xref: /freebsd/sys/net/if.c (revision fd253945ac76a54ff9c11cf02f5458561f711866)
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	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void	do_link_state_change(void *, int);
269 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int	if_getgroupmembers(struct ifgroupreq *);
271 static void	if_delgroups(struct ifnet *);
272 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
274 #ifdef VIMAGE
275 static void	if_vmove(struct ifnet *, struct vnet *);
276 #endif
277 
278 #ifdef INET6
279 /*
280  * XXX: declare here to avoid to include many inet6 related files..
281  * should be more generalized?
282  */
283 extern void	nd6_setmtu(struct ifnet *);
284 #endif
285 
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
289 
290 VNET_DEFINE(int, if_index);
291 int	ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
294 
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
296 
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
299 
300 #define	V_if_indexlim		VNET(if_indexlim)
301 #define	V_ifindex_table		VNET(ifindex_table)
302 
303 /*
304  * The global network interface list (V_ifnet) and related state (such as
305  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306  * an rwlock.  Either may be acquired shared to stablize the list, but both
307  * must be acquired writable to modify the list.  This model allows us to
308  * both stablize the interface list during interrupt thread processing, but
309  * also to stablize it over long-running ioctls, without introducing priority
310  * inversions and deadlocks.
311  */
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
316 
317 /*
318  * The allocation of network interfaces is a rather non-atomic affair; we
319  * need to select an index before we are ready to expose the interface for
320  * use, so will use this pointer value to indicate reservation.
321  */
322 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
323 
324 static	if_com_alloc_t *if_com_alloc[256];
325 static	if_com_free_t *if_com_free[256];
326 
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
330 
331 struct ifnet *
332 ifnet_byindex_locked(u_short idx)
333 {
334 
335 	if (idx > V_if_index)
336 		return (NULL);
337 	if (V_ifindex_table[idx] == IFNET_HOLD)
338 		return (NULL);
339 	return (V_ifindex_table[idx]);
340 }
341 
342 struct ifnet *
343 ifnet_byindex(u_short idx)
344 {
345 	struct ifnet *ifp;
346 
347 	ifp = ifnet_byindex_locked(idx);
348 	return (ifp);
349 }
350 
351 struct ifnet *
352 ifnet_byindex_ref(u_short idx)
353 {
354 	struct ifnet *ifp;
355 
356 	IFNET_RLOCK_NOSLEEP();
357 	ifp = ifnet_byindex_locked(idx);
358 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
359 		IFNET_RUNLOCK_NOSLEEP();
360 		return (NULL);
361 	}
362 	if_ref(ifp);
363 	IFNET_RUNLOCK_NOSLEEP();
364 	return (ifp);
365 }
366 
367 /*
368  * Allocate an ifindex array entry; return 0 on success or an error on
369  * failure.
370  */
371 static u_short
372 ifindex_alloc(void **old)
373 {
374 	u_short idx;
375 
376 	IFNET_WLOCK_ASSERT();
377 	/*
378 	 * Try to find an empty slot below V_if_index.  If we fail, take the
379 	 * next slot.
380 	 */
381 	for (idx = 1; idx <= V_if_index; idx++) {
382 		if (V_ifindex_table[idx] == NULL)
383 			break;
384 	}
385 
386 	/* Catch if_index overflow. */
387 	if (idx >= V_if_indexlim) {
388 		*old = if_grow();
389 		return (USHRT_MAX);
390 	}
391 	if (idx > V_if_index)
392 		V_if_index = idx;
393 	return (idx);
394 }
395 
396 static void
397 ifindex_free_locked(u_short idx)
398 {
399 
400 	IFNET_WLOCK_ASSERT();
401 
402 	V_ifindex_table[idx] = NULL;
403 	while (V_if_index > 0 &&
404 	    V_ifindex_table[V_if_index] == NULL)
405 		V_if_index--;
406 }
407 
408 static void
409 ifindex_free(u_short idx)
410 {
411 
412 	IFNET_WLOCK();
413 	ifindex_free_locked(idx);
414 	IFNET_WUNLOCK();
415 }
416 
417 static void
418 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
419 {
420 
421 	V_ifindex_table[idx] = ifp;
422 }
423 
424 struct ifaddr *
425 ifaddr_byindex(u_short idx)
426 {
427 	struct ifnet *ifp;
428 	struct ifaddr *ifa = NULL;
429 
430 	IFNET_RLOCK_NOSLEEP();
431 	ifp = ifnet_byindex_locked(idx);
432 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
433 		ifa_ref(ifa);
434 	IFNET_RUNLOCK_NOSLEEP();
435 	return (ifa);
436 }
437 
438 /*
439  * Network interface utility routines.
440  *
441  * Routines with ifa_ifwith* names take sockaddr *'s as
442  * parameters.
443  */
444 
445 static void
446 vnet_if_init(const void *unused __unused)
447 {
448 	void *old;
449 
450 	CK_STAILQ_INIT(&V_ifnet);
451 	CK_STAILQ_INIT(&V_ifg_head);
452 	IFNET_WLOCK();
453 	old = if_grow();				/* create initial table */
454 	IFNET_WUNLOCK();
455 	epoch_wait_preempt(net_epoch_preempt);
456 	free(old, M_IFNET);
457 	vnet_if_clone_init();
458 }
459 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
460     NULL);
461 
462 #ifdef VIMAGE
463 static void
464 vnet_if_uninit(const void *unused __unused)
465 {
466 
467 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
468 	    "not empty", __func__, __LINE__, &V_ifnet));
469 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
470 	    "not empty", __func__, __LINE__, &V_ifg_head));
471 
472 	free((caddr_t)V_ifindex_table, M_IFNET);
473 }
474 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
475     vnet_if_uninit, NULL);
476 
477 static void
478 vnet_if_return(const void *unused __unused)
479 {
480 	struct ifnet *ifp, *nifp;
481 
482 	/* Return all inherited interfaces to their parent vnets. */
483 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
484 		if (ifp->if_home_vnet != ifp->if_vnet)
485 			if_vmove(ifp, ifp->if_home_vnet);
486 	}
487 }
488 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
489     vnet_if_return, NULL);
490 #endif
491 
492 
493 static void *
494 if_grow(void)
495 {
496 	int oldlim;
497 	u_int n;
498 	struct ifnet **e;
499 	void *old;
500 
501 	old = NULL;
502 	IFNET_WLOCK_ASSERT();
503 	oldlim = V_if_indexlim;
504 	IFNET_WUNLOCK();
505 	n = (oldlim << 1) * sizeof(*e);
506 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
507 	IFNET_WLOCK();
508 	if (V_if_indexlim != oldlim) {
509 		free(e, M_IFNET);
510 		return (NULL);
511 	}
512 	if (V_ifindex_table != NULL) {
513 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
514 		old = V_ifindex_table;
515 	}
516 	V_if_indexlim <<= 1;
517 	V_ifindex_table = e;
518 	return (old);
519 }
520 
521 /*
522  * Allocate a struct ifnet and an index for an interface.  A layer 2
523  * common structure will also be allocated if an allocation routine is
524  * registered for the passed type.
525  */
526 struct ifnet *
527 if_alloc(u_char type)
528 {
529 	struct ifnet *ifp;
530 	u_short idx;
531 	void *old;
532 
533 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
534  restart:
535 	IFNET_WLOCK();
536 	idx = ifindex_alloc(&old);
537 	if (__predict_false(idx == USHRT_MAX)) {
538 		IFNET_WUNLOCK();
539 		epoch_wait_preempt(net_epoch_preempt);
540 		free(old, M_IFNET);
541 		goto restart;
542 	}
543 	ifnet_setbyindex(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 	CK_STAILQ_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 static void
613 if_destroy(epoch_context_t ctx)
614 {
615 	struct ifnet *ifp;
616 
617 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
618 	if_free_internal(ifp);
619 }
620 
621 /*
622  * Deregister an interface and free the associated storage.
623  */
624 void
625 if_free(struct ifnet *ifp)
626 {
627 
628 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
629 
630 	CURVNET_SET_QUIET(ifp->if_vnet);
631 	IFNET_WLOCK();
632 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
633 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
634 
635 	ifindex_free_locked(ifp->if_index);
636 	IFNET_WUNLOCK();
637 
638 	if (refcount_release(&ifp->if_refcount))
639 		epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
640 	CURVNET_RESTORE();
641 }
642 
643 /*
644  * Interfaces to keep an ifnet type-stable despite the possibility of the
645  * driver calling if_free().  If there are additional references, we defer
646  * freeing the underlying data structure.
647  */
648 void
649 if_ref(struct ifnet *ifp)
650 {
651 
652 	/* We don't assert the ifnet list lock here, but arguably should. */
653 	refcount_acquire(&ifp->if_refcount);
654 }
655 
656 void
657 if_rele(struct ifnet *ifp)
658 {
659 
660 	if (!refcount_release(&ifp->if_refcount))
661 		return;
662 	epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
663 }
664 
665 void
666 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
667 {
668 
669 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
670 
671 	if (ifq->ifq_maxlen == 0)
672 		ifq->ifq_maxlen = ifqmaxlen;
673 
674 	ifq->altq_type = 0;
675 	ifq->altq_disc = NULL;
676 	ifq->altq_flags &= ALTQF_CANTCHANGE;
677 	ifq->altq_tbr  = NULL;
678 	ifq->altq_ifp  = ifp;
679 }
680 
681 void
682 ifq_delete(struct ifaltq *ifq)
683 {
684 	mtx_destroy(&ifq->ifq_mtx);
685 }
686 
687 /*
688  * Perform generic interface initialization tasks and attach the interface
689  * to the list of "active" interfaces.  If vmove flag is set on entry
690  * to if_attach_internal(), perform only a limited subset of initialization
691  * tasks, given that we are moving from one vnet to another an ifnet which
692  * has already been fully initialized.
693  *
694  * Note that if_detach_internal() removes group membership unconditionally
695  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
696  * Thus, when if_vmove() is applied to a cloned interface, group membership
697  * is lost while a cloned one always joins a group whose name is
698  * ifc->ifc_name.  To recover this after if_detach_internal() and
699  * if_attach_internal(), the cloner should be specified to
700  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
701  * attempts to join a group whose name is ifc->ifc_name.
702  *
703  * XXX:
704  *  - The decision to return void and thus require this function to
705  *    succeed is questionable.
706  *  - We should probably do more sanity checking.  For instance we don't
707  *    do anything to insure if_xname is unique or non-empty.
708  */
709 void
710 if_attach(struct ifnet *ifp)
711 {
712 
713 	if_attach_internal(ifp, 0, NULL);
714 }
715 
716 /*
717  * Compute the least common TSO limit.
718  */
719 void
720 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
721 {
722 	/*
723 	 * 1) If there is no limit currently, take the limit from
724 	 * the network adapter.
725 	 *
726 	 * 2) If the network adapter has a limit below the current
727 	 * limit, apply it.
728 	 */
729 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
730 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
731 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
732 	}
733 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
734 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
735 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
736 	}
737 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
738 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
739 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
740 	}
741 }
742 
743 /*
744  * Update TSO limit of a network adapter.
745  *
746  * Returns zero if no change. Else non-zero.
747  */
748 int
749 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
750 {
751 	int retval = 0;
752 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
753 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
754 		retval++;
755 	}
756 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
757 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
758 		retval++;
759 	}
760 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
761 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
762 		retval++;
763 	}
764 	return (retval);
765 }
766 
767 static void
768 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
769 {
770 	unsigned socksize, ifasize;
771 	int namelen, masklen;
772 	struct sockaddr_dl *sdl;
773 	struct ifaddr *ifa;
774 
775 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
776 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
777 		    ifp->if_xname);
778 
779 #ifdef VIMAGE
780 	ifp->if_vnet = curvnet;
781 	if (ifp->if_home_vnet == NULL)
782 		ifp->if_home_vnet = curvnet;
783 #endif
784 
785 	if_addgroup(ifp, IFG_ALL);
786 
787 	/* Restore group membership for cloned interfaces. */
788 	if (vmove && ifc != NULL)
789 		if_clone_addgroup(ifp, ifc);
790 
791 	getmicrotime(&ifp->if_lastchange);
792 	ifp->if_epoch = time_uptime;
793 
794 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
795 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
796 	    ("transmit and qflush must both either be set or both be NULL"));
797 	if (ifp->if_transmit == NULL) {
798 		ifp->if_transmit = if_transmit;
799 		ifp->if_qflush = if_qflush;
800 	}
801 	if (ifp->if_input == NULL)
802 		ifp->if_input = if_input_default;
803 
804 	if (ifp->if_requestencap == NULL)
805 		ifp->if_requestencap = if_requestencap_default;
806 
807 	if (!vmove) {
808 #ifdef MAC
809 		mac_ifnet_create(ifp);
810 #endif
811 
812 		/*
813 		 * Create a Link Level name for this device.
814 		 */
815 		namelen = strlen(ifp->if_xname);
816 		/*
817 		 * Always save enough space for any possiable name so we
818 		 * can do a rename in place later.
819 		 */
820 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
821 		socksize = masklen + ifp->if_addrlen;
822 		if (socksize < sizeof(*sdl))
823 			socksize = sizeof(*sdl);
824 		socksize = roundup2(socksize, sizeof(long));
825 		ifasize = sizeof(*ifa) + 2 * socksize;
826 		ifa = ifa_alloc(ifasize, M_WAITOK);
827 		sdl = (struct sockaddr_dl *)(ifa + 1);
828 		sdl->sdl_len = socksize;
829 		sdl->sdl_family = AF_LINK;
830 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
831 		sdl->sdl_nlen = namelen;
832 		sdl->sdl_index = ifp->if_index;
833 		sdl->sdl_type = ifp->if_type;
834 		ifp->if_addr = ifa;
835 		ifa->ifa_ifp = ifp;
836 		ifa->ifa_rtrequest = link_rtrequest;
837 		ifa->ifa_addr = (struct sockaddr *)sdl;
838 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
839 		ifa->ifa_netmask = (struct sockaddr *)sdl;
840 		sdl->sdl_len = masklen;
841 		while (namelen != 0)
842 			sdl->sdl_data[--namelen] = 0xff;
843 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
844 		/* Reliably crash if used uninitialized. */
845 		ifp->if_broadcastaddr = NULL;
846 
847 		if (ifp->if_type == IFT_ETHER) {
848 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
849 			    M_WAITOK | M_ZERO);
850 		}
851 
852 #if defined(INET) || defined(INET6)
853 		/* Use defaults for TSO, if nothing is set */
854 		if (ifp->if_hw_tsomax == 0 &&
855 		    ifp->if_hw_tsomaxsegcount == 0 &&
856 		    ifp->if_hw_tsomaxsegsize == 0) {
857 			/*
858 			 * The TSO defaults needs to be such that an
859 			 * NFS mbuf list of 35 mbufs totalling just
860 			 * below 64K works and that a chain of mbufs
861 			 * can be defragged into at most 32 segments:
862 			 */
863 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
864 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
865 			ifp->if_hw_tsomaxsegcount = 35;
866 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
867 
868 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
869 			if (ifp->if_capabilities & IFCAP_TSO) {
870 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
871 				    ifp->if_hw_tsomax,
872 				    ifp->if_hw_tsomaxsegcount,
873 				    ifp->if_hw_tsomaxsegsize);
874 			}
875 		}
876 #endif
877 	}
878 #ifdef VIMAGE
879 	else {
880 		/*
881 		 * Update the interface index in the link layer address
882 		 * of the interface.
