifp);
976 		ifa_free(ifa);
977 	}
978 }
979 
980 /*
981  * Remove any multicast network addresses from an interface when an ifnet
982  * is going away.
983  */
984 static void
985 if_purgemaddrs(struct ifnet *ifp)
986 {
987 	struct ifmultiaddr *ifma;
988 
989 	IF_ADDR_WLOCK(ifp);
990 	while (!TAILQ_EMPTY(&ifp->if_multiaddrs)) {
991 		ifma = TAILQ_FIRST(&ifp->if_multiaddrs);
992 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
993 		if_delmulti_locked(ifp, ifma, 1);
994 	}
995 	IF_ADDR_WUNLOCK(ifp);
996 }
997 
998 /*
999  * Detach an interface, removing it from the list of "active" interfaces.
1000  * If vmove flag is set on entry to if_detach_internal(), perform only a
1001  * limited subset of cleanup tasks, given that we are moving an ifnet from
1002  * one vnet to another, where it must be fully operational.
1003  *
1004  * XXXRW: There are some significant questions about event ordering, and
1005  * how to prevent things from starting to use the interface during detach.
1006  */
1007 void
1008 if_detach(struct ifnet *ifp)
1009 {
1010 
1011 	CURVNET_SET_QUIET(ifp->if_vnet);
1012 	if_detach_internal(ifp, 0, NULL);
1013 	CURVNET_RESTORE();
1014 }
1015 
1016 /*
1017  * The vmove flag, if set, indicates that we are called from a callpath
1018  * that is moving an interface to a different vnet instance.
1019  *
1020  * The shutdown flag, if set, indicates that we are called in the
1021  * process of shutting down a vnet instance.  Currently only the
1022  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1023  * on a vnet instance shutdown without this flag being set, e.g., when
1024  * the cloned interfaces are destoyed as first thing of teardown.
1025  */
1026 static int
1027 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1028 {
1029 	struct ifaddr *ifa;
1030 	int i;
1031 	struct domain *dp;
1032  	struct ifnet *iter;
1033  	int found = 0;
1034 #ifdef VIMAGE
1035 	int shutdown;
1036 
1037 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1038 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1039 #endif
1040 	IFNET_WLOCK();
1041 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
1042 		if (iter == ifp) {
1043 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
1044 			found = 1;
1045 			break;
1046 		}
1047 	IFNET_WUNLOCK();
1048 	if (!found) {
1049 		/*
1050 		 * While we would want to panic here, we cannot
1051 		 * guarantee that the interface is indeed still on
1052 		 * the list given we don't hold locks all the way.
1053 		 */
1054 		return (ENOENT);
1055 #if 0
1056 		if (vmove)
1057 			panic("%s: ifp=%p not on the ifnet tailq %p",
1058 			    __func__, ifp, &V_ifnet);
1059 		else
1060 			return; /* XXX this should panic as well? */
1061 #endif
1062 	}
1063 
1064 	/*
1065 	 * At this point we know the interface still was on the ifnet list
1066 	 * and we removed it so we are in a stable state.
1067 	 */
1068 #ifdef VIMAGE
1069 	curvnet->vnet_ifcnt--;
1070 #endif
1071 
1072 	/*
1073 	 * In any case (destroy or vmove) detach us from the groups
1074 	 * and remove/wait for pending events on the taskq.
1075 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1076 	 */
1077 	if_delgroups(ifp);
1078 
1079 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1080 
1081 	/*
1082 	 * Check if this is a cloned interface or not. Must do even if
1083 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1084 	 * the if_clone_addgroup() will have a corrupted string overwise
1085 	 * from a gibberish pointer.
1086 	 */
1087 	if (vmove && ifcp != NULL)
1088 		*ifcp = if_clone_findifc(ifp);
1089 
1090 	if_down(ifp);
1091 
1092 #ifdef VIMAGE
1093 	/*
1094 	 * On VNET shutdown abort here as the stack teardown will do all
1095 	 * the work top-down for us.
1096 	 */
1097 	if (shutdown) {
1098 		/*
1099 		 * In case of a vmove we are done here without error.
1100 		 * If we would signal an error it would lead to the same
1101 		 * abort as if we did not find the ifnet anymore.
1102 		 * if_detach() calls us in void context and does not care
1103 		 * about an early abort notification, so life is splendid :)
1104 		 */
1105 		goto finish_vnet_shutdown;
1106 	}
1107 #endif
1108 
1109 	/*
1110 	 * At this point we are not tearing down a VNET and are either
1111 	 * going to destroy or vmove the interface and have to cleanup
1112 	 * accordingly.
1113 	 */
1114 
1115 	/*
1116 	 * Remove routes and flush queues.
1117 	 */
1118 #ifdef ALTQ
1119 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1120 		altq_disable(&ifp->if_snd);
1121 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1122 		altq_detach(&ifp->if_snd);
1123 #endif
1124 
1125 	if_purgeaddrs(ifp);
1126 
1127 #ifdef INET
1128 	in_ifdetach(ifp);
1129 #endif
1130 
1131 #ifdef INET6
1132 	/*
1133 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1134 	 * before removing routing entries below, since IPv6 interface direct
1135 	 * routes are expected to be removed by the IPv6-specific kernel API.
1136 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1137 	 */
1138 	in6_ifdetach(ifp);
1139 #endif
1140 	if_purgemaddrs(ifp);
1141 
1142 	/* Announce that the interface is gone. */
1143 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1144 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1145 	if (IS_DEFAULT_VNET(curvnet))
1146 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1147 
1148 	if (!vmove) {
1149 		/*
1150 		 * Prevent further calls into the device driver via ifnet.
1151 		 */
1152 		if_dead(ifp);
1153 
1154 		/*
1155 		 * Remove link ifaddr pointer and maybe decrement if_index.
1156 		 * Clean up all addresses.
1157 		 */
1158 		free(ifp->if_hw_addr, M_IFADDR);
1159 		ifp->if_hw_addr = NULL;
1160 		ifp->if_addr = NULL;
1161 
1162 		/* We can now free link ifaddr. */
1163 		IF_ADDR_WLOCK(ifp);
1164 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1165 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1166 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1167 			IF_ADDR_WUNLOCK(ifp);
1168 			ifa_free(ifa);
1169 		} else
1170 			IF_ADDR_WUNLOCK(ifp);
1171 	}
1172 
1173 	rt_flushifroutes(ifp);
1174 
1175 #ifdef VIMAGE
1176 finish_vnet_shutdown:
1177 #endif
1178 	/*
1179 	 * We cannot hold the lock over dom_ifdetach calls as they might
1180 	 * sleep, for example trying to drain a callout, thus open up the
1181 	 * theoretical race with re-attaching.
1182 	 */
1183 	IF_AFDATA_LOCK(ifp);
1184 	i = ifp->if_afdata_initialized;
1185 	ifp->if_afdata_initialized = 0;
1186 	IF_AFDATA_UNLOCK(ifp);
1187 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1188 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1189 			(*dp->dom_ifdetach)(ifp,
1190 			    ifp->if_afdata[dp->dom_family]);
1191 			ifp->if_afdata[dp->dom_family] = NULL;
1192 		}
1193 	}
1194 
1195 	return (0);
1196 }
1197 
1198 #ifdef VIMAGE
1199 /*
1200  * if_vmove() performs a limited version of if_detach() in current
1201  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1202  * An attempt is made to shrink if_index in current vnet, find an
1203  * unused if_index in target vnet and calls if_grow() if necessary,
1204  * and finally find an unused if_xname for the target vnet.
1205  */
1206 static void
1207 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1208 {
1209 	struct if_clone *ifc;
1210 	u_int bif_dlt, bif_hdrlen;
1211 	int rc;
1212 
1213  	/*
1214 	 * if_detach_internal() will call the eventhandler to notify
1215 	 * interface departure.  That will detach if_bpf.  We need to
1216 	 * safe the dlt and hdrlen so we can re-attach it later.
1217 	 */
1218 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1219 
1220 	/*
1221 	 * Detach from current vnet, but preserve LLADDR info, do not
1222 	 * mark as dead etc. so that the ifnet can be reattached later.
1223 	 * If we cannot find it, we lost the race to someone else.
1224 	 */
1225 	rc = if_detach_internal(ifp, 1, &ifc);
1226 	if (rc != 0)
1227 		return;
1228 
1229 	/*
1230 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1231 	 * the if_index for that vnet if possible.
1232 	 *
1233 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1234 	 * or we'd lock on one vnet and unlock on another.
1235 	 */
1236 	IFNET_WLOCK();
1237 	ifindex_free_locked(ifp->if_index);
1238 	IFNET_WUNLOCK();
1239 
1240 	/*
1241 	 * Perform interface-specific reassignment tasks, if provided by
1242 	 * the driver.
1243 	 */
1244 	if (ifp->if_reassign != NULL)
1245 		ifp->if_reassign(ifp, new_vnet, NULL);
1246 
1247 	/*
1248 	 * Switch to the context of the target vnet.
1249 	 */
1250 	CURVNET_SET_QUIET(new_vnet);
1251 
1252 	IFNET_WLOCK();
1253 	ifp->if_index = ifindex_alloc();
1254 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1255 	IFNET_WUNLOCK();
1256 
1257 	if_attach_internal(ifp, 1, ifc);
1258 
1259 	if (ifp->if_bpf == NULL)
1260 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1261 
1262 	CURVNET_RESTORE();
1263 }
1264 
1265 /*
1266  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1267  */
1268 static int
1269 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1270 {
1271 	struct prison *pr;
1272 	struct ifnet *difp;
1273 	int shutdown;
1274 
1275 	/* Try to find the prison within our visibility. */
1276 	sx_slock(&allprison_lock);
1277 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1278 	sx_sunlock(&allprison_lock);
1279 	if (pr == NULL)
1280 		return (ENXIO);
1281 	prison_hold_locked(pr);
1282 	mtx_unlock(&pr->pr_mtx);
1283 
1284 	/* Do not try to move the iface from and to the same prison. */
1285 	if (pr->pr_vnet == ifp->if_vnet) {
1286 		prison_free(pr);
1287 		return (EEXIST);
1288 	}
1289 
1290 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1291 	/* XXX Lock interfaces to avoid races. */
1292 	CURVNET_SET_QUIET(pr->pr_vnet);
1293 	difp = ifunit(ifname);
1294 	if (difp != NULL) {
1295 		CURVNET_RESTORE();
1296 		prison_free(pr);
1297 		return (EEXIST);
1298 	}
1299 
1300 	/* Make sure the VNET is stable. */
1301 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1302 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1303 	if (shutdown) {
1304 		CURVNET_RESTORE();
1305 		prison_free(pr);
1306 		return (EBUSY);
1307 	}
1308 	CURVNET_RESTORE();
1309 
1310 	/* Move the interface into the child jail/vnet. */
1311 	if_vmove(ifp, pr->pr_vnet);
1312 
1313 	/* Report the new if_xname back to the userland. */
1314 	sprintf(ifname, "%s", ifp->if_xname);
1315 
1316 	prison_free(pr);
1317 	return (0);
1318 }
1319 
1320 static int
1321 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1322 {
1323 	struct prison *pr;
1324 	struct vnet *vnet_dst;
1325 	struct ifnet *ifp;
1326  	int shutdown;
1327 
1328 	/* Try to find the prison within our visibility. */
1329 	sx_slock(&allprison_lock);
1330 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1331 	sx_sunlock(&allprison_lock);
1332 	if (pr == NULL)
1333 		return (ENXIO);
1334 	prison_hold_locked(pr);
1335 	mtx_unlock(&pr->pr_mtx);
1336 
1337 	/* Make sure the named iface exists in the source prison/vnet. */
1338 	CURVNET_SET(pr->pr_vnet);
1339 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1340 	if (ifp == NULL) {
1341 		CURVNET_RESTORE();
1342 		prison_free(pr);
1343 		return (ENXIO);
1344 	}
1345 
1346 	/* Do not try to move the iface from and to the same prison. */
1347 	vnet_dst = TD_TO_VNET(td);
1348 	if (vnet_dst == ifp->if_vnet) {
1349 		CURVNET_RESTORE();
1350 		prison_free(pr);
1351 		return (EEXIST);
1352 	}
1353 
1354 	/* Make sure the VNET is stable. */
1355 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1356 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1357 	if (shutdown) {
1358 		CURVNET_RESTORE();
1359 		prison_free(pr);
1360 		return (EBUSY);
1361 	}
1362 
1363 	/* Get interface back from child jail/vnet. */
1364 	if_vmove(ifp, vnet_dst);
1365 	CURVNET_RESTORE();
1366 
1367 	/* Report the new if_xname back to the userland. */
1368 	sprintf(ifname, "%s", ifp->if_xname);
1369 
1370 	prison_free(pr);
1371 	return (0);
1372 }
1373 #endif /* VIMAGE */
1374 
1375 /*
1376  * Add a group to an interface
1377  */
1378 int
1379 if_addgroup(struct ifnet *ifp, const char *groupname)
1380 {
1381 	struct ifg_list		*ifgl;
1382 	struct ifg_group	*ifg = NULL;
1383 	struct ifg_member	*ifgm;
1384 	int 			 new = 0;
1385 
1386 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1387 	    groupname[strlen(groupname) - 1] <= '9')
1388 		return (EINVAL);
1389 
1390 	IFNET_WLOCK();
1391 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1392 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1393 			IFNET_WUNLOCK();
1394 			return (EEXIST);
1395 		}
1396 
1397 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1398 	    M_NOWAIT)) == NULL) {
