xref: /freebsd/sys/net/if.c (revision 983afe3373c427a080f06dccec820b20891be186)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/epoch.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysent.h>
61 #include <sys/taskqueue.h>
62 #include <sys/domain.h>
63 #include <sys/jail.h>
64 #include <sys/priv.h>
65 
66 #include <machine/stdarg.h>
67 #include <vm/uma.h>
68 
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if.h>
72 #include <net/if_arp.h>
73 #include <net/if_clone.h>
74 #include <net/if_dl.h>
75 #include <net/if_types.h>
76 #include <net/if_var.h>
77 #include <net/if_media.h>
78 #include <net/if_vlan_var.h>
79 #include <net/radix.h>
80 #include <net/route.h>
81 #include <net/vnet.h>
82 
83 #if defined(INET) || defined(INET6)
84 #include <net/ethernet.h>
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip_carp.h>
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #include <netinet/netdump/netdump.h>
92 #endif /* INET */
93 #ifdef INET6
94 #include <netinet6/in6_var.h>
95 #include <netinet6/in6_ifattach.h>
96 #endif /* INET6 */
97 #endif /* INET || INET6 */
98 
99 #include <security/mac/mac_framework.h>
100 
101 /*
102  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
103  * and ifr_ifru when it is used in SIOCGIFCONF.
104  */
105 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
106     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
107 
108 __read_mostly epoch_t net_epoch_preempt;
109 __read_mostly epoch_t net_epoch;
110 #ifdef COMPAT_FREEBSD32
111 #include <sys/mount.h>
112 #include <compat/freebsd32/freebsd32.h>
113 
114 struct ifreq_buffer32 {
115 	uint32_t	length;		/* (size_t) */
116 	uint32_t	buffer;		/* (void *) */
117 };
118 
119 /*
120  * Interface request structure used for socket
121  * ioctl's.  All interface ioctl's must have parameter
122  * definitions which begin with ifr_name.  The
123  * remainder may be interface specific.
124  */
125 struct ifreq32 {
126 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
127 	union {
128 		struct sockaddr	ifru_addr;
129 		struct sockaddr	ifru_dstaddr;
130 		struct sockaddr	ifru_broadaddr;
131 		struct ifreq_buffer32 ifru_buffer;
132 		short		ifru_flags[2];
133 		short		ifru_index;
134 		int		ifru_jid;
135 		int		ifru_metric;
136 		int		ifru_mtu;
137 		int		ifru_phys;
138 		int		ifru_media;
139 		uint32_t	ifru_data;
140 		int		ifru_cap[2];
141 		u_int		ifru_fib;
142 		u_char		ifru_vlan_pcp;
143 	} ifr_ifru;
144 };
145 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
146 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
147     __offsetof(struct ifreq32, ifr_ifru));
148 
149 struct ifgroupreq32 {
150 	char	ifgr_name[IFNAMSIZ];
151 	u_int	ifgr_len;
152 	union {
153 		char		ifgru_group[IFNAMSIZ];
154 		uint32_t	ifgru_groups;
155 	} ifgr_ifgru;
156 };
157 
158 struct ifmediareq32 {
159 	char		ifm_name[IFNAMSIZ];
160 	int		ifm_current;
161 	int		ifm_mask;
162 	int		ifm_status;
163 	int		ifm_active;
164 	int		ifm_count;
165 	uint32_t	ifm_ulist;	/* (int *) */
166 };
167 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
168 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
169 
170 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
171     case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
172 #else /* !COMPAT_FREEBSD32 */
173 #define _CASE_IOC_IFGROUPREQ_32(cmd)
174 #endif /* !COMPAT_FREEBSD32 */
175 
176 #define CASE_IOC_IFGROUPREQ(cmd)	\
177     _CASE_IOC_IFGROUPREQ_32(cmd)	\
178     case (cmd)
179 
180 union ifreq_union {
181 	struct ifreq	ifr;
182 #ifdef COMPAT_FREEBSD32
183 	struct ifreq32	ifr32;
184 #endif
185 };
186 
187 union ifgroupreq_union {
188 	struct ifgroupreq ifgr;
189 #ifdef COMPAT_FREEBSD32
190 	struct ifgroupreq32 ifgr32;
191 #endif
192 };
193 
194 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
195 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
196 
197 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
198     &ifqmaxlen, 0, "max send queue size");
199 
200 /* Log link state change events */
201 static int log_link_state_change = 1;
202 
203 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
204 	&log_link_state_change, 0,
205 	"log interface link state change events");
206 
207 /* Log promiscuous mode change events */
208 static int log_promisc_mode_change = 1;
209 
210 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
211 	&log_promisc_mode_change, 1,
212 	"log promiscuous mode change events");
213 
214 /* Interface description */
215 static unsigned int ifdescr_maxlen = 1024;
216 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
217 	&ifdescr_maxlen, 0,
218 	"administrative maximum length for interface description");
219 
220 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
221 
222 /* global sx for non-critical path ifdescr */
223 static struct sx ifdescr_sx;
224 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
225 
226 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
227 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
228 /* These are external hooks for CARP. */
229 void	(*carp_linkstate_p)(struct ifnet *ifp);
230 void	(*carp_demote_adj_p)(int, char *);
231 int	(*carp_master_p)(struct ifaddr *);
232 #if defined(INET) || defined(INET6)
233 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
234 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
235     const struct sockaddr *sa);
236 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
237 int	(*carp_attach_p)(struct ifaddr *, int);
238 void	(*carp_detach_p)(struct ifaddr *, bool);
239 #endif
240 #ifdef INET
241 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
242 #endif
243 #ifdef INET6
244 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
245 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
246     const struct in6_addr *taddr);
247 #endif
248 
249 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
250 
251 /*
252  * XXX: Style; these should be sorted alphabetically, and unprototyped
253  * static functions should be prototyped. Currently they are sorted by
254  * declaration order.
255  */
256 static void	if_attachdomain(void *);
257 static void	if_attachdomain1(struct ifnet *);
258 static int	ifconf(u_long, caddr_t);
259 static void	if_grow(void);
260 static void	if_input_default(struct ifnet *, struct mbuf *);
261 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
262 static void	if_route(struct ifnet *, int flag, int fam);
263 static int	if_setflag(struct ifnet *, int, int, int *, int);
264 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
265 static void	if_unroute(struct ifnet *, int flag, int fam);
266 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
267 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
268 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
269 static void	do_link_state_change(void *, int);
270 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
271 static int	if_getgroupmembers(struct ifgroupreq *);
272 static void	if_delgroups(struct ifnet *);
273 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
274 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
275 #ifdef VIMAGE
276 static void	if_vmove(struct ifnet *, struct vnet *);
277 #endif
278 
279 #ifdef INET6
280 /*
281  * XXX: declare here to avoid to include many inet6 related files..
282  * should be more generalized?
283  */
284 extern void	nd6_setmtu(struct ifnet *);
285 #endif
286 
287 /* ipsec helper hooks */
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
289 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
290 
291 VNET_DEFINE(int, if_index);
292 int	ifqmaxlen = IFQ_MAXLEN;
293 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
294 VNET_DEFINE(struct ifgrouphead, ifg_head);
295 
296 static VNET_DEFINE(int, if_indexlim) = 8;
297 
298 /* Table of ifnet by index. */
299 VNET_DEFINE(struct ifnet **, ifindex_table);
300 
301 #define	V_if_indexlim		VNET(if_indexlim)
302 #define	V_ifindex_table		VNET(ifindex_table)
303 
304 /*
305  * The global network interface list (V_ifnet) and related state (such as
306  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
307  * an rwlock.  Either may be acquired shared to stablize the list, but both
308  * must be acquired writable to modify the list.  This model allows us to
309  * both stablize the interface list during interrupt thread processing, but
310  * also to stablize it over long-running ioctls, without introducing priority
311  * inversions and deadlocks.
312  */
313 struct rwlock ifnet_rwlock;
314 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
315 struct sx ifnet_sxlock;
316 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
317 
318 /*
319  * The allocation of network interfaces is a rather non-atomic affair; we
320  * need to select an index before we are ready to expose the interface for
321  * use, so will use this pointer value to indicate reservation.
322  */
323 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
324 
325 static	if_com_alloc_t *if_com_alloc[256];
326 static	if_com_free_t *if_com_free[256];
327 
328 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
329 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
330 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
331 
332 struct ifnet *
333 ifnet_byindex_locked(u_short idx)
334 {
335 
336 	if (idx > V_if_index)
337 		return (NULL);
338 	if (V_ifindex_table[idx] == IFNET_HOLD)
339 		return (NULL);
340 	return (V_ifindex_table[idx]);
341 }
342 
343 struct ifnet *
344 ifnet_byindex(u_short idx)
345 {
346 	struct ifnet *ifp;
347 
348 	IFNET_RLOCK_NOSLEEP();
349 	ifp = ifnet_byindex_locked(idx);
350 	IFNET_RUNLOCK_NOSLEEP();
351 	return (ifp);
352 }
353 
354 struct ifnet *
355 ifnet_byindex_ref(u_short idx)
356 {
357 	struct ifnet *ifp;
358 
359 	IFNET_RLOCK_NOSLEEP();
360 	ifp = ifnet_byindex_locked(idx);
361 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
362 		IFNET_RUNLOCK_NOSLEEP();
363 		return (NULL);
364 	}
365 	if_ref(ifp);
366 	IFNET_RUNLOCK_NOSLEEP();
367 	return (ifp);
368 }
369 
370 /*
371  * Allocate an ifindex array entry; return 0 on success or an error on
372  * failure.
373  */
374 static u_short
375 ifindex_alloc(void)
376 {
377 	u_short idx;
378 
379 	IFNET_WLOCK_ASSERT();
380 retry:
381 	/*
382 	 * Try to find an empty slot below V_if_index.  If we fail, take the
383 	 * next slot.
384 	 */
385 	for (idx = 1; idx <= V_if_index; idx++) {
386 		if (V_ifindex_table[idx] == NULL)
387 			break;
388 	}
389 
390 	/* Catch if_index overflow. */
391 	if (idx >= V_if_indexlim) {
392 		if_grow();
393 		goto retry;
394 	}
395 	if (idx > V_if_index)
396 		V_if_index = idx;
397 	return (idx);
398 }
399 
400 static void
401 ifindex_free_locked(u_short idx)
402 {
403 
404 	IFNET_WLOCK_ASSERT();
405 
406 	V_ifindex_table[idx] = NULL;
407 	while (V_if_index > 0 &&
408 	    V_ifindex_table[V_if_index] == NULL)
409 		V_if_index--;
410 }
411 
412 static void
413 ifindex_free(u_short idx)
414 {
415 
416 	IFNET_WLOCK();
417 	ifindex_free_locked(idx);
418 	IFNET_WUNLOCK();
419 }
420 
421 static void
422 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
423 {
424 
425 	IFNET_WLOCK_ASSERT();
426 
427 	V_ifindex_table[idx] = ifp;
428 }
429 
430 static void
431 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
432 {
433 
434 	IFNET_WLOCK();
435 	ifnet_setbyindex_locked(idx, ifp);
436 	IFNET_WUNLOCK();
437 }
438 
439 struct ifaddr *
440 ifaddr_byindex(u_short idx)
441 {
442 	struct ifnet *ifp;
443 	struct ifaddr *ifa = NULL;
444 
445 	IFNET_RLOCK_NOSLEEP();
446 	ifp = ifnet_byindex_locked(idx);
447 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
448 		ifa_ref(ifa);
449 	IFNET_RUNLOCK_NOSLEEP();
450 	return (ifa);
451 }
452 
453 /*
454  * Network interface utility routines.
455  *
456  * Routines with ifa_ifwith* names take sockaddr *'s as
457  * parameters.
458  */
459 
460 static void
461 vnet_if_init(const void *unused __unused)
462 {
463 
464 	TAILQ_INIT(&V_ifnet);
465 	TAILQ_INIT(&V_ifg_head);
466 	IFNET_WLOCK();
467 	if_grow();				/* create initial table */
468 	IFNET_WUNLOCK();
469 	vnet_if_clone_init();
470 }
471 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
472     NULL);
473 
474 #ifdef VIMAGE
475 static void
476 vnet_if_uninit(const void *unused __unused)
477 {
478 
479 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
480 	    "not empty", __func__, __LINE__, &V_ifnet));
481 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
482 	    "not empty", __func__, __LINE__, &V_ifg_head));
483 
484 	free((caddr_t)V_ifindex_table, M_IFNET);
485 }
486 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
487     vnet_if_uninit, NULL);
488 
489 static void
490 vnet_if_return(const void *unused __unused)
491 {
492 	struct ifnet *ifp, *nifp;
493 
494 	/* Return all inherited interfaces to their parent vnets. */
495 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
496 		if (ifp->if_home_vnet != ifp->if_vnet)
497 			if_vmove(ifp, ifp->if_home_vnet);
498 	}
499 }
500 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
501     vnet_if_return, NULL);
502 #endif
503 
504 static void
505 if_grow(void)
506 {
507 	int oldlim;
508 	u_int n;
509 	struct ifnet **e;
510 
511 	IFNET_WLOCK_ASSERT();
512 	oldlim = V_if_indexlim;
513 	IFNET_WUNLOCK();
514 	n = (oldlim << 1) * sizeof(*e);
515 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
516 	IFNET_WLOCK();
517 	if (V_if_indexlim != oldlim) {
518 		free(e, M_IFNET);
519 		return;
520 	}
521 	if (V_ifindex_table != NULL) {
522 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
523 		free((caddr_t)V_ifindex_table, M_IFNET);
524 	}
525 	V_if_indexlim <<= 1;
526 	V_ifindex_table = e;
527 }
528 
529 /*
530  * Allocate a struct ifnet and an index for an interface.  A layer 2
531  * common structure will also be allocated if an allocation routine is
532  * registered for the passed type.
533  */
534 struct ifnet *
535 if_alloc(u_char type)
536 {
537 	struct ifnet *ifp;
538 	u_short idx;
539 
540 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
541 	IFNET_WLOCK();
542 	idx = ifindex_alloc();
543 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
544 	IFNET_WUNLOCK();
545 	ifp->if_index = idx;
546 	ifp->if_type = type;
547 	ifp->if_alloctype = type;
548 #ifdef VIMAGE
549 	ifp->if_vnet = curvnet;
550 #endif
551 	if (if_com_alloc[type] != NULL) {
552 		ifp->if_l2com = if_com_alloc[type](type, ifp);
553 		if (ifp->if_l2com == NULL) {
554 			free(ifp, M_IFNET);
555 			ifindex_free(idx);
556 			return (NULL);
557 		}
558 	}
559 
560 	IF_ADDR_LOCK_INIT(ifp);
561 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
562 	ifp->if_afdata_initialized = 0;
563 	IF_AFDATA_LOCK_INIT(ifp);
564 	CK_STAILQ_INIT(&ifp->if_addrhead);
565 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
566 	TAILQ_INIT(&ifp->if_groups);
567 #ifdef MAC
568 	mac_ifnet_init(ifp);
569 #endif
570 	ifq_init(&ifp->if_snd, ifp);
571 
572 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
573 	for (int i = 0; i < IFCOUNTERS; i++)
574 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
575 	ifp->if_get_counter = if_get_counter_default;
576 	ifp->if_pcp = IFNET_PCP_NONE;
577 	ifnet_setbyindex(ifp->if_index, ifp);
578 	return (ifp);
579 }
580 
581 /*
582  * Do the actual work of freeing a struct ifnet, and layer 2 common
583  * structure.  This call is made when the last reference to an
584  * interface is released.
585  */
586 static void
587 if_free_internal(struct ifnet *ifp)
588 {
589 
590 	KASSERT((ifp->if_flags & IFF_DYING),
591 	    ("if_free_internal: interface not dying"));
592 
593 	if (if_com_free[ifp->if_alloctype] != NULL)
594 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
595 		    ifp->if_alloctype);
596 
597 #ifdef MAC
598 	mac_ifnet_destroy(ifp);
599 #endif /* MAC */
600 	if (ifp->if_description != NULL)
601 		free(ifp->if_description, M_IFDESCR);
602 	IF_AFDATA_DESTROY(ifp);
603 	IF_ADDR_LOCK_DESTROY(ifp);
604 	ifq_delete(&ifp->if_snd);
605 
606 	for (int i = 0; i < IFCOUNTERS; i++)
607 		counter_u64_free(ifp->if_counters[i]);
608 
609 	free(ifp, M_IFNET);
610 }
611 
612 /*
613  * Deregister an interface and free the associated storage.
