xref: /freebsd/sys/net/if.c (revision cc9e6590773dba57440750c124173ed531349a06)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org>
5  * Copyright (c) 1980, 1986, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)if.c	8.5 (Berkeley) 1/9/95
33  * $FreeBSD$
34  */
35 
36 #include "opt_bpf.h"
37 #include "opt_inet6.h"
38 #include "opt_inet.h"
39 #include "opt_ddb.h"
40 
41 #include <sys/param.h>
42 #include <sys/capsicum.h>
43 #include <sys/conf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/malloc.h>
46 #include <sys/domainset.h>
47 #include <sys/sbuf.h>
48 #include <sys/bus.h>
49 #include <sys/epoch.h>
50 #include <sys/mbuf.h>
51 #include <sys/systm.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/protosw.h>
57 #include <sys/kernel.h>
58 #include <sys/lock.h>
59 #include <sys/refcount.h>
60 #include <sys/module.h>
61 #include <sys/rwlock.h>
62 #include <sys/sockio.h>
63 #include <sys/syslog.h>
64 #include <sys/sysctl.h>
65 #include <sys/sysent.h>
66 #include <sys/taskqueue.h>
67 #include <sys/domain.h>
68 #include <sys/jail.h>
69 #include <sys/priv.h>
70 
71 #ifdef DDB
72 #include <ddb/ddb.h>
73 #endif
74 
75 #include <machine/stdarg.h>
76 #include <vm/uma.h>
77 
78 #include <net/bpf.h>
79 #include <net/ethernet.h>
80 #include <net/if.h>
81 #include <net/if_arp.h>
82 #include <net/if_clone.h>
83 #include <net/if_dl.h>
84 #include <net/if_types.h>
85 #include <net/if_var.h>
86 #include <net/if_media.h>
87 #include <net/if_mib.h>
88 #include <net/if_vlan_var.h>
89 #include <net/radix.h>
90 #include <net/route.h>
91 #include <net/route/route_ctl.h>
92 #include <net/vnet.h>
93 
94 #if defined(INET) || defined(INET6)
95 #include <net/ethernet.h>
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip.h>
99 #include <netinet/ip_carp.h>
100 #ifdef INET
101 #include <net/debugnet.h>
102 #include <netinet/if_ether.h>
103 #endif /* INET */
104 #ifdef INET6
105 #include <netinet6/in6_var.h>
106 #include <netinet6/in6_ifattach.h>
107 #endif /* INET6 */
108 #endif /* INET || INET6 */
109 
110 #include <security/mac/mac_framework.h>
111 
112 /*
113  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
114  * and ifr_ifru when it is used in SIOCGIFCONF.
115  */
116 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
117     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
118 
119 __read_mostly epoch_t net_epoch_preempt;
120 #ifdef COMPAT_FREEBSD32
121 #include <sys/mount.h>
122 #include <compat/freebsd32/freebsd32.h>
123 
124 struct ifreq_buffer32 {
125 	uint32_t	length;		/* (size_t) */
126 	uint32_t	buffer;		/* (void *) */
127 };
128 
129 /*
130  * Interface request structure used for socket
131  * ioctl's.  All interface ioctl's must have parameter
132  * definitions which begin with ifr_name.  The
133  * remainder may be interface specific.
134  */
135 struct ifreq32 {
136 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
137 	union {
138 		struct sockaddr	ifru_addr;
139 		struct sockaddr	ifru_dstaddr;
140 		struct sockaddr	ifru_broadaddr;
141 		struct ifreq_buffer32 ifru_buffer;
142 		short		ifru_flags[2];
143 		short		ifru_index;
144 		int		ifru_jid;
145 		int		ifru_metric;
146 		int		ifru_mtu;
147 		int		ifru_phys;
148 		int		ifru_media;
149 		uint32_t	ifru_data;
150 		int		ifru_cap[2];
151 		u_int		ifru_fib;
152 		u_char		ifru_vlan_pcp;
153 	} ifr_ifru;
154 };
155 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
156 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
157     __offsetof(struct ifreq32, ifr_ifru));
158 
159 struct ifconf32 {
160 	int32_t	ifc_len;
161 	union {
162 		uint32_t	ifcu_buf;
163 		uint32_t	ifcu_req;
164 	} ifc_ifcu;
165 };
166 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
167 
168 struct ifdrv32 {
169 	char		ifd_name[IFNAMSIZ];
170 	uint32_t	ifd_cmd;
171 	uint32_t	ifd_len;
172 	uint32_t	ifd_data;
173 };
174 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
175 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
176 
177 struct ifgroupreq32 {
178 	char	ifgr_name[IFNAMSIZ];
179 	u_int	ifgr_len;
180 	union {
181 		char		ifgru_group[IFNAMSIZ];
182 		uint32_t	ifgru_groups;
183 	} ifgr_ifgru;
184 };
185 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
186 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
187 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
188 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
189 
190 struct ifmediareq32 {
191 	char		ifm_name[IFNAMSIZ];
192 	int		ifm_current;
193 	int		ifm_mask;
194 	int		ifm_status;
195 	int		ifm_active;
196 	int		ifm_count;
197 	uint32_t	ifm_ulist;	/* (int *) */
198 };
199 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
200 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
201 #endif /* COMPAT_FREEBSD32 */
202 
203 union ifreq_union {
204 	struct ifreq	ifr;
205 #ifdef COMPAT_FREEBSD32
206 	struct ifreq32	ifr32;
207 #endif
208 };
209 
210 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
211     "Link layers");
212 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
213     "Generic link-management");
214 
215 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
216     &ifqmaxlen, 0, "max send queue size");
217 
218 /* Log link state change events */
219 static int log_link_state_change = 1;
220 
221 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
222 	&log_link_state_change, 0,
223 	"log interface link state change events");
224 
225 /* Log promiscuous mode change events */
226 static int log_promisc_mode_change = 1;
227 
228 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
229 	&log_promisc_mode_change, 1,
230 	"log promiscuous mode change events");
231 
232 /* Interface description */
233 static unsigned int ifdescr_maxlen = 1024;
234 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
235 	&ifdescr_maxlen, 0,
236 	"administrative maximum length for interface description");
237 
238 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
239 
240 /* global sx for non-critical path ifdescr */
241 static struct sx ifdescr_sx;
242 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
243 
244 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
245 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
246 /* These are external hooks for CARP. */
247 void	(*carp_linkstate_p)(struct ifnet *ifp);
248 void	(*carp_demote_adj_p)(int, char *);
249 int	(*carp_master_p)(struct ifaddr *);
250 #if defined(INET) || defined(INET6)
251 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
252 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
253     const struct sockaddr *sa);
254 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
255 int	(*carp_attach_p)(struct ifaddr *, int);
256 void	(*carp_detach_p)(struct ifaddr *, bool);
257 #endif
258 #ifdef INET
259 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
260 #endif
261 #ifdef INET6
262 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
263 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
264     const struct in6_addr *taddr);
265 #endif
266 
267 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
268 
269 /*
270  * XXX: Style; these should be sorted alphabetically, and unprototyped
271  * static functions should be prototyped. Currently they are sorted by
272  * declaration order.
273  */
274 static void	if_attachdomain(void *);
275 static void	if_attachdomain1(struct ifnet *);
276 static int	ifconf(u_long, caddr_t);
277 static void	if_input_default(struct ifnet *, struct mbuf *);
278 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
279 static void	if_route(struct ifnet *, int flag, int fam);
280 static int	if_setflag(struct ifnet *, int, int, int *, int);
281 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
282 static void	if_unroute(struct ifnet *, int flag, int fam);
283 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
284 static void	do_link_state_change(void *, int);
285 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
286 static int	if_getgroupmembers(struct ifgroupreq *);
287 static void	if_delgroups(struct ifnet *);
288 static void	if_attach_internal(struct ifnet *, bool);
289 static int	if_detach_internal(struct ifnet *, bool);
290 static void	if_siocaddmulti(void *, int);
291 static void	if_link_ifnet(struct ifnet *);
292 static bool	if_unlink_ifnet(struct ifnet *, bool);
293 #ifdef VIMAGE
294 static int	if_vmove(struct ifnet *, struct vnet *);
295 #endif
296 
297 #ifdef INET6
298 /*
299  * XXX: declare here to avoid to include many inet6 related files..
300  * should be more generalized?
301  */
302 extern void	nd6_setmtu(struct ifnet *);
303 #endif
304 
305 /* ipsec helper hooks */
306 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
307 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
308 
309 int	ifqmaxlen = IFQ_MAXLEN;
310 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
311 VNET_DEFINE(struct ifgrouphead, ifg_head);
312 
313 /* Table of ifnet by index. */
314 static int if_index;
315 static int if_indexlim = 8;
316 static struct ifindex_entry {
317 	struct ifnet	*ife_ifnet;
318 	uint16_t	ife_gencnt;
319 } *ifindex_table;
320 
321 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
322     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
323     "Variables global to all interfaces");
324 static int
325 sysctl_ifcount(SYSCTL_HANDLER_ARGS)
326 {
327 	int rv = 0;
328 
329 	IFNET_RLOCK();
330 	for (int i = 1; i <= if_index; i++)
331 		if (ifindex_table[i].ife_ifnet != NULL &&
332 		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
333 			rv = i;
334 	IFNET_RUNLOCK();
335 
336 	return (sysctl_handle_int(oidp, &rv, 0, req));
337 }
338 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
339     CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
340     "Maximum known interface index");
341 
342 /*
343  * The global network interface list (V_ifnet) and related state (such as
344  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
345  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
346  */
347 struct sx ifnet_sxlock;
348 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
349 
350 struct sx ifnet_detach_sxlock;
351 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
352     SX_RECURSE);
353 
354 #ifdef VIMAGE
355 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
356     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
357 #endif
358 
359 static	if_com_alloc_t *if_com_alloc[256];
360 static	if_com_free_t *if_com_free[256];
361 
362 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
363 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
364 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
365 
366 struct ifnet *
367 ifnet_byindex(u_int idx)
368 {
369 	struct ifnet *ifp;
370 
371 	NET_EPOCH_ASSERT();
372 
373 	if (__predict_false(idx > if_index))
374 		return (NULL);
375 
376 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
377 
378 	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
379 		ifp = NULL;
380 
381 	return (ifp);
382 }
383 
384 struct ifnet *
385 ifnet_byindex_ref(u_int idx)
386 {
387 	struct ifnet *ifp;
388 
389 	ifp = ifnet_byindex(idx);
390 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
391 		return (NULL);
392 	if (!if_try_ref(ifp))
393 		return (NULL);
394 	return (ifp);
395 }
396 
397 struct ifnet *
398 ifnet_byindexgen(uint16_t idx, uint16_t gen)
399 {
400 	struct ifnet *ifp;
401 
402 	NET_EPOCH_ASSERT();
403 
404 	if (__predict_false(idx > if_index))
405 		return (NULL);
406 
407 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
408 
409 	if (ifindex_table[idx].ife_gencnt == gen)
410 		return (ifp);
411 	else
412 		return (NULL);
413 }
414 
415 /*
416  * Network interface utility routines.
417  *
418  * Routines with ifa_ifwith* names take sockaddr *'s as
419  * parameters.
420  */
421 
422 static void
423 if_init(void *arg __unused)
424 {
425 
426 	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
427 	    M_IFNET, M_WAITOK | M_ZERO);
428 }
429 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
430 
431 static void
432 vnet_if_init(const void *unused __unused)
433 {
434 
435 	CK_STAILQ_INIT(&V_ifnet);
436 	CK_STAILQ_INIT(&V_ifg_head);
437 	vnet_if_clone_init();
438 }
439 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
440     NULL);
441 
442 static void
443 if_link_ifnet(struct ifnet *ifp)
444 {
445 
446 	IFNET_WLOCK();
447 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
448 #ifdef VIMAGE
449 	curvnet->vnet_ifcnt++;
450 #endif
451 	IFNET_WUNLOCK();
452 }
453 
454 static bool
455 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
456 {
457 	struct ifnet *iter;
458 	int found = 0;
459 
460 	IFNET_WLOCK();
461 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
462 		if (iter == ifp) {
463 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
464 			if (!vmove)
465 				ifp->if_flags |= IFF_DYING;
466 			found = 1;
467 			break;
468 		}
469 #ifdef VIMAGE
470 	curvnet->vnet_ifcnt--;
471 #endif
472 	IFNET_WUNLOCK();
473 
474 	return (found);
475 }
476 
477 #ifdef VIMAGE
478 static void
479 vnet_if_return(const void *unused __unused)
480 {
481 	struct ifnet *ifp, *nifp;
482 	struct ifnet **pending;
483 	int found __diagused;
484 	int i;
485 
486 	i = 0;
487 
488 	/*
489 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
490 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
491 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
492 	 * complete. We can't do that from within NET_EPOCH.
493 	 *
494 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
495 	 * read/write lock. We cannot hold the lock as we call if_vmove()
496 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
497 	 * ctx lock.
498 	 */
499 	IFNET_WLOCK();
500 
501 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
502 	    M_IFNET, M_WAITOK | M_ZERO);
503 
504 	/* Return all inherited interfaces to their parent vnets. */
505 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
506 		if (ifp->if_home_vnet != ifp->if_vnet) {
507 			found = if_unlink_ifnet(ifp, true);
508 			MPASS(found);
509 
510 			pending[i++] = ifp;
511 		}
512 	}
513 	IFNET_WUNLOCK();
514 
515 	for (int j = 0; j < i; j++) {
516 		if_vmove(pending[j], pending[j]->if_home_vnet);
517 	}
518 
519 	free(pending, M_IFNET);
520 }
521 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
522     vnet_if_return, NULL);
523 #endif
524 
525 /*
526  * Allocate a struct ifnet and an index for an interface.  A layer 2
527  * common structure will also be allocated if an allocation routine is
528  * registered for the passed type.
529  */
530 static struct ifnet *
531 if_alloc_domain(u_char type, int numa_domain)
532 {
533 	struct ifnet *ifp;
534 	u_short idx;
535 
536 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
537 	if (numa_domain == IF_NODOM)
538 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
539 		    M_WAITOK | M_ZERO);
540 	else
541 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
542 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
543 	ifp->if_type = type;
544 	ifp->if_alloctype = type;
545 	ifp->if_numa_domain = numa_domain;
546 #ifdef VIMAGE
547 	ifp->if_vnet = curvnet;
548 #endif
549 	if (if_com_alloc[type] != NULL) {
550 		ifp->if_l2com = if_com_alloc[type](type, ifp);
551 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
552 		    type));
553 	}
554 
555 	IF_ADDR_LOCK_INIT(ifp);
556 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
557 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
558 	ifp->if_afdata_initialized = 0;
559 	IF_AFDATA_LOCK_INIT(ifp);
560 	CK_STAILQ_INIT(&ifp->if_addrhead);
561 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
562 	CK_STAILQ_INIT(&ifp->if_groups);
563 #ifdef MAC
564 	mac_ifnet_init(ifp);
565 #endif
566 	ifq_init(&ifp->if_snd, ifp);
567 
568 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
569 	for (int i = 0; i < IFCOUNTERS; i++)
570 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
571 	ifp->if_get_counter = if_get_counter_default;
572 	ifp->if_pcp = IFNET_PCP_NONE;
573 
574 	/* Allocate an ifindex array entry. */
575 	IFNET_WLOCK();
576 	/*
577 	 * Try to find an empty slot below if_index.  If we fail, take the
578 	 * next slot.
579 	 */
580 	for (idx = 1; idx <= if_index; idx++) {
581 		if (ifindex_table[idx].ife_ifnet == NULL)
582 			break;
583 	}
584 
585 	/* Catch if_index overflow. */
586 	if (idx >= if_indexlim) {
587 		struct ifindex_entry *new, *old;
588 		int newlim;
589 
590 		newlim = if_indexlim * 2;
591 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
592 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
593 		old = ifindex_table;
594 		ck_pr_store_ptr(&ifindex_table, new);
595 		if_indexlim = newlim;
596 		epoch_wait_preempt(net_epoch_preempt);
597 		free(old, M_IFNET);
598 	}
599 	if (idx > if_index)
600 		if_index = idx;
601 
602 	ifp->if_index = idx;
603 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
604 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
605 	IFNET_WUNLOCK();
606 
607 	return (ifp);
608 }
609 
610 struct ifnet *
611 if_alloc_dev(u_char type, device_t dev)
612 {
613 	int numa_domain;
614 
615 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
616 		return (if_alloc_domain(type, IF_NODOM));
617 	return (if_alloc_domain(type, numa_domain));
618 }
619 
620 struct ifnet *
621 if_alloc(u_char type)
622 {
623 
624 	return (if_alloc_domain(type, IF_NODOM));
625 }
626 /*
627  * Do the actual work of freeing a struct ifnet, and layer 2 common
628  * structure.  This call is made when the network epoch guarantees
629  * us that nobody holds a pointer to the interface.