1399 	    	IFNET_WUNLOCK();
1400 		return (ENOMEM);
1401 	}
1402 
1403 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1404 	    M_TEMP, M_NOWAIT)) == NULL) {
1405 		free(ifgl, M_TEMP);
1406 		IFNET_WUNLOCK();
1407 		return (ENOMEM);
1408 	}
1409 
1410 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1411 		if (!strcmp(ifg->ifg_group, groupname))
1412 			break;
1413 
1414 	if (ifg == NULL) {
1415 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1416 		    M_TEMP, M_NOWAIT)) == NULL) {
1417 			free(ifgl, M_TEMP);
1418 			free(ifgm, M_TEMP);
1419 			IFNET_WUNLOCK();
1420 			return (ENOMEM);
1421 		}
1422 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1423 		ifg->ifg_refcnt = 0;
1424 		TAILQ_INIT(&ifg->ifg_members);
1425 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1426 		new = 1;
1427 	}
1428 
1429 	ifg->ifg_refcnt++;
1430 	ifgl->ifgl_group = ifg;
1431 	ifgm->ifgm_ifp = ifp;
1432 
1433 	IF_ADDR_WLOCK(ifp);
1434 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1435 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1436 	IF_ADDR_WUNLOCK(ifp);
1437 
1438 	IFNET_WUNLOCK();
1439 
1440 	if (new)
1441 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1442 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1443 
1444 	return (0);
1445 }
1446 
1447 /*
1448  * Remove a group from an interface
1449  */
1450 int
1451 if_delgroup(struct ifnet *ifp, const char *groupname)
1452 {
1453 	struct ifg_list		*ifgl;
1454 	struct ifg_member	*ifgm;
1455 
1456 	IFNET_WLOCK();
1457 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1458 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1459 			break;
1460 	if (ifgl == NULL) {
1461 		IFNET_WUNLOCK();
1462 		return (ENOENT);
1463 	}
1464 
1465 	IF_ADDR_WLOCK(ifp);
1466 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1467 	IF_ADDR_WUNLOCK(ifp);
1468 
1469 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1470 		if (ifgm->ifgm_ifp == ifp)
1471 			break;
1472 
1473 	if (ifgm != NULL) {
1474 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1475 		free(ifgm, M_TEMP);
1476 	}
1477 
1478 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1479 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1480 		IFNET_WUNLOCK();
1481 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1482 		free(ifgl->ifgl_group, M_TEMP);
1483 	} else
1484 		IFNET_WUNLOCK();
1485 
1486 	free(ifgl, M_TEMP);
1487 
1488 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1489 
1490 	return (0);
1491 }
1492 
1493 /*
1494  * Remove an interface from all groups
1495  */
1496 static void
1497 if_delgroups(struct ifnet *ifp)
1498 {
1499 	struct ifg_list		*ifgl;
1500 	struct ifg_member	*ifgm;
1501 	char groupname[IFNAMSIZ];
1502 
1503 	IFNET_WLOCK();
1504 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1505 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1506 
1507 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1508 
1509 		IF_ADDR_WLOCK(ifp);
1510 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1511 		IF_ADDR_WUNLOCK(ifp);
1512 
1513 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1514 			if (ifgm->ifgm_ifp == ifp)
1515 				break;
1516 
1517 		if (ifgm != NULL) {
1518 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1519 			    ifgm_next);
1520 			free(ifgm, M_TEMP);
1521 		}
1522 
1523 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1524 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1525 			IFNET_WUNLOCK();
1526 			EVENTHANDLER_INVOKE(group_detach_event,
1527 			    ifgl->ifgl_group);
1528 			free(ifgl->ifgl_group, M_TEMP);
1529 		} else
1530 			IFNET_WUNLOCK();
1531 
1532 		free(ifgl, M_TEMP);
1533 
1534 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1535 
1536 		IFNET_WLOCK();
1537 	}
1538 	IFNET_WUNLOCK();
1539 }
1540 
1541 static char *
1542 ifgr_group_get(void *ifgrp)
1543 {
1544 	union ifgroupreq_union *ifgrup;
1545 
1546 	ifgrup = ifgrp;
1547 #ifdef COMPAT_FREEBSD32
1548 	if (SV_CURPROC_FLAG(SV_ILP32))
1549 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1550 #endif
1551 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1552 }
1553 
1554 static struct ifg_req *
1555 ifgr_groups_get(void *ifgrp)
1556 {
1557 	union ifgroupreq_union *ifgrup;
1558 
1559 	ifgrup = ifgrp;
1560 #ifdef COMPAT_FREEBSD32
1561 	if (SV_CURPROC_FLAG(SV_ILP32))
1562 		return ((struct ifg_req *)(uintptr_t)
1563 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1564 #endif
1565 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1566 }
1567 
1568 /*
1569  * Stores all groups from an interface in memory pointed to by ifgr.
1570  */
1571 static int
1572 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1573 {
1574 	int			 len, error;
1575 	struct ifg_list		*ifgl;
1576 	struct ifg_req		 ifgrq, *ifgp;
1577 
1578 	if (ifgr->ifgr_len == 0) {
1579 		IF_ADDR_RLOCK(ifp);
1580 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1581 			ifgr->ifgr_len += sizeof(struct ifg_req);
1582 		IF_ADDR_RUNLOCK(ifp);
1583 		return (0);
1584 	}
1585 
1586 	len = ifgr->ifgr_len;
1587 	ifgp = ifgr_groups_get(ifgr);
1588 	/* XXX: wire */
1589 	IF_ADDR_RLOCK(ifp);
1590 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1591 		if (len < sizeof(ifgrq)) {
1592 			IF_ADDR_RUNLOCK(ifp);
1593 			return (EINVAL);
1594 		}
1595 		bzero(&ifgrq, sizeof ifgrq);
1596 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1597 		    sizeof(ifgrq.ifgrq_group));
1598 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1599 		    	IF_ADDR_RUNLOCK(ifp);
1600 			return (error);
1601 		}
1602 		len -= sizeof(ifgrq);
1603 		ifgp++;
1604 	}
1605 	IF_ADDR_RUNLOCK(ifp);
1606 
1607 	return (0);
1608 }
1609 
1610 /*
1611  * Stores all members of a group in memory pointed to by igfr
1612  */
1613 static int
1614 if_getgroupmembers(struct ifgroupreq *ifgr)
1615 {
1616 	struct ifg_group	*ifg;
1617 	struct ifg_member	*ifgm;
1618 	struct ifg_req		 ifgrq, *ifgp;
1619 	int			 len, error;
1620 
1621 	IFNET_RLOCK();
1622 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1623 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1624 			break;
1625 	if (ifg == NULL) {
1626 		IFNET_RUNLOCK();
1627 		return (ENOENT);
1628 	}
1629 
1630 	if (ifgr->ifgr_len == 0) {
1631 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1632 			ifgr->ifgr_len += sizeof(ifgrq);
1633 		IFNET_RUNLOCK();
1634 		return (0);
1635 	}
1636 
1637 	len = ifgr->ifgr_len;
1638 	ifgp = ifgr_groups_get(ifgr);
1639 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1640 		if (len < sizeof(ifgrq)) {
1641 			IFNET_RUNLOCK();
1642 			return (EINVAL);
1643 		}
1644 		bzero(&ifgrq, sizeof ifgrq);
1645 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1646 		    sizeof(ifgrq.ifgrq_member));
1647 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1648 			IFNET_RUNLOCK();
1649 			return (error);
1650 		}
1651 		len -= sizeof(ifgrq);
1652 		ifgp++;
1653 	}
1654 	IFNET_RUNLOCK();
1655 
1656 	return (0);
1657 }
1658 
1659 /*
1660  * Return counter values from counter(9)s stored in ifnet.
1661  */
1662 uint64_t
1663 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1664 {
1665 
1666 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1667 
1668 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1669 }
1670 
1671 /*
1672  * Increase an ifnet counter. Usually used for counters shared
1673  * between the stack and a driver, but function supports them all.
1674  */
1675 void
1676 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1677 {
1678 
1679 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1680 
1681 	counter_u64_add(ifp->if_counters[cnt], inc);
1682 }
1683 
1684 /*
1685  * Copy data from ifnet to userland API structure if_data.
1686  */
1687 void
1688 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1689 {
1690 
1691 	ifd->ifi_type = ifp->if_type;
1692 	ifd->ifi_physical = 0;
1693 	ifd->ifi_addrlen = ifp->if_addrlen;
1694 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1695 	ifd->ifi_link_state = ifp->if_link_state;
1696 	ifd->ifi_vhid = 0;
1697 	ifd->ifi_datalen = sizeof(struct if_data);
1698 	ifd->ifi_mtu = ifp->if_mtu;
1699 	ifd->ifi_metric = ifp->if_metric;
1700 	ifd->ifi_baudrate = ifp->if_baudrate;
1701 	ifd->ifi_hwassist = ifp->if_hwassist;
1702 	ifd->ifi_epoch = ifp->if_epoch;
1703 	ifd->ifi_lastchange = ifp->if_lastchange;
1704 
1705 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1706 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1707 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1708 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1709 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1710 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1711 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1712 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1713 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1714 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1715 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1716 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1717 }
1718 
1719 /*
1720  * Wrapper functions for struct ifnet address list locking macros.  These are
1721  * used by kernel modules to avoid encoding programming interface or binary
1722  * interface assumptions that may be violated when kernel-internal locking
1723  * approaches change.
1724  */
1725 void
1726 if_addr_rlock(struct ifnet *ifp)
1727 {
1728 
1729 	IF_ADDR_RLOCK(ifp);
1730 }
1731 
1732 void
1733 if_addr_runlock(struct ifnet *ifp)
1734 {
1735 
1736 	IF_ADDR_RUNLOCK(ifp);
1737 }
1738 
1739 void
1740 if_maddr_rlock(if_t ifp)
1741 {
1742 
1743 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1744 }
1745 
1746 void
1747 if_maddr_runlock(if_t ifp)
1748 {
1749 
1750 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1751 }
1752 
1753 /*
1754  * Initialization, destruction and refcounting functions for ifaddrs.
1755  */
1756 struct ifaddr *
1757 ifa_alloc(size_t size, int flags)
1758 {
1759 	struct ifaddr *ifa;
1760 
1761 	KASSERT(size >= sizeof(struct ifaddr),
1762 	    ("%s: invalid size %zu", __func__, size));
1763 
1764 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1765 	if (ifa == NULL)
1766 		return (NULL);
1767 
1768 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1769 		goto fail;
1770 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1771 		goto fail;
1772 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1773 		goto fail;
1774 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1775 		goto fail;
1776 
1777 	refcount_init(&ifa->ifa_refcnt, 1);
1778 
1779 	return (ifa);
1780 
1781 fail:
1782 	/* free(NULL) is okay */
1783 	counter_u64_free(ifa->ifa_opackets);
1784 	counter_u64_free(ifa->ifa_ipackets);
1785 	counter_u64_free(ifa->ifa_obytes);
1786 	counter_u64_free(ifa->ifa_ibytes);
1787 	free(ifa, M_IFADDR);
1788 
1789 	return (NULL);
1790 }
1791 
1792 void
1793 ifa_ref(struct ifaddr *ifa)
1794 {
1795 
1796 	refcount_acquire(&ifa->ifa_refcnt);
1797 }
1798 
1799 void
1800 ifa_free(struct ifaddr *ifa)
1801 {
1802 
1803 	if (refcount_release(&ifa->ifa_refcnt)) {
1804 		counter_u64_free(ifa->ifa_opackets);
1805 		counter_u64_free(ifa->ifa_ipackets);
1806 		counter_u64_free(ifa->ifa_obytes);
1807 		counter_u64_free(ifa->ifa_ibytes);
1808 		free(ifa, M_IFADDR);
1809 	}
1810 }
1811 
1812 static int
1813 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1814     struct sockaddr *ia)
1815 {
1816 	int error;
1817 	struct rt_addrinfo info;
1818 	struct sockaddr_dl null_sdl;
1819 	struct ifnet *ifp;
1820 
1821 	ifp = ifa->ifa_ifp;
1822 
1823 	bzero(&info, sizeof(info));
1824 	if (cmd != RTM_DELETE)
1825 		info.rti_ifp = V_loif;
1826 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1827 	info.rti_info[RTAX_DST] = ia;
1828 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1829 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1830 
1831 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1832 
1833 	if (error != 0)
1834 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1835 		    __func__, otype, if_name(ifp), error);
1836 
1837 	return (error);
1838 }
1839 
1840 int
1841 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1842 {
1843 
1844 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1845 }
1846 
1847 int
1848 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1849 {
1850 
1851 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1852 }
1853 
1854 int
1855 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1856 {
1857 
1858 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1859 }
1860 
1861 /*
1862  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1863  * structs used to represent other address families, it is necessary
1864  * to perform a different comparison.
