xref: /freebsd/sys/net/if.c (revision cd8537910406e68d4719136a5b0cf6d23bb1b23b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_bpf.h"
36 #include "opt_inet6.h"
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/conf.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/domainset.h>
44 #include <sys/sbuf.h>
45 #include <sys/bus.h>
46 #include <sys/epoch.h>
47 #include <sys/mbuf.h>
48 #include <sys/systm.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/refcount.h>
57 #include <sys/module.h>
58 #include <sys/rwlock.h>
59 #include <sys/sockio.h>
60 #include <sys/syslog.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/taskqueue.h>
64 #include <sys/domain.h>
65 #include <sys/jail.h>
66 #include <sys/priv.h>
67 
68 #include <machine/stdarg.h>
69 #include <vm/uma.h>
70 
71 #include <net/bpf.h>
72 #include <net/ethernet.h>
73 #include <net/if.h>
74 #include <net/if_arp.h>
75 #include <net/if_clone.h>
76 #include <net/if_dl.h>
77 #include <net/if_types.h>
78 #include <net/if_var.h>
79 #include <net/if_media.h>
80 #include <net/if_vlan_var.h>
81 #include <net/radix.h>
82 #include <net/route.h>
83 #include <net/route/route_ctl.h>
84 #include <net/vnet.h>
85 
86 #if defined(INET) || defined(INET6)
87 #include <net/ethernet.h>
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip.h>
91 #include <netinet/ip_carp.h>
92 #ifdef INET
93 #include <net/debugnet.h>
94 #include <netinet/if_ether.h>
95 #endif /* INET */
96 #ifdef INET6
97 #include <netinet6/in6_var.h>
98 #include <netinet6/in6_ifattach.h>
99 #endif /* INET6 */
100 #endif /* INET || INET6 */
101 
102 #include <security/mac/mac_framework.h>
103 
104 /*
105  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
106  * and ifr_ifru when it is used in SIOCGIFCONF.
107  */
108 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
109     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
110 
111 __read_mostly epoch_t net_epoch_preempt;
112 #ifdef COMPAT_FREEBSD32
113 #include <sys/mount.h>
114 #include <compat/freebsd32/freebsd32.h>
115 
116 struct ifreq_buffer32 {
117 	uint32_t	length;		/* (size_t) */
118 	uint32_t	buffer;		/* (void *) */
119 };
120 
121 /*
122  * Interface request structure used for socket
123  * ioctl's.  All interface ioctl's must have parameter
124  * definitions which begin with ifr_name.  The
125  * remainder may be interface specific.
126  */
127 struct ifreq32 {
128 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
129 	union {
130 		struct sockaddr	ifru_addr;
131 		struct sockaddr	ifru_dstaddr;
132 		struct sockaddr	ifru_broadaddr;
133 		struct ifreq_buffer32 ifru_buffer;
134 		short		ifru_flags[2];
135 		short		ifru_index;
136 		int		ifru_jid;
137 		int		ifru_metric;
138 		int		ifru_mtu;
139 		int		ifru_phys;
140 		int		ifru_media;
141 		uint32_t	ifru_data;
142 		int		ifru_cap[2];
143 		u_int		ifru_fib;
144 		u_char		ifru_vlan_pcp;
145 	} ifr_ifru;
146 };
147 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
148 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
149     __offsetof(struct ifreq32, ifr_ifru));
150 
151 struct ifgroupreq32 {
152 	char	ifgr_name[IFNAMSIZ];
153 	u_int	ifgr_len;
154 	union {
155 		char		ifgru_group[IFNAMSIZ];
156 		uint32_t	ifgru_groups;
157 	} ifgr_ifgru;
158 };
159 
160 struct ifmediareq32 {
161 	char		ifm_name[IFNAMSIZ];
162 	int		ifm_current;
163 	int		ifm_mask;
164 	int		ifm_status;
165 	int		ifm_active;
166 	int		ifm_count;
167 	uint32_t	ifm_ulist;	/* (int *) */
168 };
169 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
170 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
171 
172 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
173     _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
174 #else /* !COMPAT_FREEBSD32 */
175 #define _CASE_IOC_IFGROUPREQ_32(cmd)
176 #endif /* !COMPAT_FREEBSD32 */
177 
178 #define CASE_IOC_IFGROUPREQ(cmd)	\
179     _CASE_IOC_IFGROUPREQ_32(cmd)	\
180     (cmd)
181 
182 union ifreq_union {
183 	struct ifreq	ifr;
184 #ifdef COMPAT_FREEBSD32
185 	struct ifreq32	ifr32;
186 #endif
187 };
188 
189 union ifgroupreq_union {
190 	struct ifgroupreq ifgr;
191 #ifdef COMPAT_FREEBSD32
192 	struct ifgroupreq32 ifgr32;
193 #endif
194 };
195 
196 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
197     "Link layers");
198 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
199     "Generic link-management");
200 
201 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
202     &ifqmaxlen, 0, "max send queue size");
203 
204 /* Log link state change events */
205 static int log_link_state_change = 1;
206 
207 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
208 	&log_link_state_change, 0,
209 	"log interface link state change events");
210 
211 /* Log promiscuous mode change events */
212 static int log_promisc_mode_change = 1;
213 
214 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
215 	&log_promisc_mode_change, 1,
216 	"log promiscuous mode change events");
217 
218 /* Interface description */
219 static unsigned int ifdescr_maxlen = 1024;
220 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
221 	&ifdescr_maxlen, 0,
222 	"administrative maximum length for interface description");
223 
224 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
225 
226 /* global sx for non-critical path ifdescr */
227 static struct sx ifdescr_sx;
228 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
229 
230 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
231 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
232 /* These are external hooks for CARP. */
233 void	(*carp_linkstate_p)(struct ifnet *ifp);
234 void	(*carp_demote_adj_p)(int, char *);
235 int	(*carp_master_p)(struct ifaddr *);
236 #if defined(INET) || defined(INET6)
237 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
238 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
239     const struct sockaddr *sa);
240 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
241 int	(*carp_attach_p)(struct ifaddr *, int);
242 void	(*carp_detach_p)(struct ifaddr *, bool);
243 #endif
244 #ifdef INET
245 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
246 #endif
247 #ifdef INET6
248 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
249 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
250     const struct in6_addr *taddr);
251 #endif
252 
253 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
254 
255 /*
256  * XXX: Style; these should be sorted alphabetically, and unprototyped
257  * static functions should be prototyped. Currently they are sorted by
258  * declaration order.
259  */
260 static void	if_attachdomain(void *);
261 static void	if_attachdomain1(struct ifnet *);
262 static int	ifconf(u_long, caddr_t);
263 static void	*if_grow(void);
264 static void	if_input_default(struct ifnet *, struct mbuf *);
265 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
266 static void	if_route(struct ifnet *, int flag, int fam);
267 static int	if_setflag(struct ifnet *, int, int, int *, int);
268 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
269 static void	if_unroute(struct ifnet *, int flag, int fam);
270 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
271 static void	do_link_state_change(void *, int);
272 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
273 static int	if_getgroupmembers(struct ifgroupreq *);
274 static void	if_delgroups(struct ifnet *);
275 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
276 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
277 static void	if_siocaddmulti(void *, int);
278 static void	if_link_ifnet(struct ifnet *);
279 static bool	if_unlink_ifnet(struct ifnet *, bool);
280 #ifdef VIMAGE
281 static int	if_vmove(struct ifnet *, struct vnet *);
282 #endif
283 
284 #ifdef INET6
285 /*
286  * XXX: declare here to avoid to include many inet6 related files..
287  * should be more generalized?
288  */
289 extern void	nd6_setmtu(struct ifnet *);
290 #endif
291 
292 /* ipsec helper hooks */
293 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
294 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
295 
296 VNET_DEFINE(int, if_index);
297 int	ifqmaxlen = IFQ_MAXLEN;
298 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
299 VNET_DEFINE(struct ifgrouphead, ifg_head);
300 
301 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
302 
303 /* Table of ifnet by index. */
304 VNET_DEFINE(struct ifnet **, ifindex_table);
305 
306 #define	V_if_indexlim		VNET(if_indexlim)
307 #define	V_ifindex_table		VNET(ifindex_table)
308 
309 /*
310  * The global network interface list (V_ifnet) and related state (such as
311  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
312  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
313  */
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 #ifdef VIMAGE
325 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
326     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
327 #endif
328 
329 static	if_com_alloc_t *if_com_alloc[256];
330 static	if_com_free_t *if_com_free[256];
331 
332 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
333 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
334 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
335 
336 struct ifnet *
337 ifnet_byindex(u_short idx)
338 {
339 	struct ifnet *ifp;
340 
341 	if (__predict_false(idx > V_if_index))
342 		return (NULL);
343 
344 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
345 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
346 }
347 
348 struct ifnet *
349 ifnet_byindex_ref(u_short idx)
350 {
351 	struct ifnet *ifp;
352 
353 	NET_EPOCH_ASSERT();
354 
355 	ifp = ifnet_byindex(idx);
356 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
357 		return (NULL);
358 	if_ref(ifp);
359 	return (ifp);
360 }
361 
362 /*
363  * Allocate an ifindex array entry; return 0 on success or an error on
364  * failure.
365  */
366 static u_short
367 ifindex_alloc(void **old)
368 {
369 	u_short idx;
370 
371 	IFNET_WLOCK_ASSERT();
372 	/*
373 	 * Try to find an empty slot below V_if_index.  If we fail, take the
374 	 * next slot.
375 	 */
376 	for (idx = 1; idx <= V_if_index; idx++) {
377 		if (V_ifindex_table[idx] == NULL)
378 			break;
379 	}
380 
381 	/* Catch if_index overflow. */
382 	if (idx >= V_if_indexlim) {
383 		*old = if_grow();
384 		return (USHRT_MAX);
385 	}
386 	if (idx > V_if_index)
387 		V_if_index = idx;
388 	return (idx);
389 }
390 
391 static void
392 ifindex_free_locked(u_short idx)
393 {
394 
395 	IFNET_WLOCK_ASSERT();
396 
397 	V_ifindex_table[idx] = NULL;
398 	while (V_if_index > 0 &&
399 	    V_ifindex_table[V_if_index] == NULL)
400 		V_if_index--;
401 }
402 
403 static void
404 ifindex_free(u_short idx)
405 {
406 
407 	IFNET_WLOCK();
408 	ifindex_free_locked(idx);
409 	IFNET_WUNLOCK();
410 }
411 
412 static void
413 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
414 {
415 
416 	V_ifindex_table[idx] = ifp;
417 }
418 
419 struct ifaddr *
420 ifaddr_byindex(u_short idx)
421 {
422 	struct ifnet *ifp;
423 	struct ifaddr *ifa = NULL;
424 
425 	NET_EPOCH_ASSERT();
426 
427 	ifp = ifnet_byindex(idx);
428 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
429 		ifa_ref(ifa);
430 	return (ifa);
431 }
432 
433 /*
434  * Network interface utility routines.
435  *
436  * Routines with ifa_ifwith* names take sockaddr *'s as
437  * parameters.
438  */
439 
440 static void
441 vnet_if_init(const void *unused __unused)
442 {
443 	void *old;
444 
445 	CK_STAILQ_INIT(&V_ifnet);
446 	CK_STAILQ_INIT(&V_ifg_head);
447 	IFNET_WLOCK();
448 	old = if_grow();				/* create initial table */
449 	IFNET_WUNLOCK();
450 	epoch_wait_preempt(net_epoch_preempt);
451 	free(old, M_IFNET);
452 	vnet_if_clone_init();
453 }
454 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
455     NULL);
456 
457 #ifdef VIMAGE
458 static void
459 vnet_if_uninit(const void *unused __unused)
460 {
461 
462 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
463 	    "not empty", __func__, __LINE__, &V_ifnet));
464 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
465 	    "not empty", __func__, __LINE__, &V_ifg_head));
466 
467 	free((caddr_t)V_ifindex_table, M_IFNET);
468 }
469 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
470     vnet_if_uninit, NULL);
471 #endif
472 
473 static void
474 if_link_ifnet(struct ifnet *ifp)
475 {
476 
477 	IFNET_WLOCK();
478 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
479 #ifdef VIMAGE
480 	curvnet->vnet_ifcnt++;
481 #endif
482 	IFNET_WUNLOCK();
483 }
484 
485 static bool
486 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
487 {
488 	struct ifnet *iter;
489 	int found = 0;
490 
491 	IFNET_WLOCK();
492 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
493 		if (iter == ifp) {
494 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
495 			if (!vmove)
496 				ifp->if_flags |= IFF_DYING;
497 			found = 1;
498 			break;
499 		}
500 #ifdef VIMAGE
501 	curvnet->vnet_ifcnt--;
502 #endif
503 	IFNET_WUNLOCK();
504 
505 	return (found);
506 }
507 
508 #ifdef VIMAGE
509 static void
510 vnet_if_return(const void *unused __unused)
511 {
512 	struct ifnet *ifp, *nifp;
513 	struct ifnet **pending;
514 	int found, i;
515 
516 	i = 0;
517 
518 	/*
519 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
520 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
521 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
522 	 * complete. We can't do that from within NET_EPOCH.
523 	 *
524 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
525 	 * read/write lock. We cannot hold the lock as we call if_vmove()
526 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
527 	 * ctx lock.
528 	 */
529 	IFNET_WLOCK();
530 
531 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
532 	    M_IFNET, M_WAITOK | M_ZERO);
533 
534 	/* Return all inherited interfaces to their parent vnets. */
535 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
536 		if (ifp->if_home_vnet != ifp->if_vnet) {
537 			found = if_unlink_ifnet(ifp, true);
538 			MPASS(found);
539 
540 			pending[i++] = ifp;
541 		}
542 	}
543 	IFNET_WUNLOCK();
544 
545 	for (int j = 0; j < i; j++) {
546 		if_vmove(pending[j], pending[j]->if_home_vnet);
547 	}
548 
549 	free(pending, M_IFNET);
550 }
551 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
552     vnet_if_return, NULL);
553 #endif
554 
555 static void *
556 if_grow(void)
557 {
558 	int oldlim;
559 	u_int n;
560 	struct ifnet **e;
561 	void *old;
562 
563 	old = NULL;
564 	IFNET_WLOCK_ASSERT();
565 	oldlim = V_if_indexlim;
566 	IFNET_WUNLOCK();
567 	n = (oldlim << 1) * sizeof(*e);
568 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
569 	IFNET_WLOCK();
570 	if (V_if_indexlim != oldlim) {
571 		free(e, M_IFNET);
572 		return (NULL);
573 	}
574 	if (V_ifindex_table != NULL) {
575 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
576 		old = V_ifindex_table;
577 	}
578 	V_if_indexlim <<= 1;
579 	V_ifindex_table = e;
580 	return (old);
581 }
582 
583 /*
584  * Allocate a struct ifnet and an index for an interface.  A layer 2
585  * common structure will also be allocated if an allocation routine is
586  * registered for the passed type.
587  */
588 struct ifnet *
589 if_alloc_domain(u_char type, int numa_domain)
590 {
591 	struct ifnet *ifp;
592 	u_short idx;
593 	void *old;
594 
595 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
596 	if (numa_domain == IF_NODOM)
597 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
598 		    M_WAITOK | M_ZERO);
599 	else
600 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
601 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
602  restart:
603 	IFNET_WLOCK();
604 	idx = ifindex_alloc(&old);
605 	if (__predict_false(idx == USHRT_MAX)) {
606 		IFNET_WUNLOCK();
607 		epoch_wait_preempt(net_epoch_preempt);
608 		free(old, M_IFNET);
609 		goto restart;
610 	}
611 	ifnet_setbyindex(idx, IFNET_HOLD);
612 	IFNET_WUNLOCK();
613 	ifp->if_index = idx;
614 	ifp->if_type = type;
615 	ifp->if_alloctype = type;
616 	ifp->if_numa_domain = numa_domain;
617 #ifdef VIMAGE
618 	ifp->if_vnet = curvnet;
619 #endif
620 	if (if_com_alloc[type] != NULL) {
621 		ifp->if_l2com = if_com_alloc[type](type, ifp);
622 		if (ifp->if_l2com == NULL) {
623 			free(ifp, M_IFNET);
624 			ifindex_free(idx);
625 			return (NULL);
626 		}
627 	}
628 
629 	IF_ADDR_LOCK_INIT(ifp);
630 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
631 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
632 	ifp->if_afdata_initialized = 0;
633 	IF_AFDATA_LOCK_INIT(ifp);
634 	CK_STAILQ_INIT(&ifp->if_addrhead);
635 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
636 	CK_STAILQ_INIT(&ifp->if_groups);
637 #ifdef MAC
638 	mac_ifnet_init(ifp);
639 #endif
640 	ifq_init(&ifp->if_snd, ifp);
641 
642 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
643 	for (int i = 0; i < IFCOUNTERS; i++)
644 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
645 	ifp->if_get_counter = if_get_counter_default;
646 	ifp->if_pcp = IFNET_PCP_NONE;
647 	ifnet_setbyindex(ifp->if_index, ifp);
648 	return (ifp);
649 }
650 
651 struct ifnet *
652 if_alloc_dev(u_char type, device_t dev)
653 {
654 	int numa_domain;
655 
656 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
657 		return (if_alloc_domain(type, IF_NODOM));
658 	return (if_alloc_domain(type, numa_domain));
659 }
660 
661 struct ifnet *
662 if_alloc(u_char type)
663 {
664 
665 	return (if_alloc_domain(type, IF_NODOM));
666 }
667 /*
668  * Do the actual work of freeing a struct ifnet, and layer 2 common
669  * structure.  This call is made when the last reference to an
670  * interface is released.