614  */
615 void
616 if_free(struct ifnet *ifp)
617 {
618 
619 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
620 
621 	CURVNET_SET_QUIET(ifp->if_vnet);
622 	IFNET_WLOCK();
623 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
624 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
625 
626 	ifindex_free_locked(ifp->if_index);
627 	IFNET_WUNLOCK();
628 
629 	if (refcount_release(&ifp->if_refcount))
630 		if_free_internal(ifp);
631 	CURVNET_RESTORE();
632 }
633 
634 /*
635  * Interfaces to keep an ifnet type-stable despite the possibility of the
636  * driver calling if_free().  If there are additional references, we defer
637  * freeing the underlying data structure.
638  */
639 void
640 if_ref(struct ifnet *ifp)
641 {
642 
643 	/* We don't assert the ifnet list lock here, but arguably should. */
644 	refcount_acquire(&ifp->if_refcount);
645 }
646 
647 void
648 if_rele(struct ifnet *ifp)
649 {
650 
651 	if (!refcount_release(&ifp->if_refcount))
652 		return;
653 	if_free_internal(ifp);
654 }
655 
656 void
657 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
658 {
659 
660 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
661 
662 	if (ifq->ifq_maxlen == 0)
663 		ifq->ifq_maxlen = ifqmaxlen;
664 
665 	ifq->altq_type = 0;
666 	ifq->altq_disc = NULL;
667 	ifq->altq_flags &= ALTQF_CANTCHANGE;
668 	ifq->altq_tbr  = NULL;
669 	ifq->altq_ifp  = ifp;
670 }
671 
672 void
673 ifq_delete(struct ifaltq *ifq)
674 {
675 	mtx_destroy(&ifq->ifq_mtx);
676 }
677 
678 /*
679  * Perform generic interface initialization tasks and attach the interface
680  * to the list of "active" interfaces.  If vmove flag is set on entry
681  * to if_attach_internal(), perform only a limited subset of initialization
682  * tasks, given that we are moving from one vnet to another an ifnet which
683  * has already been fully initialized.
684  *
685  * Note that if_detach_internal() removes group membership unconditionally
686  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
687  * Thus, when if_vmove() is applied to a cloned interface, group membership
688  * is lost while a cloned one always joins a group whose name is
689  * ifc->ifc_name.  To recover this after if_detach_internal() and
690  * if_attach_internal(), the cloner should be specified to
691  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
692  * attempts to join a group whose name is ifc->ifc_name.
693  *
694  * XXX:
695  *  - The decision to return void and thus require this function to
696  *    succeed is questionable.
697  *  - We should probably do more sanity checking.  For instance we don't
698  *    do anything to insure if_xname is unique or non-empty.
699  */
700 void
701 if_attach(struct ifnet *ifp)
702 {
703 
704 	if_attach_internal(ifp, 0, NULL);
705 }
706 
707 /*
708  * Compute the least common TSO limit.
709  */
710 void
711 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
712 {
713 	/*
714 	 * 1) If there is no limit currently, take the limit from
715 	 * the network adapter.
716 	 *
717 	 * 2) If the network adapter has a limit below the current
718 	 * limit, apply it.
719 	 */
720 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
721 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
722 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
723 	}
724 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
725 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
726 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
727 	}
728 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
729 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
730 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
731 	}
732 }
733 
734 /*
735  * Update TSO limit of a network adapter.
736  *
737  * Returns zero if no change. Else non-zero.
738  */
739 int
740 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
741 {
742 	int retval = 0;
743 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
744 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
745 		retval++;
746 	}
747 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
748 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
749 		retval++;
750 	}
751 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
752 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
753 		retval++;
754 	}
755 	return (retval);
756 }
757 
758 static void
759 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
760 {
761 	unsigned socksize, ifasize;
762 	int namelen, masklen;
763 	struct sockaddr_dl *sdl;
764 	struct ifaddr *ifa;
765 
766 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
767 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
768 		    ifp->if_xname);
769 
770 #ifdef VIMAGE
771 	ifp->if_vnet = curvnet;
772 	if (ifp->if_home_vnet == NULL)
773 		ifp->if_home_vnet = curvnet;
774 #endif
775 
776 	if_addgroup(ifp, IFG_ALL);
777 
778 	/* Restore group membership for cloned interfaces. */
779 	if (vmove && ifc != NULL)
780 		if_clone_addgroup(ifp, ifc);
781 
782 	getmicrotime(&ifp->if_lastchange);
783 	ifp->if_epoch = time_uptime;
784 
785 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
786 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
787 	    ("transmit and qflush must both either be set or both be NULL"));
788 	if (ifp->if_transmit == NULL) {
789 		ifp->if_transmit = if_transmit;
790 		ifp->if_qflush = if_qflush;
791 	}
792 	if (ifp->if_input == NULL)
793 		ifp->if_input = if_input_default;
794 
795 	if (ifp->if_requestencap == NULL)
796 		ifp->if_requestencap = if_requestencap_default;
797 
798 	if (!vmove) {
799 #ifdef MAC
800 		mac_ifnet_create(ifp);
801 #endif
802 
803 		/*
804 		 * Create a Link Level name for this device.
805 		 */
806 		namelen = strlen(ifp->if_xname);
807 		/*
808 		 * Always save enough space for any possiable name so we
809 		 * can do a rename in place later.
810 		 */
811 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
812 		socksize = masklen + ifp->if_addrlen;
813 		if (socksize < sizeof(*sdl))
814 			socksize = sizeof(*sdl);
815 		socksize = roundup2(socksize, sizeof(long));
816 		ifasize = sizeof(*ifa) + 2 * socksize;
817 		ifa = ifa_alloc(ifasize, M_WAITOK);
818 		sdl = (struct sockaddr_dl *)(ifa + 1);
819 		sdl->sdl_len = socksize;
820 		sdl->sdl_family = AF_LINK;
821 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
822 		sdl->sdl_nlen = namelen;
823 		sdl->sdl_index = ifp->if_index;
824 		sdl->sdl_type = ifp->if_type;
825 		ifp->if_addr = ifa;
826 		ifa->ifa_ifp = ifp;
827 		ifa->ifa_rtrequest = link_rtrequest;
828 		ifa->ifa_addr = (struct sockaddr *)sdl;
829 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
830 		ifa->ifa_netmask = (struct sockaddr *)sdl;
831 		sdl->sdl_len = masklen;
832 		while (namelen != 0)
833 			sdl->sdl_data[--namelen] = 0xff;
834 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
835 		/* Reliably crash if used uninitialized. */
836 		ifp->if_broadcastaddr = NULL;
837 
838 		if (ifp->if_type == IFT_ETHER) {
839 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
840 			    M_WAITOK | M_ZERO);
841 		}
842 
843 #if defined(INET) || defined(INET6)
844 		/* Use defaults for TSO, if nothing is set */
845 		if (ifp->if_hw_tsomax == 0 &&
846 		    ifp->if_hw_tsomaxsegcount == 0 &&
847 		    ifp->if_hw_tsomaxsegsize == 0) {
848 			/*
849 			 * The TSO defaults needs to be such that an
850 			 * NFS mbuf list of 35 mbufs totalling just
851 			 * below 64K works and that a chain of mbufs
852 			 * can be defragged into at most 32 segments:
853 			 */
854 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
855 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
856 			ifp->if_hw_tsomaxsegcount = 35;
857 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
858 
859 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
860 			if (ifp->if_capabilities & IFCAP_TSO) {
861 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
862 				    ifp->if_hw_tsomax,
863 				    ifp->if_hw_tsomaxsegcount,
864 				    ifp->if_hw_tsomaxsegsize);
865 			}
866 		}
867 #endif
868 	}
869 #ifdef VIMAGE
870 	else {
871 		/*
872 		 * Update the interface index in the link layer address
873 		 * of the interface.
874 		 */
875 		for (ifa = ifp->if_addr; ifa != NULL;
876 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
877 			if (ifa->ifa_addr->sa_family == AF_LINK) {
878 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
879 				sdl->sdl_index = ifp->if_index;
880 			}
881 		}
882 	}
883 #endif
884 
885 	IFNET_WLOCK();
886 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
887 #ifdef VIMAGE
888 	curvnet->vnet_ifcnt++;
889 #endif
890 	IFNET_WUNLOCK();
891 
892 	if (domain_init_status >= 2)
893 		if_attachdomain1(ifp);
894 
895 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
896 	if (IS_DEFAULT_VNET(curvnet))
897 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
898 
899 	/* Announce the interface. */
900 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
901 }
902 
903 static void
904 if_epochalloc(void *dummy __unused)
905 {
906 
907 	net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT);
908 	net_epoch = epoch_alloc(0);
909 }
910 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
911     if_epochalloc, NULL);
912 
913 static void
914 if_attachdomain(void *dummy)
915 {
916 	struct ifnet *ifp;
917 
918 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
919 		if_attachdomain1(ifp);
920 }
921 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
922     if_attachdomain, NULL);
923 
924 static void
925 if_attachdomain1(struct ifnet *ifp)
926 {
927 	struct domain *dp;
928 
929 	/*
930 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
931 	 * cannot lock ifp->if_afdata initialization, entirely.
932 	 */
933 	IF_AFDATA_LOCK(ifp);
934 	if (ifp->if_afdata_initialized >= domain_init_status) {
935 		IF_AFDATA_UNLOCK(ifp);
936 		log(LOG_WARNING, "%s called more than once on %s\n",
937 		    __func__, ifp->if_xname);
938 		return;
939 	}
940 	ifp->if_afdata_initialized = domain_init_status;
941 	IF_AFDATA_UNLOCK(ifp);
942 
943 	/* address family dependent data region */
944 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
945 	for (dp = domains; dp; dp = dp->dom_next) {
946 		if (dp->dom_ifattach)
947 			ifp->if_afdata[dp->dom_family] =
948 			    (*dp->dom_ifattach)(ifp);
949 	}
950 }
951 
952 /*
953  * Remove any unicast or broadcast network addresses from an interface.
954  */
955 void
956 if_purgeaddrs(struct ifnet *ifp)
957 {
958 	struct ifaddr *ifa, *next;
959 
960 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
961 	CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
962 		if (ifa->ifa_addr->sa_family == AF_LINK)
963 			continue;
964 #ifdef INET
965 		/* XXX: Ugly!! ad hoc just for INET */
966 		if (ifa->ifa_addr->sa_family == AF_INET) {
967 			struct ifaliasreq ifr;
968 
969 			bzero(&ifr, sizeof(ifr));
970 			ifr.ifra_addr = *ifa->ifa_addr;
971 			if (ifa->ifa_dstaddr)
972 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
973 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
974 			    NULL) == 0)
975 				continue;
976 		}
977 #endif /* INET */
978 #ifdef INET6
979 		if (ifa->ifa_addr->sa_family == AF_INET6) {
980 			in6_purgeaddr(ifa);
981 			/* ifp_addrhead is already updated */
982 			continue;
983 		}
984 #endif /* INET6 */
985 		IF_ADDR_WLOCK(ifp);
986 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
987 		IF_ADDR_WUNLOCK(ifp);
988 		ifa_free(ifa);
989 	}
990 }
991 
992 /*
993  * Remove any multicast network addresses from an interface when an ifnet
994  * is going away.
995  */
996 static void
997 if_purgemaddrs(struct ifnet *ifp)
998 {
999 	struct ifmultiaddr *ifma;
1000 
1001 	IF_ADDR_WLOCK(ifp);
1002 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1003 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1004 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1005 		if_delmulti_locked(ifp, ifma, 1);
1006 	}
1007 	IF_ADDR_WUNLOCK(ifp);
1008 }
1009 
1010 /*
1011  * Detach an interface, removing it from the list of "active" interfaces.
1012  * If vmove flag is set on entry to if_detach_internal(), perform only a
1013  * limited subset of cleanup tasks, given that we are moving an ifnet from
1014  * one vnet to another, where it must be fully operational.
1015  *
1016  * XXXRW: There are some significant questions about event ordering, and
1017  * how to prevent things from starting to use the interface during detach.
1018  */
1019 void
1020 if_detach(struct ifnet *ifp)
1021 {
1022 
1023 	CURVNET_SET_QUIET(ifp->if_vnet);
1024 	if_detach_internal(ifp, 0, NULL);
1025 	CURVNET_RESTORE();
1026 }
1027 
1028 /*
1029  * The vmove flag, if set, indicates that we are called from a callpath
1030  * that is moving an interface to a different vnet instance.
1031  *
1032  * The shutdown flag, if set, indicates that we are called in the
1033  * process of shutting down a vnet instance.  Currently only the
1034  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1035  * on a vnet instance shutdown without this flag being set, e.g., when
1036  * the cloned interfaces are destoyed as first thing of teardown.
1037  */
1038 static int
1039 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1040 {
1041 	struct ifaddr *ifa;
1042 	int i;
1043 	struct domain *dp;
1044  	struct ifnet *iter;
1045  	int found = 0;
1046 #ifdef VIMAGE
1047 	int shutdown;
1048 
1049 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1050 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1051 #endif
1052 	IFNET_WLOCK();
1053 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
1054 		if (iter == ifp) {
1055 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
1056 			found = 1;
1057 			break;
1058 		}
1059 	IFNET_WUNLOCK();
1060 	if (!found) {
1061 		/*
1062 		 * While we would want to panic here, we cannot
1063 		 * guarantee that the interface is indeed still on
1064 		 * the list given we don't hold locks all the way.
1065 		 */
1066 		return (ENOENT);
1067 #if 0
1068 		if (vmove)
1069 			panic("%s: ifp=%p not on the ifnet tailq %p",
1070 			    __func__, ifp, &V_ifnet);
1071 		else
1072 			return; /* XXX this should panic as well? */
1073 #endif
1074 	}
1075 
1076 	/*
1077 	 * At this point we know the interface still was on the ifnet list
1078 	 * and we removed it so we are in a stable state.
1079 	 */
1080 #ifdef VIMAGE
1081 	curvnet->vnet_ifcnt--;
1082 #endif
1083 
1084 	/*
1085 	 * In any case (destroy or vmove) detach us from the groups
1086 	 * and remove/wait for pending events on the taskq.
1087 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1088 	 */
1089 	if_delgroups(ifp);
1090 
1091 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1092 
1093 	/*
1094 	 * Check if this is a cloned interface or not. Must do even if
1095 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1096 	 * the if_clone_addgroup() will have a corrupted string overwise
1097 	 * from a gibberish pointer.
1098 	 */
1099 	if (vmove && ifcp != NULL)
1100 		*ifcp = if_clone_findifc(ifp);
1101 
1102 	if_down(ifp);
1103 
1104 #ifdef VIMAGE
1105 	/*
1106 	 * On VNET shutdown abort here as the stack teardown will do all
1107 	 * the work top-down for us.
1108 	 */
1109 	if (shutdown) {
1110 		/*
1111 		 * In case of a vmove we are done here without error.
1112 		 * If we would signal an error it would lead to the same
1113 		 * abort as if we did not find the ifnet anymore.
1114 		 * if_detach() calls us in void context and does not care
1115 		 * about an early abort notification, so life is splendid :)
1116 		 */
1117 		goto finish_vnet_shutdown;
1118 	}
1119 #endif
1120 
1121 	/*
1122 	 * At this point we are not tearing down a VNET and are either
1123 	 * going to destroy or vmove the interface and have to cleanup
1124 	 * accordingly.
1125 	 */
1126 
1127 	/*
1128 	 * Remove routes and flush queues.
1129 	 */
1130 #ifdef ALTQ
1131 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1132 		altq_disable(&ifp->if_snd);
1133 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1134 		altq_detach(&ifp->if_snd);
1135 #endif
1136 
1137 	if_purgeaddrs(ifp);
1138 
1139 #ifdef INET
1140 	in_ifdetach(ifp);
1141 #endif
1142 
1143 #ifdef INET6
1144 	/*
1145 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1146 	 * before removing routing entries below, since IPv6 interface direct
1147 	 * routes are expected to be removed by the IPv6-specific kernel API.
1148 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1149 	 */
1150 	in6_ifdetach(ifp);
1151 #endif
1152 	if_purgemaddrs(ifp);
1153 
1154 	/* Announce that the interface is gone. */
1155 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1156 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1157 	if (IS_DEFAULT_VNET(curvnet))
1158 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1159 
1160 	if (!vmove) {
1161 		/*
1162 		 * Prevent further calls into the device driver via ifnet.
1163 		 */
1164 		if_dead(ifp);
1165 
1166 		/*
1167 		 * Remove link ifaddr pointer and maybe decrement if_index.
1168 		 * Clean up all addresses.