1865  */
1866 
1867 #define	sa_dl_equal(a1, a2)	\
1868 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1869 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1870 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1871 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1872 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1873 
1874 /*
1875  * Locate an interface based on a complete address.
1876  */
1877 /*ARGSUSED*/
1878 static struct ifaddr *
1879 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1880 {
1881 	struct ifnet *ifp;
1882 	struct ifaddr *ifa;
1883 
1884 	IFNET_RLOCK_NOSLEEP();
1885 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1886 		IF_ADDR_RLOCK(ifp);
1887 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1888 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1889 				continue;
1890 			if (sa_equal(addr, ifa->ifa_addr)) {
1891 				if (getref)
1892 					ifa_ref(ifa);
1893 				IF_ADDR_RUNLOCK(ifp);
1894 				goto done;
1895 			}
1896 			/* IP6 doesn't have broadcast */
1897 			if ((ifp->if_flags & IFF_BROADCAST) &&
1898 			    ifa->ifa_broadaddr &&
1899 			    ifa->ifa_broadaddr->sa_len != 0 &&
1900 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1901 				if (getref)
1902 					ifa_ref(ifa);
1903 				IF_ADDR_RUNLOCK(ifp);
1904 				goto done;
1905 			}
1906 		}
1907 		IF_ADDR_RUNLOCK(ifp);
1908 	}
1909 	ifa = NULL;
1910 done:
1911 	IFNET_RUNLOCK_NOSLEEP();
1912 	return (ifa);
1913 }
1914 
1915 struct ifaddr *
1916 ifa_ifwithaddr(const struct sockaddr *addr)
1917 {
1918 
1919 	return (ifa_ifwithaddr_internal(addr, 1));
1920 }
1921 
1922 int
1923 ifa_ifwithaddr_check(const struct sockaddr *addr)
1924 {
1925 
1926 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1927 }
1928 
1929 /*
1930  * Locate an interface based on the broadcast address.
1931  */
1932 /* ARGSUSED */
1933 struct ifaddr *
1934 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1935 {
1936 	struct ifnet *ifp;
1937 	struct ifaddr *ifa;
1938 
1939 	IFNET_RLOCK_NOSLEEP();
1940 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1941 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1942 			continue;
1943 		IF_ADDR_RLOCK(ifp);
1944 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1945 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1946 				continue;
1947 			if ((ifp->if_flags & IFF_BROADCAST) &&
1948 			    ifa->ifa_broadaddr &&
1949 			    ifa->ifa_broadaddr->sa_len != 0 &&
1950 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1951 				ifa_ref(ifa);
1952 				IF_ADDR_RUNLOCK(ifp);
1953 				goto done;
1954 			}
1955 		}
1956 		IF_ADDR_RUNLOCK(ifp);
1957 	}
1958 	ifa = NULL;
1959 done:
1960 	IFNET_RUNLOCK_NOSLEEP();
1961 	return (ifa);
1962 }
1963 
1964 /*
1965  * Locate the point to point interface with a given destination address.
1966  */
1967 /*ARGSUSED*/
1968 struct ifaddr *
1969 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1970 {
1971 	struct ifnet *ifp;
1972 	struct ifaddr *ifa;
1973 
1974 	IFNET_RLOCK_NOSLEEP();
1975 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1976 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1977 			continue;
1978 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1979 			continue;
1980 		IF_ADDR_RLOCK(ifp);
1981 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1982 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1983 				continue;
1984 			if (ifa->ifa_dstaddr != NULL &&
1985 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1986 				ifa_ref(ifa);
1987 				IF_ADDR_RUNLOCK(ifp);
1988 				goto done;
1989 			}
1990 		}
1991 		IF_ADDR_RUNLOCK(ifp);
1992 	}
1993 	ifa = NULL;
1994 done:
1995 	IFNET_RUNLOCK_NOSLEEP();
1996 	return (ifa);
1997 }
1998 
1999 /*
2000  * Find an interface on a specific network.  If many, choice
2001  * is most specific found.
2002  */
2003 struct ifaddr *
2004 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2005 {
2006 	struct ifnet *ifp;
2007 	struct ifaddr *ifa;
2008 	struct ifaddr *ifa_maybe = NULL;
2009 	u_int af = addr->sa_family;
2010 	const char *addr_data = addr->sa_data, *cplim;
2011 
2012 	/*
2013 	 * AF_LINK addresses can be looked up directly by their index number,
2014 	 * so do that if we can.
2015 	 */
2016 	if (af == AF_LINK) {
2017 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2018 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2019 		return (ifaddr_byindex(sdl->sdl_index));
2020 	}
2021 
2022 	/*
2023 	 * Scan though each interface, looking for ones that have addresses
2024 	 * in this address family and the requested fib.  Maintain a reference
2025 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2026 	 * kept it stable when we move onto the next interface.
2027 	 */
2028 	IFNET_RLOCK_NOSLEEP();
2029 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2030 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2031 			continue;
2032 		IF_ADDR_RLOCK(ifp);
2033 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2034 			const char *cp, *cp2, *cp3;
2035 
2036 			if (ifa->ifa_addr->sa_family != af)
2037 next:				continue;
2038 			if (af == AF_INET &&
2039 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2040 				/*
2041 				 * This is a bit broken as it doesn't
2042 				 * take into account that the remote end may
2043 				 * be a single node in the network we are
2044 				 * looking for.
2045 				 * The trouble is that we don't know the
2046 				 * netmask for the remote end.
2047 				 */
2048 				if (ifa->ifa_dstaddr != NULL &&
2049 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2050 					ifa_ref(ifa);
2051 					IF_ADDR_RUNLOCK(ifp);
2052 					goto done;
2053 				}
2054 			} else {
2055 				/*
2056 				 * Scan all the bits in the ifa's address.
2057 				 * If a bit dissagrees with what we are
2058 				 * looking for, mask it with the netmask
2059 				 * to see if it really matters.
2060 				 * (A byte at a time)
2061 				 */
2062 				if (ifa->ifa_netmask == 0)
2063 					continue;
2064 				cp = addr_data;
2065 				cp2 = ifa->ifa_addr->sa_data;
2066 				cp3 = ifa->ifa_netmask->sa_data;
2067 				cplim = ifa->ifa_netmask->sa_len
2068 					+ (char *)ifa->ifa_netmask;
2069 				while (cp3 < cplim)
2070 					if ((*cp++ ^ *cp2++) & *cp3++)
2071 						goto next; /* next address! */
2072 				/*
2073 				 * If the netmask of what we just found
2074 				 * is more specific than what we had before
2075 				 * (if we had one), or if the virtual status
2076 				 * of new prefix is better than of the old one,
2077 				 * then remember the new one before continuing
2078 				 * to search for an even better one.
2079 				 */
2080 				if (ifa_maybe == NULL ||
2081 				    ifa_preferred(ifa_maybe, ifa) ||
2082 				    rn_refines((caddr_t)ifa->ifa_netmask,
2083 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2084 					if (ifa_maybe != NULL)
2085 						ifa_free(ifa_maybe);
2086 					ifa_maybe = ifa;
2087 					ifa_ref(ifa_maybe);
2088 				}
2089 			}
2090 		}
2091 		IF_ADDR_RUNLOCK(ifp);
2092 	}
2093 	ifa = ifa_maybe;
2094 	ifa_maybe = NULL;
2095 done:
2096 	IFNET_RUNLOCK_NOSLEEP();
2097 	if (ifa_maybe != NULL)
2098 		ifa_free(ifa_maybe);
2099 	return (ifa);
2100 }
2101 
2102 /*
2103  * Find an interface address specific to an interface best matching
2104  * a given address.
2105  */
2106 struct ifaddr *
2107 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2108 {
2109 	struct ifaddr *ifa;
2110 	const char *cp, *cp2, *cp3;
2111 	char *cplim;
2112 	struct ifaddr *ifa_maybe = NULL;
2113 	u_int af = addr->sa_family;
2114 
2115 	if (af >= AF_MAX)
2116 		return (NULL);
2117 	IF_ADDR_RLOCK(ifp);
2118 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2119 		if (ifa->ifa_addr->sa_family != af)
2120 			continue;
2121 		if (ifa_maybe == NULL)
2122 			ifa_maybe = ifa;
2123 		if (ifa->ifa_netmask == 0) {
2124 			if (sa_equal(addr, ifa->ifa_addr) ||
2125 			    (ifa->ifa_dstaddr &&
2126 			    sa_equal(addr, ifa->ifa_dstaddr)))
2127 				goto done;
2128 			continue;
2129 		}
2130 		if (ifp->if_flags & IFF_POINTOPOINT) {
2131 			if (sa_equal(addr, ifa->ifa_dstaddr))
2132 				goto done;
2133 		} else {
2134 			cp = addr->sa_data;
2135 			cp2 = ifa->ifa_addr->sa_data;
2136 			cp3 = ifa->ifa_netmask->sa_data;
2137 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2138 			for (; cp3 < cplim; cp3++)
2139 				if ((*cp++ ^ *cp2++) & *cp3)
2140 					break;
2141 			if (cp3 == cplim)
2142 				goto done;
2143 		}
2144 	}
2145 	ifa = ifa_maybe;
2146 done:
2147 	if (ifa != NULL)
2148 		ifa_ref(ifa);
2149 	IF_ADDR_RUNLOCK(ifp);
2150 	return (ifa);
2151 }
2152 
2153 /*
2154  * See whether new ifa is better than current one:
2155  * 1) A non-virtual one is preferred over virtual.
2156  * 2) A virtual in master state preferred over any other state.
2157  *
2158  * Used in several address selecting functions.
2159  */
2160 int
2161 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2162 {
2163 
2164 	return (cur->ifa_carp && (!next->ifa_carp ||
2165 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2166 }
2167 
2168 #include <net/if_llatbl.h>
2169 
2170 /*
2171  * Default action when installing a route with a Link Level gateway.
2172  * Lookup an appropriate real ifa to point to.
2173  * This should be moved to /sys/net/link.c eventually.
2174  */
2175 static void
2176 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2177 {
2178 	struct ifaddr *ifa, *oifa;
2179 	struct sockaddr *dst;
2180 	struct ifnet *ifp;
2181 
2182 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2183 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2184 		return;
2185 	ifa = ifaof_ifpforaddr(dst, ifp);
2186 	if (ifa) {
2187 		oifa = rt->rt_ifa;
2188 		rt->rt_ifa = ifa;
2189 		ifa_free(oifa);
2190 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2191 			ifa->ifa_rtrequest(cmd, rt, info);
2192 	}
2193 }
2194 
2195 struct sockaddr_dl *
2196 link_alloc_sdl(size_t size, int flags)
2197 {
2198 
2199 	return (malloc(size, M_TEMP, flags));
2200 }
2201 
2202 void
2203 link_free_sdl(struct sockaddr *sa)
2204 {
2205 	free(sa, M_TEMP);
2206 }
2207 
2208 /*
2209  * Fills in given sdl with interface basic info.
2210  * Returns pointer to filled sdl.
2211  */
2212 struct sockaddr_dl *
2213 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2214 {
2215 	struct sockaddr_dl *sdl;
2216 
2217 	sdl = (struct sockaddr_dl *)paddr;
2218 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2219 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2220 	sdl->sdl_family = AF_LINK;
2221 	sdl->sdl_index = ifp->if_index;
2222 	sdl->sdl_type = iftype;
2223 
2224 	return (sdl);
2225 }
2226 
2227 /*
2228  * Mark an interface down and notify protocols of
2229  * the transition.
2230  */
2231 static void
2232 if_unroute(struct ifnet *ifp, int flag, int fam)
2233 {
2234 	struct ifaddr *ifa;
2235 
2236 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2237 
2238 	ifp->if_flags &= ~flag;
2239 	getmicrotime(&ifp->if_lastchange);
2240 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2241 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2242 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2243 	ifp->if_qflush(ifp);
2244 
2245 	if (ifp->if_carp)
2246 		(*carp_linkstate_p)(ifp);
2247 	rt_ifmsg(ifp);
2248 }
2249 
2250 /*
2251  * Mark an interface up and notify protocols of
2252  * the transition.