671  */
672 static void
673 if_free_internal(struct ifnet *ifp)
674 {
675 
676 	KASSERT((ifp->if_flags & IFF_DYING),
677 	    ("if_free_internal: interface not dying"));
678 
679 	if (if_com_free[ifp->if_alloctype] != NULL)
680 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
681 		    ifp->if_alloctype);
682 
683 #ifdef MAC
684 	mac_ifnet_destroy(ifp);
685 #endif /* MAC */
686 	IF_AFDATA_DESTROY(ifp);
687 	IF_ADDR_LOCK_DESTROY(ifp);
688 	ifq_delete(&ifp->if_snd);
689 
690 	for (int i = 0; i < IFCOUNTERS; i++)
691 		counter_u64_free(ifp->if_counters[i]);
692 
693 	free(ifp->if_description, M_IFDESCR);
694 	free(ifp->if_hw_addr, M_IFADDR);
695 	free(ifp, M_IFNET);
696 }
697 
698 static void
699 if_destroy(epoch_context_t ctx)
700 {
701 	struct ifnet *ifp;
702 
703 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
704 	if_free_internal(ifp);
705 }
706 
707 /*
708  * Deregister an interface and free the associated storage.
709  */
710 void
711 if_free(struct ifnet *ifp)
712 {
713 
714 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
715 
716 	CURVNET_SET_QUIET(ifp->if_vnet);
717 	IFNET_WLOCK();
718 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
719 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
720 
721 	ifindex_free_locked(ifp->if_index);
722 	IFNET_WUNLOCK();
723 
724 	if (refcount_release(&ifp->if_refcount))
725 		NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
726 	CURVNET_RESTORE();
727 }
728 
729 /*
730  * Interfaces to keep an ifnet type-stable despite the possibility of the
731  * driver calling if_free().  If there are additional references, we defer
732  * freeing the underlying data structure.
733  */
734 void
735 if_ref(struct ifnet *ifp)
736 {
737 
738 	/* We don't assert the ifnet list lock here, but arguably should. */
739 	refcount_acquire(&ifp->if_refcount);
740 }
741 
742 void
743 if_rele(struct ifnet *ifp)
744 {
745 
746 	if (!refcount_release(&ifp->if_refcount))
747 		return;
748 	NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
749 }
750 
751 void
752 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
753 {
754 
755 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
756 
757 	if (ifq->ifq_maxlen == 0)
758 		ifq->ifq_maxlen = ifqmaxlen;
759 
760 	ifq->altq_type = 0;
761 	ifq->altq_disc = NULL;
762 	ifq->altq_flags &= ALTQF_CANTCHANGE;
763 	ifq->altq_tbr  = NULL;
764 	ifq->altq_ifp  = ifp;
765 }
766 
767 void
768 ifq_delete(struct ifaltq *ifq)
769 {
770 	mtx_destroy(&ifq->ifq_mtx);
771 }
772 
773 /*
774  * Perform generic interface initialization tasks and attach the interface
775  * to the list of "active" interfaces.  If vmove flag is set on entry
776  * to if_attach_internal(), perform only a limited subset of initialization
777  * tasks, given that we are moving from one vnet to another an ifnet which
778  * has already been fully initialized.
779  *
780  * Note that if_detach_internal() removes group membership unconditionally
781  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
782  * Thus, when if_vmove() is applied to a cloned interface, group membership
783  * is lost while a cloned one always joins a group whose name is
784  * ifc->ifc_name.  To recover this after if_detach_internal() and
785  * if_attach_internal(), the cloner should be specified to
786  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
787  * attempts to join a group whose name is ifc->ifc_name.
788  *
789  * XXX:
790  *  - The decision to return void and thus require this function to
791  *    succeed is questionable.
792  *  - We should probably do more sanity checking.  For instance we don't
793  *    do anything to insure if_xname is unique or non-empty.
794  */
795 void
796 if_attach(struct ifnet *ifp)
797 {
798 
799 	if_attach_internal(ifp, 0, NULL);
800 }
801 
802 /*
803  * Compute the least common TSO limit.
804  */
805 void
806 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
807 {
808 	/*
809 	 * 1) If there is no limit currently, take the limit from
810 	 * the network adapter.
811 	 *
812 	 * 2) If the network adapter has a limit below the current
813 	 * limit, apply it.
814 	 */
815 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
816 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
817 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
818 	}
819 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
820 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
821 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
822 	}
823 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
824 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
825 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
826 	}
827 }
828 
829 /*
830  * Update TSO limit of a network adapter.
831  *
832  * Returns zero if no change. Else non-zero.
833  */
834 int
835 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
836 {
837 	int retval = 0;
838 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
839 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
840 		retval++;
841 	}
842 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
843 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
844 		retval++;
845 	}
846 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
847 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
848 		retval++;
849 	}
850 	return (retval);
851 }
852 
853 static void
854 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
855 {
856 	unsigned socksize, ifasize;
857 	int namelen, masklen;
858 	struct sockaddr_dl *sdl;
859 	struct ifaddr *ifa;
860 
861 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
862 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
863 		    ifp->if_xname);
864 
865 #ifdef VIMAGE
866 	ifp->if_vnet = curvnet;
867 	if (ifp->if_home_vnet == NULL)
868 		ifp->if_home_vnet = curvnet;
869 #endif
870 
871 	if_addgroup(ifp, IFG_ALL);
872 
873 	/* Restore group membership for cloned interfaces. */
874 	if (vmove && ifc != NULL)
875 		if_clone_addgroup(ifp, ifc);
876 
877 	getmicrotime(&ifp->if_lastchange);
878 	ifp->if_epoch = time_uptime;
879 
880 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
881 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
882 	    ("transmit and qflush must both either be set or both be NULL"));
883 	if (ifp->if_transmit == NULL) {
884 		ifp->if_transmit = if_transmit;
885 		ifp->if_qflush = if_qflush;
886 	}
887 	if (ifp->if_input == NULL)
888 		ifp->if_input = if_input_default;
889 
890 	if (ifp->if_requestencap == NULL)
891 		ifp->if_requestencap = if_requestencap_default;
892 
893 	if (!vmove) {
894 #ifdef MAC
895 		mac_ifnet_create(ifp);
896 #endif
897 
898 		/*
899 		 * Create a Link Level name for this device.
900 		 */
901 		namelen = strlen(ifp->if_xname);
902 		/*
903 		 * Always save enough space for any possiable name so we
904 		 * can do a rename in place later.
905 		 */
906 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
907 		socksize = masklen + ifp->if_addrlen;
908 		if (socksize < sizeof(*sdl))
909 			socksize = sizeof(*sdl);
910 		socksize = roundup2(socksize, sizeof(long));
911 		ifasize = sizeof(*ifa) + 2 * socksize;
912 		ifa = ifa_alloc(ifasize, M_WAITOK);
913 		sdl = (struct sockaddr_dl *)(ifa + 1);
914 		sdl->sdl_len = socksize;
915 		sdl->sdl_family = AF_LINK;
916 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
917 		sdl->sdl_nlen = namelen;
918 		sdl->sdl_index = ifp->if_index;
919 		sdl->sdl_type = ifp->if_type;
920 		ifp->if_addr = ifa;
921 		ifa->ifa_ifp = ifp;
922 		ifa->ifa_addr = (struct sockaddr *)sdl;
923 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
924 		ifa->ifa_netmask = (struct sockaddr *)sdl;
925 		sdl->sdl_len = masklen;
926 		while (namelen != 0)
927 			sdl->sdl_data[--namelen] = 0xff;
928 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
929 		/* Reliably crash if used uninitialized. */
930 		ifp->if_broadcastaddr = NULL;
931 
932 		if (ifp->if_type == IFT_ETHER) {
933 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
934 			    M_WAITOK | M_ZERO);
935 		}
936 
937 #if defined(INET) || defined(INET6)
938 		/* Use defaults for TSO, if nothing is set */
939 		if (ifp->if_hw_tsomax == 0 &&
940 		    ifp->if_hw_tsomaxsegcount == 0 &&
941 		    ifp->if_hw_tsomaxsegsize == 0) {
942 			/*
943 			 * The TSO defaults needs to be such that an
944 			 * NFS mbuf list of 35 mbufs totalling just
945 			 * below 64K works and that a chain of mbufs
946 			 * can be defragged into at most 32 segments:
947 			 */
948 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
949 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
950 			ifp->if_hw_tsomaxsegcount = 35;
951 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
952 
953 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
954 			if (ifp->if_capabilities & IFCAP_TSO) {
955 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
956 				    ifp->if_hw_tsomax,
957 				    ifp->if_hw_tsomaxsegcount,
958 				    ifp->if_hw_tsomaxsegsize);
959 			}
960 		}
961 #endif
962 	}
963 #ifdef VIMAGE
964 	else {
965 		/*
966 		 * Update the interface index in the link layer address
967 		 * of the interface.
968 		 */
969 		for (ifa = ifp->if_addr; ifa != NULL;
970 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
971 			if (ifa->ifa_addr->sa_family == AF_LINK) {
972 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
973 				sdl->sdl_index = ifp->if_index;
974 			}
975 		}
976 	}
977 #endif
978 
979 	if_link_ifnet(ifp);
980 
981 	if (domain_init_status >= 2)
982 		if_attachdomain1(ifp);
983 
984 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
985 	if (IS_DEFAULT_VNET(curvnet))
986 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
987 
988 	/* Announce the interface. */
989 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
990 }
991 
992 static void
993 if_epochalloc(void *dummy __unused)
994 {
995 
996 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
997 }
998 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
999 
1000 static void
1001 if_attachdomain(void *dummy)
1002 {
1003 	struct ifnet *ifp;
1004 
1005 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
1006 		if_attachdomain1(ifp);
1007 }
1008 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
1009     if_attachdomain, NULL);
1010 
1011 static void
1012 if_attachdomain1(struct ifnet *ifp)
1013 {
1014 	struct domain *dp;
1015 
1016 	/*
1017 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
1018 	 * cannot lock ifp->if_afdata initialization, entirely.
1019 	 */
1020 	IF_AFDATA_LOCK(ifp);
1021 	if (ifp->if_afdata_initialized >= domain_init_status) {
1022 		IF_AFDATA_UNLOCK(ifp);
1023 		log(LOG_WARNING, "%s called more than once on %s\n",
1024 		    __func__, ifp->if_xname);
1025 		return;
1026 	}
1027 	ifp->if_afdata_initialized = domain_init_status;
1028 	IF_AFDATA_UNLOCK(ifp);
1029 
1030 	/* address family dependent data region */
1031 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1032 	for (dp = domains; dp; dp = dp->dom_next) {
1033 		if (dp->dom_ifattach)
1034 			ifp->if_afdata[dp->dom_family] =
1035 			    (*dp->dom_ifattach)(ifp);
1036 	}
1037 }
1038 
1039 /*
1040  * Remove any unicast or broadcast network addresses from an interface.
1041  */
1042 void
1043 if_purgeaddrs(struct ifnet *ifp)
1044 {
1045 	struct ifaddr *ifa;
1046 
1047 	while (1) {
1048 		struct epoch_tracker et;
1049 
1050 		NET_EPOCH_ENTER(et);
1051 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1052 			if (ifa->ifa_addr->sa_family != AF_LINK)
1053 				break;
1054 		}
1055 		NET_EPOCH_EXIT(et);
1056 
1057 		if (ifa == NULL)
1058 			break;
1059 #ifdef INET
1060 		/* XXX: Ugly!! ad hoc just for INET */
1061 		if (ifa->ifa_addr->sa_family == AF_INET) {
1062 			struct ifaliasreq ifr;
1063 
1064 			bzero(&ifr, sizeof(ifr));
1065 			ifr.ifra_addr = *ifa->ifa_addr;
1066 			if (ifa->ifa_dstaddr)
1067 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1068 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1069 			    NULL) == 0)
1070 				continue;
1071 		}
1072 #endif /* INET */
1073 #ifdef INET6
1074 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1075 			in6_purgeaddr(ifa);
1076 			/* ifp_addrhead is already updated */
1077 			continue;
1078 		}
1079 #endif /* INET6 */
1080 		IF_ADDR_WLOCK(ifp);
1081 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1082 		IF_ADDR_WUNLOCK(ifp);
1083 		ifa_free(ifa);
1084 	}
1085 }
1086 
1087 /*
1088  * Remove any multicast network addresses from an interface when an ifnet
1089  * is going away.
1090  */
1091 static void
1092 if_purgemaddrs(struct ifnet *ifp)
1093 {
1094 	struct ifmultiaddr *ifma;
1095 
1096 	IF_ADDR_WLOCK(ifp);
1097 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1098 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1099 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1100 		if_delmulti_locked(ifp, ifma, 1);
1101 	}
1102 	IF_ADDR_WUNLOCK(ifp);
1103 }
1104 
1105 /*
1106  * Detach an interface, removing it from the list of "active" interfaces.
1107  * If vmove flag is set on entry to if_detach_internal(), perform only a
1108  * limited subset of cleanup tasks, given that we are moving an ifnet from
1109  * one vnet to another, where it must be fully operational.
1110  *
1111  * XXXRW: There are some significant questions about event ordering, and
1112  * how to prevent things from starting to use the interface during detach.
1113  */
1114 void
1115 if_detach(struct ifnet *ifp)
1116 {
1117 	bool found;
1118 
1119 	CURVNET_SET_QUIET(ifp->if_vnet);
1120 	found = if_unlink_ifnet(ifp, false);
1121 	if (found)
1122 		if_detach_internal(ifp, 0, NULL);
1123 	CURVNET_RESTORE();
1124 }
1125 
1126 /*
1127  * The vmove flag, if set, indicates that we are called from a callpath
1128  * that is moving an interface to a different vnet instance.
1129  *
1130  * The shutdown flag, if set, indicates that we are called in the
1131  * process of shutting down a vnet instance.  Currently only the
1132  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1133  * on a vnet instance shutdown without this flag being set, e.g., when
1134  * the cloned interfaces are destoyed as first thing of teardown.
1135  */
1136 static int
1137 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1138 {
1139 	struct ifaddr *ifa;
1140 	int i;
1141 	struct domain *dp;
1142 #ifdef VIMAGE
1143 	bool shutdown;
1144 
1145 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1146 #endif
1147 
1148 	/*
1149 	 * At this point we know the interface still was on the ifnet list
1150 	 * and we removed it so we are in a stable state.
1151 	 */
1152 	epoch_wait_preempt(net_epoch_preempt);
1153 
1154 	/*
1155 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1156 	 * fixes issues about late destruction of multicast options
1157 	 * which lead to leave group calls, which in turn access the
1158 	 * belonging ifnet structure:
1159 	 */
1160 	epoch_drain_callbacks(net_epoch_preempt);
1161 
1162 	/*
1163 	 * In any case (destroy or vmove) detach us from the groups
1164 	 * and remove/wait for pending events on the taskq.
1165 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1166 	 */
1167 	if_delgroups(ifp);
1168 
1169 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1170 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1171 
1172 	/*
1173 	 * Check if this is a cloned interface or not. Must do even if
1174 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1175 	 * the if_clone_addgroup() will have a corrupted string overwise
1176 	 * from a gibberish pointer.
1177 	 */
1178 	if (vmove && ifcp != NULL)
1179 		*ifcp = if_clone_findifc(ifp);
1180 
1181 	if_down(ifp);
1182 
1183 #ifdef VIMAGE
1184 	/*
1185 	 * On VNET shutdown abort here as the stack teardown will do all
1186 	 * the work top-down for us.