630  */
631 static void
632 if_free_deferred(epoch_context_t ctx)
633 {
634 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
635 
636 	KASSERT((ifp->if_flags & IFF_DYING),
637 	    ("%s: interface not dying", __func__));
638 
639 	if (if_com_free[ifp->if_alloctype] != NULL)
640 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
641 		    ifp->if_alloctype);
642 
643 #ifdef MAC
644 	mac_ifnet_destroy(ifp);
645 #endif /* MAC */
646 	IF_AFDATA_DESTROY(ifp);
647 	IF_ADDR_LOCK_DESTROY(ifp);
648 	ifq_delete(&ifp->if_snd);
649 
650 	for (int i = 0; i < IFCOUNTERS; i++)
651 		counter_u64_free(ifp->if_counters[i]);
652 
653 	free(ifp->if_description, M_IFDESCR);
654 	free(ifp->if_hw_addr, M_IFADDR);
655 	free(ifp, M_IFNET);
656 }
657 
658 /*
659  * Deregister an interface and free the associated storage.
660  */
661 void
662 if_free(struct ifnet *ifp)
663 {
664 
665 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
666 
667 	/*
668 	 * XXXGL: An interface index is really an alias to ifp pointer.
669 	 * Why would we clear the alias now, and not in the deferred
670 	 * context?  Indeed there is nothing wrong with some network
671 	 * thread obtaining ifp via ifnet_byindex() inside the network
672 	 * epoch and then dereferencing ifp while we perform if_free(),
673 	 * and after if_free() finished, too.
674 	 *
675 	 * This early index freeing was important back when ifindex was
676 	 * virtualized and interface would outlive the vnet.
677 	 */
678 	IFNET_WLOCK();
679 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
680 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
681 	ifindex_table[ifp->if_index].ife_gencnt++;
682 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
683 		if_index--;
684 	IFNET_WUNLOCK();
685 
686 	if (refcount_release(&ifp->if_refcount))
687 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
688 }
689 
690 /*
691  * Interfaces to keep an ifnet type-stable despite the possibility of the
692  * driver calling if_free().  If there are additional references, we defer
693  * freeing the underlying data structure.
694  */
695 void
696 if_ref(struct ifnet *ifp)
697 {
698 	u_int old __diagused;
699 
700 	/* We don't assert the ifnet list lock here, but arguably should. */
701 	old = refcount_acquire(&ifp->if_refcount);
702 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
703 }
704 
705 bool
706 if_try_ref(struct ifnet *ifp)
707 {
708 	NET_EPOCH_ASSERT();
709 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
710 }
711 
712 void
713 if_rele(struct ifnet *ifp)
714 {
715 
716 	if (!refcount_release(&ifp->if_refcount))
717 		return;
718 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
719 }
720 
721 void
722 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
723 {
724 
725 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
726 
727 	if (ifq->ifq_maxlen == 0)
728 		ifq->ifq_maxlen = ifqmaxlen;
729 
730 	ifq->altq_type = 0;
731 	ifq->altq_disc = NULL;
732 	ifq->altq_flags &= ALTQF_CANTCHANGE;
733 	ifq->altq_tbr  = NULL;
734 	ifq->altq_ifp  = ifp;
735 }
736 
737 void
738 ifq_delete(struct ifaltq *ifq)
739 {
740 	mtx_destroy(&ifq->ifq_mtx);
741 }
742 
743 /*
744  * Perform generic interface initialization tasks and attach the interface
745  * to the list of "active" interfaces.  If vmove flag is set on entry
746  * to if_attach_internal(), perform only a limited subset of initialization
747  * tasks, given that we are moving from one vnet to another an ifnet which
748  * has already been fully initialized.
749  *
750  * Note that if_detach_internal() removes group membership unconditionally
751  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
752  * Thus, when if_vmove() is applied to a cloned interface, group membership
753  * is lost while a cloned one always joins a group whose name is
754  * ifc->ifc_name.  To recover this after if_detach_internal() and
755  * if_attach_internal(), the cloner should be specified to
756  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
757  * attempts to join a group whose name is ifc->ifc_name.
758  *
759  * XXX:
760  *  - The decision to return void and thus require this function to
761  *    succeed is questionable.
762  *  - We should probably do more sanity checking.  For instance we don't
763  *    do anything to insure if_xname is unique or non-empty.
764  */
765 void
766 if_attach(struct ifnet *ifp)
767 {
768 
769 	if_attach_internal(ifp, false);
770 }
771 
772 /*
773  * Compute the least common TSO limit.
774  */
775 void
776 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
777 {
778 	/*
779 	 * 1) If there is no limit currently, take the limit from
780 	 * the network adapter.
781 	 *
782 	 * 2) If the network adapter has a limit below the current
783 	 * limit, apply it.
784 	 */
785 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
786 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
787 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
788 	}
789 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
790 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
791 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
792 	}
793 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
794 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
795 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
796 	}
797 }
798 
799 /*
800  * Update TSO limit of a network adapter.
801  *
802  * Returns zero if no change. Else non-zero.
803  */
804 int
805 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
806 {
807 	int retval = 0;
808 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
809 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
810 		retval++;
811 	}
812 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
813 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
814 		retval++;
815 	}
816 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
817 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
818 		retval++;
819 	}
820 	return (retval);
821 }
822 
823 static void
824 if_attach_internal(struct ifnet *ifp, bool vmove)
825 {
826 	unsigned socksize, ifasize;
827 	int namelen, masklen;
828 	struct sockaddr_dl *sdl;
829 	struct ifaddr *ifa;
830 
831 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
832 
833 #ifdef VIMAGE
834 	ifp->if_vnet = curvnet;
835 	if (ifp->if_home_vnet == NULL)
836 		ifp->if_home_vnet = curvnet;
837 #endif
838 
839 	if_addgroup(ifp, IFG_ALL);
840 
841 #ifdef VIMAGE
842 	/* Restore group membership for cloned interface. */
843 	if (vmove)
844 		if_clone_restoregroup(ifp);
845 #endif
846 
847 	getmicrotime(&ifp->if_lastchange);
848 	ifp->if_epoch = time_uptime;
849 
850 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
851 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
852 	    ("transmit and qflush must both either be set or both be NULL"));
853 	if (ifp->if_transmit == NULL) {
854 		ifp->if_transmit = if_transmit;
855 		ifp->if_qflush = if_qflush;
856 	}
857 	if (ifp->if_input == NULL)
858 		ifp->if_input = if_input_default;
859 
860 	if (ifp->if_requestencap == NULL)
861 		ifp->if_requestencap = if_requestencap_default;
862 
863 	if (!vmove) {
864 #ifdef MAC
865 		mac_ifnet_create(ifp);
866 #endif
867 
868 		/*
869 		 * Create a Link Level name for this device.
870 		 */
871 		namelen = strlen(ifp->if_xname);
872 		/*
873 		 * Always save enough space for any possiable name so we
874 		 * can do a rename in place later.
875 		 */
876 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
877 		socksize = masklen + ifp->if_addrlen;
878 		if (socksize < sizeof(*sdl))
879 			socksize = sizeof(*sdl);
880 		socksize = roundup2(socksize, sizeof(long));
881 		ifasize = sizeof(*ifa) + 2 * socksize;
882 		ifa = ifa_alloc(ifasize, M_WAITOK);
883 		sdl = (struct sockaddr_dl *)(ifa + 1);
884 		sdl->sdl_len = socksize;
885 		sdl->sdl_family = AF_LINK;
886 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
887 		sdl->sdl_nlen = namelen;
888 		sdl->sdl_index = ifp->if_index;
889 		sdl->sdl_type = ifp->if_type;
890 		ifp->if_addr = ifa;
891 		ifa->ifa_ifp = ifp;
892 		ifa->ifa_addr = (struct sockaddr *)sdl;
893 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
894 		ifa->ifa_netmask = (struct sockaddr *)sdl;
895 		sdl->sdl_len = masklen;
896 		while (namelen != 0)
897 			sdl->sdl_data[--namelen] = 0xff;
898 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
899 		/* Reliably crash if used uninitialized. */
900 		ifp->if_broadcastaddr = NULL;
901 
902 		if (ifp->if_type == IFT_ETHER) {
903 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
904 			    M_WAITOK | M_ZERO);
905 		}
906 
907 #if defined(INET) || defined(INET6)
908 		/* Use defaults for TSO, if nothing is set */
909 		if (ifp->if_hw_tsomax == 0 &&
910 		    ifp->if_hw_tsomaxsegcount == 0 &&
911 		    ifp->if_hw_tsomaxsegsize == 0) {
912 			/*
913 			 * The TSO defaults needs to be such that an
914 			 * NFS mbuf list of 35 mbufs totalling just
915 			 * below 64K works and that a chain of mbufs
916 			 * can be defragged into at most 32 segments:
917 			 */
918 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
919 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
920 			ifp->if_hw_tsomaxsegcount = 35;
921 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
922 
923 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
924 			if (ifp->if_capabilities & IFCAP_TSO) {
925 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
926 				    ifp->if_hw_tsomax,
927 				    ifp->if_hw_tsomaxsegcount,
928 				    ifp->if_hw_tsomaxsegsize);
929 			}
930 		}
931 #endif
932 	}
933 #ifdef VIMAGE
934 	else {
935 		/*
936 		 * Update the interface index in the link layer address
937 		 * of the interface.
938 		 */
939 		for (ifa = ifp->if_addr; ifa != NULL;
940 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
941 			if (ifa->ifa_addr->sa_family == AF_LINK) {
942 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
943 				sdl->sdl_index = ifp->if_index;
944 			}
945 		}
946 	}
947 #endif
948 
949 	if_link_ifnet(ifp);
950 
951 	if (domain_init_status >= 2)
952 		if_attachdomain1(ifp);
953 
954 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
955 	if (IS_DEFAULT_VNET(curvnet))
956 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
957 
958 	/* Announce the interface. */
959 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
960 }
961 
962 static void
963 if_epochalloc(void *dummy __unused)
964 {
965 
966 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
967 }
968 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
969 
970 static void
971 if_attachdomain(void *dummy)
972 {
973 	struct ifnet *ifp;
974 
975 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
976 		if_attachdomain1(ifp);
977 }
978 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
979     if_attachdomain, NULL);
980 
981 static void
982 if_attachdomain1(struct ifnet *ifp)
983 {
984 	struct domain *dp;
985 
986 	/*
987 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
988 	 * cannot lock ifp->if_afdata initialization, entirely.
989 	 */
990 	IF_AFDATA_LOCK(ifp);
991 	if (ifp->if_afdata_initialized >= domain_init_status) {
992 		IF_AFDATA_UNLOCK(ifp);
993 		log(LOG_WARNING, "%s called more than once on %s\n",
994 		    __func__, ifp->if_xname);
995 		return;
996 	}
997 	ifp->if_afdata_initialized = domain_init_status;
998 	IF_AFDATA_UNLOCK(ifp);
999 
1000 	/* address family dependent data region */
1001 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1002 	for (dp = domains; dp; dp = dp->dom_next) {
1003 		if (dp->dom_ifattach)
1004 			ifp->if_afdata[dp->dom_family] =
1005 			    (*dp->dom_ifattach)(ifp);
1006 	}
1007 }
1008 
1009 /*
1010  * Remove any unicast or broadcast network addresses from an interface.
1011  */
1012 void
1013 if_purgeaddrs(struct ifnet *ifp)
1014 {
1015 	struct ifaddr *ifa;
1016 
1017 	while (1) {
1018 		struct epoch_tracker et;
1019 
1020 		NET_EPOCH_ENTER(et);
1021 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1022 			if (ifa->ifa_addr->sa_family != AF_LINK)
1023 				break;
1024 		}
1025 		NET_EPOCH_EXIT(et);
1026 
1027 		if (ifa == NULL)
1028 			break;
1029 #ifdef INET
1030 		/* XXX: Ugly!! ad hoc just for INET */
1031 		if (ifa->ifa_addr->sa_family == AF_INET) {
1032 			struct ifaliasreq ifr;
1033 
1034 			bzero(&ifr, sizeof(ifr));
1035 			ifr.ifra_addr = *ifa->ifa_addr;
1036 			if (ifa->ifa_dstaddr)
1037 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1038 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1039 			    NULL) == 0)
1040 				continue;
1041 		}
1042 #endif /* INET */
1043 #ifdef INET6
1044 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1045 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
1046 			/* ifp_addrhead is already updated */
1047 			continue;
1048 		}
1049 #endif /* INET6 */
1050 		IF_ADDR_WLOCK(ifp);
1051 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1052 		IF_ADDR_WUNLOCK(ifp);
1053 		ifa_free(ifa);
1054 	}
1055 }
1056 
1057 /*
1058  * Remove any multicast network addresses from an interface when an ifnet
1059  * is going away.
1060  */
1061 static void
1062 if_purgemaddrs(struct ifnet *ifp)
1063 {
1064 	struct ifmultiaddr *ifma;
1065 
1066 	IF_ADDR_WLOCK(ifp);
1067 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1068 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1069 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1070 		if_delmulti_locked(ifp, ifma, 1);
1071 	}
1072 	IF_ADDR_WUNLOCK(ifp);
1073 }
1074 
1075 /*
1076  * Detach an interface, removing it from the list of "active" interfaces.
1077  * If vmove flag is set on entry to if_detach_internal(), perform only a
1078  * limited subset of cleanup tasks, given that we are moving an ifnet from
1079  * one vnet to another, where it must be fully operational.
1080  *
1081  * XXXRW: There are some significant questions about event ordering, and
1082  * how to prevent things from starting to use the interface during detach.
1083  */
1084 void
1085 if_detach(struct ifnet *ifp)
1086 {
1087 	bool found;
1088 
1089 	CURVNET_SET_QUIET(ifp->if_vnet);
1090 	found = if_unlink_ifnet(ifp, false);
1091 	if (found) {
1092 		sx_xlock(&ifnet_detach_sxlock);
1093 		if_detach_internal(ifp, false);
1094 		sx_xunlock(&ifnet_detach_sxlock);
1095 	}
1096 	CURVNET_RESTORE();
1097 }
1098 
1099 /*
1100  * The vmove flag, if set, indicates that we are called from a callpath
1101  * that is moving an interface to a different vnet instance.
1102  *
1103  * The shutdown flag, if set, indicates that we are called in the
1104  * process of shutting down a vnet instance.  Currently only the
1105  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1106  * on a vnet instance shutdown without this flag being set, e.g., when
1107  * the cloned interfaces are destoyed as first thing of teardown.
1108  */
1109 static int
1110 if_detach_internal(struct ifnet *ifp, bool vmove)
1111 {
1112 	struct ifaddr *ifa;
1113 	int i;
1114 	struct domain *dp;
1115 #ifdef VIMAGE
1116 	bool shutdown;
1117 
1118 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1119 #endif
1120 
1121 	/*
1122 	 * At this point we know the interface still was on the ifnet list
1123 	 * and we removed it so we are in a stable state.
1124 	 */
1125 	epoch_wait_preempt(net_epoch_preempt);
1126 
1127 	/*
1128 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1129 	 * fixes issues about late destruction of multicast options
1130 	 * which lead to leave group calls, which in turn access the
1131 	 * belonging ifnet structure:
1132 	 */
1133 	epoch_drain_callbacks(net_epoch_preempt);
1134 
1135 	/*
1136 	 * In any case (destroy or vmove) detach us from the groups
1137 	 * and remove/wait for pending events on the taskq.
1138 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1139 	 */
1140 	if_delgroups(ifp);
1141 
1142 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1143 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1144 
1145 	if_down(ifp);
1146 
1147 #ifdef VIMAGE
1148 	/*
1149 	 * On VNET shutdown abort here as the stack teardown will do all
1150 	 * the work top-down for us.
1151 	 */
1152 	if (shutdown) {
1153 		/* Give interface users the chance to clean up. */
1154 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1155 
1156 		/*
1157 		 * In case of a vmove we are done here without error.
1158 		 * If we would signal an error it would lead to the same
1159 		 * abort as if we did not find the ifnet anymore.
1160 		 * if_detach() calls us in void context and does not care
1161 		 * about an early abort notification, so life is splendid :)
1162 		 */
1163 		goto finish_vnet_shutdown;
1164 	}
1165 #endif
1166 
1167 	/*
1168 	 * At this point we are not tearing down a VNET and are either
1169 	 * going to destroy or vmove the interface and have to cleanup
1170 	 * accordingly.
1171 	 */
1172 
1173 	/*
1174 	 * Remove routes and flush queues.
1175 	 */
1176 #ifdef ALTQ
1177 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1178 		altq_disable(&ifp->if_snd);
1179 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1180 		altq_detach(&ifp->if_snd);
1181 #endif
1182 
1183 	if_purgeaddrs(ifp);
1184 
1185 #ifdef INET
1186 	in_ifdetach(ifp);
1187 #endif
1188 
1189 #ifdef INET6
1190 	/*
1191 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1192 	 * before removing routing entries below, since IPv6 interface direct
1193 	 * routes are expected to be removed by the IPv6-specific kernel API.
