xref: /freebsd/sys/net/if.c (revision 732a02b4e77866604a120a275c082bb6221bd2ff)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_bpf.h"
36 #include "opt_inet6.h"
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/conf.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/domainset.h>
44 #include <sys/sbuf.h>
45 #include <sys/bus.h>
46 #include <sys/epoch.h>
47 #include <sys/mbuf.h>
48 #include <sys/systm.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/refcount.h>
57 #include <sys/module.h>
58 #include <sys/rwlock.h>
59 #include <sys/sockio.h>
60 #include <sys/syslog.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/taskqueue.h>
64 #include <sys/domain.h>
65 #include <sys/jail.h>
66 #include <sys/priv.h>
67 
68 #include <machine/stdarg.h>
69 #include <vm/uma.h>
70 
71 #include <net/bpf.h>
72 #include <net/ethernet.h>
73 #include <net/if.h>
74 #include <net/if_arp.h>
75 #include <net/if_clone.h>
76 #include <net/if_dl.h>
77 #include <net/if_types.h>
78 #include <net/if_var.h>
79 #include <net/if_media.h>
80 #include <net/if_vlan_var.h>
81 #include <net/radix.h>
82 #include <net/route.h>
83 #include <net/vnet.h>
84 
85 #if defined(INET) || defined(INET6)
86 #include <net/ethernet.h>
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip.h>
90 #include <netinet/ip_carp.h>
91 #ifdef INET
92 #include <net/debugnet.h>
93 #include <netinet/if_ether.h>
94 #endif /* INET */
95 #ifdef INET6
96 #include <netinet6/in6_var.h>
97 #include <netinet6/in6_ifattach.h>
98 #endif /* INET6 */
99 #endif /* INET || INET6 */
100 
101 #include <security/mac/mac_framework.h>
102 
103 /*
104  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
105  * and ifr_ifru when it is used in SIOCGIFCONF.
106  */
107 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
108     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
109 
110 __read_mostly epoch_t net_epoch_preempt;
111 #ifdef COMPAT_FREEBSD32
112 #include <sys/mount.h>
113 #include <compat/freebsd32/freebsd32.h>
114 
115 struct ifreq_buffer32 {
116 	uint32_t	length;		/* (size_t) */
117 	uint32_t	buffer;		/* (void *) */
118 };
119 
120 /*
121  * Interface request structure used for socket
122  * ioctl's.  All interface ioctl's must have parameter
123  * definitions which begin with ifr_name.  The
124  * remainder may be interface specific.
125  */
126 struct ifreq32 {
127 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
128 	union {
129 		struct sockaddr	ifru_addr;
130 		struct sockaddr	ifru_dstaddr;
131 		struct sockaddr	ifru_broadaddr;
132 		struct ifreq_buffer32 ifru_buffer;
133 		short		ifru_flags[2];
134 		short		ifru_index;
135 		int		ifru_jid;
136 		int		ifru_metric;
137 		int		ifru_mtu;
138 		int		ifru_phys;
139 		int		ifru_media;
140 		uint32_t	ifru_data;
141 		int		ifru_cap[2];
142 		u_int		ifru_fib;
143 		u_char		ifru_vlan_pcp;
144 	} ifr_ifru;
145 };
146 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
147 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
148     __offsetof(struct ifreq32, ifr_ifru));
149 
150 struct ifgroupreq32 {
151 	char	ifgr_name[IFNAMSIZ];
152 	u_int	ifgr_len;
153 	union {
154 		char		ifgru_group[IFNAMSIZ];
155 		uint32_t	ifgru_groups;
156 	} ifgr_ifgru;
157 };
158 
159 struct ifmediareq32 {
160 	char		ifm_name[IFNAMSIZ];
161 	int		ifm_current;
162 	int		ifm_mask;
163 	int		ifm_status;
164 	int		ifm_active;
165 	int		ifm_count;
166 	uint32_t	ifm_ulist;	/* (int *) */
167 };
168 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
169 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
170 
171 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
172     _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
173 #else /* !COMPAT_FREEBSD32 */
174 #define _CASE_IOC_IFGROUPREQ_32(cmd)
175 #endif /* !COMPAT_FREEBSD32 */
176 
177 #define CASE_IOC_IFGROUPREQ(cmd)	\
178     _CASE_IOC_IFGROUPREQ_32(cmd)	\
179     (cmd)
180 
181 union ifreq_union {
182 	struct ifreq	ifr;
183 #ifdef COMPAT_FREEBSD32
184 	struct ifreq32	ifr32;
185 #endif
186 };
187 
188 union ifgroupreq_union {
189 	struct ifgroupreq ifgr;
190 #ifdef COMPAT_FREEBSD32
191 	struct ifgroupreq32 ifgr32;
192 #endif
193 };
194 
195 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
196     "Link layers");
197 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
198     "Generic link-management");
199 
200 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
201     &ifqmaxlen, 0, "max send queue size");
