xref: /freebsd/sys/net/if.c (revision 5def4c47d4bd90b209b9b4a4ba9faec15846d8fd)
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/capsicum.h>
41 #include <sys/conf.h>
42 #include <sys/eventhandler.h>
43 #include <sys/malloc.h>
44 #include <sys/domainset.h>
45 #include <sys/sbuf.h>
46 #include <sys/bus.h>
47 #include <sys/epoch.h>
48 #include <sys/mbuf.h>
49 #include <sys/systm.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/protosw.h>
55 #include <sys/kernel.h>
56 #include <sys/lock.h>
57 #include <sys/refcount.h>
58 #include <sys/module.h>
59 #include <sys/rwlock.h>
60 #include <sys/sockio.h>
61 #include <sys/syslog.h>
62 #include <sys/sysctl.h>
63 #include <sys/sysent.h>
64 #include <sys/taskqueue.h>
65 #include <sys/domain.h>
66 #include <sys/jail.h>
67 #include <sys/priv.h>
68 
69 #include <machine/stdarg.h>
70 #include <vm/uma.h>
71 
72 #include <net/bpf.h>
73 #include <net/ethernet.h>
74 #include <net/if.h>
75 #include <net/if_arp.h>
76 #include <net/if_clone.h>
77 #include <net/if_dl.h>
78 #include <net/if_types.h>
79 #include <net/if_var.h>
80 #include <net/if_media.h>
81 #include <net/if_vlan_var.h>
82 #include <net/radix.h>
83 #include <net/route.h>
84 #include <net/route/route_ctl.h>
85 #include <net/vnet.h>
86 
87 #if defined(INET) || defined(INET6)
88 #include <net/ethernet.h>
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/ip_carp.h>
93 #ifdef INET
94 #include <net/debugnet.h>
95 #include <netinet/if_ether.h>
96 #endif /* INET */
97 #ifdef INET6
98 #include <netinet6/in6_var.h>
99 #include <netinet6/in6_ifattach.h>
100 #endif /* INET6 */
101 #endif /* INET || INET6 */
102 
103 #include <security/mac/mac_framework.h>
104 
105 /*
106  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
107  * and ifr_ifru when it is used in SIOCGIFCONF.
108  */
109 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
110     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
111 
112 __read_mostly epoch_t net_epoch_preempt;
113 #ifdef COMPAT_FREEBSD32
114 #include <sys/mount.h>
115 #include <compat/freebsd32/freebsd32.h>
116 
117 struct ifreq_buffer32 {
118 	uint32_t	length;		/* (size_t) */
119 	uint32_t	buffer;		/* (void *) */
120 };
121 
122 /*
123  * Interface request structure used for socket
124  * ioctl's.  All interface ioctl's must have parameter
125  * definitions which begin with ifr_name.  The
126  * remainder may be interface specific.
127  */
128 struct ifreq32 {
129 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
130 	union {
131 		struct sockaddr	ifru_addr;
132 		struct sockaddr	ifru_dstaddr;
133 		struct sockaddr	ifru_broadaddr;
134 		struct ifreq_buffer32 ifru_buffer;
135 		short		ifru_flags[2];
136 		short		ifru_index;
137 		int		ifru_jid;
138 		int		ifru_metric;
139 		int		ifru_mtu;
140 		int		ifru_phys;
141 		int		ifru_media;
142 		uint32_t	ifru_data;
143 		int		ifru_cap[2];
144 		u_int		ifru_fib;
145 		u_char		ifru_vlan_pcp;
146 	} ifr_ifru;
147 };
148 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
149 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
150     __offsetof(struct ifreq32, ifr_ifru));
151 
152 struct ifconf32 {
153 	int32_t	ifc_len;
154 	union {
155 		uint32_t	ifcu_buf;
156 		uint32_t	ifcu_req;
157 	} ifc_ifcu;
158 };
159 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
160 
161 struct ifdrv32 {
162 	char		ifd_name[IFNAMSIZ];
163 	uint32_t	ifd_cmd;
164 	uint32_t	ifd_len;
165 	uint32_t	ifd_data;
166 };
167 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
168 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
169 
170 struct ifgroupreq32 {
171 	char	ifgr_name[IFNAMSIZ];
172 	u_int	ifgr_len;
173 	union {
174 		char		ifgru_group[IFNAMSIZ];
175 		uint32_t	ifgru_groups;
176 	} ifgr_ifgru;
177 };
178 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
179 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
180 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
181 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
182 
183 struct ifmediareq32 {
184 	char		ifm_name[IFNAMSIZ];
185 	int		ifm_current;
186 	int		ifm_mask;
187 	int		ifm_status;
188 	int		ifm_active;
189 	int		ifm_count;
190 	uint32_t	ifm_ulist;	/* (int *) */
191 };
192 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
193 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
194 #endif /* COMPAT_FREEBSD32 */
195 
196 union ifreq_union {
197 	struct ifreq	ifr;
198 #ifdef COMPAT_FREEBSD32
199 	struct ifreq32	ifr32;
200 #endif
201 };
202 
203 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
204     "Link layers");
205 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
206     "Generic link-management");
207 
208 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
209     &ifqmaxlen, 0, "max send queue size");
210 
211 /* Log link state change events */
212 static int log_link_state_change = 1;
213 
214 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
215 	&log_link_state_change, 0,
216 	"log interface link state change events");
217 
218 /* Log promiscuous mode change events */
219 static int log_promisc_mode_change = 1;
220 
221 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
222 	&log_promisc_mode_change, 1,
223 	"log promiscuous mode change events");
224 
225 /* Interface description */
226 static unsigned int ifdescr_maxlen = 1024;
227 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
228 	&ifdescr_maxlen, 0,
229 	"administrative maximum length for interface description");
230 
231 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
232 
233 /* global sx for non-critical path ifdescr */
234 static struct sx ifdescr_sx;
235 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
236 
237 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
238 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
239 /* These are external hooks for CARP. */
240 void	(*carp_linkstate_p)(struct ifnet *ifp);
241 void	(*carp_demote_adj_p)(int, char *);
242 int	(*carp_master_p)(struct ifaddr *);
243 #if defined(INET) || defined(INET6)
244 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
245 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
246     const struct sockaddr *sa);
247 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
248 int	(*carp_attach_p)(struct ifaddr *, int);
249 void	(*carp_detach_p)(struct ifaddr *, bool);
250 #endif
251 #ifdef INET
252 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
253 #endif
254 #ifdef INET6
255 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
256 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
257     const struct in6_addr *taddr);
258 #endif
259 
260 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
261 
262 /*
263  * XXX: Style; these should be sorted alphabetically, and unprototyped
264  * static functions should be prototyped. Currently they are sorted by
265  * declaration order.
266  */
267 static void	if_attachdomain(void *);
268 static void	if_attachdomain1(struct ifnet *);
269 static int	ifconf(u_long, caddr_t);
270 static void	*if_grow(void);
271 static void	if_input_default(struct ifnet *, struct mbuf *);
272 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
273 static void	if_route(struct ifnet *, int flag, int fam);
274 static int	if_setflag(struct ifnet *, int, int, int *, int);
275 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
276 static void	if_unroute(struct ifnet *, int flag, int fam);
277 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
278 static void	do_link_state_change(void *, int);
279 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
280 static int	if_getgroupmembers(struct ifgroupreq *);
281 static void	if_delgroups(struct ifnet *);
282 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
283 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
284 static void	if_siocaddmulti(void *, int);
285 static void	if_link_ifnet(struct ifnet *);
286 static bool	if_unlink_ifnet(struct ifnet *, bool);
287 #ifdef VIMAGE
288 static int	if_vmove(struct ifnet *, struct vnet *);
289 #endif
290 
291 #ifdef INET6
292 /*
293  * XXX: declare here to avoid to include many inet6 related files..
294  * should be more generalized?
295  */
296 extern void	nd6_setmtu(struct ifnet *);
297 #endif
298 
299 /* ipsec helper hooks */
300 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
301 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
302 
303 VNET_DEFINE(int, if_index);
304 int	ifqmaxlen = IFQ_MAXLEN;
305 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
306 VNET_DEFINE(struct ifgrouphead, ifg_head);
307 
308 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
309 
310 /* Table of ifnet by index. */
311 VNET_DEFINE(struct ifnet **, ifindex_table);
312 
313 #define	V_if_indexlim		VNET(if_indexlim)
314 #define	V_ifindex_table		VNET(ifindex_table)
315 
316 /*
317  * The global network interface list (V_ifnet) and related state (such as
318  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
319  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
320  */
321 struct sx ifnet_sxlock;
322 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
323 
324 struct sx ifnet_detach_sxlock;
325 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
326     SX_RECURSE);
327 
328 /*
329  * The allocation of network interfaces is a rather non-atomic affair; we
330  * need to select an index before we are ready to expose the interface for
331  * use, so will use this pointer value to indicate reservation.
332  */
333 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
334 
335 #ifdef VIMAGE
336 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
337     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
338 #endif
339 
340 static	if_com_alloc_t *if_com_alloc[256];
341 static	if_com_free_t *if_com_free[256];
342 
343 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
344 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
345 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
346 
347 struct ifnet *
348 ifnet_byindex(u_short idx)
349 {
350 	struct ifnet *ifp;
351 
352 	if (__predict_false(idx > V_if_index))
353 		return (NULL);
354 
355 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
356 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
357 }
358 
359 struct ifnet *
360 ifnet_byindex_ref(u_short idx)
361 {
362 	struct ifnet *ifp;
363 
364 	NET_EPOCH_ASSERT();
365 
366 	ifp = ifnet_byindex(idx);
367 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
368 		return (NULL);
369 	if (!if_try_ref(ifp))
370 		return (NULL);
371 	return (ifp);
372 }
373 
374 /*
375  * Allocate an ifindex array entry; return 0 on success or an error on
376  * failure.
377  */
378 static u_short
379 ifindex_alloc(void **old)
380 {
381 	u_short idx;
382 
383 	IFNET_WLOCK_ASSERT();
384 	/*
385 	 * Try to find an empty slot below V_if_index.  If we fail, take the
386 	 * next slot.
387 	 */
388 	for (idx = 1; idx <= V_if_index; idx++) {
389 		if (V_ifindex_table[idx] == NULL)
390 			break;
391 	}
392 
393 	/* Catch if_index overflow. */
394 	if (idx >= V_if_indexlim) {
395 		*old = if_grow();
396 		return (USHRT_MAX);
397 	}
398 	if (idx > V_if_index)
399 		V_if_index = idx;
400 	return (idx);
401 }
402 
403 static void
404 ifindex_free_locked(u_short idx)
405 {
406 
407 	IFNET_WLOCK_ASSERT();
408 
409 	V_ifindex_table[idx] = NULL;
410 	while (V_if_index > 0 &&
411 	    V_ifindex_table[V_if_index] == NULL)
412 		V_if_index--;
413 }
414 
415 static void
416 ifindex_free(u_short idx)
417 {
418 
419 	IFNET_WLOCK();
420 	ifindex_free_locked(idx);
421 	IFNET_WUNLOCK();
422 }
423 
424 static void
425 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
426 {
427 
428 	V_ifindex_table[idx] = ifp;
429 }
430 
431 struct ifaddr *
432 ifaddr_byindex(u_short idx)
433 {
434 	struct ifnet *ifp;
435 	struct ifaddr *ifa = NULL;
436 
437 	NET_EPOCH_ASSERT();
438 
439 	ifp = ifnet_byindex(idx);
440 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
441 		ifa_ref(ifa);
442 	return (ifa);
443 }
444 
445 /*
446  * Network interface utility routines.
447  *
448  * Routines with ifa_ifwith* names take sockaddr *'s as
449  * parameters.
450  */
451 
452 static void
453 vnet_if_init(const void *unused __unused)
454 {
455 	void *old;
456 
457 	CK_STAILQ_INIT(&V_ifnet);
458 	CK_STAILQ_INIT(&V_ifg_head);
459 	IFNET_WLOCK();
460 	old = if_grow();				/* create initial table */
461 	IFNET_WUNLOCK();
462 	epoch_wait_preempt(net_epoch_preempt);
463 	free(old, M_IFNET);
464 	vnet_if_clone_init();
465 }
466 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
467     NULL);
468 
469 #ifdef VIMAGE
470 static void
471 vnet_if_uninit(const void *unused __unused)
472 {
473 
474 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
475 	    "not empty", __func__, __LINE__, &V_ifnet));
476 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
477 	    "not empty", __func__, __LINE__, &V_ifg_head));
478 
479 	free((caddr_t)V_ifindex_table, M_IFNET);
480 }
481 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
482     vnet_if_uninit, NULL);
483 #endif
484 
485 static void
486 if_link_ifnet(struct ifnet *ifp)
487 {
488 
489 	IFNET_WLOCK();
490 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
491 #ifdef VIMAGE
492 	curvnet->vnet_ifcnt++;
493 #endif
494 	IFNET_WUNLOCK();
495 }
496 
497 static bool
498 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
499 {
500 	struct ifnet *iter;
501 	int found = 0;
502 
503 	IFNET_WLOCK();
504 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
505 		if (iter == ifp) {
506 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
507 			if (!vmove)
508 				ifp->if_flags |= IFF_DYING;
509 			found = 1;
510 			break;
511 		}
512 #ifdef VIMAGE
513 	curvnet->vnet_ifcnt--;
514 #endif
515 	IFNET_WUNLOCK();
516 
517 	return (found);
518 }
519 
520 #ifdef VIMAGE
521 static void
522 vnet_if_return(const void *unused __unused)
523 {
524 	struct ifnet *ifp, *nifp;
525 	struct ifnet **pending;
526 	int found, i;
527 
528 	i = 0;
529 
530 	/*
531 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
532 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
533 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
534 	 * complete. We can't do that from within NET_EPOCH.
535 	 *
536 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
537 	 * read/write lock. We cannot hold the lock as we call if_vmove()
538 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
539 	 * ctx lock.
540 	 */
541 	IFNET_WLOCK();
542 
543 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
544 	    M_IFNET, M_WAITOK | M_ZERO);
545 
546 	/* Return all inherited interfaces to their parent vnets. */
547 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
548 		if (ifp->if_home_vnet != ifp->if_vnet) {
549 			found = if_unlink_ifnet(ifp, true);
550 			MPASS(found);
551 
552 			pending[i++] = ifp;
553 		}
554 	}
555 	IFNET_WUNLOCK();
556 
557 	for (int j = 0; j < i; j++) {
558 		if_vmove(pending[j], pending[j]->if_home_vnet);
559 	}
560 
561 	free(pending, M_IFNET);
562 }
563 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
564     vnet_if_return, NULL);
565 #endif
566 
567 static void *
568 if_grow(void)
569 {
570 	int oldlim;
571 	u_int n;
572 	struct ifnet **e;
573 	void *old;
574 
575 	old = NULL;
576 	IFNET_WLOCK_ASSERT();
577 	oldlim = V_if_indexlim;
578 	IFNET_WUNLOCK();
579 	n = (oldlim << 1) * sizeof(*e);
580 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
581 	IFNET_WLOCK();
582 	if (V_if_indexlim != oldlim) {
583 		free(e, M_IFNET);
584 		return (NULL);
585 	}
586 	if (V_ifindex_table != NULL) {
587 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
588 		old = V_ifindex_table;
589 	}
590 	V_if_indexlim <<= 1;
591 	V_ifindex_table = e;
592 	return (old);
593 }
594 
595 /*
596  * Allocate a struct ifnet and an index for an interface.  A layer 2
597  * common structure will also be allocated if an allocation routine is
598  * registered for the passed type.