1194 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1195 	 */
1196 	in6_ifdetach(ifp);
1197 #endif
1198 	if_purgemaddrs(ifp);
1199 
1200 	/* Announce that the interface is gone. */
1201 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1202 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1203 	if (IS_DEFAULT_VNET(curvnet))
1204 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1205 
1206 	if (!vmove) {
1207 		/*
1208 		 * Prevent further calls into the device driver via ifnet.
1209 		 */
1210 		if_dead(ifp);
1211 
1212 		/*
1213 		 * Clean up all addresses.
1214 		 */
1215 		IF_ADDR_WLOCK(ifp);
1216 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1217 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1218 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1219 			IF_ADDR_WUNLOCK(ifp);
1220 			ifa_free(ifa);
1221 		} else
1222 			IF_ADDR_WUNLOCK(ifp);
1223 	}
1224 
1225 	rt_flushifroutes(ifp);
1226 
1227 #ifdef VIMAGE
1228 finish_vnet_shutdown:
1229 #endif
1230 	/*
1231 	 * We cannot hold the lock over dom_ifdetach calls as they might
1232 	 * sleep, for example trying to drain a callout, thus open up the
1233 	 * theoretical race with re-attaching.
1234 	 */
1235 	IF_AFDATA_LOCK(ifp);
1236 	i = ifp->if_afdata_initialized;
1237 	ifp->if_afdata_initialized = 0;
1238 	IF_AFDATA_UNLOCK(ifp);
1239 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1240 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1241 			(*dp->dom_ifdetach)(ifp,
1242 			    ifp->if_afdata[dp->dom_family]);
1243 			ifp->if_afdata[dp->dom_family] = NULL;
1244 		}
1245 	}
1246 
1247 	return (0);
1248 }
1249 
1250 #ifdef VIMAGE
1251 /*
1252  * if_vmove() performs a limited version of if_detach() in current
1253  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1254  */
1255 static int
1256 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1257 {
1258 #ifdef DEV_BPF
1259 	u_int bif_dlt, bif_hdrlen;
1260 #endif
1261 	int rc;
1262 
1263 #ifdef DEV_BPF
1264  	/*
1265 	 * if_detach_internal() will call the eventhandler to notify
1266 	 * interface departure.  That will detach if_bpf.  We need to
1267 	 * safe the dlt and hdrlen so we can re-attach it later.
1268 	 */
1269 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1270 #endif
1271 
1272 	/*
1273 	 * Detach from current vnet, but preserve LLADDR info, do not
1274 	 * mark as dead etc. so that the ifnet can be reattached later.
1275 	 * If we cannot find it, we lost the race to someone else.
1276 	 */
1277 	rc = if_detach_internal(ifp, true);
1278 	if (rc != 0)
1279 		return (rc);
1280 
1281 	/*
1282 	 * Perform interface-specific reassignment tasks, if provided by
1283 	 * the driver.
1284 	 */
1285 	if (ifp->if_reassign != NULL)
1286 		ifp->if_reassign(ifp, new_vnet, NULL);
1287 
1288 	/*
1289 	 * Switch to the context of the target vnet.
1290 	 */
1291 	CURVNET_SET_QUIET(new_vnet);
1292 	if_attach_internal(ifp, true);
1293 
1294 #ifdef DEV_BPF
1295 	if (ifp->if_bpf == NULL)
1296 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1297 #endif
1298 
1299 	CURVNET_RESTORE();
1300 	return (0);
1301 }
1302 
1303 /*
1304  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1305  */
1306 static int
1307 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1308 {
1309 	struct prison *pr;
1310 	struct ifnet *difp;
1311 	int error;
1312 	bool found __diagused;
1313 	bool shutdown;
1314 
1315 	/* Try to find the prison within our visibility. */
1316 	sx_slock(&allprison_lock);
1317 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1318 	sx_sunlock(&allprison_lock);
1319 	if (pr == NULL)
1320 		return (ENXIO);
1321 	prison_hold_locked(pr);
1322 	mtx_unlock(&pr->pr_mtx);
1323 
1324 	/* Do not try to move the iface from and to the same prison. */
1325 	if (pr->pr_vnet == ifp->if_vnet) {
1326 		prison_free(pr);
1327 		return (EEXIST);
1328 	}
1329 
1330 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1331 	/* XXX Lock interfaces to avoid races. */
1332 	CURVNET_SET_QUIET(pr->pr_vnet);
1333 	difp = ifunit(ifname);
1334 	if (difp != NULL) {
1335 		CURVNET_RESTORE();
1336 		prison_free(pr);
1337 		return (EEXIST);
1338 	}
1339 
1340 	/* Make sure the VNET is stable. */
1341 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1342 	if (shutdown) {
1343 		CURVNET_RESTORE();
1344 		prison_free(pr);
1345 		return (EBUSY);
1346 	}
1347 	CURVNET_RESTORE();
1348 
1349 	found = if_unlink_ifnet(ifp, true);
1350 	MPASS(found);
1351 
1352 	/* Move the interface into the child jail/vnet. */
1353 	error = if_vmove(ifp, pr->pr_vnet);
1354 
1355 	/* Report the new if_xname back to the userland on success. */
1356 	if (error == 0)
1357 		sprintf(ifname, "%s", ifp->if_xname);
1358 
1359 	prison_free(pr);
1360 	return (error);
1361 }
1362 
1363 static int
1364 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1365 {
1366 	struct prison *pr;
1367 	struct vnet *vnet_dst;
1368 	struct ifnet *ifp;
1369 	int error, found __diagused;
1370  	bool shutdown;
1371 
1372 	/* Try to find the prison within our visibility. */
1373 	sx_slock(&allprison_lock);
1374 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1375 	sx_sunlock(&allprison_lock);
1376 	if (pr == NULL)
1377 		return (ENXIO);
1378 	prison_hold_locked(pr);
1379 	mtx_unlock(&pr->pr_mtx);
1380 
1381 	/* Make sure the named iface exists in the source prison/vnet. */
1382 	CURVNET_SET(pr->pr_vnet);
1383 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1384 	if (ifp == NULL) {
1385 		CURVNET_RESTORE();
1386 		prison_free(pr);
1387 		return (ENXIO);
1388 	}
1389 
1390 	/* Do not try to move the iface from and to the same prison. */
1391 	vnet_dst = TD_TO_VNET(td);
1392 	if (vnet_dst == ifp->if_vnet) {
1393 		CURVNET_RESTORE();
1394 		prison_free(pr);
1395 		return (EEXIST);
1396 	}
1397 
1398 	/* Make sure the VNET is stable. */
1399 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1400 	if (shutdown) {
1401 		CURVNET_RESTORE();
1402 		prison_free(pr);
1403 		return (EBUSY);
1404 	}
1405 
1406 	/* Get interface back from child jail/vnet. */
1407 	found = if_unlink_ifnet(ifp, true);
1408 	MPASS(found);
1409 	error = if_vmove(ifp, vnet_dst);
1410 	CURVNET_RESTORE();
1411 
1412 	/* Report the new if_xname back to the userland on success. */
1413 	if (error == 0)
1414 		sprintf(ifname, "%s", ifp->if_xname);
1415 
1416 	prison_free(pr);
1417 	return (error);
1418 }
1419 #endif /* VIMAGE */
1420 
1421 /*
1422  * Add a group to an interface
1423  */
1424 int
1425 if_addgroup(struct ifnet *ifp, const char *groupname)
1426 {
1427 	struct ifg_list		*ifgl;
1428 	struct ifg_group	*ifg = NULL;
1429 	struct ifg_member	*ifgm;
1430 	int 			 new = 0;
1431 
1432 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1433 	    groupname[strlen(groupname) - 1] <= '9')
1434 		return (EINVAL);
1435 
1436 	IFNET_WLOCK();
1437 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1438 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1439 			IFNET_WUNLOCK();
1440 			return (EEXIST);
1441 		}
1442 
1443 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1444 	    	IFNET_WUNLOCK();
1445 		return (ENOMEM);
1446 	}
1447 
1448 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1449 		free(ifgl, M_TEMP);
1450 		IFNET_WUNLOCK();
1451 		return (ENOMEM);
1452 	}
1453 
1454 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1455 		if (!strcmp(ifg->ifg_group, groupname))
1456 			break;
1457 
1458 	if (ifg == NULL) {
1459 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1460 			free(ifgl, M_TEMP);
1461 			free(ifgm, M_TEMP);
1462 			IFNET_WUNLOCK();
1463 			return (ENOMEM);
1464 		}
1465 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1466 		ifg->ifg_refcnt = 0;
1467 		CK_STAILQ_INIT(&ifg->ifg_members);
1468 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1469 		new = 1;
1470 	}
1471 
1472 	ifg->ifg_refcnt++;
1473 	ifgl->ifgl_group = ifg;
1474 	ifgm->ifgm_ifp = ifp;
1475 
1476 	IF_ADDR_WLOCK(ifp);
1477 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1478 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1479 	IF_ADDR_WUNLOCK(ifp);
1480 
1481 	IFNET_WUNLOCK();
1482 
1483 	if (new)
1484 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1485 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1486 
1487 	return (0);
1488 }
1489 
1490 /*
1491  * Helper function to remove a group out of an interface.  Expects the global
1492  * ifnet lock to be write-locked, and drops it before returning.
1493  */
1494 static void
1495 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1496     const char *groupname)
1497 {
1498 	struct ifg_member *ifgm;
1499 	bool freeifgl;
1500 
1501 	IFNET_WLOCK_ASSERT();
1502 
1503 	IF_ADDR_WLOCK(ifp);
1504 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1505 	IF_ADDR_WUNLOCK(ifp);
1506 
1507 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1508 		if (ifgm->ifgm_ifp == ifp) {
1509 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1510 			    ifg_member, ifgm_next);
1511 			break;
1512 		}
1513 	}
1514 
1515 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1516 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1517 		    ifg_next);
1518 		freeifgl = true;
1519 	} else {
1520 		freeifgl = false;
1521 	}
1522 	IFNET_WUNLOCK();
1523 
1524 	epoch_wait_preempt(net_epoch_preempt);
1525 	if (freeifgl) {
1526 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1527 		free(ifgl->ifgl_group, M_TEMP);
1528 	}
1529 	free(ifgm, M_TEMP);
1530 	free(ifgl, M_TEMP);
1531 
1532 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1533 }
1534 
1535 /*
1536  * Remove a group from an interface
1537  */
1538 int
1539 if_delgroup(struct ifnet *ifp, const char *groupname)
1540 {
1541 	struct ifg_list *ifgl;
1542 
1543 	IFNET_WLOCK();
1544 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1545 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1546 			break;
1547 	if (ifgl == NULL) {
1548 		IFNET_WUNLOCK();
1549 		return (ENOENT);
1550 	}
1551 
1552 	_if_delgroup_locked(ifp, ifgl, groupname);
1553 
1554 	return (0);
1555 }
1556 
1557 /*
1558  * Remove an interface from all groups
1559  */
1560 static void
1561 if_delgroups(struct ifnet *ifp)
1562 {
1563 	struct ifg_list *ifgl;
1564 	char groupname[IFNAMSIZ];
1565 
1566 	IFNET_WLOCK();
1567 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1568 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1569 		_if_delgroup_locked(ifp, ifgl, groupname);
1570 		IFNET_WLOCK();
1571 	}
1572 	IFNET_WUNLOCK();
1573 }
1574 
1575 /*
1576  * Stores all groups from an interface in memory pointed to by ifgr.
1577  */
1578 static int
1579 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1580 {
1581 	int			 len, error;
1582 	struct ifg_list		*ifgl;
1583 	struct ifg_req		 ifgrq, *ifgp;
1584 
1585 	NET_EPOCH_ASSERT();
1586 
1587 	if (ifgr->ifgr_len == 0) {
1588 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1589 			ifgr->ifgr_len += sizeof(struct ifg_req);
1590 		return (0);
1591 	}
1592 
1593 	len = ifgr->ifgr_len;
1594 	ifgp = ifgr->ifgr_groups;
1595 	/* XXX: wire */
1596 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1597 		if (len < sizeof(ifgrq))
1598 			return (EINVAL);
1599 		bzero(&ifgrq, sizeof ifgrq);
1600 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1601 		    sizeof(ifgrq.ifgrq_group));
1602 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1603 			return (error);
1604 		len -= sizeof(ifgrq);
1605 		ifgp++;
1606 	}
1607 
1608 	return (0);
1609 }
1610 
1611 /*
1612  * Stores all members of a group in memory pointed to by igfr
1613  */
1614 static int
1615 if_getgroupmembers(struct ifgroupreq *ifgr)
1616 {
1617 	struct ifg_group	*ifg;
1618 	struct ifg_member	*ifgm;
1619 	struct ifg_req		 ifgrq, *ifgp;
1620 	int			 len, error;
1621 
1622 	IFNET_RLOCK();
1623 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1624 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1625 			break;
1626 	if (ifg == NULL) {
1627 		IFNET_RUNLOCK();
1628 		return (ENOENT);
1629 	}
1630 
1631 	if (ifgr->ifgr_len == 0) {
1632 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1633 			ifgr->ifgr_len += sizeof(ifgrq);
1634 		IFNET_RUNLOCK();
1635 		return (0);
1636 	}
1637 
1638 	len = ifgr->ifgr_len;
1639 	ifgp = ifgr->ifgr_groups;
1640 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1641 		if (len < sizeof(ifgrq)) {
1642 			IFNET_RUNLOCK();
1643 			return (EINVAL);
1644 		}
1645 		bzero(&ifgrq, sizeof ifgrq);
1646 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1647 		    sizeof(ifgrq.ifgrq_member));
1648 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1649 			IFNET_RUNLOCK();
1650 			return (error);
1651 		}
1652 		len -= sizeof(ifgrq);
1653 		ifgp++;
1654 	}
1655 	IFNET_RUNLOCK();
1656 
1657 	return (0);
1658 }
1659 
1660 /*
1661  * Return counter values from counter(9)s stored in ifnet.
1662  */
1663 uint64_t
1664 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1665 {
1666 
1667 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1668 
1669 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1670 }
1671 
1672 /*
1673  * Increase an ifnet counter. Usually used for counters shared
1674  * between the stack and a driver, but function supports them all.
1675  */
1676 void
1677 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1678 {
1679 
1680 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1681 
1682 	counter_u64_add(ifp->if_counters[cnt], inc);
1683 }
1684 
1685 /*
1686  * Copy data from ifnet to userland API structure if_data.
1687  */
1688 void
1689 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1690 {
1691 
1692 	ifd->ifi_type = ifp->if_type;
1693 	ifd->ifi_physical = 0;
1694 	ifd->ifi_addrlen = ifp->if_addrlen;
1695 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1696 	ifd->ifi_link_state = ifp->if_link_state;
1697 	ifd->ifi_vhid = 0;
1698 	ifd->ifi_datalen = sizeof(struct if_data);
1699 	ifd->ifi_mtu = ifp->if_mtu;
1700 	ifd->ifi_metric = ifp->if_metric;
1701 	ifd->ifi_baudrate = ifp->if_baudrate;
1702 	ifd->ifi_hwassist = ifp->if_hwassist;
1703 	ifd->ifi_epoch = ifp->if_epoch;
1704 	ifd->ifi_lastchange = ifp->if_lastchange;
1705 
1706 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1707 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1708 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1709 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1710 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1711 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1712 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1713 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1714 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1715 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1716 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1717 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1718 }
1719 
1720 /*
1721  * Initialization, destruction and refcounting functions for ifaddrs.
1722  */
1723 struct ifaddr *
1724 ifa_alloc(size_t size, int flags)
1725 {
1726 	struct ifaddr *ifa;
1727 
1728 	KASSERT(size >= sizeof(struct ifaddr),
1729 	    ("%s: invalid size %zu", __func__, size));
1730 
1731 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1732 	if (ifa == NULL)
1733 		return (NULL);
1734 
1735 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1736 		goto fail;
1737 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1738 		goto fail;
1739 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1740 		goto fail;
1741 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1742 		goto fail;
1743 
1744 	refcount_init(&ifa->ifa_refcnt, 1);
1745 
1746 	return (ifa);
1747 
1748 fail:
1749 	/* free(NULL) is okay */
1750 	counter_u64_free(ifa->ifa_opackets);
1751 	counter_u64_free(ifa->ifa_ipackets);
1752 	counter_u64_free(ifa->ifa_obytes);
1753 	counter_u64_free(ifa->ifa_ibytes);
1754 	free(ifa, M_IFADDR);
1755 
1756 	return (NULL);
1757 }
1758 
1759 void
1760 ifa_ref(struct ifaddr *ifa)
1761 {
1762 	u_int old __diagused;
1763 
1764 	old = refcount_acquire(&ifa->ifa_refcnt);
1765 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1766 }
1767 
1768 int
1769 ifa_try_ref(struct ifaddr *ifa)
1770 {
1771 
1772 	NET_EPOCH_ASSERT();
1773 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1774 }
1775 
1776 static void
1777 ifa_destroy(epoch_context_t ctx)
1778 {
1779 	struct ifaddr *ifa;
1780 
1781 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1782 	counter_u64_free(ifa->ifa_opackets);
1783 	counter_u64_free(ifa->ifa_ipackets);
1784 	counter_u64_free(ifa->ifa_obytes);
1785 	counter_u64_free(ifa->ifa_ibytes);
1786 	free(ifa, M_IFADDR);
1787 }
1788 
1789 void
1790 ifa_free(struct ifaddr *ifa)
1791 {
1792 
1793 	if (refcount_release(&ifa->ifa_refcnt))
1794 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1795 }
1796 
1797 /*
1798  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1799  * structs used to represent other address families, it is necessary
1800  * to perform a different comparison.