202 
203 /* Log link state change events */
204 static int log_link_state_change = 1;
205 
206 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
207 	&log_link_state_change, 0,
208 	"log interface link state change events");
209 
210 /* Log promiscuous mode change events */
211 static int log_promisc_mode_change = 1;
212 
213 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
214 	&log_promisc_mode_change, 1,
215 	"log promiscuous mode change events");
216 
217 /* Interface description */
218 static unsigned int ifdescr_maxlen = 1024;
219 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
220 	&ifdescr_maxlen, 0,
221 	"administrative maximum length for interface description");
222 
223 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
224 
225 /* global sx for non-critical path ifdescr */
226 static struct sx ifdescr_sx;
227 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
228 
229 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
230 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
231 /* These are external hooks for CARP. */
232 void	(*carp_linkstate_p)(struct ifnet *ifp);
233 void	(*carp_demote_adj_p)(int, char *);
234 int	(*carp_master_p)(struct ifaddr *);
235 #if defined(INET) || defined(INET6)
236 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
237 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
238     const struct sockaddr *sa);
239 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
240 int	(*carp_attach_p)(struct ifaddr *, int);
241 void	(*carp_detach_p)(struct ifaddr *, bool);
242 #endif
243 #ifdef INET
244 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
245 #endif
246 #ifdef INET6
247 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
248 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
249     const struct in6_addr *taddr);
250 #endif
251 
252 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
253 
254 /*
255  * XXX: Style; these should be sorted alphabetically, and unprototyped
256  * static functions should be prototyped. Currently they are sorted by
257  * declaration order.
258  */
259 static void	if_attachdomain(void *);
260 static void	if_attachdomain1(struct ifnet *);
261 static int	ifconf(u_long, caddr_t);
262 static void	*if_grow(void);
263 static void	if_input_default(struct ifnet *, struct mbuf *);
264 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
265 static void	if_route(struct ifnet *, int flag, int fam);
266 static int	if_setflag(struct ifnet *, int, int, int *, int);
267 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
268 static void	if_unroute(struct ifnet *, int flag, int fam);
269 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
270 static void	do_link_state_change(void *, int);
271 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
272 static int	if_getgroupmembers(struct ifgroupreq *);
273 static void	if_delgroups(struct ifnet *);
274 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
275 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
276 static void	if_siocaddmulti(void *, int);
277 #ifdef VIMAGE
278 static int	if_vmove(struct ifnet *, struct vnet *);
279 #endif
280 
281 #ifdef INET6
282 /*
283  * XXX: declare here to avoid to include many inet6 related files..
284  * should be more generalized?
285  */
286 extern void	nd6_setmtu(struct ifnet *);
287 #endif
288 
289 /* ipsec helper hooks */
290 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
291 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
292 
293 VNET_DEFINE(int, if_index);
294 int	ifqmaxlen = IFQ_MAXLEN;
295 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
296 VNET_DEFINE(struct ifgrouphead, ifg_head);
297 
298 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
299 
300 /* Table of ifnet by index. */
301 VNET_DEFINE(struct ifnet **, ifindex_table);
302 
303 #define	V_if_indexlim		VNET(if_indexlim)
304 #define	V_ifindex_table		VNET(ifindex_table)
305 
306 /*
307  * The global network interface list (V_ifnet) and related state (such as
308  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
309  * an rwlock.  Either may be acquired shared to stablize the list, but both
310  * must be acquired writable to modify the list.  This model allows us to
311  * both stablize the interface list during interrupt thread processing, but
312  * also to stablize it over long-running ioctls, without introducing priority
313  * inversions and deadlocks.