599  */
600 struct ifnet *
601 if_alloc_domain(u_char type, int numa_domain)
602 {
603 	struct ifnet *ifp;
604 	u_short idx;
605 	void *old;
606 
607 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
608 	if (numa_domain == IF_NODOM)
609 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
610 		    M_WAITOK | M_ZERO);
611 	else
612 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
613 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
614  restart:
615 	IFNET_WLOCK();
616 	idx = ifindex_alloc(&old);
617 	if (__predict_false(idx == USHRT_MAX)) {
618 		IFNET_WUNLOCK();
619 		epoch_wait_preempt(net_epoch_preempt);
620 		free(old, M_IFNET);
621 		goto restart;
622 	}
623 	ifnet_setbyindex(idx, IFNET_HOLD);
624 	IFNET_WUNLOCK();
625 	ifp->if_index = idx;
626 	ifp->if_type = type;
627 	ifp->if_alloctype = type;
628 	ifp->if_numa_domain = numa_domain;
629 #ifdef VIMAGE
630 	ifp->if_vnet = curvnet;
631 #endif
632 	if (if_com_alloc[type] != NULL) {
633 		ifp->if_l2com = if_com_alloc[type](type, ifp);
634 		if (ifp->if_l2com == NULL) {
635 			free(ifp, M_IFNET);
636 			ifindex_free(idx);
637 			return (NULL);
638 		}
639 	}
640 
641 	IF_ADDR_LOCK_INIT(ifp);
642 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
643 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
644 	ifp->if_afdata_initialized = 0;
645 	IF_AFDATA_LOCK_INIT(ifp);
646 	CK_STAILQ_INIT(&ifp->if_addrhead);
647 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
648 	CK_STAILQ_INIT(&ifp->if_groups);
649 #ifdef MAC
650 	mac_ifnet_init(ifp);
651 #endif
652 	ifq_init(&ifp->if_snd, ifp);
653 
654 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
655 	for (int i = 0; i < IFCOUNTERS; i++)
656 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
657 	ifp->if_get_counter = if_get_counter_default;
658 	ifp->if_pcp = IFNET_PCP_NONE;
659 	ifnet_setbyindex(ifp->if_index, ifp);
660 	return (ifp);
661 }
662 
663 struct ifnet *
664 if_alloc_dev(u_char type, device_t dev)
665 {
666 	int numa_domain;
667 
668 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
669 		return (if_alloc_domain(type, IF_NODOM));
670 	return (if_alloc_domain(type, numa_domain));
671 }
672 
673 struct ifnet *
674 if_alloc(u_char type)
675 {
676 
677 	return (if_alloc_domain(type, IF_NODOM));
678 }
679 /*
680  * Do the actual work of freeing a struct ifnet, and layer 2 common
681  * structure.  This call is made when the last reference to an
682  * interface is released.
683  */
684 static void
685 if_free_internal(struct ifnet *ifp)
686 {
687 
688 	KASSERT((ifp->if_flags & IFF_DYING),
689 	    ("if_free_internal: interface not dying"));
690 
691 	if (if_com_free[ifp->if_alloctype] != NULL)
692 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
693 		    ifp->if_alloctype);
694 
695 #ifdef MAC
696 	mac_ifnet_destroy(ifp);
697 #endif /* MAC */
698 	IF_AFDATA_DESTROY(ifp);
699 	IF_ADDR_LOCK_DESTROY(ifp);
700 	ifq_delete(&ifp->if_snd);
701 
702 	for (int i = 0; i < IFCOUNTERS; i++)
703 		counter_u64_free(ifp->if_counters[i]);
704 
705 	free(ifp->if_description, M_IFDESCR);
706 	free(ifp->if_hw_addr, M_IFADDR);
707 	free(ifp, M_IFNET);
708 }
709 
710 static void
711 if_destroy(epoch_context_t ctx)
712 {
713 	struct ifnet *ifp;
714 
715 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
716 	if_free_internal(ifp);
717 }
718 
719 /*
720  * Deregister an interface and free the associated storage.
721  */
722 void
723 if_free(struct ifnet *ifp)
724 {
725 
726 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
727 
728 	CURVNET_SET_QUIET(ifp->if_vnet);
729 	IFNET_WLOCK();
730 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
731 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
732 
733 	ifindex_free_locked(ifp->if_index);
734 	IFNET_WUNLOCK();
735 
736 	if (refcount_release(&ifp->if_refcount))
737 		NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
738 	CURVNET_RESTORE();
739 }
740 
741 /*
742  * Interfaces to keep an ifnet type-stable despite the possibility of the
743  * driver calling if_free().  If there are additional references, we defer
744  * freeing the underlying data structure.
745  */
746 void
747 if_ref(struct ifnet *ifp)
748 {
749 	u_int old;
750 
751 	/* We don't assert the ifnet list lock here, but arguably should. */
752 	old = refcount_acquire(&ifp->if_refcount);
753 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
754 }
755 
756 bool
757 if_try_ref(struct ifnet *ifp)
758 {
759 	NET_EPOCH_ASSERT();
760 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
761 }
762 
763 void
764 if_rele(struct ifnet *ifp)
765 {
766 
767 	if (!refcount_release(&ifp->if_refcount))
768 		return;
769 	NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
770 }
771 
772 void
773 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
774 {
775 
776 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
777 
778 	if (ifq->ifq_maxlen == 0)
779 		ifq->ifq_maxlen = ifqmaxlen;
780 
781 	ifq->altq_type = 0;
782 	ifq->altq_disc = NULL;
783 	ifq->altq_flags &= ALTQF_CANTCHANGE;
784 	ifq->altq_tbr  = NULL;
785 	ifq->altq_ifp  = ifp;
786 }
787 
788 void
789 ifq_delete(struct ifaltq *ifq)
790 {
791 	mtx_destroy(&ifq->ifq_mtx);
792 }
793 
794 /*
795  * Perform generic interface initialization tasks and attach the interface
796  * to the list of "active" interfaces.  If vmove flag is set on entry
797  * to if_attach_internal(), perform only a limited subset of initialization
798  * tasks, given that we are moving from one vnet to another an ifnet which
799  * has already been fully initialized.
800  *
801  * Note that if_detach_internal() removes group membership unconditionally
802  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
803  * Thus, when if_vmove() is applied to a cloned interface, group membership
804  * is lost while a cloned one always joins a group whose name is
805  * ifc->ifc_name.  To recover this after if_detach_internal() and
806  * if_attach_internal(), the cloner should be specified to
807  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
808  * attempts to join a group whose name is ifc->ifc_name.
809  *
810  * XXX:
811  *  - The decision to return void and thus require this function to
812  *    succeed is questionable.
813  *  - We should probably do more sanity checking.  For instance we don't
814  *    do anything to insure if_xname is unique or non-empty.
815  */
816 void
817 if_attach(struct ifnet *ifp)
818 {
819 
820 	if_attach_internal(ifp, 0, NULL);
821 }
822 
823 /*
824  * Compute the least common TSO limit.
825  */
826 void
827 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
828 {
829 	/*
830 	 * 1) If there is no limit currently, take the limit from
831 	 * the network adapter.
832 	 *
833 	 * 2) If the network adapter has a limit below the current
834 	 * limit, apply it.
835 	 */
836 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
837 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
838 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
839 	}
840 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
841 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
842 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
843 	}
844 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
845 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
846 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
847 	}
848 }
849 
850 /*
851  * Update TSO limit of a network adapter.
852  *
853  * Returns zero if no change. Else non-zero.
854  */
855 int
856 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
857 {
858 	int retval = 0;
859 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
860 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
861 		retval++;
862 	}
863 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
864 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
865 		retval++;
866 	}
867 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
868 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
869 		retval++;
870 	}
871 	return (retval);
872 }
873 
874 static void
875 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
876 {
877 	unsigned socksize, ifasize;
878 	int namelen, masklen;
879 	struct sockaddr_dl *sdl;
880 	struct ifaddr *ifa;
881 
882 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
883 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
884 		    ifp->if_xname);
885 
886 #ifdef VIMAGE
887 	ifp->if_vnet = curvnet;
888 	if (ifp->if_home_vnet == NULL)
889 		ifp->if_home_vnet = curvnet;
890 #endif
891 
892 	if_addgroup(ifp, IFG_ALL);
893 
894 	/* Restore group membership for cloned interfaces. */
895 	if (vmove && ifc != NULL)
896 		if_clone_addgroup(ifp, ifc);
897 
898 	getmicrotime(&ifp->if_lastchange);
899 	ifp->if_epoch = time_uptime;
900 
901 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
902 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
903 	    ("transmit and qflush must both either be set or both be NULL"));
904 	if (ifp->if_transmit == NULL) {
905 		ifp->if_transmit = if_transmit;
906 		ifp->if_qflush = if_qflush;
907 	}
908 	if (ifp->if_input == NULL)
909 		ifp->if_input = if_input_default;
910 
911 	if (ifp->if_requestencap == NULL)
912 		ifp->if_requestencap = if_requestencap_default;
913 
914 	if (!vmove) {
915 #ifdef MAC
916 		mac_ifnet_create(ifp);
917 #endif
918 
919 		/*
920 		 * Create a Link Level name for this device.
921 		 */
922 		namelen = strlen(ifp->if_xname);
923 		/*
924 		 * Always save enough space for any possiable name so we
925 		 * can do a rename in place later.
926 		 */
927 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
928 		socksize = masklen + ifp->if_addrlen;
929 		if (socksize < sizeof(*sdl))
930 			socksize = sizeof(*sdl);
931 		socksize = roundup2(socksize, sizeof(long));
932 		ifasize = sizeof(*ifa) + 2 * socksize;
933 		ifa = ifa_alloc(ifasize, M_WAITOK);
934 		sdl = (struct sockaddr_dl *)(ifa + 1);
935 		sdl->sdl_len = socksize;
936 		sdl->sdl_family = AF_LINK;
937 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
938 		sdl->sdl_nlen = namelen;
939 		sdl->sdl_index = ifp->if_index;
940 		sdl->sdl_type = ifp->if_type;
941 		ifp->if_addr = ifa;
942 		ifa->ifa_ifp = ifp;
943 		ifa->ifa_addr = (struct sockaddr *)sdl;
944 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
945 		ifa->ifa_netmask = (struct sockaddr *)sdl;
946 		sdl->sdl_len = masklen;
947 		while (namelen != 0)
948 			sdl->sdl_data[--namelen] = 0xff;
949 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
950 		/* Reliably crash if used uninitialized. */
951 		ifp->if_broadcastaddr = NULL;
952 
953 		if (ifp->if_type == IFT_ETHER) {
954 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
955 			    M_WAITOK | M_ZERO);
956 		}
957 
958 #if defined(INET) || defined(INET6)
959 		/* Use defaults for TSO, if nothing is set */
960 		if (ifp->if_hw_tsomax == 0 &&
961 		    ifp->if_hw_tsomaxsegcount == 0 &&
962 		    ifp->if_hw_tsomaxsegsize == 0) {
963 			/*
964 			 * The TSO defaults needs to be such that an
965 			 * NFS mbuf list of 35 mbufs totalling just
966 			 * below 64K works and that a chain of mbufs
967 			 * can be defragged into at most 32 segments:
968 			 */
969 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
970 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
971 			ifp->if_hw_tsomaxsegcount = 35;
972 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
973 
974 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
975 			if (ifp->if_capabilities & IFCAP_TSO) {
976 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
977 				    ifp->if_hw_tsomax,
978 				    ifp->if_hw_tsomaxsegcount,
979 				    ifp->if_hw_tsomaxsegsize);
980 			}
981 		}
982 #endif
983 	}
984 #ifdef VIMAGE
985 	else {
986 		/*
987 		 * Update the interface index in the link layer address
988 		 * of the interface.
989 		 */
990 		for (ifa = ifp->if_addr; ifa != NULL;
991 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
992 			if (ifa->ifa_addr->sa_family == AF_LINK) {
993 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
994 				sdl->sdl_index = ifp->if_index;
995 			}
996 		}
997 	}
998 #endif
999 
1000 	if_link_ifnet(ifp);
1001 
1002 	if (domain_init_status >= 2)
1003 		if_attachdomain1(ifp);
1004 
1005 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1006 	if (IS_DEFAULT_VNET(curvnet))
1007 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
1008 
1009 	/* Announce the interface. */
1010 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1011 }
1012 
1013 static void
1014 if_epochalloc(void *dummy __unused)
1015 {
1016 
1017 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
1018 }
1019 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
1020 
1021 static void
1022 if_attachdomain(void *dummy)
1023 {
1024 	struct ifnet *ifp;
1025 
1026 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
1027 		if_attachdomain1(ifp);
1028 }
1029 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
1030     if_attachdomain, NULL);
1031 
1032 static void
1033 if_attachdomain1(struct ifnet *ifp)
1034 {
1035 	struct domain *dp;
1036 
1037 	/*
1038 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
1039 	 * cannot lock ifp->if_afdata initialization, entirely.
1040 	 */
1041 	IF_AFDATA_LOCK(ifp);
1042 	if (ifp->if_afdata_initialized >= domain_init_status) {
1043 		IF_AFDATA_UNLOCK(ifp);
1044 		log(LOG_WARNING, "%s called more than once on %s\n",
1045 		    __func__, ifp->if_xname);
1046 		return;
1047 	}
1048 	ifp->if_afdata_initialized = domain_init_status;
1049 	IF_AFDATA_UNLOCK(ifp);
1050 
1051 	/* address family dependent data region */
1052 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1053 	for (dp = domains; dp; dp = dp->dom_next) {
1054 		if (dp->dom_ifattach)
1055 			ifp->if_afdata[dp->dom_family] =
1056 			    (*dp->dom_ifattach)(ifp);
1057 	}
1058 }
1059 
1060 /*
1061  * Remove any unicast or broadcast network addresses from an interface.
1062  */
1063 void
1064 if_purgeaddrs(struct ifnet *ifp)
1065 {
1066 	struct ifaddr *ifa;
1067 
1068 	while (1) {
1069 		struct epoch_tracker et;
1070 
1071 		NET_EPOCH_ENTER(et);
1072 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1073 			if (ifa->ifa_addr->sa_family != AF_LINK)
1074 				break;
1075 		}
1076 		NET_EPOCH_EXIT(et);
1077 
1078 		if (ifa == NULL)
1079 			break;
1080 #ifdef INET
1081 		/* XXX: Ugly!! ad hoc just for INET */
1082 		if (ifa->ifa_addr->sa_family == AF_INET) {
1083 			struct ifaliasreq ifr;
1084 
1085 			bzero(&ifr, sizeof(ifr));
1086 			ifr.ifra_addr = *ifa->ifa_addr;
1087 			if (ifa->ifa_dstaddr)
1088 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1089 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1090 			    NULL) == 0)
1091 				continue;
1092 		}
1093 #endif /* INET */
1094 #ifdef INET6
1095 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1096 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
1097 			/* ifp_addrhead is already updated */
1098 			continue;
1099 		}
1100 #endif /* INET6 */
1101 		IF_ADDR_WLOCK(ifp);
1102 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1103 		IF_ADDR_WUNLOCK(ifp);
1104 		ifa_free(ifa);
1105 	}
1106 }
1107 
1108 /*
1109  * Remove any multicast network addresses from an interface when an ifnet
1110  * is going away.
1111  */
1112 static void
1113 if_purgemaddrs(struct ifnet *ifp)
1114 {
1115 	struct ifmultiaddr *ifma;
1116 
1117 	IF_ADDR_WLOCK(ifp);
1118 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1119 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1120 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1121 		if_delmulti_locked(ifp, ifma, 1);
1122 	}
1123 	IF_ADDR_WUNLOCK(ifp);
1124 }
1125 
1126 /*
1127  * Detach an interface, removing it from the list of "active" interfaces.
1128  * If vmove flag is set on entry to if_detach_internal(), perform only a
1129  * limited subset of cleanup tasks, given that we are moving an ifnet from
1130  * one vnet to another, where it must be fully operational.
1131  *
1132  * XXXRW: There are some significant questions about event ordering, and
1133  * how to prevent things from starting to use the interface during detach.
1134  */
1135 void
1136 if_detach(struct ifnet *ifp)
1137 {
1138 	bool found;
1139 
1140 	CURVNET_SET_QUIET(ifp->if_vnet);
1141 	found = if_unlink_ifnet(ifp, false);
1142 	if (found) {
1143 		sx_xlock(&ifnet_detach_sxlock);
1144 		if_detach_internal(ifp, 0, NULL);
1145 		sx_xunlock(&ifnet_detach_sxlock);
1146 	}
1147 	CURVNET_RESTORE();
1148 }
1149 
1150 /*
1151  * The vmove flag, if set, indicates that we are called from a callpath
1152  * that is moving an interface to a different vnet instance.
1153  *
1154  * The shutdown flag, if set, indicates that we are called in the
1155  * process of shutting down a vnet instance.  Currently only the
1156  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1157  * on a vnet instance shutdown without this flag being set, e.g., when
1158  * the cloned interfaces are destoyed as first thing of teardown.
1159  */
1160 static int
1161 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1162 {
1163 	struct ifaddr *ifa;
1164 	int i;
1165 	struct domain *dp;
1166 #ifdef VIMAGE
1167 	bool shutdown;
1168 
1169 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1170 #endif
1171 
1172 	/*
1173 	 * At this point we know the interface still was on the ifnet list
1174 	 * and we removed it so we are in a stable state.
1175 	 */
1176 	epoch_wait_preempt(net_epoch_preempt);
1177 
1178 	/*
1179 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1180 	 * fixes issues about late destruction of multicast options
1181 	 * which lead to leave group calls, which in turn access the
1182 	 * belonging ifnet structure:
1183 	 */
1184 	epoch_drain_callbacks(net_epoch_preempt);
1185 
1186 	/*
1187 	 * In any case (destroy or vmove) detach us from the groups
1188 	 * and remove/wait for pending events on the taskq.