1169 		 */
1170 		free(ifp->if_hw_addr, M_IFADDR);
1171 		ifp->if_hw_addr = NULL;
1172 		ifp->if_addr = NULL;
1173 
1174 		/* We can now free link ifaddr. */
1175 		IF_ADDR_WLOCK(ifp);
1176 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1177 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1178 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1179 			IF_ADDR_WUNLOCK(ifp);
1180 			ifa_free(ifa);
1181 		} else
1182 			IF_ADDR_WUNLOCK(ifp);
1183 	}
1184 
1185 	rt_flushifroutes(ifp);
1186 
1187 #ifdef VIMAGE
1188 finish_vnet_shutdown:
1189 #endif
1190 	/*
1191 	 * We cannot hold the lock over dom_ifdetach calls as they might
1192 	 * sleep, for example trying to drain a callout, thus open up the
1193 	 * theoretical race with re-attaching.
1194 	 */
1195 	IF_AFDATA_LOCK(ifp);
1196 	i = ifp->if_afdata_initialized;
1197 	ifp->if_afdata_initialized = 0;
1198 	IF_AFDATA_UNLOCK(ifp);
1199 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1200 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1201 			(*dp->dom_ifdetach)(ifp,
1202 			    ifp->if_afdata[dp->dom_family]);
1203 			ifp->if_afdata[dp->dom_family] = NULL;
1204 		}
1205 	}
1206 
1207 	return (0);
1208 }
1209 
1210 #ifdef VIMAGE
1211 /*
1212  * if_vmove() performs a limited version of if_detach() in current
1213  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1214  * An attempt is made to shrink if_index in current vnet, find an
1215  * unused if_index in target vnet and calls if_grow() if necessary,
1216  * and finally find an unused if_xname for the target vnet.
1217  */
1218 static void
1219 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1220 {
1221 	struct if_clone *ifc;
1222 	u_int bif_dlt, bif_hdrlen;
1223 	int rc;
1224 
1225  	/*
1226 	 * if_detach_internal() will call the eventhandler to notify
1227 	 * interface departure.  That will detach if_bpf.  We need to
1228 	 * safe the dlt and hdrlen so we can re-attach it later.
1229 	 */
1230 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1231 
1232 	/*
1233 	 * Detach from current vnet, but preserve LLADDR info, do not
1234 	 * mark as dead etc. so that the ifnet can be reattached later.
1235 	 * If we cannot find it, we lost the race to someone else.
1236 	 */
1237 	rc = if_detach_internal(ifp, 1, &ifc);
1238 	if (rc != 0)
1239 		return;
1240 
1241 	/*
1242 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1243 	 * the if_index for that vnet if possible.
1244 	 *
1245 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1246 	 * or we'd lock on one vnet and unlock on another.
1247 	 */
1248 	IFNET_WLOCK();
1249 	ifindex_free_locked(ifp->if_index);
1250 	IFNET_WUNLOCK();
1251 
1252 	/*
1253 	 * Perform interface-specific reassignment tasks, if provided by
1254 	 * the driver.
1255 	 */
1256 	if (ifp->if_reassign != NULL)
1257 		ifp->if_reassign(ifp, new_vnet, NULL);
1258 
1259 	/*
1260 	 * Switch to the context of the target vnet.
1261 	 */
1262 	CURVNET_SET_QUIET(new_vnet);
1263 
1264 	IFNET_WLOCK();
1265 	ifp->if_index = ifindex_alloc();
1266 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1267 	IFNET_WUNLOCK();
1268 
1269 	if_attach_internal(ifp, 1, ifc);
1270 
1271 	if (ifp->if_bpf == NULL)
1272 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1273 
1274 	CURVNET_RESTORE();
1275 }
1276 
1277 /*
1278  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1279  */
1280 static int
1281 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1282 {
1283 	struct prison *pr;
1284 	struct ifnet *difp;
1285 	int shutdown;
1286 
1287 	/* Try to find the prison within our visibility. */
1288 	sx_slock(&allprison_lock);
1289 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1290 	sx_sunlock(&allprison_lock);
1291 	if (pr == NULL)
1292 		return (ENXIO);
1293 	prison_hold_locked(pr);
1294 	mtx_unlock(&pr->pr_mtx);
1295 
1296 	/* Do not try to move the iface from and to the same prison. */
1297 	if (pr->pr_vnet == ifp->if_vnet) {
1298 		prison_free(pr);
1299 		return (EEXIST);
1300 	}
1301 
1302 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1303 	/* XXX Lock interfaces to avoid races. */
1304 	CURVNET_SET_QUIET(pr->pr_vnet);
1305 	difp = ifunit(ifname);
1306 	if (difp != NULL) {
1307 		CURVNET_RESTORE();
1308 		prison_free(pr);
1309 		return (EEXIST);
1310 	}
1311 
1312 	/* Make sure the VNET is stable. */
1313 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1314 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1315 	if (shutdown) {
1316 		CURVNET_RESTORE();
1317 		prison_free(pr);
1318 		return (EBUSY);
1319 	}
1320 	CURVNET_RESTORE();
1321 
1322 	/* Move the interface into the child jail/vnet. */
1323 	if_vmove(ifp, pr->pr_vnet);
1324 
1325 	/* Report the new if_xname back to the userland. */
1326 	sprintf(ifname, "%s", ifp->if_xname);
1327 
1328 	prison_free(pr);
1329 	return (0);
1330 }
1331 
1332 static int
1333 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1334 {
1335 	struct prison *pr;
1336 	struct vnet *vnet_dst;
1337 	struct ifnet *ifp;
1338  	int shutdown;
1339 
1340 	/* Try to find the prison within our visibility. */
1341 	sx_slock(&allprison_lock);
1342 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1343 	sx_sunlock(&allprison_lock);
1344 	if (pr == NULL)
1345 		return (ENXIO);
1346 	prison_hold_locked(pr);
1347 	mtx_unlock(&pr->pr_mtx);
1348 
1349 	/* Make sure the named iface exists in the source prison/vnet. */
1350 	CURVNET_SET(pr->pr_vnet);
1351 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1352 	if (ifp == NULL) {
1353 		CURVNET_RESTORE();
1354 		prison_free(pr);
1355 		return (ENXIO);
1356 	}
1357 
1358 	/* Do not try to move the iface from and to the same prison. */
1359 	vnet_dst = TD_TO_VNET(td);
1360 	if (vnet_dst == ifp->if_vnet) {
1361 		CURVNET_RESTORE();
1362 		prison_free(pr);
1363 		return (EEXIST);
1364 	}
1365 
1366 	/* Make sure the VNET is stable. */
1367 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1368 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1369 	if (shutdown) {
1370 		CURVNET_RESTORE();
1371 		prison_free(pr);
1372 		return (EBUSY);
1373 	}
1374 
1375 	/* Get interface back from child jail/vnet. */
1376 	if_vmove(ifp, vnet_dst);
1377 	CURVNET_RESTORE();
1378 
1379 	/* Report the new if_xname back to the userland. */
1380 	sprintf(ifname, "%s", ifp->if_xname);
1381 
1382 	prison_free(pr);
1383 	return (0);
1384 }
1385 #endif /* VIMAGE */
1386 
1387 /*
1388  * Add a group to an interface
1389  */
1390 int
1391 if_addgroup(struct ifnet *ifp, const char *groupname)
1392 {
1393 	struct ifg_list		*ifgl;
1394 	struct ifg_group	*ifg = NULL;
1395 	struct ifg_member	*ifgm;
1396 	int 			 new = 0;
1397 
1398 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1399 	    groupname[strlen(groupname) - 1] <= '9')
1400 		return (EINVAL);
1401 
1402 	IFNET_WLOCK();
1403 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1404 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1405 			IFNET_WUNLOCK();
1406 			return (EEXIST);
1407 		}
1408 
1409 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1410 	    M_NOWAIT)) == NULL) {
1411 	    	IFNET_WUNLOCK();
1412 		return (ENOMEM);
1413 	}
1414 
1415 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1416 	    M_TEMP, M_NOWAIT)) == NULL) {
1417 		free(ifgl, M_TEMP);
1418 		IFNET_WUNLOCK();
1419 		return (ENOMEM);
1420 	}
1421 
1422 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1423 		if (!strcmp(ifg->ifg_group, groupname))
1424 			break;
1425 
1426 	if (ifg == NULL) {
1427 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1428 		    M_TEMP, M_NOWAIT)) == NULL) {
1429 			free(ifgl, M_TEMP);
1430 			free(ifgm, M_TEMP);
1431 			IFNET_WUNLOCK();
1432 			return (ENOMEM);
1433 		}
1434 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1435 		ifg->ifg_refcnt = 0;
1436 		TAILQ_INIT(&ifg->ifg_members);
1437 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1438 		new = 1;
1439 	}
1440 
1441 	ifg->ifg_refcnt++;
1442 	ifgl->ifgl_group = ifg;
1443 	ifgm->ifgm_ifp = ifp;
1444 
1445 	IF_ADDR_WLOCK(ifp);
1446 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1447 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1448 	IF_ADDR_WUNLOCK(ifp);
1449 
1450 	IFNET_WUNLOCK();
1451 
1452 	if (new)
1453 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1454 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1455 
1456 	return (0);
1457 }
1458 
1459 /*
1460  * Remove a group from an interface
1461  */
1462 int
1463 if_delgroup(struct ifnet *ifp, const char *groupname)
1464 {
1465 	struct ifg_list		*ifgl;
1466 	struct ifg_member	*ifgm;
1467 
1468 	IFNET_WLOCK();
1469 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1470 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1471 			break;
1472 	if (ifgl == NULL) {
1473 		IFNET_WUNLOCK();
1474 		return (ENOENT);
1475 	}
1476 
1477 	IF_ADDR_WLOCK(ifp);
1478 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1479 	IF_ADDR_WUNLOCK(ifp);
1480 
1481 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1482 		if (ifgm->ifgm_ifp == ifp)
1483 			break;
1484 
1485 	if (ifgm != NULL) {
1486 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1487 		free(ifgm, M_TEMP);
1488 	}
1489 
1490 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1491 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1492 		IFNET_WUNLOCK();
1493 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1494 		free(ifgl->ifgl_group, M_TEMP);
1495 	} else
1496 		IFNET_WUNLOCK();
1497 
1498 	free(ifgl, M_TEMP);
1499 
1500 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1501 
1502 	return (0);
1503 }
1504 
1505 /*
1506  * Remove an interface from all groups
1507  */
1508 static void
1509 if_delgroups(struct ifnet *ifp)
1510 {
1511 	struct ifg_list		*ifgl;
1512 	struct ifg_member	*ifgm;
1513 	char groupname[IFNAMSIZ];
1514 
1515 	IFNET_WLOCK();
1516 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1517 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1518 
1519 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1520 
1521 		IF_ADDR_WLOCK(ifp);
1522 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1523 		IF_ADDR_WUNLOCK(ifp);
1524 
1525 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1526 			if (ifgm->ifgm_ifp == ifp)
1527 				break;
1528 
1529 		if (ifgm != NULL) {
1530 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1531 			    ifgm_next);
1532 			free(ifgm, M_TEMP);
1533 		}
1534 
1535 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1536 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1537 			IFNET_WUNLOCK();
1538 			EVENTHANDLER_INVOKE(group_detach_event,
1539 			    ifgl->ifgl_group);
1540 			free(ifgl->ifgl_group, M_TEMP);
1541 		} else
1542 			IFNET_WUNLOCK();
1543 
1544 		free(ifgl, M_TEMP);
1545 
1546 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1547 
1548 		IFNET_WLOCK();
1549 	}
1550 	IFNET_WUNLOCK();
1551 }
1552 
1553 static char *
1554 ifgr_group_get(void *ifgrp)
1555 {
1556 	union ifgroupreq_union *ifgrup;
1557 
1558 	ifgrup = ifgrp;
1559 #ifdef COMPAT_FREEBSD32
1560 	if (SV_CURPROC_FLAG(SV_ILP32))
1561 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1562 #endif
1563 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1564 }
1565 
1566 static struct ifg_req *
1567 ifgr_groups_get(void *ifgrp)
1568 {
1569 	union ifgroupreq_union *ifgrup;
1570 
1571 	ifgrup = ifgrp;
1572 #ifdef COMPAT_FREEBSD32
1573 	if (SV_CURPROC_FLAG(SV_ILP32))
1574 		return ((struct ifg_req *)(uintptr_t)
1575 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1576 #endif
1577 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1578 }
1579 
1580 /*
1581  * Stores all groups from an interface in memory pointed to by ifgr.
1582  */
1583 static int
1584 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1585 {
1586 	int			 len, error;
1587 	struct ifg_list		*ifgl;
1588 	struct ifg_req		 ifgrq, *ifgp;
1589 
1590 	if (ifgr->ifgr_len == 0) {
1591 		IF_ADDR_RLOCK(ifp);
1592 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1593 			ifgr->ifgr_len += sizeof(struct ifg_req);
1594 		IF_ADDR_RUNLOCK(ifp);
1595 		return (0);
1596 	}
1597 
1598 	len = ifgr->ifgr_len;
1599 	ifgp = ifgr_groups_get(ifgr);
1600 	/* XXX: wire */
1601 	IF_ADDR_RLOCK(ifp);
1602 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1603 		if (len < sizeof(ifgrq)) {
1604 			IF_ADDR_RUNLOCK(ifp);
1605 			return (EINVAL);
1606 		}
1607 		bzero(&ifgrq, sizeof ifgrq);
1608 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1609 		    sizeof(ifgrq.ifgrq_group));
1610 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1611 		    	IF_ADDR_RUNLOCK(ifp);
1612 			return (error);
1613 		}
1614 		len -= sizeof(ifgrq);
1615 		ifgp++;
1616 	}
1617 	IF_ADDR_RUNLOCK(ifp);
1618 
1619 	return (0);
1620 }
1621 
1622 /*
1623  * Stores all members of a group in memory pointed to by igfr
1624  */
1625 static int
1626 if_getgroupmembers(struct ifgroupreq *ifgr)
1627 {
1628 	struct ifg_group	*ifg;
1629 	struct ifg_member	*ifgm;
1630 	struct ifg_req		 ifgrq, *ifgp;
1631 	int			 len, error;
1632 
1633 	IFNET_RLOCK();
1634 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1635 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1636 			break;
1637 	if (ifg == NULL) {
1638 		IFNET_RUNLOCK();
1639 		return (ENOENT);
1640 	}
1641 
1642 	if (ifgr->ifgr_len == 0) {
1643 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1644 			ifgr->ifgr_len += sizeof(ifgrq);
1645 		IFNET_RUNLOCK();
1646 		return (0);
1647 	}
1648 
1649 	len = ifgr->ifgr_len;
1650 	ifgp = ifgr_groups_get(ifgr);
1651 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1652 		if (len < sizeof(ifgrq)) {
1653 			IFNET_RUNLOCK();
1654 			return (EINVAL);
1655 		}
1656 		bzero(&ifgrq, sizeof ifgrq);
1657 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1658 		    sizeof(ifgrq.ifgrq_member));
1659 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1660 			IFNET_RUNLOCK();
1661 			return (error);
1662 		}
1663 		len -= sizeof(ifgrq);
1664 		ifgp++;
1665 	}
1666 	IFNET_RUNLOCK();
1667 
1668 	return (0);
1669 }
1670 
1671 /*
1672  * Return counter values from counter(9)s stored in ifnet.
1673  */
1674 uint64_t
1675 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1676 {
1677 
1678 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1679 
1680 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1681 }
1682 
1683 /*
1684  * Increase an ifnet counter. Usually used for counters shared
1685  * between the stack and a driver, but function supports them all.
1686  */
1687 void
1688 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1689 {
1690 
1691 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1692 
1693 	counter_u64_add(ifp->if_counters[cnt], inc);
1694 }
1695 
1696 /*
1697  * Copy data from ifnet to userland API structure if_data.
1698  */
1699 void
1700 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1701 {
1702 
1703 	ifd->ifi_type = ifp->if_type;
1704 	ifd->ifi_physical = 0;
1705 	ifd->ifi_addrlen = ifp->if_addrlen;
1706 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1707 	ifd->ifi_link_state = ifp->if_link_state;
1708 	ifd->ifi_vhid = 0;
1709 	ifd->ifi_datalen = sizeof(struct if_data);
1710 	ifd->ifi_mtu = ifp->if_mtu;
1711 	ifd->ifi_metric = ifp->if_metric;
1712 	ifd->ifi_baudrate = ifp->if_baudrate;
1713 	ifd->ifi_hwassist = ifp->if_hwassist;
1714 	ifd->ifi_epoch = ifp->if_epoch;
1715 	ifd->ifi_lastchange = ifp->if_lastchange;
1716 
1717 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1718 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1719 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1720 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1721 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1722 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1723 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1724 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1725 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1726 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1727 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1728 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1729 }
1730 
1731 /*
1732  * Wrapper functions for struct ifnet address list locking macros.  These are
1733  * used by kernel modules to avoid encoding programming interface or binary
1734  * interface assumptions that may be violated when kernel-internal locking
1735  * approaches change.