883 		 */
884 		for (ifa = ifp->if_addr; ifa != NULL;
885 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
886 			if (ifa->ifa_addr->sa_family == AF_LINK) {
887 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
888 				sdl->sdl_index = ifp->if_index;
889 			}
890 		}
891 	}
892 #endif
893 
894 	IFNET_WLOCK();
895 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
896 #ifdef VIMAGE
897 	curvnet->vnet_ifcnt++;
898 #endif
899 	IFNET_WUNLOCK();
900 
901 	if (domain_init_status >= 2)
902 		if_attachdomain1(ifp);
903 
904 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
905 	if (IS_DEFAULT_VNET(curvnet))
906 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
907 
908 	/* Announce the interface. */
909 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
910 }
911 
912 static void
913 if_epochalloc(void *dummy __unused)
914 {
915 
916 	net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT);
917 	net_epoch = epoch_alloc(0);
918 }
919 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
920     if_epochalloc, NULL);
921 
922 static void
923 if_attachdomain(void *dummy)
924 {
925 	struct ifnet *ifp;
926 
927 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
928 		if_attachdomain1(ifp);
929 }
930 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
931     if_attachdomain, NULL);
932 
933 static void
934 if_attachdomain1(struct ifnet *ifp)
935 {
936 	struct domain *dp;
937 
938 	/*
939 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
940 	 * cannot lock ifp->if_afdata initialization, entirely.
941 	 */
942 	IF_AFDATA_LOCK(ifp);
943 	if (ifp->if_afdata_initialized >= domain_init_status) {
944 		IF_AFDATA_UNLOCK(ifp);
945 		log(LOG_WARNING, "%s called more than once on %s\n",
946 		    __func__, ifp->if_xname);
947 		return;
948 	}
949 	ifp->if_afdata_initialized = domain_init_status;
950 	IF_AFDATA_UNLOCK(ifp);
951 
952 	/* address family dependent data region */
953 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
954 	for (dp = domains; dp; dp = dp->dom_next) {
955 		if (dp->dom_ifattach)
956 			ifp->if_afdata[dp->dom_family] =
957 			    (*dp->dom_ifattach)(ifp);
958 	}
959 }
960 
961 /*
962  * Remove any unicast or broadcast network addresses from an interface.
963  */
964 void
965 if_purgeaddrs(struct ifnet *ifp)
966 {
967 	struct ifaddr *ifa;
968 
969 	while (1) {
970 		NET_EPOCH_ENTER();
971 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
972 			if (ifa->ifa_addr->sa_family != AF_LINK)
973 				break;
974 		}
975 		NET_EPOCH_EXIT();
976 
977 		if (ifa == NULL)
978 			break;
979 #ifdef INET
980 		/* XXX: Ugly!! ad hoc just for INET */
981 		if (ifa->ifa_addr->sa_family == AF_INET) {
982 			struct ifaliasreq ifr;
983 
984 			bzero(&ifr, sizeof(ifr));
985 			ifr.ifra_addr = *ifa->ifa_addr;
986 			if (ifa->ifa_dstaddr)
987 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
988 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
989 			    NULL) == 0)
990 				continue;
991 		}
992 #endif /* INET */
993 #ifdef INET6
994 		if (ifa->ifa_addr->sa_family == AF_INET6) {
995 			in6_purgeaddr(ifa);
996 			/* ifp_addrhead is already updated */
997 			continue;
998 		}
999 #endif /* INET6 */
1000 		IF_ADDR_WLOCK(ifp);
1001 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1002 		IF_ADDR_WUNLOCK(ifp);
1003 		ifa_free(ifa);
1004 	}
1005 }
1006 
1007 /*
1008  * Remove any multicast network addresses from an interface when an ifnet
1009  * is going away.
1010  */
1011 static void
1012 if_purgemaddrs(struct ifnet *ifp)
1013 {
1014 	struct ifmultiaddr *ifma;
1015 
1016 	IF_ADDR_WLOCK(ifp);
1017 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1018 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1019 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1020 		if_delmulti_locked(ifp, ifma, 1);
1021 	}
1022 	IF_ADDR_WUNLOCK(ifp);
1023 }
1024 
1025 /*
1026  * Detach an interface, removing it from the list of "active" interfaces.
1027  * If vmove flag is set on entry to if_detach_internal(), perform only a
1028  * limited subset of cleanup tasks, given that we are moving an ifnet from
1029  * one vnet to another, where it must be fully operational.
1030  *
1031  * XXXRW: There are some significant questions about event ordering, and
1032  * how to prevent things from starting to use the interface during detach.
1033  */
1034 void
1035 if_detach(struct ifnet *ifp)
1036 {
1037 
1038 	CURVNET_SET_QUIET(ifp->if_vnet);
1039 	if_detach_internal(ifp, 0, NULL);
1040 	CURVNET_RESTORE();
1041 }
1042 
1043 /*
1044  * The vmove flag, if set, indicates that we are called from a callpath
1045  * that is moving an interface to a different vnet instance.
1046  *
1047  * The shutdown flag, if set, indicates that we are called in the
1048  * process of shutting down a vnet instance.  Currently only the
1049  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1050  * on a vnet instance shutdown without this flag being set, e.g., when
1051  * the cloned interfaces are destoyed as first thing of teardown.
1052  */
1053 static int
1054 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1055 {
1056 	struct ifaddr *ifa;
1057 	int i;
1058 	struct domain *dp;
1059  	struct ifnet *iter;
1060  	int found = 0;
1061 #ifdef VIMAGE
1062 	int shutdown;
1063 
1064 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1065 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1066 #endif
1067 	IFNET_WLOCK();
1068 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1069 		if (iter == ifp) {
1070 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1071 			found = 1;
1072 			break;
1073 		}
1074 	IFNET_WUNLOCK();
1075 	if (!found) {
1076 		/*
1077 		 * While we would want to panic here, we cannot
1078 		 * guarantee that the interface is indeed still on
1079 		 * the list given we don't hold locks all the way.
1080 		 */
1081 		return (ENOENT);
1082 #if 0
1083 		if (vmove)
1084 			panic("%s: ifp=%p not on the ifnet tailq %p",
1085 			    __func__, ifp, &V_ifnet);
1086 		else
1087 			return; /* XXX this should panic as well? */
1088 #endif
1089 	}
1090 
1091 	/*
1092 	 * At this point we know the interface still was on the ifnet list
1093 	 * and we removed it so we are in a stable state.
1094 	 */
1095 #ifdef VIMAGE
1096 	curvnet->vnet_ifcnt--;
1097 #endif
1098 	epoch_wait_preempt(net_epoch_preempt);
1099 	/*
1100 	 * In any case (destroy or vmove) detach us from the groups
1101 	 * and remove/wait for pending events on the taskq.
1102 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1103 	 */
1104 	if_delgroups(ifp);
1105 
1106 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1107 
1108 	/*
1109 	 * Check if this is a cloned interface or not. Must do even if
1110 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1111 	 * the if_clone_addgroup() will have a corrupted string overwise
1112 	 * from a gibberish pointer.
1113 	 */
1114 	if (vmove && ifcp != NULL)
1115 		*ifcp = if_clone_findifc(ifp);
1116 
1117 	if_down(ifp);
1118 
1119 #ifdef VIMAGE
1120 	/*
1121 	 * On VNET shutdown abort here as the stack teardown will do all
1122 	 * the work top-down for us.
1123 	 */
1124 	if (shutdown) {
1125 		/* Give interface users the chance to clean up. */
1126 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1127 
1128 		/*
1129 		 * In case of a vmove we are done here without error.
1130 		 * If we would signal an error it would lead to the same
1131 		 * abort as if we did not find the ifnet anymore.
1132 		 * if_detach() calls us in void context and does not care
1133 		 * about an early abort notification, so life is splendid :)
1134 		 */
1135 		goto finish_vnet_shutdown;
1136 	}
1137 #endif
1138 
1139 	/*
1140 	 * At this point we are not tearing down a VNET and are either
1141 	 * going to destroy or vmove the interface and have to cleanup
1142 	 * accordingly.
1143 	 */
1144 
1145 	/*
1146 	 * Remove routes and flush queues.
1147 	 */
1148 #ifdef ALTQ
1149 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1150 		altq_disable(&ifp->if_snd);
1151 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1152 		altq_detach(&ifp->if_snd);
1153 #endif
1154 
1155 	if_purgeaddrs(ifp);
1156 
1157 #ifdef INET
1158 	in_ifdetach(ifp);
1159 #endif
1160 
1161 #ifdef INET6
1162 	/*
1163 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1164 	 * before removing routing entries below, since IPv6 interface direct
1165 	 * routes are expected to be removed by the IPv6-specific kernel API.
1166 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1167 	 */
1168 	in6_ifdetach(ifp);
1169 #endif
1170 	if_purgemaddrs(ifp);
1171 
1172 	/* Announce that the interface is gone. */
1173 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1174 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1175 	if (IS_DEFAULT_VNET(curvnet))
1176 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1177 
1178 	if (!vmove) {
1179 		/*
1180 		 * Prevent further calls into the device driver via ifnet.
1181 		 */
1182 		if_dead(ifp);
1183 
1184 		/*
1185 		 * Remove link ifaddr pointer and maybe decrement if_index.
1186 		 * Clean up all addresses.
1187 		 */
1188 		free(ifp->if_hw_addr, M_IFADDR);
1189 		ifp->if_hw_addr = NULL;
1190 		ifp->if_addr = NULL;
1191 
1192 		/* We can now free link ifaddr. */
1193 		IF_ADDR_WLOCK(ifp);
1194 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1195 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1196 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1197 			IF_ADDR_WUNLOCK(ifp);
1198 			ifa_free(ifa);
1199 		} else
1200 			IF_ADDR_WUNLOCK(ifp);
1201 	}
1202 
1203 	rt_flushifroutes(ifp);
1204 
1205 #ifdef VIMAGE
1206 finish_vnet_shutdown:
1207 #endif
1208 	/*
1209 	 * We cannot hold the lock over dom_ifdetach calls as they might
1210 	 * sleep, for example trying to drain a callout, thus open up the
1211 	 * theoretical race with re-attaching.
1212 	 */
1213 	IF_AFDATA_LOCK(ifp);
1214 	i = ifp->if_afdata_initialized;
1215 	ifp->if_afdata_initialized = 0;
1216 	IF_AFDATA_UNLOCK(ifp);
1217 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1218 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1219 			(*dp->dom_ifdetach)(ifp,
1220 			    ifp->if_afdata[dp->dom_family]);
1221 			ifp->if_afdata[dp->dom_family] = NULL;
1222 		}
1223 	}
1224 
1225 	return (0);
1226 }
1227 
1228 #ifdef VIMAGE
1229 /*
1230  * if_vmove() performs a limited version of if_detach() in current
1231  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1232  * An attempt is made to shrink if_index in current vnet, find an
1233  * unused if_index in target vnet and calls if_grow() if necessary,
1234  * and finally find an unused if_xname for the target vnet.
1235  */
1236 static void
1237 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1238 {
1239 	struct if_clone *ifc;
1240 	u_int bif_dlt, bif_hdrlen;
1241 	void *old;
1242 	int rc;
1243 
1244  	/*
1245 	 * if_detach_internal() will call the eventhandler to notify
1246 	 * interface departure.  That will detach if_bpf.  We need to
1247 	 * safe the dlt and hdrlen so we can re-attach it later.
1248 	 */
1249 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1250 
1251 	/*
1252 	 * Detach from current vnet, but preserve LLADDR info, do not
1253 	 * mark as dead etc. so that the ifnet can be reattached later.
1254 	 * If we cannot find it, we lost the race to someone else.
1255 	 */
1256 	rc = if_detach_internal(ifp, 1, &ifc);
1257 	if (rc != 0)
1258 		return;
1259 
1260 	/*
1261 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1262 	 * the if_index for that vnet if possible.
1263 	 *
1264 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1265 	 * or we'd lock on one vnet and unlock on another.
1266 	 */
1267 	IFNET_WLOCK();
1268 	ifindex_free_locked(ifp->if_index);
1269 	IFNET_WUNLOCK();
1270 
1271 	/*
1272 	 * Perform interface-specific reassignment tasks, if provided by
1273 	 * the driver.
1274 	 */
1275 	if (ifp->if_reassign != NULL)
1276 		ifp->if_reassign(ifp, new_vnet, NULL);
1277 
1278 	/*
1279 	 * Switch to the context of the target vnet.