2253  */
2254 static void
2255 if_route(struct ifnet *ifp, int flag, int fam)
2256 {
2257 	struct ifaddr *ifa;
2258 
2259 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2260 
2261 	ifp->if_flags |= flag;
2262 	getmicrotime(&ifp->if_lastchange);
2263 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2264 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2265 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2266 	if (ifp->if_carp)
2267 		(*carp_linkstate_p)(ifp);
2268 	rt_ifmsg(ifp);
2269 #ifdef INET6
2270 	in6_if_up(ifp);
2271 #endif
2272 }
2273 
2274 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2275 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2276 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2277 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2278 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2279 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2280 void	*(*vlan_cookie_p)(struct ifnet *);
2281 
2282 /*
2283  * Handle a change in the interface link state. To avoid LORs
2284  * between driver lock and upper layer locks, as well as possible
2285  * recursions, we post event to taskqueue, and all job
2286  * is done in static do_link_state_change().
2287  */
2288 void
2289 if_link_state_change(struct ifnet *ifp, int link_state)
2290 {
2291 	/* Return if state hasn't changed. */
2292 	if (ifp->if_link_state == link_state)
2293 		return;
2294 
2295 	ifp->if_link_state = link_state;
2296 
2297 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2298 }
2299 
2300 static void
2301 do_link_state_change(void *arg, int pending)
2302 {
2303 	struct ifnet *ifp = (struct ifnet *)arg;
2304 	int link_state = ifp->if_link_state;
2305 	CURVNET_SET(ifp->if_vnet);
2306 
2307 	/* Notify that the link state has changed. */
2308 	rt_ifmsg(ifp);
2309 	if (ifp->if_vlantrunk != NULL)
2310 		(*vlan_link_state_p)(ifp);
2311 
2312 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2313 	    ifp->if_l2com != NULL)
2314 		(*ng_ether_link_state_p)(ifp, link_state);
2315 	if (ifp->if_carp)
2316 		(*carp_linkstate_p)(ifp);
2317 	if (ifp->if_bridge)
2318 		(*bridge_linkstate_p)(ifp);
2319 	if (ifp->if_lagg)
2320 		(*lagg_linkstate_p)(ifp, link_state);
2321 
2322 	if (IS_DEFAULT_VNET(curvnet))
2323 		devctl_notify("IFNET", ifp->if_xname,
2324 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2325 		    NULL);
2326 	if (pending > 1)
2327 		if_printf(ifp, "%d link states coalesced\n", pending);
2328 	if (log_link_state_change)
2329 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2330 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2331 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2332 	CURVNET_RESTORE();
2333 }
2334 
2335 /*
2336  * Mark an interface down and notify protocols of
2337  * the transition.
2338  */
2339 void
2340 if_down(struct ifnet *ifp)
2341 {
2342 
2343 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2344 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2345 }
2346 
2347 /*
2348  * Mark an interface up and notify protocols of
2349  * the transition.
2350  */
2351 void
2352 if_up(struct ifnet *ifp)
2353 {
2354 
2355 	if_route(ifp, IFF_UP, AF_UNSPEC);
2356 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2357 }
2358 
2359 /*
2360  * Flush an interface queue.
2361  */
2362 void
2363 if_qflush(struct ifnet *ifp)
2364 {
2365 	struct mbuf *m, *n;
2366 	struct ifaltq *ifq;
2367 
2368 	ifq = &ifp->if_snd;
2369 	IFQ_LOCK(ifq);
2370 #ifdef ALTQ
2371 	if (ALTQ_IS_ENABLED(ifq))
2372 		ALTQ_PURGE(ifq);
2373 #endif
2374 	n = ifq->ifq_head;
2375 	while ((m = n) != NULL) {
2376 		n = m->m_nextpkt;
2377 		m_freem(m);
2378 	}
2379 	ifq->ifq_head = 0;
2380 	ifq->ifq_tail = 0;
2381 	ifq->ifq_len = 0;
2382 	IFQ_UNLOCK(ifq);
2383 }
2384 
2385 /*
2386  * Map interface name to interface structure pointer, with or without
2387  * returning a reference.
2388  */
2389 struct ifnet *
2390 ifunit_ref(const char *name)
2391 {
2392 	struct ifnet *ifp;
2393 
2394 	IFNET_RLOCK_NOSLEEP();
2395 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2396 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2397 		    !(ifp->if_flags & IFF_DYING))
2398 			break;
2399 	}
2400 	if (ifp != NULL)
2401 		if_ref(ifp);
2402 	IFNET_RUNLOCK_NOSLEEP();
2403 	return (ifp);
2404 }
2405 
2406 struct ifnet *
2407 ifunit(const char *name)
2408 {
2409 	struct ifnet *ifp;
2410 
2411 	IFNET_RLOCK_NOSLEEP();
2412 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2413 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2414 			break;
2415 	}
2416 	IFNET_RUNLOCK_NOSLEEP();
2417 	return (ifp);
2418 }
2419 
2420 static void *
2421 ifr_buffer_get_buffer(void *data)
2422 {
2423 	union ifreq_union *ifrup;
2424 
2425 	ifrup = data;
2426 #ifdef COMPAT_FREEBSD32
2427 	if (SV_CURPROC_FLAG(SV_ILP32))
2428 		return ((void *)(uintptr_t)
2429 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2430 #endif
2431 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2432 }
2433 
2434 static void
2435 ifr_buffer_set_buffer_null(void *data)
2436 {
2437 	union ifreq_union *ifrup;
2438 
2439 	ifrup = data;
2440 #ifdef COMPAT_FREEBSD32
2441 	if (SV_CURPROC_FLAG(SV_ILP32))
2442 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2443 	else
2444 #endif
2445 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2446 }
2447 
2448 static size_t
2449 ifr_buffer_get_length(void *data)
2450 {
2451 	union ifreq_union *ifrup;
2452 
2453 	ifrup = data;
2454 #ifdef COMPAT_FREEBSD32
2455 	if (SV_CURPROC_FLAG(SV_ILP32))
2456 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2457 #endif
2458 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2459 }
2460 
2461 static void
2462 ifr_buffer_set_length(void *data, size_t len)
2463 {
2464 	union ifreq_union *ifrup;
2465 
2466 	ifrup = data;
2467 #ifdef COMPAT_FREEBSD32
2468 	if (SV_CURPROC_FLAG(SV_ILP32))
2469 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2470 	else
2471 #endif
2472 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2473 }
2474 
2475 void *
2476 ifr_data_get_ptr(void *ifrp)
2477 {
2478 	union ifreq_union *ifrup;
2479 
2480 	ifrup = ifrp;
2481 #ifdef COMPAT_FREEBSD32
2482 	if (SV_CURPROC_FLAG(SV_ILP32))
2483 		return ((void *)(uintptr_t)
2484 		    ifrup->ifr32.ifr_ifru.ifru_data);
2485 #endif
2486 		return (ifrup->ifr.ifr_ifru.ifru_data);
2487 }
2488 
2489 /*
2490  * Hardware specific interface ioctls.
2491  */
2492 static int
2493 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2494 {
2495 	struct ifreq *ifr;
2496 	int error = 0, do_ifup = 0;
2497 	int new_flags, temp_flags;
2498 	size_t namelen, onamelen;
2499 	size_t descrlen;
2500 	char *descrbuf, *odescrbuf;
2501 	char new_name[IFNAMSIZ];
2502 	struct ifaddr *ifa;
2503 	struct sockaddr_dl *sdl;
2504 
2505 	ifr = (struct ifreq *)data;
2506 	switch (cmd) {
2507 	case SIOCGIFINDEX:
2508 		ifr->ifr_index = ifp->if_index;
2509 		break;
2510 
2511 	case SIOCGIFFLAGS:
2512 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2513 		ifr->ifr_flags = temp_flags & 0xffff;
2514 		ifr->ifr_flagshigh = temp_flags >> 16;
2515 		break;
2516 
2517 	case SIOCGIFCAP:
2518 		ifr->ifr_reqcap = ifp->if_capabilities;
2519 		ifr->ifr_curcap = ifp->if_capenable;
2520 		break;
2521 
2522 #ifdef MAC
2523 	case SIOCGIFMAC:
2524 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2525 		break;
2526 #endif
2527 
2528 	case SIOCGIFMETRIC:
2529 		ifr->ifr_metric = ifp->if_metric;
2530 		break;
2531 
2532 	case SIOCGIFMTU:
2533 		ifr->ifr_mtu = ifp->if_mtu;
2534 		break;
2535 
2536 	case SIOCGIFPHYS:
2537 		/* XXXGL: did this ever worked? */
2538 		ifr->ifr_phys = 0;
2539 		break;
2540 
2541 	case SIOCGIFDESCR:
2542 		error = 0;
2543 		sx_slock(&ifdescr_sx);
2544 		if (ifp->if_description == NULL)
2545 			error = ENOMSG;
2546 		else {
2547 			/* space for terminating nul */
2548 			descrlen = strlen(ifp->if_description) + 1;
2549 			if (ifr_buffer_get_length(ifr) < descrlen)
2550 				ifr_buffer_set_buffer_null(ifr);
2551 			else
2552 				error = copyout(ifp->if_description,
2553 				    ifr_buffer_get_buffer(ifr), descrlen);
2554 			ifr_buffer_set_length(ifr, descrlen);
2555 		}
2556 		sx_sunlock(&ifdescr_sx);
2557 		break;
2558 
2559 	case SIOCSIFDESCR:
2560 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2561 		if (error)
2562 			return (error);
2563 
2564 		/*
2565 		 * Copy only (length-1) bytes to make sure that
2566 		 * if_description is always nul terminated.  The
2567 		 * length parameter is supposed to count the
2568 		 * terminating nul in.
2569 		 */
2570 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2571 			return (ENAMETOOLONG);
2572 		else if (ifr_buffer_get_length(ifr) == 0)
2573 			descrbuf = NULL;
2574 		else {
2575 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2576 			    M_IFDESCR, M_WAITOK | M_ZERO);
2577 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2578 			    ifr_buffer_get_length(ifr) - 1);
2579 			if (error) {
2580 				free(descrbuf, M_IFDESCR);
2581 				break;
2582 			}
2583 		}
2584 
2585 		sx_xlock(&ifdescr_sx);
2586 		odescrbuf = ifp->if_description;
2587 		ifp->if_description = descrbuf;
2588 		sx_xunlock(&ifdescr_sx);
2589 
2590 		getmicrotime(&ifp->if_lastchange);
2591 		free(odescrbuf, M_IFDESCR);
2592 		break;
2593 
2594 	case SIOCGIFFIB:
2595 		ifr->ifr_fib = ifp->if_fib;
2596 		break;
2597 
2598 	case SIOCSIFFIB:
2599 		error = priv_check(td, PRIV_NET_SETIFFIB);
2600 		if (error)
2601 			return (error);
2602 		if (ifr->ifr_fib >= rt_numfibs)
2603 			return (EINVAL);
2604 
2605 		ifp->if_fib = ifr->ifr_fib;
2606 		break;
2607 
2608 	case SIOCSIFFLAGS:
2609 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2610 		if (error)
2611 			return (error);
2612 		/*
2613 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2614 		 * check, so we don't need special handling here yet.
2615 		 */
2616 		new_flags = (ifr->ifr_flags & 0xffff) |
2617 		    (ifr->ifr_flagshigh << 16);
2618 		if (ifp->if_flags & IFF_UP &&
2619 		    (new_flags & IFF_UP) == 0) {
2620 			if_down(ifp);
2621 		} else if (new_flags & IFF_UP &&
2622 		    (ifp->if_flags & IFF_UP) == 0) {
2623 			do_ifup = 1;
2624 		}
2625 		/* See if permanently promiscuous mode bit is about to flip */
2626 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2627 			if (new_flags & IFF_PPROMISC)
2628 				ifp->if_flags |= IFF_PROMISC;
2629 			else if (ifp->if_pcount == 0)
2630 				ifp->if_flags &= ~IFF_PROMISC;
2631 			if (log_promisc_mode_change)
2632                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2633                                     ifp->if_xname,
2634                                     ((new_flags & IFF_PPROMISC) ?