1187 	 */
1188 	if (shutdown) {
1189 		/* Give interface users the chance to clean up. */
1190 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1191 
1192 		/*
1193 		 * In case of a vmove we are done here without error.
1194 		 * If we would signal an error it would lead to the same
1195 		 * abort as if we did not find the ifnet anymore.
1196 		 * if_detach() calls us in void context and does not care
1197 		 * about an early abort notification, so life is splendid :)
1198 		 */
1199 		goto finish_vnet_shutdown;
1200 	}
1201 #endif
1202 
1203 	/*
1204 	 * At this point we are not tearing down a VNET and are either
1205 	 * going to destroy or vmove the interface and have to cleanup
1206 	 * accordingly.
1207 	 */
1208 
1209 	/*
1210 	 * Remove routes and flush queues.
1211 	 */
1212 #ifdef ALTQ
1213 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1214 		altq_disable(&ifp->if_snd);
1215 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1216 		altq_detach(&ifp->if_snd);
1217 #endif
1218 
1219 	if_purgeaddrs(ifp);
1220 
1221 #ifdef INET
1222 	in_ifdetach(ifp);
1223 #endif
1224 
1225 #ifdef INET6
1226 	/*
1227 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1228 	 * before removing routing entries below, since IPv6 interface direct
1229 	 * routes are expected to be removed by the IPv6-specific kernel API.
1230 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1231 	 */
1232 	in6_ifdetach(ifp);
1233 #endif
1234 	if_purgemaddrs(ifp);
1235 
1236 	/* Announce that the interface is gone. */
1237 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1238 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1239 	if (IS_DEFAULT_VNET(curvnet))
1240 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1241 
1242 	if (!vmove) {
1243 		/*
1244 		 * Prevent further calls into the device driver via ifnet.
1245 		 */
1246 		if_dead(ifp);
1247 
1248 		/*
1249 		 * Clean up all addresses.
1250 		 */
1251 		IF_ADDR_WLOCK(ifp);
1252 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1253 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1254 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1255 			IF_ADDR_WUNLOCK(ifp);
1256 			ifa_free(ifa);
1257 		} else
1258 			IF_ADDR_WUNLOCK(ifp);
1259 	}
1260 
1261 	rt_flushifroutes(ifp);
1262 
1263 #ifdef VIMAGE
1264 finish_vnet_shutdown:
1265 #endif
1266 	/*
1267 	 * We cannot hold the lock over dom_ifdetach calls as they might
1268 	 * sleep, for example trying to drain a callout, thus open up the
1269 	 * theoretical race with re-attaching.
1270 	 */
1271 	IF_AFDATA_LOCK(ifp);
1272 	i = ifp->if_afdata_initialized;
1273 	ifp->if_afdata_initialized = 0;
1274 	IF_AFDATA_UNLOCK(ifp);
1275 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1276 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1277 			(*dp->dom_ifdetach)(ifp,
1278 			    ifp->if_afdata[dp->dom_family]);
1279 			ifp->if_afdata[dp->dom_family] = NULL;
1280 		}
1281 	}
1282 
1283 	return (0);
1284 }
1285 
1286 #ifdef VIMAGE
1287 /*
1288  * if_vmove() performs a limited version of if_detach() in current
1289  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1290  * An attempt is made to shrink if_index in current vnet, find an
1291  * unused if_index in target vnet and calls if_grow() if necessary,
1292  * and finally find an unused if_xname for the target vnet.
1293  */
1294 static int
1295 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1296 {
1297 	struct if_clone *ifc;
1298 #ifdef DEV_BPF
1299 	u_int bif_dlt, bif_hdrlen;
1300 #endif
1301 	void *old;
1302 	int rc;
1303 
1304 #ifdef DEV_BPF
1305  	/*
1306 	 * if_detach_internal() will call the eventhandler to notify
1307 	 * interface departure.  That will detach if_bpf.  We need to
1308 	 * safe the dlt and hdrlen so we can re-attach it later.
1309 	 */
1310 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1311 #endif
1312 
1313 	/*
1314 	 * Detach from current vnet, but preserve LLADDR info, do not
1315 	 * mark as dead etc. so that the ifnet can be reattached later.
1316 	 * If we cannot find it, we lost the race to someone else.
1317 	 */
1318 	rc = if_detach_internal(ifp, 1, &ifc);
1319 	if (rc != 0)
1320 		return (rc);
1321 
1322 	/*
1323 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1324 	 * the if_index for that vnet if possible.
1325 	 *
1326 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1327 	 * or we'd lock on one vnet and unlock on another.
1328 	 */
1329 	IFNET_WLOCK();
1330 	ifindex_free_locked(ifp->if_index);
1331 	IFNET_WUNLOCK();
1332 
1333 
1334 	/* Don't re-attach DYING interfaces. */
1335 	if (ifp->if_flags & IFF_DYING)
1336 		return (0);
1337 
1338 	/*
1339 	 * Perform interface-specific reassignment tasks, if provided by
1340 	 * the driver.
1341 	 */
1342 	if (ifp->if_reassign != NULL)
1343 		ifp->if_reassign(ifp, new_vnet, NULL);
1344 
1345 	/*
1346 	 * Switch to the context of the target vnet.
1347 	 */
1348 	CURVNET_SET_QUIET(new_vnet);
1349  restart:
1350 	IFNET_WLOCK();
1351 	ifp->if_index = ifindex_alloc(&old);
1352 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1353 		IFNET_WUNLOCK();
1354 		epoch_wait_preempt(net_epoch_preempt);
1355 		free(old, M_IFNET);
1356 		goto restart;
1357 	}
1358 	ifnet_setbyindex(ifp->if_index, ifp);
1359 	IFNET_WUNLOCK();
1360 
1361 	if_attach_internal(ifp, 1, ifc);
1362 
1363 #ifdef DEV_BPF
1364 	if (ifp->if_bpf == NULL)
1365 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1366 #endif
1367 
1368 	CURVNET_RESTORE();
1369 	return (0);
1370 }
1371 
1372 /*
1373  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1374  */
1375 static int
1376 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1377 {
1378 	struct prison *pr;
1379 	struct ifnet *difp;
1380 	int error;
1381 	bool found;
1382 	bool shutdown;
1383 
1384 	/* Try to find the prison within our visibility. */
1385 	sx_slock(&allprison_lock);
1386 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1387 	sx_sunlock(&allprison_lock);
1388 	if (pr == NULL)
1389 		return (ENXIO);
1390 	prison_hold_locked(pr);
1391 	mtx_unlock(&pr->pr_mtx);
1392 
1393 	/* Do not try to move the iface from and to the same prison. */
1394 	if (pr->pr_vnet == ifp->if_vnet) {
1395 		prison_free(pr);
1396 		return (EEXIST);
1397 	}
1398 
1399 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1400 	/* XXX Lock interfaces to avoid races. */
1401 	CURVNET_SET_QUIET(pr->pr_vnet);
1402 	difp = ifunit(ifname);
1403 	if (difp != NULL) {
1404 		CURVNET_RESTORE();
1405 		prison_free(pr);
1406 		return (EEXIST);
1407 	}
1408 
1409 	/* Make sure the VNET is stable. */
1410 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1411 	if (shutdown) {
1412 		CURVNET_RESTORE();
1413 		prison_free(pr);
1414 		return (EBUSY);
1415 	}
1416 	CURVNET_RESTORE();
1417 
1418 	found = if_unlink_ifnet(ifp, true);
1419 	MPASS(found);
1420 
1421 	/* Move the interface into the child jail/vnet. */
1422 	error = if_vmove(ifp, pr->pr_vnet);
1423 
1424 	/* Report the new if_xname back to the userland on success. */
1425 	if (error == 0)
1426 		sprintf(ifname, "%s", ifp->if_xname);
1427 
1428 	prison_free(pr);
1429 	return (error);
1430 }
1431 
1432 static int
1433 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1434 {
1435 	struct prison *pr;
1436 	struct vnet *vnet_dst;
1437 	struct ifnet *ifp;
1438 	int error, found;
1439  	bool shutdown;
1440 
1441 	/* Try to find the prison within our visibility. */
1442 	sx_slock(&allprison_lock);
1443 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1444 	sx_sunlock(&allprison_lock);
1445 	if (pr == NULL)
1446 		return (ENXIO);
1447 	prison_hold_locked(pr);
1448 	mtx_unlock(&pr->pr_mtx);
1449 
1450 	/* Make sure the named iface exists in the source prison/vnet. */
1451 	CURVNET_SET(pr->pr_vnet);
1452 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1453 	if (ifp == NULL) {
1454 		CURVNET_RESTORE();
1455 		prison_free(pr);
1456 		return (ENXIO);
1457 	}
1458 
1459 	/* Do not try to move the iface from and to the same prison. */
1460 	vnet_dst = TD_TO_VNET(td);
1461 	if (vnet_dst == ifp->if_vnet) {
1462 		CURVNET_RESTORE();
1463 		prison_free(pr);
1464 		return (EEXIST);
1465 	}
1466 
1467 	/* Make sure the VNET is stable. */
1468 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1469 	if (shutdown) {
1470 		CURVNET_RESTORE();
1471 		prison_free(pr);
1472 		return (EBUSY);
1473 	}
1474 
1475 	/* Get interface back from child jail/vnet. */
1476 	found = if_unlink_ifnet(ifp, true);
1477 	MPASS(found);
1478 	error = if_vmove(ifp, vnet_dst);
1479 	CURVNET_RESTORE();
1480 
1481 	/* Report the new if_xname back to the userland on success. */
1482 	if (error == 0)
1483 		sprintf(ifname, "%s", ifp->if_xname);
1484 
1485 	prison_free(pr);
1486 	return (error);
1487 }
1488 #endif /* VIMAGE */
1489 
1490 /*
1491  * Add a group to an interface
1492  */
1493 int
1494 if_addgroup(struct ifnet *ifp, const char *groupname)
1495 {
1496 	struct ifg_list		*ifgl;
1497 	struct ifg_group	*ifg = NULL;
1498 	struct ifg_member	*ifgm;
1499 	int 			 new = 0;
1500 
1501 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1502 	    groupname[strlen(groupname) - 1] <= '9')
1503 		return (EINVAL);
1504 
1505 	IFNET_WLOCK();
1506 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1507 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1508 			IFNET_WUNLOCK();
1509 			return (EEXIST);
1510 		}
1511 
1512 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1513 	    	IFNET_WUNLOCK();
1514 		return (ENOMEM);
1515 	}
1516 
1517 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1518 		free(ifgl, M_TEMP);
1519 		IFNET_WUNLOCK();
1520 		return (ENOMEM);
1521 	}
1522 
1523 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1524 		if (!strcmp(ifg->ifg_group, groupname))
1525 			break;
1526 
1527 	if (ifg == NULL) {
1528 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1529 			free(ifgl, M_TEMP);
1530 			free(ifgm, M_TEMP);
1531 			IFNET_WUNLOCK();
1532 			return (ENOMEM);
1533 		}
1534 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1535 		ifg->ifg_refcnt = 0;
1536 		CK_STAILQ_INIT(&ifg->ifg_members);
1537 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1538 		new = 1;
1539 	}
1540 
1541 	ifg->ifg_refcnt++;
1542 	ifgl->ifgl_group = ifg;
1543 	ifgm->ifgm_ifp = ifp;
1544 
1545 	IF_ADDR_WLOCK(ifp);
1546 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1547 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1548 	IF_ADDR_WUNLOCK(ifp);
1549 
1550 	IFNET_WUNLOCK();
1551 
1552 	if (new)
1553 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1554 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1555 
1556 	return (0);
1557 }
1558 
1559 /*
1560  * Helper function to remove a group out of an interface.  Expects the global
1561  * ifnet lock to be write-locked, and drops it before returning.
1562  */
1563 static void
1564 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1565     const char *groupname)
1566 {
1567 	struct ifg_member *ifgm;
1568 	bool freeifgl;
1569 
1570 	IFNET_WLOCK_ASSERT();
1571 
1572 	IF_ADDR_WLOCK(ifp);
1573 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1574 	IF_ADDR_WUNLOCK(ifp);
1575 
1576 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1577 		if (ifgm->ifgm_ifp == ifp) {
1578 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1579 			    ifg_member, ifgm_next);
1580 			break;
1581 		}
1582 	}
1583 
1584 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1585 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1586 		    ifg_next);
1587 		freeifgl = true;
1588 	} else {
1589 		freeifgl = false;
1590 	}
1591 	IFNET_WUNLOCK();
1592 
1593 	epoch_wait_preempt(net_epoch_preempt);
1594 	if (freeifgl) {
1595 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1596 		free(ifgl->ifgl_group, M_TEMP);
1597 	}
1598 	free(ifgm, M_TEMP);
1599 	free(ifgl, M_TEMP);
1600 
1601 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1602 }
1603 
1604 /*
1605  * Remove a group from an interface
1606  */
1607 int
1608 if_delgroup(struct ifnet *ifp, const char *groupname)
1609 {
1610 	struct ifg_list *ifgl;
1611 
1612 	IFNET_WLOCK();
1613 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1614 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1615 			break;
1616 	if (ifgl == NULL) {
1617 		IFNET_WUNLOCK();
1618 		return (ENOENT);
1619 	}
1620 
1621 	_if_delgroup_locked(ifp, ifgl, groupname);
1622 
1623 	return (0);
1624 }
1625 
1626 /*
1627  * Remove an interface from all groups
1628  */
1629 static void
1630 if_delgroups(struct ifnet *ifp)
1631 {
1632 	struct ifg_list *ifgl;
1633 	char groupname[IFNAMSIZ];
1634 
1635 	IFNET_WLOCK();
1636 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1637 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1638 		_if_delgroup_locked(ifp, ifgl, groupname);
1639 		IFNET_WLOCK();
1640 	}
1641 	IFNET_WUNLOCK();
1642 }
1643 
1644 static char *
1645 ifgr_group_get(void *ifgrp)
1646 {
1647 	union ifgroupreq_union *ifgrup;
1648 
1649 	ifgrup = ifgrp;
1650 #ifdef COMPAT_FREEBSD32
1651 	if (SV_CURPROC_FLAG(SV_ILP32))
1652 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1653 #endif
1654 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1655 }
1656 
1657 static struct ifg_req *
1658 ifgr_groups_get(void *ifgrp)
1659 {
1660 	union ifgroupreq_union *ifgrup;
1661 
1662 	ifgrup = ifgrp;
1663 #ifdef COMPAT_FREEBSD32
1664 	if (SV_CURPROC_FLAG(SV_ILP32))
1665 		return ((struct ifg_req *)(uintptr_t)
1666 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1667 #endif
1668 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1669 }
1670 
1671 /*
1672  * Stores all groups from an interface in memory pointed to by ifgr.
1673  */
1674 static int
1675 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1676 {
1677 	int			 len, error;
1678 	struct ifg_list		*ifgl;
1679 	struct ifg_req		 ifgrq, *ifgp;
1680 
1681 	NET_EPOCH_ASSERT();
1682 
1683 	if (ifgr->ifgr_len == 0) {
1684 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1685 			ifgr->ifgr_len += sizeof(struct ifg_req);
1686 		return (0);
1687 	}
1688 
1689 	len = ifgr->ifgr_len;
1690 	ifgp = ifgr_groups_get(ifgr);
1691 	/* XXX: wire */
1692 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1693 		if (len < sizeof(ifgrq))
1694 			return (EINVAL);
1695 		bzero(&ifgrq, sizeof ifgrq);
1696 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1697 		    sizeof(ifgrq.ifgrq_group));
1698 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1699 			return (error);
1700 		len -= sizeof(ifgrq);
1701 		ifgp++;
1702 	}
1703 
1704 	return (0);
1705 }
1706 
1707 /*
1708  * Stores all members of a group in memory pointed to by igfr
1709  */
1710 static int
1711 if_getgroupmembers(struct ifgroupreq *ifgr)
1712 {
1713 	struct ifg_group	*ifg;
1714 	struct ifg_member	*ifgm;
1715 	struct ifg_req		 ifgrq, *ifgp;
1716 	int			 len, error;
1717 
1718 	IFNET_RLOCK();
1719 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1720 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1721 			break;
1722 	if (ifg == NULL) {
1723 		IFNET_RUNLOCK();
1724 		return (ENOENT);
1725 	}
1726 
1727 	if (ifgr->ifgr_len == 0) {
1728 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1729 			ifgr->ifgr_len += sizeof(ifgrq);
1730 		IFNET_RUNLOCK();
1731 		return (0);
1732 	}
1733 
1734 	len = ifgr->ifgr_len;
1735 	ifgp = ifgr_groups_get(ifgr);
1736 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1737 		if (len < sizeof(ifgrq)) {
1738 			IFNET_RUNLOCK();
1739 			return (EINVAL);
1740 		}
1741 		bzero(&ifgrq, sizeof ifgrq);
1742 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1743 		    sizeof(ifgrq.ifgrq_member));
1744 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1745 			IFNET_RUNLOCK();
1746 			return (error);
1747 		}
1748 		len -= sizeof(ifgrq);
1749 		ifgp++;
1750 	}
1751 	IFNET_RUNLOCK();
1752 
1753 	return (0);
1754 }
1755 
1756 /*
1757  * Return counter values from counter(9)s stored in ifnet.