1280 	 */
1281 	CURVNET_SET_QUIET(new_vnet);
1282  restart:
1283 	IFNET_WLOCK();
1284 	ifp->if_index = ifindex_alloc(&old);
1285 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1286 		IFNET_WUNLOCK();
1287 		epoch_wait_preempt(net_epoch_preempt);
1288 		free(old, M_IFNET);
1289 		goto restart;
1290 	}
1291 	ifnet_setbyindex(ifp->if_index, ifp);
1292 	IFNET_WUNLOCK();
1293 
1294 	if_attach_internal(ifp, 1, ifc);
1295 
1296 	if (ifp->if_bpf == NULL)
1297 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1298 
1299 	CURVNET_RESTORE();
1300 }
1301 
1302 /*
1303  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1304  */
1305 static int
1306 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1307 {
1308 	struct prison *pr;
1309 	struct ifnet *difp;
1310 	int shutdown;
1311 
1312 	/* Try to find the prison within our visibility. */
1313 	sx_slock(&allprison_lock);
1314 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1315 	sx_sunlock(&allprison_lock);
1316 	if (pr == NULL)
1317 		return (ENXIO);
1318 	prison_hold_locked(pr);
1319 	mtx_unlock(&pr->pr_mtx);
1320 
1321 	/* Do not try to move the iface from and to the same prison. */
1322 	if (pr->pr_vnet == ifp->if_vnet) {
1323 		prison_free(pr);
1324 		return (EEXIST);
1325 	}
1326 
1327 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1328 	/* XXX Lock interfaces to avoid races. */
1329 	CURVNET_SET_QUIET(pr->pr_vnet);
1330 	difp = ifunit(ifname);
1331 	if (difp != NULL) {
1332 		CURVNET_RESTORE();
1333 		prison_free(pr);
1334 		return (EEXIST);
1335 	}
1336 
1337 	/* Make sure the VNET is stable. */
1338 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1339 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1340 	if (shutdown) {
1341 		CURVNET_RESTORE();
1342 		prison_free(pr);
1343 		return (EBUSY);
1344 	}
1345 	CURVNET_RESTORE();
1346 
1347 	/* Move the interface into the child jail/vnet. */
1348 	if_vmove(ifp, pr->pr_vnet);
1349 
1350 	/* Report the new if_xname back to the userland. */
1351 	sprintf(ifname, "%s", ifp->if_xname);
1352 
1353 	prison_free(pr);
1354 	return (0);
1355 }
1356 
1357 static int
1358 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1359 {
1360 	struct prison *pr;
1361 	struct vnet *vnet_dst;
1362 	struct ifnet *ifp;
1363  	int shutdown;
1364 
1365 	/* Try to find the prison within our visibility. */
1366 	sx_slock(&allprison_lock);
1367 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1368 	sx_sunlock(&allprison_lock);
1369 	if (pr == NULL)
1370 		return (ENXIO);
1371 	prison_hold_locked(pr);
1372 	mtx_unlock(&pr->pr_mtx);
1373 
1374 	/* Make sure the named iface exists in the source prison/vnet. */
1375 	CURVNET_SET(pr->pr_vnet);
1376 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1377 	if (ifp == NULL) {
1378 		CURVNET_RESTORE();
1379 		prison_free(pr);
1380 		return (ENXIO);
1381 	}
1382 
1383 	/* Do not try to move the iface from and to the same prison. */
1384 	vnet_dst = TD_TO_VNET(td);
1385 	if (vnet_dst == ifp->if_vnet) {
1386 		CURVNET_RESTORE();
1387 		prison_free(pr);
1388 		return (EEXIST);
1389 	}
1390 
1391 	/* Make sure the VNET is stable. */
1392 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1393 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1394 	if (shutdown) {
1395 		CURVNET_RESTORE();
1396 		prison_free(pr);
1397 		return (EBUSY);
1398 	}
1399 
1400 	/* Get interface back from child jail/vnet. */
1401 	if_vmove(ifp, vnet_dst);
1402 	CURVNET_RESTORE();
1403 
1404 	/* Report the new if_xname back to the userland. */
1405 	sprintf(ifname, "%s", ifp->if_xname);
1406 
1407 	prison_free(pr);
1408 	return (0);
1409 }
1410 #endif /* VIMAGE */
1411 
1412 /*
1413  * Add a group to an interface
1414  */
1415 int
1416 if_addgroup(struct ifnet *ifp, const char *groupname)
1417 {
1418 	struct ifg_list		*ifgl;
1419 	struct ifg_group	*ifg = NULL;
1420 	struct ifg_member	*ifgm;
1421 	int 			 new = 0;
1422 
1423 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1424 	    groupname[strlen(groupname) - 1] <= '9')
1425 		return (EINVAL);
1426 
1427 	IFNET_WLOCK();
1428 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1429 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1430 			IFNET_WUNLOCK();
1431 			return (EEXIST);
1432 		}
1433 
1434 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1435 	    M_NOWAIT)) == NULL) {
1436 	    	IFNET_WUNLOCK();
1437 		return (ENOMEM);
1438 	}
1439 
1440 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1441 	    M_TEMP, M_NOWAIT)) == NULL) {
1442 		free(ifgl, M_TEMP);
1443 		IFNET_WUNLOCK();
1444 		return (ENOMEM);
1445 	}
1446 
1447 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1448 		if (!strcmp(ifg->ifg_group, groupname))
1449 			break;
1450 
1451 	if (ifg == NULL) {
1452 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1453 		    M_TEMP, M_NOWAIT)) == NULL) {
1454 			free(ifgl, M_TEMP);
1455 			free(ifgm, M_TEMP);
1456 			IFNET_WUNLOCK();
1457 			return (ENOMEM);
1458 		}
1459 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1460 		ifg->ifg_refcnt = 0;
1461 		CK_STAILQ_INIT(&ifg->ifg_members);
1462 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1463 		new = 1;
1464 	}
1465 
1466 	ifg->ifg_refcnt++;
1467 	ifgl->ifgl_group = ifg;
1468 	ifgm->ifgm_ifp = ifp;
1469 
1470 	IF_ADDR_WLOCK(ifp);
1471 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1472 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1473 	IF_ADDR_WUNLOCK(ifp);
1474 
1475 	IFNET_WUNLOCK();
1476 
1477 	if (new)
1478 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1479 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1480 
1481 	return (0);
1482 }
1483 
1484 /*
1485  * Remove a group from an interface
1486  */
1487 int
1488 if_delgroup(struct ifnet *ifp, const char *groupname)
1489 {
1490 	struct ifg_list		*ifgl;
1491 	struct ifg_member	*ifgm;
1492 	int freeifgl;
1493 
1494 	IFNET_WLOCK();
1495 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1496 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1497 			break;
1498 	if (ifgl == NULL) {
1499 		IFNET_WUNLOCK();
1500 		return (ENOENT);
1501 	}
1502 
1503 	freeifgl = 0;
1504 	IF_ADDR_WLOCK(ifp);
1505 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1506 	IF_ADDR_WUNLOCK(ifp);
1507 
1508 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1509 		if (ifgm->ifgm_ifp == ifp)
1510 			break;
1511 
1512 	if (ifgm != NULL)
1513 		CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1514 
1515 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1516 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1517 		freeifgl = 1;
1518 	}
1519 	IFNET_WUNLOCK();
1520 
1521 	epoch_wait_preempt(net_epoch_preempt);
1522 	if (freeifgl) {
1523 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1524 		free(ifgl->ifgl_group, M_TEMP);
1525 	}
1526 	free(ifgm, M_TEMP);
1527 	free(ifgl, M_TEMP);
1528 
1529 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1530 
1531 	return (0);
1532 }
1533 
1534 /*
1535  * Remove an interface from all groups
1536  */
1537 static void
1538 if_delgroups(struct ifnet *ifp)
1539 {
1540 	struct ifg_list		*ifgl;
1541 	struct ifg_member	*ifgm;
1542 	char groupname[IFNAMSIZ];
1543 	int ifglfree;
1544 
1545 	IFNET_WLOCK();
1546 	while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1547 		ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1548 
1549 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1550 
1551 		IF_ADDR_WLOCK(ifp);
1552 		CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1553 		IF_ADDR_WUNLOCK(ifp);
1554 
1555 		CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1556 			if (ifgm->ifgm_ifp == ifp)
1557 				break;
1558 
1559 		if (ifgm != NULL)
1560 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1561 			    ifgm_next);
1562 		ifglfree = 0;
1563 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1564 			CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1565 			ifglfree = 1;
1566 		}
1567 
1568 		IFNET_WUNLOCK();
1569 		epoch_wait_preempt(net_epoch_preempt);
1570 		free(ifgm, M_TEMP);
1571 		if (ifglfree) {
1572 			EVENTHANDLER_INVOKE(group_detach_event,
1573 								ifgl->ifgl_group);
1574 			free(ifgl->ifgl_group, M_TEMP);
1575 		}
1576 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1577 
1578 		IFNET_WLOCK();
1579 	}
1580 	IFNET_WUNLOCK();
1581 }
1582 
1583 static char *
1584 ifgr_group_get(void *ifgrp)
1585 {
1586 	union ifgroupreq_union *ifgrup;
1587 
1588 	ifgrup = ifgrp;
1589 #ifdef COMPAT_FREEBSD32
1590 	if (SV_CURPROC_FLAG(SV_ILP32))
1591 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1592 #endif
1593 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1594 }
1595 
1596 static struct ifg_req *
1597 ifgr_groups_get(void *ifgrp)
1598 {
1599 	union ifgroupreq_union *ifgrup;
1600 
1601 	ifgrup = ifgrp;
1602 #ifdef COMPAT_FREEBSD32
1603 	if (SV_CURPROC_FLAG(SV_ILP32))
1604 		return ((struct ifg_req *)(uintptr_t)
1605 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1606 #endif
1607 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1608 }
1609 
1610 /*
1611  * Stores all groups from an interface in memory pointed to by ifgr.
1612  */
1613 static int
1614 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1615 {
1616 	int			 len, error;
1617 	struct ifg_list		*ifgl;
1618 	struct ifg_req		 ifgrq, *ifgp;
1619 
1620 	if (ifgr->ifgr_len == 0) {
1621 		IF_ADDR_RLOCK(ifp);
1622 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1623 			ifgr->ifgr_len += sizeof(struct ifg_req);
1624 		IF_ADDR_RUNLOCK(ifp);
1625 		return (0);
1626 	}
1627 
1628 	len = ifgr->ifgr_len;
1629 	ifgp = ifgr_groups_get(ifgr);
1630 	/* XXX: wire */
1631 	IF_ADDR_RLOCK(ifp);
1632 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1633 		if (len < sizeof(ifgrq)) {
1634 			IF_ADDR_RUNLOCK(ifp);
1635 			return (EINVAL);
1636 		}
1637 		bzero(&ifgrq, sizeof ifgrq);
1638 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1639 		    sizeof(ifgrq.ifgrq_group));
1640 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1641 		    	IF_ADDR_RUNLOCK(ifp);
1642 			return (error);
1643 		}
1644 		len -= sizeof(ifgrq);
1645 		ifgp++;
1646 	}
1647 	IF_ADDR_RUNLOCK(ifp);
1648 
1649 	return (0);
1650 }
1651 
1652 /*
1653  * Stores all members of a group in memory pointed to by igfr
1654  */
1655 static int
1656 if_getgroupmembers(struct ifgroupreq *ifgr)
1657 {
1658 	struct ifg_group	*ifg;
1659 	struct ifg_member	*ifgm;
1660 	struct ifg_req		 ifgrq, *ifgp;
1661 	int			 len, error;
1662 
1663 	IFNET_RLOCK();
1664 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1665 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1666 			break;
1667 	if (ifg == NULL) {
1668 		IFNET_RUNLOCK();
1669 		return (ENOENT);
1670 	}
1671 
1672 	if (ifgr->ifgr_len == 0) {
1673 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1674 			ifgr->ifgr_len += sizeof(ifgrq);
1675 		IFNET_RUNLOCK();
1676 		return (0);
1677 	}
1678 
1679 	len = ifgr->ifgr_len;
1680 	ifgp = ifgr_groups_get(ifgr);
1681 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1682 		if (len < sizeof(ifgrq)) {
1683 			IFNET_RUNLOCK();
1684 			return (EINVAL);
1685 		}
1686 		bzero(&ifgrq, sizeof ifgrq);
1687 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1688 		    sizeof(ifgrq.ifgrq_member));
1689 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1690 			IFNET_RUNLOCK();
1691 			return (error);
1692 		}
1693 		len -= sizeof(ifgrq);
1694 		ifgp++;
1695 	}
1696 	IFNET_RUNLOCK();
1697 
1698 	return (0);
1699 }
1700 
1701 /*
1702  * Return counter values from counter(9)s stored in ifnet.
1703  */
1704 uint64_t
1705 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1706 {
1707 
1708 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1709 
1710 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1711 }
1712 
1713 /*
1714  * Increase an ifnet counter. Usually used for counters shared
1715  * between the stack and a driver, but function supports them all.
1716  */
1717 void
1718 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1719 {
1720 
1721 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1722 
1723 	counter_u64_add(ifp->if_counters[cnt], inc);
1724 }
1725 
1726 /*
1727  * Copy data from ifnet to userland API structure if_data.
1728  */
1729 void
1730 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1731 {
1732 
1733 	ifd->ifi_type = ifp->if_type;
1734 	ifd->ifi_physical = 0;
1735 	ifd->ifi_addrlen = ifp->if_addrlen;
1736 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1737 	ifd->ifi_link_state = ifp->if_link_state;
1738 	ifd->ifi_vhid = 0;
1739 	ifd->ifi_datalen = sizeof(struct if_data);
1740 	ifd->ifi_mtu = ifp->if_mtu;
1741 	ifd->ifi_metric = ifp->if_metric;
1742 	ifd->ifi_baudrate = ifp->if_baudrate;
1743 	ifd->ifi_hwassist = ifp->if_hwassist;
1744 	ifd->ifi_epoch = ifp->if_epoch;
1745 	ifd->ifi_lastchange = ifp->if_lastchange;
1746 
1747 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1748 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1749 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1750 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1751 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1752 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1753 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1754 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1755 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1756 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1757 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1758 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1759 }
1760 
1761 /*
1762  * Wrapper functions for struct ifnet address list locking macros.  These are
1763  * used by kernel modules to avoid encoding programming interface or binary
1764  * interface assumptions that may be violated when kernel-internal locking
1765  * approaches change.
1766  */
1767 void
1768 if_addr_rlock(struct ifnet *ifp)
1769 {
1770 
1771 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1772 }
1773 
1774 void
1775 if_addr_runlock(struct ifnet *ifp)
1776 {
1777 
1778 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1779 }
1780 
1781 void
1782 if_maddr_rlock(if_t ifp)
1783 {
1784 
1785 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1786 }
1787 
1788 void
1789 if_maddr_runlock(if_t ifp)
1790 {
1791 
1792 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1793 }
1794 
1795 /*
1796  * Initialization, destruction and refcounting functions for ifaddrs.
1797  */
1798 struct ifaddr *
1799 ifa_alloc(size_t size, int flags)
1800 {
1801 	struct ifaddr *ifa;
1802 
1803 	KASSERT(size >= sizeof(struct ifaddr),
1804 	    ("%s: invalid size %zu", __func__, size));
1805 
1806 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1807 	if (ifa == NULL)
1808 		return (NULL);
1809 
1810 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1811 		goto fail;
1812 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1813 		goto fail;
1814 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1815 		goto fail;
1816 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1817 		goto fail;
1818 
1819 	refcount_init(&ifa->ifa_refcnt, 1);
1820 
1821 	return (ifa);
1822 
1823 fail:
1824 	/* free(NULL) is okay */
1825 	counter_u64_free(ifa->ifa_opackets);
1826 	counter_u64_free(ifa->ifa_ipackets);
1827 	counter_u64_free(ifa->ifa_obytes);
1828 	counter_u64_free(ifa->ifa_ibytes);
1829 	free(ifa, M_IFADDR);
1830 
1831 	return (NULL);
1832 }
1833 
1834 void
1835 ifa_ref(struct ifaddr *ifa)
1836 {
1837 
1838 	refcount_acquire(&ifa->ifa_refcnt);
1839 }
1840 
1841 static void
1842 ifa_destroy(epoch_context_t ctx)
1843 {
1844 	struct ifaddr *ifa;
1845 
1846 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1847 	counter_u64_free(ifa->ifa_opackets);
1848 	counter_u64_free(ifa->ifa_ipackets);
1849 	counter_u64_free(ifa->ifa_obytes);
1850 	counter_u64_free(ifa->ifa_ibytes);
1851 	free(ifa, M_IFADDR);
1852 }
1853 
1854 void
1855 ifa_free(struct ifaddr *ifa)
1856 {
1857 
1858 	if (refcount_release(&ifa->ifa_refcnt))
1859 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1860 }
1861 
1862 
1863 static int
1864 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1865     struct sockaddr *ia)
1866 {
1867 	int error;
1868 	struct rt_addrinfo info;
1869 	struct sockaddr_dl null_sdl;
1870 	struct ifnet *ifp;
1871 
1872 	ifp = ifa->ifa_ifp;
1873 
1874 	bzero(&info, sizeof(info));
1875 	if (cmd != RTM_DELETE)
1876 		info.rti_ifp = V_loif;
1877 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1878 	info.rti_info[RTAX_DST] = ia;
1879 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1880 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1881 
1882 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1883 
1884 	if (error != 0 &&
1885 	    !(cmd == RTM_ADD && error == EEXIST) &&
1886 	    !(cmd == RTM_DELETE && error == ENOENT))
1887 		if_printf(ifp, "%s failed: %d\n", otype, error);
1888 
1889 	return (error);
1890 }
1891 
1892 int
1893 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1894 {
1895 
1896 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1897 }
1898 
1899 int
1900 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1901 {
1902 
1903 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1904 }
1905 
1906 int
1907 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1908 {
1909 
1910 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1911 }
1912 
1913 /*
1914  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1915  * structs used to represent other address families, it is necessary
1916  * to perform a different comparison.
1917  */
1918 
1919 #define	sa_dl_equal(a1, a2)	\
1920 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1921 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1922 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1923 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1924 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1925 
1926 /*
1927  * Locate an interface based on a complete address.
1928  */
1929 /*ARGSUSED*/
1930 struct ifaddr *
1931 ifa_ifwithaddr(const struct sockaddr *addr)
1932 {
1933 	struct ifnet *ifp;
1934 	struct ifaddr *ifa;
1935 
1936 	MPASS(in_epoch(net_epoch_preempt));
1937 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1938 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1939 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1940 				continue;
1941 			if (sa_equal(addr, ifa->ifa_addr)) {
1942 				goto done;
1943 			}
1944 			/* IP6 doesn't have broadcast */
1945 			if ((ifp->if_flags & IFF_BROADCAST) &&
1946 			    ifa->ifa_broadaddr &&
1947 			    ifa->ifa_broadaddr->sa_len != 0 &&
1948 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1949 				goto done;
1950 			}
1951 		}
1952 	}
1953 	ifa = NULL;
1954 done:
1955 	return (ifa);
1956 }
1957 
1958 int
1959 ifa_ifwithaddr_check(const struct sockaddr *addr)
1960 {
1961 	int rc;
1962 
1963 	NET_EPOCH_ENTER();
1964 	rc = (ifa_ifwithaddr(addr) != NULL);
1965 	NET_EPOCH_EXIT();
1966 	return (rc);
1967 }
1968 
1969 /*
1970  * Locate an interface based on the broadcast address.