1736  */
1737 void
1738 if_addr_rlock(struct ifnet *ifp)
1739 {
1740 
1741 	IF_ADDR_RLOCK(ifp);
1742 }
1743 
1744 void
1745 if_addr_runlock(struct ifnet *ifp)
1746 {
1747 
1748 	IF_ADDR_RUNLOCK(ifp);
1749 }
1750 
1751 void
1752 if_maddr_rlock(if_t ifp)
1753 {
1754 
1755 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1756 }
1757 
1758 void
1759 if_maddr_runlock(if_t ifp)
1760 {
1761 
1762 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1763 }
1764 
1765 /*
1766  * Initialization, destruction and refcounting functions for ifaddrs.
1767  */
1768 struct ifaddr *
1769 ifa_alloc(size_t size, int flags)
1770 {
1771 	struct ifaddr *ifa;
1772 
1773 	KASSERT(size >= sizeof(struct ifaddr),
1774 	    ("%s: invalid size %zu", __func__, size));
1775 
1776 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1777 	if (ifa == NULL)
1778 		return (NULL);
1779 
1780 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1781 		goto fail;
1782 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1783 		goto fail;
1784 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1785 		goto fail;
1786 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1787 		goto fail;
1788 
1789 	refcount_init(&ifa->ifa_refcnt, 1);
1790 
1791 	return (ifa);
1792 
1793 fail:
1794 	/* free(NULL) is okay */
1795 	counter_u64_free(ifa->ifa_opackets);
1796 	counter_u64_free(ifa->ifa_ipackets);
1797 	counter_u64_free(ifa->ifa_obytes);
1798 	counter_u64_free(ifa->ifa_ibytes);
1799 	free(ifa, M_IFADDR);
1800 
1801 	return (NULL);
1802 }
1803 
1804 void
1805 ifa_ref(struct ifaddr *ifa)
1806 {
1807 
1808 	refcount_acquire(&ifa->ifa_refcnt);
1809 }
1810 
1811 static void
1812 ifa_destroy(epoch_context_t ctx)
1813 {
1814 	struct ifaddr *ifa;
1815 
1816 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1817 	counter_u64_free(ifa->ifa_opackets);
1818 	counter_u64_free(ifa->ifa_ipackets);
1819 	counter_u64_free(ifa->ifa_obytes);
1820 	counter_u64_free(ifa->ifa_ibytes);
1821 	free(ifa, M_IFADDR);
1822 }
1823 
1824 void
1825 ifa_free(struct ifaddr *ifa)
1826 {
1827 
1828 	if (refcount_release(&ifa->ifa_refcnt))
1829 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1830 }
1831 
1832 
1833 static int
1834 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1835     struct sockaddr *ia)
1836 {
1837 	int error;
1838 	struct rt_addrinfo info;
1839 	struct sockaddr_dl null_sdl;
1840 	struct ifnet *ifp;
1841 
1842 	ifp = ifa->ifa_ifp;
1843 
1844 	bzero(&info, sizeof(info));
1845 	if (cmd != RTM_DELETE)
1846 		info.rti_ifp = V_loif;
1847 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1848 	info.rti_info[RTAX_DST] = ia;
1849 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1850 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1851 
1852 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1853 
1854 	if (error != 0 &&
1855 	    !(cmd == RTM_ADD && error == EEXIST) &&
1856 	    !(cmd == RTM_DELETE && error == ENOENT))
1857 		if_printf(ifp, "%s failed: %d\n", otype, error);
1858 
1859 	return (error);
1860 }
1861 
1862 int
1863 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1864 {
1865 
1866 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1867 }
1868 
1869 int
1870 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1871 {
1872 
1873 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1874 }
1875 
1876 int
1877 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1878 {
1879 
1880 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1881 }
1882 
1883 /*
1884  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1885  * structs used to represent other address families, it is necessary
1886  * to perform a different comparison.
1887  */
1888 
1889 #define	sa_dl_equal(a1, a2)	\
1890 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1891 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1892 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1893 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1894 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1895 
1896 /*
1897  * Locate an interface based on a complete address.
1898  */
1899 /*ARGSUSED*/
1900 static struct ifaddr *
1901 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1902 {
1903 	struct ifnet *ifp;
1904 	struct ifaddr *ifa;
1905 
1906 	IFNET_RLOCK_NOSLEEP();
1907 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1908 		IF_ADDR_RLOCK(ifp);
1909 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1910 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1911 				continue;
1912 			if (sa_equal(addr, ifa->ifa_addr)) {
1913 				if (getref)
1914 					ifa_ref(ifa);
1915 				IF_ADDR_RUNLOCK(ifp);
1916 				goto done;
1917 			}
1918 			/* IP6 doesn't have broadcast */
1919 			if ((ifp->if_flags & IFF_BROADCAST) &&
1920 			    ifa->ifa_broadaddr &&
1921 			    ifa->ifa_broadaddr->sa_len != 0 &&
1922 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1923 				if (getref)
1924 					ifa_ref(ifa);
1925 				IF_ADDR_RUNLOCK(ifp);
1926 				goto done;
1927 			}
1928 		}
1929 		IF_ADDR_RUNLOCK(ifp);
1930 	}
1931 	ifa = NULL;
1932 done:
1933 	IFNET_RUNLOCK_NOSLEEP();
1934 	return (ifa);
1935 }
1936 
1937 struct ifaddr *
1938 ifa_ifwithaddr(const struct sockaddr *addr)
1939 {
1940 
1941 	return (ifa_ifwithaddr_internal(addr, 1));
1942 }
1943 
1944 int
1945 ifa_ifwithaddr_check(const struct sockaddr *addr)
1946 {
1947 
1948 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1949 }
1950 
1951 /*
1952  * Locate an interface based on the broadcast address.
1953  */
1954 /* ARGSUSED */
1955 struct ifaddr *
1956 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1957 {
1958 	struct ifnet *ifp;
1959 	struct ifaddr *ifa;
1960 
1961 	IFNET_RLOCK_NOSLEEP();
1962 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1963 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1964 			continue;
1965 		IF_ADDR_RLOCK(ifp);
1966 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1967 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1968 				continue;
1969 			if ((ifp->if_flags & IFF_BROADCAST) &&
1970 			    ifa->ifa_broadaddr &&
1971 			    ifa->ifa_broadaddr->sa_len != 0 &&
1972 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1973 				ifa_ref(ifa);
1974 				IF_ADDR_RUNLOCK(ifp);
1975 				goto done;
1976 			}
1977 		}
1978 		IF_ADDR_RUNLOCK(ifp);
1979 	}
1980 	ifa = NULL;
1981 done:
1982 	IFNET_RUNLOCK_NOSLEEP();
1983 	return (ifa);
1984 }
1985 
1986 /*
1987  * Locate the point to point interface with a given destination address.
1988  */
1989 /*ARGSUSED*/
1990 struct ifaddr *
1991 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1992 {
1993 	struct ifnet *ifp;
1994 	struct ifaddr *ifa;
1995 
1996 	IFNET_RLOCK_NOSLEEP();
1997 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1998 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1999 			continue;
2000 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2001 			continue;
2002 		IF_ADDR_RLOCK(ifp);
2003 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2004 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2005 				continue;
2006 			if (ifa->ifa_dstaddr != NULL &&
2007 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2008 				ifa_ref(ifa);
2009 				IF_ADDR_RUNLOCK(ifp);
2010 				goto done;
2011 			}
2012 		}
2013 		IF_ADDR_RUNLOCK(ifp);
2014 	}
2015 	ifa = NULL;
2016 done:
2017 	IFNET_RUNLOCK_NOSLEEP();
2018 	return (ifa);
2019 }
2020 
2021 /*
2022  * Find an interface on a specific network.  If many, choice
2023  * is most specific found.
2024  */
2025 struct ifaddr *
2026 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2027 {
2028 	struct ifnet *ifp;
2029 	struct ifaddr *ifa;
2030 	struct ifaddr *ifa_maybe = NULL;
2031 	u_int af = addr->sa_family;
2032 	const char *addr_data = addr->sa_data, *cplim;
2033 
2034 	/*
2035 	 * AF_LINK addresses can be looked up directly by their index number,
2036 	 * so do that if we can.
2037 	 */
2038 	if (af == AF_LINK) {
2039 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2040 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2041 		return (ifaddr_byindex(sdl->sdl_index));
2042 	}
2043 
2044 	/*
2045 	 * Scan though each interface, looking for ones that have addresses
2046 	 * in this address family and the requested fib.  Maintain a reference
2047 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2048 	 * kept it stable when we move onto the next interface.
2049 	 */
2050 	IFNET_RLOCK_NOSLEEP();
2051 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2052 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2053 			continue;
2054 		IF_ADDR_RLOCK(ifp);
2055 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2056 			const char *cp, *cp2, *cp3;
2057 
2058 			if (ifa->ifa_addr->sa_family != af)
2059 next:				continue;
2060 			if (af == AF_INET &&
2061 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2062 				/*
2063 				 * This is a bit broken as it doesn't
2064 				 * take into account that the remote end may
2065 				 * be a single node in the network we are
2066 				 * looking for.
2067 				 * The trouble is that we don't know the
2068 				 * netmask for the remote end.
2069 				 */
2070 				if (ifa->ifa_dstaddr != NULL &&
2071 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2072 					ifa_ref(ifa);
2073 					IF_ADDR_RUNLOCK(ifp);
2074 					goto done;
2075 				}
2076 			} else {
2077 				/*
2078 				 * Scan all the bits in the ifa's address.
2079 				 * If a bit dissagrees with what we are
2080 				 * looking for, mask it with the netmask
2081 				 * to see if it really matters.
2082 				 * (A byte at a time)
2083 				 */
2084 				if (ifa->ifa_netmask == 0)
2085 					continue;
2086 				cp = addr_data;
2087 				cp2 = ifa->ifa_addr->sa_data;
2088 				cp3 = ifa->ifa_netmask->sa_data;
2089 				cplim = ifa->ifa_netmask->sa_len
2090 					+ (char *)ifa->ifa_netmask;
2091 				while (cp3 < cplim)
2092 					if ((*cp++ ^ *cp2++) & *cp3++)
2093 						goto next; /* next address! */
2094 				/*
2095 				 * If the netmask of what we just found
2096 				 * is more specific than what we had before
2097 				 * (if we had one), or if the virtual status
2098 				 * of new prefix is better than of the old one,
2099 				 * then remember the new one before continuing
2100 				 * to search for an even better one.
2101 				 */
2102 				if (ifa_maybe == NULL ||
2103 				    ifa_preferred(ifa_maybe, ifa) ||
2104 				    rn_refines((caddr_t)ifa->ifa_netmask,
2105 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2106 					if (ifa_maybe != NULL)
2107 						ifa_free(ifa_maybe);
2108 					ifa_maybe = ifa;
2109 					ifa_ref(ifa_maybe);
2110 				}
2111 			}
2112 		}
2113 		IF_ADDR_RUNLOCK(ifp);
2114 	}
2115 	ifa = ifa_maybe;
2116 	ifa_maybe = NULL;
2117 done:
2118 	IFNET_RUNLOCK_NOSLEEP();
2119 	if (ifa_maybe != NULL)
2120 		ifa_free(ifa_maybe);
2121 	return (ifa);
2122 }
2123 
2124 /*
2125  * Find an interface address specific to an interface best matching
2126  * a given address.
2127  */
2128 struct ifaddr *
2129 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2130 {
2131 	struct ifaddr *ifa;
2132 	const char *cp, *cp2, *cp3;
2133 	char *cplim;
2134 	struct ifaddr *ifa_maybe = NULL;
2135 	u_int af = addr->sa_family;
2136 
2137 	if (af >= AF_MAX)
2138 		return (NULL);
2139 	IF_ADDR_RLOCK(ifp);
2140 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2141 		if (ifa->ifa_addr->sa_family != af)
2142 			continue;
2143 		if (ifa_maybe == NULL)
2144 			ifa_maybe = ifa;
2145 		if (ifa->ifa_netmask == 0) {
2146 			if (sa_equal(addr, ifa->ifa_addr) ||
2147 			    (ifa->ifa_dstaddr &&
2148 			    sa_equal(addr, ifa->ifa_dstaddr)))
2149 				goto done;
2150 			continue;
2151 		}
2152 		if (ifp->if_flags & IFF_POINTOPOINT) {
2153 			if (sa_equal(addr, ifa->ifa_dstaddr))
2154 				goto done;
2155 		} else {
2156 			cp = addr->sa_data;
2157 			cp2 = ifa->ifa_addr->sa_data;
2158 			cp3 = ifa->ifa_netmask->sa_data;
2159 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2160 			for (; cp3 < cplim; cp3++)
2161 				if ((*cp++ ^ *cp2++) & *cp3)
2162 					break;
2163 			if (cp3 == cplim)
2164 				goto done;
2165 		}
2166 	}
2167 	ifa = ifa_maybe;
2168 done:
2169 	if (ifa != NULL)
2170 		ifa_ref(ifa);
2171 	IF_ADDR_RUNLOCK(ifp);
2172 	return (ifa);
2173 }
2174 
2175 /*
2176  * See whether new ifa is better than current one:
2177  * 1) A non-virtual one is preferred over virtual.
2178  * 2) A virtual in master state preferred over any other state.
2179  *
2180  * Used in several address selecting functions.
2181  */
2182 int
2183 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2184 {
2185 
2186 	return (cur->ifa_carp && (!next->ifa_carp ||
2187 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2188 }
2189 
2190 #include <net/if_llatbl.h>
2191 
2192 /*
2193  * Default action when installing a route with a Link Level gateway.
2194  * Lookup an appropriate real ifa to point to.
2195  * This should be moved to /sys/net/link.c eventually.
2196  */
2197 static void
2198 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2199 {
2200 	struct ifaddr *ifa, *oifa;
2201 	struct sockaddr *dst;
2202 	struct ifnet *ifp;
2203 
2204 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2205 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2206 		return;
2207 	ifa = ifaof_ifpforaddr(dst, ifp);
2208 	if (ifa) {
2209 		oifa = rt->rt_ifa;
2210 		rt->rt_ifa = ifa;
2211 		ifa_free(oifa);
2212 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2213 			ifa->ifa_rtrequest(cmd, rt, info);
2214 	}
2215 }
2216 
2217 struct sockaddr_dl *
2218 link_alloc_sdl(size_t size, int flags)
2219 {
2220 
2221 	return (malloc(size, M_TEMP, flags));
2222 }
2223 
2224 void
2225 link_free_sdl(struct sockaddr *sa)
2226 {
2227 	free(sa, M_TEMP);
2228 }
2229 
2230 /*
2231  * Fills in given sdl with interface basic info.
2232  * Returns pointer to filled sdl.
2233  */
2234 struct sockaddr_dl *
2235 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2236 {
2237 	struct sockaddr_dl *sdl;
2238 
2239 	sdl = (struct sockaddr_dl *)paddr;
2240 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2241 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2242 	sdl->sdl_family = AF_LINK;
2243 	sdl->sdl_index = ifp->if_index;
2244 	sdl->sdl_type = iftype;
2245 
2246 	return (sdl);
2247 }
2248 
2249 /*
2250  * Mark an interface down and notify protocols of
2251  * the transition.
2252  */
2253 static void
2254 if_unroute(struct ifnet *ifp, int flag, int fam)
2255 {
2256 	struct ifaddr *ifa;
2257 
2258 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2259 
2260 	ifp->if_flags &= ~flag;
2261 	getmicrotime(&ifp->if_lastchange);
2262 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2263 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2264 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2265 	ifp->if_qflush(ifp);
2266 
2267 	if (ifp->if_carp)
2268 		(*carp_linkstate_p)(ifp);
2269 	rt_ifmsg(ifp);
2270 }
2271 
2272 /*
2273  * Mark an interface up and notify protocols of
2274  * the transition.
2275  */
2276 static void
2277 if_route(struct ifnet *ifp, int flag, int fam)
2278 {
2279 	struct ifaddr *ifa;
2280 
2281 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2282 
2283 	ifp->if_flags |= flag;
2284 	getmicrotime(&ifp->if_lastchange);
2285 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2286 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2287 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2288 	if (ifp->if_carp)
2289 		(*carp_linkstate_p)(ifp);
2290 	rt_ifmsg(ifp);
2291 #ifdef INET6
2292 	in6_if_up(ifp);
2293 #endif
2294 }
2295 
2296 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2297 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2298 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2299 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2300 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2301 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2302 void	*(*vlan_cookie_p)(struct ifnet *);
2303 
2304 /*
2305  * Handle a change in the interface link state. To avoid LORs
2306  * between driver lock and upper layer locks, as well as possible
2307  * recursions, we post event to taskqueue, and all job
2308  * is done in static do_link_state_change().