1801  */
1802 
1803 #define	sa_dl_equal(a1, a2)	\
1804 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1805 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1806 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1807 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1808 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1809 
1810 /*
1811  * Locate an interface based on a complete address.
1812  */
1813 /*ARGSUSED*/
1814 struct ifaddr *
1815 ifa_ifwithaddr(const struct sockaddr *addr)
1816 {
1817 	struct ifnet *ifp;
1818 	struct ifaddr *ifa;
1819 
1820 	NET_EPOCH_ASSERT();
1821 
1822 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1823 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1824 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1825 				continue;
1826 			if (sa_equal(addr, ifa->ifa_addr)) {
1827 				goto done;
1828 			}
1829 			/* IP6 doesn't have broadcast */
1830 			if ((ifp->if_flags & IFF_BROADCAST) &&
1831 			    ifa->ifa_broadaddr &&
1832 			    ifa->ifa_broadaddr->sa_len != 0 &&
1833 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1834 				goto done;
1835 			}
1836 		}
1837 	}
1838 	ifa = NULL;
1839 done:
1840 	return (ifa);
1841 }
1842 
1843 int
1844 ifa_ifwithaddr_check(const struct sockaddr *addr)
1845 {
1846 	struct epoch_tracker et;
1847 	int rc;
1848 
1849 	NET_EPOCH_ENTER(et);
1850 	rc = (ifa_ifwithaddr(addr) != NULL);
1851 	NET_EPOCH_EXIT(et);
1852 	return (rc);
1853 }
1854 
1855 /*
1856  * Locate an interface based on the broadcast address.
1857  */
1858 /* ARGSUSED */
1859 struct ifaddr *
1860 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1861 {
1862 	struct ifnet *ifp;
1863 	struct ifaddr *ifa;
1864 
1865 	NET_EPOCH_ASSERT();
1866 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1867 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1868 			continue;
1869 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1870 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1871 				continue;
1872 			if ((ifp->if_flags & IFF_BROADCAST) &&
1873 			    ifa->ifa_broadaddr &&
1874 			    ifa->ifa_broadaddr->sa_len != 0 &&
1875 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1876 				goto done;
1877 			}
1878 		}
1879 	}
1880 	ifa = NULL;
1881 done:
1882 	return (ifa);
1883 }
1884 
1885 /*
1886  * Locate the point to point interface with a given destination address.
1887  */
1888 /*ARGSUSED*/
1889 struct ifaddr *
1890 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1891 {
1892 	struct ifnet *ifp;
1893 	struct ifaddr *ifa;
1894 
1895 	NET_EPOCH_ASSERT();
1896 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1897 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1898 			continue;
1899 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1900 			continue;
1901 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1902 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1903 				continue;
1904 			if (ifa->ifa_dstaddr != NULL &&
1905 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1906 				goto done;
1907 			}
1908 		}
1909 	}
1910 	ifa = NULL;
1911 done:
1912 	return (ifa);
1913 }
1914 
1915 /*
1916  * Find an interface on a specific network.  If many, choice
1917  * is most specific found.
1918  */
1919 struct ifaddr *
1920 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1921 {
1922 	struct ifnet *ifp;
1923 	struct ifaddr *ifa;
1924 	struct ifaddr *ifa_maybe = NULL;
1925 	u_int af = addr->sa_family;
1926 	const char *addr_data = addr->sa_data, *cplim;
1927 
1928 	NET_EPOCH_ASSERT();
1929 	/*
1930 	 * AF_LINK addresses can be looked up directly by their index number,
1931 	 * so do that if we can.
1932 	 */
1933 	if (af == AF_LINK) {
1934 		ifp = ifnet_byindex(
1935 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1936 		return (ifp ? ifp->if_addr : NULL);
1937 	}
1938 
1939 	/*
1940 	 * Scan though each interface, looking for ones that have addresses
1941 	 * in this address family and the requested fib.
1942 	 */
1943 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1944 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1945 			continue;
1946 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1947 			const char *cp, *cp2, *cp3;
1948 
1949 			if (ifa->ifa_addr->sa_family != af)
1950 next:				continue;
1951 			if (af == AF_INET &&
1952 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1953 				/*
1954 				 * This is a bit broken as it doesn't
1955 				 * take into account that the remote end may
1956 				 * be a single node in the network we are
1957 				 * looking for.
1958 				 * The trouble is that we don't know the
1959 				 * netmask for the remote end.
1960 				 */
1961 				if (ifa->ifa_dstaddr != NULL &&
1962 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1963 					goto done;
1964 				}
1965 			} else {
1966 				/*
1967 				 * Scan all the bits in the ifa's address.
1968 				 * If a bit dissagrees with what we are
1969 				 * looking for, mask it with the netmask
1970 				 * to see if it really matters.
1971 				 * (A byte at a time)
1972 				 */
1973 				if (ifa->ifa_netmask == 0)
1974 					continue;
1975 				cp = addr_data;
1976 				cp2 = ifa->ifa_addr->sa_data;
1977 				cp3 = ifa->ifa_netmask->sa_data;
1978 				cplim = ifa->ifa_netmask->sa_len
1979 					+ (char *)ifa->ifa_netmask;
1980 				while (cp3 < cplim)
1981 					if ((*cp++ ^ *cp2++) & *cp3++)
1982 						goto next; /* next address! */
1983 				/*
1984 				 * If the netmask of what we just found
1985 				 * is more specific than what we had before
1986 				 * (if we had one), or if the virtual status
1987 				 * of new prefix is better than of the old one,
1988 				 * then remember the new one before continuing
1989 				 * to search for an even better one.
1990 				 */
1991 				if (ifa_maybe == NULL ||
1992 				    ifa_preferred(ifa_maybe, ifa) ||
1993 				    rn_refines((caddr_t)ifa->ifa_netmask,
1994 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1995 					ifa_maybe = ifa;
1996 				}
1997 			}
1998 		}
1999 	}
2000 	ifa = ifa_maybe;
2001 	ifa_maybe = NULL;
2002 done:
2003 	return (ifa);
2004 }
2005 
2006 /*
2007  * Find an interface address specific to an interface best matching
2008  * a given address.
2009  */
2010 struct ifaddr *
2011 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2012 {
2013 	struct ifaddr *ifa;
2014 	const char *cp, *cp2, *cp3;
2015 	char *cplim;
2016 	struct ifaddr *ifa_maybe = NULL;
2017 	u_int af = addr->sa_family;
2018 
2019 	if (af >= AF_MAX)
2020 		return (NULL);
2021 
2022 	NET_EPOCH_ASSERT();
2023 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2024 		if (ifa->ifa_addr->sa_family != af)
2025 			continue;
2026 		if (ifa_maybe == NULL)
2027 			ifa_maybe = ifa;
2028 		if (ifa->ifa_netmask == 0) {
2029 			if (sa_equal(addr, ifa->ifa_addr) ||
2030 			    (ifa->ifa_dstaddr &&
2031 			    sa_equal(addr, ifa->ifa_dstaddr)))
2032 				goto done;
2033 			continue;
2034 		}
2035 		if (ifp->if_flags & IFF_POINTOPOINT) {
2036 			if (sa_equal(addr, ifa->ifa_dstaddr))
2037 				goto done;
2038 		} else {
2039 			cp = addr->sa_data;
2040 			cp2 = ifa->ifa_addr->sa_data;
2041 			cp3 = ifa->ifa_netmask->sa_data;
2042 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2043 			for (; cp3 < cplim; cp3++)
2044 				if ((*cp++ ^ *cp2++) & *cp3)
2045 					break;
2046 			if (cp3 == cplim)
2047 				goto done;
2048 		}
2049 	}
2050 	ifa = ifa_maybe;
2051 done:
2052 	return (ifa);
2053 }
2054 
2055 /*
2056  * See whether new ifa is better than current one:
2057  * 1) A non-virtual one is preferred over virtual.
2058  * 2) A virtual in master state preferred over any other state.
2059  *
2060  * Used in several address selecting functions.
2061  */
2062 int
2063 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2064 {
2065 
2066 	return (cur->ifa_carp && (!next->ifa_carp ||
2067 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2068 }
2069 
2070 struct sockaddr_dl *
2071 link_alloc_sdl(size_t size, int flags)
2072 {
2073 
2074 	return (malloc(size, M_TEMP, flags));
2075 }
2076 
2077 void
2078 link_free_sdl(struct sockaddr *sa)
2079 {
2080 	free(sa, M_TEMP);
2081 }
2082 
2083 /*
2084  * Fills in given sdl with interface basic info.
2085  * Returns pointer to filled sdl.
2086  */
2087 struct sockaddr_dl *
2088 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2089 {
2090 	struct sockaddr_dl *sdl;
2091 
2092 	sdl = (struct sockaddr_dl *)paddr;
2093 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2094 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2095 	sdl->sdl_family = AF_LINK;
2096 	sdl->sdl_index = ifp->if_index;
2097 	sdl->sdl_type = iftype;
2098 
2099 	return (sdl);
2100 }
2101 
2102 /*
2103  * Mark an interface down and notify protocols of
2104  * the transition.
2105  */
2106 static void
2107 if_unroute(struct ifnet *ifp, int flag, int fam)
2108 {
2109 	struct ifaddr *ifa;
2110 	struct epoch_tracker et;
2111 
2112 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2113 
2114 	ifp->if_flags &= ~flag;
2115 	getmicrotime(&ifp->if_lastchange);
2116 	NET_EPOCH_ENTER(et);
2117 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2118 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2119 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2120 	NET_EPOCH_EXIT(et);
2121 	ifp->if_qflush(ifp);
2122 
2123 	if (ifp->if_carp)
2124 		(*carp_linkstate_p)(ifp);
2125 	rt_ifmsg(ifp);
2126 }
2127 
2128 /*
2129  * Mark an interface up and notify protocols of
2130  * the transition.
2131  */
2132 static void
2133 if_route(struct ifnet *ifp, int flag, int fam)
2134 {
2135 	struct ifaddr *ifa;
2136 	struct epoch_tracker et;
2137 
2138 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2139 
2140 	ifp->if_flags |= flag;
2141 	getmicrotime(&ifp->if_lastchange);
2142 	NET_EPOCH_ENTER(et);
2143 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2144 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2145 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2146 	NET_EPOCH_EXIT(et);
2147 	if (ifp->if_carp)
2148 		(*carp_linkstate_p)(ifp);
2149 	rt_ifmsg(ifp);
2150 #ifdef INET6
2151 	in6_if_up(ifp);
2152 #endif
2153 }
2154 
2155 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2156 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2157 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2158 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2159 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2160 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2161 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2162 void	*(*vlan_cookie_p)(struct ifnet *);
2163 
2164 /*
2165  * Handle a change in the interface link state. To avoid LORs
2166  * between driver lock and upper layer locks, as well as possible
2167  * recursions, we post event to taskqueue, and all job
2168  * is done in static do_link_state_change().
2169  */
2170 void
2171 if_link_state_change(struct ifnet *ifp, int link_state)
2172 {
2173 	/* Return if state hasn't changed. */
2174 	if (ifp->if_link_state == link_state)
2175 		return;
2176 
2177 	ifp->if_link_state = link_state;
2178 
2179 	/* XXXGL: reference ifp? */
2180 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2181 }
2182 
2183 static void
2184 do_link_state_change(void *arg, int pending)
2185 {
2186 	struct ifnet *ifp;
2187 	int link_state;
2188 
2189 	ifp = arg;
2190 	link_state = ifp->if_link_state;
2191 
2192 	CURVNET_SET(ifp->if_vnet);
2193 	rt_ifmsg(ifp);
2194 	if (ifp->if_vlantrunk != NULL)
2195 		(*vlan_link_state_p)(ifp);
2196 
2197 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2198 	    ifp->if_l2com != NULL)
2199 		(*ng_ether_link_state_p)(ifp, link_state);
2200 	if (ifp->if_carp)
2201 		(*carp_linkstate_p)(ifp);
2202 	if (ifp->if_bridge)
2203 		ifp->if_bridge_linkstate(ifp);
2204 	if (ifp->if_lagg)
2205 		(*lagg_linkstate_p)(ifp, link_state);
2206 
2207 	if (IS_DEFAULT_VNET(curvnet))
2208 		devctl_notify("IFNET", ifp->if_xname,
2209 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2210 		    NULL);
2211 	if (pending > 1)
2212 		if_printf(ifp, "%d link states coalesced\n", pending);
2213 	if (log_link_state_change)
2214 		if_printf(ifp, "link state changed to %s\n",
2215 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2216 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2217 	CURVNET_RESTORE();
2218 }
2219 
2220 /*
2221  * Mark an interface down and notify protocols of
2222  * the transition.
2223  */
2224 void
2225 if_down(struct ifnet *ifp)
2226 {
2227 
2228 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2229 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2230 }
2231 
2232 /*
2233  * Mark an interface up and notify protocols of
2234  * the transition.
2235  */
2236 void
2237 if_up(struct ifnet *ifp)
2238 {
2239 
2240 	if_route(ifp, IFF_UP, AF_UNSPEC);
2241 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2242 }
2243 
2244 /*
2245  * Flush an interface queue.
2246  */
2247 void
2248 if_qflush(struct ifnet *ifp)
2249 {
2250 	struct mbuf *m, *n;
2251 	struct ifaltq *ifq;
2252 
2253 	ifq = &ifp->if_snd;
2254 	IFQ_LOCK(ifq);
2255 #ifdef ALTQ
2256 	if (ALTQ_IS_ENABLED(ifq))
2257 		ALTQ_PURGE(ifq);
2258 #endif
2259 	n = ifq->ifq_head;
2260 	while ((m = n) != NULL) {
2261 		n = m->m_nextpkt;
2262 		m_freem(m);
2263 	}
2264 	ifq->ifq_head = 0;
2265 	ifq->ifq_tail = 0;
2266 	ifq->ifq_len = 0;
2267 	IFQ_UNLOCK(ifq);
2268 }
2269 
2270 /*
2271  * Map interface name to interface structure pointer, with or without
2272  * returning a reference.
2273  */
2274 struct ifnet *
2275 ifunit_ref(const char *name)
2276 {
2277 	struct epoch_tracker et;
2278 	struct ifnet *ifp;
2279 
2280 	NET_EPOCH_ENTER(et);
2281 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2282 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2283 		    !(ifp->if_flags & IFF_DYING))
2284 			break;
2285 	}
2286 	if (ifp != NULL)
2287 		if_ref(ifp);
2288 	NET_EPOCH_EXIT(et);
2289 	return (ifp);
2290 }
2291 
2292 struct ifnet *
2293 ifunit(const char *name)
2294 {
2295 	struct epoch_tracker et;
2296 	struct ifnet *ifp;
2297 
2298 	NET_EPOCH_ENTER(et);
2299 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2300 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2301 			break;
2302 	}
2303 	NET_EPOCH_EXIT(et);
2304 	return (ifp);
2305 }
2306 
2307 void *
2308 ifr_buffer_get_buffer(void *data)
2309 {
2310 	union ifreq_union *ifrup;
2311 
2312 	ifrup = data;
2313 #ifdef COMPAT_FREEBSD32
2314 	if (SV_CURPROC_FLAG(SV_ILP32))
2315 		return ((void *)(uintptr_t)
2316 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2317 #endif
2318 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2319 }
2320 
2321 static void
2322 ifr_buffer_set_buffer_null(void *data)
2323 {
2324 	union ifreq_union *ifrup;
2325 
2326 	ifrup = data;
2327 #ifdef COMPAT_FREEBSD32
2328 	if (SV_CURPROC_FLAG(SV_ILP32))
2329 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2330 	else
2331 #endif
2332 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2333 }
2334 
2335 size_t
2336 ifr_buffer_get_length(void *data)
2337 {
2338 	union ifreq_union *ifrup;
2339 
2340 	ifrup = data;
2341 #ifdef COMPAT_FREEBSD32
2342 	if (SV_CURPROC_FLAG(SV_ILP32))
2343 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2344 #endif
2345 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2346 }
2347 
2348 static void
2349 ifr_buffer_set_length(void *data, size_t len)
2350 {
2351 	union ifreq_union *ifrup;
2352 
2353 	ifrup = data;
2354 #ifdef COMPAT_FREEBSD32
2355 	if (SV_CURPROC_FLAG(SV_ILP32))
2356 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2357 	else
2358 #endif
2359 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2360 }
2361 
2362 void *
2363 ifr_data_get_ptr(void *ifrp)
2364 {
2365 	union ifreq_union *ifrup;
2366 
2367 	ifrup = ifrp;
2368 #ifdef COMPAT_FREEBSD32
2369 	if (SV_CURPROC_FLAG(SV_ILP32))
2370 		return ((void *)(uintptr_t)
2371 		    ifrup->ifr32.ifr_ifru.ifru_data);
2372 #endif
2373 		return (ifrup->ifr.ifr_ifru.ifru_data);
2374 }
2375 
2376 /*
2377  * Hardware specific interface ioctls.