1971  */
1972 /* ARGSUSED */
1973 struct ifaddr *
1974 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1975 {
1976 	struct ifnet *ifp;
1977 	struct ifaddr *ifa;
1978 
1979 	MPASS(in_epoch(net_epoch_preempt));
1980 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1981 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1982 			continue;
1983 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1984 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1985 				continue;
1986 			if ((ifp->if_flags & IFF_BROADCAST) &&
1987 			    ifa->ifa_broadaddr &&
1988 			    ifa->ifa_broadaddr->sa_len != 0 &&
1989 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1990 				goto done;
1991 			}
1992 		}
1993 	}
1994 	ifa = NULL;
1995 done:
1996 	return (ifa);
1997 }
1998 
1999 /*
2000  * Locate the point to point interface with a given destination address.
2001  */
2002 /*ARGSUSED*/
2003 struct ifaddr *
2004 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2005 {
2006 	struct ifnet *ifp;
2007 	struct ifaddr *ifa;
2008 
2009 	MPASS(in_epoch(net_epoch_preempt));
2010 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2011 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2012 			continue;
2013 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2014 			continue;
2015 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2016 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2017 				continue;
2018 			if (ifa->ifa_dstaddr != NULL &&
2019 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2020 				goto done;
2021 			}
2022 		}
2023 	}
2024 	ifa = NULL;
2025 done:
2026 	return (ifa);
2027 }
2028 
2029 /*
2030  * Find an interface on a specific network.  If many, choice
2031  * is most specific found.
2032  */
2033 struct ifaddr *
2034 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2035 {
2036 	struct ifnet *ifp;
2037 	struct ifaddr *ifa;
2038 	struct ifaddr *ifa_maybe = NULL;
2039 	u_int af = addr->sa_family;
2040 	const char *addr_data = addr->sa_data, *cplim;
2041 
2042 	MPASS(in_epoch(net_epoch_preempt));
2043 	/*
2044 	 * AF_LINK addresses can be looked up directly by their index number,
2045 	 * so do that if we can.
2046 	 */
2047 	if (af == AF_LINK) {
2048 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2049 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2050 		return (ifaddr_byindex(sdl->sdl_index));
2051 	}
2052 
2053 	/*
2054 	 * Scan though each interface, looking for ones that have addresses
2055 	 * in this address family and the requested fib.  Maintain a reference
2056 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2057 	 * kept it stable when we move onto the next interface.
2058 	 */
2059 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2060 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2061 			continue;
2062 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2063 			const char *cp, *cp2, *cp3;
2064 
2065 			if (ifa->ifa_addr->sa_family != af)
2066 next:				continue;
2067 			if (af == AF_INET &&
2068 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2069 				/*
2070 				 * This is a bit broken as it doesn't
2071 				 * take into account that the remote end may
2072 				 * be a single node in the network we are
2073 				 * looking for.
2074 				 * The trouble is that we don't know the
2075 				 * netmask for the remote end.
2076 				 */
2077 				if (ifa->ifa_dstaddr != NULL &&
2078 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2079 					goto done;
2080 				}
2081 			} else {
2082 				/*
2083 				 * Scan all the bits in the ifa's address.
2084 				 * If a bit dissagrees with what we are
2085 				 * looking for, mask it with the netmask
2086 				 * to see if it really matters.
2087 				 * (A byte at a time)
2088 				 */
2089 				if (ifa->ifa_netmask == 0)
2090 					continue;
2091 				cp = addr_data;
2092 				cp2 = ifa->ifa_addr->sa_data;
2093 				cp3 = ifa->ifa_netmask->sa_data;
2094 				cplim = ifa->ifa_netmask->sa_len
2095 					+ (char *)ifa->ifa_netmask;
2096 				while (cp3 < cplim)
2097 					if ((*cp++ ^ *cp2++) & *cp3++)
2098 						goto next; /* next address! */
2099 				/*
2100 				 * If the netmask of what we just found
2101 				 * is more specific than what we had before
2102 				 * (if we had one), or if the virtual status
2103 				 * of new prefix is better than of the old one,
2104 				 * then remember the new one before continuing
2105 				 * to search for an even better one.
2106 				 */
2107 				if (ifa_maybe == NULL ||
2108 				    ifa_preferred(ifa_maybe, ifa) ||
2109 				    rn_refines((caddr_t)ifa->ifa_netmask,
2110 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2111 					ifa_maybe = ifa;
2112 				}
2113 			}
2114 		}
2115 	}
2116 	ifa = ifa_maybe;
2117 	ifa_maybe = NULL;
2118 done:
2119 	return (ifa);
2120 }
2121 
2122 /*
2123  * Find an interface address specific to an interface best matching
2124  * a given address.
2125  */
2126 struct ifaddr *
2127 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2128 {
2129 	struct ifaddr *ifa;
2130 	const char *cp, *cp2, *cp3;
2131 	char *cplim;
2132 	struct ifaddr *ifa_maybe = NULL;
2133 	u_int af = addr->sa_family;
2134 
2135 	if (af >= AF_MAX)
2136 		return (NULL);
2137 
2138 	MPASS(in_epoch(net_epoch_preempt));
2139 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2140 		if (ifa->ifa_addr->sa_family != af)
2141 			continue;
2142 		if (ifa_maybe == NULL)
2143 			ifa_maybe = ifa;
2144 		if (ifa->ifa_netmask == 0) {
2145 			if (sa_equal(addr, ifa->ifa_addr) ||
2146 			    (ifa->ifa_dstaddr &&
2147 			    sa_equal(addr, ifa->ifa_dstaddr)))
2148 				goto done;
2149 			continue;
2150 		}
2151 		if (ifp->if_flags & IFF_POINTOPOINT) {
2152 			if (sa_equal(addr, ifa->ifa_dstaddr))
2153 				goto done;
2154 		} else {
2155 			cp = addr->sa_data;
2156 			cp2 = ifa->ifa_addr->sa_data;
2157 			cp3 = ifa->ifa_netmask->sa_data;
2158 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2159 			for (; cp3 < cplim; cp3++)
2160 				if ((*cp++ ^ *cp2++) & *cp3)
2161 					break;
2162 			if (cp3 == cplim)
2163 				goto done;
2164 		}
2165 	}
2166 	ifa = ifa_maybe;
2167 done:
2168 	return (ifa);
2169 }
2170 
2171 /*
2172  * See whether new ifa is better than current one:
2173  * 1) A non-virtual one is preferred over virtual.
2174  * 2) A virtual in master state preferred over any other state.
2175  *
2176  * Used in several address selecting functions.
2177  */
2178 int
2179 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2180 {
2181 
2182 	return (cur->ifa_carp && (!next->ifa_carp ||
2183 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2184 }
2185 
2186 #include <net/if_llatbl.h>
2187 
2188 /*
2189  * Default action when installing a route with a Link Level gateway.
2190  * Lookup an appropriate real ifa to point to.
2191  * This should be moved to /sys/net/link.c eventually.
2192  */
2193 static void
2194 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2195 {
2196 	struct ifaddr *ifa, *oifa;
2197 	struct sockaddr *dst;
2198 	struct ifnet *ifp;
2199 
2200 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2201 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2202 		return;
2203 	NET_EPOCH_ENTER();
2204 	ifa = ifaof_ifpforaddr(dst, ifp);
2205 	if (ifa) {
2206 		oifa = rt->rt_ifa;
2207 		if (oifa != ifa) {
2208 			ifa_free(oifa);
2209 			ifa_ref(ifa);
2210 		}
2211 		rt->rt_ifa = ifa;
2212 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2213 			ifa->ifa_rtrequest(cmd, rt, info);
2214 	}
2215 	NET_EPOCH_EXIT();
2216 }
2217 
2218 struct sockaddr_dl *
2219 link_alloc_sdl(size_t size, int flags)
2220 {
2221 
2222 	return (malloc(size, M_TEMP, flags));
2223 }
2224 
2225 void
2226 link_free_sdl(struct sockaddr *sa)
2227 {
2228 	free(sa, M_TEMP);
2229 }
2230 
2231 /*
2232  * Fills in given sdl with interface basic info.
2233  * Returns pointer to filled sdl.
2234  */
2235 struct sockaddr_dl *
2236 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2237 {
2238 	struct sockaddr_dl *sdl;
2239 
2240 	sdl = (struct sockaddr_dl *)paddr;
2241 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2242 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2243 	sdl->sdl_family = AF_LINK;
2244 	sdl->sdl_index = ifp->if_index;
2245 	sdl->sdl_type = iftype;
2246 
2247 	return (sdl);
2248 }
2249 
2250 /*
2251  * Mark an interface down and notify protocols of
2252  * the transition.
2253  */
2254 static void
2255 if_unroute(struct ifnet *ifp, int flag, int fam)
2256 {
2257 	struct ifaddr *ifa;
2258 
2259 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2260 
2261 	ifp->if_flags &= ~flag;
2262 	getmicrotime(&ifp->if_lastchange);
2263 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2264 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2265 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2266 	ifp->if_qflush(ifp);
2267 
2268 	if (ifp->if_carp)
2269 		(*carp_linkstate_p)(ifp);
2270 	rt_ifmsg(ifp);
2271 }
2272 
2273 /*
2274  * Mark an interface up and notify protocols of
2275  * the transition.
2276  */
2277 static void
2278 if_route(struct ifnet *ifp, int flag, int fam)
2279 {
2280 	struct ifaddr *ifa;
2281 
2282 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2283 
2284 	ifp->if_flags |= flag;
2285 	getmicrotime(&ifp->if_lastchange);
2286 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2287 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2288 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2289 	if (ifp->if_carp)
2290 		(*carp_linkstate_p)(ifp);
2291 	rt_ifmsg(ifp);
2292 #ifdef INET6
2293 	in6_if_up(ifp);
2294 #endif
2295 }
2296 
2297 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2298 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2299 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2300 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2301 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2302 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2303 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2304 void	*(*vlan_cookie_p)(struct ifnet *);
2305 
2306 /*
2307  * Handle a change in the interface link state. To avoid LORs
2308  * between driver lock and upper layer locks, as well as possible
2309  * recursions, we post event to taskqueue, and all job
2310  * is done in static do_link_state_change().
2311  */
2312 void
2313 if_link_state_change(struct ifnet *ifp, int link_state)
2314 {
2315 	/* Return if state hasn't changed. */
2316 	if (ifp->if_link_state == link_state)
2317 		return;
2318 
2319 	ifp->if_link_state = link_state;
2320 
2321 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2322 }
2323 
2324 static void
2325 do_link_state_change(void *arg, int pending)
2326 {
2327 	struct ifnet *ifp = (struct ifnet *)arg;
2328 	int link_state = ifp->if_link_state;
2329 	CURVNET_SET(ifp->if_vnet);
2330 
2331 	/* Notify that the link state has changed. */
2332 	rt_ifmsg(ifp);
2333 	if (ifp->if_vlantrunk != NULL)
2334 		(*vlan_link_state_p)(ifp);
2335 
2336 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2337 	    ifp->if_l2com != NULL)
2338 		(*ng_ether_link_state_p)(ifp, link_state);
2339 	if (ifp->if_carp)
2340 		(*carp_linkstate_p)(ifp);
2341 	if (ifp->if_bridge)
2342 		ifp->if_bridge_linkstate(ifp);
2343 	if (ifp->if_lagg)
2344 		(*lagg_linkstate_p)(ifp, link_state);
2345 
2346 	if (IS_DEFAULT_VNET(curvnet))
2347 		devctl_notify("IFNET", ifp->if_xname,
2348 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2349 		    NULL);
2350 	if (pending > 1)
2351 		if_printf(ifp, "%d link states coalesced\n", pending);
2352 	if (log_link_state_change)
2353 		if_printf(ifp, "link state changed to %s\n",
2354 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2355 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2356 	CURVNET_RESTORE();
2357 }
2358 
2359 /*
2360  * Mark an interface down and notify protocols of
2361  * the transition.
2362  */
2363 void
2364 if_down(struct ifnet *ifp)
2365 {
2366 
2367 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2368 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2369 }
2370 
2371 /*
2372  * Mark an interface up and notify protocols of
2373  * the transition.
2374  */
2375 void
2376 if_up(struct ifnet *ifp)
2377 {
2378 
2379 	if_route(ifp, IFF_UP, AF_UNSPEC);
2380 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2381 }
2382 
2383 /*
2384  * Flush an interface queue.
2385  */
2386 void
2387 if_qflush(struct ifnet *ifp)
2388 {
2389 	struct mbuf *m, *n;
2390 	struct ifaltq *ifq;
2391 
2392 	ifq = &ifp->if_snd;
2393 	IFQ_LOCK(ifq);
2394 #ifdef ALTQ
2395 	if (ALTQ_IS_ENABLED(ifq))
2396 		ALTQ_PURGE(ifq);
2397 #endif
2398 	n = ifq->ifq_head;
2399 	while ((m = n) != NULL) {
2400 		n = m->m_nextpkt;
2401 		m_freem(m);
2402 	}
2403 	ifq->ifq_head = 0;
2404 	ifq->ifq_tail = 0;
2405 	ifq->ifq_len = 0;
2406 	IFQ_UNLOCK(ifq);
2407 }
2408 
2409 /*
2410  * Map interface name to interface structure pointer, with or without
2411  * returning a reference.
2412  */
2413 struct ifnet *
2414 ifunit_ref(const char *name)
2415 {
2416 	struct ifnet *ifp;
2417 
2418 	IFNET_RLOCK_NOSLEEP();
2419 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2420 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2421 		    !(ifp->if_flags & IFF_DYING))
2422 			break;
2423 	}
2424 	if (ifp != NULL)
2425 		if_ref(ifp);
2426 	IFNET_RUNLOCK_NOSLEEP();
2427 	return (ifp);
2428 }
2429 
2430 struct ifnet *
2431 ifunit(const char *name)
2432 {
2433 	struct ifnet *ifp;
2434 
2435 	IFNET_RLOCK_NOSLEEP();
2436 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2437 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2438 			break;
2439 	}
2440 	IFNET_RUNLOCK_NOSLEEP();
2441 	return (ifp);
2442 }
2443 
2444 static void *
2445 ifr_buffer_get_buffer(void *data)
2446 {
2447 	union ifreq_union *ifrup;
2448 
2449 	ifrup = data;
2450 #ifdef COMPAT_FREEBSD32
2451 	if (SV_CURPROC_FLAG(SV_ILP32))
2452 		return ((void *)(uintptr_t)
2453 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2454 #endif
2455 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2456 }
2457 
2458 static void
2459 ifr_buffer_set_buffer_null(void *data)
2460 {
2461 	union ifreq_union *ifrup;
2462 
2463 	ifrup = data;
2464 #ifdef COMPAT_FREEBSD32
2465 	if (SV_CURPROC_FLAG(SV_ILP32))
2466 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2467 	else
2468 #endif
2469 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2470 }
2471 
2472 static size_t
2473 ifr_buffer_get_length(void *data)
2474 {
2475 	union ifreq_union *ifrup;
2476 
2477 	ifrup = data;
2478 #ifdef COMPAT_FREEBSD32
2479 	if (SV_CURPROC_FLAG(SV_ILP32))
2480 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2481 #endif
2482 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2483 }
2484 
2485 static void
2486 ifr_buffer_set_length(void *data, size_t len)
2487 {
2488 	union ifreq_union *ifrup;
2489 
2490 	ifrup = data;
2491 #ifdef COMPAT_FREEBSD32
2492 	if (SV_CURPROC_FLAG(SV_ILP32))
2493 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2494 	else
2495 #endif
2496 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2497 }
2498 
2499 void *
2500 ifr_data_get_ptr(void *ifrp)
2501 {
2502 	union ifreq_union *ifrup;
2503 
2504 	ifrup = ifrp;
2505 #ifdef COMPAT_FREEBSD32
2506 	if (SV_CURPROC_FLAG(SV_ILP32))
2507 		return ((void *)(uintptr_t)
2508 		    ifrup->ifr32.ifr_ifru.ifru_data);
2509 #endif
2510 		return (ifrup->ifr.ifr_ifru.ifru_data);
2511 }
2512 
2513 /*
2514  * Hardware specific interface ioctls.