2309  */
2310 void
2311 if_link_state_change(struct ifnet *ifp, int link_state)
2312 {
2313 	/* Return if state hasn't changed. */
2314 	if (ifp->if_link_state == link_state)
2315 		return;
2316 
2317 	ifp->if_link_state = link_state;
2318 
2319 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2320 }
2321 
2322 static void
2323 do_link_state_change(void *arg, int pending)
2324 {
2325 	struct ifnet *ifp = (struct ifnet *)arg;
2326 	int link_state = ifp->if_link_state;
2327 	CURVNET_SET(ifp->if_vnet);
2328 
2329 	/* Notify that the link state has changed. */
2330 	rt_ifmsg(ifp);
2331 	if (ifp->if_vlantrunk != NULL)
2332 		(*vlan_link_state_p)(ifp);
2333 
2334 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2335 	    ifp->if_l2com != NULL)
2336 		(*ng_ether_link_state_p)(ifp, link_state);
2337 	if (ifp->if_carp)
2338 		(*carp_linkstate_p)(ifp);
2339 	if (ifp->if_bridge)
2340 		ifp->if_bridge_linkstate(ifp);
2341 	if (ifp->if_lagg)
2342 		(*lagg_linkstate_p)(ifp, link_state);
2343 
2344 	if (IS_DEFAULT_VNET(curvnet))
2345 		devctl_notify("IFNET", ifp->if_xname,
2346 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2347 		    NULL);
2348 	if (pending > 1)
2349 		if_printf(ifp, "%d link states coalesced\n", pending);
2350 	if (log_link_state_change)
2351 		if_printf(ifp, "link state changed to %s\n",
2352 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2353 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2354 	CURVNET_RESTORE();
2355 }
2356 
2357 /*
2358  * Mark an interface down and notify protocols of
2359  * the transition.
2360  */
2361 void
2362 if_down(struct ifnet *ifp)
2363 {
2364 
2365 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2366 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2367 }
2368 
2369 /*
2370  * Mark an interface up and notify protocols of
2371  * the transition.
2372  */
2373 void
2374 if_up(struct ifnet *ifp)
2375 {
2376 
2377 	if_route(ifp, IFF_UP, AF_UNSPEC);
2378 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2379 }
2380 
2381 /*
2382  * Flush an interface queue.
2383  */
2384 void
2385 if_qflush(struct ifnet *ifp)
2386 {
2387 	struct mbuf *m, *n;
2388 	struct ifaltq *ifq;
2389 
2390 	ifq = &ifp->if_snd;
2391 	IFQ_LOCK(ifq);
2392 #ifdef ALTQ
2393 	if (ALTQ_IS_ENABLED(ifq))
2394 		ALTQ_PURGE(ifq);
2395 #endif
2396 	n = ifq->ifq_head;
2397 	while ((m = n) != NULL) {
2398 		n = m->m_nextpkt;
2399 		m_freem(m);
2400 	}
2401 	ifq->ifq_head = 0;
2402 	ifq->ifq_tail = 0;
2403 	ifq->ifq_len = 0;
2404 	IFQ_UNLOCK(ifq);
2405 }
2406 
2407 /*
2408  * Map interface name to interface structure pointer, with or without
2409  * returning a reference.
2410  */
2411 struct ifnet *
2412 ifunit_ref(const char *name)
2413 {
2414 	struct ifnet *ifp;
2415 
2416 	IFNET_RLOCK_NOSLEEP();
2417 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2418 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2419 		    !(ifp->if_flags & IFF_DYING))
2420 			break;
2421 	}
2422 	if (ifp != NULL)
2423 		if_ref(ifp);
2424 	IFNET_RUNLOCK_NOSLEEP();
2425 	return (ifp);
2426 }
2427 
2428 struct ifnet *
2429 ifunit(const char *name)
2430 {
2431 	struct ifnet *ifp;
2432 
2433 	IFNET_RLOCK_NOSLEEP();
2434 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2435 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2436 			break;
2437 	}
2438 	IFNET_RUNLOCK_NOSLEEP();
2439 	return (ifp);
2440 }
2441 
2442 static void *
2443 ifr_buffer_get_buffer(void *data)
2444 {
2445 	union ifreq_union *ifrup;
2446 
2447 	ifrup = data;
2448 #ifdef COMPAT_FREEBSD32
2449 	if (SV_CURPROC_FLAG(SV_ILP32))
2450 		return ((void *)(uintptr_t)
2451 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2452 #endif
2453 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2454 }
2455 
2456 static void
2457 ifr_buffer_set_buffer_null(void *data)
2458 {
2459 	union ifreq_union *ifrup;
2460 
2461 	ifrup = data;
2462 #ifdef COMPAT_FREEBSD32
2463 	if (SV_CURPROC_FLAG(SV_ILP32))
2464 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2465 	else
2466 #endif
2467 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2468 }
2469 
2470 static size_t
2471 ifr_buffer_get_length(void *data)
2472 {
2473 	union ifreq_union *ifrup;
2474 
2475 	ifrup = data;
2476 #ifdef COMPAT_FREEBSD32
2477 	if (SV_CURPROC_FLAG(SV_ILP32))
2478 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2479 #endif
2480 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2481 }
2482 
2483 static void
2484 ifr_buffer_set_length(void *data, size_t len)
2485 {
2486 	union ifreq_union *ifrup;
2487 
2488 	ifrup = data;
2489 #ifdef COMPAT_FREEBSD32
2490 	if (SV_CURPROC_FLAG(SV_ILP32))
2491 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2492 	else
2493 #endif
2494 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2495 }
2496 
2497 void *
2498 ifr_data_get_ptr(void *ifrp)
2499 {
2500 	union ifreq_union *ifrup;
2501 
2502 	ifrup = ifrp;
2503 #ifdef COMPAT_FREEBSD32
2504 	if (SV_CURPROC_FLAG(SV_ILP32))
2505 		return ((void *)(uintptr_t)
2506 		    ifrup->ifr32.ifr_ifru.ifru_data);
2507 #endif
2508 		return (ifrup->ifr.ifr_ifru.ifru_data);
2509 }
2510 
2511 /*
2512  * Hardware specific interface ioctls.
2513  */
2514 static int
2515 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2516 {
2517 	struct ifreq *ifr;
2518 	int error = 0, do_ifup = 0;
2519 	int new_flags, temp_flags;
2520 	size_t namelen, onamelen;
2521 	size_t descrlen;
2522 	char *descrbuf, *odescrbuf;
2523 	char new_name[IFNAMSIZ];
2524 	struct ifaddr *ifa;
2525 	struct sockaddr_dl *sdl;
2526 
2527 	ifr = (struct ifreq *)data;
2528 	switch (cmd) {
2529 	case SIOCGIFINDEX:
2530 		ifr->ifr_index = ifp->if_index;
2531 		break;
2532 
2533 	case SIOCGIFFLAGS:
2534 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2535 		ifr->ifr_flags = temp_flags & 0xffff;
2536 		ifr->ifr_flagshigh = temp_flags >> 16;
2537 		break;
2538 
2539 	case SIOCGIFCAP:
2540 		ifr->ifr_reqcap = ifp->if_capabilities;
2541 		ifr->ifr_curcap = ifp->if_capenable;
2542 		break;
2543 
2544 #ifdef MAC
2545 	case SIOCGIFMAC:
2546 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2547 		break;
2548 #endif
2549 
2550 	case SIOCGIFMETRIC:
2551 		ifr->ifr_metric = ifp->if_metric;
2552 		break;
2553 
2554 	case SIOCGIFMTU:
2555 		ifr->ifr_mtu = ifp->if_mtu;
2556 		break;
2557 
2558 	case SIOCGIFPHYS:
2559 		/* XXXGL: did this ever worked? */
2560 		ifr->ifr_phys = 0;
2561 		break;
2562 
2563 	case SIOCGIFDESCR:
2564 		error = 0;
2565 		sx_slock(&ifdescr_sx);
2566 		if (ifp->if_description == NULL)
2567 			error = ENOMSG;
2568 		else {
2569 			/* space for terminating nul */
2570 			descrlen = strlen(ifp->if_description) + 1;
2571 			if (ifr_buffer_get_length(ifr) < descrlen)
2572 				ifr_buffer_set_buffer_null(ifr);
2573 			else
2574 				error = copyout(ifp->if_description,
2575 				    ifr_buffer_get_buffer(ifr), descrlen);
2576 			ifr_buffer_set_length(ifr, descrlen);
2577 		}
2578 		sx_sunlock(&ifdescr_sx);
2579 		break;
2580 
2581 	case SIOCSIFDESCR:
2582 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2583 		if (error)
2584 			return (error);
2585 
2586 		/*
2587 		 * Copy only (length-1) bytes to make sure that
2588 		 * if_description is always nul terminated.  The
2589 		 * length parameter is supposed to count the
2590 		 * terminating nul in.
2591 		 */
2592 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2593 			return (ENAMETOOLONG);
2594 		else if (ifr_buffer_get_length(ifr) == 0)
2595 			descrbuf = NULL;
2596 		else {
2597 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2598 			    M_IFDESCR, M_WAITOK | M_ZERO);
2599 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2600 			    ifr_buffer_get_length(ifr) - 1);
2601 			if (error) {
2602 				free(descrbuf, M_IFDESCR);
2603 				break;
2604 			}
2605 		}
2606 
2607 		sx_xlock(&ifdescr_sx);
2608 		odescrbuf = ifp->if_description;
2609 		ifp->if_description = descrbuf;
2610 		sx_xunlock(&ifdescr_sx);
2611 
2612 		getmicrotime(&ifp->if_lastchange);
2613 		free(odescrbuf, M_IFDESCR);
2614 		break;
2615 
2616 	case SIOCGIFFIB:
2617 		ifr->ifr_fib = ifp->if_fib;
2618 		break;
2619 
2620 	case SIOCSIFFIB:
2621 		error = priv_check(td, PRIV_NET_SETIFFIB);
2622 		if (error)
2623 			return (error);
2624 		if (ifr->ifr_fib >= rt_numfibs)
2625 			return (EINVAL);
2626 
2627 		ifp->if_fib = ifr->ifr_fib;
2628 		break;
2629 
2630 	case SIOCSIFFLAGS:
2631 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2632 		if (error)
2633 			return (error);
2634 		/*
2635 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2636 		 * check, so we don't need special handling here yet.
2637 		 */
2638 		new_flags = (ifr->ifr_flags & 0xffff) |
2639 		    (ifr->ifr_flagshigh << 16);
2640 		if (ifp->if_flags & IFF_UP &&
2641 		    (new_flags & IFF_UP) == 0) {
2642 			if_down(ifp);
2643 		} else if (new_flags & IFF_UP &&
2644 		    (ifp->if_flags & IFF_UP) == 0) {
2645 			do_ifup = 1;
2646 		}
2647 		/* See if permanently promiscuous mode bit is about to flip */
2648 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2649 			if (new_flags & IFF_PPROMISC)
2650 				ifp->if_flags |= IFF_PROMISC;
2651 			else if (ifp->if_pcount == 0)
2652 				ifp->if_flags &= ~IFF_PROMISC;
2653 			if (log_promisc_mode_change)
2654                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2655                                     ((new_flags & IFF_PPROMISC) ?
2656                                      "enabled" : "disabled"));
2657 		}
2658 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2659 			(new_flags &~ IFF_CANTCHANGE);
2660 		if (ifp->if_ioctl) {
2661 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2662 		}
2663 		if (do_ifup)
2664 			if_up(ifp);
2665 		getmicrotime(&ifp->if_lastchange);
2666 		break;
2667 
2668 	case SIOCSIFCAP:
2669 		error = priv_check(td, PRIV_NET_SETIFCAP);
2670 		if (error)
2671 			return (error);
2672 		if (ifp->if_ioctl == NULL)
2673 			return (EOPNOTSUPP);
2674 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2675 			return (EINVAL);
2676 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2677 		if (error == 0)
2678 			getmicrotime(&ifp->if_lastchange);
2679 		break;
2680 
2681 #ifdef MAC
2682 	case SIOCSIFMAC:
2683 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2684 		break;
2685 #endif
2686 
2687 	case SIOCSIFNAME:
2688 		error = priv_check(td, PRIV_NET_SETIFNAME);
2689 		if (error)
2690 			return (error);
2691 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2692 		    NULL);
2693 		if (error != 0)
2694 			return (error);
2695 		if (new_name[0] == '\0')
2696 			return (EINVAL);
2697 		if (new_name[IFNAMSIZ-1] != '\0') {
2698 			new_name[IFNAMSIZ-1] = '\0';
2699 			if (strlen(new_name) == IFNAMSIZ-1)
2700 				return (EINVAL);
2701 		}
2702 		if (ifunit(new_name) != NULL)
2703 			return (EEXIST);
2704 
2705 		/*
2706 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2707 		 * and there are numerous other races with the
2708 		 * ifunit() checks not being atomic with namespace
2709 		 * changes (renames, vmoves, if_attach, etc).
2710 		 */
2711 		ifp->if_flags |= IFF_RENAMING;
2712 
2713 		/* Announce the departure of the interface. */
2714 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2715 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2716 
2717 		if_printf(ifp, "changing name to '%s'\n", new_name);
2718 
2719 		IF_ADDR_WLOCK(ifp);
2720 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2721 		ifa = ifp->if_addr;
2722 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2723 		namelen = strlen(new_name);
2724 		onamelen = sdl->sdl_nlen;
2725 		/*
2726 		 * Move the address if needed.  This is safe because we
2727 		 * allocate space for a name of length IFNAMSIZ when we
2728 		 * create this in if_attach().
2729 		 */
2730 		if (namelen != onamelen) {
2731 			bcopy(sdl->sdl_data + onamelen,
2732 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2733 		}
2734 		bcopy(new_name, sdl->sdl_data, namelen);
2735 		sdl->sdl_nlen = namelen;
2736 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2737 		bzero(sdl->sdl_data, onamelen);
2738 		while (namelen != 0)
2739 			sdl->sdl_data[--namelen] = 0xff;
2740 		IF_ADDR_WUNLOCK(ifp);
2741 
2742 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2743 		/* Announce the return of the interface. */
2744 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2745 
2746 		ifp->if_flags &= ~IFF_RENAMING;
2747 		break;
2748 
2749 #ifdef VIMAGE
2750 	case SIOCSIFVNET:
2751 		error = priv_check(td, PRIV_NET_SETIFVNET);
2752 		if (error)
2753 			return (error);
2754 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2755 		break;
2756 #endif
2757 
2758 	case SIOCSIFMETRIC:
2759 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2760 		if (error)
2761 			return (error);
2762 		ifp->if_metric = ifr->ifr_metric;
2763 		getmicrotime(&ifp->if_lastchange);
2764 		break;
2765 
2766 	case SIOCSIFPHYS:
2767 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2768 		if (error)
2769 			return (error);
2770 		if (ifp->if_ioctl == NULL)
2771 			return (EOPNOTSUPP);
2772 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2773 		if (error == 0)
2774 			getmicrotime(&ifp->if_lastchange);
2775 		break;
2776 
2777 	case SIOCSIFMTU:
2778 	{
2779 		u_long oldmtu = ifp->if_mtu;
2780 
2781 		error = priv_check(td, PRIV_NET_SETIFMTU);
2782 		if (error)
2783 			return (error);
2784 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2785 			return (EINVAL);
2786 		if (ifp->if_ioctl == NULL)
2787 			return (EOPNOTSUPP);
2788 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2789 		if (error == 0) {
2790 			getmicrotime(&ifp->if_lastchange);
2791 			rt_ifmsg(ifp);
2792 #ifdef INET
2793 			NETDUMP_REINIT(ifp);
2794 #endif
2795 		}
2796 		/*
2797 		 * If the link MTU changed, do network layer specific procedure.