1758  */
1759 uint64_t
1760 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1761 {
1762 
1763 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1764 
1765 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1766 }
1767 
1768 /*
1769  * Increase an ifnet counter. Usually used for counters shared
1770  * between the stack and a driver, but function supports them all.
1771  */
1772 void
1773 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1774 {
1775 
1776 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1777 
1778 	counter_u64_add(ifp->if_counters[cnt], inc);
1779 }
1780 
1781 /*
1782  * Copy data from ifnet to userland API structure if_data.
1783  */
1784 void
1785 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1786 {
1787 
1788 	ifd->ifi_type = ifp->if_type;
1789 	ifd->ifi_physical = 0;
1790 	ifd->ifi_addrlen = ifp->if_addrlen;
1791 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1792 	ifd->ifi_link_state = ifp->if_link_state;
1793 	ifd->ifi_vhid = 0;
1794 	ifd->ifi_datalen = sizeof(struct if_data);
1795 	ifd->ifi_mtu = ifp->if_mtu;
1796 	ifd->ifi_metric = ifp->if_metric;
1797 	ifd->ifi_baudrate = ifp->if_baudrate;
1798 	ifd->ifi_hwassist = ifp->if_hwassist;
1799 	ifd->ifi_epoch = ifp->if_epoch;
1800 	ifd->ifi_lastchange = ifp->if_lastchange;
1801 
1802 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1803 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1804 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1805 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1806 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1807 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1808 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1809 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1810 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1811 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1812 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1813 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1814 }
1815 
1816 /*
1817  * Initialization, destruction and refcounting functions for ifaddrs.
1818  */
1819 struct ifaddr *
1820 ifa_alloc(size_t size, int flags)
1821 {
1822 	struct ifaddr *ifa;
1823 
1824 	KASSERT(size >= sizeof(struct ifaddr),
1825 	    ("%s: invalid size %zu", __func__, size));
1826 
1827 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1828 	if (ifa == NULL)
1829 		return (NULL);
1830 
1831 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1832 		goto fail;
1833 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1834 		goto fail;
1835 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1836 		goto fail;
1837 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1838 		goto fail;
1839 
1840 	refcount_init(&ifa->ifa_refcnt, 1);
1841 
1842 	return (ifa);
1843 
1844 fail:
1845 	/* free(NULL) is okay */
1846 	counter_u64_free(ifa->ifa_opackets);
1847 	counter_u64_free(ifa->ifa_ipackets);
1848 	counter_u64_free(ifa->ifa_obytes);
1849 	counter_u64_free(ifa->ifa_ibytes);
1850 	free(ifa, M_IFADDR);
1851 
1852 	return (NULL);
1853 }
1854 
1855 void
1856 ifa_ref(struct ifaddr *ifa)
1857 {
1858 
1859 	refcount_acquire(&ifa->ifa_refcnt);
1860 }
1861 
1862 static void
1863 ifa_destroy(epoch_context_t ctx)
1864 {
1865 	struct ifaddr *ifa;
1866 
1867 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1868 	counter_u64_free(ifa->ifa_opackets);
1869 	counter_u64_free(ifa->ifa_ipackets);
1870 	counter_u64_free(ifa->ifa_obytes);
1871 	counter_u64_free(ifa->ifa_ibytes);
1872 	free(ifa, M_IFADDR);
1873 }
1874 
1875 void
1876 ifa_free(struct ifaddr *ifa)
1877 {
1878 
1879 	if (refcount_release(&ifa->ifa_refcnt))
1880 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1881 }
1882 
1883 /*
1884  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1885  * structs used to represent other address families, it is necessary
1886  * to perform a different comparison.
1887  */
1888 
1889 #define	sa_dl_equal(a1, a2)	\
1890 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1891 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1892 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1893 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1894 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1895 
1896 /*
1897  * Locate an interface based on a complete address.
1898  */
1899 /*ARGSUSED*/
1900 struct ifaddr *
1901 ifa_ifwithaddr(const struct sockaddr *addr)
1902 {
1903 	struct ifnet *ifp;
1904 	struct ifaddr *ifa;
1905 
1906 	NET_EPOCH_ASSERT();
1907 
1908 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1909 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1910 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1911 				continue;
1912 			if (sa_equal(addr, ifa->ifa_addr)) {
1913 				goto done;
1914 			}
1915 			/* IP6 doesn't have broadcast */
1916 			if ((ifp->if_flags & IFF_BROADCAST) &&
1917 			    ifa->ifa_broadaddr &&
1918 			    ifa->ifa_broadaddr->sa_len != 0 &&
1919 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1920 				goto done;
1921 			}
1922 		}
1923 	}
1924 	ifa = NULL;
1925 done:
1926 	return (ifa);
1927 }
1928 
1929 int
1930 ifa_ifwithaddr_check(const struct sockaddr *addr)
1931 {
1932 	struct epoch_tracker et;
1933 	int rc;
1934 
1935 	NET_EPOCH_ENTER(et);
1936 	rc = (ifa_ifwithaddr(addr) != NULL);
1937 	NET_EPOCH_EXIT(et);
1938 	return (rc);
1939 }
1940 
1941 /*
1942  * Locate an interface based on the broadcast address.
1943  */
1944 /* ARGSUSED */
1945 struct ifaddr *
1946 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1947 {
1948 	struct ifnet *ifp;
1949 	struct ifaddr *ifa;
1950 
1951 	NET_EPOCH_ASSERT();
1952 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1953 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1954 			continue;
1955 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1956 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1957 				continue;
1958 			if ((ifp->if_flags & IFF_BROADCAST) &&
1959 			    ifa->ifa_broadaddr &&
1960 			    ifa->ifa_broadaddr->sa_len != 0 &&
1961 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1962 				goto done;
1963 			}
1964 		}
1965 	}
1966 	ifa = NULL;
1967 done:
1968 	return (ifa);
1969 }
1970 
1971 /*
1972  * Locate the point to point interface with a given destination address.
1973  */
1974 /*ARGSUSED*/
1975 struct ifaddr *
1976 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1977 {
1978 	struct ifnet *ifp;
1979 	struct ifaddr *ifa;
1980 
1981 	NET_EPOCH_ASSERT();
1982 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1983 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1984 			continue;
1985 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1986 			continue;
1987 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1988 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1989 				continue;
1990 			if (ifa->ifa_dstaddr != NULL &&
1991 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1992 				goto done;
1993 			}
1994 		}
1995 	}
1996 	ifa = NULL;
1997 done:
1998 	return (ifa);
1999 }
2000 
2001 /*
2002  * Find an interface on a specific network.  If many, choice
2003  * is most specific found.
2004  */
2005 struct ifaddr *
2006 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2007 {
2008 	struct ifnet *ifp;
2009 	struct ifaddr *ifa;
2010 	struct ifaddr *ifa_maybe = NULL;
2011 	u_int af = addr->sa_family;
2012 	const char *addr_data = addr->sa_data, *cplim;
2013 
2014 	NET_EPOCH_ASSERT();
2015 	/*
2016 	 * AF_LINK addresses can be looked up directly by their index number,
2017 	 * so do that if we can.
2018 	 */
2019 	if (af == AF_LINK) {
2020 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2021 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2022 		return (ifaddr_byindex(sdl->sdl_index));
2023 	}
2024 
2025 	/*
2026 	 * Scan though each interface, looking for ones that have addresses
2027 	 * in this address family and the requested fib.
2028 	 */
2029 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2030 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2031 			continue;
2032 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2033 			const char *cp, *cp2, *cp3;
2034 
2035 			if (ifa->ifa_addr->sa_family != af)
2036 next:				continue;
2037 			if (af == AF_INET &&
2038 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2039 				/*
2040 				 * This is a bit broken as it doesn't
2041 				 * take into account that the remote end may
2042 				 * be a single node in the network we are
2043 				 * looking for.
2044 				 * The trouble is that we don't know the
2045 				 * netmask for the remote end.
2046 				 */
2047 				if (ifa->ifa_dstaddr != NULL &&
2048 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2049 					goto done;
2050 				}
2051 			} else {
2052 				/*
2053 				 * Scan all the bits in the ifa's address.
2054 				 * If a bit dissagrees with what we are
2055 				 * looking for, mask it with the netmask
2056 				 * to see if it really matters.
2057 				 * (A byte at a time)
2058 				 */
2059 				if (ifa->ifa_netmask == 0)
2060 					continue;
2061 				cp = addr_data;
2062 				cp2 = ifa->ifa_addr->sa_data;
2063 				cp3 = ifa->ifa_netmask->sa_data;
2064 				cplim = ifa->ifa_netmask->sa_len
2065 					+ (char *)ifa->ifa_netmask;
2066 				while (cp3 < cplim)
2067 					if ((*cp++ ^ *cp2++) & *cp3++)
2068 						goto next; /* next address! */
2069 				/*
2070 				 * If the netmask of what we just found
2071 				 * is more specific than what we had before
2072 				 * (if we had one), or if the virtual status
2073 				 * of new prefix is better than of the old one,
2074 				 * then remember the new one before continuing
2075 				 * to search for an even better one.
2076 				 */
2077 				if (ifa_maybe == NULL ||
2078 				    ifa_preferred(ifa_maybe, ifa) ||
2079 				    rn_refines((caddr_t)ifa->ifa_netmask,
2080 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2081 					ifa_maybe = ifa;
2082 				}
2083 			}
2084 		}
2085 	}
2086 	ifa = ifa_maybe;
2087 	ifa_maybe = NULL;
2088 done:
2089 	return (ifa);
2090 }
2091 
2092 /*
2093  * Find an interface address specific to an interface best matching
2094  * a given address.
2095  */
2096 struct ifaddr *
2097 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2098 {
2099 	struct ifaddr *ifa;
2100 	const char *cp, *cp2, *cp3;
2101 	char *cplim;
2102 	struct ifaddr *ifa_maybe = NULL;
2103 	u_int af = addr->sa_family;
2104 
2105 	if (af >= AF_MAX)
2106 		return (NULL);
2107 
2108 	NET_EPOCH_ASSERT();
2109 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2110 		if (ifa->ifa_addr->sa_family != af)
2111 			continue;
2112 		if (ifa_maybe == NULL)
2113 			ifa_maybe = ifa;
2114 		if (ifa->ifa_netmask == 0) {
2115 			if (sa_equal(addr, ifa->ifa_addr) ||
2116 			    (ifa->ifa_dstaddr &&
2117 			    sa_equal(addr, ifa->ifa_dstaddr)))
2118 				goto done;
2119 			continue;
2120 		}
2121 		if (ifp->if_flags & IFF_POINTOPOINT) {
2122 			if (sa_equal(addr, ifa->ifa_dstaddr))
2123 				goto done;
2124 		} else {
2125 			cp = addr->sa_data;
2126 			cp2 = ifa->ifa_addr->sa_data;
2127 			cp3 = ifa->ifa_netmask->sa_data;
2128 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2129 			for (; cp3 < cplim; cp3++)
2130 				if ((*cp++ ^ *cp2++) & *cp3)
2131 					break;
2132 			if (cp3 == cplim)
2133 				goto done;
2134 		}
2135 	}
2136 	ifa = ifa_maybe;
2137 done:
2138 	return (ifa);
2139 }
2140 
2141 /*
2142  * See whether new ifa is better than current one:
2143  * 1) A non-virtual one is preferred over virtual.
2144  * 2) A virtual in master state preferred over any other state.
2145  *
2146  * Used in several address selecting functions.
2147  */
2148 int
2149 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2150 {
2151 
2152 	return (cur->ifa_carp && (!next->ifa_carp ||
2153 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2154 }
2155 
2156 struct sockaddr_dl *
2157 link_alloc_sdl(size_t size, int flags)
2158 {
2159 
2160 	return (malloc(size, M_TEMP, flags));
2161 }
2162 
2163 void
2164 link_free_sdl(struct sockaddr *sa)
2165 {
2166 	free(sa, M_TEMP);
2167 }
2168 
2169 /*
2170  * Fills in given sdl with interface basic info.
2171  * Returns pointer to filled sdl.
2172  */
2173 struct sockaddr_dl *
2174 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2175 {
2176 	struct sockaddr_dl *sdl;
2177 
2178 	sdl = (struct sockaddr_dl *)paddr;
2179 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2180 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2181 	sdl->sdl_family = AF_LINK;
2182 	sdl->sdl_index = ifp->if_index;
2183 	sdl->sdl_type = iftype;
2184 
2185 	return (sdl);
2186 }
2187 
2188 /*
2189  * Mark an interface down and notify protocols of
2190  * the transition.
2191  */
2192 static void
2193 if_unroute(struct ifnet *ifp, int flag, int fam)
2194 {
2195 	struct ifaddr *ifa;
2196 
2197 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2198 
2199 	ifp->if_flags &= ~flag;
2200 	getmicrotime(&ifp->if_lastchange);
2201 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2202 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2203 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2204 	ifp->if_qflush(ifp);
2205 
2206 	if (ifp->if_carp)
2207 		(*carp_linkstate_p)(ifp);
2208 	rt_ifmsg(ifp);
2209 }
2210 
2211 /*
2212  * Mark an interface up and notify protocols of
2213  * the transition.
2214  */
2215 static void
2216 if_route(struct ifnet *ifp, int flag, int fam)
2217 {
2218 	struct ifaddr *ifa;
2219 
2220 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2221 
2222 	ifp->if_flags |= flag;
2223 	getmicrotime(&ifp->if_lastchange);
2224 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2225 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2226 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2227 	if (ifp->if_carp)
2228 		(*carp_linkstate_p)(ifp);
2229 	rt_ifmsg(ifp);
2230 #ifdef INET6
2231 	in6_if_up(ifp);
2232 #endif
2233 }
2234 
2235 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2236 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2237 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2238 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2239 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2240 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2241 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2242 void	*(*vlan_cookie_p)(struct ifnet *);
2243 
2244 /*
2245  * Handle a change in the interface link state. To avoid LORs
2246  * between driver lock and upper layer locks, as well as possible
2247  * recursions, we post event to taskqueue, and all job
2248  * is done in static do_link_state_change().
2249  */
2250 void
2251 if_link_state_change(struct ifnet *ifp, int link_state)
2252 {
2253 	/* Return if state hasn't changed. */
2254 	if (ifp->if_link_state == link_state)
2255 		return;
2256 
2257 	ifp->if_link_state = link_state;
2258 
2259 	/* XXXGL: reference ifp? */
2260 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2261 }
2262 
2263 static void
2264 do_link_state_change(void *arg, int pending)
2265 {
2266 	struct ifnet *ifp;
2267 	int link_state;
2268 
2269 	ifp = arg;
2270 	link_state = ifp->if_link_state;
2271 
2272 	CURVNET_SET(ifp->if_vnet);
2273 	rt_ifmsg(ifp);
2274 	if (ifp->if_vlantrunk != NULL)
2275 		(*vlan_link_state_p)(ifp);
2276 
2277 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2278 	    ifp->if_l2com != NULL)
2279 		(*ng_ether_link_state_p)(ifp, link_state);
2280 	if (ifp->if_carp)
2281 		(*carp_linkstate_p)(ifp);
2282 	if (ifp->if_bridge)
2283 		ifp->if_bridge_linkstate(ifp);
2284 	if (ifp->if_lagg)
2285 		(*lagg_linkstate_p)(ifp, link_state);
2286 
2287 	if (IS_DEFAULT_VNET(curvnet))
2288 		devctl_notify("IFNET", ifp->if_xname,
2289 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2290 		    NULL);
2291 	if (pending > 1)
2292 		if_printf(ifp, "%d link states coalesced\n", pending);
2293 	if (log_link_state_change)
2294 		if_printf(ifp, "link state changed to %s\n",
2295 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2296 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2297 	CURVNET_RESTORE();
2298 }
2299 
2300 /*
2301  * Mark an interface down and notify protocols of
2302  * the transition.