1189 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1190 	 */
1191 	if_delgroups(ifp);
1192 
1193 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1194 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1195 
1196 	/*
1197 	 * Check if this is a cloned interface or not. Must do even if
1198 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1199 	 * the if_clone_addgroup() will have a corrupted string overwise
1200 	 * from a gibberish pointer.
1201 	 */
1202 	if (vmove && ifcp != NULL)
1203 		*ifcp = if_clone_findifc(ifp);
1204 
1205 	if_down(ifp);
1206 
1207 #ifdef VIMAGE
1208 	/*
1209 	 * On VNET shutdown abort here as the stack teardown will do all
1210 	 * the work top-down for us.
1211 	 */
1212 	if (shutdown) {
1213 		/* Give interface users the chance to clean up. */
1214 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1215 
1216 		/*
1217 		 * In case of a vmove we are done here without error.
1218 		 * If we would signal an error it would lead to the same
1219 		 * abort as if we did not find the ifnet anymore.
1220 		 * if_detach() calls us in void context and does not care
1221 		 * about an early abort notification, so life is splendid :)
1222 		 */
1223 		goto finish_vnet_shutdown;
1224 	}
1225 #endif
1226 
1227 	/*
1228 	 * At this point we are not tearing down a VNET and are either
1229 	 * going to destroy or vmove the interface and have to cleanup
1230 	 * accordingly.
1231 	 */
1232 
1233 	/*
1234 	 * Remove routes and flush queues.
1235 	 */
1236 #ifdef ALTQ
1237 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1238 		altq_disable(&ifp->if_snd);
1239 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1240 		altq_detach(&ifp->if_snd);
1241 #endif
1242 
1243 	if_purgeaddrs(ifp);
1244 
1245 #ifdef INET
1246 	in_ifdetach(ifp);
1247 #endif
1248 
1249 #ifdef INET6
1250 	/*
1251 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1252 	 * before removing routing entries below, since IPv6 interface direct
1253 	 * routes are expected to be removed by the IPv6-specific kernel API.
1254 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1255 	 */
1256 	in6_ifdetach(ifp);
1257 #endif
1258 	if_purgemaddrs(ifp);
1259 
1260 	/* Announce that the interface is gone. */
1261 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1262 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1263 	if (IS_DEFAULT_VNET(curvnet))
1264 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1265 
1266 	if (!vmove) {
1267 		/*
1268 		 * Prevent further calls into the device driver via ifnet.
1269 		 */
1270 		if_dead(ifp);
1271 
1272 		/*
1273 		 * Clean up all addresses.
1274 		 */
1275 		IF_ADDR_WLOCK(ifp);
1276 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1277 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1278 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1279 			IF_ADDR_WUNLOCK(ifp);
1280 			ifa_free(ifa);
1281 		} else
1282 			IF_ADDR_WUNLOCK(ifp);
1283 	}
1284 
1285 	rt_flushifroutes(ifp);
1286 
1287 #ifdef VIMAGE
1288 finish_vnet_shutdown:
1289 #endif
1290 	/*
1291 	 * We cannot hold the lock over dom_ifdetach calls as they might
1292 	 * sleep, for example trying to drain a callout, thus open up the
1293 	 * theoretical race with re-attaching.
1294 	 */
1295 	IF_AFDATA_LOCK(ifp);
1296 	i = ifp->if_afdata_initialized;
1297 	ifp->if_afdata_initialized = 0;
1298 	IF_AFDATA_UNLOCK(ifp);
1299 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1300 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1301 			(*dp->dom_ifdetach)(ifp,
1302 			    ifp->if_afdata[dp->dom_family]);
1303 			ifp->if_afdata[dp->dom_family] = NULL;
1304 		}
1305 	}
1306 
1307 	return (0);
1308 }
1309 
1310 #ifdef VIMAGE
1311 /*
1312  * if_vmove() performs a limited version of if_detach() in current
1313  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1314  * An attempt is made to shrink if_index in current vnet, find an
1315  * unused if_index in target vnet and calls if_grow() if necessary,
1316  * and finally find an unused if_xname for the target vnet.
1317  */
1318 static int
1319 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1320 {
1321 	struct if_clone *ifc;
1322 #ifdef DEV_BPF
1323 	u_int bif_dlt, bif_hdrlen;
1324 #endif
1325 	void *old;
1326 	int rc;
1327 
1328 #ifdef DEV_BPF
1329  	/*
1330 	 * if_detach_internal() will call the eventhandler to notify
1331 	 * interface departure.  That will detach if_bpf.  We need to
1332 	 * safe the dlt and hdrlen so we can re-attach it later.
1333 	 */
1334 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1335 #endif
1336 
1337 	/*
1338 	 * Detach from current vnet, but preserve LLADDR info, do not
1339 	 * mark as dead etc. so that the ifnet can be reattached later.
1340 	 * If we cannot find it, we lost the race to someone else.
1341 	 */
1342 	rc = if_detach_internal(ifp, 1, &ifc);
1343 	if (rc != 0)
1344 		return (rc);
1345 
1346 	/*
1347 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1348 	 * the if_index for that vnet if possible.
1349 	 *
1350 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1351 	 * or we'd lock on one vnet and unlock on another.
1352 	 */
1353 	IFNET_WLOCK();
1354 	ifindex_free_locked(ifp->if_index);
1355 	IFNET_WUNLOCK();
1356 
1357 	/*
1358 	 * Perform interface-specific reassignment tasks, if provided by
1359 	 * the driver.
1360 	 */
1361 	if (ifp->if_reassign != NULL)
1362 		ifp->if_reassign(ifp, new_vnet, NULL);
1363 
1364 	/*
1365 	 * Switch to the context of the target vnet.
1366 	 */
1367 	CURVNET_SET_QUIET(new_vnet);
1368  restart:
1369 	IFNET_WLOCK();
1370 	ifp->if_index = ifindex_alloc(&old);
1371 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1372 		IFNET_WUNLOCK();
1373 		epoch_wait_preempt(net_epoch_preempt);
1374 		free(old, M_IFNET);
1375 		goto restart;
1376 	}
1377 	ifnet_setbyindex(ifp->if_index, ifp);
1378 	IFNET_WUNLOCK();
1379 
1380 	if_attach_internal(ifp, 1, ifc);
1381 
1382 #ifdef DEV_BPF
1383 	if (ifp->if_bpf == NULL)
1384 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1385 #endif
1386 
1387 	CURVNET_RESTORE();
1388 	return (0);
1389 }
1390 
1391 /*
1392  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1393  */
1394 static int
1395 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1396 {
1397 	struct prison *pr;
1398 	struct ifnet *difp;
1399 	int error;
1400 	bool found;
1401 	bool shutdown;
1402 
1403 	/* Try to find the prison within our visibility. */
1404 	sx_slock(&allprison_lock);
1405 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1406 	sx_sunlock(&allprison_lock);
1407 	if (pr == NULL)
1408 		return (ENXIO);
1409 	prison_hold_locked(pr);
1410 	mtx_unlock(&pr->pr_mtx);
1411 
1412 	/* Do not try to move the iface from and to the same prison. */
1413 	if (pr->pr_vnet == ifp->if_vnet) {
1414 		prison_free(pr);
1415 		return (EEXIST);
1416 	}
1417 
1418 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1419 	/* XXX Lock interfaces to avoid races. */
1420 	CURVNET_SET_QUIET(pr->pr_vnet);
1421 	difp = ifunit(ifname);
1422 	if (difp != NULL) {
1423 		CURVNET_RESTORE();
1424 		prison_free(pr);
1425 		return (EEXIST);
1426 	}
1427 
1428 	/* Make sure the VNET is stable. */
1429 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1430 	if (shutdown) {
1431 		CURVNET_RESTORE();
1432 		prison_free(pr);
1433 		return (EBUSY);
1434 	}
1435 	CURVNET_RESTORE();
1436 
1437 	found = if_unlink_ifnet(ifp, true);
1438 	MPASS(found);
1439 
1440 	/* Move the interface into the child jail/vnet. */
1441 	error = if_vmove(ifp, pr->pr_vnet);
1442 
1443 	/* Report the new if_xname back to the userland on success. */
1444 	if (error == 0)
1445 		sprintf(ifname, "%s", ifp->if_xname);
1446 
1447 	prison_free(pr);
1448 	return (error);
1449 }
1450 
1451 static int
1452 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1453 {
1454 	struct prison *pr;
1455 	struct vnet *vnet_dst;
1456 	struct ifnet *ifp;
1457 	int error, found;
1458  	bool shutdown;
1459 
1460 	/* Try to find the prison within our visibility. */
1461 	sx_slock(&allprison_lock);
1462 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1463 	sx_sunlock(&allprison_lock);
1464 	if (pr == NULL)
1465 		return (ENXIO);
1466 	prison_hold_locked(pr);
1467 	mtx_unlock(&pr->pr_mtx);
1468 
1469 	/* Make sure the named iface exists in the source prison/vnet. */
1470 	CURVNET_SET(pr->pr_vnet);
1471 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1472 	if (ifp == NULL) {
1473 		CURVNET_RESTORE();
1474 		prison_free(pr);
1475 		return (ENXIO);
1476 	}
1477 
1478 	/* Do not try to move the iface from and to the same prison. */
1479 	vnet_dst = TD_TO_VNET(td);
1480 	if (vnet_dst == ifp->if_vnet) {
1481 		CURVNET_RESTORE();
1482 		prison_free(pr);
1483 		return (EEXIST);
1484 	}
1485 
1486 	/* Make sure the VNET is stable. */
1487 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1488 	if (shutdown) {
1489 		CURVNET_RESTORE();
1490 		prison_free(pr);
1491 		return (EBUSY);
1492 	}
1493 
1494 	/* Get interface back from child jail/vnet. */
1495 	found = if_unlink_ifnet(ifp, true);
1496 	MPASS(found);
1497 	error = if_vmove(ifp, vnet_dst);
1498 	CURVNET_RESTORE();
1499 
1500 	/* Report the new if_xname back to the userland on success. */
1501 	if (error == 0)
1502 		sprintf(ifname, "%s", ifp->if_xname);
1503 
1504 	prison_free(pr);
1505 	return (error);
1506 }
1507 #endif /* VIMAGE */
1508 
1509 /*
1510  * Add a group to an interface
1511  */
1512 int
1513 if_addgroup(struct ifnet *ifp, const char *groupname)
1514 {
1515 	struct ifg_list		*ifgl;
1516 	struct ifg_group	*ifg = NULL;
1517 	struct ifg_member	*ifgm;
1518 	int 			 new = 0;
1519 
1520 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1521 	    groupname[strlen(groupname) - 1] <= '9')
1522 		return (EINVAL);
1523 
1524 	IFNET_WLOCK();
1525 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1526 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1527 			IFNET_WUNLOCK();
1528 			return (EEXIST);
1529 		}
1530 
1531 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1532 	    	IFNET_WUNLOCK();
1533 		return (ENOMEM);
1534 	}
1535 
1536 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1537 		free(ifgl, M_TEMP);
1538 		IFNET_WUNLOCK();
1539 		return (ENOMEM);
1540 	}
1541 
1542 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1543 		if (!strcmp(ifg->ifg_group, groupname))
1544 			break;
1545 
1546 	if (ifg == NULL) {
1547 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1548 			free(ifgl, M_TEMP);
1549 			free(ifgm, M_TEMP);
1550 			IFNET_WUNLOCK();
1551 			return (ENOMEM);
1552 		}
1553 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1554 		ifg->ifg_refcnt = 0;
1555 		CK_STAILQ_INIT(&ifg->ifg_members);
1556 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1557 		new = 1;
1558 	}
1559 
1560 	ifg->ifg_refcnt++;
1561 	ifgl->ifgl_group = ifg;
1562 	ifgm->ifgm_ifp = ifp;
1563 
1564 	IF_ADDR_WLOCK(ifp);
1565 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1566 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1567 	IF_ADDR_WUNLOCK(ifp);
1568 
1569 	IFNET_WUNLOCK();
1570 
1571 	if (new)
1572 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1573 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1574 
1575 	return (0);
1576 }
1577 
1578 /*
1579  * Helper function to remove a group out of an interface.  Expects the global
1580  * ifnet lock to be write-locked, and drops it before returning.
1581  */
1582 static void
1583 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1584     const char *groupname)
1585 {
1586 	struct ifg_member *ifgm;
1587 	bool freeifgl;
1588 
1589 	IFNET_WLOCK_ASSERT();
1590 
1591 	IF_ADDR_WLOCK(ifp);
1592 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1593 	IF_ADDR_WUNLOCK(ifp);
1594 
1595 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1596 		if (ifgm->ifgm_ifp == ifp) {
1597 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1598 			    ifg_member, ifgm_next);
1599 			break;
1600 		}
1601 	}
1602 
1603 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1604 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1605 		    ifg_next);
1606 		freeifgl = true;
1607 	} else {
1608 		freeifgl = false;
1609 	}
1610 	IFNET_WUNLOCK();
1611 
1612 	epoch_wait_preempt(net_epoch_preempt);
1613 	if (freeifgl) {
1614 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1615 		free(ifgl->ifgl_group, M_TEMP);
1616 	}
1617 	free(ifgm, M_TEMP);
1618 	free(ifgl, M_TEMP);
1619 
1620 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1621 }
1622 
1623 /*
1624  * Remove a group from an interface
1625  */
1626 int
1627 if_delgroup(struct ifnet *ifp, const char *groupname)
1628 {
1629 	struct ifg_list *ifgl;
1630 
1631 	IFNET_WLOCK();
1632 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1633 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1634 			break;
1635 	if (ifgl == NULL) {
1636 		IFNET_WUNLOCK();
1637 		return (ENOENT);
1638 	}
1639 
1640 	_if_delgroup_locked(ifp, ifgl, groupname);
1641 
1642 	return (0);
1643 }
1644 
1645 /*
1646  * Remove an interface from all groups
1647  */
1648 static void
1649 if_delgroups(struct ifnet *ifp)
1650 {
1651 	struct ifg_list *ifgl;
1652 	char groupname[IFNAMSIZ];
1653 
1654 	IFNET_WLOCK();
1655 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1656 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1657 		_if_delgroup_locked(ifp, ifgl, groupname);
1658 		IFNET_WLOCK();
1659 	}
1660 	IFNET_WUNLOCK();
1661 }
1662 
1663 /*
1664  * Stores all groups from an interface in memory pointed to by ifgr.
1665  */
1666 static int
1667 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1668 {
1669 	int			 len, error;
1670 	struct ifg_list		*ifgl;
1671 	struct ifg_req		 ifgrq, *ifgp;
1672 
1673 	NET_EPOCH_ASSERT();
1674 
1675 	if (ifgr->ifgr_len == 0) {
1676 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1677 			ifgr->ifgr_len += sizeof(struct ifg_req);
1678 		return (0);
1679 	}
1680 
1681 	len = ifgr->ifgr_len;
1682 	ifgp = ifgr->ifgr_groups;
1683 	/* XXX: wire */
1684 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1685 		if (len < sizeof(ifgrq))
1686 			return (EINVAL);
1687 		bzero(&ifgrq, sizeof ifgrq);
1688 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1689 		    sizeof(ifgrq.ifgrq_group));
1690 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1691 			return (error);
1692 		len -= sizeof(ifgrq);
1693 		ifgp++;
1694 	}
1695 
1696 	return (0);
1697 }
1698 
1699 /*
1700  * Stores all members of a group in memory pointed to by igfr
1701  */
1702 static int
1703 if_getgroupmembers(struct ifgroupreq *ifgr)
1704 {
1705 	struct ifg_group	*ifg;
1706 	struct ifg_member	*ifgm;
1707 	struct ifg_req		 ifgrq, *ifgp;
1708 	int			 len, error;
1709 
1710 	IFNET_RLOCK();
1711 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1712 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1713 			break;
1714 	if (ifg == NULL) {
1715 		IFNET_RUNLOCK();
1716 		return (ENOENT);
1717 	}
1718 
1719 	if (ifgr->ifgr_len == 0) {
1720 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1721 			ifgr->ifgr_len += sizeof(ifgrq);
1722 		IFNET_RUNLOCK();
1723 		return (0);
1724 	}
1725 
1726 	len = ifgr->ifgr_len;
1727 	ifgp = ifgr->ifgr_groups;
1728 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1729 		if (len < sizeof(ifgrq)) {
1730 			IFNET_RUNLOCK();
1731 			return (EINVAL);
1732 		}
1733 		bzero(&ifgrq, sizeof ifgrq);
1734 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1735 		    sizeof(ifgrq.ifgrq_member));
1736 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1737 			IFNET_RUNLOCK();
1738 			return (error);
1739 		}
1740 		len -= sizeof(ifgrq);
1741 		ifgp++;
1742 	}
1743 	IFNET_RUNLOCK();
1744 
1745 	return (0);
1746 }
1747 
1748 /*
1749  * Return counter values from counter(9)s stored in ifnet.