2798 		 */
2799 		if (ifp->if_mtu != oldmtu) {
2800 #ifdef INET6
2801 			nd6_setmtu(ifp);
2802 #endif
2803 			rt_updatemtu(ifp);
2804 		}
2805 		break;
2806 	}
2807 
2808 	case SIOCADDMULTI:
2809 	case SIOCDELMULTI:
2810 		if (cmd == SIOCADDMULTI)
2811 			error = priv_check(td, PRIV_NET_ADDMULTI);
2812 		else
2813 			error = priv_check(td, PRIV_NET_DELMULTI);
2814 		if (error)
2815 			return (error);
2816 
2817 		/* Don't allow group membership on non-multicast interfaces. */
2818 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2819 			return (EOPNOTSUPP);
2820 
2821 		/* Don't let users screw up protocols' entries. */
2822 		if (ifr->ifr_addr.sa_family != AF_LINK)
2823 			return (EINVAL);
2824 
2825 		if (cmd == SIOCADDMULTI) {
2826 			struct ifmultiaddr *ifma;
2827 
2828 			/*
2829 			 * Userland is only permitted to join groups once
2830 			 * via the if_addmulti() KPI, because it cannot hold
2831 			 * struct ifmultiaddr * between calls. It may also
2832 			 * lose a race while we check if the membership
2833 			 * already exists.
2834 			 */
2835 			IF_ADDR_RLOCK(ifp);
2836 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2837 			IF_ADDR_RUNLOCK(ifp);
2838 			if (ifma != NULL)
2839 				error = EADDRINUSE;
2840 			else
2841 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2842 		} else {
2843 			error = if_delmulti(ifp, &ifr->ifr_addr);
2844 		}
2845 		if (error == 0)
2846 			getmicrotime(&ifp->if_lastchange);
2847 		break;
2848 
2849 	case SIOCSIFPHYADDR:
2850 	case SIOCDIFPHYADDR:
2851 #ifdef INET6
2852 	case SIOCSIFPHYADDR_IN6:
2853 #endif
2854 	case SIOCSIFMEDIA:
2855 	case SIOCSIFGENERIC:
2856 		error = priv_check(td, PRIV_NET_HWIOCTL);
2857 		if (error)
2858 			return (error);
2859 		if (ifp->if_ioctl == NULL)
2860 			return (EOPNOTSUPP);
2861 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2862 		if (error == 0)
2863 			getmicrotime(&ifp->if_lastchange);
2864 		break;
2865 
2866 	case SIOCGIFSTATUS:
2867 	case SIOCGIFPSRCADDR:
2868 	case SIOCGIFPDSTADDR:
2869 	case SIOCGIFMEDIA:
2870 	case SIOCGIFXMEDIA:
2871 	case SIOCGIFGENERIC:
2872 	case SIOCGIFRSSKEY:
2873 	case SIOCGIFRSSHASH:
2874 		if (ifp->if_ioctl == NULL)
2875 			return (EOPNOTSUPP);
2876 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2877 		break;
2878 
2879 	case SIOCSIFLLADDR:
2880 		error = priv_check(td, PRIV_NET_SETLLADDR);
2881 		if (error)
2882 			return (error);
2883 		error = if_setlladdr(ifp,
2884 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2885 		break;
2886 
2887 	case SIOCGHWADDR:
2888 		error = if_gethwaddr(ifp, ifr);
2889 		break;
2890 
2891 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2892 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2893 		if (error)
2894 			return (error);
2895 		if ((error = if_addgroup(ifp,
2896 		    ifgr_group_get((struct ifgroupreq *)data))))
2897 			return (error);
2898 		break;
2899 
2900 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2901 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2902 			return (error);
2903 		break;
2904 
2905 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2906 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2907 		if (error)
2908 			return (error);
2909 		if ((error = if_delgroup(ifp,
2910 		    ifgr_group_get((struct ifgroupreq *)data))))
2911 			return (error);
2912 		break;
2913 
2914 	default:
2915 		error = ENOIOCTL;
2916 		break;
2917 	}
2918 	return (error);
2919 }
2920 
2921 #ifdef COMPAT_FREEBSD32
2922 struct ifconf32 {
2923 	int32_t	ifc_len;
2924 	union {
2925 		uint32_t	ifcu_buf;
2926 		uint32_t	ifcu_req;
2927 	} ifc_ifcu;
2928 };
2929 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2930 #endif
2931 
2932 #ifdef COMPAT_FREEBSD32
2933 static void
2934 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2935 {
2936 	struct ifmediareq32 *ifmr32;
2937 
2938 	ifmr32 = (struct ifmediareq32 *)data;
2939 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2940 	    sizeof(ifmr->ifm_name));
2941 	ifmr->ifm_current = ifmr32->ifm_current;
2942 	ifmr->ifm_mask = ifmr32->ifm_mask;
2943 	ifmr->ifm_status = ifmr32->ifm_status;
2944 	ifmr->ifm_active = ifmr32->ifm_active;
2945 	ifmr->ifm_count = ifmr32->ifm_count;
2946 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2947 }
2948 
2949 static void
2950 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2951 {
2952 	struct ifmediareq32 *ifmr32;
2953 
2954 	ifmr32 = (struct ifmediareq32 *)data;
2955 	ifmr32->ifm_current = ifmr->ifm_current;
2956 	ifmr32->ifm_mask = ifmr->ifm_mask;
2957 	ifmr32->ifm_status = ifmr->ifm_status;
2958 	ifmr32->ifm_active = ifmr->ifm_active;
2959 	ifmr32->ifm_count = ifmr->ifm_count;
2960 }
2961 #endif
2962 
2963 /*
2964  * Interface ioctls.
2965  */
2966 int
2967 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2968 {
2969 #ifdef COMPAT_FREEBSD32
2970 	caddr_t saved_data = NULL;
2971 	struct ifmediareq ifmr;
2972 #endif
2973 	struct ifmediareq *ifmrp;
2974 	struct ifnet *ifp;
2975 	struct ifreq *ifr;
2976 	int error;
2977 	int oif_flags;
2978 #ifdef VIMAGE
2979 	int shutdown;
2980 #endif
2981 
2982 	CURVNET_SET(so->so_vnet);
2983 #ifdef VIMAGE
2984 	/* Make sure the VNET is stable. */
2985 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2986 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2987 	if (shutdown) {
2988 		CURVNET_RESTORE();
2989 		return (EBUSY);
2990 	}
2991 #endif
2992 
2993 
2994 	switch (cmd) {
2995 	case SIOCGIFCONF:
2996 		error = ifconf(cmd, data);
2997 		CURVNET_RESTORE();
2998 		return (error);
2999 
3000 #ifdef COMPAT_FREEBSD32
3001 	case SIOCGIFCONF32:
3002 		{
3003 			struct ifconf32 *ifc32;
3004 			struct ifconf ifc;
3005 
3006 			ifc32 = (struct ifconf32 *)data;
3007 			ifc.ifc_len = ifc32->ifc_len;
3008 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3009 
3010 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
3011 			CURVNET_RESTORE();
3012 			if (error == 0)
3013 				ifc32->ifc_len = ifc.ifc_len;
3014 			return (error);
3015 		}
3016 #endif
3017 	}
3018 
3019 	ifmrp = NULL;
3020 #ifdef COMPAT_FREEBSD32
3021 	switch (cmd) {
3022 	case SIOCGIFMEDIA32:
3023 	case SIOCGIFXMEDIA32:
3024 		ifmrp = &ifmr;
3025 		ifmr_init(ifmrp, data);
3026 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3027 		saved_data = data;
3028 		data = (caddr_t)ifmrp;
3029 	}
3030 #endif
3031 
3032 	ifr = (struct ifreq *)data;
3033 	switch (cmd) {
3034 #ifdef VIMAGE
3035 	case SIOCSIFRVNET:
3036 		error = priv_check(td, PRIV_NET_SETIFVNET);
3037 		if (error == 0)
3038 			error = if_vmove_reclaim(td, ifr->ifr_name,
3039 			    ifr->ifr_jid);
3040 		goto out_noref;
3041 #endif
3042 	case SIOCIFCREATE:
3043 	case SIOCIFCREATE2:
3044 		error = priv_check(td, PRIV_NET_IFCREATE);
3045 		if (error == 0)
3046 			error = if_clone_create(ifr->ifr_name,
3047 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3048 			    ifr_data_get_ptr(ifr) : NULL);
3049 		goto out_noref;
3050 	case SIOCIFDESTROY:
3051 		error = priv_check(td, PRIV_NET_IFDESTROY);
3052 		if (error == 0)
3053 			error = if_clone_destroy(ifr->ifr_name);
3054 		goto out_noref;
3055 
3056 	case SIOCIFGCLONERS:
3057 		error = if_clone_list((struct if_clonereq *)data);
3058 		goto out_noref;
3059 
3060 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3061 		error = if_getgroupmembers((struct ifgroupreq *)data);
3062 		goto out_noref;
3063 
3064 #if defined(INET) || defined(INET6)
3065 	case SIOCSVH:
3066 	case SIOCGVH:
3067 		if (carp_ioctl_p == NULL)
3068 			error = EPROTONOSUPPORT;
3069 		else
3070 			error = (*carp_ioctl_p)(ifr, cmd, td);
3071 		goto out_noref;
3072 #endif
3073 	}
3074 
3075 	ifp = ifunit_ref(ifr->ifr_name);
3076 	if (ifp == NULL) {
3077 		error = ENXIO;
3078 		goto out_noref;
3079 	}
3080 
3081 	error = ifhwioctl(cmd, ifp, data, td);
3082 	if (error != ENOIOCTL)
3083 		goto out_ref;
3084 
3085 	oif_flags = ifp->if_flags;
3086 	if (so->so_proto == NULL) {
3087 		error = EOPNOTSUPP;
3088 		goto out_ref;
3089 	}
3090 
3091 	/*
3092 	 * Pass the request on to the socket control method, and if the
3093 	 * latter returns EOPNOTSUPP, directly to the interface.
3094 	 *
3095 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3096 	 * trust SIOCSIFADDR et al to come from an already privileged
3097 	 * layer, and do not perform any credentials checks or input
3098 	 * validation.
3099 	 */
3100 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3101 	    ifp, td));
3102 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3103 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3104 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3105 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3106 
3107 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3108 #ifdef INET6
3109 		if (ifp->if_flags & IFF_UP)
3110 			in6_if_up(ifp);
3111 #endif
3112 	}
3113 
3114 out_ref:
3115 	if_rele(ifp);
3116 out_noref:
3117 #ifdef COMPAT_FREEBSD32
3118 	if (ifmrp != NULL) {
3119 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3120 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3121 		     cmd));
3122 		data = saved_data;
3123 		ifmr_update(ifmrp, data);
3124 	}
3125 #endif
3126 	CURVNET_RESTORE();
3127 	return (error);
3128 }
3129 
3130 /*
3131  * The code common to handling reference counted flags,
3132  * e.g., in ifpromisc() and if_allmulti().
3133  * The "pflag" argument can specify a permanent mode flag to check,
3134  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3135  *
3136  * Only to be used on stack-owned flags, not driver-owned flags.
3137  */
3138 static int
3139 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3140 {
3141 	struct ifreq ifr;
3142 	int error;
3143 	int oldflags, oldcount;
3144 
3145 	/* Sanity checks to catch programming errors */
3146 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3147 	    ("%s: setting driver-owned flag %d", __func__, flag));
3148 
3149 	if (onswitch)
3150 		KASSERT(*refcount >= 0,
3151 		    ("%s: increment negative refcount %d for flag %d",
3152 		    __func__, *refcount, flag));
3153 	else
3154 		KASSERT(*refcount > 0,
3155 		    ("%s: decrement non-positive refcount %d for flag %d",
3156 		    __func__, *refcount, flag));
3157 
3158 	/* In case this mode is permanent, just touch refcount */
3159 	if (ifp->if_flags & pflag) {
3160 		*refcount += onswitch ? 1 : -1;
3161 		return (0);
3162 	}
3163 
3164 	/* Save ifnet parameters for if_ioctl() may fail */
3165 	oldcount = *refcount;
3166 	oldflags = ifp->if_flags;
3167 
3168 	/*
3169 	 * See if we aren't the only and touching refcount is enough.
3170 	 * Actually toggle interface flag if we are the first or last.
3171 	 */
3172 	if (onswitch) {
3173 		if ((*refcount)++)
3174 			return (0);
3175 		ifp->if_flags |= flag;
3176 	} else {
3177 		if (--(*refcount))
3178 			return (0);
3179 		ifp->if_flags &= ~flag;
3180 	}
3181 
3182 	/* Call down the driver since we've changed interface flags */
3183 	if (ifp->if_ioctl == NULL) {
3184 		error = EOPNOTSUPP;
3185 		goto recover;
3186 	}
3187 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3188 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3189 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3190 	if (error)
3191 		goto recover;
3192 	/* Notify userland that interface flags have changed */
3193 	rt_ifmsg(ifp);
3194 	return (0);
3195 
3196 recover:
3197 	/* Recover after driver error */
3198 	*refcount = oldcount;
3199 	ifp->if_flags = oldflags;
3200 	return (error);
3201 }
3202 
3203 /*
3204  * Set/clear promiscuous mode on interface ifp based on the truth value
3205  * of pswitch.  The calls are reference counted so that only the first
3206  * "on" request actually has an effect, as does the final "off" request.
3207  * Results are undefined if the "off" and "on" requests are not matched.
3208  */
3209 int
3210 ifpromisc(struct ifnet *ifp, int pswitch)
3211 {
3212 	int error;
3213 	int oldflags = ifp->if_flags;
3214 
3215 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3216 			   &ifp->if_pcount, pswitch);
3217 	/* If promiscuous mode status has changed, log a message */
3218 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3219             log_promisc_mode_change)
3220 		if_printf(ifp, "promiscuous mode %s\n",
3221 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3222 	return (error);
3223 }
3224 
3225 /*
3226  * Return interface configuration
3227  * of system.  List may be used
3228  * in later ioctl's (above) to get
3229  * other information.
3230  */
3231 /*ARGSUSED*/
3232 static int
3233 ifconf(u_long cmd, caddr_t data)
3234 {
3235 	struct ifconf *ifc = (struct ifconf *)data;
3236 	struct ifnet *ifp;
3237 	struct ifaddr *ifa;
3238 	struct ifreq ifr;
3239 	struct sbuf *sb;
3240 	int error, full = 0, valid_len, max_len;
3241 
3242 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3243 	max_len = MAXPHYS - 1;
3244 
3245 	/* Prevent hostile input from being able to crash the system */
3246 	if (ifc->ifc_len <= 0)
3247 		return (EINVAL);
3248 
3249 again:
3250 	if (ifc->ifc_len <= max_len) {
3251 		max_len = ifc->ifc_len;
3252 		full = 1;
3253 	}
3254 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3255 	max_len = 0;
3256 	valid_len = 0;
3257 
3258 	IFNET_RLOCK();
3259 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3260 		int addrs;
3261 
3262 		/*
3263 		 * Zero the ifr to make sure we don't disclose the contents
3264 		 * of the stack.
3265 		 */
3266 		memset(&ifr, 0, sizeof(ifr));
3267 
3268 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3269 		    >= sizeof(ifr.ifr_name)) {
3270 			sbuf_delete(sb);
3271 			IFNET_RUNLOCK();
3272 			return (ENAMETOOLONG);
3273 		}
3274 
3275 		addrs = 0;
3276 		IF_ADDR_RLOCK(ifp);
3277 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3278 			struct sockaddr *sa = ifa->ifa_addr;
3279 
3280 			if (prison_if(curthread->td_ucred, sa) != 0)
3281 				continue;
3282 			addrs++;
3283 			if (sa->sa_len <= sizeof(*sa)) {
3284 				if (sa->sa_len < sizeof(*sa)) {
3285 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3286 					    sizeof(ifr.ifr_ifru.ifru_addr));
3287 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3288 					    sa->sa_len);
3289 				} else
3290 					ifr.ifr_ifru.ifru_addr = *sa;
3291 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3292 				max_len += sizeof(ifr);
3293 			} else {
3294 				sbuf_bcat(sb, &ifr,
3295 				    offsetof(struct ifreq, ifr_addr));
3296 				max_len += offsetof(struct ifreq, ifr_addr);
3297 				sbuf_bcat(sb, sa, sa->sa_len);
3298 				max_len += sa->sa_len;
3299 			}
3300 
3301 			if (sbuf_error(sb) == 0)
3302 				valid_len = sbuf_len(sb);
3303 		}
3304 		IF_ADDR_RUNLOCK(ifp);
3305 		if (addrs == 0) {
3306 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3307 			max_len += sizeof(ifr);
3308 
3309 			if (sbuf_error(sb) == 0)
3310 				valid_len = sbuf_len(sb);
3311 		}
3312 	}
3313 	IFNET_RUNLOCK();
3314 
3315 	/*
3316 	 * If we didn't allocate enough space (uncommon), try again.  If
3317 	 * we have already allocated as much space as we are allowed,
3318 	 * return what we've got.