2303  */
2304 void
2305 if_down(struct ifnet *ifp)
2306 {
2307 
2308 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2309 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2310 }
2311 
2312 /*
2313  * Mark an interface up and notify protocols of
2314  * the transition.
2315  */
2316 void
2317 if_up(struct ifnet *ifp)
2318 {
2319 
2320 	if_route(ifp, IFF_UP, AF_UNSPEC);
2321 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2322 }
2323 
2324 /*
2325  * Flush an interface queue.
2326  */
2327 void
2328 if_qflush(struct ifnet *ifp)
2329 {
2330 	struct mbuf *m, *n;
2331 	struct ifaltq *ifq;
2332 
2333 	ifq = &ifp->if_snd;
2334 	IFQ_LOCK(ifq);
2335 #ifdef ALTQ
2336 	if (ALTQ_IS_ENABLED(ifq))
2337 		ALTQ_PURGE(ifq);
2338 #endif
2339 	n = ifq->ifq_head;
2340 	while ((m = n) != NULL) {
2341 		n = m->m_nextpkt;
2342 		m_freem(m);
2343 	}
2344 	ifq->ifq_head = 0;
2345 	ifq->ifq_tail = 0;
2346 	ifq->ifq_len = 0;
2347 	IFQ_UNLOCK(ifq);
2348 }
2349 
2350 /*
2351  * Map interface name to interface structure pointer, with or without
2352  * returning a reference.
2353  */
2354 struct ifnet *
2355 ifunit_ref(const char *name)
2356 {
2357 	struct epoch_tracker et;
2358 	struct ifnet *ifp;
2359 
2360 	NET_EPOCH_ENTER(et);
2361 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2362 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2363 		    !(ifp->if_flags & IFF_DYING))
2364 			break;
2365 	}
2366 	if (ifp != NULL)
2367 		if_ref(ifp);
2368 	NET_EPOCH_EXIT(et);
2369 	return (ifp);
2370 }
2371 
2372 struct ifnet *
2373 ifunit(const char *name)
2374 {
2375 	struct epoch_tracker et;
2376 	struct ifnet *ifp;
2377 
2378 	NET_EPOCH_ENTER(et);
2379 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2380 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2381 			break;
2382 	}
2383 	NET_EPOCH_EXIT(et);
2384 	return (ifp);
2385 }
2386 
2387 void *
2388 ifr_buffer_get_buffer(void *data)
2389 {
2390 	union ifreq_union *ifrup;
2391 
2392 	ifrup = data;
2393 #ifdef COMPAT_FREEBSD32
2394 	if (SV_CURPROC_FLAG(SV_ILP32))
2395 		return ((void *)(uintptr_t)
2396 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2397 #endif
2398 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2399 }
2400 
2401 static void
2402 ifr_buffer_set_buffer_null(void *data)
2403 {
2404 	union ifreq_union *ifrup;
2405 
2406 	ifrup = data;
2407 #ifdef COMPAT_FREEBSD32
2408 	if (SV_CURPROC_FLAG(SV_ILP32))
2409 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2410 	else
2411 #endif
2412 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2413 }
2414 
2415 size_t
2416 ifr_buffer_get_length(void *data)
2417 {
2418 	union ifreq_union *ifrup;
2419 
2420 	ifrup = data;
2421 #ifdef COMPAT_FREEBSD32
2422 	if (SV_CURPROC_FLAG(SV_ILP32))
2423 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2424 #endif
2425 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2426 }
2427 
2428 static void
2429 ifr_buffer_set_length(void *data, size_t len)
2430 {
2431 	union ifreq_union *ifrup;
2432 
2433 	ifrup = data;
2434 #ifdef COMPAT_FREEBSD32
2435 	if (SV_CURPROC_FLAG(SV_ILP32))
2436 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2437 	else
2438 #endif
2439 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2440 }
2441 
2442 void *
2443 ifr_data_get_ptr(void *ifrp)
2444 {
2445 	union ifreq_union *ifrup;
2446 
2447 	ifrup = ifrp;
2448 #ifdef COMPAT_FREEBSD32
2449 	if (SV_CURPROC_FLAG(SV_ILP32))
2450 		return ((void *)(uintptr_t)
2451 		    ifrup->ifr32.ifr_ifru.ifru_data);
2452 #endif
2453 		return (ifrup->ifr.ifr_ifru.ifru_data);
2454 }
2455 
2456 /*
2457  * Hardware specific interface ioctls.
2458  */
2459 int
2460 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2461 {
2462 	struct ifreq *ifr;
2463 	int error = 0, do_ifup = 0;
2464 	int new_flags, temp_flags;
2465 	size_t namelen, onamelen;
2466 	size_t descrlen;
2467 	char *descrbuf, *odescrbuf;
2468 	char new_name[IFNAMSIZ];
2469 	struct ifaddr *ifa;
2470 	struct sockaddr_dl *sdl;
2471 
2472 	ifr = (struct ifreq *)data;
2473 	switch (cmd) {
2474 	case SIOCGIFINDEX:
2475 		ifr->ifr_index = ifp->if_index;
2476 		break;
2477 
2478 	case SIOCGIFFLAGS:
2479 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2480 		ifr->ifr_flags = temp_flags & 0xffff;
2481 		ifr->ifr_flagshigh = temp_flags >> 16;
2482 		break;
2483 
2484 	case SIOCGIFCAP:
2485 		ifr->ifr_reqcap = ifp->if_capabilities;
2486 		ifr->ifr_curcap = ifp->if_capenable;
2487 		break;
2488 
2489 	case SIOCGIFDATA:
2490 	{
2491 		struct if_data ifd;
2492 
2493 		/* Ensure uninitialised padding is not leaked. */
2494 		memset(&ifd, 0, sizeof(ifd));
2495 
2496 		if_data_copy(ifp, &ifd);
2497 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2498 		break;
2499 	}
2500 
2501 #ifdef MAC
2502 	case SIOCGIFMAC:
2503 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2504 		break;
2505 #endif
2506 
2507 	case SIOCGIFMETRIC:
2508 		ifr->ifr_metric = ifp->if_metric;
2509 		break;
2510 
2511 	case SIOCGIFMTU:
2512 		ifr->ifr_mtu = ifp->if_mtu;
2513 		break;
2514 
2515 	case SIOCGIFPHYS:
2516 		/* XXXGL: did this ever worked? */
2517 		ifr->ifr_phys = 0;
2518 		break;
2519 
2520 	case SIOCGIFDESCR:
2521 		error = 0;
2522 		sx_slock(&ifdescr_sx);
2523 		if (ifp->if_description == NULL)
2524 			error = ENOMSG;
2525 		else {
2526 			/* space for terminating nul */
2527 			descrlen = strlen(ifp->if_description) + 1;
2528 			if (ifr_buffer_get_length(ifr) < descrlen)
2529 				ifr_buffer_set_buffer_null(ifr);
2530 			else
2531 				error = copyout(ifp->if_description,
2532 				    ifr_buffer_get_buffer(ifr), descrlen);
2533 			ifr_buffer_set_length(ifr, descrlen);
2534 		}
2535 		sx_sunlock(&ifdescr_sx);
2536 		break;
2537 
2538 	case SIOCSIFDESCR:
2539 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2540 		if (error)
2541 			return (error);
2542 
2543 		/*
2544 		 * Copy only (length-1) bytes to make sure that
2545 		 * if_description is always nul terminated.  The
2546 		 * length parameter is supposed to count the
2547 		 * terminating nul in.
2548 		 */
2549 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2550 			return (ENAMETOOLONG);
2551 		else if (ifr_buffer_get_length(ifr) == 0)
2552 			descrbuf = NULL;
2553 		else {
2554 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2555 			    M_IFDESCR, M_WAITOK | M_ZERO);
2556 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2557 			    ifr_buffer_get_length(ifr) - 1);
2558 			if (error) {
2559 				free(descrbuf, M_IFDESCR);
2560 				break;
2561 			}
2562 		}
2563 
2564 		sx_xlock(&ifdescr_sx);
2565 		odescrbuf = ifp->if_description;
2566 		ifp->if_description = descrbuf;
2567 		sx_xunlock(&ifdescr_sx);
2568 
2569 		getmicrotime(&ifp->if_lastchange);
2570 		free(odescrbuf, M_IFDESCR);
2571 		break;
2572 
2573 	case SIOCGIFFIB:
2574 		ifr->ifr_fib = ifp->if_fib;
2575 		break;
2576 
2577 	case SIOCSIFFIB:
2578 		error = priv_check(td, PRIV_NET_SETIFFIB);
2579 		if (error)
2580 			return (error);
2581 		if (ifr->ifr_fib >= rt_numfibs)
2582 			return (EINVAL);
2583 
2584 		ifp->if_fib = ifr->ifr_fib;
2585 		break;
2586 
2587 	case SIOCSIFFLAGS:
2588 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2589 		if (error)
2590 			return (error);
2591 		/*
2592 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2593 		 * check, so we don't need special handling here yet.
2594 		 */
2595 		new_flags = (ifr->ifr_flags & 0xffff) |
2596 		    (ifr->ifr_flagshigh << 16);
2597 		if (ifp->if_flags & IFF_UP &&
2598 		    (new_flags & IFF_UP) == 0) {
2599 			if_down(ifp);
2600 		} else if (new_flags & IFF_UP &&
2601 		    (ifp->if_flags & IFF_UP) == 0) {
2602 			do_ifup = 1;
2603 		}
2604 		/* See if permanently promiscuous mode bit is about to flip */
2605 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2606 			if (new_flags & IFF_PPROMISC)
2607 				ifp->if_flags |= IFF_PROMISC;
2608 			else if (ifp->if_pcount == 0)
2609 				ifp->if_flags &= ~IFF_PROMISC;
2610 			if (log_promisc_mode_change)
2611                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2612                                     ((new_flags & IFF_PPROMISC) ?
2613                                      "enabled" : "disabled"));
2614 		}
2615 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2616 			(new_flags &~ IFF_CANTCHANGE);
2617 		if (ifp->if_ioctl) {
2618 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2619 		}
2620 		if (do_ifup)
2621 			if_up(ifp);
2622 		getmicrotime(&ifp->if_lastchange);
2623 		break;
2624 
2625 	case SIOCSIFCAP:
2626 		error = priv_check(td, PRIV_NET_SETIFCAP);
2627 		if (error)
2628 			return (error);
2629 		if (ifp->if_ioctl == NULL)
2630 			return (EOPNOTSUPP);
2631 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2632 			return (EINVAL);
2633 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2634 		if (error == 0)
2635 			getmicrotime(&ifp->if_lastchange);
2636 		break;
2637 
2638 #ifdef MAC
2639 	case SIOCSIFMAC:
2640 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2641 		break;
2642 #endif
2643 
2644 	case SIOCSIFNAME:
2645 		error = priv_check(td, PRIV_NET_SETIFNAME);
2646 		if (error)
2647 			return (error);
2648 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2649 		    NULL);
2650 		if (error != 0)
2651 			return (error);
2652 		if (new_name[0] == '\0')
2653 			return (EINVAL);
2654 		if (new_name[IFNAMSIZ-1] != '\0') {
2655 			new_name[IFNAMSIZ-1] = '\0';
2656 			if (strlen(new_name) == IFNAMSIZ-1)
2657 				return (EINVAL);
2658 		}
2659 		if (strcmp(new_name, ifp->if_xname) == 0)
2660 			break;
2661 		if (ifunit(new_name) != NULL)
2662 			return (EEXIST);
2663 
2664 		/*
2665 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2666 		 * and there are numerous other races with the
2667 		 * ifunit() checks not being atomic with namespace
2668 		 * changes (renames, vmoves, if_attach, etc).
2669 		 */
2670 		ifp->if_flags |= IFF_RENAMING;
2671 
2672 		/* Announce the departure of the interface. */
2673 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2674 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2675 
2676 		if_printf(ifp, "changing name to '%s'\n", new_name);
2677 
2678 		IF_ADDR_WLOCK(ifp);
2679 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2680 		ifa = ifp->if_addr;
2681 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2682 		namelen = strlen(new_name);
2683 		onamelen = sdl->sdl_nlen;
2684 		/*
2685 		 * Move the address if needed.  This is safe because we
2686 		 * allocate space for a name of length IFNAMSIZ when we
2687 		 * create this in if_attach().
2688 		 */
2689 		if (namelen != onamelen) {
2690 			bcopy(sdl->sdl_data + onamelen,
2691 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2692 		}
2693 		bcopy(new_name, sdl->sdl_data, namelen);
2694 		sdl->sdl_nlen = namelen;
2695 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2696 		bzero(sdl->sdl_data, onamelen);
2697 		while (namelen != 0)
2698 			sdl->sdl_data[--namelen] = 0xff;
2699 		IF_ADDR_WUNLOCK(ifp);
2700 
2701 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2702 		/* Announce the return of the interface. */
2703 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2704 
2705 		ifp->if_flags &= ~IFF_RENAMING;
2706 		break;
2707 
2708 #ifdef VIMAGE
2709 	case SIOCSIFVNET:
2710 		error = priv_check(td, PRIV_NET_SETIFVNET);
2711 		if (error)
2712 			return (error);
2713 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2714 		break;
2715 #endif
2716 
2717 	case SIOCSIFMETRIC:
2718 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2719 		if (error)
2720 			return (error);
2721 		ifp->if_metric = ifr->ifr_metric;
2722 		getmicrotime(&ifp->if_lastchange);
2723 		break;
2724 
2725 	case SIOCSIFPHYS:
2726 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2727 		if (error)
2728 			return (error);
2729 		if (ifp->if_ioctl == NULL)
2730 			return (EOPNOTSUPP);
2731 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2732 		if (error == 0)
2733 			getmicrotime(&ifp->if_lastchange);
2734 		break;
2735 
2736 	case SIOCSIFMTU:
2737 	{
2738 		u_long oldmtu = ifp->if_mtu;
2739 
2740 		error = priv_check(td, PRIV_NET_SETIFMTU);
2741 		if (error)
2742 			return (error);
2743 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2744 			return (EINVAL);
2745 		if (ifp->if_ioctl == NULL)
2746 			return (EOPNOTSUPP);
2747 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2748 		if (error == 0) {
2749 			getmicrotime(&ifp->if_lastchange);
2750 			rt_ifmsg(ifp);
2751 #ifdef INET
2752 			DEBUGNET_NOTIFY_MTU(ifp);
2753 #endif
2754 		}
2755 		/*
2756 		 * If the link MTU changed, do network layer specific procedure.
2757 		 */
2758 		if (ifp->if_mtu != oldmtu) {
2759 #ifdef INET6
2760 			nd6_setmtu(ifp);
2761 #endif
2762 			rt_updatemtu(ifp);
2763 		}
2764 		break;
2765 	}
2766 
2767 	case SIOCADDMULTI:
2768 	case SIOCDELMULTI:
2769 		if (cmd == SIOCADDMULTI)
2770 			error = priv_check(td, PRIV_NET_ADDMULTI);
2771 		else
2772 			error = priv_check(td, PRIV_NET_DELMULTI);
2773 		if (error)
2774 			return (error);
2775 
2776 		/* Don't allow group membership on non-multicast interfaces. */
2777 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2778 			return (EOPNOTSUPP);
2779 
2780 		/* Don't let users screw up protocols' entries. */
2781 		if (ifr->ifr_addr.sa_family != AF_LINK)
2782 			return (EINVAL);
2783 
2784 		if (cmd == SIOCADDMULTI) {
2785 			struct epoch_tracker et;
2786 			struct ifmultiaddr *ifma;
2787 
2788 			/*
2789 			 * Userland is only permitted to join groups once
2790 			 * via the if_addmulti() KPI, because it cannot hold
2791 			 * struct ifmultiaddr * between calls. It may also
2792 			 * lose a race while we check if the membership
2793 			 * already exists.