1750  */
1751 uint64_t
1752 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1753 {
1754 
1755 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1756 
1757 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1758 }
1759 
1760 /*
1761  * Increase an ifnet counter. Usually used for counters shared
1762  * between the stack and a driver, but function supports them all.
1763  */
1764 void
1765 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1766 {
1767 
1768 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1769 
1770 	counter_u64_add(ifp->if_counters[cnt], inc);
1771 }
1772 
1773 /*
1774  * Copy data from ifnet to userland API structure if_data.
1775  */
1776 void
1777 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1778 {
1779 
1780 	ifd->ifi_type = ifp->if_type;
1781 	ifd->ifi_physical = 0;
1782 	ifd->ifi_addrlen = ifp->if_addrlen;
1783 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1784 	ifd->ifi_link_state = ifp->if_link_state;
1785 	ifd->ifi_vhid = 0;
1786 	ifd->ifi_datalen = sizeof(struct if_data);
1787 	ifd->ifi_mtu = ifp->if_mtu;
1788 	ifd->ifi_metric = ifp->if_metric;
1789 	ifd->ifi_baudrate = ifp->if_baudrate;
1790 	ifd->ifi_hwassist = ifp->if_hwassist;
1791 	ifd->ifi_epoch = ifp->if_epoch;
1792 	ifd->ifi_lastchange = ifp->if_lastchange;
1793 
1794 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1795 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1796 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1797 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1798 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1799 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1800 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1801 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1802 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1803 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1804 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1805 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1806 }
1807 
1808 /*
1809  * Initialization, destruction and refcounting functions for ifaddrs.
1810  */
1811 struct ifaddr *
1812 ifa_alloc(size_t size, int flags)
1813 {
1814 	struct ifaddr *ifa;
1815 
1816 	KASSERT(size >= sizeof(struct ifaddr),
1817 	    ("%s: invalid size %zu", __func__, size));
1818 
1819 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1820 	if (ifa == NULL)
1821 		return (NULL);
1822 
1823 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1824 		goto fail;
1825 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1826 		goto fail;
1827 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1828 		goto fail;
1829 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1830 		goto fail;
1831 
1832 	refcount_init(&ifa->ifa_refcnt, 1);
1833 
1834 	return (ifa);
1835 
1836 fail:
1837 	/* free(NULL) is okay */
1838 	counter_u64_free(ifa->ifa_opackets);
1839 	counter_u64_free(ifa->ifa_ipackets);
1840 	counter_u64_free(ifa->ifa_obytes);
1841 	counter_u64_free(ifa->ifa_ibytes);
1842 	free(ifa, M_IFADDR);
1843 
1844 	return (NULL);
1845 }
1846 
1847 void
1848 ifa_ref(struct ifaddr *ifa)
1849 {
1850 	u_int old;
1851 
1852 	old = refcount_acquire(&ifa->ifa_refcnt);
1853 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1854 }
1855 
1856 int
1857 ifa_try_ref(struct ifaddr *ifa)
1858 {
1859 
1860 	NET_EPOCH_ASSERT();
1861 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1862 }
1863 
1864 static void
1865 ifa_destroy(epoch_context_t ctx)
1866 {
1867 	struct ifaddr *ifa;
1868 
1869 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1870 	counter_u64_free(ifa->ifa_opackets);
1871 	counter_u64_free(ifa->ifa_ipackets);
1872 	counter_u64_free(ifa->ifa_obytes);
1873 	counter_u64_free(ifa->ifa_ibytes);
1874 	free(ifa, M_IFADDR);
1875 }
1876 
1877 void
1878 ifa_free(struct ifaddr *ifa)
1879 {
1880 
1881 	if (refcount_release(&ifa->ifa_refcnt))
1882 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1883 }
1884 
1885 /*
1886  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1887  * structs used to represent other address families, it is necessary
1888  * to perform a different comparison.
1889  */
1890 
1891 #define	sa_dl_equal(a1, a2)	\
1892 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1893 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1894 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1895 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1896 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1897 
1898 /*
1899  * Locate an interface based on a complete address.
1900  */
1901 /*ARGSUSED*/
1902 struct ifaddr *
1903 ifa_ifwithaddr(const struct sockaddr *addr)
1904 {
1905 	struct ifnet *ifp;
1906 	struct ifaddr *ifa;
1907 
1908 	NET_EPOCH_ASSERT();
1909 
1910 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1911 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1912 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1913 				continue;
1914 			if (sa_equal(addr, ifa->ifa_addr)) {
1915 				goto done;
1916 			}
1917 			/* IP6 doesn't have broadcast */
1918 			if ((ifp->if_flags & IFF_BROADCAST) &&
1919 			    ifa->ifa_broadaddr &&
1920 			    ifa->ifa_broadaddr->sa_len != 0 &&
1921 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1922 				goto done;
1923 			}
1924 		}
1925 	}
1926 	ifa = NULL;
1927 done:
1928 	return (ifa);
1929 }
1930 
1931 int
1932 ifa_ifwithaddr_check(const struct sockaddr *addr)
1933 {
1934 	struct epoch_tracker et;
1935 	int rc;
1936 
1937 	NET_EPOCH_ENTER(et);
1938 	rc = (ifa_ifwithaddr(addr) != NULL);
1939 	NET_EPOCH_EXIT(et);
1940 	return (rc);
1941 }
1942 
1943 /*
1944  * Locate an interface based on the broadcast address.
1945  */
1946 /* ARGSUSED */
1947 struct ifaddr *
1948 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1949 {
1950 	struct ifnet *ifp;
1951 	struct ifaddr *ifa;
1952 
1953 	NET_EPOCH_ASSERT();
1954 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1955 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1956 			continue;
1957 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1958 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1959 				continue;
1960 			if ((ifp->if_flags & IFF_BROADCAST) &&
1961 			    ifa->ifa_broadaddr &&
1962 			    ifa->ifa_broadaddr->sa_len != 0 &&
1963 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1964 				goto done;
1965 			}
1966 		}
1967 	}
1968 	ifa = NULL;
1969 done:
1970 	return (ifa);
1971 }
1972 
1973 /*
1974  * Locate the point to point interface with a given destination address.
1975  */
1976 /*ARGSUSED*/
1977 struct ifaddr *
1978 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1979 {
1980 	struct ifnet *ifp;
1981 	struct ifaddr *ifa;
1982 
1983 	NET_EPOCH_ASSERT();
1984 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1985 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1986 			continue;
1987 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1988 			continue;
1989 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1990 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1991 				continue;
1992 			if (ifa->ifa_dstaddr != NULL &&
1993 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1994 				goto done;
1995 			}
1996 		}
1997 	}
1998 	ifa = NULL;
1999 done:
2000 	return (ifa);
2001 }
2002 
2003 /*
2004  * Find an interface on a specific network.  If many, choice
2005  * is most specific found.
2006  */
2007 struct ifaddr *
2008 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2009 {
2010 	struct ifnet *ifp;
2011 	struct ifaddr *ifa;
2012 	struct ifaddr *ifa_maybe = NULL;
2013 	u_int af = addr->sa_family;
2014 	const char *addr_data = addr->sa_data, *cplim;
2015 
2016 	NET_EPOCH_ASSERT();
2017 	/*
2018 	 * AF_LINK addresses can be looked up directly by their index number,
2019 	 * so do that if we can.
2020 	 */
2021 	if (af == AF_LINK) {
2022 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2023 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2024 		return (ifaddr_byindex(sdl->sdl_index));
2025 	}
2026 
2027 	/*
2028 	 * Scan though each interface, looking for ones that have addresses
2029 	 * in this address family and the requested fib.
2030 	 */
2031 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2032 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2033 			continue;
2034 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2035 			const char *cp, *cp2, *cp3;
2036 
2037 			if (ifa->ifa_addr->sa_family != af)
2038 next:				continue;
2039 			if (af == AF_INET &&
2040 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2041 				/*
2042 				 * This is a bit broken as it doesn't
2043 				 * take into account that the remote end may
2044 				 * be a single node in the network we are
2045 				 * looking for.
2046 				 * The trouble is that we don't know the
2047 				 * netmask for the remote end.
2048 				 */
2049 				if (ifa->ifa_dstaddr != NULL &&
2050 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2051 					goto done;
2052 				}
2053 			} else {
2054 				/*
2055 				 * Scan all the bits in the ifa's address.
2056 				 * If a bit dissagrees with what we are
2057 				 * looking for, mask it with the netmask
2058 				 * to see if it really matters.
2059 				 * (A byte at a time)
2060 				 */
2061 				if (ifa->ifa_netmask == 0)
2062 					continue;
2063 				cp = addr_data;
2064 				cp2 = ifa->ifa_addr->sa_data;
2065 				cp3 = ifa->ifa_netmask->sa_data;
2066 				cplim = ifa->ifa_netmask->sa_len
2067 					+ (char *)ifa->ifa_netmask;
2068 				while (cp3 < cplim)
2069 					if ((*cp++ ^ *cp2++) & *cp3++)
2070 						goto next; /* next address! */
2071 				/*
2072 				 * If the netmask of what we just found
2073 				 * is more specific than what we had before
2074 				 * (if we had one), or if the virtual status
2075 				 * of new prefix is better than of the old one,
2076 				 * then remember the new one before continuing
2077 				 * to search for an even better one.
2078 				 */
2079 				if (ifa_maybe == NULL ||
2080 				    ifa_preferred(ifa_maybe, ifa) ||
2081 				    rn_refines((caddr_t)ifa->ifa_netmask,
2082 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2083 					ifa_maybe = ifa;
2084 				}
2085 			}
2086 		}
2087 	}
2088 	ifa = ifa_maybe;
2089 	ifa_maybe = NULL;
2090 done:
2091 	return (ifa);
2092 }
2093 
2094 /*
2095  * Find an interface address specific to an interface best matching
2096  * a given address.
2097  */
2098 struct ifaddr *
2099 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2100 {
2101 	struct ifaddr *ifa;
2102 	const char *cp, *cp2, *cp3;
2103 	char *cplim;
2104 	struct ifaddr *ifa_maybe = NULL;
2105 	u_int af = addr->sa_family;
2106 
2107 	if (af >= AF_MAX)
2108 		return (NULL);
2109 
2110 	NET_EPOCH_ASSERT();
2111 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2112 		if (ifa->ifa_addr->sa_family != af)
2113 			continue;
2114 		if (ifa_maybe == NULL)
2115 			ifa_maybe = ifa;
2116 		if (ifa->ifa_netmask == 0) {
2117 			if (sa_equal(addr, ifa->ifa_addr) ||
2118 			    (ifa->ifa_dstaddr &&
2119 			    sa_equal(addr, ifa->ifa_dstaddr)))
2120 				goto done;
2121 			continue;
2122 		}
2123 		if (ifp->if_flags & IFF_POINTOPOINT) {
2124 			if (sa_equal(addr, ifa->ifa_dstaddr))
2125 				goto done;
2126 		} else {
2127 			cp = addr->sa_data;
2128 			cp2 = ifa->ifa_addr->sa_data;
2129 			cp3 = ifa->ifa_netmask->sa_data;
2130 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2131 			for (; cp3 < cplim; cp3++)
2132 				if ((*cp++ ^ *cp2++) & *cp3)
2133 					break;
2134 			if (cp3 == cplim)
2135 				goto done;
2136 		}
2137 	}
2138 	ifa = ifa_maybe;
2139 done:
2140 	return (ifa);
2141 }
2142 
2143 /*
2144  * See whether new ifa is better than current one:
2145  * 1) A non-virtual one is preferred over virtual.
2146  * 2) A virtual in master state preferred over any other state.
2147  *
2148  * Used in several address selecting functions.
2149  */
2150 int
2151 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2152 {
2153 
2154 	return (cur->ifa_carp && (!next->ifa_carp ||
2155 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2156 }
2157 
2158 struct sockaddr_dl *
2159 link_alloc_sdl(size_t size, int flags)
2160 {
2161 
2162 	return (malloc(size, M_TEMP, flags));
2163 }
2164 
2165 void
2166 link_free_sdl(struct sockaddr *sa)
2167 {
2168 	free(sa, M_TEMP);
2169 }
2170 
2171 /*
2172  * Fills in given sdl with interface basic info.
2173  * Returns pointer to filled sdl.
2174  */
2175 struct sockaddr_dl *
2176 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2177 {
2178 	struct sockaddr_dl *sdl;
2179 
2180 	sdl = (struct sockaddr_dl *)paddr;
2181 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2182 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2183 	sdl->sdl_family = AF_LINK;
2184 	sdl->sdl_index = ifp->if_index;
2185 	sdl->sdl_type = iftype;
2186 
2187 	return (sdl);
2188 }
2189 
2190 /*
2191  * Mark an interface down and notify protocols of
2192  * the transition.
2193  */
2194 static void
2195 if_unroute(struct ifnet *ifp, int flag, int fam)
2196 {
2197 	struct ifaddr *ifa;
2198 
2199 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2200 
2201 	ifp->if_flags &= ~flag;
2202 	getmicrotime(&ifp->if_lastchange);
2203 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2204 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2205 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2206 	ifp->if_qflush(ifp);
2207 
2208 	if (ifp->if_carp)
2209 		(*carp_linkstate_p)(ifp);
2210 	rt_ifmsg(ifp);
2211 }
2212 
2213 /*
2214  * Mark an interface up and notify protocols of
2215  * the transition.
2216  */
2217 static void
2218 if_route(struct ifnet *ifp, int flag, int fam)
2219 {
2220 	struct ifaddr *ifa;
2221 
2222 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2223 
2224 	ifp->if_flags |= flag;
2225 	getmicrotime(&ifp->if_lastchange);
2226 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2227 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2228 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2229 	if (ifp->if_carp)
2230 		(*carp_linkstate_p)(ifp);
2231 	rt_ifmsg(ifp);
2232 #ifdef INET6
2233 	in6_if_up(ifp);
2234 #endif
2235 }
2236 
2237 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2238 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2239 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2240 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2241 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2242 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2243 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2244 void	*(*vlan_cookie_p)(struct ifnet *);
2245 
2246 /*
2247  * Handle a change in the interface link state. To avoid LORs
2248  * between driver lock and upper layer locks, as well as possible
2249  * recursions, we post event to taskqueue, and all job
2250  * is done in static do_link_state_change().
2251  */
2252 void
2253 if_link_state_change(struct ifnet *ifp, int link_state)
2254 {
2255 	/* Return if state hasn't changed. */
2256 	if (ifp->if_link_state == link_state)
2257 		return;
2258 
2259 	ifp->if_link_state = link_state;
2260 
2261 	/* XXXGL: reference ifp? */
2262 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2263 }
2264 
2265 static void
2266 do_link_state_change(void *arg, int pending)
2267 {
2268 	struct ifnet *ifp;
2269 	int link_state;
2270 
2271 	ifp = arg;
2272 	link_state = ifp->if_link_state;
2273 
2274 	CURVNET_SET(ifp->if_vnet);
2275 	rt_ifmsg(ifp);
2276 	if (ifp->if_vlantrunk != NULL)
2277 		(*vlan_link_state_p)(ifp);
2278 
2279 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2280 	    ifp->if_l2com != NULL)
2281 		(*ng_ether_link_state_p)(ifp, link_state);
2282 	if (ifp->if_carp)
2283 		(*carp_linkstate_p)(ifp);
2284 	if (ifp->if_bridge)
2285 		ifp->if_bridge_linkstate(ifp);
2286 	if (ifp->if_lagg)
2287 		(*lagg_linkstate_p)(ifp, link_state);
2288 
2289 	if (IS_DEFAULT_VNET(curvnet))
2290 		devctl_notify("IFNET", ifp->if_xname,
2291 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2292 		    NULL);
2293 	if (pending > 1)
2294 		if_printf(ifp, "%d link states coalesced\n", pending);
2295 	if (log_link_state_change)
2296 		if_printf(ifp, "link state changed to %s\n",
2297 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2298 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2299 	CURVNET_RESTORE();
2300 }
2301 
2302 /*
2303  * Mark an interface down and notify protocols of
2304  * the transition.
2305  */
2306 void
2307 if_down(struct ifnet *ifp)
2308 {
2309 
2310 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2311 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2312 }
2313 
2314 /*
2315  * Mark an interface up and notify protocols of
2316  * the transition.
2317  */
2318 void
2319 if_up(struct ifnet *ifp)
2320 {
2321 
2322 	if_route(ifp, IFF_UP, AF_UNSPEC);
2323 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2324 }
2325 
2326 /*
2327  * Flush an interface queue.