314  */
315 struct rwlock ifnet_rwlock;
316 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
317 struct sx ifnet_sxlock;
318 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
319 
320 /*
321  * The allocation of network interfaces is a rather non-atomic affair; we
322  * need to select an index before we are ready to expose the interface for
323  * use, so will use this pointer value to indicate reservation.
324  */
325 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
326 
327 #ifdef VIMAGE
328 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
329     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
330 #endif
331 
332 static	if_com_alloc_t *if_com_alloc[256];
333 static	if_com_free_t *if_com_free[256];
334 
335 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
336 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
337 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
338 
339 struct ifnet *
340 ifnet_byindex(u_short idx)
341 {
342 	struct ifnet *ifp;
343 
344 	if (__predict_false(idx > V_if_index))
345 		return (NULL);
346 
347 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
348 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
349 }
350 
351 struct ifnet *
352 ifnet_byindex_ref(u_short idx)
353 {
354 	struct ifnet *ifp;
355 
356 	NET_EPOCH_ASSERT();
357 
358 	ifp = ifnet_byindex(idx);
359 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
360 		return (NULL);
361 	if_ref(ifp);
362 	return (ifp);
363 }
364 
365 /*
366  * Allocate an ifindex array entry; return 0 on success or an error on
367  * failure.
368  */
369 static u_short
370 ifindex_alloc(void **old)
371 {
372 	u_short idx;
373 
374 	IFNET_WLOCK_ASSERT();
375 	/*
376 	 * Try to find an empty slot below V_if_index.  If we fail, take the
377 	 * next slot.
378 	 */
379 	for (idx = 1; idx <= V_if_index; idx++) {
380 		if (V_ifindex_table[idx] == NULL)
381 			break;
382 	}
383 
384 	/* Catch if_index overflow. */
385 	if (idx >= V_if_indexlim) {
386 		*old = if_grow();
387 		return (USHRT_MAX);
388 	}
389 	if (idx > V_if_index)
390 		V_if_index = idx;
391 	return (idx);
392 }
393 
394 static void
395 ifindex_free_locked(u_short idx)
396 {
397 
398 	IFNET_WLOCK_ASSERT();
399 
400 	V_ifindex_table[idx] = NULL;
401 	while (V_if_index > 0 &&
402 	    V_ifindex_table[V_if_index] == NULL)
403 		V_if_index--;
404 }
405 
406 static void
407 ifindex_free(u_short idx)
408 {
409 
410 	IFNET_WLOCK();
411 	ifindex_free_locked(idx);
412 	IFNET_WUNLOCK();
413 }
414 
415 static void
416 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
417 {
418 
419 	V_ifindex_table[idx] = ifp;
420 }
421 
422 struct ifaddr *
423 ifaddr_byindex(u_short idx)
424 {
425 	struct ifnet *ifp;
426 	struct ifaddr *ifa = NULL;
427 
428 	NET_EPOCH_ASSERT();
429 
430 	ifp = ifnet_byindex(idx);
431 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
432 		ifa_ref(ifa);
433 	return (ifa);
434 }
435 
436 /*
437  * Network interface utility routines.
438  *
439  * Routines with ifa_ifwith* names take sockaddr *'s as
440  * parameters.