3319 	 */
3320 	if (valid_len != max_len && !full) {
3321 		sbuf_delete(sb);
3322 		goto again;
3323 	}
3324 
3325 	ifc->ifc_len = valid_len;
3326 	sbuf_finish(sb);
3327 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3328 	sbuf_delete(sb);
3329 	return (error);
3330 }
3331 
3332 /*
3333  * Just like ifpromisc(), but for all-multicast-reception mode.
3334  */
3335 int
3336 if_allmulti(struct ifnet *ifp, int onswitch)
3337 {
3338 
3339 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3340 }
3341 
3342 struct ifmultiaddr *
3343 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3344 {
3345 	struct ifmultiaddr *ifma;
3346 
3347 	IF_ADDR_LOCK_ASSERT(ifp);
3348 
3349 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3350 		if (sa->sa_family == AF_LINK) {
3351 			if (sa_dl_equal(ifma->ifma_addr, sa))
3352 				break;
3353 		} else {
3354 			if (sa_equal(ifma->ifma_addr, sa))
3355 				break;
3356 		}
3357 	}
3358 
3359 	return ifma;
3360 }
3361 
3362 /*
3363  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3364  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3365  * the ifnet multicast address list here, so the caller must do that and
3366  * other setup work (such as notifying the device driver).  The reference
3367  * count is initialized to 1.
3368  */
3369 static struct ifmultiaddr *
3370 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3371     int mflags)
3372 {
3373 	struct ifmultiaddr *ifma;
3374 	struct sockaddr *dupsa;
3375 
3376 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3377 	    M_ZERO);
3378 	if (ifma == NULL)
3379 		return (NULL);
3380 
3381 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3382 	if (dupsa == NULL) {
3383 		free(ifma, M_IFMADDR);
3384 		return (NULL);
3385 	}
3386 	bcopy(sa, dupsa, sa->sa_len);
3387 	ifma->ifma_addr = dupsa;
3388 
3389 	ifma->ifma_ifp = ifp;
3390 	ifma->ifma_refcount = 1;
3391 	ifma->ifma_protospec = NULL;
3392 
3393 	if (llsa == NULL) {
3394 		ifma->ifma_lladdr = NULL;
3395 		return (ifma);
3396 	}
3397 
3398 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3399 	if (dupsa == NULL) {
3400 		free(ifma->ifma_addr, M_IFMADDR);
3401 		free(ifma, M_IFMADDR);
3402 		return (NULL);
3403 	}
3404 	bcopy(llsa, dupsa, llsa->sa_len);
3405 	ifma->ifma_lladdr = dupsa;
3406 
3407 	return (ifma);
3408 }
3409 
3410 /*
3411  * if_freemulti: free ifmultiaddr structure and possibly attached related
3412  * addresses.  The caller is responsible for implementing reference
3413  * counting, notifying the driver, handling routing messages, and releasing
3414  * any dependent link layer state.
3415  */
3416 #ifdef MCAST_VERBOSE
3417 extern void kdb_backtrace(void);
3418 #endif
3419 static void
3420 if_freemulti_internal(struct ifmultiaddr *ifma)
3421 {
3422 
3423 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3424 	    ifma->ifma_refcount));
3425 
3426 	if (ifma->ifma_lladdr != NULL)
3427 		free(ifma->ifma_lladdr, M_IFMADDR);
3428 #ifdef MCAST_VERBOSE
3429 	kdb_backtrace();
3430 	printf("%s freeing ifma: %p\n", __func__, ifma);
3431 #endif
3432 	free(ifma->ifma_addr, M_IFMADDR);
3433 	free(ifma, M_IFMADDR);
3434 }
3435 
3436 static void
3437 if_destroymulti(epoch_context_t ctx)
3438 {
3439 	struct ifmultiaddr *ifma;
3440 
3441 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3442 	if_freemulti_internal(ifma);
3443 }
3444 
3445 void
3446 if_freemulti(struct ifmultiaddr *ifma)
3447 {
3448 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3449 	    ifma->ifma_refcount));
3450 
3451 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3452 }
3453 
3454 
3455 /*
3456  * Register an additional multicast address with a network interface.
3457  *
3458  * - If the address is already present, bump the reference count on the
3459  *   address and return.
3460  * - If the address is not link-layer, look up a link layer address.
3461  * - Allocate address structures for one or both addresses, and attach to the
3462  *   multicast address list on the interface.  If automatically adding a link
3463  *   layer address, the protocol address will own a reference to the link
3464  *   layer address, to be freed when it is freed.
3465  * - Notify the network device driver of an addition to the multicast address
3466  *   list.
3467  *
3468  * 'sa' points to caller-owned memory with the desired multicast address.
3469  *
3470  * 'retifma' will be used to return a pointer to the resulting multicast
3471  * address reference, if desired.
3472  */
3473 int
3474 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3475     struct ifmultiaddr **retifma)
3476 {
3477 	struct ifmultiaddr *ifma, *ll_ifma;
3478 	struct sockaddr *llsa;
3479 	struct sockaddr_dl sdl;
3480 	int error;
3481 
3482 #ifdef INET
3483 	IN_MULTI_LIST_UNLOCK_ASSERT();
3484 #endif
3485 #ifdef INET6
3486 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3487 #endif
3488 	/*
3489 	 * If the address is already present, return a new reference to it;
3490 	 * otherwise, allocate storage and set up a new address.
3491 	 */
3492 	IF_ADDR_WLOCK(ifp);
3493 	ifma = if_findmulti(ifp, sa);
3494 	if (ifma != NULL) {
3495 		ifma->ifma_refcount++;
3496 		if (retifma != NULL)
3497 			*retifma = ifma;
3498 		IF_ADDR_WUNLOCK(ifp);
3499 		return (0);
3500 	}
3501 
3502 	/*
3503 	 * The address isn't already present; resolve the protocol address
3504 	 * into a link layer address, and then look that up, bump its
3505 	 * refcount or allocate an ifma for that also.
3506 	 * Most link layer resolving functions returns address data which
3507 	 * fits inside default sockaddr_dl structure. However callback
3508 	 * can allocate another sockaddr structure, in that case we need to
3509 	 * free it later.
3510 	 */
3511 	llsa = NULL;
3512 	ll_ifma = NULL;
3513 	if (ifp->if_resolvemulti != NULL) {
3514 		/* Provide called function with buffer size information */
3515 		sdl.sdl_len = sizeof(sdl);
3516 		llsa = (struct sockaddr *)&sdl;
3517 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3518 		if (error)
3519 			goto unlock_out;
3520 	}
3521 
3522 	/*
3523 	 * Allocate the new address.  Don't hook it up yet, as we may also
3524 	 * need to allocate a link layer multicast address.
3525 	 */
3526 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3527 	if (ifma == NULL) {
3528 		error = ENOMEM;
3529 		goto free_llsa_out;
3530 	}
3531 
3532 	/*
3533 	 * If a link layer address is found, we'll need to see if it's
3534 	 * already present in the address list, or allocate is as well.
3535 	 * When this block finishes, the link layer address will be on the
3536 	 * list.
3537 	 */
3538 	if (llsa != NULL) {
3539 		ll_ifma = if_findmulti(ifp, llsa);
3540 		if (ll_ifma == NULL) {
3541 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3542 			if (ll_ifma == NULL) {
3543 				--ifma->ifma_refcount;
3544 				if_freemulti(ifma);
3545 				error = ENOMEM;
3546 				goto free_llsa_out;
3547 			}
3548 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3549 			    ifma_link);
3550 		} else
3551 			ll_ifma->ifma_refcount++;
3552 		ifma->ifma_llifma = ll_ifma;
3553 	}
3554 
3555 	/*
3556 	 * We now have a new multicast address, ifma, and possibly a new or
3557 	 * referenced link layer address.  Add the primary address to the
3558 	 * ifnet address list.
3559 	 */
3560 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3561 
3562 	if (retifma != NULL)
3563 		*retifma = ifma;
3564 
3565 	/*
3566 	 * Must generate the message while holding the lock so that 'ifma'
3567 	 * pointer is still valid.
3568 	 */
3569 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3570 	IF_ADDR_WUNLOCK(ifp);
3571 
3572 	/*
3573 	 * We are certain we have added something, so call down to the
3574 	 * interface to let them know about it.
3575 	 */
3576 	if (ifp->if_ioctl != NULL) {
3577 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3578 	}
3579 
3580 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3581 		link_free_sdl(llsa);
3582 
3583 	return (0);
3584 
3585 free_llsa_out:
3586 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3587 		link_free_sdl(llsa);
3588 
3589 unlock_out:
3590 	IF_ADDR_WUNLOCK(ifp);
3591 	return (error);
3592 }
3593 
3594 /*
3595  * Delete a multicast group membership by network-layer group address.
3596  *
3597  * Returns ENOENT if the entry could not be found. If ifp no longer
3598  * exists, results are undefined. This entry point should only be used
3599  * from subsystems which do appropriate locking to hold ifp for the
3600  * duration of the call.
3601  * Network-layer protocol domains must use if_delmulti_ifma().
3602  */
3603 int
3604 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3605 {
3606 	struct ifmultiaddr *ifma;
3607 	int lastref;
3608 #ifdef INVARIANTS
3609 	struct ifnet *oifp;
3610 
3611 	IFNET_RLOCK_NOSLEEP();
3612 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3613 		if (ifp == oifp)
3614 			break;
3615 	if (ifp != oifp)
3616 		ifp = NULL;
3617 	IFNET_RUNLOCK_NOSLEEP();
3618 
3619 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3620 #endif
3621 	if (ifp == NULL)
3622 		return (ENOENT);
3623 
3624 	IF_ADDR_WLOCK(ifp);
3625 	lastref = 0;
3626 	ifma = if_findmulti(ifp, sa);
3627 	if (ifma != NULL)
3628 		lastref = if_delmulti_locked(ifp, ifma, 0);
3629 	IF_ADDR_WUNLOCK(ifp);
3630 
3631 	if (ifma == NULL)
3632 		return (ENOENT);
3633 
3634 	if (lastref && ifp->if_ioctl != NULL) {
3635 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3636 	}
3637 
3638 	return (0);
3639 }
3640 
3641 /*
3642  * Delete all multicast group membership for an interface.
3643  * Should be used to quickly flush all multicast filters.
3644  */
3645 void
3646 if_delallmulti(struct ifnet *ifp)
3647 {
3648 	struct ifmultiaddr *ifma;
3649 	struct ifmultiaddr *next;
3650 
3651 	IF_ADDR_WLOCK(ifp);
3652 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3653 		if_delmulti_locked(ifp, ifma, 0);
3654 	IF_ADDR_WUNLOCK(ifp);
3655 }
3656 
3657 void
3658 if_delmulti_ifma(struct ifmultiaddr *ifma)
3659 {
3660 	if_delmulti_ifma_flags(ifma, 0);
3661 }
3662 
3663 /*
3664  * Delete a multicast group membership by group membership pointer.
3665  * Network-layer protocol domains must use this routine.
3666  *
3667  * It is safe to call this routine if the ifp disappeared.
3668  */
3669 void
3670 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3671 {
3672 	struct ifnet *ifp;
3673 	int lastref;
3674 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3675 #ifdef INET
3676 	IN_MULTI_LIST_UNLOCK_ASSERT();
3677 #endif
3678 	ifp = ifma->ifma_ifp;
3679 #ifdef DIAGNOSTIC
3680 	if (ifp == NULL) {
3681 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3682 	} else {
3683 		struct ifnet *oifp;
3684 
3685 		IFNET_RLOCK_NOSLEEP();
3686 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3687 			if (ifp == oifp)
3688 				break;
3689 		if (ifp != oifp) {
3690 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3691 			ifp = NULL;
3692 		}
3693 		IFNET_RUNLOCK_NOSLEEP();
3694 	}
3695 #endif
3696 	/*
3697 	 * If and only if the ifnet instance exists: Acquire the address lock.
3698 	 */
3699 	if (ifp != NULL)
3700 		IF_ADDR_WLOCK(ifp);
3701 
3702 	lastref = if_delmulti_locked(ifp, ifma, flags);
3703 
3704 	if (ifp != NULL) {
3705 		/*
3706 		 * If and only if the ifnet instance exists:
3707 		 *  Release the address lock.
3708 		 *  If the group was left: update the hardware hash filter.
3709 		 */
3710 		IF_ADDR_WUNLOCK(ifp);
3711 		if (lastref && ifp->if_ioctl != NULL) {
3712 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3713 		}
3714 	}
3715 }
3716 
3717 /*
3718  * Perform deletion of network-layer and/or link-layer multicast address.
3719  *
3720  * Return 0 if the reference count was decremented.
3721  * Return 1 if the final reference was released, indicating that the
3722  * hardware hash filter should be reprogrammed.
3723  */
3724 static int
3725 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3726 {
3727 	struct ifmultiaddr *ll_ifma;
3728 
3729 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3730 		KASSERT(ifma->ifma_ifp == ifp,
3731 		    ("%s: inconsistent ifp %p", __func__, ifp));
3732 		IF_ADDR_WLOCK_ASSERT(ifp);
3733 	}
3734 
3735 	ifp = ifma->ifma_ifp;
3736 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3737 
3738 	/*
3739 	 * If the ifnet is detaching, null out references to ifnet,
3740 	 * so that upper protocol layers will notice, and not attempt
3741 	 * to obtain locks for an ifnet which no longer exists. The
3742 	 * routing socket announcement must happen before the ifnet
3743 	 * instance is detached from the system.
3744 	 */
3745 	if (detaching) {
3746 #ifdef DIAGNOSTIC
3747 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3748 #endif
3749 		/*
3750 		 * ifp may already be nulled out if we are being reentered
3751 		 * to delete the ll_ifma.
3752 		 */
3753 		if (ifp != NULL) {
3754 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3755 			ifma->ifma_ifp = NULL;
3756 		}
3757 	}
3758 
3759 	if (--ifma->ifma_refcount > 0)
3760 		return 0;
3761 
3762 	if (ifp != NULL && detaching == 0)
3763 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3764 
3765 	/*
3766 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3767 	 * have been associated with it. Release it first if so.
3768 	 */
3769 	ll_ifma = ifma->ifma_llifma;
3770 	if (ll_ifma != NULL) {
3771 		KASSERT(ifma->ifma_lladdr != NULL,
3772 		    ("%s: llifma w/o lladdr", __func__));
3773 		if (detaching)
3774 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3775 		if (--ll_ifma->ifma_refcount == 0) {
3776 			if (ifp != NULL) {
3777 				CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3778 				    ifma_link);
3779 			}
3780 			if_freemulti(ll_ifma);
3781 		}
3782 	}
3783 #ifdef INVARIANTS
3784 	if (ifp) {
3785 		struct ifmultiaddr *ifmatmp;
3786 
3787 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3788 			MPASS(ifma != ifmatmp);
3789 	}
3790 #endif
3791 	if_freemulti(ifma);
3792 	/*
3793 	 * The last reference to this instance of struct ifmultiaddr
3794 	 * was released; the hardware should be notified of this change.
3795 	 */
3796 	return 1;
3797 }
3798 
3799 /*
3800  * Set the link layer address on an interface.
3801  *
3802  * At this time we only support certain types of interfaces,
3803  * and we don't allow the length of the address to change.
3804  *
3805  * Set noinline to be dtrace-friendly
3806  */
3807 __noinline int
3808 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3809 {
3810 	struct sockaddr_dl *sdl;
3811 	struct ifaddr *ifa;
3812 	struct ifreq ifr;
3813 
3814 	IF_ADDR_RLOCK(ifp);
3815 	ifa = ifp->if_addr;
3816 	if (ifa == NULL) {
3817 		IF_ADDR_RUNLOCK(ifp);
3818 		return (EINVAL);
3819 	}
3820 	ifa_ref(ifa);
3821 	IF_ADDR_RUNLOCK(ifp);
3822 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3823 	if (sdl == NULL) {
3824 		ifa_free(ifa);
3825 		return (EINVAL);
3826 	}
3827 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3828 		ifa_free(ifa);
3829 		return (EINVAL);
3830 	}
3831 	switch (ifp->if_type) {
3832 	case IFT_ETHER:
3833 	case IFT_XETHER:
3834 	case IFT_L2VLAN:
3835 	case IFT_BRIDGE:
3836 	case IFT_IEEE8023ADLAG:
3837 		bcopy(lladdr, LLADDR(sdl), len);
3838 		ifa_free(ifa);
3839 		break;
3840 	default:
3841 		ifa_free(ifa);
3842 		return (ENODEV);
3843 	}
3844 
3845 	/*
3846 	 * If the interface is already up, we need
3847 	 * to re-init it in order to reprogram its
3848 	 * address filter.