2328  */
2329 void
2330 if_qflush(struct ifnet *ifp)
2331 {
2332 	struct mbuf *m, *n;
2333 	struct ifaltq *ifq;
2334 
2335 	ifq = &ifp->if_snd;
2336 	IFQ_LOCK(ifq);
2337 #ifdef ALTQ
2338 	if (ALTQ_IS_ENABLED(ifq))
2339 		ALTQ_PURGE(ifq);
2340 #endif
2341 	n = ifq->ifq_head;
2342 	while ((m = n) != NULL) {
2343 		n = m->m_nextpkt;
2344 		m_freem(m);
2345 	}
2346 	ifq->ifq_head = 0;
2347 	ifq->ifq_tail = 0;
2348 	ifq->ifq_len = 0;
2349 	IFQ_UNLOCK(ifq);
2350 }
2351 
2352 /*
2353  * Map interface name to interface structure pointer, with or without
2354  * returning a reference.
2355  */
2356 struct ifnet *
2357 ifunit_ref(const char *name)
2358 {
2359 	struct epoch_tracker et;
2360 	struct ifnet *ifp;
2361 
2362 	NET_EPOCH_ENTER(et);
2363 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2364 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2365 		    !(ifp->if_flags & IFF_DYING))
2366 			break;
2367 	}
2368 	if (ifp != NULL)
2369 		if_ref(ifp);
2370 	NET_EPOCH_EXIT(et);
2371 	return (ifp);
2372 }
2373 
2374 struct ifnet *
2375 ifunit(const char *name)
2376 {
2377 	struct epoch_tracker et;
2378 	struct ifnet *ifp;
2379 
2380 	NET_EPOCH_ENTER(et);
2381 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2382 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2383 			break;
2384 	}
2385 	NET_EPOCH_EXIT(et);
2386 	return (ifp);
2387 }
2388 
2389 void *
2390 ifr_buffer_get_buffer(void *data)
2391 {
2392 	union ifreq_union *ifrup;
2393 
2394 	ifrup = data;
2395 #ifdef COMPAT_FREEBSD32
2396 	if (SV_CURPROC_FLAG(SV_ILP32))
2397 		return ((void *)(uintptr_t)
2398 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2399 #endif
2400 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2401 }
2402 
2403 static void
2404 ifr_buffer_set_buffer_null(void *data)
2405 {
2406 	union ifreq_union *ifrup;
2407 
2408 	ifrup = data;
2409 #ifdef COMPAT_FREEBSD32
2410 	if (SV_CURPROC_FLAG(SV_ILP32))
2411 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2412 	else
2413 #endif
2414 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2415 }
2416 
2417 size_t
2418 ifr_buffer_get_length(void *data)
2419 {
2420 	union ifreq_union *ifrup;
2421 
2422 	ifrup = data;
2423 #ifdef COMPAT_FREEBSD32
2424 	if (SV_CURPROC_FLAG(SV_ILP32))
2425 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2426 #endif
2427 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2428 }
2429 
2430 static void
2431 ifr_buffer_set_length(void *data, size_t len)
2432 {
2433 	union ifreq_union *ifrup;
2434 
2435 	ifrup = data;
2436 #ifdef COMPAT_FREEBSD32
2437 	if (SV_CURPROC_FLAG(SV_ILP32))
2438 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2439 	else
2440 #endif
2441 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2442 }
2443 
2444 void *
2445 ifr_data_get_ptr(void *ifrp)
2446 {
2447 	union ifreq_union *ifrup;
2448 
2449 	ifrup = ifrp;
2450 #ifdef COMPAT_FREEBSD32
2451 	if (SV_CURPROC_FLAG(SV_ILP32))
2452 		return ((void *)(uintptr_t)
2453 		    ifrup->ifr32.ifr_ifru.ifru_data);
2454 #endif
2455 		return (ifrup->ifr.ifr_ifru.ifru_data);
2456 }
2457 
2458 /*
2459  * Hardware specific interface ioctls.
2460  */
2461 int
2462 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2463 {
2464 	struct ifreq *ifr;
2465 	int error = 0, do_ifup = 0;
2466 	int new_flags, temp_flags;
2467 	size_t namelen, onamelen;
2468 	size_t descrlen;
2469 	char *descrbuf, *odescrbuf;
2470 	char new_name[IFNAMSIZ];
2471 	char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8];
2472 	struct ifaddr *ifa;
2473 	struct sockaddr_dl *sdl;
2474 
2475 	ifr = (struct ifreq *)data;
2476 	switch (cmd) {
2477 	case SIOCGIFINDEX:
2478 		ifr->ifr_index = ifp->if_index;
2479 		break;
2480 
2481 	case SIOCGIFFLAGS:
2482 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2483 		ifr->ifr_flags = temp_flags & 0xffff;
2484 		ifr->ifr_flagshigh = temp_flags >> 16;
2485 		break;
2486 
2487 	case SIOCGIFCAP:
2488 		ifr->ifr_reqcap = ifp->if_capabilities;
2489 		ifr->ifr_curcap = ifp->if_capenable;
2490 		break;
2491 
2492 	case SIOCGIFDATA:
2493 	{
2494 		struct if_data ifd;
2495 
2496 		/* Ensure uninitialised padding is not leaked. */
2497 		memset(&ifd, 0, sizeof(ifd));
2498 
2499 		if_data_copy(ifp, &ifd);
2500 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2501 		break;
2502 	}
2503 
2504 #ifdef MAC
2505 	case SIOCGIFMAC:
2506 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2507 		break;
2508 #endif
2509 
2510 	case SIOCGIFMETRIC:
2511 		ifr->ifr_metric = ifp->if_metric;
2512 		break;
2513 
2514 	case SIOCGIFMTU:
2515 		ifr->ifr_mtu = ifp->if_mtu;
2516 		break;
2517 
2518 	case SIOCGIFPHYS:
2519 		/* XXXGL: did this ever worked? */
2520 		ifr->ifr_phys = 0;
2521 		break;
2522 
2523 	case SIOCGIFDESCR:
2524 		error = 0;
2525 		sx_slock(&ifdescr_sx);
2526 		if (ifp->if_description == NULL)
2527 			error = ENOMSG;
2528 		else {
2529 			/* space for terminating nul */
2530 			descrlen = strlen(ifp->if_description) + 1;
2531 			if (ifr_buffer_get_length(ifr) < descrlen)
2532 				ifr_buffer_set_buffer_null(ifr);
2533 			else
2534 				error = copyout(ifp->if_description,
2535 				    ifr_buffer_get_buffer(ifr), descrlen);
2536 			ifr_buffer_set_length(ifr, descrlen);
2537 		}
2538 		sx_sunlock(&ifdescr_sx);
2539 		break;
2540 
2541 	case SIOCSIFDESCR:
2542 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2543 		if (error)
2544 			return (error);
2545 
2546 		/*
2547 		 * Copy only (length-1) bytes to make sure that
2548 		 * if_description is always nul terminated.  The
2549 		 * length parameter is supposed to count the
2550 		 * terminating nul in.
2551 		 */
2552 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2553 			return (ENAMETOOLONG);
2554 		else if (ifr_buffer_get_length(ifr) == 0)
2555 			descrbuf = NULL;
2556 		else {
2557 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2558 			    M_IFDESCR, M_WAITOK | M_ZERO);
2559 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2560 			    ifr_buffer_get_length(ifr) - 1);
2561 			if (error) {
2562 				free(descrbuf, M_IFDESCR);
2563 				break;
2564 			}
2565 		}
2566 
2567 		sx_xlock(&ifdescr_sx);
2568 		odescrbuf = ifp->if_description;
2569 		ifp->if_description = descrbuf;
2570 		sx_xunlock(&ifdescr_sx);
2571 
2572 		getmicrotime(&ifp->if_lastchange);
2573 		free(odescrbuf, M_IFDESCR);
2574 		break;
2575 
2576 	case SIOCGIFFIB:
2577 		ifr->ifr_fib = ifp->if_fib;
2578 		break;
2579 
2580 	case SIOCSIFFIB:
2581 		error = priv_check(td, PRIV_NET_SETIFFIB);
2582 		if (error)
2583 			return (error);
2584 		if (ifr->ifr_fib >= rt_numfibs)
2585 			return (EINVAL);
2586 
2587 		ifp->if_fib = ifr->ifr_fib;
2588 		break;
2589 
2590 	case SIOCSIFFLAGS:
2591 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2592 		if (error)
2593 			return (error);
2594 		/*
2595 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2596 		 * check, so we don't need special handling here yet.
2597 		 */
2598 		new_flags = (ifr->ifr_flags & 0xffff) |
2599 		    (ifr->ifr_flagshigh << 16);
2600 		if (ifp->if_flags & IFF_UP &&
2601 		    (new_flags & IFF_UP) == 0) {
2602 			if_down(ifp);
2603 		} else if (new_flags & IFF_UP &&
2604 		    (ifp->if_flags & IFF_UP) == 0) {
2605 			do_ifup = 1;
2606 		}
2607 		/* See if permanently promiscuous mode bit is about to flip */
2608 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2609 			if (new_flags & IFF_PPROMISC)
2610 				ifp->if_flags |= IFF_PROMISC;
2611 			else if (ifp->if_pcount == 0)
2612 				ifp->if_flags &= ~IFF_PROMISC;
2613 			if (log_promisc_mode_change)
2614                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2615                                     ((new_flags & IFF_PPROMISC) ?
2616                                      "enabled" : "disabled"));
2617 		}
2618 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2619 			(new_flags &~ IFF_CANTCHANGE);
2620 		if (ifp->if_ioctl) {
2621 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2622 		}
2623 		if (do_ifup)
2624 			if_up(ifp);
2625 		getmicrotime(&ifp->if_lastchange);
2626 		break;
2627 
2628 	case SIOCSIFCAP:
2629 		error = priv_check(td, PRIV_NET_SETIFCAP);
2630 		if (error)
2631 			return (error);
2632 		if (ifp->if_ioctl == NULL)
2633 			return (EOPNOTSUPP);
2634 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2635 			return (EINVAL);
2636 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2637 		if (error == 0)
2638 			getmicrotime(&ifp->if_lastchange);
2639 		break;
2640 
2641 #ifdef MAC
2642 	case SIOCSIFMAC:
2643 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2644 		break;
2645 #endif
2646 
2647 	case SIOCSIFNAME:
2648 		error = priv_check(td, PRIV_NET_SETIFNAME);
2649 		if (error)
2650 			return (error);
2651 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2652 		    NULL);
2653 		if (error != 0)
2654 			return (error);
2655 		if (new_name[0] == '\0')
2656 			return (EINVAL);
2657 		if (strcmp(new_name, ifp->if_xname) == 0)
2658 			break;
2659 		if (ifunit(new_name) != NULL)
2660 			return (EEXIST);
2661 
2662 		/*
2663 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2664 		 * and there are numerous other races with the
2665 		 * ifunit() checks not being atomic with namespace
2666 		 * changes (renames, vmoves, if_attach, etc).
2667 		 */
2668 		ifp->if_flags |= IFF_RENAMING;
2669 
2670 		/* Announce the departure of the interface. */
2671 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2672 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2673 
2674 		if_printf(ifp, "changing name to '%s'\n", new_name);
2675 
2676 		IF_ADDR_WLOCK(ifp);
2677 		strlcpy(old_name, ifp->if_xname, sizeof(old_name));
2678 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2679 		ifa = ifp->if_addr;
2680 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2681 		namelen = strlen(new_name);
2682 		onamelen = sdl->sdl_nlen;
2683 		/*
2684 		 * Move the address if needed.  This is safe because we
2685 		 * allocate space for a name of length IFNAMSIZ when we
2686 		 * create this in if_attach().
2687 		 */
2688 		if (namelen != onamelen) {
2689 			bcopy(sdl->sdl_data + onamelen,
2690 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2691 		}
2692 		bcopy(new_name, sdl->sdl_data, namelen);
2693 		sdl->sdl_nlen = namelen;
2694 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2695 		bzero(sdl->sdl_data, onamelen);
2696 		while (namelen != 0)
2697 			sdl->sdl_data[--namelen] = 0xff;
2698 		IF_ADDR_WUNLOCK(ifp);
2699 
2700 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2701 		/* Announce the return of the interface. */
2702 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2703 
2704 		ifp->if_flags &= ~IFF_RENAMING;
2705 
2706 		snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
2707 		devctl_notify("IFNET", old_name, "RENAME", strbuf);
2708 		break;
2709 
2710 #ifdef VIMAGE
2711 	case SIOCSIFVNET:
2712 		error = priv_check(td, PRIV_NET_SETIFVNET);
2713 		if (error)
2714 			return (error);
2715 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2716 		break;
2717 #endif
2718 
2719 	case SIOCSIFMETRIC:
2720 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2721 		if (error)
2722 			return (error);
2723 		ifp->if_metric = ifr->ifr_metric;
2724 		getmicrotime(&ifp->if_lastchange);
2725 		break;
2726 
2727 	case SIOCSIFPHYS:
2728 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2729 		if (error)
2730 			return (error);
2731 		if (ifp->if_ioctl == NULL)
2732 			return (EOPNOTSUPP);
2733 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2734 		if (error == 0)
2735 			getmicrotime(&ifp->if_lastchange);
2736 		break;
2737 
2738 	case SIOCSIFMTU:
2739 	{
2740 		u_long oldmtu = ifp->if_mtu;
2741 
2742 		error = priv_check(td, PRIV_NET_SETIFMTU);
2743 		if (error)
2744 			return (error);
2745 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2746 			return (EINVAL);
2747 		if (ifp->if_ioctl == NULL)
2748 			return (EOPNOTSUPP);
2749 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2750 		if (error == 0) {
2751 			getmicrotime(&ifp->if_lastchange);
2752 			rt_ifmsg(ifp);
2753 #ifdef INET
2754 			DEBUGNET_NOTIFY_MTU(ifp);
2755 #endif
2756 		}
2757 		/*
2758 		 * If the link MTU changed, do network layer specific procedure.
2759 		 */
2760 		if (ifp->if_mtu != oldmtu) {
2761 #ifdef INET6
2762 			nd6_setmtu(ifp);
2763 #endif
2764 			rt_updatemtu(ifp);
2765 		}
2766 		break;
2767 	}
2768 
2769 	case SIOCADDMULTI:
2770 	case SIOCDELMULTI:
2771 		if (cmd == SIOCADDMULTI)
2772 			error = priv_check(td, PRIV_NET_ADDMULTI);
2773 		else
2774 			error = priv_check(td, PRIV_NET_DELMULTI);
2775 		if (error)
2776 			return (error);
2777 
2778 		/* Don't allow group membership on non-multicast interfaces. */
2779 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2780 			return (EOPNOTSUPP);
2781 
2782 		/* Don't let users screw up protocols' entries. */
2783 		if (ifr->ifr_addr.sa_family != AF_LINK)
2784 			return (EINVAL);
2785 
2786 		if (cmd == SIOCADDMULTI) {
2787 			struct epoch_tracker et;
2788 			struct ifmultiaddr *ifma;
2789 
2790 			/*
2791 			 * Userland is only permitted to join groups once
2792 			 * via the if_addmulti() KPI, because it cannot hold
2793 			 * struct ifmultiaddr * between calls. It may also
2794 			 * lose a race while we check if the membership
2795 			 * already exists.
2796 			 */
2797 			NET_EPOCH_ENTER(et);
2798 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2799 			NET_EPOCH_EXIT(et);
2800 			if (ifma != NULL)
2801 				error = EADDRINUSE;
2802 			else
2803 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2804 		} else {
2805 			error = if_delmulti(ifp, &ifr->ifr_addr);
2806 		}
2807 		if (error == 0)
2808 			getmicrotime(&ifp->if_lastchange);
2809 		break;
2810 
2811 	case SIOCSIFPHYADDR:
2812 	case SIOCDIFPHYADDR:
2813 #ifdef INET6
2814 	case SIOCSIFPHYADDR_IN6:
2815 #endif
2816 	case SIOCSIFMEDIA:
2817 	case SIOCSIFGENERIC:
2818 		error = priv_check(td, PRIV_NET_HWIOCTL);
2819 		if (error)
2820 			return (error);
2821 		if (ifp->if_ioctl == NULL)
2822 			return (EOPNOTSUPP);
2823 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2824 		if (error == 0)
2825 			getmicrotime(&ifp->if_lastchange);
2826 		break;
2827 
2828 	case SIOCGIFSTATUS:
2829 	case SIOCGIFPSRCADDR:
2830 	case SIOCGIFPDSTADDR:
2831 	case SIOCGIFMEDIA:
2832 	case SIOCGIFXMEDIA:
2833 	case SIOCGIFGENERIC:
2834 	case SIOCGIFRSSKEY:
2835 	case SIOCGIFRSSHASH:
2836 	case SIOCGIFDOWNREASON:
2837 		if (ifp->if_ioctl == NULL)
2838 			return (EOPNOTSUPP);
2839 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2840 		break;
2841 
2842 	case SIOCSIFLLADDR:
2843 		error = priv_check(td, PRIV_NET_SETLLADDR);
2844 		if (error)
2845 			return (error);
2846 		error = if_setlladdr(ifp,
2847 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2848 		break;
2849 
2850 	case SIOCGHWADDR:
2851 		error = if_gethwaddr(ifp, ifr);
2852 		break;
2853 
2854 	case SIOCAIFGROUP:
2855 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2856 		if (error)
2857 			return (error);
2858 		error = if_addgroup(ifp,
2859 		    ((struct ifgroupreq *)data)->ifgr_group);
2860 		if (error != 0)
2861 			return (error);
2862 		break;
2863 
2864 	case SIOCGIFGROUP:
2865 	{
2866 		struct epoch_tracker et;
2867 
2868 		NET_EPOCH_ENTER(et);
2869 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2870 		NET_EPOCH_EXIT(et);
2871 		break;
2872 	}
2873 
2874 	case SIOCDIFGROUP:
2875 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2876 		if (error)
2877 			return (error);
2878 		error = if_delgroup(ifp,
2879 		    ((struct ifgroupreq *)data)->ifgr_group);
2880 		if (error != 0)
2881 			return (error);
2882 		break;
2883 
2884 	default:
2885 		error = ENOIOCTL;
2886 		break;
2887 	}
2888 	return (error);
2889 }
2890 
2891 /*
2892  * Interface ioctls.