2378  */
2379 int
2380 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2381 {
2382 	struct ifreq *ifr;
2383 	int error = 0, do_ifup = 0;
2384 	int new_flags, temp_flags;
2385 	size_t namelen, onamelen;
2386 	size_t descrlen;
2387 	char *descrbuf, *odescrbuf;
2388 	char new_name[IFNAMSIZ];
2389 	char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8];
2390 	struct ifaddr *ifa;
2391 	struct sockaddr_dl *sdl;
2392 
2393 	ifr = (struct ifreq *)data;
2394 	switch (cmd) {
2395 	case SIOCGIFINDEX:
2396 		ifr->ifr_index = ifp->if_index;
2397 		break;
2398 
2399 	case SIOCGIFFLAGS:
2400 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2401 		ifr->ifr_flags = temp_flags & 0xffff;
2402 		ifr->ifr_flagshigh = temp_flags >> 16;
2403 		break;
2404 
2405 	case SIOCGIFCAP:
2406 		ifr->ifr_reqcap = ifp->if_capabilities;
2407 		ifr->ifr_curcap = ifp->if_capenable;
2408 		break;
2409 
2410 	case SIOCGIFDATA:
2411 	{
2412 		struct if_data ifd;
2413 
2414 		/* Ensure uninitialised padding is not leaked. */
2415 		memset(&ifd, 0, sizeof(ifd));
2416 
2417 		if_data_copy(ifp, &ifd);
2418 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2419 		break;
2420 	}
2421 
2422 #ifdef MAC
2423 	case SIOCGIFMAC:
2424 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2425 		break;
2426 #endif
2427 
2428 	case SIOCGIFMETRIC:
2429 		ifr->ifr_metric = ifp->if_metric;
2430 		break;
2431 
2432 	case SIOCGIFMTU:
2433 		ifr->ifr_mtu = ifp->if_mtu;
2434 		break;
2435 
2436 	case SIOCGIFPHYS:
2437 		/* XXXGL: did this ever worked? */
2438 		ifr->ifr_phys = 0;
2439 		break;
2440 
2441 	case SIOCGIFDESCR:
2442 		error = 0;
2443 		sx_slock(&ifdescr_sx);
2444 		if (ifp->if_description == NULL)
2445 			error = ENOMSG;
2446 		else {
2447 			/* space for terminating nul */
2448 			descrlen = strlen(ifp->if_description) + 1;
2449 			if (ifr_buffer_get_length(ifr) < descrlen)
2450 				ifr_buffer_set_buffer_null(ifr);
2451 			else
2452 				error = copyout(ifp->if_description,
2453 				    ifr_buffer_get_buffer(ifr), descrlen);
2454 			ifr_buffer_set_length(ifr, descrlen);
2455 		}
2456 		sx_sunlock(&ifdescr_sx);
2457 		break;
2458 
2459 	case SIOCSIFDESCR:
2460 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2461 		if (error)
2462 			return (error);
2463 
2464 		/*
2465 		 * Copy only (length-1) bytes to make sure that
2466 		 * if_description is always nul terminated.  The
2467 		 * length parameter is supposed to count the
2468 		 * terminating nul in.
2469 		 */
2470 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2471 			return (ENAMETOOLONG);
2472 		else if (ifr_buffer_get_length(ifr) == 0)
2473 			descrbuf = NULL;
2474 		else {
2475 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2476 			    M_IFDESCR, M_WAITOK | M_ZERO);
2477 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2478 			    ifr_buffer_get_length(ifr) - 1);
2479 			if (error) {
2480 				free(descrbuf, M_IFDESCR);
2481 				break;
2482 			}
2483 		}
2484 
2485 		sx_xlock(&ifdescr_sx);
2486 		odescrbuf = ifp->if_description;
2487 		ifp->if_description = descrbuf;
2488 		sx_xunlock(&ifdescr_sx);
2489 
2490 		getmicrotime(&ifp->if_lastchange);
2491 		free(odescrbuf, M_IFDESCR);
2492 		break;
2493 
2494 	case SIOCGIFFIB:
2495 		ifr->ifr_fib = ifp->if_fib;
2496 		break;
2497 
2498 	case SIOCSIFFIB:
2499 		error = priv_check(td, PRIV_NET_SETIFFIB);
2500 		if (error)
2501 			return (error);
2502 		if (ifr->ifr_fib >= rt_numfibs)
2503 			return (EINVAL);
2504 
2505 		ifp->if_fib = ifr->ifr_fib;
2506 		break;
2507 
2508 	case SIOCSIFFLAGS:
2509 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2510 		if (error)
2511 			return (error);
2512 		/*
2513 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2514 		 * check, so we don't need special handling here yet.
2515 		 */
2516 		new_flags = (ifr->ifr_flags & 0xffff) |
2517 		    (ifr->ifr_flagshigh << 16);
2518 		if (ifp->if_flags & IFF_UP &&
2519 		    (new_flags & IFF_UP) == 0) {
2520 			if_down(ifp);
2521 		} else if (new_flags & IFF_UP &&
2522 		    (ifp->if_flags & IFF_UP) == 0) {
2523 			do_ifup = 1;
2524 		}
2525 		/* See if permanently promiscuous mode bit is about to flip */
2526 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2527 			if (new_flags & IFF_PPROMISC)
2528 				ifp->if_flags |= IFF_PROMISC;
2529 			else if (ifp->if_pcount == 0)
2530 				ifp->if_flags &= ~IFF_PROMISC;
2531 			if (log_promisc_mode_change)
2532                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2533                                     ((new_flags & IFF_PPROMISC) ?
2534                                      "enabled" : "disabled"));
2535 		}
2536 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2537 			(new_flags &~ IFF_CANTCHANGE);
2538 		if (ifp->if_ioctl) {
2539 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2540 		}
2541 		if (do_ifup)
2542 			if_up(ifp);
2543 		getmicrotime(&ifp->if_lastchange);
2544 		break;
2545 
2546 	case SIOCSIFCAP:
2547 		error = priv_check(td, PRIV_NET_SETIFCAP);
2548 		if (error)
2549 			return (error);
2550 		if (ifp->if_ioctl == NULL)
2551 			return (EOPNOTSUPP);
2552 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2553 			return (EINVAL);
2554 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2555 		if (error == 0)
2556 			getmicrotime(&ifp->if_lastchange);
2557 		break;
2558 
2559 #ifdef MAC
2560 	case SIOCSIFMAC:
2561 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2562 		break;
2563 #endif
2564 
2565 	case SIOCSIFNAME:
2566 		error = priv_check(td, PRIV_NET_SETIFNAME);
2567 		if (error)
2568 			return (error);
2569 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2570 		    NULL);
2571 		if (error != 0)
2572 			return (error);
2573 		if (new_name[0] == '\0')
2574 			return (EINVAL);
2575 		if (strcmp(new_name, ifp->if_xname) == 0)
2576 			break;
2577 		if (ifunit(new_name) != NULL)
2578 			return (EEXIST);
2579 
2580 		/*
2581 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2582 		 * and there are numerous other races with the
2583 		 * ifunit() checks not being atomic with namespace
2584 		 * changes (renames, vmoves, if_attach, etc).
2585 		 */
2586 		ifp->if_flags |= IFF_RENAMING;
2587 
2588 		/* Announce the departure of the interface. */
2589 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2590 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2591 
2592 		if_printf(ifp, "changing name to '%s'\n", new_name);
2593 
2594 		IF_ADDR_WLOCK(ifp);
2595 		strlcpy(old_name, ifp->if_xname, sizeof(old_name));
2596 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2597 		ifa = ifp->if_addr;
2598 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2599 		namelen = strlen(new_name);
2600 		onamelen = sdl->sdl_nlen;
2601 		/*
2602 		 * Move the address if needed.  This is safe because we
2603 		 * allocate space for a name of length IFNAMSIZ when we
2604 		 * create this in if_attach().
2605 		 */
2606 		if (namelen != onamelen) {
2607 			bcopy(sdl->sdl_data + onamelen,
2608 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2609 		}
2610 		bcopy(new_name, sdl->sdl_data, namelen);
2611 		sdl->sdl_nlen = namelen;
2612 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2613 		bzero(sdl->sdl_data, onamelen);
2614 		while (namelen != 0)
2615 			sdl->sdl_data[--namelen] = 0xff;
2616 		IF_ADDR_WUNLOCK(ifp);
2617 
2618 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2619 		/* Announce the return of the interface. */
2620 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2621 
2622 		ifp->if_flags &= ~IFF_RENAMING;
2623 
2624 		snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
2625 		devctl_notify("IFNET", old_name, "RENAME", strbuf);
2626 		break;
2627 
2628 #ifdef VIMAGE
2629 	case SIOCSIFVNET:
2630 		error = priv_check(td, PRIV_NET_SETIFVNET);
2631 		if (error)
2632 			return (error);
2633 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2634 		break;
2635 #endif
2636 
2637 	case SIOCSIFMETRIC:
2638 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2639 		if (error)
2640 			return (error);
2641 		ifp->if_metric = ifr->ifr_metric;
2642 		getmicrotime(&ifp->if_lastchange);
2643 		break;
2644 
2645 	case SIOCSIFPHYS:
2646 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2647 		if (error)
2648 			return (error);
2649 		if (ifp->if_ioctl == NULL)
2650 			return (EOPNOTSUPP);
2651 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2652 		if (error == 0)
2653 			getmicrotime(&ifp->if_lastchange);
2654 		break;
2655 
2656 	case SIOCSIFMTU:
2657 	{
2658 		u_long oldmtu = ifp->if_mtu;
2659 
2660 		error = priv_check(td, PRIV_NET_SETIFMTU);
2661 		if (error)
2662 			return (error);
2663 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2664 			return (EINVAL);
2665 		if (ifp->if_ioctl == NULL)
2666 			return (EOPNOTSUPP);
2667 		/* Disallow MTU changes on bridge member interfaces. */
2668 		if (ifp->if_bridge)
2669 			return (EOPNOTSUPP);
2670 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2671 		if (error == 0) {
2672 			getmicrotime(&ifp->if_lastchange);
2673 			rt_ifmsg(ifp);
2674 #ifdef INET
2675 			DEBUGNET_NOTIFY_MTU(ifp);
2676 #endif
2677 		}
2678 		/*
2679 		 * If the link MTU changed, do network layer specific procedure.
2680 		 */
2681 		if (ifp->if_mtu != oldmtu) {
2682 #ifdef INET6
2683 			nd6_setmtu(ifp);
2684 #endif
2685 			rt_updatemtu(ifp);
2686 		}
2687 		break;
2688 	}
2689 
2690 	case SIOCADDMULTI:
2691 	case SIOCDELMULTI:
2692 		if (cmd == SIOCADDMULTI)
2693 			error = priv_check(td, PRIV_NET_ADDMULTI);
2694 		else
2695 			error = priv_check(td, PRIV_NET_DELMULTI);
2696 		if (error)
2697 			return (error);
2698 
2699 		/* Don't allow group membership on non-multicast interfaces. */
2700 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2701 			return (EOPNOTSUPP);
2702 
2703 		/* Don't let users screw up protocols' entries. */
2704 		if (ifr->ifr_addr.sa_family != AF_LINK)
2705 			return (EINVAL);
2706 
2707 		if (cmd == SIOCADDMULTI) {
2708 			struct epoch_tracker et;
2709 			struct ifmultiaddr *ifma;
2710 
2711 			/*
2712 			 * Userland is only permitted to join groups once
2713 			 * via the if_addmulti() KPI, because it cannot hold
2714 			 * struct ifmultiaddr * between calls. It may also
2715 			 * lose a race while we check if the membership
2716 			 * already exists.
2717 			 */
2718 			NET_EPOCH_ENTER(et);
2719 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2720 			NET_EPOCH_EXIT(et);
2721 			if (ifma != NULL)
2722 				error = EADDRINUSE;
2723 			else
2724 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2725 		} else {
2726 			error = if_delmulti(ifp, &ifr->ifr_addr);
2727 		}
2728 		if (error == 0)
2729 			getmicrotime(&ifp->if_lastchange);
2730 		break;
2731 
2732 	case SIOCSIFPHYADDR:
2733 	case SIOCDIFPHYADDR:
2734 #ifdef INET6
2735 	case SIOCSIFPHYADDR_IN6:
2736 #endif
2737 	case SIOCSIFMEDIA:
2738 	case SIOCSIFGENERIC:
2739 		error = priv_check(td, PRIV_NET_HWIOCTL);
2740 		if (error)
2741 			return (error);
2742 		if (ifp->if_ioctl == NULL)
2743 			return (EOPNOTSUPP);
2744 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2745 		if (error == 0)
2746 			getmicrotime(&ifp->if_lastchange);
2747 		break;
2748 
2749 	case SIOCGIFSTATUS:
2750 	case SIOCGIFPSRCADDR:
2751 	case SIOCGIFPDSTADDR:
2752 	case SIOCGIFMEDIA:
2753 	case SIOCGIFXMEDIA:
2754 	case SIOCGIFGENERIC:
2755 	case SIOCGIFRSSKEY:
2756 	case SIOCGIFRSSHASH:
2757 	case SIOCGIFDOWNREASON:
2758 		if (ifp->if_ioctl == NULL)
2759 			return (EOPNOTSUPP);
2760 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2761 		break;
2762 
2763 	case SIOCSIFLLADDR:
2764 		error = priv_check(td, PRIV_NET_SETLLADDR);
2765 		if (error)
2766 			return (error);
2767 		error = if_setlladdr(ifp,
2768 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2769 		break;
2770 
2771 	case SIOCGHWADDR:
2772 		error = if_gethwaddr(ifp, ifr);
2773 		break;
2774 
2775 	case SIOCAIFGROUP:
2776 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2777 		if (error)
2778 			return (error);
2779 		error = if_addgroup(ifp,
2780 		    ((struct ifgroupreq *)data)->ifgr_group);
2781 		if (error != 0)
2782 			return (error);
2783 		break;
2784 
2785 	case SIOCGIFGROUP:
2786 	{
2787 		struct epoch_tracker et;
2788 
2789 		NET_EPOCH_ENTER(et);
2790 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2791 		NET_EPOCH_EXIT(et);
2792 		break;
2793 	}
2794 
2795 	case SIOCDIFGROUP:
2796 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2797 		if (error)
2798 			return (error);
2799 		error = if_delgroup(ifp,
2800 		    ((struct ifgroupreq *)data)->ifgr_group);
2801 		if (error != 0)
2802 			return (error);
2803 		break;
2804 
2805 	default:
2806 		error = ENOIOCTL;
2807 		break;
2808 	}
2809 	return (error);
2810 }
2811 
2812 /*
2813  * Interface ioctls.