2794 			 */
2795 			NET_EPOCH_ENTER(et);
2796 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2797 			NET_EPOCH_EXIT(et);
2798 			if (ifma != NULL)
2799 				error = EADDRINUSE;
2800 			else
2801 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2802 		} else {
2803 			error = if_delmulti(ifp, &ifr->ifr_addr);
2804 		}
2805 		if (error == 0)
2806 			getmicrotime(&ifp->if_lastchange);
2807 		break;
2808 
2809 	case SIOCSIFPHYADDR:
2810 	case SIOCDIFPHYADDR:
2811 #ifdef INET6
2812 	case SIOCSIFPHYADDR_IN6:
2813 #endif
2814 	case SIOCSIFMEDIA:
2815 	case SIOCSIFGENERIC:
2816 		error = priv_check(td, PRIV_NET_HWIOCTL);
2817 		if (error)
2818 			return (error);
2819 		if (ifp->if_ioctl == NULL)
2820 			return (EOPNOTSUPP);
2821 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2822 		if (error == 0)
2823 			getmicrotime(&ifp->if_lastchange);
2824 		break;
2825 
2826 	case SIOCGIFSTATUS:
2827 	case SIOCGIFPSRCADDR:
2828 	case SIOCGIFPDSTADDR:
2829 	case SIOCGIFMEDIA:
2830 	case SIOCGIFXMEDIA:
2831 	case SIOCGIFGENERIC:
2832 	case SIOCGIFRSSKEY:
2833 	case SIOCGIFRSSHASH:
2834 	case SIOCGIFDOWNREASON:
2835 		if (ifp->if_ioctl == NULL)
2836 			return (EOPNOTSUPP);
2837 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2838 		break;
2839 
2840 	case SIOCSIFLLADDR:
2841 		error = priv_check(td, PRIV_NET_SETLLADDR);
2842 		if (error)
2843 			return (error);
2844 		error = if_setlladdr(ifp,
2845 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2846 		break;
2847 
2848 	case SIOCGHWADDR:
2849 		error = if_gethwaddr(ifp, ifr);
2850 		break;
2851 
2852 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2853 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2854 		if (error)
2855 			return (error);
2856 		if ((error = if_addgroup(ifp,
2857 		    ifgr_group_get((struct ifgroupreq *)data))))
2858 			return (error);
2859 		break;
2860 
2861 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2862 	{
2863 		struct epoch_tracker et;
2864 
2865 		NET_EPOCH_ENTER(et);
2866 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2867 		NET_EPOCH_EXIT(et);
2868 		break;
2869 	}
2870 
2871 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2872 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2873 		if (error)
2874 			return (error);
2875 		if ((error = if_delgroup(ifp,
2876 		    ifgr_group_get((struct ifgroupreq *)data))))
2877 			return (error);
2878 		break;
2879 
2880 	default:
2881 		error = ENOIOCTL;
2882 		break;
2883 	}
2884 	return (error);
2885 }
2886 
2887 #ifdef COMPAT_FREEBSD32
2888 struct ifconf32 {
2889 	int32_t	ifc_len;
2890 	union {
2891 		uint32_t	ifcu_buf;
2892 		uint32_t	ifcu_req;
2893 	} ifc_ifcu;
2894 };
2895 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2896 #endif
2897 
2898 #ifdef COMPAT_FREEBSD32
2899 static void
2900 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2901 {
2902 	struct ifmediareq32 *ifmr32;
2903 
2904 	ifmr32 = (struct ifmediareq32 *)data;
2905 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2906 	    sizeof(ifmr->ifm_name));
2907 	ifmr->ifm_current = ifmr32->ifm_current;
2908 	ifmr->ifm_mask = ifmr32->ifm_mask;
2909 	ifmr->ifm_status = ifmr32->ifm_status;
2910 	ifmr->ifm_active = ifmr32->ifm_active;
2911 	ifmr->ifm_count = ifmr32->ifm_count;
2912 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2913 }
2914 
2915 static void
2916 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2917 {
2918 	struct ifmediareq32 *ifmr32;
2919 
2920 	ifmr32 = (struct ifmediareq32 *)data;
2921 	ifmr32->ifm_current = ifmr->ifm_current;
2922 	ifmr32->ifm_mask = ifmr->ifm_mask;
2923 	ifmr32->ifm_status = ifmr->ifm_status;
2924 	ifmr32->ifm_active = ifmr->ifm_active;
2925 	ifmr32->ifm_count = ifmr->ifm_count;
2926 }
2927 #endif
2928 
2929 /*
2930  * Interface ioctls.
2931  */
2932 int
2933 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2934 {
2935 #ifdef COMPAT_FREEBSD32
2936 	caddr_t saved_data = NULL;
2937 	struct ifmediareq ifmr;
2938 	struct ifmediareq *ifmrp = NULL;
2939 #endif
2940 	struct ifnet *ifp;
2941 	struct ifreq *ifr;
2942 	int error;
2943 	int oif_flags;
2944 #ifdef VIMAGE
2945 	bool shutdown;
2946 #endif
2947 
2948 	CURVNET_SET(so->so_vnet);
2949 #ifdef VIMAGE
2950 	/* Make sure the VNET is stable. */
2951 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2952 	if (shutdown) {
2953 		CURVNET_RESTORE();
2954 		return (EBUSY);
2955 	}
2956 #endif
2957 
2958 	switch (cmd) {
2959 	case SIOCGIFCONF:
2960 		error = ifconf(cmd, data);
2961 		goto out_noref;
2962 
2963 #ifdef COMPAT_FREEBSD32
2964 	case SIOCGIFCONF32:
2965 		{
2966 			struct ifconf32 *ifc32;
2967 			struct ifconf ifc;
2968 
2969 			ifc32 = (struct ifconf32 *)data;
2970 			ifc.ifc_len = ifc32->ifc_len;
2971 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2972 
2973 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2974 			if (error == 0)
2975 				ifc32->ifc_len = ifc.ifc_len;
2976 			goto out_noref;
2977 		}
2978 #endif
2979 	}
2980 
2981 #ifdef COMPAT_FREEBSD32
2982 	switch (cmd) {
2983 	case SIOCGIFMEDIA32:
2984 	case SIOCGIFXMEDIA32:
2985 		ifmrp = &ifmr;
2986 		ifmr_init(ifmrp, data);
2987 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2988 		saved_data = data;
2989 		data = (caddr_t)ifmrp;
2990 	}
2991 #endif
2992 
2993 	ifr = (struct ifreq *)data;
2994 	switch (cmd) {
2995 #ifdef VIMAGE
2996 	case SIOCSIFRVNET:
2997 		error = priv_check(td, PRIV_NET_SETIFVNET);
2998 		if (error == 0)
2999 			error = if_vmove_reclaim(td, ifr->ifr_name,
3000 			    ifr->ifr_jid);
3001 		goto out_noref;
3002 #endif
3003 	case SIOCIFCREATE:
3004 	case SIOCIFCREATE2:
3005 		error = priv_check(td, PRIV_NET_IFCREATE);
3006 		if (error == 0)
3007 			error = if_clone_create(ifr->ifr_name,
3008 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3009 			    ifr_data_get_ptr(ifr) : NULL);
3010 		goto out_noref;
3011 	case SIOCIFDESTROY:
3012 		error = priv_check(td, PRIV_NET_IFDESTROY);
3013 		if (error == 0)
3014 			error = if_clone_destroy(ifr->ifr_name);
3015 		goto out_noref;
3016 
3017 	case SIOCIFGCLONERS:
3018 		error = if_clone_list((struct if_clonereq *)data);
3019 		goto out_noref;
3020 
3021 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3022 		error = if_getgroupmembers((struct ifgroupreq *)data);
3023 		goto out_noref;
3024 
3025 #if defined(INET) || defined(INET6)
3026 	case SIOCSVH:
3027 	case SIOCGVH:
3028 		if (carp_ioctl_p == NULL)
3029 			error = EPROTONOSUPPORT;
3030 		else
3031 			error = (*carp_ioctl_p)(ifr, cmd, td);
3032 		goto out_noref;
3033 #endif
3034 	}
3035 
3036 	ifp = ifunit_ref(ifr->ifr_name);
3037 	if (ifp == NULL) {
3038 		error = ENXIO;
3039 		goto out_noref;
3040 	}
3041 
3042 	error = ifhwioctl(cmd, ifp, data, td);
3043 	if (error != ENOIOCTL)
3044 		goto out_ref;
3045 
3046 	oif_flags = ifp->if_flags;
3047 	if (so->so_proto == NULL) {
3048 		error = EOPNOTSUPP;
3049 		goto out_ref;
3050 	}
3051 
3052 	/*
3053 	 * Pass the request on to the socket control method, and if the
3054 	 * latter returns EOPNOTSUPP, directly to the interface.
3055 	 *
3056 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3057 	 * trust SIOCSIFADDR et al to come from an already privileged
3058 	 * layer, and do not perform any credentials checks or input
3059 	 * validation.
3060 	 */
3061 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3062 	    ifp, td));
3063 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3064 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3065 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3066 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3067 
3068 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3069 #ifdef INET6
3070 		if (ifp->if_flags & IFF_UP)
3071 			in6_if_up(ifp);
3072 #endif
3073 	}
3074 
3075 out_ref:
3076 	if_rele(ifp);
3077 out_noref:
3078 #ifdef COMPAT_FREEBSD32
3079 	if (ifmrp != NULL) {
3080 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3081 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3082 		     cmd));
3083 		data = saved_data;
3084 		ifmr_update(ifmrp, data);
3085 	}
3086 #endif
3087 	CURVNET_RESTORE();
3088 	return (error);
3089 }
3090 
3091 /*
3092  * The code common to handling reference counted flags,
3093  * e.g., in ifpromisc() and if_allmulti().
3094  * The "pflag" argument can specify a permanent mode flag to check,
3095  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3096  *
3097  * Only to be used on stack-owned flags, not driver-owned flags.
3098  */
3099 static int
3100 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3101 {
3102 	struct ifreq ifr;
3103 	int error;
3104 	int oldflags, oldcount;
3105 
3106 	/* Sanity checks to catch programming errors */
3107 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3108 	    ("%s: setting driver-owned flag %d", __func__, flag));
3109 
3110 	if (onswitch)
3111 		KASSERT(*refcount >= 0,
3112 		    ("%s: increment negative refcount %d for flag %d",
3113 		    __func__, *refcount, flag));
3114 	else
3115 		KASSERT(*refcount > 0,
3116 		    ("%s: decrement non-positive refcount %d for flag %d",
3117 		    __func__, *refcount, flag));
3118 
3119 	/* In case this mode is permanent, just touch refcount */
3120 	if (ifp->if_flags & pflag) {
3121 		*refcount += onswitch ? 1 : -1;
3122 		return (0);
3123 	}
3124 
3125 	/* Save ifnet parameters for if_ioctl() may fail */
3126 	oldcount = *refcount;
3127 	oldflags = ifp->if_flags;
3128 
3129 	/*
3130 	 * See if we aren't the only and touching refcount is enough.
3131 	 * Actually toggle interface flag if we are the first or last.
3132 	 */
3133 	if (onswitch) {
3134 		if ((*refcount)++)
3135 			return (0);
3136 		ifp->if_flags |= flag;
3137 	} else {
3138 		if (--(*refcount))
3139 			return (0);
3140 		ifp->if_flags &= ~flag;
3141 	}
3142 
3143 	/* Call down the driver since we've changed interface flags */
3144 	if (ifp->if_ioctl == NULL) {
3145 		error = EOPNOTSUPP;
3146 		goto recover;
3147 	}
3148 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3149 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3150 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3151 	if (error)
3152 		goto recover;
3153 	/* Notify userland that interface flags have changed */
3154 	rt_ifmsg(ifp);
3155 	return (0);
3156 
3157 recover:
3158 	/* Recover after driver error */
3159 	*refcount = oldcount;
3160 	ifp->if_flags = oldflags;
3161 	return (error);
3162 }
3163 
3164 /*
3165  * Set/clear promiscuous mode on interface ifp based on the truth value
3166  * of pswitch.  The calls are reference counted so that only the first
3167  * "on" request actually has an effect, as does the final "off" request.
3168  * Results are undefined if the "off" and "on" requests are not matched.
3169  */
3170 int
3171 ifpromisc(struct ifnet *ifp, int pswitch)
3172 {
3173 	int error;
3174 	int oldflags = ifp->if_flags;
3175 
3176 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3177 			   &ifp->if_pcount, pswitch);
3178 	/* If promiscuous mode status has changed, log a message */
3179 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3180             log_promisc_mode_change)
3181 		if_printf(ifp, "promiscuous mode %s\n",
3182 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3183 	return (error);
3184 }
3185 
3186 /*
3187  * Return interface configuration
3188  * of system.  List may be used
3189  * in later ioctl's (above) to get
3190  * other information.
3191  */
3192 /*ARGSUSED*/
3193 static int
3194 ifconf(u_long cmd, caddr_t data)
3195 {
3196 	struct ifconf *ifc = (struct ifconf *)data;
3197 	struct ifnet *ifp;
3198 	struct ifaddr *ifa;
3199 	struct ifreq ifr;
3200 	struct sbuf *sb;
3201 	int error, full = 0, valid_len, max_len;
3202 
3203 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3204 	max_len = maxphys - 1;
3205 
3206 	/* Prevent hostile input from being able to crash the system */
3207 	if (ifc->ifc_len <= 0)
3208 		return (EINVAL);
3209 
3210 again:
3211 	if (ifc->ifc_len <= max_len) {
3212 		max_len = ifc->ifc_len;
3213 		full = 1;
3214 	}
3215 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3216 	max_len = 0;
3217 	valid_len = 0;
3218 
3219 	IFNET_RLOCK();
3220 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3221 		struct epoch_tracker et;
3222 		int addrs;
3223 
3224 		/*
3225 		 * Zero the ifr to make sure we don't disclose the contents
3226 		 * of the stack.
3227 		 */
3228 		memset(&ifr, 0, sizeof(ifr));
3229 
3230 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3231 		    >= sizeof(ifr.ifr_name)) {
3232 			sbuf_delete(sb);
3233 			IFNET_RUNLOCK();
3234 			return (ENAMETOOLONG);
3235 		}
3236 
3237 		addrs = 0;
3238 		NET_EPOCH_ENTER(et);
3239 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3240 			struct sockaddr *sa = ifa->ifa_addr;
3241 
3242 			if (prison_if(curthread->td_ucred, sa) != 0)
3243 				continue;
3244 			addrs++;
3245 			if (sa->sa_len <= sizeof(*sa)) {
3246 				if (sa->sa_len < sizeof(*sa)) {
3247 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3248 					    sizeof(ifr.ifr_ifru.ifru_addr));
3249 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3250 					    sa->sa_len);
3251 				} else
3252 					ifr.ifr_ifru.ifru_addr = *sa;
3253 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3254 				max_len += sizeof(ifr);
3255 			} else {
3256 				sbuf_bcat(sb, &ifr,
3257 				    offsetof(struct ifreq, ifr_addr));
3258 				max_len += offsetof(struct ifreq, ifr_addr);
3259 				sbuf_bcat(sb, sa, sa->sa_len);
3260 				max_len += sa->sa_len;
3261 			}
3262 
3263 			if (sbuf_error(sb) == 0)
3264 				valid_len = sbuf_len(sb);
3265 		}
3266 		NET_EPOCH_EXIT(et);
3267 		if (addrs == 0) {
3268 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3269 			max_len += sizeof(ifr);
3270 
3271 			if (sbuf_error(sb) == 0)
3272 				valid_len = sbuf_len(sb);
3273 		}
3274 	}
3275 	IFNET_RUNLOCK();
3276 
3277 	/*
3278 	 * If we didn't allocate enough space (uncommon), try again.  If
3279 	 * we have already allocated as much space as we are allowed,
3280 	 * return what we've got.
3281 	 */
3282 	if (valid_len != max_len && !full) {
3283 		sbuf_delete(sb);
3284 		goto again;
3285 	}
3286 
3287 	ifc->ifc_len = valid_len;
3288 	sbuf_finish(sb);
3289 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3290 	sbuf_delete(sb);
3291 	return (error);
3292 }
3293 
3294 /*
3295  * Just like ifpromisc(), but for all-multicast-reception mode.
3296  */
3297 int
3298 if_allmulti(struct ifnet *ifp, int onswitch)
3299 {
3300 
3301 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3302 }
3303 
3304 struct ifmultiaddr *
3305 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3306 {
3307 	struct ifmultiaddr *ifma;
3308 
3309 	IF_ADDR_LOCK_ASSERT(ifp);
3310 
3311 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3312 		if (sa->sa_family == AF_LINK) {
3313 			if (sa_dl_equal(ifma->ifma_addr, sa))
3314 				break;
3315 		} else {
3316 			if (sa_equal(ifma->ifma_addr, sa))
3317 				break;
3318 		}
3319 	}
3320 
3321 	return ifma;
3322 }
3323 
3324 /*
3325  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3326  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3327  * the ifnet multicast address list here, so the caller must do that and
3328  * other setup work (such as notifying the device driver).  The reference
3329  * count is initialized to 1.