2893  */
2894 int
2895 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2896 {
2897 #ifdef COMPAT_FREEBSD32
2898 	union {
2899 		struct ifconf ifc;
2900 		struct ifdrv ifd;
2901 		struct ifgroupreq ifgr;
2902 		struct ifmediareq ifmr;
2903 	} thunk;
2904 	u_long saved_cmd;
2905 	struct ifconf32 *ifc32;
2906 	struct ifdrv32 *ifd32;
2907 	struct ifgroupreq32 *ifgr32;
2908 	struct ifmediareq32 *ifmr32;
2909 #endif
2910 	struct ifnet *ifp;
2911 	struct ifreq *ifr;
2912 	int error;
2913 	int oif_flags;
2914 #ifdef VIMAGE
2915 	bool shutdown;
2916 #endif
2917 
2918 	CURVNET_SET(so->so_vnet);
2919 #ifdef VIMAGE
2920 	/* Make sure the VNET is stable. */
2921 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2922 	if (shutdown) {
2923 		CURVNET_RESTORE();
2924 		return (EBUSY);
2925 	}
2926 #endif
2927 
2928 #ifdef COMPAT_FREEBSD32
2929 	saved_cmd = cmd;
2930 	switch (cmd) {
2931 	case SIOCGIFCONF32:
2932 		ifc32 = (struct ifconf32 *)data;
2933 		thunk.ifc.ifc_len = ifc32->ifc_len;
2934 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2935 		data = (caddr_t)&thunk.ifc;
2936 		cmd = SIOCGIFCONF;
2937 		break;
2938 	case SIOCGDRVSPEC32:
2939 	case SIOCSDRVSPEC32:
2940 		ifd32 = (struct ifdrv32 *)data;
2941 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2942 		    sizeof(thunk.ifd.ifd_name));
2943 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2944 		thunk.ifd.ifd_len = ifd32->ifd_len;
2945 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2946 		data = (caddr_t)&thunk.ifd;
2947 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2948 		break;
2949 	case SIOCAIFGROUP32:
2950 	case SIOCGIFGROUP32:
2951 	case SIOCDIFGROUP32:
2952 	case SIOCGIFGMEMB32:
2953 		ifgr32 = (struct ifgroupreq32 *)data;
2954 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2955 		    sizeof(thunk.ifgr.ifgr_name));
2956 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2957 		switch (cmd) {
2958 		case SIOCAIFGROUP32:
2959 		case SIOCDIFGROUP32:
2960 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2961 			    sizeof(thunk.ifgr.ifgr_group));
2962 			break;
2963 		case SIOCGIFGROUP32:
2964 		case SIOCGIFGMEMB32:
2965 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2966 			break;
2967 		}
2968 		data = (caddr_t)&thunk.ifgr;
2969 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2970 		break;
2971 	case SIOCGIFMEDIA32:
2972 	case SIOCGIFXMEDIA32:
2973 		ifmr32 = (struct ifmediareq32 *)data;
2974 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2975 		    sizeof(thunk.ifmr.ifm_name));
2976 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2977 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2978 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2979 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2980 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2981 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2982 		data = (caddr_t)&thunk.ifmr;
2983 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2984 		break;
2985 	}
2986 #endif
2987 
2988 	switch (cmd) {
2989 	case SIOCGIFCONF:
2990 		error = ifconf(cmd, data);
2991 		goto out_noref;
2992 	}
2993 
2994 	ifr = (struct ifreq *)data;
2995 	switch (cmd) {
2996 #ifdef VIMAGE
2997 	case SIOCSIFRVNET:
2998 		error = priv_check(td, PRIV_NET_SETIFVNET);
2999 		if (error == 0)
3000 			error = if_vmove_reclaim(td, ifr->ifr_name,
3001 			    ifr->ifr_jid);
3002 		goto out_noref;
3003 #endif
3004 	case SIOCIFCREATE:
3005 	case SIOCIFCREATE2:
3006 		error = priv_check(td, PRIV_NET_IFCREATE);
3007 		if (error == 0)
3008 			error = if_clone_create(ifr->ifr_name,
3009 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3010 			    ifr_data_get_ptr(ifr) : NULL);
3011 		goto out_noref;
3012 	case SIOCIFDESTROY:
3013 		error = priv_check(td, PRIV_NET_IFDESTROY);
3014 
3015 		if (error == 0) {
3016 			sx_xlock(&ifnet_detach_sxlock);
3017 			error = if_clone_destroy(ifr->ifr_name);
3018 			sx_xunlock(&ifnet_detach_sxlock);
3019 		}
3020 		goto out_noref;
3021 
3022 	case SIOCIFGCLONERS:
3023 		error = if_clone_list((struct if_clonereq *)data);
3024 		goto out_noref;
3025 
3026 	case SIOCGIFGMEMB:
3027 		error = if_getgroupmembers((struct ifgroupreq *)data);
3028 		goto out_noref;
3029 
3030 #if defined(INET) || defined(INET6)
3031 	case SIOCSVH:
3032 	case SIOCGVH:
3033 		if (carp_ioctl_p == NULL)
3034 			error = EPROTONOSUPPORT;
3035 		else
3036 			error = (*carp_ioctl_p)(ifr, cmd, td);
3037 		goto out_noref;
3038 #endif
3039 	}
3040 
3041 	ifp = ifunit_ref(ifr->ifr_name);
3042 	if (ifp == NULL) {
3043 		error = ENXIO;
3044 		goto out_noref;
3045 	}
3046 
3047 	error = ifhwioctl(cmd, ifp, data, td);
3048 	if (error != ENOIOCTL)
3049 		goto out_ref;
3050 
3051 	oif_flags = ifp->if_flags;
3052 	if (so->so_proto == NULL) {
3053 		error = EOPNOTSUPP;
3054 		goto out_ref;
3055 	}
3056 
3057 	/*
3058 	 * Pass the request on to the socket control method, and if the
3059 	 * latter returns EOPNOTSUPP, directly to the interface.
3060 	 *
3061 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3062 	 * trust SIOCSIFADDR et al to come from an already privileged
3063 	 * layer, and do not perform any credentials checks or input
3064 	 * validation.
3065 	 */
3066 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3067 	    ifp, td));
3068 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3069 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3070 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3071 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3072 
3073 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3074 #ifdef INET6
3075 		if (ifp->if_flags & IFF_UP)
3076 			in6_if_up(ifp);
3077 #endif
3078 	}
3079 
3080 out_ref:
3081 	if_rele(ifp);
3082 out_noref:
3083 	CURVNET_RESTORE();
3084 #ifdef COMPAT_FREEBSD32
3085 	if (error != 0)
3086 		return (error);
3087 	switch (saved_cmd) {
3088 	case SIOCGIFCONF32:
3089 		ifc32->ifc_len = thunk.ifc.ifc_len;
3090 		break;
3091 	case SIOCGDRVSPEC32:
3092 		/*
3093 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
3094 		 * the struct so just assert that ifd_len (the only
3095 		 * field it might make sense to update) hasn't
3096 		 * changed.
3097 		 */
3098 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
3099 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
3100 			thunk.ifd.ifd_len));
3101 		break;
3102 	case SIOCGIFGROUP32:
3103 	case SIOCGIFGMEMB32:
3104 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3105 		break;
3106 	case SIOCGIFMEDIA32:
3107 	case SIOCGIFXMEDIA32:
3108 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3109 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3110 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3111 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3112 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3113 		break;
3114 	}
3115 #endif
3116 	return (error);
3117 }
3118 
3119 /*
3120  * The code common to handling reference counted flags,
3121  * e.g., in ifpromisc() and if_allmulti().
3122  * The "pflag" argument can specify a permanent mode flag to check,
3123  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3124  *
3125  * Only to be used on stack-owned flags, not driver-owned flags.
3126  */
3127 static int
3128 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3129 {
3130 	struct ifreq ifr;
3131 	int error;
3132 	int oldflags, oldcount;
3133 
3134 	/* Sanity checks to catch programming errors */
3135 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3136 	    ("%s: setting driver-owned flag %d", __func__, flag));
3137 
3138 	if (onswitch)
3139 		KASSERT(*refcount >= 0,
3140 		    ("%s: increment negative refcount %d for flag %d",
3141 		    __func__, *refcount, flag));
3142 	else
3143 		KASSERT(*refcount > 0,
3144 		    ("%s: decrement non-positive refcount %d for flag %d",
3145 		    __func__, *refcount, flag));
3146 
3147 	/* In case this mode is permanent, just touch refcount */
3148 	if (ifp->if_flags & pflag) {
3149 		*refcount += onswitch ? 1 : -1;
3150 		return (0);
3151 	}
3152 
3153 	/* Save ifnet parameters for if_ioctl() may fail */
3154 	oldcount = *refcount;
3155 	oldflags = ifp->if_flags;
3156 
3157 	/*
3158 	 * See if we aren't the only and touching refcount is enough.
3159 	 * Actually toggle interface flag if we are the first or last.
3160 	 */
3161 	if (onswitch) {
3162 		if ((*refcount)++)
3163 			return (0);
3164 		ifp->if_flags |= flag;
3165 	} else {
3166 		if (--(*refcount))
3167 			return (0);
3168 		ifp->if_flags &= ~flag;
3169 	}
3170 
3171 	/* Call down the driver since we've changed interface flags */
3172 	if (ifp->if_ioctl == NULL) {
3173 		error = EOPNOTSUPP;
3174 		goto recover;
3175 	}
3176 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3177 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3178 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3179 	if (error)
3180 		goto recover;
3181 	/* Notify userland that interface flags have changed */
3182 	rt_ifmsg(ifp);
3183 	return (0);
3184 
3185 recover:
3186 	/* Recover after driver error */
3187 	*refcount = oldcount;
3188 	ifp->if_flags = oldflags;
3189 	return (error);
3190 }
3191 
3192 /*
3193  * Set/clear promiscuous mode on interface ifp based on the truth value
3194  * of pswitch.  The calls are reference counted so that only the first
3195  * "on" request actually has an effect, as does the final "off" request.
3196  * Results are undefined if the "off" and "on" requests are not matched.
3197  */
3198 int
3199 ifpromisc(struct ifnet *ifp, int pswitch)
3200 {
3201 	int error;
3202 	int oldflags = ifp->if_flags;
3203 
3204 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3205 			   &ifp->if_pcount, pswitch);
3206 	/* If promiscuous mode status has changed, log a message */
3207 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3208             log_promisc_mode_change)
3209 		if_printf(ifp, "promiscuous mode %s\n",
3210 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3211 	return (error);
3212 }
3213 
3214 /*
3215  * Return interface configuration
3216  * of system.  List may be used
3217  * in later ioctl's (above) to get
3218  * other information.
3219  */
3220 /*ARGSUSED*/
3221 static int
3222 ifconf(u_long cmd, caddr_t data)
3223 {
3224 	struct ifconf *ifc = (struct ifconf *)data;
3225 	struct ifnet *ifp;
3226 	struct ifaddr *ifa;
3227 	struct ifreq ifr;
3228 	struct sbuf *sb;
3229 	int error, full = 0, valid_len, max_len;
3230 
3231 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3232 	max_len = maxphys - 1;
3233 
3234 	/* Prevent hostile input from being able to crash the system */
3235 	if (ifc->ifc_len <= 0)
3236 		return (EINVAL);
3237 
3238 again:
3239 	if (ifc->ifc_len <= max_len) {
3240 		max_len = ifc->ifc_len;
3241 		full = 1;
3242 	}
3243 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3244 	max_len = 0;
3245 	valid_len = 0;
3246 
3247 	IFNET_RLOCK();
3248 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3249 		struct epoch_tracker et;
3250 		int addrs;
3251 
3252 		/*
3253 		 * Zero the ifr to make sure we don't disclose the contents
3254 		 * of the stack.
3255 		 */
3256 		memset(&ifr, 0, sizeof(ifr));
3257 
3258 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3259 		    >= sizeof(ifr.ifr_name)) {
3260 			sbuf_delete(sb);
3261 			IFNET_RUNLOCK();
3262 			return (ENAMETOOLONG);
3263 		}
3264 
3265 		addrs = 0;
3266 		NET_EPOCH_ENTER(et);
3267 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3268 			struct sockaddr *sa = ifa->ifa_addr;
3269 
3270 			if (prison_if(curthread->td_ucred, sa) != 0)
3271 				continue;
3272 			addrs++;
3273 			if (sa->sa_len <= sizeof(*sa)) {
3274 				if (sa->sa_len < sizeof(*sa)) {
3275 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3276 					    sizeof(ifr.ifr_ifru.ifru_addr));
3277 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3278 					    sa->sa_len);
3279 				} else
3280 					ifr.ifr_ifru.ifru_addr = *sa;
3281 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3282 				max_len += sizeof(ifr);
3283 			} else {
3284 				sbuf_bcat(sb, &ifr,
3285 				    offsetof(struct ifreq, ifr_addr));
3286 				max_len += offsetof(struct ifreq, ifr_addr);
3287 				sbuf_bcat(sb, sa, sa->sa_len);
3288 				max_len += sa->sa_len;
3289 			}
3290 
3291 			if (sbuf_error(sb) == 0)
3292 				valid_len = sbuf_len(sb);
3293 		}
3294 		NET_EPOCH_EXIT(et);
3295 		if (addrs == 0) {
3296 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3297 			max_len += sizeof(ifr);
3298 
3299 			if (sbuf_error(sb) == 0)
3300 				valid_len = sbuf_len(sb);
3301 		}
3302 	}
3303 	IFNET_RUNLOCK();
3304 
3305 	/*
3306 	 * If we didn't allocate enough space (uncommon), try again.  If
3307 	 * we have already allocated as much space as we are allowed,
3308 	 * return what we've got.
3309 	 */
3310 	if (valid_len != max_len && !full) {
3311 		sbuf_delete(sb);
3312 		goto again;
3313 	}
3314 
3315 	ifc->ifc_len = valid_len;
3316 	sbuf_finish(sb);
3317 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3318 	sbuf_delete(sb);
3319 	return (error);
3320 }
3321 
3322 /*
3323  * Just like ifpromisc(), but for all-multicast-reception mode.
3324  */
3325 int
3326 if_allmulti(struct ifnet *ifp, int onswitch)
3327 {
3328 
3329 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3330 }
3331 
3332 struct ifmultiaddr *
3333 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3334 {
3335 	struct ifmultiaddr *ifma;
3336 
3337 	IF_ADDR_LOCK_ASSERT(ifp);
3338 
3339 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3340 		if (sa->sa_family == AF_LINK) {
3341 			if (sa_dl_equal(ifma->ifma_addr, sa))
3342 				break;
3343 		} else {
3344 			if (sa_equal(ifma->ifma_addr, sa))
3345 				break;
3346 		}
3347 	}
3348 
3349 	return ifma;
3350 }
3351 
3352 /*
3353  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3354  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3355  * the ifnet multicast address list here, so the caller must do that and
3356  * other setup work (such as notifying the device driver).  The reference
3357  * count is initialized to 1.