2814  */
2815 int
2816 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2817 {
2818 #ifdef COMPAT_FREEBSD32
2819 	union {
2820 		struct ifconf ifc;
2821 		struct ifdrv ifd;
2822 		struct ifgroupreq ifgr;
2823 		struct ifmediareq ifmr;
2824 	} thunk;
2825 	u_long saved_cmd;
2826 	struct ifconf32 *ifc32;
2827 	struct ifdrv32 *ifd32;
2828 	struct ifgroupreq32 *ifgr32;
2829 	struct ifmediareq32 *ifmr32;
2830 #endif
2831 	struct ifnet *ifp;
2832 	struct ifreq *ifr;
2833 	int error;
2834 	int oif_flags;
2835 #ifdef VIMAGE
2836 	bool shutdown;
2837 #endif
2838 
2839 	CURVNET_SET(so->so_vnet);
2840 #ifdef VIMAGE
2841 	/* Make sure the VNET is stable. */
2842 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2843 	if (shutdown) {
2844 		CURVNET_RESTORE();
2845 		return (EBUSY);
2846 	}
2847 #endif
2848 
2849 #ifdef COMPAT_FREEBSD32
2850 	saved_cmd = cmd;
2851 	switch (cmd) {
2852 	case SIOCGIFCONF32:
2853 		ifc32 = (struct ifconf32 *)data;
2854 		thunk.ifc.ifc_len = ifc32->ifc_len;
2855 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2856 		data = (caddr_t)&thunk.ifc;
2857 		cmd = SIOCGIFCONF;
2858 		break;
2859 	case SIOCGDRVSPEC32:
2860 	case SIOCSDRVSPEC32:
2861 		ifd32 = (struct ifdrv32 *)data;
2862 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2863 		    sizeof(thunk.ifd.ifd_name));
2864 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2865 		thunk.ifd.ifd_len = ifd32->ifd_len;
2866 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2867 		data = (caddr_t)&thunk.ifd;
2868 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2869 		break;
2870 	case SIOCAIFGROUP32:
2871 	case SIOCGIFGROUP32:
2872 	case SIOCDIFGROUP32:
2873 	case SIOCGIFGMEMB32:
2874 		ifgr32 = (struct ifgroupreq32 *)data;
2875 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2876 		    sizeof(thunk.ifgr.ifgr_name));
2877 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2878 		switch (cmd) {
2879 		case SIOCAIFGROUP32:
2880 		case SIOCDIFGROUP32:
2881 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2882 			    sizeof(thunk.ifgr.ifgr_group));
2883 			break;
2884 		case SIOCGIFGROUP32:
2885 		case SIOCGIFGMEMB32:
2886 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2887 			break;
2888 		}
2889 		data = (caddr_t)&thunk.ifgr;
2890 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2891 		break;
2892 	case SIOCGIFMEDIA32:
2893 	case SIOCGIFXMEDIA32:
2894 		ifmr32 = (struct ifmediareq32 *)data;
2895 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2896 		    sizeof(thunk.ifmr.ifm_name));
2897 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2898 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2899 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2900 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2901 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2902 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2903 		data = (caddr_t)&thunk.ifmr;
2904 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2905 		break;
2906 	}
2907 #endif
2908 
2909 	switch (cmd) {
2910 	case SIOCGIFCONF:
2911 		error = ifconf(cmd, data);
2912 		goto out_noref;
2913 	}
2914 
2915 	ifr = (struct ifreq *)data;
2916 	switch (cmd) {
2917 #ifdef VIMAGE
2918 	case SIOCSIFRVNET:
2919 		error = priv_check(td, PRIV_NET_SETIFVNET);
2920 		if (error == 0)
2921 			error = if_vmove_reclaim(td, ifr->ifr_name,
2922 			    ifr->ifr_jid);
2923 		goto out_noref;
2924 #endif
2925 	case SIOCIFCREATE:
2926 	case SIOCIFCREATE2:
2927 		error = priv_check(td, PRIV_NET_IFCREATE);
2928 		if (error == 0)
2929 			error = if_clone_create(ifr->ifr_name,
2930 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
2931 			    ifr_data_get_ptr(ifr) : NULL);
2932 		goto out_noref;
2933 	case SIOCIFDESTROY:
2934 		error = priv_check(td, PRIV_NET_IFDESTROY);
2935 
2936 		if (error == 0) {
2937 			sx_xlock(&ifnet_detach_sxlock);
2938 			error = if_clone_destroy(ifr->ifr_name);
2939 			sx_xunlock(&ifnet_detach_sxlock);
2940 		}
2941 		goto out_noref;
2942 
2943 	case SIOCIFGCLONERS:
2944 		error = if_clone_list((struct if_clonereq *)data);
2945 		goto out_noref;
2946 
2947 	case SIOCGIFGMEMB:
2948 		error = if_getgroupmembers((struct ifgroupreq *)data);
2949 		goto out_noref;
2950 
2951 #if defined(INET) || defined(INET6)
2952 	case SIOCSVH:
2953 	case SIOCGVH:
2954 		if (carp_ioctl_p == NULL)
2955 			error = EPROTONOSUPPORT;
2956 		else
2957 			error = (*carp_ioctl_p)(ifr, cmd, td);
2958 		goto out_noref;
2959 #endif
2960 	}
2961 
2962 	ifp = ifunit_ref(ifr->ifr_name);
2963 	if (ifp == NULL) {
2964 		error = ENXIO;
2965 		goto out_noref;
2966 	}
2967 
2968 	error = ifhwioctl(cmd, ifp, data, td);
2969 	if (error != ENOIOCTL)
2970 		goto out_ref;
2971 
2972 	oif_flags = ifp->if_flags;
2973 	if (so->so_proto == NULL) {
2974 		error = EOPNOTSUPP;
2975 		goto out_ref;
2976 	}
2977 
2978 	/*
2979 	 * Pass the request on to the socket control method, and if the
2980 	 * latter returns EOPNOTSUPP, directly to the interface.
2981 	 *
2982 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2983 	 * trust SIOCSIFADDR et al to come from an already privileged
2984 	 * layer, and do not perform any credentials checks or input
2985 	 * validation.
2986 	 */
2987 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2988 	    ifp, td));
2989 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2990 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2991 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2992 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2993 
2994 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2995 #ifdef INET6
2996 		if (ifp->if_flags & IFF_UP)
2997 			in6_if_up(ifp);
2998 #endif
2999 	}
3000 
3001 out_ref:
3002 	if_rele(ifp);
3003 out_noref:
3004 	CURVNET_RESTORE();
3005 #ifdef COMPAT_FREEBSD32
3006 	if (error != 0)
3007 		return (error);
3008 	switch (saved_cmd) {
3009 	case SIOCGIFCONF32:
3010 		ifc32->ifc_len = thunk.ifc.ifc_len;
3011 		break;
3012 	case SIOCGDRVSPEC32:
3013 		/*
3014 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
3015 		 * the struct so just assert that ifd_len (the only
3016 		 * field it might make sense to update) hasn't
3017 		 * changed.
3018 		 */
3019 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
3020 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
3021 			thunk.ifd.ifd_len));
3022 		break;
3023 	case SIOCGIFGROUP32:
3024 	case SIOCGIFGMEMB32:
3025 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3026 		break;
3027 	case SIOCGIFMEDIA32:
3028 	case SIOCGIFXMEDIA32:
3029 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3030 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3031 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3032 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3033 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3034 		break;
3035 	}
3036 #endif
3037 	return (error);
3038 }
3039 
3040 /*
3041  * The code common to handling reference counted flags,
3042  * e.g., in ifpromisc() and if_allmulti().
3043  * The "pflag" argument can specify a permanent mode flag to check,
3044  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3045  *
3046  * Only to be used on stack-owned flags, not driver-owned flags.
3047  */
3048 static int
3049 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3050 {
3051 	struct ifreq ifr;
3052 	int error;
3053 	int oldflags, oldcount;
3054 
3055 	/* Sanity checks to catch programming errors */
3056 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3057 	    ("%s: setting driver-owned flag %d", __func__, flag));
3058 
3059 	if (onswitch)
3060 		KASSERT(*refcount >= 0,
3061 		    ("%s: increment negative refcount %d for flag %d",
3062 		    __func__, *refcount, flag));
3063 	else
3064 		KASSERT(*refcount > 0,
3065 		    ("%s: decrement non-positive refcount %d for flag %d",
3066 		    __func__, *refcount, flag));
3067 
3068 	/* In case this mode is permanent, just touch refcount */
3069 	if (ifp->if_flags & pflag) {
3070 		*refcount += onswitch ? 1 : -1;
3071 		return (0);
3072 	}
3073 
3074 	/* Save ifnet parameters for if_ioctl() may fail */
3075 	oldcount = *refcount;
3076 	oldflags = ifp->if_flags;
3077 
3078 	/*
3079 	 * See if we aren't the only and touching refcount is enough.
3080 	 * Actually toggle interface flag if we are the first or last.
3081 	 */
3082 	if (onswitch) {
3083 		if ((*refcount)++)
3084 			return (0);
3085 		ifp->if_flags |= flag;
3086 	} else {
3087 		if (--(*refcount))
3088 			return (0);
3089 		ifp->if_flags &= ~flag;
3090 	}
3091 
3092 	/* Call down the driver since we've changed interface flags */
3093 	if (ifp->if_ioctl == NULL) {
3094 		error = EOPNOTSUPP;
3095 		goto recover;
3096 	}
3097 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3098 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3099 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3100 	if (error)
3101 		goto recover;
3102 	/* Notify userland that interface flags have changed */
3103 	rt_ifmsg(ifp);
3104 	return (0);
3105 
3106 recover:
3107 	/* Recover after driver error */
3108 	*refcount = oldcount;
3109 	ifp->if_flags = oldflags;
3110 	return (error);
3111 }
3112 
3113 /*
3114  * Set/clear promiscuous mode on interface ifp based on the truth value
3115  * of pswitch.  The calls are reference counted so that only the first
3116  * "on" request actually has an effect, as does the final "off" request.
3117  * Results are undefined if the "off" and "on" requests are not matched.
3118  */
3119 int
3120 ifpromisc(struct ifnet *ifp, int pswitch)
3121 {
3122 	int error;
3123 	int oldflags = ifp->if_flags;
3124 
3125 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3126 			   &ifp->if_pcount, pswitch);
3127 	/* If promiscuous mode status has changed, log a message */
3128 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3129             log_promisc_mode_change)
3130 		if_printf(ifp, "promiscuous mode %s\n",
3131 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3132 	return (error);
3133 }
3134 
3135 /*
3136  * Return interface configuration
3137  * of system.  List may be used
3138  * in later ioctl's (above) to get
3139  * other information.
3140  */
3141 /*ARGSUSED*/
3142 static int
3143 ifconf(u_long cmd, caddr_t data)
3144 {
3145 	struct ifconf *ifc = (struct ifconf *)data;
3146 	struct ifnet *ifp;
3147 	struct ifaddr *ifa;
3148 	struct ifreq ifr;
3149 	struct sbuf *sb;
3150 	int error, full = 0, valid_len, max_len;
3151 
3152 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3153 	max_len = maxphys - 1;
3154 
3155 	/* Prevent hostile input from being able to crash the system */
3156 	if (ifc->ifc_len <= 0)
3157 		return (EINVAL);
3158 
3159 again:
3160 	if (ifc->ifc_len <= max_len) {
3161 		max_len = ifc->ifc_len;
3162 		full = 1;
3163 	}
3164 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3165 	max_len = 0;
3166 	valid_len = 0;
3167 
3168 	IFNET_RLOCK();
3169 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3170 		struct epoch_tracker et;
3171 		int addrs;
3172 
3173 		/*
3174 		 * Zero the ifr to make sure we don't disclose the contents
3175 		 * of the stack.
3176 		 */
3177 		memset(&ifr, 0, sizeof(ifr));
3178 
3179 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3180 		    >= sizeof(ifr.ifr_name)) {
3181 			sbuf_delete(sb);
3182 			IFNET_RUNLOCK();
3183 			return (ENAMETOOLONG);
3184 		}
3185 
3186 		addrs = 0;
3187 		NET_EPOCH_ENTER(et);
3188 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3189 			struct sockaddr *sa = ifa->ifa_addr;
3190 
3191 			if (prison_if(curthread->td_ucred, sa) != 0)
3192 				continue;
3193 			addrs++;
3194 			if (sa->sa_len <= sizeof(*sa)) {
3195 				if (sa->sa_len < sizeof(*sa)) {
3196 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3197 					    sizeof(ifr.ifr_ifru.ifru_addr));
3198 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3199 					    sa->sa_len);
3200 				} else
3201 					ifr.ifr_ifru.ifru_addr = *sa;
3202 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3203 				max_len += sizeof(ifr);
3204 			} else {
3205 				sbuf_bcat(sb, &ifr,
3206 				    offsetof(struct ifreq, ifr_addr));
3207 				max_len += offsetof(struct ifreq, ifr_addr);
3208 				sbuf_bcat(sb, sa, sa->sa_len);
3209 				max_len += sa->sa_len;
3210 			}
3211 
3212 			if (sbuf_error(sb) == 0)
3213 				valid_len = sbuf_len(sb);
3214 		}
3215 		NET_EPOCH_EXIT(et);
3216 		if (addrs == 0) {
3217 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3218 			max_len += sizeof(ifr);
3219 
3220 			if (sbuf_error(sb) == 0)
3221 				valid_len = sbuf_len(sb);
3222 		}
3223 	}
3224 	IFNET_RUNLOCK();
3225 
3226 	/*
3227 	 * If we didn't allocate enough space (uncommon), try again.  If
3228 	 * we have already allocated as much space as we are allowed,
3229 	 * return what we've got.
3230 	 */
3231 	if (valid_len != max_len && !full) {
3232 		sbuf_delete(sb);
3233 		goto again;
3234 	}
3235 
3236 	ifc->ifc_len = valid_len;
3237 	sbuf_finish(sb);
3238 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3239 	sbuf_delete(sb);
3240 	return (error);
3241 }
3242 
3243 /*
3244  * Just like ifpromisc(), but for all-multicast-reception mode.
3245  */
3246 int
3247 if_allmulti(struct ifnet *ifp, int onswitch)
3248 {
3249 
3250 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3251 }
3252 
3253 struct ifmultiaddr *
3254 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3255 {
3256 	struct ifmultiaddr *ifma;
3257 
3258 	IF_ADDR_LOCK_ASSERT(ifp);
3259 
3260 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3261 		if (sa->sa_family == AF_LINK) {
3262 			if (sa_dl_equal(ifma->ifma_addr, sa))
3263 				break;
3264 		} else {
3265 			if (sa_equal(ifma->ifma_addr, sa))
3266 				break;
3267 		}
3268 	}
3269 
3270 	return ifma;
3271 }
3272 
3273 /*
3274  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3275  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3276  * the ifnet multicast address list here, so the caller must do that and
3277  * other setup work (such as notifying the device driver).  The reference
3278  * count is initialized to 1.
3279  */
3280 static struct ifmultiaddr *
3281 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3282     int mflags)
3283 {
3284 	struct ifmultiaddr *ifma;
3285 	struct sockaddr *dupsa;
3286 
3287 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3288 	    M_ZERO);
3289 	if (ifma == NULL)
3290 		return (NULL);
3291 
3292 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3293 	if (dupsa == NULL) {
3294 		free(ifma, M_IFMADDR);
3295 		return (NULL);
3296 	}
3297 	bcopy(sa, dupsa, sa->sa_len);
3298 	ifma->ifma_addr = dupsa;
3299 
3300 	ifma->ifma_ifp = ifp;
3301 	ifma->ifma_refcount = 1;
3302 	ifma->ifma_protospec = NULL;
3303 
3304 	if (llsa == NULL) {
3305 		ifma->ifma_lladdr = NULL;
3306 		return (ifma);
3307 	}
3308 
3309 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3310 	if (dupsa == NULL) {
3311 		free(ifma->ifma_addr, M_IFMADDR);
3312 		free(ifma, M_IFMADDR);
3313 		return (NULL);
3314 	}
3315 	bcopy(llsa, dupsa, llsa->sa_len);
3316 	ifma->ifma_lladdr = dupsa;
3317 
3318 	return (ifma);
3319 }
3320 
3321 /*
3322  * if_freemulti: free ifmultiaddr structure and possibly attached related
3323  * addresses.  The caller is responsible for implementing reference
3324  * counting, notifying the driver, handling routing messages, and releasing
3325  * any dependent link layer state.
3326  */
3327 #ifdef MCAST_VERBOSE
3328 extern void kdb_backtrace(void);
3329 #endif
3330 static void
3331 if_freemulti_internal(struct ifmultiaddr *ifma)
3332 {
3333 
3334 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3335 	    ifma->ifma_refcount));
3336 
3337 	if (ifma->ifma_lladdr != NULL)
3338 		free(ifma->ifma_lladdr, M_IFMADDR);
3339 #ifdef MCAST_VERBOSE
3340 	kdb_backtrace();
3341 	printf("%s freeing ifma: %p\n", __func__, ifma);
3342 #endif
3343 	free(ifma->ifma_addr, M_IFMADDR);
3344 	free(ifma, M_IFMADDR);
3345 }
3346 
3347 static void
3348 if_destroymulti(epoch_context_t ctx)
3349 {
3350 	struct ifmultiaddr *ifma;
3351 
3352 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3353 	if_freemulti_internal(ifma);
3354 }
3355 
3356 void
3357 if_freemulti(struct ifmultiaddr *ifma)
3358 {
3359 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3360 	    ifma->ifma_refcount));
3361 
3362 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3363 }
3364 
3365 /*
3366  * Register an additional multicast address with a network interface.
3367  *
3368  * - If the address is already present, bump the reference count on the
3369  *   address and return.
3370  * - If the address is not link-layer, look up a link layer address.
3371  * - Allocate address structures for one or both addresses, and attach to the
3372  *   multicast address list on the interface.  If automatically adding a link
3373  *   layer address, the protocol address will own a reference to the link
3374  *   layer address, to be freed when it is freed.