2515  */
2516 int
2517 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2518 {
2519 	struct ifreq *ifr;
2520 	int error = 0, do_ifup = 0;
2521 	int new_flags, temp_flags;
2522 	size_t namelen, onamelen;
2523 	size_t descrlen;
2524 	char *descrbuf, *odescrbuf;
2525 	char new_name[IFNAMSIZ];
2526 	struct ifaddr *ifa;
2527 	struct sockaddr_dl *sdl;
2528 
2529 	ifr = (struct ifreq *)data;
2530 	switch (cmd) {
2531 	case SIOCGIFINDEX:
2532 		ifr->ifr_index = ifp->if_index;
2533 		break;
2534 
2535 	case SIOCGIFFLAGS:
2536 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2537 		ifr->ifr_flags = temp_flags & 0xffff;
2538 		ifr->ifr_flagshigh = temp_flags >> 16;
2539 		break;
2540 
2541 	case SIOCGIFCAP:
2542 		ifr->ifr_reqcap = ifp->if_capabilities;
2543 		ifr->ifr_curcap = ifp->if_capenable;
2544 		break;
2545 
2546 #ifdef MAC
2547 	case SIOCGIFMAC:
2548 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2549 		break;
2550 #endif
2551 
2552 	case SIOCGIFMETRIC:
2553 		ifr->ifr_metric = ifp->if_metric;
2554 		break;
2555 
2556 	case SIOCGIFMTU:
2557 		ifr->ifr_mtu = ifp->if_mtu;
2558 		break;
2559 
2560 	case SIOCGIFPHYS:
2561 		/* XXXGL: did this ever worked? */
2562 		ifr->ifr_phys = 0;
2563 		break;
2564 
2565 	case SIOCGIFDESCR:
2566 		error = 0;
2567 		sx_slock(&ifdescr_sx);
2568 		if (ifp->if_description == NULL)
2569 			error = ENOMSG;
2570 		else {
2571 			/* space for terminating nul */
2572 			descrlen = strlen(ifp->if_description) + 1;
2573 			if (ifr_buffer_get_length(ifr) < descrlen)
2574 				ifr_buffer_set_buffer_null(ifr);
2575 			else
2576 				error = copyout(ifp->if_description,
2577 				    ifr_buffer_get_buffer(ifr), descrlen);
2578 			ifr_buffer_set_length(ifr, descrlen);
2579 		}
2580 		sx_sunlock(&ifdescr_sx);
2581 		break;
2582 
2583 	case SIOCSIFDESCR:
2584 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2585 		if (error)
2586 			return (error);
2587 
2588 		/*
2589 		 * Copy only (length-1) bytes to make sure that
2590 		 * if_description is always nul terminated.  The
2591 		 * length parameter is supposed to count the
2592 		 * terminating nul in.
2593 		 */
2594 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2595 			return (ENAMETOOLONG);
2596 		else if (ifr_buffer_get_length(ifr) == 0)
2597 			descrbuf = NULL;
2598 		else {
2599 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2600 			    M_IFDESCR, M_WAITOK | M_ZERO);
2601 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2602 			    ifr_buffer_get_length(ifr) - 1);
2603 			if (error) {
2604 				free(descrbuf, M_IFDESCR);
2605 				break;
2606 			}
2607 		}
2608 
2609 		sx_xlock(&ifdescr_sx);
2610 		odescrbuf = ifp->if_description;
2611 		ifp->if_description = descrbuf;
2612 		sx_xunlock(&ifdescr_sx);
2613 
2614 		getmicrotime(&ifp->if_lastchange);
2615 		free(odescrbuf, M_IFDESCR);
2616 		break;
2617 
2618 	case SIOCGIFFIB:
2619 		ifr->ifr_fib = ifp->if_fib;
2620 		break;
2621 
2622 	case SIOCSIFFIB:
2623 		error = priv_check(td, PRIV_NET_SETIFFIB);
2624 		if (error)
2625 			return (error);
2626 		if (ifr->ifr_fib >= rt_numfibs)
2627 			return (EINVAL);
2628 
2629 		ifp->if_fib = ifr->ifr_fib;
2630 		break;
2631 
2632 	case SIOCSIFFLAGS:
2633 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2634 		if (error)
2635 			return (error);
2636 		/*
2637 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2638 		 * check, so we don't need special handling here yet.
2639 		 */
2640 		new_flags = (ifr->ifr_flags & 0xffff) |
2641 		    (ifr->ifr_flagshigh << 16);
2642 		if (ifp->if_flags & IFF_UP &&
2643 		    (new_flags & IFF_UP) == 0) {
2644 			if_down(ifp);
2645 		} else if (new_flags & IFF_UP &&
2646 		    (ifp->if_flags & IFF_UP) == 0) {
2647 			do_ifup = 1;
2648 		}
2649 		/* See if permanently promiscuous mode bit is about to flip */
2650 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2651 			if (new_flags & IFF_PPROMISC)
2652 				ifp->if_flags |= IFF_PROMISC;
2653 			else if (ifp->if_pcount == 0)
2654 				ifp->if_flags &= ~IFF_PROMISC;
2655 			if (log_promisc_mode_change)
2656                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2657                                     ((new_flags & IFF_PPROMISC) ?
2658                                      "enabled" : "disabled"));
2659 		}
2660 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2661 			(new_flags &~ IFF_CANTCHANGE);
2662 		if (ifp->if_ioctl) {
2663 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2664 		}
2665 		if (do_ifup)
2666 			if_up(ifp);
2667 		getmicrotime(&ifp->if_lastchange);
2668 		break;
2669 
2670 	case SIOCSIFCAP:
2671 		error = priv_check(td, PRIV_NET_SETIFCAP);
2672 		if (error)
2673 			return (error);
2674 		if (ifp->if_ioctl == NULL)
2675 			return (EOPNOTSUPP);
2676 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2677 			return (EINVAL);
2678 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2679 		if (error == 0)
2680 			getmicrotime(&ifp->if_lastchange);
2681 		break;
2682 
2683 #ifdef MAC
2684 	case SIOCSIFMAC:
2685 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2686 		break;
2687 #endif
2688 
2689 	case SIOCSIFNAME:
2690 		error = priv_check(td, PRIV_NET_SETIFNAME);
2691 		if (error)
2692 			return (error);
2693 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2694 		    NULL);
2695 		if (error != 0)
2696 			return (error);
2697 		if (new_name[0] == '\0')
2698 			return (EINVAL);
2699 		if (new_name[IFNAMSIZ-1] != '\0') {
2700 			new_name[IFNAMSIZ-1] = '\0';
2701 			if (strlen(new_name) == IFNAMSIZ-1)
2702 				return (EINVAL);
2703 		}
2704 		if (ifunit(new_name) != NULL)
2705 			return (EEXIST);
2706 
2707 		/*
2708 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2709 		 * and there are numerous other races with the
2710 		 * ifunit() checks not being atomic with namespace
2711 		 * changes (renames, vmoves, if_attach, etc).
2712 		 */
2713 		ifp->if_flags |= IFF_RENAMING;
2714 
2715 		/* Announce the departure of the interface. */
2716 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2717 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2718 
2719 		if_printf(ifp, "changing name to '%s'\n", new_name);
2720 
2721 		IF_ADDR_WLOCK(ifp);
2722 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2723 		ifa = ifp->if_addr;
2724 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2725 		namelen = strlen(new_name);
2726 		onamelen = sdl->sdl_nlen;
2727 		/*
2728 		 * Move the address if needed.  This is safe because we
2729 		 * allocate space for a name of length IFNAMSIZ when we
2730 		 * create this in if_attach().
2731 		 */
2732 		if (namelen != onamelen) {
2733 			bcopy(sdl->sdl_data + onamelen,
2734 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2735 		}
2736 		bcopy(new_name, sdl->sdl_data, namelen);
2737 		sdl->sdl_nlen = namelen;
2738 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2739 		bzero(sdl->sdl_data, onamelen);
2740 		while (namelen != 0)
2741 			sdl->sdl_data[--namelen] = 0xff;
2742 		IF_ADDR_WUNLOCK(ifp);
2743 
2744 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2745 		/* Announce the return of the interface. */
2746 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2747 
2748 		ifp->if_flags &= ~IFF_RENAMING;
2749 		break;
2750 
2751 #ifdef VIMAGE
2752 	case SIOCSIFVNET:
2753 		error = priv_check(td, PRIV_NET_SETIFVNET);
2754 		if (error)
2755 			return (error);
2756 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2757 		break;
2758 #endif
2759 
2760 	case SIOCSIFMETRIC:
2761 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2762 		if (error)
2763 			return (error);
2764 		ifp->if_metric = ifr->ifr_metric;
2765 		getmicrotime(&ifp->if_lastchange);
2766 		break;
2767 
2768 	case SIOCSIFPHYS:
2769 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2770 		if (error)
2771 			return (error);
2772 		if (ifp->if_ioctl == NULL)
2773 			return (EOPNOTSUPP);
2774 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2775 		if (error == 0)
2776 			getmicrotime(&ifp->if_lastchange);
2777 		break;
2778 
2779 	case SIOCSIFMTU:
2780 	{
2781 		u_long oldmtu = ifp->if_mtu;
2782 
2783 		error = priv_check(td, PRIV_NET_SETIFMTU);
2784 		if (error)
2785 			return (error);
2786 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2787 			return (EINVAL);
2788 		if (ifp->if_ioctl == NULL)
2789 			return (EOPNOTSUPP);
2790 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2791 		if (error == 0) {
2792 			getmicrotime(&ifp->if_lastchange);
2793 			rt_ifmsg(ifp);
2794 #ifdef INET
2795 			NETDUMP_REINIT(ifp);
2796 #endif
2797 		}
2798 		/*
2799 		 * If the link MTU changed, do network layer specific procedure.
2800 		 */
2801 		if (ifp->if_mtu != oldmtu) {
2802 #ifdef INET6
2803 			nd6_setmtu(ifp);
2804 #endif
2805 			rt_updatemtu(ifp);
2806 		}
2807 		break;
2808 	}
2809 
2810 	case SIOCADDMULTI:
2811 	case SIOCDELMULTI:
2812 		if (cmd == SIOCADDMULTI)
2813 			error = priv_check(td, PRIV_NET_ADDMULTI);
2814 		else
2815 			error = priv_check(td, PRIV_NET_DELMULTI);
2816 		if (error)
2817 			return (error);
2818 
2819 		/* Don't allow group membership on non-multicast interfaces. */
2820 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2821 			return (EOPNOTSUPP);
2822 
2823 		/* Don't let users screw up protocols' entries. */
2824 		if (ifr->ifr_addr.sa_family != AF_LINK)
2825 			return (EINVAL);
2826 
2827 		if (cmd == SIOCADDMULTI) {
2828 			struct ifmultiaddr *ifma;
2829 
2830 			/*
2831 			 * Userland is only permitted to join groups once
2832 			 * via the if_addmulti() KPI, because it cannot hold
2833 			 * struct ifmultiaddr * between calls. It may also
2834 			 * lose a race while we check if the membership
2835 			 * already exists.
2836 			 */
2837 			IF_ADDR_RLOCK(ifp);
2838 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2839 			IF_ADDR_RUNLOCK(ifp);
2840 			if (ifma != NULL)
2841 				error = EADDRINUSE;
2842 			else
2843 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2844 		} else {
2845 			error = if_delmulti(ifp, &ifr->ifr_addr);
2846 		}
2847 		if (error == 0)
2848 			getmicrotime(&ifp->if_lastchange);
2849 		break;
2850 
2851 	case SIOCSIFPHYADDR:
2852 	case SIOCDIFPHYADDR:
2853 #ifdef INET6
2854 	case SIOCSIFPHYADDR_IN6:
2855 #endif
2856 	case SIOCSIFMEDIA:
2857 	case SIOCSIFGENERIC:
2858 		error = priv_check(td, PRIV_NET_HWIOCTL);
2859 		if (error)
2860 			return (error);
2861 		if (ifp->if_ioctl == NULL)
2862 			return (EOPNOTSUPP);
2863 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2864 		if (error == 0)
2865 			getmicrotime(&ifp->if_lastchange);
2866 		break;
2867 
2868 	case SIOCGIFSTATUS:
2869 	case SIOCGIFPSRCADDR:
2870 	case SIOCGIFPDSTADDR:
2871 	case SIOCGIFMEDIA:
2872 	case SIOCGIFXMEDIA:
2873 	case SIOCGIFGENERIC:
2874 	case SIOCGIFRSSKEY:
2875 	case SIOCGIFRSSHASH:
2876 		if (ifp->if_ioctl == NULL)
2877 			return (EOPNOTSUPP);
2878 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2879 		break;
2880 
2881 	case SIOCSIFLLADDR:
2882 		error = priv_check(td, PRIV_NET_SETLLADDR);
2883 		if (error)
2884 			return (error);
2885 		error = if_setlladdr(ifp,
2886 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2887 		break;
2888 
2889 	case SIOCGHWADDR:
2890 		error = if_gethwaddr(ifp, ifr);
2891 		break;
2892 
2893 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2894 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2895 		if (error)
2896 			return (error);
2897 		if ((error = if_addgroup(ifp,
2898 		    ifgr_group_get((struct ifgroupreq *)data))))
2899 			return (error);
2900 		break;
2901 
2902 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2903 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2904 			return (error);
2905 		break;
2906 
2907 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2908 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2909 		if (error)
2910 			return (error);
2911 		if ((error = if_delgroup(ifp,
2912 		    ifgr_group_get((struct ifgroupreq *)data))))
2913 			return (error);
2914 		break;
2915 
2916 	default:
2917 		error = ENOIOCTL;
2918 		break;
2919 	}
2920 	return (error);
2921 }
2922 
2923 #ifdef COMPAT_FREEBSD32
2924 struct ifconf32 {
2925 	int32_t	ifc_len;
2926 	union {
2927 		uint32_t	ifcu_buf;
2928 		uint32_t	ifcu_req;
2929 	} ifc_ifcu;
2930 };
2931 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2932 #endif
2933 
2934 #ifdef COMPAT_FREEBSD32
2935 static void
2936 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2937 {
2938 	struct ifmediareq32 *ifmr32;
2939 
2940 	ifmr32 = (struct ifmediareq32 *)data;
2941 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2942 	    sizeof(ifmr->ifm_name));
2943 	ifmr->ifm_current = ifmr32->ifm_current;
2944 	ifmr->ifm_mask = ifmr32->ifm_mask;
2945 	ifmr->ifm_status = ifmr32->ifm_status;
2946 	ifmr->ifm_active = ifmr32->ifm_active;
2947 	ifmr->ifm_count = ifmr32->ifm_count;
2948 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2949 }
2950 
2951 static void
2952 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2953 {
2954 	struct ifmediareq32 *ifmr32;
2955 
2956 	ifmr32 = (struct ifmediareq32 *)data;
2957 	ifmr32->ifm_current = ifmr->ifm_current;
2958 	ifmr32->ifm_mask = ifmr->ifm_mask;
2959 	ifmr32->ifm_status = ifmr->ifm_status;
2960 	ifmr32->ifm_active = ifmr->ifm_active;
2961 	ifmr32->ifm_count = ifmr->ifm_count;
2962 }
2963 #endif
2964 
2965 /*
2966  * Interface ioctls.