3849 	 */
3850 	if ((ifp->if_flags & IFF_UP) != 0) {
3851 		if (ifp->if_ioctl) {
3852 			ifp->if_flags &= ~IFF_UP;
3853 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3854 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3855 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3856 			ifp->if_flags |= IFF_UP;
3857 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3858 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3859 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3860 		}
3861 	}
3862 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3863 	return (0);
3864 }
3865 
3866 /*
3867  * Compat function for handling basic encapsulation requests.
3868  * Not converted stacks (FDDI, IB, ..) supports traditional
3869  * output model: ARP (and other similar L2 protocols) are handled
3870  * inside output routine, arpresolve/nd6_resolve() returns MAC
3871  * address instead of full prepend.
3872  *
3873  * This function creates calculated header==MAC for IPv4/IPv6 and
3874  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3875  * address families.
3876  */
3877 static int
3878 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3879 {
3880 
3881 	if (req->rtype != IFENCAP_LL)
3882 		return (EOPNOTSUPP);
3883 
3884 	if (req->bufsize < req->lladdr_len)
3885 		return (ENOMEM);
3886 
3887 	switch (req->family) {
3888 	case AF_INET:
3889 	case AF_INET6:
3890 		break;
3891 	default:
3892 		return (EAFNOSUPPORT);
3893 	}
3894 
3895 	/* Copy lladdr to storage as is */
3896 	memmove(req->buf, req->lladdr, req->lladdr_len);
3897 	req->bufsize = req->lladdr_len;
3898 	req->lladdr_off = 0;
3899 
3900 	return (0);
3901 }
3902 
3903 /*
3904  * Get the link layer address that was read from the hardware at attach.
3905  *
3906  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3907  * their component interfaces as IFT_IEEE8023ADLAG.
3908  */
3909 int
3910 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3911 {
3912 
3913 	if (ifp->if_hw_addr == NULL)
3914 		return (ENODEV);
3915 
3916 	switch (ifp->if_type) {
3917 	case IFT_ETHER:
3918 	case IFT_IEEE8023ADLAG:
3919 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3920 		return (0);
3921 	default:
3922 		return (ENODEV);
3923 	}
3924 }
3925 
3926 /*
3927  * The name argument must be a pointer to storage which will last as
3928  * long as the interface does.  For physical devices, the result of
3929  * device_get_name(dev) is a good choice and for pseudo-devices a
3930  * static string works well.
3931  */
3932 void
3933 if_initname(struct ifnet *ifp, const char *name, int unit)
3934 {
3935 	ifp->if_dname = name;
3936 	ifp->if_dunit = unit;
3937 	if (unit != IF_DUNIT_NONE)
3938 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3939 	else
3940 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3941 }
3942 
3943 int
3944 if_printf(struct ifnet *ifp, const char *fmt, ...)
3945 {
3946 	char if_fmt[256];
3947 	va_list ap;
3948 
3949 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3950 	va_start(ap, fmt);
3951 	vlog(LOG_INFO, if_fmt, ap);
3952 	va_end(ap);
3953 	return (0);
3954 }
3955 
3956 void
3957 if_start(struct ifnet *ifp)
3958 {
3959 
3960 	(*(ifp)->if_start)(ifp);
3961 }
3962 
3963 /*
3964  * Backwards compatibility interface for drivers
3965  * that have not implemented it
3966  */
3967 static int
3968 if_transmit(struct ifnet *ifp, struct mbuf *m)
3969 {
3970 	int error;
3971 
3972 	IFQ_HANDOFF(ifp, m, error);
3973 	return (error);
3974 }
3975 
3976 static void
3977 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3978 {
3979 
3980 	m_freem(m);
3981 }
3982 
3983 int
3984 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3985 {
3986 	int active = 0;
3987 
3988 	IF_LOCK(ifq);
3989 	if (_IF_QFULL(ifq)) {
3990 		IF_UNLOCK(ifq);
3991 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3992 		m_freem(m);
3993 		return (0);
3994 	}
3995 	if (ifp != NULL) {
3996 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3997 		if (m->m_flags & (M_BCAST|M_MCAST))
3998 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3999 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4000 	}
4001 	_IF_ENQUEUE(ifq, m);
4002 	IF_UNLOCK(ifq);
4003 	if (ifp != NULL && !active)
4004 		(*(ifp)->if_start)(ifp);
4005 	return (1);
4006 }
4007 
4008 void
4009 if_register_com_alloc(u_char type,
4010     if_com_alloc_t *a, if_com_free_t *f)
4011 {
4012 
4013 	KASSERT(if_com_alloc[type] == NULL,
4014 	    ("if_register_com_alloc: %d already registered", type));
4015 	KASSERT(if_com_free[type] == NULL,
4016 	    ("if_register_com_alloc: %d free already registered", type));
4017 
4018 	if_com_alloc[type] = a;
4019 	if_com_free[type] = f;
4020 }
4021 
4022 void
4023 if_deregister_com_alloc(u_char type)
4024 {
4025 
4026 	KASSERT(if_com_alloc[type] != NULL,
4027 	    ("if_deregister_com_alloc: %d not registered", type));
4028 	KASSERT(if_com_free[type] != NULL,
4029 	    ("if_deregister_com_alloc: %d free not registered", type));
4030 	if_com_alloc[type] = NULL;
4031 	if_com_free[type] = NULL;
4032 }
4033 
4034 /* API for driver access to network stack owned ifnet.*/
4035 uint64_t
4036 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4037 {
4038 	uint64_t oldbrate;
4039 
4040 	oldbrate = ifp->if_baudrate;
4041 	ifp->if_baudrate = baudrate;
4042 	return (oldbrate);
4043 }
4044 
4045 uint64_t
4046 if_getbaudrate(if_t ifp)
4047 {
4048 
4049 	return (((struct ifnet *)ifp)->if_baudrate);
4050 }
4051 
4052 int
4053 if_setcapabilities(if_t ifp, int capabilities)
4054 {
4055 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4056 	return (0);
4057 }
4058 
4059 int
4060 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4061 {
4062 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4063 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4064 
4065 	return (0);
4066 }
4067 
4068 int
4069 if_getcapabilities(if_t ifp)
4070 {
4071 	return ((struct ifnet *)ifp)->if_capabilities;
4072 }
4073 
4074 int
4075 if_setcapenable(if_t ifp, int capabilities)
4076 {
4077 	((struct ifnet *)ifp)->if_capenable = capabilities;
4078 	return (0);
4079 }
4080 
4081 int
4082 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4083 {
4084 	if(setcap)
4085 		((struct ifnet *)ifp)->if_capenable |= setcap;
4086 	if(clearcap)
4087 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4088 
4089 	return (0);
4090 }
4091 
4092 const char *
4093 if_getdname(if_t ifp)
4094 {
4095 	return ((struct ifnet *)ifp)->if_dname;
4096 }
4097 
4098 int
4099 if_togglecapenable(if_t ifp, int togglecap)
4100 {
4101 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4102 	return (0);
4103 }
4104 
4105 int
4106 if_getcapenable(if_t ifp)
4107 {
4108 	return ((struct ifnet *)ifp)->if_capenable;
4109 }
4110 
4111 /*
4112  * This is largely undesirable because it ties ifnet to a device, but does
4113  * provide flexiblity for an embedded product vendor. Should be used with
4114  * the understanding that it violates the interface boundaries, and should be
4115  * a last resort only.
4116  */
4117 int
4118 if_setdev(if_t ifp, void *dev)
4119 {
4120 	return (0);
4121 }
4122 
4123 int
4124 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4125 {
4126 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4127 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4128 
4129 	return (0);
4130 }
4131 
4132 int
4133 if_getdrvflags(if_t ifp)
4134 {
4135 	return ((struct ifnet *)ifp)->if_drv_flags;
4136 }
4137 
4138 int
4139 if_setdrvflags(if_t ifp, int flags)
4140 {
4141 	((struct ifnet *)ifp)->if_drv_flags = flags;
4142 	return (0);
4143 }
4144 
4145 
4146 int
4147 if_setflags(if_t ifp, int flags)
4148 {
4149 	((struct ifnet *)ifp)->if_flags = flags;
4150 	return (0);
4151 }
4152 
4153 int
4154 if_setflagbits(if_t ifp, int set, int clear)
4155 {
4156 	((struct ifnet *)ifp)->if_flags |= set;
4157 	((struct ifnet *)ifp)->if_flags &= ~clear;
4158 
4159 	return (0);
4160 }
4161 
4162 int
4163 if_getflags(if_t ifp)
4164 {
4165 	return ((struct ifnet *)ifp)->if_flags;
4166 }
4167 
4168 int
4169 if_clearhwassist(if_t ifp)
4170 {
4171 	((struct ifnet *)ifp)->if_hwassist = 0;
4172 	return (0);
4173 }
4174 
4175 int
4176 if_sethwassistbits(if_t ifp, int toset, int toclear)
4177 {
4178 	((struct ifnet *)ifp)->if_hwassist |= toset;
4179 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4180 
4181 	return (0);
4182 }
4183 
4184 int
4185 if_sethwassist(if_t ifp, int hwassist_bit)
4186 {
4187 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4188 	return (0);
4189 }
4190 
4191 int
4192 if_gethwassist(if_t ifp)
4193 {
4194 	return ((struct ifnet *)ifp)->if_hwassist;
4195 }
4196 
4197 int
4198 if_setmtu(if_t ifp, int mtu)
4199 {
4200 	((struct ifnet *)ifp)->if_mtu = mtu;
4201 	return (0);
4202 }
4203 
4204 int
4205 if_getmtu(if_t ifp)
4206 {
4207 	return ((struct ifnet *)ifp)->if_mtu;
4208 }
4209 
4210 int
4211 if_getmtu_family(if_t ifp, int family)
4212 {
4213 	struct domain *dp;
4214 
4215 	for (dp = domains; dp; dp = dp->dom_next) {
4216 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4217 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4218 	}
4219 
4220 	return (((struct ifnet *)ifp)->if_mtu);
4221 }
4222 
4223 int
4224 if_setsoftc(if_t ifp, void *softc)
4225 {
4226 	((struct ifnet *)ifp)->if_softc = softc;
4227 	return (0);
4228 }
4229 
4230 void *
4231 if_getsoftc(if_t ifp)
4232 {
4233 	return ((struct ifnet *)ifp)->if_softc;
4234 }
4235 
4236 void
4237 if_setrcvif(struct mbuf *m, if_t ifp)
4238 {
4239 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4240 }
4241 
4242 void
4243 if_setvtag(struct mbuf *m, uint16_t tag)
4244 {
4245 	m->m_pkthdr.ether_vtag = tag;
4246 }
4247 
4248 uint16_t
4249 if_getvtag(struct mbuf *m)
4250 {
4251 
4252 	return (m->m_pkthdr.ether_vtag);
4253 }
4254 
4255 int
4256 if_sendq_empty(if_t ifp)
4257 {
4258 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4259 }
4260 
4261 struct ifaddr *
4262 if_getifaddr(if_t ifp)
4263 {
4264 	return ((struct ifnet *)ifp)->if_addr;
4265 }
4266 
4267 int
4268 if_getamcount(if_t ifp)
4269 {
4270 	return ((struct ifnet *)ifp)->if_amcount;
4271 }
4272 
4273 
4274 int
4275 if_setsendqready(if_t ifp)
4276 {
4277 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4278 	return (0);
4279 }
4280 
4281 int
4282 if_setsendqlen(if_t ifp, int tx_desc_count)
4283 {
4284 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4285 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4286 
4287 	return (0);
4288 }
4289 
4290 int
4291 if_vlantrunkinuse(if_t ifp)
4292 {
4293 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4294 }
4295 
4296 int
4297 if_input(if_t ifp, struct mbuf* sendmp)
4298 {
4299 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4300 	return (0);
4301 
4302 }
4303 
4304 /* XXX */
4305 #ifndef ETH_ADDR_LEN
4306 #define ETH_ADDR_LEN 6
4307 #endif
4308 
4309 int
4310 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4311 {
4312 	struct ifmultiaddr *ifma;
4313 	uint8_t *lmta = (uint8_t *)mta;
4314 	int mcnt = 0;
4315 
4316 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4317 		if (ifma->ifma_addr->sa_family != AF_LINK)
4318 			continue;
4319 
4320 		if (mcnt == max)
4321 			break;
4322 
4323 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4324 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4325 		mcnt++;
4326 	}
4327 	*cnt = mcnt;
4328 
4329 	return (0);
4330 }
4331 
4332 int
4333 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4334 {
4335 	int error;
4336 
4337 	if_maddr_rlock(ifp);
4338 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4339 	if_maddr_runlock(ifp);
4340 	return (error);
4341 }
4342 
4343 int
4344 if_multiaddr_count(if_t ifp, int max)
4345 {
4346 	struct ifmultiaddr *ifma;
4347 	int count;
4348 
4349 	count = 0;
4350 	if_maddr_rlock(ifp);
4351 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4352 		if (ifma->ifma_addr->sa_family != AF_LINK)
4353 			continue;
4354 		count++;
4355 		if (count == max)
4356 			break;
4357 	}
4358 	if_maddr_runlock(ifp);
4359 	return (count);
4360 }
4361 
4362 int
4363 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4364 {
4365 	struct ifmultiaddr *ifma;
4366 	int cnt = 0;
4367 
4368 	if_maddr_rlock(ifp);
4369 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4370 		cnt += filter(arg, ifma, cnt);
4371 	if_maddr_runlock(ifp);
4372 	return (cnt);
4373 }
4374 
4375 struct mbuf *
4376 if_dequeue(if_t ifp)
4377 {
4378 	struct mbuf *m;
4379 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4380 
4381 	return (m);
4382 }
4383 
4384 int
4385 if_sendq_prepend(if_t ifp, struct mbuf *m)
4386 {
4387 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4388 	return (0);
4389 }
4390 
4391 int
4392 if_setifheaderlen(if_t ifp, int len)
4393 {
4394 	((struct ifnet *)ifp)->if_hdrlen = len;
4395 	return (0);
4396 }
4397 
4398 caddr_t
4399 if_getlladdr(if_t ifp)
4400 {
4401 	return (IF_LLADDR((struct ifnet *)ifp));
4402 }
4403 
4404 void *
4405 if_gethandle(u_char type)
4406 {
4407 	return (if_alloc(type));
4408 }
4409 
4410 void
4411 if_bpfmtap(if_t ifh, struct mbuf *m)
4412 {
4413 	struct ifnet *ifp = (struct ifnet *)ifh;
4414 
4415 	BPF_MTAP(ifp, m);
4416 }
4417 
4418 void
4419 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4420 {
4421 	struct ifnet *ifp = (struct ifnet *)ifh;
4422 
4423 	ETHER_BPF_MTAP(ifp, m);
4424 }
4425 
4426 void
4427 if_vlancap(if_t ifh)
4428 {
4429 	struct ifnet *ifp = (struct ifnet *)ifh;
4430 	VLAN_CAPABILITIES(ifp);
4431 }
4432 
4433 int
4434 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4435 {
4436 
4437 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4438         return (0);
4439 }
4440 
4441 int
4442 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4443 {
4444 
4445 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4446         return (0);
4447 }
4448 
4449 int
4450 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4451 {
4452 
4453 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4454         return (0);
4455 }
4456 
4457 u_int
4458 if_gethwtsomax(if_t ifp)
4459 {
4460 
4461 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4462 }
4463 
4464 u_int
4465 if_gethwtsomaxsegcount(if_t ifp)
4466 {
4467 
4468 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4469 }
4470 
4471 u_int
4472 if_gethwtsomaxsegsize(if_t ifp)
4473 {
4474 
4475 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4476 }
4477 
4478 void
4479 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4480 {
4481 	((struct ifnet *)ifp)->if_init = init_fn;
4482 }
4483 
4484 void
4485 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4486 {
4487 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4488 }
4489 
4490 void
4491 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4492 {
4493 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4494 }
4495 
4496 void
4497 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4498 {
4499 	((struct ifnet *)ifp)->if_transmit = start_fn;
4500 }
4501 
4502 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4503 {
4504 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4505 
4506 }
4507 
4508 void
4509 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4510 {
4511 
4512 	ifp->if_get_counter = fn;
4513 }
4514 
4515 /* Revisit these - These are inline functions originally. */
4516 int
4517 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4518 {
4519 	return drbr_inuse(ifh, br);
4520 }
4521 
4522 struct mbuf*
4523 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4524 {
4525 	return drbr_dequeue(ifh, br);
4526 }
4527 
4528 int
4529 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4530 {
4531 	return drbr_needs_enqueue(ifh, br);
4532 }
4533 
4534 int
4535 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4536 {
4537 	return drbr_enqueue(ifh, br, m);
4538 
4539 }
4540