3330  */
3331 static struct ifmultiaddr *
3332 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3333     int mflags)
3334 {
3335 	struct ifmultiaddr *ifma;
3336 	struct sockaddr *dupsa;
3337 
3338 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3339 	    M_ZERO);
3340 	if (ifma == NULL)
3341 		return (NULL);
3342 
3343 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3344 	if (dupsa == NULL) {
3345 		free(ifma, M_IFMADDR);
3346 		return (NULL);
3347 	}
3348 	bcopy(sa, dupsa, sa->sa_len);
3349 	ifma->ifma_addr = dupsa;
3350 
3351 	ifma->ifma_ifp = ifp;
3352 	ifma->ifma_refcount = 1;
3353 	ifma->ifma_protospec = NULL;
3354 
3355 	if (llsa == NULL) {
3356 		ifma->ifma_lladdr = NULL;
3357 		return (ifma);
3358 	}
3359 
3360 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3361 	if (dupsa == NULL) {
3362 		free(ifma->ifma_addr, M_IFMADDR);
3363 		free(ifma, M_IFMADDR);
3364 		return (NULL);
3365 	}
3366 	bcopy(llsa, dupsa, llsa->sa_len);
3367 	ifma->ifma_lladdr = dupsa;
3368 
3369 	return (ifma);
3370 }
3371 
3372 /*
3373  * if_freemulti: free ifmultiaddr structure and possibly attached related
3374  * addresses.  The caller is responsible for implementing reference
3375  * counting, notifying the driver, handling routing messages, and releasing
3376  * any dependent link layer state.
3377  */
3378 #ifdef MCAST_VERBOSE
3379 extern void kdb_backtrace(void);
3380 #endif
3381 static void
3382 if_freemulti_internal(struct ifmultiaddr *ifma)
3383 {
3384 
3385 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3386 	    ifma->ifma_refcount));
3387 
3388 	if (ifma->ifma_lladdr != NULL)
3389 		free(ifma->ifma_lladdr, M_IFMADDR);
3390 #ifdef MCAST_VERBOSE
3391 	kdb_backtrace();
3392 	printf("%s freeing ifma: %p\n", __func__, ifma);
3393 #endif
3394 	free(ifma->ifma_addr, M_IFMADDR);
3395 	free(ifma, M_IFMADDR);
3396 }
3397 
3398 static void
3399 if_destroymulti(epoch_context_t ctx)
3400 {
3401 	struct ifmultiaddr *ifma;
3402 
3403 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3404 	if_freemulti_internal(ifma);
3405 }
3406 
3407 void
3408 if_freemulti(struct ifmultiaddr *ifma)
3409 {
3410 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3411 	    ifma->ifma_refcount));
3412 
3413 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3414 }
3415 
3416 /*
3417  * Register an additional multicast address with a network interface.
3418  *
3419  * - If the address is already present, bump the reference count on the
3420  *   address and return.
3421  * - If the address is not link-layer, look up a link layer address.
3422  * - Allocate address structures for one or both addresses, and attach to the
3423  *   multicast address list on the interface.  If automatically adding a link
3424  *   layer address, the protocol address will own a reference to the link
3425  *   layer address, to be freed when it is freed.
3426  * - Notify the network device driver of an addition to the multicast address
3427  *   list.
3428  *
3429  * 'sa' points to caller-owned memory with the desired multicast address.
3430  *
3431  * 'retifma' will be used to return a pointer to the resulting multicast
3432  * address reference, if desired.
3433  */
3434 int
3435 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3436     struct ifmultiaddr **retifma)
3437 {
3438 	struct ifmultiaddr *ifma, *ll_ifma;
3439 	struct sockaddr *llsa;
3440 	struct sockaddr_dl sdl;
3441 	int error;
3442 
3443 #ifdef INET
3444 	IN_MULTI_LIST_UNLOCK_ASSERT();
3445 #endif
3446 #ifdef INET6
3447 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3448 #endif
3449 	/*
3450 	 * If the address is already present, return a new reference to it;
3451 	 * otherwise, allocate storage and set up a new address.
3452 	 */
3453 	IF_ADDR_WLOCK(ifp);
3454 	ifma = if_findmulti(ifp, sa);
3455 	if (ifma != NULL) {
3456 		ifma->ifma_refcount++;
3457 		if (retifma != NULL)
3458 			*retifma = ifma;
3459 		IF_ADDR_WUNLOCK(ifp);
3460 		return (0);
3461 	}
3462 
3463 	/*
3464 	 * The address isn't already present; resolve the protocol address
3465 	 * into a link layer address, and then look that up, bump its
3466 	 * refcount or allocate an ifma for that also.
3467 	 * Most link layer resolving functions returns address data which
3468 	 * fits inside default sockaddr_dl structure. However callback
3469 	 * can allocate another sockaddr structure, in that case we need to
3470 	 * free it later.
3471 	 */
3472 	llsa = NULL;
3473 	ll_ifma = NULL;
3474 	if (ifp->if_resolvemulti != NULL) {
3475 		/* Provide called function with buffer size information */
3476 		sdl.sdl_len = sizeof(sdl);
3477 		llsa = (struct sockaddr *)&sdl;
3478 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3479 		if (error)
3480 			goto unlock_out;
3481 	}
3482 
3483 	/*
3484 	 * Allocate the new address.  Don't hook it up yet, as we may also
3485 	 * need to allocate a link layer multicast address.
3486 	 */
3487 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3488 	if (ifma == NULL) {
3489 		error = ENOMEM;
3490 		goto free_llsa_out;
3491 	}
3492 
3493 	/*
3494 	 * If a link layer address is found, we'll need to see if it's
3495 	 * already present in the address list, or allocate is as well.
3496 	 * When this block finishes, the link layer address will be on the
3497 	 * list.
3498 	 */
3499 	if (llsa != NULL) {
3500 		ll_ifma = if_findmulti(ifp, llsa);
3501 		if (ll_ifma == NULL) {
3502 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3503 			if (ll_ifma == NULL) {
3504 				--ifma->ifma_refcount;
3505 				if_freemulti(ifma);
3506 				error = ENOMEM;
3507 				goto free_llsa_out;
3508 			}
3509 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3510 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3511 			    ifma_link);
3512 		} else
3513 			ll_ifma->ifma_refcount++;
3514 		ifma->ifma_llifma = ll_ifma;
3515 	}
3516 
3517 	/*
3518 	 * We now have a new multicast address, ifma, and possibly a new or
3519 	 * referenced link layer address.  Add the primary address to the
3520 	 * ifnet address list.
3521 	 */
3522 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3523 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3524 
3525 	if (retifma != NULL)
3526 		*retifma = ifma;
3527 
3528 	/*
3529 	 * Must generate the message while holding the lock so that 'ifma'
3530 	 * pointer is still valid.
3531 	 */
3532 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3533 	IF_ADDR_WUNLOCK(ifp);
3534 
3535 	/*
3536 	 * We are certain we have added something, so call down to the
3537 	 * interface to let them know about it.
3538 	 */
3539 	if (ifp->if_ioctl != NULL) {
3540 		if (THREAD_CAN_SLEEP())
3541 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3542 		else
3543 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3544 	}
3545 
3546 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3547 		link_free_sdl(llsa);
3548 
3549 	return (0);
3550 
3551 free_llsa_out:
3552 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3553 		link_free_sdl(llsa);
3554 
3555 unlock_out:
3556 	IF_ADDR_WUNLOCK(ifp);
3557 	return (error);
3558 }
3559 
3560 static void
3561 if_siocaddmulti(void *arg, int pending)
3562 {
3563 	struct ifnet *ifp;
3564 
3565 	ifp = arg;
3566 #ifdef DIAGNOSTIC
3567 	if (pending > 1)
3568 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3569 #endif
3570 	CURVNET_SET(ifp->if_vnet);
3571 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3572 	CURVNET_RESTORE();
3573 }
3574 
3575 /*
3576  * Delete a multicast group membership by network-layer group address.
3577  *
3578  * Returns ENOENT if the entry could not be found. If ifp no longer
3579  * exists, results are undefined. This entry point should only be used
3580  * from subsystems which do appropriate locking to hold ifp for the
3581  * duration of the call.
3582  * Network-layer protocol domains must use if_delmulti_ifma().
3583  */
3584 int
3585 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3586 {
3587 	struct ifmultiaddr *ifma;
3588 	int lastref;
3589 
3590 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3591 
3592 	IF_ADDR_WLOCK(ifp);
3593 	lastref = 0;
3594 	ifma = if_findmulti(ifp, sa);
3595 	if (ifma != NULL)
3596 		lastref = if_delmulti_locked(ifp, ifma, 0);
3597 	IF_ADDR_WUNLOCK(ifp);
3598 
3599 	if (ifma == NULL)
3600 		return (ENOENT);
3601 
3602 	if (lastref && ifp->if_ioctl != NULL) {
3603 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3604 	}
3605 
3606 	return (0);
3607 }
3608 
3609 /*
3610  * Delete all multicast group membership for an interface.
3611  * Should be used to quickly flush all multicast filters.
3612  */
3613 void
3614 if_delallmulti(struct ifnet *ifp)
3615 {
3616 	struct ifmultiaddr *ifma;
3617 	struct ifmultiaddr *next;
3618 
3619 	IF_ADDR_WLOCK(ifp);
3620 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3621 		if_delmulti_locked(ifp, ifma, 0);
3622 	IF_ADDR_WUNLOCK(ifp);
3623 }
3624 
3625 void
3626 if_delmulti_ifma(struct ifmultiaddr *ifma)
3627 {
3628 	if_delmulti_ifma_flags(ifma, 0);
3629 }
3630 
3631 /*
3632  * Delete a multicast group membership by group membership pointer.
3633  * Network-layer protocol domains must use this routine.
3634  *
3635  * It is safe to call this routine if the ifp disappeared.
3636  */
3637 void
3638 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3639 {
3640 	struct ifnet *ifp;
3641 	int lastref;
3642 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3643 #ifdef INET
3644 	IN_MULTI_LIST_UNLOCK_ASSERT();
3645 #endif
3646 	ifp = ifma->ifma_ifp;
3647 #ifdef DIAGNOSTIC
3648 	if (ifp == NULL) {
3649 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3650 	} else {
3651 		struct epoch_tracker et;
3652 		struct ifnet *oifp;
3653 
3654 		NET_EPOCH_ENTER(et);
3655 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3656 			if (ifp == oifp)
3657 				break;
3658 		NET_EPOCH_EXIT(et);
3659 		if (ifp != oifp)
3660 			ifp = NULL;
3661 	}
3662 #endif
3663 	/*
3664 	 * If and only if the ifnet instance exists: Acquire the address lock.
3665 	 */
3666 	if (ifp != NULL)
3667 		IF_ADDR_WLOCK(ifp);
3668 
3669 	lastref = if_delmulti_locked(ifp, ifma, flags);
3670 
3671 	if (ifp != NULL) {
3672 		/*
3673 		 * If and only if the ifnet instance exists:
3674 		 *  Release the address lock.
3675 		 *  If the group was left: update the hardware hash filter.
3676 		 */
3677 		IF_ADDR_WUNLOCK(ifp);
3678 		if (lastref && ifp->if_ioctl != NULL) {
3679 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3680 		}
3681 	}
3682 }
3683 
3684 /*
3685  * Perform deletion of network-layer and/or link-layer multicast address.
3686  *
3687  * Return 0 if the reference count was decremented.
3688  * Return 1 if the final reference was released, indicating that the
3689  * hardware hash filter should be reprogrammed.
3690  */
3691 static int
3692 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3693 {
3694 	struct ifmultiaddr *ll_ifma;
3695 
3696 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3697 		KASSERT(ifma->ifma_ifp == ifp,
3698 		    ("%s: inconsistent ifp %p", __func__, ifp));
3699 		IF_ADDR_WLOCK_ASSERT(ifp);
3700 	}
3701 
3702 	ifp = ifma->ifma_ifp;
3703 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3704 
3705 	/*
3706 	 * If the ifnet is detaching, null out references to ifnet,
3707 	 * so that upper protocol layers will notice, and not attempt
3708 	 * to obtain locks for an ifnet which no longer exists. The
3709 	 * routing socket announcement must happen before the ifnet
3710 	 * instance is detached from the system.
3711 	 */
3712 	if (detaching) {
3713 #ifdef DIAGNOSTIC
3714 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3715 #endif
3716 		/*
3717 		 * ifp may already be nulled out if we are being reentered
3718 		 * to delete the ll_ifma.
3719 		 */
3720 		if (ifp != NULL) {
3721 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3722 			ifma->ifma_ifp = NULL;
3723 		}
3724 	}
3725 
3726 	if (--ifma->ifma_refcount > 0)
3727 		return 0;
3728 
3729 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3730 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3731 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3732 	}
3733 	/*
3734 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3735 	 * have been associated with it. Release it first if so.
3736 	 */
3737 	ll_ifma = ifma->ifma_llifma;
3738 	if (ll_ifma != NULL) {
3739 		KASSERT(ifma->ifma_lladdr != NULL,
3740 		    ("%s: llifma w/o lladdr", __func__));
3741 		if (detaching)
3742 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3743 		if (--ll_ifma->ifma_refcount == 0) {
3744 			if (ifp != NULL) {
3745 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3746 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3747 						ifma_link);
3748 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3749 				}
3750 			}
3751 			if_freemulti(ll_ifma);
3752 		}
3753 	}
3754 #ifdef INVARIANTS
3755 	if (ifp) {
3756 		struct ifmultiaddr *ifmatmp;
3757 
3758 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3759 			MPASS(ifma != ifmatmp);
3760 	}
3761 #endif
3762 	if_freemulti(ifma);
3763 	/*
3764 	 * The last reference to this instance of struct ifmultiaddr
3765 	 * was released; the hardware should be notified of this change.
3766 	 */
3767 	return 1;
3768 }
3769 
3770 /*
3771  * Set the link layer address on an interface.
3772  *
3773  * At this time we only support certain types of interfaces,
3774  * and we don't allow the length of the address to change.
3775  *
3776  * Set noinline to be dtrace-friendly
3777  */
3778 __noinline int
3779 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3780 {
3781 	struct sockaddr_dl *sdl;
3782 	struct ifaddr *ifa;
3783 	struct ifreq ifr;
3784 
3785 	ifa = ifp->if_addr;
3786 	if (ifa == NULL)
3787 		return (EINVAL);
3788 
3789 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3790 	if (sdl == NULL)
3791 		return (EINVAL);
3792 
3793 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3794 		return (EINVAL);
3795 
3796 	switch (ifp->if_type) {
3797 	case IFT_ETHER:
3798 	case IFT_XETHER:
3799 	case IFT_L2VLAN:
3800 	case IFT_BRIDGE:
3801 	case IFT_IEEE8023ADLAG:
3802 		bcopy(lladdr, LLADDR(sdl), len);
3803 		break;
3804 	default:
3805 		return (ENODEV);
3806 	}
3807 
3808 	/*
3809 	 * If the interface is already up, we need
3810 	 * to re-init it in order to reprogram its
3811 	 * address filter.
3812 	 */
3813 	if ((ifp->if_flags & IFF_UP) != 0) {
3814 		if (ifp->if_ioctl) {
3815 			ifp->if_flags &= ~IFF_UP;
3816 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3817 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3818 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3819 			ifp->if_flags |= IFF_UP;
3820 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3821 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3822 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3823 		}
3824 	}
3825 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3826 
3827 	return (0);
3828 }
3829 
3830 /*
3831  * Compat function for handling basic encapsulation requests.
3832  * Not converted stacks (FDDI, IB, ..) supports traditional
3833  * output model: ARP (and other similar L2 protocols) are handled
3834  * inside output routine, arpresolve/nd6_resolve() returns MAC
3835  * address instead of full prepend.
3836  *
3837  * This function creates calculated header==MAC for IPv4/IPv6 and
3838  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3839  * address families.
3840  */
3841 static int
3842 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3843 {
3844 
3845 	if (req->rtype != IFENCAP_LL)
3846 		return (EOPNOTSUPP);
3847 
3848 	if (req->bufsize < req->lladdr_len)
3849 		return (ENOMEM);
3850 
3851 	switch (req->family) {
3852 	case AF_INET:
3853 	case AF_INET6:
3854 		break;
3855 	default:
3856 		return (EAFNOSUPPORT);
3857 	}
3858 
3859 	/* Copy lladdr to storage as is */
3860 	memmove(req->buf, req->lladdr, req->lladdr_len);
3861 	req->bufsize = req->lladdr_len;
3862 	req->lladdr_off = 0;
3863 
3864 	return (0);
3865 }
3866 
3867 /*
3868  * Tunnel interfaces can nest, also they may cause infinite recursion
3869  * calls when misconfigured. We'll prevent this by detecting loops.