2967  */
2968 int
2969 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2970 {
2971 #ifdef COMPAT_FREEBSD32
2972 	caddr_t saved_data = NULL;
2973 	struct ifmediareq ifmr;
2974 	struct ifmediareq *ifmrp;
2975 #endif
2976 	struct ifnet *ifp;
2977 	struct ifreq *ifr;
2978 	int error;
2979 	int oif_flags;
2980 #ifdef VIMAGE
2981 	int shutdown;
2982 #endif
2983 
2984 	CURVNET_SET(so->so_vnet);
2985 #ifdef VIMAGE
2986 	/* Make sure the VNET is stable. */
2987 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2988 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2989 	if (shutdown) {
2990 		CURVNET_RESTORE();
2991 		return (EBUSY);
2992 	}
2993 #endif
2994 
2995 
2996 	switch (cmd) {
2997 	case SIOCGIFCONF:
2998 		error = ifconf(cmd, data);
2999 		CURVNET_RESTORE();
3000 		return (error);
3001 
3002 #ifdef COMPAT_FREEBSD32
3003 	case SIOCGIFCONF32:
3004 		{
3005 			struct ifconf32 *ifc32;
3006 			struct ifconf ifc;
3007 
3008 			ifc32 = (struct ifconf32 *)data;
3009 			ifc.ifc_len = ifc32->ifc_len;
3010 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3011 
3012 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
3013 			CURVNET_RESTORE();
3014 			if (error == 0)
3015 				ifc32->ifc_len = ifc.ifc_len;
3016 			return (error);
3017 		}
3018 #endif
3019 	}
3020 
3021 #ifdef COMPAT_FREEBSD32
3022 	ifmrp = NULL;
3023 	switch (cmd) {
3024 	case SIOCGIFMEDIA32:
3025 	case SIOCGIFXMEDIA32:
3026 		ifmrp = &ifmr;
3027 		ifmr_init(ifmrp, data);
3028 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3029 		saved_data = data;
3030 		data = (caddr_t)ifmrp;
3031 	}
3032 #endif
3033 
3034 	ifr = (struct ifreq *)data;
3035 	switch (cmd) {
3036 #ifdef VIMAGE
3037 	case SIOCSIFRVNET:
3038 		error = priv_check(td, PRIV_NET_SETIFVNET);
3039 		if (error == 0)
3040 			error = if_vmove_reclaim(td, ifr->ifr_name,
3041 			    ifr->ifr_jid);
3042 		goto out_noref;
3043 #endif
3044 	case SIOCIFCREATE:
3045 	case SIOCIFCREATE2:
3046 		error = priv_check(td, PRIV_NET_IFCREATE);
3047 		if (error == 0)
3048 			error = if_clone_create(ifr->ifr_name,
3049 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3050 			    ifr_data_get_ptr(ifr) : NULL);
3051 		goto out_noref;
3052 	case SIOCIFDESTROY:
3053 		error = priv_check(td, PRIV_NET_IFDESTROY);
3054 		if (error == 0)
3055 			error = if_clone_destroy(ifr->ifr_name);
3056 		goto out_noref;
3057 
3058 	case SIOCIFGCLONERS:
3059 		error = if_clone_list((struct if_clonereq *)data);
3060 		goto out_noref;
3061 
3062 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3063 		error = if_getgroupmembers((struct ifgroupreq *)data);
3064 		goto out_noref;
3065 
3066 #if defined(INET) || defined(INET6)
3067 	case SIOCSVH:
3068 	case SIOCGVH:
3069 		if (carp_ioctl_p == NULL)
3070 			error = EPROTONOSUPPORT;
3071 		else
3072 			error = (*carp_ioctl_p)(ifr, cmd, td);
3073 		goto out_noref;
3074 #endif
3075 	}
3076 
3077 	ifp = ifunit_ref(ifr->ifr_name);
3078 	if (ifp == NULL) {
3079 		error = ENXIO;
3080 		goto out_noref;
3081 	}
3082 
3083 	error = ifhwioctl(cmd, ifp, data, td);
3084 	if (error != ENOIOCTL)
3085 		goto out_ref;
3086 
3087 	oif_flags = ifp->if_flags;
3088 	if (so->so_proto == NULL) {
3089 		error = EOPNOTSUPP;
3090 		goto out_ref;
3091 	}
3092 
3093 	/*
3094 	 * Pass the request on to the socket control method, and if the
3095 	 * latter returns EOPNOTSUPP, directly to the interface.
3096 	 *
3097 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3098 	 * trust SIOCSIFADDR et al to come from an already privileged
3099 	 * layer, and do not perform any credentials checks or input
3100 	 * validation.
3101 	 */
3102 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3103 	    ifp, td));
3104 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3105 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3106 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3107 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3108 
3109 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3110 #ifdef INET6
3111 		if (ifp->if_flags & IFF_UP)
3112 			in6_if_up(ifp);
3113 #endif
3114 	}
3115 
3116 out_ref:
3117 	if_rele(ifp);
3118 out_noref:
3119 #ifdef COMPAT_FREEBSD32
3120 	if (ifmrp != NULL) {
3121 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3122 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3123 		     cmd));
3124 		data = saved_data;
3125 		ifmr_update(ifmrp, data);
3126 	}
3127 #endif
3128 	CURVNET_RESTORE();
3129 	return (error);
3130 }
3131 
3132 /*
3133  * The code common to handling reference counted flags,
3134  * e.g., in ifpromisc() and if_allmulti().
3135  * The "pflag" argument can specify a permanent mode flag to check,
3136  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3137  *
3138  * Only to be used on stack-owned flags, not driver-owned flags.
3139  */
3140 static int
3141 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3142 {
3143 	struct ifreq ifr;
3144 	int error;
3145 	int oldflags, oldcount;
3146 
3147 	/* Sanity checks to catch programming errors */
3148 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3149 	    ("%s: setting driver-owned flag %d", __func__, flag));
3150 
3151 	if (onswitch)
3152 		KASSERT(*refcount >= 0,
3153 		    ("%s: increment negative refcount %d for flag %d",
3154 		    __func__, *refcount, flag));
3155 	else
3156 		KASSERT(*refcount > 0,
3157 		    ("%s: decrement non-positive refcount %d for flag %d",
3158 		    __func__, *refcount, flag));
3159 
3160 	/* In case this mode is permanent, just touch refcount */
3161 	if (ifp->if_flags & pflag) {
3162 		*refcount += onswitch ? 1 : -1;
3163 		return (0);
3164 	}
3165 
3166 	/* Save ifnet parameters for if_ioctl() may fail */
3167 	oldcount = *refcount;
3168 	oldflags = ifp->if_flags;
3169 
3170 	/*
3171 	 * See if we aren't the only and touching refcount is enough.
3172 	 * Actually toggle interface flag if we are the first or last.
3173 	 */
3174 	if (onswitch) {
3175 		if ((*refcount)++)
3176 			return (0);
3177 		ifp->if_flags |= flag;
3178 	} else {
3179 		if (--(*refcount))
3180 			return (0);
3181 		ifp->if_flags &= ~flag;
3182 	}
3183 
3184 	/* Call down the driver since we've changed interface flags */
3185 	if (ifp->if_ioctl == NULL) {
3186 		error = EOPNOTSUPP;
3187 		goto recover;
3188 	}
3189 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3190 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3191 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3192 	if (error)
3193 		goto recover;
3194 	/* Notify userland that interface flags have changed */
3195 	rt_ifmsg(ifp);
3196 	return (0);
3197 
3198 recover:
3199 	/* Recover after driver error */
3200 	*refcount = oldcount;
3201 	ifp->if_flags = oldflags;
3202 	return (error);
3203 }
3204 
3205 /*
3206  * Set/clear promiscuous mode on interface ifp based on the truth value
3207  * of pswitch.  The calls are reference counted so that only the first
3208  * "on" request actually has an effect, as does the final "off" request.
3209  * Results are undefined if the "off" and "on" requests are not matched.
3210  */
3211 int
3212 ifpromisc(struct ifnet *ifp, int pswitch)
3213 {
3214 	int error;
3215 	int oldflags = ifp->if_flags;
3216 
3217 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3218 			   &ifp->if_pcount, pswitch);
3219 	/* If promiscuous mode status has changed, log a message */
3220 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3221             log_promisc_mode_change)
3222 		if_printf(ifp, "promiscuous mode %s\n",
3223 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3224 	return (error);
3225 }
3226 
3227 /*
3228  * Return interface configuration
3229  * of system.  List may be used
3230  * in later ioctl's (above) to get
3231  * other information.
3232  */
3233 /*ARGSUSED*/
3234 static int
3235 ifconf(u_long cmd, caddr_t data)
3236 {
3237 	struct ifconf *ifc = (struct ifconf *)data;
3238 	struct ifnet *ifp;
3239 	struct ifaddr *ifa;
3240 	struct ifreq ifr;
3241 	struct sbuf *sb;
3242 	int error, full = 0, valid_len, max_len;
3243 
3244 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3245 	max_len = MAXPHYS - 1;
3246 
3247 	/* Prevent hostile input from being able to crash the system */
3248 	if (ifc->ifc_len <= 0)
3249 		return (EINVAL);
3250 
3251 again:
3252 	if (ifc->ifc_len <= max_len) {
3253 		max_len = ifc->ifc_len;
3254 		full = 1;
3255 	}
3256 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3257 	max_len = 0;
3258 	valid_len = 0;
3259 
3260 	IFNET_RLOCK();
3261 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3262 		int addrs;
3263 
3264 		/*
3265 		 * Zero the ifr to make sure we don't disclose the contents
3266 		 * of the stack.
3267 		 */
3268 		memset(&ifr, 0, sizeof(ifr));
3269 
3270 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3271 		    >= sizeof(ifr.ifr_name)) {
3272 			sbuf_delete(sb);
3273 			IFNET_RUNLOCK();
3274 			return (ENAMETOOLONG);
3275 		}
3276 
3277 		addrs = 0;
3278 		IF_ADDR_RLOCK(ifp);
3279 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3280 			struct sockaddr *sa = ifa->ifa_addr;
3281 
3282 			if (prison_if(curthread->td_ucred, sa) != 0)
3283 				continue;
3284 			addrs++;
3285 			if (sa->sa_len <= sizeof(*sa)) {
3286 				if (sa->sa_len < sizeof(*sa)) {
3287 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3288 					    sizeof(ifr.ifr_ifru.ifru_addr));
3289 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3290 					    sa->sa_len);
3291 				} else
3292 					ifr.ifr_ifru.ifru_addr = *sa;
3293 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3294 				max_len += sizeof(ifr);
3295 			} else {
3296 				sbuf_bcat(sb, &ifr,
3297 				    offsetof(struct ifreq, ifr_addr));
3298 				max_len += offsetof(struct ifreq, ifr_addr);
3299 				sbuf_bcat(sb, sa, sa->sa_len);
3300 				max_len += sa->sa_len;
3301 			}
3302 
3303 			if (sbuf_error(sb) == 0)
3304 				valid_len = sbuf_len(sb);
3305 		}
3306 		IF_ADDR_RUNLOCK(ifp);
3307 		if (addrs == 0) {
3308 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3309 			max_len += sizeof(ifr);
3310 
3311 			if (sbuf_error(sb) == 0)
3312 				valid_len = sbuf_len(sb);
3313 		}
3314 	}
3315 	IFNET_RUNLOCK();
3316 
3317 	/*
3318 	 * If we didn't allocate enough space (uncommon), try again.  If
3319 	 * we have already allocated as much space as we are allowed,
3320 	 * return what we've got.
3321 	 */
3322 	if (valid_len != max_len && !full) {
3323 		sbuf_delete(sb);
3324 		goto again;
3325 	}
3326 
3327 	ifc->ifc_len = valid_len;
3328 	sbuf_finish(sb);
3329 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3330 	sbuf_delete(sb);
3331 	return (error);
3332 }
3333 
3334 /*
3335  * Just like ifpromisc(), but for all-multicast-reception mode.
3336  */
3337 int
3338 if_allmulti(struct ifnet *ifp, int onswitch)
3339 {
3340 
3341 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3342 }
3343 
3344 struct ifmultiaddr *
3345 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3346 {
3347 	struct ifmultiaddr *ifma;
3348 
3349 	IF_ADDR_LOCK_ASSERT(ifp);
3350 
3351 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3352 		if (sa->sa_family == AF_LINK) {
3353 			if (sa_dl_equal(ifma->ifma_addr, sa))
3354 				break;
3355 		} else {
3356 			if (sa_equal(ifma->ifma_addr, sa))
3357 				break;
3358 		}
3359 	}
3360 
3361 	return ifma;
3362 }
3363 
3364 /*
3365  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3366  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3367  * the ifnet multicast address list here, so the caller must do that and
3368  * other setup work (such as notifying the device driver).  The reference
3369  * count is initialized to 1.
3370  */
3371 static struct ifmultiaddr *
3372 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3373     int mflags)
3374 {
3375 	struct ifmultiaddr *ifma;
3376 	struct sockaddr *dupsa;
3377 
3378 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3379 	    M_ZERO);
3380 	if (ifma == NULL)
3381 		return (NULL);
3382 
3383 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3384 	if (dupsa == NULL) {
3385 		free(ifma, M_IFMADDR);
3386 		return (NULL);
3387 	}
3388 	bcopy(sa, dupsa, sa->sa_len);
3389 	ifma->ifma_addr = dupsa;
3390 
3391 	ifma->ifma_ifp = ifp;
3392 	ifma->ifma_refcount = 1;
3393 	ifma->ifma_protospec = NULL;
3394 
3395 	if (llsa == NULL) {
3396 		ifma->ifma_lladdr = NULL;
3397 		return (ifma);
3398 	}
3399 
3400 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3401 	if (dupsa == NULL) {
3402 		free(ifma->ifma_addr, M_IFMADDR);
3403 		free(ifma, M_IFMADDR);
3404 		return (NULL);
3405 	}
3406 	bcopy(llsa, dupsa, llsa->sa_len);
3407 	ifma->ifma_lladdr = dupsa;
3408 
3409 	return (ifma);
3410 }
3411 
3412 /*
3413  * if_freemulti: free ifmultiaddr structure and possibly attached related
3414  * addresses.  The caller is responsible for implementing reference
3415  * counting, notifying the driver, handling routing messages, and releasing
3416  * any dependent link layer state.