2635                                      "enabled" : "disabled"));
2636 		}
2637 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2638 			(new_flags &~ IFF_CANTCHANGE);
2639 		if (ifp->if_ioctl) {
2640 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2641 		}
2642 		if (do_ifup)
2643 			if_up(ifp);
2644 		getmicrotime(&ifp->if_lastchange);
2645 		break;
2646 
2647 	case SIOCSIFCAP:
2648 		error = priv_check(td, PRIV_NET_SETIFCAP);
2649 		if (error)
2650 			return (error);
2651 		if (ifp->if_ioctl == NULL)
2652 			return (EOPNOTSUPP);
2653 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2654 			return (EINVAL);
2655 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2656 		if (error == 0)
2657 			getmicrotime(&ifp->if_lastchange);
2658 		break;
2659 
2660 #ifdef MAC
2661 	case SIOCSIFMAC:
2662 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2663 		break;
2664 #endif
2665 
2666 	case SIOCSIFNAME:
2667 		error = priv_check(td, PRIV_NET_SETIFNAME);
2668 		if (error)
2669 			return (error);
2670 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2671 		    NULL);
2672 		if (error != 0)
2673 			return (error);
2674 		if (new_name[0] == '\0')
2675 			return (EINVAL);
2676 		if (new_name[IFNAMSIZ-1] != '\0') {
2677 			new_name[IFNAMSIZ-1] = '\0';
2678 			if (strlen(new_name) == IFNAMSIZ-1)
2679 				return (EINVAL);
2680 		}
2681 		if (ifunit(new_name) != NULL)
2682 			return (EEXIST);
2683 
2684 		/*
2685 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2686 		 * and there are numerous other races with the
2687 		 * ifunit() checks not being atomic with namespace
2688 		 * changes (renames, vmoves, if_attach, etc).
2689 		 */
2690 		ifp->if_flags |= IFF_RENAMING;
2691 
2692 		/* Announce the departure of the interface. */
2693 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2694 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2695 
2696 		log(LOG_INFO, "%s: changing name to '%s'\n",
2697 		    ifp->if_xname, new_name);
2698 
2699 		IF_ADDR_WLOCK(ifp);
2700 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2701 		ifa = ifp->if_addr;
2702 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2703 		namelen = strlen(new_name);
2704 		onamelen = sdl->sdl_nlen;
2705 		/*
2706 		 * Move the address if needed.  This is safe because we
2707 		 * allocate space for a name of length IFNAMSIZ when we
2708 		 * create this in if_attach().
2709 		 */
2710 		if (namelen != onamelen) {
2711 			bcopy(sdl->sdl_data + onamelen,
2712 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2713 		}
2714 		bcopy(new_name, sdl->sdl_data, namelen);
2715 		sdl->sdl_nlen = namelen;
2716 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2717 		bzero(sdl->sdl_data, onamelen);
2718 		while (namelen != 0)
2719 			sdl->sdl_data[--namelen] = 0xff;
2720 		IF_ADDR_WUNLOCK(ifp);
2721 
2722 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2723 		/* Announce the return of the interface. */
2724 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2725 
2726 		ifp->if_flags &= ~IFF_RENAMING;
2727 		break;
2728 
2729 #ifdef VIMAGE
2730 	case SIOCSIFVNET:
2731 		error = priv_check(td, PRIV_NET_SETIFVNET);
2732 		if (error)
2733 			return (error);
2734 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2735 		break;
2736 #endif
2737 
2738 	case SIOCSIFMETRIC:
2739 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2740 		if (error)
2741 			return (error);
2742 		ifp->if_metric = ifr->ifr_metric;
2743 		getmicrotime(&ifp->if_lastchange);
2744 		break;
2745 
2746 	case SIOCSIFPHYS:
2747 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2748 		if (error)
2749 			return (error);
2750 		if (ifp->if_ioctl == NULL)
2751 			return (EOPNOTSUPP);
2752 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2753 		if (error == 0)
2754 			getmicrotime(&ifp->if_lastchange);
2755 		break;
2756 
2757 	case SIOCSIFMTU:
2758 	{
2759 		u_long oldmtu = ifp->if_mtu;
2760 
2761 		error = priv_check(td, PRIV_NET_SETIFMTU);
2762 		if (error)
2763 			return (error);
2764 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2765 			return (EINVAL);
2766 		if (ifp->if_ioctl == NULL)
2767 			return (EOPNOTSUPP);
2768 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2769 		if (error == 0) {
2770 			getmicrotime(&ifp->if_lastchange);
2771 			rt_ifmsg(ifp);
2772 #ifdef INET
2773 			NETDUMP_REINIT(ifp);
2774 #endif
2775 		}
2776 		/*
2777 		 * If the link MTU changed, do network layer specific procedure.
2778 		 */
2779 		if (ifp->if_mtu != oldmtu) {
2780 #ifdef INET6
2781 			nd6_setmtu(ifp);
2782 #endif
2783 			rt_updatemtu(ifp);
2784 		}
2785 		break;
2786 	}
2787 
2788 	case SIOCADDMULTI:
2789 	case SIOCDELMULTI:
2790 		if (cmd == SIOCADDMULTI)
2791 			error = priv_check(td, PRIV_NET_ADDMULTI);
2792 		else
2793 			error = priv_check(td, PRIV_NET_DELMULTI);
2794 		if (error)
2795 			return (error);
2796 
2797 		/* Don't allow group membership on non-multicast interfaces. */
2798 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2799 			return (EOPNOTSUPP);
2800 
2801 		/* Don't let users screw up protocols' entries. */
2802 		if (ifr->ifr_addr.sa_family != AF_LINK)
2803 			return (EINVAL);
2804 
2805 		if (cmd == SIOCADDMULTI) {
2806 			struct ifmultiaddr *ifma;
2807 
2808 			/*
2809 			 * Userland is only permitted to join groups once
2810 			 * via the if_addmulti() KPI, because it cannot hold
2811 			 * struct ifmultiaddr * between calls. It may also
2812 			 * lose a race while we check if the membership
2813 			 * already exists.
2814 			 */
2815 			IF_ADDR_RLOCK(ifp);
2816 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2817 			IF_ADDR_RUNLOCK(ifp);
2818 			if (ifma != NULL)
2819 				error = EADDRINUSE;
2820 			else
2821 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2822 		} else {
2823 			error = if_delmulti(ifp, &ifr->ifr_addr);
2824 		}
2825 		if (error == 0)
2826 			getmicrotime(&ifp->if_lastchange);
2827 		break;
2828 
2829 	case SIOCSIFPHYADDR:
2830 	case SIOCDIFPHYADDR:
2831 #ifdef INET6
2832 	case SIOCSIFPHYADDR_IN6:
2833 #endif
2834 	case SIOCSIFMEDIA:
2835 	case SIOCSIFGENERIC:
2836 		error = priv_check(td, PRIV_NET_HWIOCTL);
2837 		if (error)
2838 			return (error);
2839 		if (ifp->if_ioctl == NULL)
2840 			return (EOPNOTSUPP);
2841 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2842 		if (error == 0)
2843 			getmicrotime(&ifp->if_lastchange);
2844 		break;
2845 
2846 	case SIOCGIFSTATUS:
2847 	case SIOCGIFPSRCADDR:
2848 	case SIOCGIFPDSTADDR:
2849 	case SIOCGIFMEDIA:
2850 	case SIOCGIFXMEDIA:
2851 	case SIOCGIFGENERIC:
2852 	case SIOCGIFRSSKEY:
2853 	case SIOCGIFRSSHASH:
2854 		if (ifp->if_ioctl == NULL)
2855 			return (EOPNOTSUPP);
2856 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2857 		break;
2858 
2859 	case SIOCSIFLLADDR:
2860 		error = priv_check(td, PRIV_NET_SETLLADDR);
2861 		if (error)
2862 			return (error);
2863 		error = if_setlladdr(ifp,
2864 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2865 		break;
2866 
2867 	case SIOCGHWADDR:
2868 		error = if_gethwaddr(ifp, ifr);
2869 		break;
2870 
2871 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2872 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2873 		if (error)
2874 			return (error);
2875 		if ((error = if_addgroup(ifp,
2876 		    ifgr_group_get((struct ifgroupreq *)data))))
2877 			return (error);
2878 		break;
2879 
2880 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2881 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2882 			return (error);
2883 		break;
2884 
2885 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2886 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2887 		if (error)
2888 			return (error);
2889 		if ((error = if_delgroup(ifp,
2890 		    ifgr_group_get((struct ifgroupreq *)data))))
2891 			return (error);
2892 		break;
2893 
2894 	default:
2895 		error = ENOIOCTL;
2896 		break;
2897 	}
2898 	return (error);
2899 }
2900 
2901 #ifdef COMPAT_FREEBSD32
2902 struct ifconf32 {
2903 	int32_t	ifc_len;
2904 	union {
2905 		uint32_t	ifcu_buf;
2906 		uint32_t	ifcu_req;
2907 	} ifc_ifcu;
2908 };
2909 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2910 #endif
2911 
2912 #ifdef COMPAT_FREEBSD32
2913 static void
2914 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2915 {
2916 	struct ifmediareq32 *ifmr32;
2917 
2918 	ifmr32 = (struct ifmediareq32 *)data;
2919 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2920 	    sizeof(ifmr->ifm_name));
2921 	ifmr->ifm_current = ifmr32->ifm_current;
2922 	ifmr->ifm_mask = ifmr32->ifm_mask;
2923 	ifmr->ifm_status = ifmr32->ifm_status;
2924 	ifmr->ifm_active = ifmr32->ifm_active;
2925 	ifmr->ifm_count = ifmr32->ifm_count;
2926 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2927 }
2928 
2929 static void
2930 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2931 {
2932 	struct ifmediareq32 *ifmr32;
2933 
2934 	ifmr32 = (struct ifmediareq32 *)data;
2935 	ifmr32->ifm_current = ifmr->ifm_current;
2936 	ifmr32->ifm_mask = ifmr->ifm_mask;
2937 	ifmr32->ifm_status = ifmr->ifm_status;
2938 	ifmr32->ifm_active = ifmr->ifm_active;
2939 	ifmr32->ifm_count = ifmr->ifm_count;
2940 }
2941 #endif
2942 
2943 /*
2944  * Interface ioctls.
2945  */
2946 int
2947 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2948 {
2949 #ifdef COMPAT_FREEBSD32
2950 	caddr_t saved_data;
2951 	struct ifmediareq ifmr;
2952 #endif
2953 	struct ifmediareq *ifmrp;
2954 	struct ifnet *ifp;
2955 	struct ifreq *ifr;
2956 	int error;
2957 	int oif_flags;
2958 #ifdef VIMAGE
2959 	int shutdown;
2960 #endif
2961 
2962 	CURVNET_SET(so->so_vnet);
2963 #ifdef VIMAGE
2964 	/* Make sure the VNET is stable. */
2965 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2966 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2967 	if (shutdown) {
2968 		CURVNET_RESTORE();
2969 		return (EBUSY);
2970 	}
2971 #endif
2972 
2973 
2974 	switch (cmd) {
2975 	case SIOCGIFCONF:
2976 		error = ifconf(cmd, data);
2977 		CURVNET_RESTORE();
2978 		return (error);
2979 
2980 #ifdef COMPAT_FREEBSD32
2981 	case SIOCGIFCONF32:
2982 		{
2983 			struct ifconf32 *ifc32;
2984 			struct ifconf ifc;
2985 
2986 			ifc32 = (struct ifconf32 *)data;
2987 			ifc.ifc_len = ifc32->ifc_len;
2988 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2989 
2990 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2991 			CURVNET_RESTORE();
2992 			if (error == 0)
2993 				ifc32->ifc_len = ifc.ifc_len;
2994 			return (error);
2995 		}
2996 #endif
2997 	}
2998 
2999 	ifmrp = NULL;
3000 #ifdef COMPAT_FREEBSD32
3001 	switch (cmd) {
3002 	case SIOCGIFMEDIA32:
3003 	case SIOCGIFXMEDIA32:
3004 		ifmrp = &ifmr;
3005 		ifmr_init(ifmrp, data);
3006 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3007 		saved_data = data;
3008 		data = (caddr_t)ifmrp;
3009 	}
3010 #endif
3011 
3012 	ifr = (struct ifreq *)data;
3013 	switch (cmd) {
3014 #ifdef VIMAGE
3015 	case SIOCSIFRVNET:
3016 		error = priv_check(td, PRIV_NET_SETIFVNET);
3017 		if (error == 0)
3018 			error = if_vmove_reclaim(td, ifr->ifr_name,
3019 			    ifr->ifr_jid);
3020 		goto out_noref;
3021 #endif
3022 	case SIOCIFCREATE:
3023 	case SIOCIFCREATE2:
3024 		error = priv_check(td, PRIV_NET_IFCREATE);
3025 		if (error == 0)
3026 			error = if_clone_create(ifr->ifr_name,
3027 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3028 			    ifr_data_get_ptr(ifr) : NULL);
3029 		goto out_noref;
3030 	case SIOCIFDESTROY:
3031 		error = priv_check(td, PRIV_NET_IFDESTROY);
3032 		if (error == 0)
3033 			error = if_clone_destroy(ifr->ifr_name);
3034 		goto out_noref;
3035 
3036 	case SIOCIFGCLONERS:
3037 		error = if_clone_list((struct if_clonereq *)data);
3038 		goto out_noref;
3039 
3040 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3041 		error = if_getgroupmembers((struct ifgroupreq *)data);
3042 		goto out_noref;
3043 
3044 #if defined(INET) || defined(INET6)
3045 	case SIOCSVH:
3046 	case SIOCGVH:
3047 		if (carp_ioctl_p == NULL)
3048 			error = EPROTONOSUPPORT;
3049 		else
3050 			error = (*carp_ioctl_p)(ifr, cmd, td);
3051 		goto out_noref;
3052 #endif
3053 	}
3054 
3055 	ifp = ifunit_ref(ifr->ifr_name);
3056 	if (ifp == NULL) {
3057 		error = ENXIO;
3058 		goto out_noref;
3059 	}
3060 
3061 	error = ifhwioctl(cmd, ifp, data, td);
3062 	if (error != ENOIOCTL)
3063 		goto out_ref;
3064 
3065 	oif_flags = ifp->if_flags;
3066 	if (so->so_proto == NULL) {
3067 		error = EOPNOTSUPP;
3068 		goto out_ref;
3069 	}
3070 
3071 	/*
3072 	 * Pass the request on to the socket control method, and if the
3073 	 * latter returns EOPNOTSUPP, directly to the interface.