3358  */
3359 static struct ifmultiaddr *
3360 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3361     int mflags)
3362 {
3363 	struct ifmultiaddr *ifma;
3364 	struct sockaddr *dupsa;
3365 
3366 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3367 	    M_ZERO);
3368 	if (ifma == NULL)
3369 		return (NULL);
3370 
3371 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3372 	if (dupsa == NULL) {
3373 		free(ifma, M_IFMADDR);
3374 		return (NULL);
3375 	}
3376 	bcopy(sa, dupsa, sa->sa_len);
3377 	ifma->ifma_addr = dupsa;
3378 
3379 	ifma->ifma_ifp = ifp;
3380 	ifma->ifma_refcount = 1;
3381 	ifma->ifma_protospec = NULL;
3382 
3383 	if (llsa == NULL) {
3384 		ifma->ifma_lladdr = NULL;
3385 		return (ifma);
3386 	}
3387 
3388 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3389 	if (dupsa == NULL) {
3390 		free(ifma->ifma_addr, M_IFMADDR);
3391 		free(ifma, M_IFMADDR);
3392 		return (NULL);
3393 	}
3394 	bcopy(llsa, dupsa, llsa->sa_len);
3395 	ifma->ifma_lladdr = dupsa;
3396 
3397 	return (ifma);
3398 }
3399 
3400 /*
3401  * if_freemulti: free ifmultiaddr structure and possibly attached related
3402  * addresses.  The caller is responsible for implementing reference
3403  * counting, notifying the driver, handling routing messages, and releasing
3404  * any dependent link layer state.
3405  */
3406 #ifdef MCAST_VERBOSE
3407 extern void kdb_backtrace(void);
3408 #endif
3409 static void
3410 if_freemulti_internal(struct ifmultiaddr *ifma)
3411 {
3412 
3413 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3414 	    ifma->ifma_refcount));
3415 
3416 	if (ifma->ifma_lladdr != NULL)
3417 		free(ifma->ifma_lladdr, M_IFMADDR);
3418 #ifdef MCAST_VERBOSE
3419 	kdb_backtrace();
3420 	printf("%s freeing ifma: %p\n", __func__, ifma);
3421 #endif
3422 	free(ifma->ifma_addr, M_IFMADDR);
3423 	free(ifma, M_IFMADDR);
3424 }
3425 
3426 static void
3427 if_destroymulti(epoch_context_t ctx)
3428 {
3429 	struct ifmultiaddr *ifma;
3430 
3431 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3432 	if_freemulti_internal(ifma);
3433 }
3434 
3435 void
3436 if_freemulti(struct ifmultiaddr *ifma)
3437 {
3438 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3439 	    ifma->ifma_refcount));
3440 
3441 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3442 }
3443 
3444 /*
3445  * Register an additional multicast address with a network interface.
3446  *
3447  * - If the address is already present, bump the reference count on the
3448  *   address and return.
3449  * - If the address is not link-layer, look up a link layer address.
3450  * - Allocate address structures for one or both addresses, and attach to the
3451  *   multicast address list on the interface.  If automatically adding a link
3452  *   layer address, the protocol address will own a reference to the link
3453  *   layer address, to be freed when it is freed.
3454  * - Notify the network device driver of an addition to the multicast address
3455  *   list.
3456  *
3457  * 'sa' points to caller-owned memory with the desired multicast address.
3458  *
3459  * 'retifma' will be used to return a pointer to the resulting multicast
3460  * address reference, if desired.
3461  */
3462 int
3463 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3464     struct ifmultiaddr **retifma)
3465 {
3466 	struct ifmultiaddr *ifma, *ll_ifma;
3467 	struct sockaddr *llsa;
3468 	struct sockaddr_dl sdl;
3469 	int error;
3470 
3471 #ifdef INET
3472 	IN_MULTI_LIST_UNLOCK_ASSERT();
3473 #endif
3474 #ifdef INET6
3475 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3476 #endif
3477 	/*
3478 	 * If the address is already present, return a new reference to it;
3479 	 * otherwise, allocate storage and set up a new address.
3480 	 */
3481 	IF_ADDR_WLOCK(ifp);
3482 	ifma = if_findmulti(ifp, sa);
3483 	if (ifma != NULL) {
3484 		ifma->ifma_refcount++;
3485 		if (retifma != NULL)
3486 			*retifma = ifma;
3487 		IF_ADDR_WUNLOCK(ifp);
3488 		return (0);
3489 	}
3490 
3491 	/*
3492 	 * The address isn't already present; resolve the protocol address
3493 	 * into a link layer address, and then look that up, bump its
3494 	 * refcount or allocate an ifma for that also.
3495 	 * Most link layer resolving functions returns address data which
3496 	 * fits inside default sockaddr_dl structure. However callback
3497 	 * can allocate another sockaddr structure, in that case we need to
3498 	 * free it later.
3499 	 */
3500 	llsa = NULL;
3501 	ll_ifma = NULL;
3502 	if (ifp->if_resolvemulti != NULL) {
3503 		/* Provide called function with buffer size information */
3504 		sdl.sdl_len = sizeof(sdl);
3505 		llsa = (struct sockaddr *)&sdl;
3506 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3507 		if (error)
3508 			goto unlock_out;
3509 	}
3510 
3511 	/*
3512 	 * Allocate the new address.  Don't hook it up yet, as we may also
3513 	 * need to allocate a link layer multicast address.
3514 	 */
3515 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3516 	if (ifma == NULL) {
3517 		error = ENOMEM;
3518 		goto free_llsa_out;
3519 	}
3520 
3521 	/*
3522 	 * If a link layer address is found, we'll need to see if it's
3523 	 * already present in the address list, or allocate is as well.
3524 	 * When this block finishes, the link layer address will be on the
3525 	 * list.
3526 	 */
3527 	if (llsa != NULL) {
3528 		ll_ifma = if_findmulti(ifp, llsa);
3529 		if (ll_ifma == NULL) {
3530 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3531 			if (ll_ifma == NULL) {
3532 				--ifma->ifma_refcount;
3533 				if_freemulti(ifma);
3534 				error = ENOMEM;
3535 				goto free_llsa_out;
3536 			}
3537 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3538 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3539 			    ifma_link);
3540 		} else
3541 			ll_ifma->ifma_refcount++;
3542 		ifma->ifma_llifma = ll_ifma;
3543 	}
3544 
3545 	/*
3546 	 * We now have a new multicast address, ifma, and possibly a new or
3547 	 * referenced link layer address.  Add the primary address to the
3548 	 * ifnet address list.
3549 	 */
3550 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3551 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3552 
3553 	if (retifma != NULL)
3554 		*retifma = ifma;
3555 
3556 	/*
3557 	 * Must generate the message while holding the lock so that 'ifma'
3558 	 * pointer is still valid.
3559 	 */
3560 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3561 	IF_ADDR_WUNLOCK(ifp);
3562 
3563 	/*
3564 	 * We are certain we have added something, so call down to the
3565 	 * interface to let them know about it.
3566 	 */
3567 	if (ifp->if_ioctl != NULL) {
3568 		if (THREAD_CAN_SLEEP())
3569 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3570 		else
3571 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3572 	}
3573 
3574 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3575 		link_free_sdl(llsa);
3576 
3577 	return (0);
3578 
3579 free_llsa_out:
3580 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3581 		link_free_sdl(llsa);
3582 
3583 unlock_out:
3584 	IF_ADDR_WUNLOCK(ifp);
3585 	return (error);
3586 }
3587 
3588 static void
3589 if_siocaddmulti(void *arg, int pending)
3590 {
3591 	struct ifnet *ifp;
3592 
3593 	ifp = arg;
3594 #ifdef DIAGNOSTIC
3595 	if (pending > 1)
3596 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3597 #endif
3598 	CURVNET_SET(ifp->if_vnet);
3599 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3600 	CURVNET_RESTORE();
3601 }
3602 
3603 /*
3604  * Delete a multicast group membership by network-layer group address.
3605  *
3606  * Returns ENOENT if the entry could not be found. If ifp no longer
3607  * exists, results are undefined. This entry point should only be used
3608  * from subsystems which do appropriate locking to hold ifp for the
3609  * duration of the call.
3610  * Network-layer protocol domains must use if_delmulti_ifma().
3611  */
3612 int
3613 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3614 {
3615 	struct ifmultiaddr *ifma;
3616 	int lastref;
3617 
3618 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3619 
3620 	IF_ADDR_WLOCK(ifp);
3621 	lastref = 0;
3622 	ifma = if_findmulti(ifp, sa);
3623 	if (ifma != NULL)
3624 		lastref = if_delmulti_locked(ifp, ifma, 0);
3625 	IF_ADDR_WUNLOCK(ifp);
3626 
3627 	if (ifma == NULL)
3628 		return (ENOENT);
3629 
3630 	if (lastref && ifp->if_ioctl != NULL) {
3631 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3632 	}
3633 
3634 	return (0);
3635 }
3636 
3637 /*
3638  * Delete all multicast group membership for an interface.
3639  * Should be used to quickly flush all multicast filters.
3640  */
3641 void
3642 if_delallmulti(struct ifnet *ifp)
3643 {
3644 	struct ifmultiaddr *ifma;
3645 	struct ifmultiaddr *next;
3646 
3647 	IF_ADDR_WLOCK(ifp);
3648 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3649 		if_delmulti_locked(ifp, ifma, 0);
3650 	IF_ADDR_WUNLOCK(ifp);
3651 }
3652 
3653 void
3654 if_delmulti_ifma(struct ifmultiaddr *ifma)
3655 {
3656 	if_delmulti_ifma_flags(ifma, 0);
3657 }
3658 
3659 /*
3660  * Delete a multicast group membership by group membership pointer.
3661  * Network-layer protocol domains must use this routine.
3662  *
3663  * It is safe to call this routine if the ifp disappeared.
3664  */
3665 void
3666 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3667 {
3668 	struct ifnet *ifp;
3669 	int lastref;
3670 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3671 #ifdef INET
3672 	IN_MULTI_LIST_UNLOCK_ASSERT();
3673 #endif
3674 	ifp = ifma->ifma_ifp;
3675 #ifdef DIAGNOSTIC
3676 	if (ifp == NULL) {
3677 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3678 	} else {
3679 		struct epoch_tracker et;
3680 		struct ifnet *oifp;
3681 
3682 		NET_EPOCH_ENTER(et);
3683 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3684 			if (ifp == oifp)
3685 				break;
3686 		NET_EPOCH_EXIT(et);
3687 		if (ifp != oifp)
3688 			ifp = NULL;
3689 	}
3690 #endif
3691 	/*
3692 	 * If and only if the ifnet instance exists: Acquire the address lock.
3693 	 */
3694 	if (ifp != NULL)
3695 		IF_ADDR_WLOCK(ifp);
3696 
3697 	lastref = if_delmulti_locked(ifp, ifma, flags);
3698 
3699 	if (ifp != NULL) {
3700 		/*
3701 		 * If and only if the ifnet instance exists:
3702 		 *  Release the address lock.
3703 		 *  If the group was left: update the hardware hash filter.
3704 		 */
3705 		IF_ADDR_WUNLOCK(ifp);
3706 		if (lastref && ifp->if_ioctl != NULL) {
3707 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3708 		}
3709 	}
3710 }
3711 
3712 /*
3713  * Perform deletion of network-layer and/or link-layer multicast address.
3714  *
3715  * Return 0 if the reference count was decremented.
3716  * Return 1 if the final reference was released, indicating that the
3717  * hardware hash filter should be reprogrammed.
3718  */
3719 static int
3720 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3721 {
3722 	struct ifmultiaddr *ll_ifma;
3723 
3724 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3725 		KASSERT(ifma->ifma_ifp == ifp,
3726 		    ("%s: inconsistent ifp %p", __func__, ifp));
3727 		IF_ADDR_WLOCK_ASSERT(ifp);
3728 	}
3729 
3730 	ifp = ifma->ifma_ifp;
3731 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3732 
3733 	/*
3734 	 * If the ifnet is detaching, null out references to ifnet,
3735 	 * so that upper protocol layers will notice, and not attempt
3736 	 * to obtain locks for an ifnet which no longer exists. The
3737 	 * routing socket announcement must happen before the ifnet
3738 	 * instance is detached from the system.
3739 	 */
3740 	if (detaching) {
3741 #ifdef DIAGNOSTIC
3742 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3743 #endif
3744 		/*
3745 		 * ifp may already be nulled out if we are being reentered
3746 		 * to delete the ll_ifma.
3747 		 */
3748 		if (ifp != NULL) {
3749 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3750 			ifma->ifma_ifp = NULL;
3751 		}
3752 	}
3753 
3754 	if (--ifma->ifma_refcount > 0)
3755 		return 0;
3756 
3757 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3758 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3759 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3760 	}
3761 	/*
3762 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3763 	 * have been associated with it. Release it first if so.
3764 	 */
3765 	ll_ifma = ifma->ifma_llifma;
3766 	if (ll_ifma != NULL) {
3767 		KASSERT(ifma->ifma_lladdr != NULL,
3768 		    ("%s: llifma w/o lladdr", __func__));
3769 		if (detaching)
3770 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3771 		if (--ll_ifma->ifma_refcount == 0) {
3772 			if (ifp != NULL) {
3773 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3774 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3775 						ifma_link);
3776 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3777 				}
3778 			}
3779 			if_freemulti(ll_ifma);
3780 		}
3781 	}
3782 #ifdef INVARIANTS
3783 	if (ifp) {
3784 		struct ifmultiaddr *ifmatmp;
3785 
3786 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3787 			MPASS(ifma != ifmatmp);
3788 	}
3789 #endif
3790 	if_freemulti(ifma);
3791 	/*
3792 	 * The last reference to this instance of struct ifmultiaddr
3793 	 * was released; the hardware should be notified of this change.
3794 	 */
3795 	return 1;
3796 }
3797 
3798 /*
3799  * Set the link layer address on an interface.
3800  *
3801  * At this time we only support certain types of interfaces,
3802  * and we don't allow the length of the address to change.
3803  *
3804  * Set noinline to be dtrace-friendly
3805  */
3806 __noinline int
3807 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3808 {
3809 	struct sockaddr_dl *sdl;
3810 	struct ifaddr *ifa;
3811 	struct ifreq ifr;
3812 
3813 	ifa = ifp->if_addr;
3814 	if (ifa == NULL)
3815 		return (EINVAL);
3816 
3817 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3818 	if (sdl == NULL)
3819 		return (EINVAL);
3820 
3821 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3822 		return (EINVAL);
3823 
3824 	switch (ifp->if_type) {
3825 	case IFT_ETHER:
3826 	case IFT_XETHER:
3827 	case IFT_L2VLAN:
3828 	case IFT_BRIDGE:
3829 	case IFT_IEEE8023ADLAG:
3830 		bcopy(lladdr, LLADDR(sdl), len);
3831 		break;
3832 	default:
3833 		return (ENODEV);
3834 	}
3835 
3836 	/*
3837 	 * If the interface is already up, we need
3838 	 * to re-init it in order to reprogram its
3839 	 * address filter.
3840 	 */
3841 	if ((ifp->if_flags & IFF_UP) != 0) {
3842 		if (ifp->if_ioctl) {
3843 			ifp->if_flags &= ~IFF_UP;
3844 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3845 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3846 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3847 			ifp->if_flags |= IFF_UP;
3848 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3849 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3850 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3851 		}
3852 	}
3853 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3854 
3855 	return (0);
3856 }
3857 
3858 /*
3859  * Compat function for handling basic encapsulation requests.
3860  * Not converted stacks (FDDI, IB, ..) supports traditional
3861  * output model: ARP (and other similar L2 protocols) are handled
3862  * inside output routine, arpresolve/nd6_resolve() returns MAC
3863  * address instead of full prepend.
3864  *
3865  * This function creates calculated header==MAC for IPv4/IPv6 and
3866  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3867  * address families.
3868  */
3869 static int
3870 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3871 {
3872 
3873 	if (req->rtype != IFENCAP_LL)
3874 		return (EOPNOTSUPP);
3875 
3876 	if (req->bufsize < req->lladdr_len)
3877 		return (ENOMEM);
3878 
3879 	switch (req->family) {
3880 	case AF_INET:
3881 	case AF_INET6:
3882 		break;
3883 	default:
3884 		return (EAFNOSUPPORT);
3885 	}
3886 
3887 	/* Copy lladdr to storage as is */
3888 	memmove(req->buf, req->lladdr, req->lladdr_len);
3889 	req->bufsize = req->lladdr_len;
3890 	req->lladdr_off = 0;
3891 
3892 	return (0);
3893 }
3894 
3895 /*
3896  * Tunnel interfaces can nest, also they may cause infinite recursion
3897  * calls when misconfigured. We'll prevent this by detecting loops.
3898  * High nesting level may cause stack exhaustion. We'll prevent this
3899  * by introducing upper limit.
3900  *
3901  * Return 0, if tunnel nesting count is equal or less than limit.