3417  */
3418 #ifdef MCAST_VERBOSE
3419 extern void kdb_backtrace(void);
3420 #endif
3421 static void
3422 if_freemulti_internal(struct ifmultiaddr *ifma)
3423 {
3424 
3425 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3426 	    ifma->ifma_refcount));
3427 
3428 	if (ifma->ifma_lladdr != NULL)
3429 		free(ifma->ifma_lladdr, M_IFMADDR);
3430 #ifdef MCAST_VERBOSE
3431 	kdb_backtrace();
3432 	printf("%s freeing ifma: %p\n", __func__, ifma);
3433 #endif
3434 	free(ifma->ifma_addr, M_IFMADDR);
3435 	free(ifma, M_IFMADDR);
3436 }
3437 
3438 static void
3439 if_destroymulti(epoch_context_t ctx)
3440 {
3441 	struct ifmultiaddr *ifma;
3442 
3443 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3444 	if_freemulti_internal(ifma);
3445 }
3446 
3447 void
3448 if_freemulti(struct ifmultiaddr *ifma)
3449 {
3450 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3451 	    ifma->ifma_refcount));
3452 
3453 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3454 }
3455 
3456 
3457 /*
3458  * Register an additional multicast address with a network interface.
3459  *
3460  * - If the address is already present, bump the reference count on the
3461  *   address and return.
3462  * - If the address is not link-layer, look up a link layer address.
3463  * - Allocate address structures for one or both addresses, and attach to the
3464  *   multicast address list on the interface.  If automatically adding a link
3465  *   layer address, the protocol address will own a reference to the link
3466  *   layer address, to be freed when it is freed.
3467  * - Notify the network device driver of an addition to the multicast address
3468  *   list.
3469  *
3470  * 'sa' points to caller-owned memory with the desired multicast address.
3471  *
3472  * 'retifma' will be used to return a pointer to the resulting multicast
3473  * address reference, if desired.
3474  */
3475 int
3476 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3477     struct ifmultiaddr **retifma)
3478 {
3479 	struct ifmultiaddr *ifma, *ll_ifma;
3480 	struct sockaddr *llsa;
3481 	struct sockaddr_dl sdl;
3482 	int error;
3483 
3484 #ifdef INET
3485 	IN_MULTI_LIST_UNLOCK_ASSERT();
3486 #endif
3487 #ifdef INET6
3488 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3489 #endif
3490 	/*
3491 	 * If the address is already present, return a new reference to it;
3492 	 * otherwise, allocate storage and set up a new address.
3493 	 */
3494 	IF_ADDR_WLOCK(ifp);
3495 	ifma = if_findmulti(ifp, sa);
3496 	if (ifma != NULL) {
3497 		ifma->ifma_refcount++;
3498 		if (retifma != NULL)
3499 			*retifma = ifma;
3500 		IF_ADDR_WUNLOCK(ifp);
3501 		return (0);
3502 	}
3503 
3504 	/*
3505 	 * The address isn't already present; resolve the protocol address
3506 	 * into a link layer address, and then look that up, bump its
3507 	 * refcount or allocate an ifma for that also.
3508 	 * Most link layer resolving functions returns address data which
3509 	 * fits inside default sockaddr_dl structure. However callback
3510 	 * can allocate another sockaddr structure, in that case we need to
3511 	 * free it later.
3512 	 */
3513 	llsa = NULL;
3514 	ll_ifma = NULL;
3515 	if (ifp->if_resolvemulti != NULL) {
3516 		/* Provide called function with buffer size information */
3517 		sdl.sdl_len = sizeof(sdl);
3518 		llsa = (struct sockaddr *)&sdl;
3519 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3520 		if (error)
3521 			goto unlock_out;
3522 	}
3523 
3524 	/*
3525 	 * Allocate the new address.  Don't hook it up yet, as we may also
3526 	 * need to allocate a link layer multicast address.
3527 	 */
3528 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3529 	if (ifma == NULL) {
3530 		error = ENOMEM;
3531 		goto free_llsa_out;
3532 	}
3533 
3534 	/*
3535 	 * If a link layer address is found, we'll need to see if it's
3536 	 * already present in the address list, or allocate is as well.
3537 	 * When this block finishes, the link layer address will be on the
3538 	 * list.
3539 	 */
3540 	if (llsa != NULL) {
3541 		ll_ifma = if_findmulti(ifp, llsa);
3542 		if (ll_ifma == NULL) {
3543 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3544 			if (ll_ifma == NULL) {
3545 				--ifma->ifma_refcount;
3546 				if_freemulti(ifma);
3547 				error = ENOMEM;
3548 				goto free_llsa_out;
3549 			}
3550 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3551 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3552 			    ifma_link);
3553 		} else
3554 			ll_ifma->ifma_refcount++;
3555 		ifma->ifma_llifma = ll_ifma;
3556 	}
3557 
3558 	/*
3559 	 * We now have a new multicast address, ifma, and possibly a new or
3560 	 * referenced link layer address.  Add the primary address to the
3561 	 * ifnet address list.
3562 	 */
3563 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3564 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3565 
3566 	if (retifma != NULL)
3567 		*retifma = ifma;
3568 
3569 	/*
3570 	 * Must generate the message while holding the lock so that 'ifma'
3571 	 * pointer is still valid.
3572 	 */
3573 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3574 	IF_ADDR_WUNLOCK(ifp);
3575 
3576 	/*
3577 	 * We are certain we have added something, so call down to the
3578 	 * interface to let them know about it.
3579 	 */
3580 	if (ifp->if_ioctl != NULL) {
3581 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3582 	}
3583 
3584 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3585 		link_free_sdl(llsa);
3586 
3587 	return (0);
3588 
3589 free_llsa_out:
3590 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3591 		link_free_sdl(llsa);
3592 
3593 unlock_out:
3594 	IF_ADDR_WUNLOCK(ifp);
3595 	return (error);
3596 }
3597 
3598 /*
3599  * Delete a multicast group membership by network-layer group address.
3600  *
3601  * Returns ENOENT if the entry could not be found. If ifp no longer
3602  * exists, results are undefined. This entry point should only be used
3603  * from subsystems which do appropriate locking to hold ifp for the
3604  * duration of the call.
3605  * Network-layer protocol domains must use if_delmulti_ifma().
3606  */
3607 int
3608 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3609 {
3610 	struct ifmultiaddr *ifma;
3611 	int lastref;
3612 #ifdef INVARIANTS
3613 	struct ifnet *oifp;
3614 
3615 	IFNET_RLOCK_NOSLEEP();
3616 	CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3617 		if (ifp == oifp)
3618 			break;
3619 	if (ifp != oifp)
3620 		ifp = NULL;
3621 	IFNET_RUNLOCK_NOSLEEP();
3622 
3623 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3624 #endif
3625 	if (ifp == NULL)
3626 		return (ENOENT);
3627 
3628 	IF_ADDR_WLOCK(ifp);
3629 	lastref = 0;
3630 	ifma = if_findmulti(ifp, sa);
3631 	if (ifma != NULL)
3632 		lastref = if_delmulti_locked(ifp, ifma, 0);
3633 	IF_ADDR_WUNLOCK(ifp);
3634 
3635 	if (ifma == NULL)
3636 		return (ENOENT);
3637 
3638 	if (lastref && ifp->if_ioctl != NULL) {
3639 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3640 	}
3641 
3642 	return (0);
3643 }
3644 
3645 /*
3646  * Delete all multicast group membership for an interface.
3647  * Should be used to quickly flush all multicast filters.
3648  */
3649 void
3650 if_delallmulti(struct ifnet *ifp)
3651 {
3652 	struct ifmultiaddr *ifma;
3653 	struct ifmultiaddr *next;
3654 
3655 	IF_ADDR_WLOCK(ifp);
3656 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3657 		if_delmulti_locked(ifp, ifma, 0);
3658 	IF_ADDR_WUNLOCK(ifp);
3659 }
3660 
3661 void
3662 if_delmulti_ifma(struct ifmultiaddr *ifma)
3663 {
3664 	if_delmulti_ifma_flags(ifma, 0);
3665 }
3666 
3667 /*
3668  * Delete a multicast group membership by group membership pointer.
3669  * Network-layer protocol domains must use this routine.
3670  *
3671  * It is safe to call this routine if the ifp disappeared.
3672  */
3673 void
3674 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3675 {
3676 	struct ifnet *ifp;
3677 	int lastref;
3678 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3679 #ifdef INET
3680 	IN_MULTI_LIST_UNLOCK_ASSERT();
3681 #endif
3682 	ifp = ifma->ifma_ifp;
3683 #ifdef DIAGNOSTIC
3684 	if (ifp == NULL) {
3685 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3686 	} else {
3687 		struct ifnet *oifp;
3688 
3689 		IFNET_RLOCK_NOSLEEP();
3690 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3691 			if (ifp == oifp)
3692 				break;
3693 		if (ifp != oifp)
3694 			ifp = NULL;
3695 		IFNET_RUNLOCK_NOSLEEP();
3696 	}
3697 #endif
3698 	/*
3699 	 * If and only if the ifnet instance exists: Acquire the address lock.
3700 	 */
3701 	if (ifp != NULL)
3702 		IF_ADDR_WLOCK(ifp);
3703 
3704 	lastref = if_delmulti_locked(ifp, ifma, flags);
3705 
3706 	if (ifp != NULL) {
3707 		/*
3708 		 * If and only if the ifnet instance exists:
3709 		 *  Release the address lock.
3710 		 *  If the group was left: update the hardware hash filter.
3711 		 */
3712 		IF_ADDR_WUNLOCK(ifp);
3713 		if (lastref && ifp->if_ioctl != NULL) {
3714 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3715 		}
3716 	}
3717 }
3718 
3719 /*
3720  * Perform deletion of network-layer and/or link-layer multicast address.
3721  *
3722  * Return 0 if the reference count was decremented.
3723  * Return 1 if the final reference was released, indicating that the
3724  * hardware hash filter should be reprogrammed.
3725  */
3726 static int
3727 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3728 {
3729 	struct ifmultiaddr *ll_ifma;
3730 
3731 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3732 		KASSERT(ifma->ifma_ifp == ifp,
3733 		    ("%s: inconsistent ifp %p", __func__, ifp));
3734 		IF_ADDR_WLOCK_ASSERT(ifp);
3735 	}
3736 
3737 	ifp = ifma->ifma_ifp;
3738 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3739 
3740 	/*
3741 	 * If the ifnet is detaching, null out references to ifnet,
3742 	 * so that upper protocol layers will notice, and not attempt
3743 	 * to obtain locks for an ifnet which no longer exists. The
3744 	 * routing socket announcement must happen before the ifnet
3745 	 * instance is detached from the system.
3746 	 */
3747 	if (detaching) {
3748 #ifdef DIAGNOSTIC
3749 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3750 #endif
3751 		/*
3752 		 * ifp may already be nulled out if we are being reentered
3753 		 * to delete the ll_ifma.
3754 		 */
3755 		if (ifp != NULL) {
3756 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3757 			ifma->ifma_ifp = NULL;
3758 		}
3759 	}
3760 
3761 	if (--ifma->ifma_refcount > 0)
3762 		return 0;
3763 
3764 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3765 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3766 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3767 	}
3768 	/*
3769 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3770 	 * have been associated with it. Release it first if so.
3771 	 */
3772 	ll_ifma = ifma->ifma_llifma;
3773 	if (ll_ifma != NULL) {
3774 		KASSERT(ifma->ifma_lladdr != NULL,
3775 		    ("%s: llifma w/o lladdr", __func__));
3776 		if (detaching)
3777 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3778 		if (--ll_ifma->ifma_refcount == 0) {
3779 			if (ifp != NULL) {
3780 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3781 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3782 						ifma_link);
3783 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3784 				}
3785 			}
3786 			if_freemulti(ll_ifma);
3787 		}
3788 	}
3789 #ifdef INVARIANTS
3790 	if (ifp) {
3791 		struct ifmultiaddr *ifmatmp;
3792 
3793 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3794 			MPASS(ifma != ifmatmp);
3795 	}
3796 #endif
3797 	if_freemulti(ifma);
3798 	/*
3799 	 * The last reference to this instance of struct ifmultiaddr
3800 	 * was released; the hardware should be notified of this change.
3801 	 */
3802 	return 1;
3803 }
3804 
3805 /*
3806  * Set the link layer address on an interface.
3807  *
3808  * At this time we only support certain types of interfaces,
3809  * and we don't allow the length of the address to change.
3810  *
3811  * Set noinline to be dtrace-friendly
3812  */
3813 __noinline int
3814 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3815 {
3816 	struct sockaddr_dl *sdl;
3817 	struct ifaddr *ifa;
3818 	struct ifreq ifr;
3819 	int rc;
3820 
3821 	rc = 0;
3822 	NET_EPOCH_ENTER();
3823 	ifa = ifp->if_addr;
3824 	if (ifa == NULL) {
3825 		rc = EINVAL;
3826 		goto out;
3827 	}
3828 
3829 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3830 	if (sdl == NULL) {
3831 		rc = EINVAL;
3832 		goto out;
3833 	}
3834 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3835 		rc = EINVAL;
3836 		goto out;
3837 	}
3838 	switch (ifp->if_type) {
3839 	case IFT_ETHER:
3840 	case IFT_XETHER:
3841 	case IFT_L2VLAN:
3842 	case IFT_BRIDGE:
3843 	case IFT_IEEE8023ADLAG:
3844 		bcopy(lladdr, LLADDR(sdl), len);
3845 		break;
3846 	default:
3847 		rc = ENODEV;
3848 		goto out;
3849 	}
3850 
3851 	/*
3852 	 * If the interface is already up, we need
3853 	 * to re-init it in order to reprogram its
3854 	 * address filter.
3855 	 */
3856 	NET_EPOCH_EXIT();
3857 	if ((ifp->if_flags & IFF_UP) != 0) {
3858 		if (ifp->if_ioctl) {
3859 			ifp->if_flags &= ~IFF_UP;
3860 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3861 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3862 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3863 			ifp->if_flags |= IFF_UP;
3864 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3865 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3866 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3867 		}
3868 	}
3869 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3870 	return (0);
3871  out:
3872 	NET_EPOCH_EXIT();
3873 	return (rc);
3874 }
3875 
3876 /*
3877  * Compat function for handling basic encapsulation requests.
3878  * Not converted stacks (FDDI, IB, ..) supports traditional
3879  * output model: ARP (and other similar L2 protocols) are handled
3880  * inside output routine, arpresolve/nd6_resolve() returns MAC
3881  * address instead of full prepend.
3882  *
3883  * This function creates calculated header==MAC for IPv4/IPv6 and
3884  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3885  * address families.
3886  */
3887 static int
3888 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3889 {
3890 
3891 	if (req->rtype != IFENCAP_LL)
3892 		return (EOPNOTSUPP);
3893 
3894 	if (req->bufsize < req->lladdr_len)
3895 		return (ENOMEM);
3896 
3897 	switch (req->family) {
3898 	case AF_INET:
3899 	case AF_INET6:
3900 		break;
3901 	default:
3902 		return (EAFNOSUPPORT);
3903 	}
3904 
3905 	/* Copy lladdr to storage as is */
3906 	memmove(req->buf, req->lladdr, req->lladdr_len);
3907 	req->bufsize = req->lladdr_len;
3908 	req->lladdr_off = 0;
3909 
3910 	return (0);
3911 }
3912 
3913 /*
3914  * Tunnel interfaces can nest, also they may cause infinite recursion
3915  * calls when misconfigured. We'll prevent this by detecting loops.
3916  * High nesting level may cause stack exhaustion. We'll prevent this
3917  * by introducing upper limit.
3918  *
3919  * Return 0, if tunnel nesting count is equal or less than limit.