3074 	 *
3075 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3076 	 * trust SIOCSIFADDR et al to come from an already privileged
3077 	 * layer, and do not perform any credentials checks or input
3078 	 * validation.
3079 	 */
3080 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3081 	    ifp, td));
3082 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3083 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3084 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3085 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3086 
3087 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3088 #ifdef INET6
3089 		if (ifp->if_flags & IFF_UP)
3090 			in6_if_up(ifp);
3091 #endif
3092 	}
3093 
3094 out_ref:
3095 	if_rele(ifp);
3096 out_noref:
3097 #ifdef COMPAT_FREEBSD32
3098 	if (ifmrp != NULL) {
3099 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3100 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3101 		     cmd));
3102 		data = saved_data;
3103 		ifmr_update(ifmrp, data);
3104 	}
3105 #endif
3106 	CURVNET_RESTORE();
3107 	return (error);
3108 }
3109 
3110 /*
3111  * The code common to handling reference counted flags,
3112  * e.g., in ifpromisc() and if_allmulti().
3113  * The "pflag" argument can specify a permanent mode flag to check,
3114  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3115  *
3116  * Only to be used on stack-owned flags, not driver-owned flags.
3117  */
3118 static int
3119 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3120 {
3121 	struct ifreq ifr;
3122 	int error;
3123 	int oldflags, oldcount;
3124 
3125 	/* Sanity checks to catch programming errors */
3126 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3127 	    ("%s: setting driver-owned flag %d", __func__, flag));
3128 
3129 	if (onswitch)
3130 		KASSERT(*refcount >= 0,
3131 		    ("%s: increment negative refcount %d for flag %d",
3132 		    __func__, *refcount, flag));
3133 	else
3134 		KASSERT(*refcount > 0,
3135 		    ("%s: decrement non-positive refcount %d for flag %d",
3136 		    __func__, *refcount, flag));
3137 
3138 	/* In case this mode is permanent, just touch refcount */
3139 	if (ifp->if_flags & pflag) {
3140 		*refcount += onswitch ? 1 : -1;
3141 		return (0);
3142 	}
3143 
3144 	/* Save ifnet parameters for if_ioctl() may fail */
3145 	oldcount = *refcount;
3146 	oldflags = ifp->if_flags;
3147 
3148 	/*
3149 	 * See if we aren't the only and touching refcount is enough.
3150 	 * Actually toggle interface flag if we are the first or last.
3151 	 */
3152 	if (onswitch) {
3153 		if ((*refcount)++)
3154 			return (0);
3155 		ifp->if_flags |= flag;
3156 	} else {
3157 		if (--(*refcount))
3158 			return (0);
3159 		ifp->if_flags &= ~flag;
3160 	}
3161 
3162 	/* Call down the driver since we've changed interface flags */
3163 	if (ifp->if_ioctl == NULL) {
3164 		error = EOPNOTSUPP;
3165 		goto recover;
3166 	}
3167 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3168 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3169 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3170 	if (error)
3171 		goto recover;
3172 	/* Notify userland that interface flags have changed */
3173 	rt_ifmsg(ifp);
3174 	return (0);
3175 
3176 recover:
3177 	/* Recover after driver error */
3178 	*refcount = oldcount;
3179 	ifp->if_flags = oldflags;
3180 	return (error);
3181 }
3182 
3183 /*
3184  * Set/clear promiscuous mode on interface ifp based on the truth value
3185  * of pswitch.  The calls are reference counted so that only the first
3186  * "on" request actually has an effect, as does the final "off" request.
3187  * Results are undefined if the "off" and "on" requests are not matched.
3188  */
3189 int
3190 ifpromisc(struct ifnet *ifp, int pswitch)
3191 {
3192 	int error;
3193 	int oldflags = ifp->if_flags;
3194 
3195 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3196 			   &ifp->if_pcount, pswitch);
3197 	/* If promiscuous mode status has changed, log a message */
3198 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3199             log_promisc_mode_change)
3200 		log(LOG_INFO, "%s: promiscuous mode %s\n",
3201 		    ifp->if_xname,
3202 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3203 	return (error);
3204 }
3205 
3206 /*
3207  * Return interface configuration
3208  * of system.  List may be used
3209  * in later ioctl's (above) to get
3210  * other information.
3211  */
3212 /*ARGSUSED*/
3213 static int
3214 ifconf(u_long cmd, caddr_t data)
3215 {
3216 	struct ifconf *ifc = (struct ifconf *)data;
3217 	struct ifnet *ifp;
3218 	struct ifaddr *ifa;
3219 	struct ifreq ifr;
3220 	struct sbuf *sb;
3221 	int error, full = 0, valid_len, max_len;
3222 
3223 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3224 	max_len = MAXPHYS - 1;
3225 
3226 	/* Prevent hostile input from being able to crash the system */
3227 	if (ifc->ifc_len <= 0)
3228 		return (EINVAL);
3229 
3230 again:
3231 	if (ifc->ifc_len <= max_len) {
3232 		max_len = ifc->ifc_len;
3233 		full = 1;
3234 	}
3235 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3236 	max_len = 0;
3237 	valid_len = 0;
3238 
3239 	IFNET_RLOCK();
3240 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3241 		int addrs;
3242 
3243 		/*
3244 		 * Zero the ifr to make sure we don't disclose the contents
3245 		 * of the stack.
3246 		 */
3247 		memset(&ifr, 0, sizeof(ifr));
3248 
3249 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3250 		    >= sizeof(ifr.ifr_name)) {
3251 			sbuf_delete(sb);
3252 			IFNET_RUNLOCK();
3253 			return (ENAMETOOLONG);
3254 		}
3255 
3256 		addrs = 0;
3257 		IF_ADDR_RLOCK(ifp);
3258 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3259 			struct sockaddr *sa = ifa->ifa_addr;
3260 
3261 			if (prison_if(curthread->td_ucred, sa) != 0)
3262 				continue;
3263 			addrs++;
3264 			if (sa->sa_len <= sizeof(*sa)) {
3265 				if (sa->sa_len < sizeof(*sa)) {
3266 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3267 					    sizeof(ifr.ifr_ifru.ifru_addr));
3268 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3269 					    sa->sa_len);
3270 				} else
3271 					ifr.ifr_ifru.ifru_addr = *sa;
3272 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3273 				max_len += sizeof(ifr);
3274 			} else {
3275 				sbuf_bcat(sb, &ifr,
3276 				    offsetof(struct ifreq, ifr_addr));
3277 				max_len += offsetof(struct ifreq, ifr_addr);
3278 				sbuf_bcat(sb, sa, sa->sa_len);
3279 				max_len += sa->sa_len;
3280 			}
3281 
3282 			if (sbuf_error(sb) == 0)
3283 				valid_len = sbuf_len(sb);
3284 		}
3285 		IF_ADDR_RUNLOCK(ifp);
3286 		if (addrs == 0) {
3287 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3288 			max_len += sizeof(ifr);
3289 
3290 			if (sbuf_error(sb) == 0)
3291 				valid_len = sbuf_len(sb);
3292 		}
3293 	}
3294 	IFNET_RUNLOCK();
3295 
3296 	/*
3297 	 * If we didn't allocate enough space (uncommon), try again.  If
3298 	 * we have already allocated as much space as we are allowed,
3299 	 * return what we've got.
3300 	 */
3301 	if (valid_len != max_len && !full) {
3302 		sbuf_delete(sb);
3303 		goto again;
3304 	}
3305 
3306 	ifc->ifc_len = valid_len;
3307 	sbuf_finish(sb);
3308 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3309 	sbuf_delete(sb);
3310 	return (error);
3311 }
3312 
3313 /*
3314  * Just like ifpromisc(), but for all-multicast-reception mode.
3315  */
3316 int
3317 if_allmulti(struct ifnet *ifp, int onswitch)
3318 {
3319 
3320 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3321 }
3322 
3323 struct ifmultiaddr *
3324 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3325 {
3326 	struct ifmultiaddr *ifma;
3327 
3328 	IF_ADDR_LOCK_ASSERT(ifp);
3329 
3330 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3331 		if (sa->sa_family == AF_LINK) {
3332 			if (sa_dl_equal(ifma->ifma_addr, sa))
3333 				break;
3334 		} else {
3335 			if (sa_equal(ifma->ifma_addr, sa))
3336 				break;
3337 		}
3338 	}
3339 
3340 	return ifma;
3341 }
3342 
3343 /*
3344  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3345  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3346  * the ifnet multicast address list here, so the caller must do that and
3347  * other setup work (such as notifying the device driver).  The reference
3348  * count is initialized to 1.
3349  */
3350 static struct ifmultiaddr *
3351 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3352     int mflags)
3353 {
3354 	struct ifmultiaddr *ifma;
3355 	struct sockaddr *dupsa;
3356 
3357 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3358 	    M_ZERO);
3359 	if (ifma == NULL)
3360 		return (NULL);
3361 
3362 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3363 	if (dupsa == NULL) {
3364 		free(ifma, M_IFMADDR);
3365 		return (NULL);
3366 	}
3367 	bcopy(sa, dupsa, sa->sa_len);
3368 	ifma->ifma_addr = dupsa;
3369 
3370 	ifma->ifma_ifp = ifp;
3371 	ifma->ifma_refcount = 1;
3372 	ifma->ifma_protospec = NULL;
3373 
3374 	if (llsa == NULL) {
3375 		ifma->ifma_lladdr = NULL;
3376 		return (ifma);
3377 	}
3378 
3379 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3380 	if (dupsa == NULL) {
3381 		free(ifma->ifma_addr, M_IFMADDR);
3382 		free(ifma, M_IFMADDR);
3383 		return (NULL);
3384 	}
3385 	bcopy(llsa, dupsa, llsa->sa_len);
3386 	ifma->ifma_lladdr = dupsa;
3387 
3388 	return (ifma);
3389 }
3390 
3391 /*
3392  * if_freemulti: free ifmultiaddr structure and possibly attached related
3393  * addresses.  The caller is responsible for implementing reference
3394  * counting, notifying the driver, handling routing messages, and releasing
3395  * any dependent link layer state.
3396  */
3397 #ifdef MCAST_VERBOSE
3398 extern void kdb_backtrace(void);
3399 #endif
3400 void
3401 if_freemulti(struct ifmultiaddr *ifma)
3402 {
3403 
3404 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3405 	    ifma->ifma_refcount));
3406 
3407 	if (ifma->ifma_lladdr != NULL)
3408 		free(ifma->ifma_lladdr, M_IFMADDR);
3409 #ifdef MCAST_VERBOSE
3410 	kdb_backtrace();
3411 	printf("%s freeing ifma: %p\n", __func__, ifma);
3412 #endif
3413 	free(ifma->ifma_addr, M_IFMADDR);
3414 	free(ifma, M_IFMADDR);
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 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3511 			    ifma_link);
3512 		} else
3513 			ll_ifma->ifma_refcount++;
3514 		ifma->ifma_llifma = ll_ifma;
3515 	}
3516 
3517 	/*
3518 	 * We now have a new multicast address, ifma, and possibly a new or
3519 	 * referenced link layer address.  Add the primary address to the
3520 	 * ifnet address list.
3521 	 */
3522 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3523 
3524 	if (retifma != NULL)
3525 		*retifma = ifma;
3526 
3527 	/*
3528 	 * Must generate the message while holding the lock so that 'ifma'
3529 	 * pointer is still valid.