3902  */
3903 int
3904 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3905     int limit)
3906 {
3907 	struct m_tag *mtag;
3908 	int count;
3909 
3910 	count = 1;
3911 	mtag = NULL;
3912 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3913 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3914 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3915 			return (EIO);
3916 		}
3917 		count++;
3918 	}
3919 	if (count > limit) {
3920 		log(LOG_NOTICE,
3921 		    "%s: if_output recursively called too many times(%d)\n",
3922 		    if_name(ifp), count);
3923 		return (EIO);
3924 	}
3925 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3926 	if (mtag == NULL)
3927 		return (ENOMEM);
3928 	*(struct ifnet **)(mtag + 1) = ifp;
3929 	m_tag_prepend(m, mtag);
3930 	return (0);
3931 }
3932 
3933 /*
3934  * Get the link layer address that was read from the hardware at attach.
3935  *
3936  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3937  * their component interfaces as IFT_IEEE8023ADLAG.
3938  */
3939 int
3940 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3941 {
3942 
3943 	if (ifp->if_hw_addr == NULL)
3944 		return (ENODEV);
3945 
3946 	switch (ifp->if_type) {
3947 	case IFT_ETHER:
3948 	case IFT_IEEE8023ADLAG:
3949 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3950 		return (0);
3951 	default:
3952 		return (ENODEV);
3953 	}
3954 }
3955 
3956 /*
3957  * The name argument must be a pointer to storage which will last as
3958  * long as the interface does.  For physical devices, the result of
3959  * device_get_name(dev) is a good choice and for pseudo-devices a
3960  * static string works well.
3961  */
3962 void
3963 if_initname(struct ifnet *ifp, const char *name, int unit)
3964 {
3965 	ifp->if_dname = name;
3966 	ifp->if_dunit = unit;
3967 	if (unit != IF_DUNIT_NONE)
3968 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3969 	else
3970 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3971 }
3972 
3973 static int
3974 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
3975 {
3976 	char if_fmt[256];
3977 
3978 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3979 	vlog(pri, if_fmt, ap);
3980 	return (0);
3981 }
3982 
3983 
3984 int
3985 if_printf(struct ifnet *ifp, const char *fmt, ...)
3986 {
3987 	va_list ap;
3988 
3989 	va_start(ap, fmt);
3990 	if_vlog(ifp, LOG_INFO, fmt, ap);
3991 	va_end(ap);
3992 	return (0);
3993 }
3994 
3995 int
3996 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
3997 {
3998 	va_list ap;
3999 
4000 	va_start(ap, fmt);
4001 	if_vlog(ifp, pri, fmt, ap);
4002 	va_end(ap);
4003 	return (0);
4004 }
4005 
4006 void
4007 if_start(struct ifnet *ifp)
4008 {
4009 
4010 	(*(ifp)->if_start)(ifp);
4011 }
4012 
4013 /*
4014  * Backwards compatibility interface for drivers
4015  * that have not implemented it
4016  */
4017 static int
4018 if_transmit(struct ifnet *ifp, struct mbuf *m)
4019 {
4020 	int error;
4021 
4022 	IFQ_HANDOFF(ifp, m, error);
4023 	return (error);
4024 }
4025 
4026 static void
4027 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4028 {
4029 
4030 	m_freem(m);
4031 }
4032 
4033 int
4034 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4035 {
4036 	int active = 0;
4037 
4038 	IF_LOCK(ifq);
4039 	if (_IF_QFULL(ifq)) {
4040 		IF_UNLOCK(ifq);
4041 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4042 		m_freem(m);
4043 		return (0);
4044 	}
4045 	if (ifp != NULL) {
4046 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4047 		if (m->m_flags & (M_BCAST|M_MCAST))
4048 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4049 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4050 	}
4051 	_IF_ENQUEUE(ifq, m);
4052 	IF_UNLOCK(ifq);
4053 	if (ifp != NULL && !active)
4054 		(*(ifp)->if_start)(ifp);
4055 	return (1);
4056 }
4057 
4058 void
4059 if_register_com_alloc(u_char type,
4060     if_com_alloc_t *a, if_com_free_t *f)
4061 {
4062 
4063 	KASSERT(if_com_alloc[type] == NULL,
4064 	    ("if_register_com_alloc: %d already registered", type));
4065 	KASSERT(if_com_free[type] == NULL,
4066 	    ("if_register_com_alloc: %d free already registered", type));
4067 
4068 	if_com_alloc[type] = a;
4069 	if_com_free[type] = f;
4070 }
4071 
4072 void
4073 if_deregister_com_alloc(u_char type)
4074 {
4075 
4076 	KASSERT(if_com_alloc[type] != NULL,
4077 	    ("if_deregister_com_alloc: %d not registered", type));
4078 	KASSERT(if_com_free[type] != NULL,
4079 	    ("if_deregister_com_alloc: %d free not registered", type));
4080 
4081 	/*
4082 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4083 	 * fixes issues about late invocation of if_destroy(), which leads
4084 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4085 	 */
4086 	epoch_drain_callbacks(net_epoch_preempt);
4087 
4088 	if_com_alloc[type] = NULL;
4089 	if_com_free[type] = NULL;
4090 }
4091 
4092 /* API for driver access to network stack owned ifnet.*/
4093 uint64_t
4094 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4095 {
4096 	uint64_t oldbrate;
4097 
4098 	oldbrate = ifp->if_baudrate;
4099 	ifp->if_baudrate = baudrate;
4100 	return (oldbrate);
4101 }
4102 
4103 uint64_t
4104 if_getbaudrate(if_t ifp)
4105 {
4106 
4107 	return (((struct ifnet *)ifp)->if_baudrate);
4108 }
4109 
4110 int
4111 if_setcapabilities(if_t ifp, int capabilities)
4112 {
4113 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4114 	return (0);
4115 }
4116 
4117 int
4118 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4119 {
4120 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4121 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4122 
4123 	return (0);
4124 }
4125 
4126 int
4127 if_getcapabilities(if_t ifp)
4128 {
4129 	return ((struct ifnet *)ifp)->if_capabilities;
4130 }
4131 
4132 int
4133 if_setcapenable(if_t ifp, int capabilities)
4134 {
4135 	((struct ifnet *)ifp)->if_capenable = capabilities;
4136 	return (0);
4137 }
4138 
4139 int
4140 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4141 {
4142 	if(setcap)
4143 		((struct ifnet *)ifp)->if_capenable |= setcap;
4144 	if(clearcap)
4145 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4146 
4147 	return (0);
4148 }
4149 
4150 const char *
4151 if_getdname(if_t ifp)
4152 {
4153 	return ((struct ifnet *)ifp)->if_dname;
4154 }
4155 
4156 int
4157 if_togglecapenable(if_t ifp, int togglecap)
4158 {
4159 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4160 	return (0);
4161 }
4162 
4163 int
4164 if_getcapenable(if_t ifp)
4165 {
4166 	return ((struct ifnet *)ifp)->if_capenable;
4167 }
4168 
4169 /*
4170  * This is largely undesirable because it ties ifnet to a device, but does
4171  * provide flexiblity for an embedded product vendor. Should be used with
4172  * the understanding that it violates the interface boundaries, and should be
4173  * a last resort only.
4174  */
4175 int
4176 if_setdev(if_t ifp, void *dev)
4177 {
4178 	return (0);
4179 }
4180 
4181 int
4182 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4183 {
4184 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4185 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4186 
4187 	return (0);
4188 }
4189 
4190 int
4191 if_getdrvflags(if_t ifp)
4192 {
4193 	return ((struct ifnet *)ifp)->if_drv_flags;
4194 }
4195 
4196 int
4197 if_setdrvflags(if_t ifp, int flags)
4198 {
4199 	((struct ifnet *)ifp)->if_drv_flags = flags;
4200 	return (0);
4201 }
4202 
4203 int
4204 if_setflags(if_t ifp, int flags)
4205 {
4206 
4207 	ifp->if_flags = flags;
4208 	return (0);
4209 }
4210 
4211 int
4212 if_setflagbits(if_t ifp, int set, int clear)
4213 {
4214 	((struct ifnet *)ifp)->if_flags |= set;
4215 	((struct ifnet *)ifp)->if_flags &= ~clear;
4216 
4217 	return (0);
4218 }
4219 
4220 int
4221 if_getflags(if_t ifp)
4222 {
4223 	return ((struct ifnet *)ifp)->if_flags;
4224 }
4225 
4226 int
4227 if_clearhwassist(if_t ifp)
4228 {
4229 	((struct ifnet *)ifp)->if_hwassist = 0;
4230 	return (0);
4231 }
4232 
4233 int
4234 if_sethwassistbits(if_t ifp, int toset, int toclear)
4235 {
4236 	((struct ifnet *)ifp)->if_hwassist |= toset;
4237 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4238 
4239 	return (0);
4240 }
4241 
4242 int
4243 if_sethwassist(if_t ifp, int hwassist_bit)
4244 {
4245 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4246 	return (0);
4247 }
4248 
4249 int
4250 if_gethwassist(if_t ifp)
4251 {
4252 	return ((struct ifnet *)ifp)->if_hwassist;
4253 }
4254 
4255 int
4256 if_setmtu(if_t ifp, int mtu)
4257 {
4258 	((struct ifnet *)ifp)->if_mtu = mtu;
4259 	return (0);
4260 }
4261 
4262 int
4263 if_getmtu(if_t ifp)
4264 {
4265 	return ((struct ifnet *)ifp)->if_mtu;
4266 }
4267 
4268 int
4269 if_getmtu_family(if_t ifp, int family)
4270 {
4271 	struct domain *dp;
4272 
4273 	for (dp = domains; dp; dp = dp->dom_next) {
4274 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4275 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4276 	}
4277 
4278 	return (((struct ifnet *)ifp)->if_mtu);
4279 }
4280 
4281 /*
4282  * Methods for drivers to access interface unicast and multicast
4283  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4284  * 'struct ifmultiaddr'.
4285  */
4286 u_int
4287 if_lladdr_count(if_t ifp)
4288 {
4289 	struct epoch_tracker et;
4290 	struct ifaddr *ifa;
4291 	u_int count;
4292 
4293 	count = 0;
4294 	NET_EPOCH_ENTER(et);
4295 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4296 		if (ifa->ifa_addr->sa_family == AF_LINK)
4297 			count++;
4298 	NET_EPOCH_EXIT(et);
4299 
4300 	return (count);
4301 }
4302 
4303 u_int
4304 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4305 {
4306 	struct epoch_tracker et;
4307 	struct ifaddr *ifa;
4308 	u_int count;
4309 
4310 	MPASS(cb);
4311 
4312 	count = 0;
4313 	NET_EPOCH_ENTER(et);
4314 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4315 		if (ifa->ifa_addr->sa_family != AF_LINK)
4316 			continue;
4317 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4318 		    count);
4319 	}
4320 	NET_EPOCH_EXIT(et);
4321 
4322 	return (count);
4323 }
4324 
4325 u_int
4326 if_llmaddr_count(if_t ifp)
4327 {
4328 	struct epoch_tracker et;
4329 	struct ifmultiaddr *ifma;
4330 	int count;
4331 
4332 	count = 0;
4333 	NET_EPOCH_ENTER(et);
4334 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4335 		if (ifma->ifma_addr->sa_family == AF_LINK)
4336 			count++;
4337 	NET_EPOCH_EXIT(et);
4338 
4339 	return (count);
4340 }
4341 
4342 u_int
4343 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4344 {
4345 	struct epoch_tracker et;
4346 	struct ifmultiaddr *ifma;
4347 	u_int count;
4348 
4349 	MPASS(cb);
4350 
4351 	count = 0;
4352 	NET_EPOCH_ENTER(et);
4353 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4354 		if (ifma->ifma_addr->sa_family != AF_LINK)
4355 			continue;
4356 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4357 		    count);
4358 	}
4359 	NET_EPOCH_EXIT(et);
4360 
4361 	return (count);
4362 }
4363 
4364 int
4365 if_setsoftc(if_t ifp, void *softc)
4366 {
4367 	((struct ifnet *)ifp)->if_softc = softc;
4368 	return (0);
4369 }
4370 
4371 void *
4372 if_getsoftc(if_t ifp)
4373 {
4374 	return ((struct ifnet *)ifp)->if_softc;
4375 }
4376 
4377 void
4378 if_setrcvif(struct mbuf *m, if_t ifp)
4379 {
4380 
4381 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4382 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4383 }
4384 
4385 void
4386 if_setvtag(struct mbuf *m, uint16_t tag)
4387 {
4388 	m->m_pkthdr.ether_vtag = tag;
4389 }
4390 
4391 uint16_t
4392 if_getvtag(struct mbuf *m)
4393 {
4394 
4395 	return (m->m_pkthdr.ether_vtag);
4396 }
4397 
4398 int
4399 if_sendq_empty(if_t ifp)
4400 {
4401 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4402 }
4403 
4404 struct ifaddr *
4405 if_getifaddr(if_t ifp)
4406 {
4407 	return ((struct ifnet *)ifp)->if_addr;
4408 }
4409 
4410 int
4411 if_getamcount(if_t ifp)
4412 {
4413 	return ((struct ifnet *)ifp)->if_amcount;
4414 }
4415 
4416 int
4417 if_setsendqready(if_t ifp)
4418 {
4419 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4420 	return (0);
4421 }
4422 
4423 int
4424 if_setsendqlen(if_t ifp, int tx_desc_count)
4425 {
4426 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4427 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4428 
4429 	return (0);
4430 }
4431 
4432 int
4433 if_vlantrunkinuse(if_t ifp)
4434 {
4435 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4436 }
4437 
4438 int
4439 if_input(if_t ifp, struct mbuf* sendmp)
4440 {
4441 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4442 	return (0);
4443 
4444 }
4445 
4446 struct mbuf *
4447 if_dequeue(if_t ifp)
4448 {
4449 	struct mbuf *m;
4450 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4451 
4452 	return (m);
4453 }
4454 
4455 int
4456 if_sendq_prepend(if_t ifp, struct mbuf *m)
4457 {
4458 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4459 	return (0);
4460 }
4461 
4462 int
4463 if_setifheaderlen(if_t ifp, int len)
4464 {
4465 	((struct ifnet *)ifp)->if_hdrlen = len;
4466 	return (0);
4467 }
4468 
4469 caddr_t
4470 if_getlladdr(if_t ifp)
4471 {
4472 	return (IF_LLADDR((struct ifnet *)ifp));
4473 }
4474 
4475 void *
4476 if_gethandle(u_char type)
4477 {
4478 	return (if_alloc(type));
4479 }
4480 
4481 void
4482 if_bpfmtap(if_t ifh, struct mbuf *m)
4483 {
4484 	struct ifnet *ifp = (struct ifnet *)ifh;
4485 
4486 	BPF_MTAP(ifp, m);
4487 }
4488 
4489 void
4490 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4491 {
4492 	struct ifnet *ifp = (struct ifnet *)ifh;
4493 
4494 	ETHER_BPF_MTAP(ifp, m);
4495 }
4496 
4497 void
4498 if_vlancap(if_t ifh)
4499 {
4500 	struct ifnet *ifp = (struct ifnet *)ifh;
4501 	VLAN_CAPABILITIES(ifp);
4502 }
4503 
4504 int
4505 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4506 {
4507 
4508 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4509         return (0);
4510 }
4511 
4512 int
4513 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4514 {
4515 
4516 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4517         return (0);
4518 }
4519 
4520 int
4521 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4522 {
4523 
4524 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4525         return (0);
4526 }
4527 
4528 u_int
4529 if_gethwtsomax(if_t ifp)
4530 {
4531 
4532 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4533 }
4534 
4535 u_int
4536 if_gethwtsomaxsegcount(if_t ifp)
4537 {
4538 
4539 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4540 }
4541 
4542 u_int
4543 if_gethwtsomaxsegsize(if_t ifp)
4544 {
4545 
4546 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4547 }
4548 
4549 void
4550 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4551 {
4552 	((struct ifnet *)ifp)->if_init = init_fn;
4553 }
4554 
4555 void
4556 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4557 {
4558 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4559 }
4560 
4561 void
4562 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4563 {
4564 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4565 }
4566 
4567 void
4568 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4569 {
4570 	((struct ifnet *)ifp)->if_transmit = start_fn;
4571 }
4572 
4573 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4574 {
4575 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4576 
4577 }
4578 
4579 void
4580 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4581 {
4582 
4583 	ifp->if_get_counter = fn;
4584 }
4585 
4586 /* Revisit these - These are inline functions originally. */
4587 int
4588 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4589 {
4590 	return drbr_inuse(ifh, br);
4591 }
4592 
4593 struct mbuf*
4594 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4595 {
4596 	return drbr_dequeue(ifh, br);
4597 }
4598 
4599 int
4600 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4601 {
4602 	return drbr_needs_enqueue(ifh, br);
4603 }
4604 
4605 int
4606 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4607 {
4608 	return drbr_enqueue(ifh, br, m);
4609 
4610 }
4611