3870  * High nesting level may cause stack exhaustion. We'll prevent this
3871  * by introducing upper limit.
3872  *
3873  * Return 0, if tunnel nesting count is equal or less than limit.
3874  */
3875 int
3876 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3877     int limit)
3878 {
3879 	struct m_tag *mtag;
3880 	int count;
3881 
3882 	count = 1;
3883 	mtag = NULL;
3884 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3885 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3886 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3887 			return (EIO);
3888 		}
3889 		count++;
3890 	}
3891 	if (count > limit) {
3892 		log(LOG_NOTICE,
3893 		    "%s: if_output recursively called too many times(%d)\n",
3894 		    if_name(ifp), count);
3895 		return (EIO);
3896 	}
3897 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3898 	if (mtag == NULL)
3899 		return (ENOMEM);
3900 	*(struct ifnet **)(mtag + 1) = ifp;
3901 	m_tag_prepend(m, mtag);
3902 	return (0);
3903 }
3904 
3905 /*
3906  * Get the link layer address that was read from the hardware at attach.
3907  *
3908  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3909  * their component interfaces as IFT_IEEE8023ADLAG.
3910  */
3911 int
3912 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3913 {
3914 
3915 	if (ifp->if_hw_addr == NULL)
3916 		return (ENODEV);
3917 
3918 	switch (ifp->if_type) {
3919 	case IFT_ETHER:
3920 	case IFT_IEEE8023ADLAG:
3921 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3922 		return (0);
3923 	default:
3924 		return (ENODEV);
3925 	}
3926 }
3927 
3928 /*
3929  * The name argument must be a pointer to storage which will last as
3930  * long as the interface does.  For physical devices, the result of
3931  * device_get_name(dev) is a good choice and for pseudo-devices a
3932  * static string works well.
3933  */
3934 void
3935 if_initname(struct ifnet *ifp, const char *name, int unit)
3936 {
3937 	ifp->if_dname = name;
3938 	ifp->if_dunit = unit;
3939 	if (unit != IF_DUNIT_NONE)
3940 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3941 	else
3942 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3943 }
3944 
3945 int
3946 if_printf(struct ifnet *ifp, const char *fmt, ...)
3947 {
3948 	char if_fmt[256];
3949 	va_list ap;
3950 
3951 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3952 	va_start(ap, fmt);
3953 	vlog(LOG_INFO, if_fmt, ap);
3954 	va_end(ap);
3955 	return (0);
3956 }
3957 
3958 void
3959 if_start(struct ifnet *ifp)
3960 {
3961 
3962 	(*(ifp)->if_start)(ifp);
3963 }
3964 
3965 /*
3966  * Backwards compatibility interface for drivers
3967  * that have not implemented it
3968  */
3969 static int
3970 if_transmit(struct ifnet *ifp, struct mbuf *m)
3971 {
3972 	int error;
3973 
3974 	IFQ_HANDOFF(ifp, m, error);
3975 	return (error);
3976 }
3977 
3978 static void
3979 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3980 {
3981 
3982 	m_freem(m);
3983 }
3984 
3985 int
3986 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3987 {
3988 	int active = 0;
3989 
3990 	IF_LOCK(ifq);
3991 	if (_IF_QFULL(ifq)) {
3992 		IF_UNLOCK(ifq);
3993 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3994 		m_freem(m);
3995 		return (0);
3996 	}
3997 	if (ifp != NULL) {
3998 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3999 		if (m->m_flags & (M_BCAST|M_MCAST))
4000 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4001 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4002 	}
4003 	_IF_ENQUEUE(ifq, m);
4004 	IF_UNLOCK(ifq);
4005 	if (ifp != NULL && !active)
4006 		(*(ifp)->if_start)(ifp);
4007 	return (1);
4008 }
4009 
4010 void
4011 if_register_com_alloc(u_char type,
4012     if_com_alloc_t *a, if_com_free_t *f)
4013 {
4014 
4015 	KASSERT(if_com_alloc[type] == NULL,
4016 	    ("if_register_com_alloc: %d already registered", type));
4017 	KASSERT(if_com_free[type] == NULL,
4018 	    ("if_register_com_alloc: %d free already registered", type));
4019 
4020 	if_com_alloc[type] = a;
4021 	if_com_free[type] = f;
4022 }
4023 
4024 void
4025 if_deregister_com_alloc(u_char type)
4026 {
4027 
4028 	KASSERT(if_com_alloc[type] != NULL,
4029 	    ("if_deregister_com_alloc: %d not registered", type));
4030 	KASSERT(if_com_free[type] != NULL,
4031 	    ("if_deregister_com_alloc: %d free not registered", type));
4032 	if_com_alloc[type] = NULL;
4033 	if_com_free[type] = NULL;
4034 }
4035 
4036 /* API for driver access to network stack owned ifnet.*/
4037 uint64_t
4038 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4039 {
4040 	uint64_t oldbrate;
4041 
4042 	oldbrate = ifp->if_baudrate;
4043 	ifp->if_baudrate = baudrate;
4044 	return (oldbrate);
4045 }
4046 
4047 uint64_t
4048 if_getbaudrate(if_t ifp)
4049 {
4050 
4051 	return (((struct ifnet *)ifp)->if_baudrate);
4052 }
4053 
4054 int
4055 if_setcapabilities(if_t ifp, int capabilities)
4056 {
4057 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4058 	return (0);
4059 }
4060 
4061 int
4062 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4063 {
4064 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4065 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4066 
4067 	return (0);
4068 }
4069 
4070 int
4071 if_getcapabilities(if_t ifp)
4072 {
4073 	return ((struct ifnet *)ifp)->if_capabilities;
4074 }
4075 
4076 int
4077 if_setcapenable(if_t ifp, int capabilities)
4078 {
4079 	((struct ifnet *)ifp)->if_capenable = capabilities;
4080 	return (0);
4081 }
4082 
4083 int
4084 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4085 {
4086 	if(setcap)
4087 		((struct ifnet *)ifp)->if_capenable |= setcap;
4088 	if(clearcap)
4089 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4090 
4091 	return (0);
4092 }
4093 
4094 const char *
4095 if_getdname(if_t ifp)
4096 {
4097 	return ((struct ifnet *)ifp)->if_dname;
4098 }
4099 
4100 int
4101 if_togglecapenable(if_t ifp, int togglecap)
4102 {
4103 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4104 	return (0);
4105 }
4106 
4107 int
4108 if_getcapenable(if_t ifp)
4109 {
4110 	return ((struct ifnet *)ifp)->if_capenable;
4111 }
4112 
4113 /*
4114  * This is largely undesirable because it ties ifnet to a device, but does
4115  * provide flexiblity for an embedded product vendor. Should be used with
4116  * the understanding that it violates the interface boundaries, and should be
4117  * a last resort only.
4118  */
4119 int
4120 if_setdev(if_t ifp, void *dev)
4121 {
4122 	return (0);
4123 }
4124 
4125 int
4126 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4127 {
4128 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4129 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4130 
4131 	return (0);
4132 }
4133 
4134 int
4135 if_getdrvflags(if_t ifp)
4136 {
4137 	return ((struct ifnet *)ifp)->if_drv_flags;
4138 }
4139 
4140 int
4141 if_setdrvflags(if_t ifp, int flags)
4142 {
4143 	((struct ifnet *)ifp)->if_drv_flags = flags;
4144 	return (0);
4145 }
4146 
4147 int
4148 if_setflags(if_t ifp, int flags)
4149 {
4150 
4151 	ifp->if_flags = flags;
4152 	return (0);
4153 }
4154 
4155 int
4156 if_setflagbits(if_t ifp, int set, int clear)
4157 {
4158 	((struct ifnet *)ifp)->if_flags |= set;
4159 	((struct ifnet *)ifp)->if_flags &= ~clear;
4160 
4161 	return (0);
4162 }
4163 
4164 int
4165 if_getflags(if_t ifp)
4166 {
4167 	return ((struct ifnet *)ifp)->if_flags;
4168 }
4169 
4170 int
4171 if_clearhwassist(if_t ifp)
4172 {
4173 	((struct ifnet *)ifp)->if_hwassist = 0;
4174 	return (0);
4175 }
4176 
4177 int
4178 if_sethwassistbits(if_t ifp, int toset, int toclear)
4179 {
4180 	((struct ifnet *)ifp)->if_hwassist |= toset;
4181 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4182 
4183 	return (0);
4184 }
4185 
4186 int
4187 if_sethwassist(if_t ifp, int hwassist_bit)
4188 {
4189 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4190 	return (0);
4191 }
4192 
4193 int
4194 if_gethwassist(if_t ifp)
4195 {
4196 	return ((struct ifnet *)ifp)->if_hwassist;
4197 }
4198 
4199 int
4200 if_setmtu(if_t ifp, int mtu)
4201 {
4202 	((struct ifnet *)ifp)->if_mtu = mtu;
4203 	return (0);
4204 }
4205 
4206 int
4207 if_getmtu(if_t ifp)
4208 {
4209 	return ((struct ifnet *)ifp)->if_mtu;
4210 }
4211 
4212 int
4213 if_getmtu_family(if_t ifp, int family)
4214 {
4215 	struct domain *dp;
4216 
4217 	for (dp = domains; dp; dp = dp->dom_next) {
4218 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4219 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4220 	}
4221 
4222 	return (((struct ifnet *)ifp)->if_mtu);
4223 }
4224 
4225 /*
4226  * Methods for drivers to access interface unicast and multicast
4227  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4228  * 'struct ifmultiaddr'.
4229  */
4230 u_int
4231 if_lladdr_count(if_t ifp)
4232 {
4233 	struct epoch_tracker et;
4234 	struct ifaddr *ifa;
4235 	u_int count;
4236 
4237 	count = 0;
4238 	NET_EPOCH_ENTER(et);
4239 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4240 		if (ifa->ifa_addr->sa_family == AF_LINK)
4241 			count++;
4242 	NET_EPOCH_EXIT(et);
4243 
4244 	return (count);
4245 }
4246 
4247 u_int
4248 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4249 {
4250 	struct epoch_tracker et;
4251 	struct ifaddr *ifa;
4252 	u_int count;
4253 
4254 	MPASS(cb);
4255 
4256 	count = 0;
4257 	NET_EPOCH_ENTER(et);
4258 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4259 		if (ifa->ifa_addr->sa_family != AF_LINK)
4260 			continue;
4261 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4262 		    count);
4263 	}
4264 	NET_EPOCH_EXIT(et);
4265 
4266 	return (count);
4267 }
4268 
4269 u_int
4270 if_llmaddr_count(if_t ifp)
4271 {
4272 	struct epoch_tracker et;
4273 	struct ifmultiaddr *ifma;
4274 	int count;
4275 
4276 	count = 0;
4277 	NET_EPOCH_ENTER(et);
4278 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4279 		if (ifma->ifma_addr->sa_family == AF_LINK)
4280 			count++;
4281 	NET_EPOCH_EXIT(et);
4282 
4283 	return (count);
4284 }
4285 
4286 u_int
4287 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4288 {
4289 	struct epoch_tracker et;
4290 	struct ifmultiaddr *ifma;
4291 	u_int count;
4292 
4293 	MPASS(cb);
4294 
4295 	count = 0;
4296 	NET_EPOCH_ENTER(et);
4297 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4298 		if (ifma->ifma_addr->sa_family != AF_LINK)
4299 			continue;
4300 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4301 		    count);
4302 	}
4303 	NET_EPOCH_EXIT(et);
4304 
4305 	return (count);
4306 }
4307 
4308 int
4309 if_setsoftc(if_t ifp, void *softc)
4310 {
4311 	((struct ifnet *)ifp)->if_softc = softc;
4312 	return (0);
4313 }
4314 
4315 void *
4316 if_getsoftc(if_t ifp)
4317 {
4318 	return ((struct ifnet *)ifp)->if_softc;
4319 }
4320 
4321 void
4322 if_setrcvif(struct mbuf *m, if_t ifp)
4323 {
4324 
4325 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4326 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4327 }
4328 
4329 void
4330 if_setvtag(struct mbuf *m, uint16_t tag)
4331 {
4332 	m->m_pkthdr.ether_vtag = tag;
4333 }
4334 
4335 uint16_t
4336 if_getvtag(struct mbuf *m)
4337 {
4338 
4339 	return (m->m_pkthdr.ether_vtag);
4340 }
4341 
4342 int
4343 if_sendq_empty(if_t ifp)
4344 {
4345 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4346 }
4347 
4348 struct ifaddr *
4349 if_getifaddr(if_t ifp)
4350 {
4351 	return ((struct ifnet *)ifp)->if_addr;
4352 }
4353 
4354 int
4355 if_getamcount(if_t ifp)
4356 {
4357 	return ((struct ifnet *)ifp)->if_amcount;
4358 }
4359 
4360 int
4361 if_setsendqready(if_t ifp)
4362 {
4363 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4364 	return (0);
4365 }
4366 
4367 int
4368 if_setsendqlen(if_t ifp, int tx_desc_count)
4369 {
4370 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4371 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4372 
4373 	return (0);
4374 }
4375 
4376 int
4377 if_vlantrunkinuse(if_t ifp)
4378 {
4379 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4380 }
4381 
4382 int
4383 if_input(if_t ifp, struct mbuf* sendmp)
4384 {
4385 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4386 	return (0);
4387 
4388 }
4389 
4390 struct mbuf *
4391 if_dequeue(if_t ifp)
4392 {
4393 	struct mbuf *m;
4394 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4395 
4396 	return (m);
4397 }
4398 
4399 int
4400 if_sendq_prepend(if_t ifp, struct mbuf *m)
4401 {
4402 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4403 	return (0);
4404 }
4405 
4406 int
4407 if_setifheaderlen(if_t ifp, int len)
4408 {
4409 	((struct ifnet *)ifp)->if_hdrlen = len;
4410 	return (0);
4411 }
4412 
4413 caddr_t
4414 if_getlladdr(if_t ifp)
4415 {
4416 	return (IF_LLADDR((struct ifnet *)ifp));
4417 }
4418 
4419 void *
4420 if_gethandle(u_char type)
4421 {
4422 	return (if_alloc(type));
4423 }
4424 
4425 void
4426 if_bpfmtap(if_t ifh, struct mbuf *m)
4427 {
4428 	struct ifnet *ifp = (struct ifnet *)ifh;
4429 
4430 	BPF_MTAP(ifp, m);
4431 }
4432 
4433 void
4434 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4435 {
4436 	struct ifnet *ifp = (struct ifnet *)ifh;
4437 
4438 	ETHER_BPF_MTAP(ifp, m);
4439 }
4440 
4441 void
4442 if_vlancap(if_t ifh)
4443 {
4444 	struct ifnet *ifp = (struct ifnet *)ifh;
4445 	VLAN_CAPABILITIES(ifp);
4446 }
4447 
4448 int
4449 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4450 {
4451 
4452 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4453         return (0);
4454 }
4455 
4456 int
4457 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4458 {
4459 
4460 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4461         return (0);
4462 }
4463 
4464 int
4465 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4466 {
4467 
4468 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4469         return (0);
4470 }
4471 
4472 u_int
4473 if_gethwtsomax(if_t ifp)
4474 {
4475 
4476 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4477 }
4478 
4479 u_int
4480 if_gethwtsomaxsegcount(if_t ifp)
4481 {
4482 
4483 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4484 }
4485 
4486 u_int
4487 if_gethwtsomaxsegsize(if_t ifp)
4488 {
4489 
4490 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4491 }
4492 
4493 void
4494 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4495 {
4496 	((struct ifnet *)ifp)->if_init = init_fn;
4497 }
4498 
4499 void
4500 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4501 {
4502 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4503 }
4504 
4505 void
4506 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4507 {
4508 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4509 }
4510 
4511 void
4512 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4513 {
4514 	((struct ifnet *)ifp)->if_transmit = start_fn;
4515 }
4516 
4517 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4518 {
4519 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4520 
4521 }
4522 
4523 void
4524 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4525 {
4526 
4527 	ifp->if_get_counter = fn;
4528 }
4529 
4530 /* Revisit these - These are inline functions originally. */
4531 int
4532 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4533 {
4534 	return drbr_inuse(ifh, br);
4535 }
4536 
4537 struct mbuf*
4538 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4539 {
4540 	return drbr_dequeue(ifh, br);
4541 }
4542 
4543 int
4544 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4545 {
4546 	return drbr_needs_enqueue(ifh, br);
4547 }
4548 
4549 int
4550 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4551 {
4552 	return drbr_enqueue(ifh, br, m);
4553 
4554 }
4555