3375  * - Notify the network device driver of an addition to the multicast address
3376  *   list.
3377  *
3378  * 'sa' points to caller-owned memory with the desired multicast address.
3379  *
3380  * 'retifma' will be used to return a pointer to the resulting multicast
3381  * address reference, if desired.
3382  */
3383 int
3384 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3385     struct ifmultiaddr **retifma)
3386 {
3387 	struct ifmultiaddr *ifma, *ll_ifma;
3388 	struct sockaddr *llsa;
3389 	struct sockaddr_dl sdl;
3390 	int error;
3391 
3392 #ifdef INET
3393 	IN_MULTI_LIST_UNLOCK_ASSERT();
3394 #endif
3395 #ifdef INET6
3396 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3397 #endif
3398 	/*
3399 	 * If the address is already present, return a new reference to it;
3400 	 * otherwise, allocate storage and set up a new address.
3401 	 */
3402 	IF_ADDR_WLOCK(ifp);
3403 	ifma = if_findmulti(ifp, sa);
3404 	if (ifma != NULL) {
3405 		ifma->ifma_refcount++;
3406 		if (retifma != NULL)
3407 			*retifma = ifma;
3408 		IF_ADDR_WUNLOCK(ifp);
3409 		return (0);
3410 	}
3411 
3412 	/*
3413 	 * The address isn't already present; resolve the protocol address
3414 	 * into a link layer address, and then look that up, bump its
3415 	 * refcount or allocate an ifma for that also.
3416 	 * Most link layer resolving functions returns address data which
3417 	 * fits inside default sockaddr_dl structure. However callback
3418 	 * can allocate another sockaddr structure, in that case we need to
3419 	 * free it later.
3420 	 */
3421 	llsa = NULL;
3422 	ll_ifma = NULL;
3423 	if (ifp->if_resolvemulti != NULL) {
3424 		/* Provide called function with buffer size information */
3425 		sdl.sdl_len = sizeof(sdl);
3426 		llsa = (struct sockaddr *)&sdl;
3427 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3428 		if (error)
3429 			goto unlock_out;
3430 	}
3431 
3432 	/*
3433 	 * Allocate the new address.  Don't hook it up yet, as we may also
3434 	 * need to allocate a link layer multicast address.
3435 	 */
3436 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3437 	if (ifma == NULL) {
3438 		error = ENOMEM;
3439 		goto free_llsa_out;
3440 	}
3441 
3442 	/*
3443 	 * If a link layer address is found, we'll need to see if it's
3444 	 * already present in the address list, or allocate is as well.
3445 	 * When this block finishes, the link layer address will be on the
3446 	 * list.
3447 	 */
3448 	if (llsa != NULL) {
3449 		ll_ifma = if_findmulti(ifp, llsa);
3450 		if (ll_ifma == NULL) {
3451 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3452 			if (ll_ifma == NULL) {
3453 				--ifma->ifma_refcount;
3454 				if_freemulti(ifma);
3455 				error = ENOMEM;
3456 				goto free_llsa_out;
3457 			}
3458 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3459 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3460 			    ifma_link);
3461 		} else
3462 			ll_ifma->ifma_refcount++;
3463 		ifma->ifma_llifma = ll_ifma;
3464 	}
3465 
3466 	/*
3467 	 * We now have a new multicast address, ifma, and possibly a new or
3468 	 * referenced link layer address.  Add the primary address to the
3469 	 * ifnet address list.
3470 	 */
3471 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3472 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3473 
3474 	if (retifma != NULL)
3475 		*retifma = ifma;
3476 
3477 	/*
3478 	 * Must generate the message while holding the lock so that 'ifma'
3479 	 * pointer is still valid.
3480 	 */
3481 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3482 	IF_ADDR_WUNLOCK(ifp);
3483 
3484 	/*
3485 	 * We are certain we have added something, so call down to the
3486 	 * interface to let them know about it.
3487 	 */
3488 	if (ifp->if_ioctl != NULL) {
3489 		if (THREAD_CAN_SLEEP())
3490 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3491 		else
3492 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3493 	}
3494 
3495 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3496 		link_free_sdl(llsa);
3497 
3498 	return (0);
3499 
3500 free_llsa_out:
3501 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3502 		link_free_sdl(llsa);
3503 
3504 unlock_out:
3505 	IF_ADDR_WUNLOCK(ifp);
3506 	return (error);
3507 }
3508 
3509 static void
3510 if_siocaddmulti(void *arg, int pending)
3511 {
3512 	struct ifnet *ifp;
3513 
3514 	ifp = arg;
3515 #ifdef DIAGNOSTIC
3516 	if (pending > 1)
3517 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3518 #endif
3519 	CURVNET_SET(ifp->if_vnet);
3520 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3521 	CURVNET_RESTORE();
3522 }
3523 
3524 /*
3525  * Delete a multicast group membership by network-layer group address.
3526  *
3527  * Returns ENOENT if the entry could not be found. If ifp no longer
3528  * exists, results are undefined. This entry point should only be used
3529  * from subsystems which do appropriate locking to hold ifp for the
3530  * duration of the call.
3531  * Network-layer protocol domains must use if_delmulti_ifma().
3532  */
3533 int
3534 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3535 {
3536 	struct ifmultiaddr *ifma;
3537 	int lastref;
3538 
3539 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3540 
3541 	IF_ADDR_WLOCK(ifp);
3542 	lastref = 0;
3543 	ifma = if_findmulti(ifp, sa);
3544 	if (ifma != NULL)
3545 		lastref = if_delmulti_locked(ifp, ifma, 0);
3546 	IF_ADDR_WUNLOCK(ifp);
3547 
3548 	if (ifma == NULL)
3549 		return (ENOENT);
3550 
3551 	if (lastref && ifp->if_ioctl != NULL) {
3552 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3553 	}
3554 
3555 	return (0);
3556 }
3557 
3558 /*
3559  * Delete all multicast group membership for an interface.
3560  * Should be used to quickly flush all multicast filters.
3561  */
3562 void
3563 if_delallmulti(struct ifnet *ifp)
3564 {
3565 	struct ifmultiaddr *ifma;
3566 	struct ifmultiaddr *next;
3567 
3568 	IF_ADDR_WLOCK(ifp);
3569 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3570 		if_delmulti_locked(ifp, ifma, 0);
3571 	IF_ADDR_WUNLOCK(ifp);
3572 }
3573 
3574 void
3575 if_delmulti_ifma(struct ifmultiaddr *ifma)
3576 {
3577 	if_delmulti_ifma_flags(ifma, 0);
3578 }
3579 
3580 /*
3581  * Delete a multicast group membership by group membership pointer.
3582  * Network-layer protocol domains must use this routine.
3583  *
3584  * It is safe to call this routine if the ifp disappeared.
3585  */
3586 void
3587 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3588 {
3589 	struct ifnet *ifp;
3590 	int lastref;
3591 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3592 #ifdef INET
3593 	IN_MULTI_LIST_UNLOCK_ASSERT();
3594 #endif
3595 	ifp = ifma->ifma_ifp;
3596 #ifdef DIAGNOSTIC
3597 	if (ifp == NULL) {
3598 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3599 	} else {
3600 		struct epoch_tracker et;
3601 		struct ifnet *oifp;
3602 
3603 		NET_EPOCH_ENTER(et);
3604 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3605 			if (ifp == oifp)
3606 				break;
3607 		NET_EPOCH_EXIT(et);
3608 		if (ifp != oifp)
3609 			ifp = NULL;
3610 	}
3611 #endif
3612 	/*
3613 	 * If and only if the ifnet instance exists: Acquire the address lock.
3614 	 */
3615 	if (ifp != NULL)
3616 		IF_ADDR_WLOCK(ifp);
3617 
3618 	lastref = if_delmulti_locked(ifp, ifma, flags);
3619 
3620 	if (ifp != NULL) {
3621 		/*
3622 		 * If and only if the ifnet instance exists:
3623 		 *  Release the address lock.
3624 		 *  If the group was left: update the hardware hash filter.
3625 		 */
3626 		IF_ADDR_WUNLOCK(ifp);
3627 		if (lastref && ifp->if_ioctl != NULL) {
3628 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3629 		}
3630 	}
3631 }
3632 
3633 /*
3634  * Perform deletion of network-layer and/or link-layer multicast address.
3635  *
3636  * Return 0 if the reference count was decremented.
3637  * Return 1 if the final reference was released, indicating that the
3638  * hardware hash filter should be reprogrammed.
3639  */
3640 static int
3641 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3642 {
3643 	struct ifmultiaddr *ll_ifma;
3644 
3645 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3646 		KASSERT(ifma->ifma_ifp == ifp,
3647 		    ("%s: inconsistent ifp %p", __func__, ifp));
3648 		IF_ADDR_WLOCK_ASSERT(ifp);
3649 	}
3650 
3651 	ifp = ifma->ifma_ifp;
3652 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3653 
3654 	/*
3655 	 * If the ifnet is detaching, null out references to ifnet,
3656 	 * so that upper protocol layers will notice, and not attempt
3657 	 * to obtain locks for an ifnet which no longer exists. The
3658 	 * routing socket announcement must happen before the ifnet
3659 	 * instance is detached from the system.
3660 	 */
3661 	if (detaching) {
3662 #ifdef DIAGNOSTIC
3663 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3664 #endif
3665 		/*
3666 		 * ifp may already be nulled out if we are being reentered
3667 		 * to delete the ll_ifma.
3668 		 */
3669 		if (ifp != NULL) {
3670 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3671 			ifma->ifma_ifp = NULL;
3672 		}
3673 	}
3674 
3675 	if (--ifma->ifma_refcount > 0)
3676 		return 0;
3677 
3678 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3679 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3680 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3681 	}
3682 	/*
3683 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3684 	 * have been associated with it. Release it first if so.
3685 	 */
3686 	ll_ifma = ifma->ifma_llifma;
3687 	if (ll_ifma != NULL) {
3688 		KASSERT(ifma->ifma_lladdr != NULL,
3689 		    ("%s: llifma w/o lladdr", __func__));
3690 		if (detaching)
3691 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3692 		if (--ll_ifma->ifma_refcount == 0) {
3693 			if (ifp != NULL) {
3694 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3695 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3696 						ifma_link);
3697 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3698 				}
3699 			}
3700 			if_freemulti(ll_ifma);
3701 		}
3702 	}
3703 #ifdef INVARIANTS
3704 	if (ifp) {
3705 		struct ifmultiaddr *ifmatmp;
3706 
3707 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3708 			MPASS(ifma != ifmatmp);
3709 	}
3710 #endif
3711 	if_freemulti(ifma);
3712 	/*
3713 	 * The last reference to this instance of struct ifmultiaddr
3714 	 * was released; the hardware should be notified of this change.
3715 	 */
3716 	return 1;
3717 }
3718 
3719 /*
3720  * Set the link layer address on an interface.
3721  *
3722  * At this time we only support certain types of interfaces,
3723  * and we don't allow the length of the address to change.
3724  *
3725  * Set noinline to be dtrace-friendly
3726  */
3727 __noinline int
3728 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3729 {
3730 	struct sockaddr_dl *sdl;
3731 	struct ifaddr *ifa;
3732 	struct ifreq ifr;
3733 
3734 	ifa = ifp->if_addr;
3735 	if (ifa == NULL)
3736 		return (EINVAL);
3737 
3738 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3739 	if (sdl == NULL)
3740 		return (EINVAL);
3741 
3742 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3743 		return (EINVAL);
3744 
3745 	switch (ifp->if_type) {
3746 	case IFT_ETHER:
3747 	case IFT_XETHER:
3748 	case IFT_L2VLAN:
3749 	case IFT_BRIDGE:
3750 	case IFT_IEEE8023ADLAG:
3751 		bcopy(lladdr, LLADDR(sdl), len);
3752 		break;
3753 	default:
3754 		return (ENODEV);
3755 	}
3756 
3757 	/*
3758 	 * If the interface is already up, we need
3759 	 * to re-init it in order to reprogram its
3760 	 * address filter.
3761 	 */
3762 	if ((ifp->if_flags & IFF_UP) != 0) {
3763 		if (ifp->if_ioctl) {
3764 			ifp->if_flags &= ~IFF_UP;
3765 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3766 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3767 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3768 			ifp->if_flags |= IFF_UP;
3769 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3770 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3771 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3772 		}
3773 	}
3774 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3775 
3776 	return (0);
3777 }
3778 
3779 /*
3780  * Compat function for handling basic encapsulation requests.
3781  * Not converted stacks (FDDI, IB, ..) supports traditional
3782  * output model: ARP (and other similar L2 protocols) are handled
3783  * inside output routine, arpresolve/nd6_resolve() returns MAC
3784  * address instead of full prepend.
3785  *
3786  * This function creates calculated header==MAC for IPv4/IPv6 and
3787  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3788  * address families.
3789  */
3790 static int
3791 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3792 {
3793 
3794 	if (req->rtype != IFENCAP_LL)
3795 		return (EOPNOTSUPP);
3796 
3797 	if (req->bufsize < req->lladdr_len)
3798 		return (ENOMEM);
3799 
3800 	switch (req->family) {
3801 	case AF_INET:
3802 	case AF_INET6:
3803 		break;
3804 	default:
3805 		return (EAFNOSUPPORT);
3806 	}
3807 
3808 	/* Copy lladdr to storage as is */
3809 	memmove(req->buf, req->lladdr, req->lladdr_len);
3810 	req->bufsize = req->lladdr_len;
3811 	req->lladdr_off = 0;
3812 
3813 	return (0);
3814 }
3815 
3816 /*
3817  * Tunnel interfaces can nest, also they may cause infinite recursion
3818  * calls when misconfigured. We'll prevent this by detecting loops.
3819  * High nesting level may cause stack exhaustion. We'll prevent this
3820  * by introducing upper limit.
3821  *
3822  * Return 0, if tunnel nesting count is equal or less than limit.
3823  */
3824 int
3825 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3826     int limit)
3827 {
3828 	struct m_tag *mtag;
3829 	int count;
3830 
3831 	count = 1;
3832 	mtag = NULL;
3833 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3834 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3835 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3836 			return (EIO);
3837 		}
3838 		count++;
3839 	}
3840 	if (count > limit) {
3841 		log(LOG_NOTICE,
3842 		    "%s: if_output recursively called too many times(%d)\n",
3843 		    if_name(ifp), count);
3844 		return (EIO);
3845 	}
3846 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3847 	if (mtag == NULL)
3848 		return (ENOMEM);
3849 	*(struct ifnet **)(mtag + 1) = ifp;
3850 	m_tag_prepend(m, mtag);
3851 	return (0);
3852 }
3853 
3854 /*
3855  * Get the link layer address that was read from the hardware at attach.
3856  *
3857  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3858  * their component interfaces as IFT_IEEE8023ADLAG.
3859  */
3860 int
3861 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3862 {
3863 
3864 	if (ifp->if_hw_addr == NULL)
3865 		return (ENODEV);
3866 
3867 	switch (ifp->if_type) {
3868 	case IFT_ETHER:
3869 	case IFT_IEEE8023ADLAG:
3870 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3871 		return (0);
3872 	default:
3873 		return (ENODEV);
3874 	}
3875 }
3876 
3877 /*
3878  * The name argument must be a pointer to storage which will last as
3879  * long as the interface does.  For physical devices, the result of
3880  * device_get_name(dev) is a good choice and for pseudo-devices a
3881  * static string works well.
3882  */
3883 void
3884 if_initname(struct ifnet *ifp, const char *name, int unit)
3885 {
3886 	ifp->if_dname = name;
3887 	ifp->if_dunit = unit;
3888 	if (unit != IF_DUNIT_NONE)
3889 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3890 	else
3891 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3892 }
3893 
3894 static int
3895 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
3896 {
3897 	char if_fmt[256];
3898 
3899 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3900 	vlog(pri, if_fmt, ap);
3901 	return (0);
3902 }
3903 
3904 
3905 int
3906 if_printf(struct ifnet *ifp, const char *fmt, ...)