3920  */
3921 int
3922 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3923     int limit)
3924 {
3925 	struct m_tag *mtag;
3926 	int count;
3927 
3928 	count = 1;
3929 	mtag = NULL;
3930 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3931 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3932 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3933 			return (EIO);
3934 		}
3935 		count++;
3936 	}
3937 	if (count > limit) {
3938 		log(LOG_NOTICE,
3939 		    "%s: if_output recursively called too many times(%d)\n",
3940 		    if_name(ifp), count);
3941 		return (EIO);
3942 	}
3943 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3944 	if (mtag == NULL)
3945 		return (ENOMEM);
3946 	*(struct ifnet **)(mtag + 1) = ifp;
3947 	m_tag_prepend(m, mtag);
3948 	return (0);
3949 }
3950 
3951 /*
3952  * Get the link layer address that was read from the hardware at attach.
3953  *
3954  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3955  * their component interfaces as IFT_IEEE8023ADLAG.
3956  */
3957 int
3958 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3959 {
3960 
3961 	if (ifp->if_hw_addr == NULL)
3962 		return (ENODEV);
3963 
3964 	switch (ifp->if_type) {
3965 	case IFT_ETHER:
3966 	case IFT_IEEE8023ADLAG:
3967 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3968 		return (0);
3969 	default:
3970 		return (ENODEV);
3971 	}
3972 }
3973 
3974 /*
3975  * The name argument must be a pointer to storage which will last as
3976  * long as the interface does.  For physical devices, the result of
3977  * device_get_name(dev) is a good choice and for pseudo-devices a
3978  * static string works well.
3979  */
3980 void
3981 if_initname(struct ifnet *ifp, const char *name, int unit)
3982 {
3983 	ifp->if_dname = name;
3984 	ifp->if_dunit = unit;
3985 	if (unit != IF_DUNIT_NONE)
3986 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3987 	else
3988 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3989 }
3990 
3991 int
3992 if_printf(struct ifnet *ifp, const char *fmt, ...)
3993 {
3994 	char if_fmt[256];
3995 	va_list ap;
3996 
3997 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3998 	va_start(ap, fmt);
3999 	vlog(LOG_INFO, if_fmt, ap);
4000 	va_end(ap);
4001 	return (0);
4002 }
4003 
4004 void
4005 if_start(struct ifnet *ifp)
4006 {
4007 
4008 	(*(ifp)->if_start)(ifp);
4009 }
4010 
4011 /*
4012  * Backwards compatibility interface for drivers
4013  * that have not implemented it
4014  */
4015 static int
4016 if_transmit(struct ifnet *ifp, struct mbuf *m)
4017 {
4018 	int error;
4019 
4020 	IFQ_HANDOFF(ifp, m, error);
4021 	return (error);
4022 }
4023 
4024 static void
4025 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4026 {
4027 
4028 	m_freem(m);
4029 }
4030 
4031 int
4032 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4033 {
4034 	int active = 0;
4035 
4036 	IF_LOCK(ifq);
4037 	if (_IF_QFULL(ifq)) {
4038 		IF_UNLOCK(ifq);
4039 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4040 		m_freem(m);
4041 		return (0);
4042 	}
4043 	if (ifp != NULL) {
4044 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4045 		if (m->m_flags & (M_BCAST|M_MCAST))
4046 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4047 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4048 	}
4049 	_IF_ENQUEUE(ifq, m);
4050 	IF_UNLOCK(ifq);
4051 	if (ifp != NULL && !active)
4052 		(*(ifp)->if_start)(ifp);
4053 	return (1);
4054 }
4055 
4056 void
4057 if_register_com_alloc(u_char type,
4058     if_com_alloc_t *a, if_com_free_t *f)
4059 {
4060 
4061 	KASSERT(if_com_alloc[type] == NULL,
4062 	    ("if_register_com_alloc: %d already registered", type));
4063 	KASSERT(if_com_free[type] == NULL,
4064 	    ("if_register_com_alloc: %d free already registered", type));
4065 
4066 	if_com_alloc[type] = a;
4067 	if_com_free[type] = f;
4068 }
4069 
4070 void
4071 if_deregister_com_alloc(u_char type)
4072 {
4073 
4074 	KASSERT(if_com_alloc[type] != NULL,
4075 	    ("if_deregister_com_alloc: %d not registered", type));
4076 	KASSERT(if_com_free[type] != NULL,
4077 	    ("if_deregister_com_alloc: %d free not registered", type));
4078 	if_com_alloc[type] = NULL;
4079 	if_com_free[type] = NULL;
4080 }
4081 
4082 /* API for driver access to network stack owned ifnet.*/
4083 uint64_t
4084 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4085 {
4086 	uint64_t oldbrate;
4087 
4088 	oldbrate = ifp->if_baudrate;
4089 	ifp->if_baudrate = baudrate;
4090 	return (oldbrate);
4091 }
4092 
4093 uint64_t
4094 if_getbaudrate(if_t ifp)
4095 {
4096 
4097 	return (((struct ifnet *)ifp)->if_baudrate);
4098 }
4099 
4100 int
4101 if_setcapabilities(if_t ifp, int capabilities)
4102 {
4103 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4104 	return (0);
4105 }
4106 
4107 int
4108 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4109 {
4110 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4111 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4112 
4113 	return (0);
4114 }
4115 
4116 int
4117 if_getcapabilities(if_t ifp)
4118 {
4119 	return ((struct ifnet *)ifp)->if_capabilities;
4120 }
4121 
4122 int
4123 if_setcapenable(if_t ifp, int capabilities)
4124 {
4125 	((struct ifnet *)ifp)->if_capenable = capabilities;
4126 	return (0);
4127 }
4128 
4129 int
4130 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4131 {
4132 	if(setcap)
4133 		((struct ifnet *)ifp)->if_capenable |= setcap;
4134 	if(clearcap)
4135 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4136 
4137 	return (0);
4138 }
4139 
4140 const char *
4141 if_getdname(if_t ifp)
4142 {
4143 	return ((struct ifnet *)ifp)->if_dname;
4144 }
4145 
4146 int
4147 if_togglecapenable(if_t ifp, int togglecap)
4148 {
4149 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4150 	return (0);
4151 }
4152 
4153 int
4154 if_getcapenable(if_t ifp)
4155 {
4156 	return ((struct ifnet *)ifp)->if_capenable;
4157 }
4158 
4159 /*
4160  * This is largely undesirable because it ties ifnet to a device, but does
4161  * provide flexiblity for an embedded product vendor. Should be used with
4162  * the understanding that it violates the interface boundaries, and should be
4163  * a last resort only.
4164  */
4165 int
4166 if_setdev(if_t ifp, void *dev)
4167 {
4168 	return (0);
4169 }
4170 
4171 int
4172 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4173 {
4174 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4175 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4176 
4177 	return (0);
4178 }
4179 
4180 int
4181 if_getdrvflags(if_t ifp)
4182 {
4183 	return ((struct ifnet *)ifp)->if_drv_flags;
4184 }
4185 
4186 int
4187 if_setdrvflags(if_t ifp, int flags)
4188 {
4189 	((struct ifnet *)ifp)->if_drv_flags = flags;
4190 	return (0);
4191 }
4192 
4193 
4194 int
4195 if_setflags(if_t ifp, int flags)
4196 {
4197 	((struct ifnet *)ifp)->if_flags = flags;
4198 	return (0);
4199 }
4200 
4201 int
4202 if_setflagbits(if_t ifp, int set, int clear)
4203 {
4204 	((struct ifnet *)ifp)->if_flags |= set;
4205 	((struct ifnet *)ifp)->if_flags &= ~clear;
4206 
4207 	return (0);
4208 }
4209 
4210 int
4211 if_getflags(if_t ifp)
4212 {
4213 	return ((struct ifnet *)ifp)->if_flags;
4214 }
4215 
4216 int
4217 if_clearhwassist(if_t ifp)
4218 {
4219 	((struct ifnet *)ifp)->if_hwassist = 0;
4220 	return (0);
4221 }
4222 
4223 int
4224 if_sethwassistbits(if_t ifp, int toset, int toclear)
4225 {
4226 	((struct ifnet *)ifp)->if_hwassist |= toset;
4227 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4228 
4229 	return (0);
4230 }
4231 
4232 int
4233 if_sethwassist(if_t ifp, int hwassist_bit)
4234 {
4235 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4236 	return (0);
4237 }
4238 
4239 int
4240 if_gethwassist(if_t ifp)
4241 {
4242 	return ((struct ifnet *)ifp)->if_hwassist;
4243 }
4244 
4245 int
4246 if_setmtu(if_t ifp, int mtu)
4247 {
4248 	((struct ifnet *)ifp)->if_mtu = mtu;
4249 	return (0);
4250 }
4251 
4252 int
4253 if_getmtu(if_t ifp)
4254 {
4255 	return ((struct ifnet *)ifp)->if_mtu;
4256 }
4257 
4258 int
4259 if_getmtu_family(if_t ifp, int family)
4260 {
4261 	struct domain *dp;
4262 
4263 	for (dp = domains; dp; dp = dp->dom_next) {
4264 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4265 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4266 	}
4267 
4268 	return (((struct ifnet *)ifp)->if_mtu);
4269 }
4270 
4271 int
4272 if_setsoftc(if_t ifp, void *softc)
4273 {
4274 	((struct ifnet *)ifp)->if_softc = softc;
4275 	return (0);
4276 }
4277 
4278 void *
4279 if_getsoftc(if_t ifp)
4280 {
4281 	return ((struct ifnet *)ifp)->if_softc;
4282 }
4283 
4284 void
4285 if_setrcvif(struct mbuf *m, if_t ifp)
4286 {
4287 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4288 }
4289 
4290 void
4291 if_setvtag(struct mbuf *m, uint16_t tag)
4292 {
4293 	m->m_pkthdr.ether_vtag = tag;
4294 }
4295 
4296 uint16_t
4297 if_getvtag(struct mbuf *m)
4298 {
4299 
4300 	return (m->m_pkthdr.ether_vtag);
4301 }
4302 
4303 int
4304 if_sendq_empty(if_t ifp)
4305 {
4306 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4307 }
4308 
4309 struct ifaddr *
4310 if_getifaddr(if_t ifp)
4311 {
4312 	return ((struct ifnet *)ifp)->if_addr;
4313 }
4314 
4315 int
4316 if_getamcount(if_t ifp)
4317 {
4318 	return ((struct ifnet *)ifp)->if_amcount;
4319 }
4320 
4321 
4322 int
4323 if_setsendqready(if_t ifp)
4324 {
4325 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4326 	return (0);
4327 }
4328 
4329 int
4330 if_setsendqlen(if_t ifp, int tx_desc_count)
4331 {
4332 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4333 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4334 
4335 	return (0);
4336 }
4337 
4338 int
4339 if_vlantrunkinuse(if_t ifp)
4340 {
4341 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4342 }
4343 
4344 int
4345 if_input(if_t ifp, struct mbuf* sendmp)
4346 {
4347 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4348 	return (0);
4349 
4350 }
4351 
4352 /* XXX */
4353 #ifndef ETH_ADDR_LEN
4354 #define ETH_ADDR_LEN 6
4355 #endif
4356 
4357 int
4358 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4359 {
4360 	struct ifmultiaddr *ifma;
4361 	uint8_t *lmta = (uint8_t *)mta;
4362 	int mcnt = 0;
4363 
4364 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4365 		if (ifma->ifma_addr->sa_family != AF_LINK)
4366 			continue;
4367 
4368 		if (mcnt == max)
4369 			break;
4370 
4371 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4372 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4373 		mcnt++;
4374 	}
4375 	*cnt = mcnt;
4376 
4377 	return (0);
4378 }
4379 
4380 int
4381 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4382 {
4383 	int error;
4384 
4385 	if_maddr_rlock(ifp);
4386 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4387 	if_maddr_runlock(ifp);
4388 	return (error);
4389 }
4390 
4391 int
4392 if_multiaddr_count(if_t ifp, int max)
4393 {
4394 	struct ifmultiaddr *ifma;
4395 	int count;
4396 
4397 	count = 0;
4398 	if_maddr_rlock(ifp);
4399 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4400 		if (ifma->ifma_addr->sa_family != AF_LINK)
4401 			continue;
4402 		count++;
4403 		if (count == max)
4404 			break;
4405 	}
4406 	if_maddr_runlock(ifp);
4407 	return (count);
4408 }
4409 
4410 int
4411 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4412 {
4413 	struct ifmultiaddr *ifma;
4414 	int cnt = 0;
4415 
4416 	if_maddr_rlock(ifp);
4417 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4418 		cnt += filter(arg, ifma, cnt);
4419 	if_maddr_runlock(ifp);
4420 	return (cnt);
4421 }
4422 
4423 struct mbuf *
4424 if_dequeue(if_t ifp)
4425 {
4426 	struct mbuf *m;
4427 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4428 
4429 	return (m);
4430 }
4431 
4432 int
4433 if_sendq_prepend(if_t ifp, struct mbuf *m)
4434 {
4435 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4436 	return (0);
4437 }
4438 
4439 int
4440 if_setifheaderlen(if_t ifp, int len)
4441 {
4442 	((struct ifnet *)ifp)->if_hdrlen = len;
4443 	return (0);
4444 }
4445 
4446 caddr_t
4447 if_getlladdr(if_t ifp)
4448 {
4449 	return (IF_LLADDR((struct ifnet *)ifp));
4450 }
4451 
4452 void *
4453 if_gethandle(u_char type)
4454 {
4455 	return (if_alloc(type));
4456 }
4457 
4458 void
4459 if_bpfmtap(if_t ifh, struct mbuf *m)
4460 {
4461 	struct ifnet *ifp = (struct ifnet *)ifh;
4462 
4463 	BPF_MTAP(ifp, m);
4464 }
4465 
4466 void
4467 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4468 {
4469 	struct ifnet *ifp = (struct ifnet *)ifh;
4470 
4471 	ETHER_BPF_MTAP(ifp, m);
4472 }
4473 
4474 void
4475 if_vlancap(if_t ifh)
4476 {
4477 	struct ifnet *ifp = (struct ifnet *)ifh;
4478 	VLAN_CAPABILITIES(ifp);
4479 }
4480 
4481 int
4482 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4483 {
4484 
4485 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4486         return (0);
4487 }
4488 
4489 int
4490 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4491 {
4492 
4493 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4494         return (0);
4495 }
4496 
4497 int
4498 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4499 {
4500 
4501 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4502         return (0);
4503 }
4504 
4505 u_int
4506 if_gethwtsomax(if_t ifp)
4507 {
4508 
4509 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4510 }
4511 
4512 u_int
4513 if_gethwtsomaxsegcount(if_t ifp)
4514 {
4515 
4516 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4517 }
4518 
4519 u_int
4520 if_gethwtsomaxsegsize(if_t ifp)
4521 {
4522 
4523 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4524 }
4525 
4526 void
4527 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4528 {
4529 	((struct ifnet *)ifp)->if_init = init_fn;
4530 }
4531 
4532 void
4533 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4534 {
4535 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4536 }
4537 
4538 void
4539 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4540 {
4541 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4542 }
4543 
4544 void
4545 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4546 {
4547 	((struct ifnet *)ifp)->if_transmit = start_fn;
4548 }
4549 
4550 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4551 {
4552 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4553 
4554 }
4555 
4556 void
4557 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4558 {
4559 
4560 	ifp->if_get_counter = fn;
4561 }
4562 
4563 /* Revisit these - These are inline functions originally. */
4564 int
4565 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4566 {
4567 	return drbr_inuse(ifh, br);
4568 }
4569 
4570 struct mbuf*
4571 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4572 {
4573 	return drbr_dequeue(ifh, br);
4574 }
4575 
4576 int
4577 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4578 {
4579 	return drbr_needs_enqueue(ifh, br);
4580 }
4581 
4582 int
4583 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4584 {
4585 	return drbr_enqueue(ifh, br, m);
4586 
4587 }
4588