3530 	 */
3531 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3532 	IF_ADDR_WUNLOCK(ifp);
3533 
3534 	/*
3535 	 * We are certain we have added something, so call down to the
3536 	 * interface to let them know about it.
3537 	 */
3538 	if (ifp->if_ioctl != NULL) {
3539 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3540 	}
3541 
3542 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3543 		link_free_sdl(llsa);
3544 
3545 	return (0);
3546 
3547 free_llsa_out:
3548 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3549 		link_free_sdl(llsa);
3550 
3551 unlock_out:
3552 	IF_ADDR_WUNLOCK(ifp);
3553 	return (error);
3554 }
3555 
3556 /*
3557  * Delete a multicast group membership by network-layer group address.
3558  *
3559  * Returns ENOENT if the entry could not be found. If ifp no longer
3560  * exists, results are undefined. This entry point should only be used
3561  * from subsystems which do appropriate locking to hold ifp for the
3562  * duration of the call.
3563  * Network-layer protocol domains must use if_delmulti_ifma().
3564  */
3565 int
3566 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3567 {
3568 	struct ifmultiaddr *ifma;
3569 	int lastref;
3570 #ifdef INVARIANTS
3571 	struct ifnet *oifp;
3572 
3573 	IFNET_RLOCK_NOSLEEP();
3574 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3575 		if (ifp == oifp)
3576 			break;
3577 	if (ifp != oifp)
3578 		ifp = NULL;
3579 	IFNET_RUNLOCK_NOSLEEP();
3580 
3581 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3582 #endif
3583 	if (ifp == NULL)
3584 		return (ENOENT);
3585 
3586 	IF_ADDR_WLOCK(ifp);
3587 	lastref = 0;
3588 	ifma = if_findmulti(ifp, sa);
3589 	if (ifma != NULL)
3590 		lastref = if_delmulti_locked(ifp, ifma, 0);
3591 	IF_ADDR_WUNLOCK(ifp);
3592 
3593 	if (ifma == NULL)
3594 		return (ENOENT);
3595 
3596 	if (lastref && ifp->if_ioctl != NULL) {
3597 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3598 	}
3599 
3600 	return (0);
3601 }
3602 
3603 /*
3604  * Delete all multicast group membership for an interface.
3605  * Should be used to quickly flush all multicast filters.
3606  */
3607 void
3608 if_delallmulti(struct ifnet *ifp)
3609 {
3610 	struct ifmultiaddr *ifma;
3611 	struct ifmultiaddr *next;
3612 
3613 	IF_ADDR_WLOCK(ifp);
3614 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3615 		if_delmulti_locked(ifp, ifma, 0);
3616 	IF_ADDR_WUNLOCK(ifp);
3617 }
3618 
3619 void
3620 if_delmulti_ifma(struct ifmultiaddr *ifma)
3621 {
3622 	if_delmulti_ifma_flags(ifma, 0);
3623 }
3624 
3625 /*
3626  * Delete a multicast group membership by group membership pointer.
3627  * Network-layer protocol domains must use this routine.
3628  *
3629  * It is safe to call this routine if the ifp disappeared.
3630  */
3631 void
3632 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3633 {
3634 	struct ifnet *ifp;
3635 	int lastref;
3636 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3637 #ifdef INET
3638 	IN_MULTI_LIST_UNLOCK_ASSERT();
3639 #endif
3640 	ifp = ifma->ifma_ifp;
3641 #ifdef DIAGNOSTIC
3642 	if (ifp == NULL) {
3643 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3644 	} else {
3645 		struct ifnet *oifp;
3646 
3647 		IFNET_RLOCK_NOSLEEP();
3648 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3649 			if (ifp == oifp)
3650 				break;
3651 		if (ifp != oifp) {
3652 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3653 			ifp = NULL;
3654 		}
3655 		IFNET_RUNLOCK_NOSLEEP();
3656 	}
3657 #endif
3658 	/*
3659 	 * If and only if the ifnet instance exists: Acquire the address lock.
3660 	 */
3661 	if (ifp != NULL)
3662 		IF_ADDR_WLOCK(ifp);
3663 
3664 	lastref = if_delmulti_locked(ifp, ifma, flags);
3665 
3666 	if (ifp != NULL) {
3667 		/*
3668 		 * If and only if the ifnet instance exists:
3669 		 *  Release the address lock.
3670 		 *  If the group was left: update the hardware hash filter.
3671 		 */
3672 		IF_ADDR_WUNLOCK(ifp);
3673 		if (lastref && ifp->if_ioctl != NULL) {
3674 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3675 		}
3676 	}
3677 }
3678 
3679 /*
3680  * Perform deletion of network-layer and/or link-layer multicast address.
3681  *
3682  * Return 0 if the reference count was decremented.
3683  * Return 1 if the final reference was released, indicating that the
3684  * hardware hash filter should be reprogrammed.
3685  */
3686 static int
3687 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3688 {
3689 	struct ifmultiaddr *ll_ifma;
3690 
3691 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3692 		KASSERT(ifma->ifma_ifp == ifp,
3693 		    ("%s: inconsistent ifp %p", __func__, ifp));
3694 		IF_ADDR_WLOCK_ASSERT(ifp);
3695 	}
3696 
3697 	ifp = ifma->ifma_ifp;
3698 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3699 
3700 	/*
3701 	 * If the ifnet is detaching, null out references to ifnet,
3702 	 * so that upper protocol layers will notice, and not attempt
3703 	 * to obtain locks for an ifnet which no longer exists. The
3704 	 * routing socket announcement must happen before the ifnet
3705 	 * instance is detached from the system.
3706 	 */
3707 	if (detaching) {
3708 #ifdef DIAGNOSTIC
3709 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3710 #endif
3711 		/*
3712 		 * ifp may already be nulled out if we are being reentered
3713 		 * to delete the ll_ifma.
3714 		 */
3715 		if (ifp != NULL) {
3716 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3717 			ifma->ifma_ifp = NULL;
3718 		}
3719 	}
3720 
3721 	if (--ifma->ifma_refcount > 0)
3722 		return 0;
3723 
3724 	if (ifp != NULL && detaching == 0)
3725 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3726 
3727 	/*
3728 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3729 	 * have been associated with it. Release it first if so.
3730 	 */
3731 	ll_ifma = ifma->ifma_llifma;
3732 	if (ll_ifma != NULL) {
3733 		KASSERT(ifma->ifma_lladdr != NULL,
3734 		    ("%s: llifma w/o lladdr", __func__));
3735 		if (detaching)
3736 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3737 		if (--ll_ifma->ifma_refcount == 0) {
3738 			if (ifp != NULL) {
3739 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3740 				    ifma_link);
3741 			}
3742 			if_freemulti(ll_ifma);
3743 		}
3744 	}
3745 #ifdef INVARIANTS
3746 	if (ifp) {
3747 		struct ifmultiaddr *ifmatmp;
3748 
3749 		TAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3750 			MPASS(ifma != ifmatmp);
3751 	}
3752 #endif
3753 	if_freemulti(ifma);
3754 	/*
3755 	 * The last reference to this instance of struct ifmultiaddr
3756 	 * was released; the hardware should be notified of this change.
3757 	 */
3758 	return 1;
3759 }
3760 
3761 /*
3762  * Set the link layer address on an interface.
3763  *
3764  * At this time we only support certain types of interfaces,
3765  * and we don't allow the length of the address to change.
3766  *
3767  * Set noinline to be dtrace-friendly
3768  */
3769 __noinline int
3770 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3771 {
3772 	struct sockaddr_dl *sdl;
3773 	struct ifaddr *ifa;
3774 	struct ifreq ifr;
3775 
3776 	IF_ADDR_RLOCK(ifp);
3777 	ifa = ifp->if_addr;
3778 	if (ifa == NULL) {
3779 		IF_ADDR_RUNLOCK(ifp);
3780 		return (EINVAL);
3781 	}
3782 	ifa_ref(ifa);
3783 	IF_ADDR_RUNLOCK(ifp);
3784 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3785 	if (sdl == NULL) {
3786 		ifa_free(ifa);
3787 		return (EINVAL);
3788 	}
3789 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3790 		ifa_free(ifa);
3791 		return (EINVAL);
3792 	}
3793 	switch (ifp->if_type) {
3794 	case IFT_ETHER:
3795 	case IFT_XETHER:
3796 	case IFT_L2VLAN:
3797 	case IFT_BRIDGE:
3798 	case IFT_IEEE8023ADLAG:
3799 		bcopy(lladdr, LLADDR(sdl), len);
3800 		ifa_free(ifa);
3801 		break;
3802 	default:
3803 		ifa_free(ifa);
3804 		return (ENODEV);
3805 	}
3806 
3807 	/*
3808 	 * If the interface is already up, we need
3809 	 * to re-init it in order to reprogram its
3810 	 * address filter.
3811 	 */
3812 	if ((ifp->if_flags & IFF_UP) != 0) {
3813 		if (ifp->if_ioctl) {
3814 			ifp->if_flags &= ~IFF_UP;
3815 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3816 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3817 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3818 			ifp->if_flags |= IFF_UP;
3819 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3820 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3821 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3822 		}
3823 	}
3824 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3825 	return (0);
3826 }
3827 
3828 /*
3829  * Compat function for handling basic encapsulation requests.
3830  * Not converted stacks (FDDI, IB, ..) supports traditional
3831  * output model: ARP (and other similar L2 protocols) are handled
3832  * inside output routine, arpresolve/nd6_resolve() returns MAC
3833  * address instead of full prepend.
3834  *
3835  * This function creates calculated header==MAC for IPv4/IPv6 and
3836  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3837  * address families.
3838  */
3839 static int
3840 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3841 {
3842 
3843 	if (req->rtype != IFENCAP_LL)
3844 		return (EOPNOTSUPP);
3845 
3846 	if (req->bufsize < req->lladdr_len)
3847 		return (ENOMEM);
3848 
3849 	switch (req->family) {
3850 	case AF_INET:
3851 	case AF_INET6:
3852 		break;
3853 	default:
3854 		return (EAFNOSUPPORT);
3855 	}
3856 
3857 	/* Copy lladdr to storage as is */
3858 	memmove(req->buf, req->lladdr, req->lladdr_len);
3859 	req->bufsize = req->lladdr_len;
3860 	req->lladdr_off = 0;
3861 
3862 	return (0);
3863 }
3864 
3865 /*
3866  * Get the link layer address that was read from the hardware at attach.
3867  *
3868  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3869  * their component interfaces as IFT_IEEE8023ADLAG.
3870  */
3871 int
3872 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3873 {
3874 
3875 	if (ifp->if_hw_addr == NULL)
3876 		return (ENODEV);
3877 
3878 	switch (ifp->if_type) {
3879 	case IFT_ETHER:
3880 	case IFT_IEEE8023ADLAG:
3881 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3882 		return (0);
3883 	default:
3884 		return (ENODEV);
3885 	}
3886 }
3887 
3888 /*
3889  * The name argument must be a pointer to storage which will last as
3890  * long as the interface does.  For physical devices, the result of
3891  * device_get_name(dev) is a good choice and for pseudo-devices a
3892  * static string works well.
3893  */
3894 void
3895 if_initname(struct ifnet *ifp, const char *name, int unit)
3896 {
3897 	ifp->if_dname = name;
3898 	ifp->if_dunit = unit;
3899 	if (unit != IF_DUNIT_NONE)
3900 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3901 	else
3902 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3903 }
3904 
3905 int
3906 if_printf(struct ifnet *ifp, const char * fmt, ...)