3907 {
3908 	va_list ap;
3909 
3910 	va_start(ap, fmt);
3911 	if_vlog(ifp, LOG_INFO, fmt, ap);
3912 	va_end(ap);
3913 	return (0);
3914 }
3915 
3916 int
3917 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
3918 {
3919 	va_list ap;
3920 
3921 	va_start(ap, fmt);
3922 	if_vlog(ifp, pri, fmt, ap);
3923 	va_end(ap);
3924 	return (0);
3925 }
3926 
3927 void
3928 if_start(struct ifnet *ifp)
3929 {
3930 
3931 	(*(ifp)->if_start)(ifp);
3932 }
3933 
3934 /*
3935  * Backwards compatibility interface for drivers
3936  * that have not implemented it
3937  */
3938 static int
3939 if_transmit(struct ifnet *ifp, struct mbuf *m)
3940 {
3941 	int error;
3942 
3943 	IFQ_HANDOFF(ifp, m, error);
3944 	return (error);
3945 }
3946 
3947 static void
3948 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3949 {
3950 
3951 	m_freem(m);
3952 }
3953 
3954 int
3955 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3956 {
3957 	int active = 0;
3958 
3959 	IF_LOCK(ifq);
3960 	if (_IF_QFULL(ifq)) {
3961 		IF_UNLOCK(ifq);
3962 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3963 		m_freem(m);
3964 		return (0);
3965 	}
3966 	if (ifp != NULL) {
3967 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3968 		if (m->m_flags & (M_BCAST|M_MCAST))
3969 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3970 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3971 	}
3972 	_IF_ENQUEUE(ifq, m);
3973 	IF_UNLOCK(ifq);
3974 	if (ifp != NULL && !active)
3975 		(*(ifp)->if_start)(ifp);
3976 	return (1);
3977 }
3978 
3979 void
3980 if_register_com_alloc(u_char type,
3981     if_com_alloc_t *a, if_com_free_t *f)
3982 {
3983 
3984 	KASSERT(if_com_alloc[type] == NULL,
3985 	    ("if_register_com_alloc: %d already registered", type));
3986 	KASSERT(if_com_free[type] == NULL,
3987 	    ("if_register_com_alloc: %d free already registered", type));
3988 
3989 	if_com_alloc[type] = a;
3990 	if_com_free[type] = f;
3991 }
3992 
3993 void
3994 if_deregister_com_alloc(u_char type)
3995 {
3996 
3997 	KASSERT(if_com_alloc[type] != NULL,
3998 	    ("if_deregister_com_alloc: %d not registered", type));
3999 	KASSERT(if_com_free[type] != NULL,
4000 	    ("if_deregister_com_alloc: %d free not registered", type));
4001 
4002 	/*
4003 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4004 	 * fixes issues about late invocation of if_destroy(), which leads
4005 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4006 	 */
4007 	epoch_drain_callbacks(net_epoch_preempt);
4008 
4009 	if_com_alloc[type] = NULL;
4010 	if_com_free[type] = NULL;
4011 }
4012 
4013 /* API for driver access to network stack owned ifnet.*/
4014 uint64_t
4015 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4016 {
4017 	uint64_t oldbrate;
4018 
4019 	oldbrate = ifp->if_baudrate;
4020 	ifp->if_baudrate = baudrate;
4021 	return (oldbrate);
4022 }
4023 
4024 uint64_t
4025 if_getbaudrate(if_t ifp)
4026 {
4027 
4028 	return (((struct ifnet *)ifp)->if_baudrate);
4029 }
4030 
4031 int
4032 if_setcapabilities(if_t ifp, int capabilities)
4033 {
4034 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4035 	return (0);
4036 }
4037 
4038 int
4039 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4040 {
4041 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4042 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4043 
4044 	return (0);
4045 }
4046 
4047 int
4048 if_getcapabilities(if_t ifp)
4049 {
4050 	return ((struct ifnet *)ifp)->if_capabilities;
4051 }
4052 
4053 int
4054 if_setcapenable(if_t ifp, int capabilities)
4055 {
4056 	((struct ifnet *)ifp)->if_capenable = capabilities;
4057 	return (0);
4058 }
4059 
4060 int
4061 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4062 {
4063 	if(setcap)
4064 		((struct ifnet *)ifp)->if_capenable |= setcap;
4065 	if(clearcap)
4066 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4067 
4068 	return (0);
4069 }
4070 
4071 const char *
4072 if_getdname(if_t ifp)
4073 {
4074 	return ((struct ifnet *)ifp)->if_dname;
4075 }
4076 
4077 int
4078 if_togglecapenable(if_t ifp, int togglecap)
4079 {
4080 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4081 	return (0);
4082 }
4083 
4084 int
4085 if_getcapenable(if_t ifp)
4086 {
4087 	return ((struct ifnet *)ifp)->if_capenable;
4088 }
4089 
4090 /*
4091  * This is largely undesirable because it ties ifnet to a device, but does
4092  * provide flexiblity for an embedded product vendor. Should be used with
4093  * the understanding that it violates the interface boundaries, and should be
4094  * a last resort only.
4095  */
4096 int
4097 if_setdev(if_t ifp, void *dev)
4098 {
4099 	return (0);
4100 }
4101 
4102 int
4103 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4104 {
4105 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4106 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4107 
4108 	return (0);
4109 }
4110 
4111 int
4112 if_getdrvflags(if_t ifp)
4113 {
4114 	return ((struct ifnet *)ifp)->if_drv_flags;
4115 }
4116 
4117 int
4118 if_setdrvflags(if_t ifp, int flags)
4119 {
4120 	((struct ifnet *)ifp)->if_drv_flags = flags;
4121 	return (0);
4122 }
4123 
4124 int
4125 if_setflags(if_t ifp, int flags)
4126 {
4127 
4128 	ifp->if_flags = flags;
4129 	return (0);
4130 }
4131 
4132 int
4133 if_setflagbits(if_t ifp, int set, int clear)
4134 {
4135 	((struct ifnet *)ifp)->if_flags |= set;
4136 	((struct ifnet *)ifp)->if_flags &= ~clear;
4137 
4138 	return (0);
4139 }
4140 
4141 int
4142 if_getflags(if_t ifp)
4143 {
4144 	return ((struct ifnet *)ifp)->if_flags;
4145 }
4146 
4147 int
4148 if_clearhwassist(if_t ifp)
4149 {
4150 	((struct ifnet *)ifp)->if_hwassist = 0;
4151 	return (0);
4152 }
4153 
4154 int
4155 if_sethwassistbits(if_t ifp, int toset, int toclear)
4156 {
4157 	((struct ifnet *)ifp)->if_hwassist |= toset;
4158 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4159 
4160 	return (0);
4161 }
4162 
4163 int
4164 if_sethwassist(if_t ifp, int hwassist_bit)
4165 {
4166 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4167 	return (0);
4168 }
4169 
4170 int
4171 if_gethwassist(if_t ifp)
4172 {
4173 	return ((struct ifnet *)ifp)->if_hwassist;
4174 }
4175 
4176 int
4177 if_setmtu(if_t ifp, int mtu)
4178 {
4179 	((struct ifnet *)ifp)->if_mtu = mtu;
4180 	return (0);
4181 }
4182 
4183 int
4184 if_getmtu(if_t ifp)
4185 {
4186 	return ((struct ifnet *)ifp)->if_mtu;
4187 }
4188 
4189 int
4190 if_getmtu_family(if_t ifp, int family)
4191 {
4192 	struct domain *dp;
4193 
4194 	for (dp = domains; dp; dp = dp->dom_next) {
4195 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4196 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4197 	}
4198 
4199 	return (((struct ifnet *)ifp)->if_mtu);
4200 }
4201 
4202 /*
4203  * Methods for drivers to access interface unicast and multicast
4204  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4205  * 'struct ifmultiaddr'.
4206  */
4207 u_int
4208 if_lladdr_count(if_t ifp)
4209 {
4210 	struct epoch_tracker et;
4211 	struct ifaddr *ifa;
4212 	u_int count;
4213 
4214 	count = 0;
4215 	NET_EPOCH_ENTER(et);
4216 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4217 		if (ifa->ifa_addr->sa_family == AF_LINK)
4218 			count++;
4219 	NET_EPOCH_EXIT(et);
4220 
4221 	return (count);
4222 }
4223 
4224 u_int
4225 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4226 {
4227 	struct epoch_tracker et;
4228 	struct ifaddr *ifa;
4229 	u_int count;
4230 
4231 	MPASS(cb);
4232 
4233 	count = 0;
4234 	NET_EPOCH_ENTER(et);
4235 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4236 		if (ifa->ifa_addr->sa_family != AF_LINK)
4237 			continue;
4238 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4239 		    count);
4240 	}
4241 	NET_EPOCH_EXIT(et);
4242 
4243 	return (count);
4244 }
4245 
4246 u_int
4247 if_llmaddr_count(if_t ifp)
4248 {
4249 	struct epoch_tracker et;
4250 	struct ifmultiaddr *ifma;
4251 	int count;
4252 
4253 	count = 0;
4254 	NET_EPOCH_ENTER(et);
4255 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4256 		if (ifma->ifma_addr->sa_family == AF_LINK)
4257 			count++;
4258 	NET_EPOCH_EXIT(et);
4259 
4260 	return (count);
4261 }
4262 
4263 u_int
4264 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4265 {
4266 	struct epoch_tracker et;
4267 	struct ifmultiaddr *ifma;
4268 	u_int count;
4269 
4270 	MPASS(cb);
4271 
4272 	count = 0;
4273 	NET_EPOCH_ENTER(et);
4274 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4275 		if (ifma->ifma_addr->sa_family != AF_LINK)
4276 			continue;
4277 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4278 		    count);
4279 	}
4280 	NET_EPOCH_EXIT(et);
4281 
4282 	return (count);
4283 }
4284 
4285 int
4286 if_setsoftc(if_t ifp, void *softc)
4287 {
4288 	((struct ifnet *)ifp)->if_softc = softc;
4289 	return (0);
4290 }
4291 
4292 void *
4293 if_getsoftc(if_t ifp)
4294 {
4295 	return ((struct ifnet *)ifp)->if_softc;
4296 }
4297 
4298 void
4299 if_setrcvif(struct mbuf *m, if_t ifp)
4300 {
4301 
4302 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4303 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4304 }
4305 
4306 void
4307 if_setvtag(struct mbuf *m, uint16_t tag)
4308 {
4309 	m->m_pkthdr.ether_vtag = tag;
4310 }
4311 
4312 uint16_t
4313 if_getvtag(struct mbuf *m)
4314 {
4315 
4316 	return (m->m_pkthdr.ether_vtag);
4317 }
4318 
4319 int
4320 if_sendq_empty(if_t ifp)
4321 {
4322 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4323 }
4324 
4325 struct ifaddr *
4326 if_getifaddr(if_t ifp)
4327 {
4328 	return ((struct ifnet *)ifp)->if_addr;
4329 }
4330 
4331 int
4332 if_getamcount(if_t ifp)
4333 {
4334 	return ((struct ifnet *)ifp)->if_amcount;
4335 }
4336 
4337 int
4338 if_setsendqready(if_t ifp)
4339 {
4340 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4341 	return (0);
4342 }
4343 
4344 int
4345 if_setsendqlen(if_t ifp, int tx_desc_count)
4346 {
4347 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4348 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4349 
4350 	return (0);
4351 }
4352 
4353 int
4354 if_vlantrunkinuse(if_t ifp)
4355 {
4356 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4357 }
4358 
4359 int
4360 if_input(if_t ifp, struct mbuf* sendmp)
4361 {
4362 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4363 	return (0);
4364 
4365 }
4366 
4367 struct mbuf *
4368 if_dequeue(if_t ifp)
4369 {
4370 	struct mbuf *m;
4371 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4372 
4373 	return (m);
4374 }
4375 
4376 int
4377 if_sendq_prepend(if_t ifp, struct mbuf *m)
4378 {
4379 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4380 	return (0);
4381 }
4382 
4383 int
4384 if_setifheaderlen(if_t ifp, int len)
4385 {
4386 	((struct ifnet *)ifp)->if_hdrlen = len;
4387 	return (0);
4388 }
4389 
4390 caddr_t
4391 if_getlladdr(if_t ifp)
4392 {
4393 	return (IF_LLADDR((struct ifnet *)ifp));
4394 }
4395 
4396 void *
4397 if_gethandle(u_char type)
4398 {
4399 	return (if_alloc(type));
4400 }
4401 
4402 void
4403 if_bpfmtap(if_t ifh, struct mbuf *m)
4404 {
4405 	struct ifnet *ifp = (struct ifnet *)ifh;
4406 
4407 	BPF_MTAP(ifp, m);
4408 }
4409 
4410 void
4411 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4412 {
4413 	struct ifnet *ifp = (struct ifnet *)ifh;
4414 
4415 	ETHER_BPF_MTAP(ifp, m);
4416 }
4417 
4418 void
4419 if_vlancap(if_t ifh)
4420 {
4421 	struct ifnet *ifp = (struct ifnet *)ifh;
4422 	VLAN_CAPABILITIES(ifp);
4423 }
4424 
4425 int
4426 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4427 {
4428 
4429 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4430         return (0);
4431 }
4432 
4433 int
4434 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4435 {
4436 
4437 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4438         return (0);
4439 }
4440 
4441 int
4442 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4443 {
4444 
4445 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4446         return (0);
4447 }
4448 
4449 u_int
4450 if_gethwtsomax(if_t ifp)
4451 {
4452 
4453 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4454 }
4455 
4456 u_int
4457 if_gethwtsomaxsegcount(if_t ifp)
4458 {
4459 
4460 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4461 }
4462 
4463 u_int
4464 if_gethwtsomaxsegsize(if_t ifp)
4465 {
4466 
4467 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4468 }
4469 
4470 void
4471 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4472 {
4473 	((struct ifnet *)ifp)->if_init = init_fn;
4474 }
4475 
4476 void
4477 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4478 {
4479 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4480 }
4481 
4482 void
4483 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4484 {
4485 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4486 }
4487 
4488 void
4489 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4490 {
4491 	((struct ifnet *)ifp)->if_transmit = start_fn;
4492 }
4493 
4494 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4495 {
4496 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4497 
4498 }
4499 
4500 void
4501 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4502 {
4503 
4504 	ifp->if_get_counter = fn;
4505 }
4506 
4507 #ifdef DDB
4508 static void
4509 if_show_ifnet(struct ifnet *ifp)
4510 {
4511 
4512 	if (ifp == NULL)
4513 		return;
4514 	db_printf("%s:\n", ifp->if_xname);
4515 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
4516 	IF_DB_PRINTF("%s", if_dname);
4517 	IF_DB_PRINTF("%d", if_dunit);
4518 	IF_DB_PRINTF("%s", if_description);
4519 	IF_DB_PRINTF("%u", if_index);
4520 	IF_DB_PRINTF("%d", if_idxgen);
4521 	IF_DB_PRINTF("%u", if_refcount);
4522 	IF_DB_PRINTF("%p", if_softc);
4523 	IF_DB_PRINTF("%p", if_l2com);
4524 	IF_DB_PRINTF("%p", if_llsoftc);
4525 	IF_DB_PRINTF("%d", if_amcount);
4526 	IF_DB_PRINTF("%p", if_addr);
4527 	IF_DB_PRINTF("%p", if_broadcastaddr);
4528 	IF_DB_PRINTF("%p", if_afdata);
4529 	IF_DB_PRINTF("%d", if_afdata_initialized);
4530 	IF_DB_PRINTF("%u", if_fib);
4531 	IF_DB_PRINTF("%p", if_vnet);
4532 	IF_DB_PRINTF("%p", if_home_vnet);
4533 	IF_DB_PRINTF("%p", if_vlantrunk);
4534 	IF_DB_PRINTF("%p", if_bpf);
4535 	IF_DB_PRINTF("%u", if_pcount);
4536 	IF_DB_PRINTF("%p", if_bridge);
4537 	IF_DB_PRINTF("%p", if_lagg);
4538 	IF_DB_PRINTF("%p", if_pf_kif);
4539 	IF_DB_PRINTF("%p", if_carp);
4540 	IF_DB_PRINTF("%p", if_label);
4541 	IF_DB_PRINTF("%p", if_netmap);
4542 	IF_DB_PRINTF("0x%08x", if_flags);
4543 	IF_DB_PRINTF("0x%08x", if_drv_flags);
4544 	IF_DB_PRINTF("0x%08x", if_capabilities);
4545 	IF_DB_PRINTF("0x%08x", if_capenable);
4546 	IF_DB_PRINTF("%p", if_snd.ifq_head);
4547 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
4548 	IF_DB_PRINTF("%d", if_snd.ifq_len);
4549 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
4550 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
4551 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
4552 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
4553 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
4554 	IF_DB_PRINTF("%d", if_snd.altq_type);
4555 	IF_DB_PRINTF("%x", if_snd.altq_flags);
4556 #undef IF_DB_PRINTF
4557 }
4558 
4559 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
4560 {
4561 
4562 	if (!have_addr) {
4563 		db_printf("usage: show ifnet <struct ifnet *>\n");
4564 		return;
4565 	}
4566 
4567 	if_show_ifnet((struct ifnet *)addr);
4568 }
4569 
4570 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
4571 {
4572 	struct ifnet *ifp;
4573 	u_short idx;
4574 
4575 	for (idx = 1; idx <= if_index; idx++) {
4576 		ifp = ifindex_table[idx].ife_ifnet;
4577 		if (ifp == NULL)
4578 			continue;
4579 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
4580 		if (db_pager_quit)
4581 			break;
4582 	}
4583 }
4584 #endif	/* DDB */
4585