3907 {
3908 	va_list ap;
3909 	int retval;
3910 
3911 	retval = printf("%s: ", ifp->if_xname);
3912 	va_start(ap, fmt);
3913 	retval += vprintf(fmt, ap);
3914 	va_end(ap);
3915 	return (retval);
3916 }
3917 
3918 void
3919 if_start(struct ifnet *ifp)
3920 {
3921 
3922 	(*(ifp)->if_start)(ifp);
3923 }
3924 
3925 /*
3926  * Backwards compatibility interface for drivers
3927  * that have not implemented it
3928  */
3929 static int
3930 if_transmit(struct ifnet *ifp, struct mbuf *m)
3931 {
3932 	int error;
3933 
3934 	IFQ_HANDOFF(ifp, m, error);
3935 	return (error);
3936 }
3937 
3938 static void
3939 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3940 {
3941 
3942 	m_freem(m);
3943 }
3944 
3945 int
3946 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3947 {
3948 	int active = 0;
3949 
3950 	IF_LOCK(ifq);
3951 	if (_IF_QFULL(ifq)) {
3952 		IF_UNLOCK(ifq);
3953 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3954 		m_freem(m);
3955 		return (0);
3956 	}
3957 	if (ifp != NULL) {
3958 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3959 		if (m->m_flags & (M_BCAST|M_MCAST))
3960 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3961 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3962 	}
3963 	_IF_ENQUEUE(ifq, m);
3964 	IF_UNLOCK(ifq);
3965 	if (ifp != NULL && !active)
3966 		(*(ifp)->if_start)(ifp);
3967 	return (1);
3968 }
3969 
3970 void
3971 if_register_com_alloc(u_char type,
3972     if_com_alloc_t *a, if_com_free_t *f)
3973 {
3974 
3975 	KASSERT(if_com_alloc[type] == NULL,
3976 	    ("if_register_com_alloc: %d already registered", type));
3977 	KASSERT(if_com_free[type] == NULL,
3978 	    ("if_register_com_alloc: %d free already registered", type));
3979 
3980 	if_com_alloc[type] = a;
3981 	if_com_free[type] = f;
3982 }
3983 
3984 void
3985 if_deregister_com_alloc(u_char type)
3986 {
3987 
3988 	KASSERT(if_com_alloc[type] != NULL,
3989 	    ("if_deregister_com_alloc: %d not registered", type));
3990 	KASSERT(if_com_free[type] != NULL,
3991 	    ("if_deregister_com_alloc: %d free not registered", type));
3992 	if_com_alloc[type] = NULL;
3993 	if_com_free[type] = NULL;
3994 }
3995 
3996 /* API for driver access to network stack owned ifnet.*/
3997 uint64_t
3998 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3999 {
4000 	uint64_t oldbrate;
4001 
4002 	oldbrate = ifp->if_baudrate;
4003 	ifp->if_baudrate = baudrate;
4004 	return (oldbrate);
4005 }
4006 
4007 uint64_t
4008 if_getbaudrate(if_t ifp)
4009 {
4010 
4011 	return (((struct ifnet *)ifp)->if_baudrate);
4012 }
4013 
4014 int
4015 if_setcapabilities(if_t ifp, int capabilities)
4016 {
4017 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4018 	return (0);
4019 }
4020 
4021 int
4022 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4023 {
4024 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4025 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4026 
4027 	return (0);
4028 }
4029 
4030 int
4031 if_getcapabilities(if_t ifp)
4032 {
4033 	return ((struct ifnet *)ifp)->if_capabilities;
4034 }
4035 
4036 int
4037 if_setcapenable(if_t ifp, int capabilities)
4038 {
4039 	((struct ifnet *)ifp)->if_capenable = capabilities;
4040 	return (0);
4041 }
4042 
4043 int
4044 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4045 {
4046 	if(setcap)
4047 		((struct ifnet *)ifp)->if_capenable |= setcap;
4048 	if(clearcap)
4049 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4050 
4051 	return (0);
4052 }
4053 
4054 const char *
4055 if_getdname(if_t ifp)
4056 {
4057 	return ((struct ifnet *)ifp)->if_dname;
4058 }
4059 
4060 int
4061 if_togglecapenable(if_t ifp, int togglecap)
4062 {
4063 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4064 	return (0);
4065 }
4066 
4067 int
4068 if_getcapenable(if_t ifp)
4069 {
4070 	return ((struct ifnet *)ifp)->if_capenable;
4071 }
4072 
4073 /*
4074  * This is largely undesirable because it ties ifnet to a device, but does
4075  * provide flexiblity for an embedded product vendor. Should be used with
4076  * the understanding that it violates the interface boundaries, and should be
4077  * a last resort only.
4078  */
4079 int
4080 if_setdev(if_t ifp, void *dev)
4081 {
4082 	return (0);
4083 }
4084 
4085 int
4086 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4087 {
4088 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4089 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4090 
4091 	return (0);
4092 }
4093 
4094 int
4095 if_getdrvflags(if_t ifp)
4096 {
4097 	return ((struct ifnet *)ifp)->if_drv_flags;
4098 }
4099 
4100 int
4101 if_setdrvflags(if_t ifp, int flags)
4102 {
4103 	((struct ifnet *)ifp)->if_drv_flags = flags;
4104 	return (0);
4105 }
4106 
4107 
4108 int
4109 if_setflags(if_t ifp, int flags)
4110 {
4111 	((struct ifnet *)ifp)->if_flags = flags;
4112 	return (0);
4113 }
4114 
4115 int
4116 if_setflagbits(if_t ifp, int set, int clear)
4117 {
4118 	((struct ifnet *)ifp)->if_flags |= set;
4119 	((struct ifnet *)ifp)->if_flags &= ~clear;
4120 
4121 	return (0);
4122 }
4123 
4124 int
4125 if_getflags(if_t ifp)
4126 {
4127 	return ((struct ifnet *)ifp)->if_flags;
4128 }
4129 
4130 int
4131 if_clearhwassist(if_t ifp)
4132 {
4133 	((struct ifnet *)ifp)->if_hwassist = 0;
4134 	return (0);
4135 }
4136 
4137 int
4138 if_sethwassistbits(if_t ifp, int toset, int toclear)
4139 {
4140 	((struct ifnet *)ifp)->if_hwassist |= toset;
4141 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4142 
4143 	return (0);
4144 }
4145 
4146 int
4147 if_sethwassist(if_t ifp, int hwassist_bit)
4148 {
4149 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4150 	return (0);
4151 }
4152 
4153 int
4154 if_gethwassist(if_t ifp)
4155 {
4156 	return ((struct ifnet *)ifp)->if_hwassist;
4157 }
4158 
4159 int
4160 if_setmtu(if_t ifp, int mtu)
4161 {
4162 	((struct ifnet *)ifp)->if_mtu = mtu;
4163 	return (0);
4164 }
4165 
4166 int
4167 if_getmtu(if_t ifp)
4168 {
4169 	return ((struct ifnet *)ifp)->if_mtu;
4170 }
4171 
4172 int
4173 if_getmtu_family(if_t ifp, int family)
4174 {
4175 	struct domain *dp;
4176 
4177 	for (dp = domains; dp; dp = dp->dom_next) {
4178 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4179 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4180 	}
4181 
4182 	return (((struct ifnet *)ifp)->if_mtu);
4183 }
4184 
4185 int
4186 if_setsoftc(if_t ifp, void *softc)
4187 {
4188 	((struct ifnet *)ifp)->if_softc = softc;
4189 	return (0);
4190 }
4191 
4192 void *
4193 if_getsoftc(if_t ifp)
4194 {
4195 	return ((struct ifnet *)ifp)->if_softc;
4196 }
4197 
4198 void
4199 if_setrcvif(struct mbuf *m, if_t ifp)
4200 {
4201 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4202 }
4203 
4204 void
4205 if_setvtag(struct mbuf *m, uint16_t tag)
4206 {
4207 	m->m_pkthdr.ether_vtag = tag;
4208 }
4209 
4210 uint16_t
4211 if_getvtag(struct mbuf *m)
4212 {
4213 
4214 	return (m->m_pkthdr.ether_vtag);
4215 }
4216 
4217 int
4218 if_sendq_empty(if_t ifp)
4219 {
4220 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4221 }
4222 
4223 struct ifaddr *
4224 if_getifaddr(if_t ifp)
4225 {
4226 	return ((struct ifnet *)ifp)->if_addr;
4227 }
4228 
4229 int
4230 if_getamcount(if_t ifp)
4231 {
4232 	return ((struct ifnet *)ifp)->if_amcount;
4233 }
4234 
4235 
4236 int
4237 if_setsendqready(if_t ifp)
4238 {
4239 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4240 	return (0);
4241 }
4242 
4243 int
4244 if_setsendqlen(if_t ifp, int tx_desc_count)
4245 {
4246 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4247 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4248 
4249 	return (0);
4250 }
4251 
4252 int
4253 if_vlantrunkinuse(if_t ifp)
4254 {
4255 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4256 }
4257 
4258 int
4259 if_input(if_t ifp, struct mbuf* sendmp)
4260 {
4261 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4262 	return (0);
4263 
4264 }
4265 
4266 /* XXX */
4267 #ifndef ETH_ADDR_LEN
4268 #define ETH_ADDR_LEN 6
4269 #endif
4270 
4271 int
4272 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4273 {
4274 	struct ifmultiaddr *ifma;
4275 	uint8_t *lmta = (uint8_t *)mta;
4276 	int mcnt = 0;
4277 
4278 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4279 		if (ifma->ifma_addr->sa_family != AF_LINK)
4280 			continue;
4281 
4282 		if (mcnt == max)
4283 			break;
4284 
4285 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4286 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4287 		mcnt++;
4288 	}
4289 	*cnt = mcnt;
4290 
4291 	return (0);
4292 }
4293 
4294 int
4295 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4296 {
4297 	int error;
4298 
4299 	if_maddr_rlock(ifp);
4300 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4301 	if_maddr_runlock(ifp);
4302 	return (error);
4303 }
4304 
4305 int
4306 if_multiaddr_count(if_t ifp, int max)
4307 {
4308 	struct ifmultiaddr *ifma;
4309 	int count;
4310 
4311 	count = 0;
4312 	if_maddr_rlock(ifp);
4313 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4314 		if (ifma->ifma_addr->sa_family != AF_LINK)
4315 			continue;
4316 		count++;
4317 		if (count == max)
4318 			break;
4319 	}
4320 	if_maddr_runlock(ifp);
4321 	return (count);
4322 }
4323 
4324 int
4325 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4326 {
4327 	struct ifmultiaddr *ifma;
4328 	int cnt = 0;
4329 
4330 	if_maddr_rlock(ifp);
4331 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4332 		cnt += filter(arg, ifma, cnt);
4333 	if_maddr_runlock(ifp);
4334 	return (cnt);
4335 }
4336 
4337 struct mbuf *
4338 if_dequeue(if_t ifp)
4339 {
4340 	struct mbuf *m;
4341 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4342 
4343 	return (m);
4344 }
4345 
4346 int
4347 if_sendq_prepend(if_t ifp, struct mbuf *m)
4348 {
4349 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4350 	return (0);
4351 }
4352 
4353 int
4354 if_setifheaderlen(if_t ifp, int len)
4355 {
4356 	((struct ifnet *)ifp)->if_hdrlen = len;
4357 	return (0);
4358 }
4359 
4360 caddr_t
4361 if_getlladdr(if_t ifp)
4362 {
4363 	return (IF_LLADDR((struct ifnet *)ifp));
4364 }
4365 
4366 void *
4367 if_gethandle(u_char type)
4368 {
4369 	return (if_alloc(type));
4370 }
4371 
4372 void
4373 if_bpfmtap(if_t ifh, struct mbuf *m)
4374 {
4375 	struct ifnet *ifp = (struct ifnet *)ifh;
4376 
4377 	BPF_MTAP(ifp, m);
4378 }
4379 
4380 void
4381 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4382 {
4383 	struct ifnet *ifp = (struct ifnet *)ifh;
4384 
4385 	ETHER_BPF_MTAP(ifp, m);
4386 }
4387 
4388 void
4389 if_vlancap(if_t ifh)
4390 {
4391 	struct ifnet *ifp = (struct ifnet *)ifh;
4392 	VLAN_CAPABILITIES(ifp);
4393 }
4394 
4395 int
4396 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4397 {
4398 
4399 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4400         return (0);
4401 }
4402 
4403 int
4404 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4405 {
4406 
4407 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4408         return (0);
4409 }
4410 
4411 int
4412 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4413 {
4414 
4415 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4416         return (0);
4417 }
4418 
4419 u_int
4420 if_gethwtsomax(if_t ifp)
4421 {
4422 
4423 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4424 }
4425 
4426 u_int
4427 if_gethwtsomaxsegcount(if_t ifp)
4428 {
4429 
4430 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4431 }
4432 
4433 u_int
4434 if_gethwtsomaxsegsize(if_t ifp)
4435 {
4436 
4437 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4438 }
4439 
4440 void
4441 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4442 {
4443 	((struct ifnet *)ifp)->if_init = init_fn;
4444 }
4445 
4446 void
4447 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4448 {
4449 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4450 }
4451 
4452 void
4453 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4454 {
4455 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4456 }
4457 
4458 void
4459 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4460 {
4461 	((struct ifnet *)ifp)->if_transmit = start_fn;
4462 }
4463 
4464 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4465 {
4466 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4467 
4468 }
4469 
4470 void
4471 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4472 {
4473 
4474 	ifp->if_get_counter = fn;
4475 }
4476 
4477 /* Revisit these - These are inline functions originally. */
4478 int
4479 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4480 {
4481 	return drbr_inuse(ifh, br);
4482 }
4483 
4484 struct mbuf*
4485 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4486 {
4487 	return drbr_dequeue(ifh, br);
4488 }
4489 
4490 int
4491 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4492 {
4493 	return drbr_needs_enqueue(ifh, br);
4494 }
4495 
4496 int
4497 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4498 {
4499 	return drbr_enqueue(ifh, br, m);
4500 
4501 }
4502