441  */
442 
443 static void
444 vnet_if_init(const void *unused __unused)
445 {
446 	void *old;
447 
448 	CK_STAILQ_INIT(&V_ifnet);
449 	CK_STAILQ_INIT(&V_ifg_head);
450 	IFNET_WLOCK();
451 	old = if_grow();				/* create initial table */
452 	IFNET_WUNLOCK();
453 	epoch_wait_preempt(net_epoch_preempt);
454 	free(old, M_IFNET);
455 	vnet_if_clone_init();
456 }
457 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
458     NULL);
459 
460 #ifdef VIMAGE
461 static void
462 vnet_if_uninit(const void *unused __unused)
463 {
464 
465 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
466 	    "not empty", __func__, __LINE__, &V_ifnet));
467 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
468 	    "not empty", __func__, __LINE__, &V_ifg_head));
469 
470 	free((caddr_t)V_ifindex_table, M_IFNET);
471 }
472 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
473     vnet_if_uninit, NULL);
474 
475 static void
476 vnet_if_return(const void *unused __unused)
477 {
478 	struct ifnet *ifp, *nifp;
479 
480 	/* Return all inherited interfaces to their parent vnets. */
481 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
482 		if (ifp->if_home_vnet != ifp->if_vnet)
483 			if_vmove(ifp, ifp->if_home_vnet);
484 	}
485 }
486 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
487     vnet_if_return, NULL);
488 #endif
489 
490 
491 static void *
492 if_grow(void)
493 {
494 	int oldlim;
495 	u_int n;
496 	struct ifnet **e;
497 	void *old;
498 
499 	old = NULL;
500 	IFNET_WLOCK_ASSERT();
501 	oldlim = V_if_indexlim;
502 	IFNET_WUNLOCK();
503 	n = (oldlim << 1) * sizeof(*e);
504 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
505 	IFNET_WLOCK();
506 	if (V_if_indexlim != oldlim) {
507 		free(e, M_IFNET);
508 		return (NULL);
509 	}
510 	if (V_ifindex_table != NULL) {
511 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
512 		old = V_ifindex_table;
513 	}
514 	V_if_indexlim <<= 1;
515 	V_ifindex_table = e;
516 	return (old);
517 }
518 
519 /*
520  * Allocate a struct ifnet and an index for an interface.  A layer 2
521  * common structure will also be allocated if an allocation routine is
522  * registered for the passed type.
523  */
524 struct ifnet *
525 if_alloc_domain(u_char type, int numa_domain)
526 {
527 	struct ifnet *ifp;
528 	u_short idx;
529 	void *old;
530 
531 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
532 	if (numa_domain == IF_NODOM)
533 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
534 		    M_WAITOK | M_ZERO);
535 	else
536 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
537 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
538  restart:
539 	IFNET_WLOCK();
540 	idx = ifindex_alloc(&old);
541 	if (__predict_false(idx == USHRT_MAX)) {
542 		IFNET_WUNLOCK();
543 		epoch_wait_preempt(net_epoch_preempt);
544 		free(old, M_IFNET);
545 		goto restart;
546 	}
547 	ifnet_setbyindex(idx, IFNET_HOLD);
548 	IFNET_WUNLOCK();
549 	ifp->if_index = idx;
550 	ifp->if_type = type;
551 	ifp->if_alloctype = type;
552 	ifp->if_numa_domain = numa_domain;
553 #ifdef VIMAGE
554 	ifp->if_vnet = curvnet;
555 #endif
556 	if (if_com_alloc[type] != NULL) {
557 		ifp->if_l2com = if_com_alloc[type](type, ifp);
558 		if (ifp->if_l2com == NULL) {
559 			free(ifp, M_IFNET);
560 			ifindex_free(idx);
561 			return (NULL);
562 		}
563 	}
564 
565 	IF_ADDR_LOCK_INIT(ifp);
566 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
567 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
568 	ifp->if_afdata_initialized = 0;
569 	IF_AFDATA_LOCK_INIT(ifp);
570 	CK_STAILQ_INIT(&ifp->if_addrhead);
571 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
572 	CK_STAILQ_INIT(&ifp->if_groups);
573 #ifdef MAC
574 	mac_ifnet_init(ifp);
575 #endif
576 	ifq_init(&ifp->if_snd, ifp);
577 
578 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
579 	for (int i = 0; i < IFCOUNTERS; i++)
580 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
581 	ifp->if_get_counter = if_get_counter_default;
582 	ifp->if_pcp = IFNET_PCP_NONE;
583 	ifnet_setbyindex(ifp->if_index, ifp);
584 	return (ifp);
585 }
586 
587 struct ifnet *
588 if_alloc_dev(u_char type, device_t dev)
589 {
590 	int numa_domain;
591 
592 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
593 		return (if_alloc_domain(type, IF_NODOM));
594 	return (if_alloc_domain(type, numa_domain));
595 }
596 
597 struct ifnet *
598 if_alloc(u_char type)
599 {
600 
601 	return (if_alloc_domain(type, IF_NODOM));
602 }
603 /*
604  * Do the actual work of freeing a struct ifnet, and layer 2 common
605  * structure.  This call is made when the last reference to an
606  * interface is released.
607  */
608 static void
609 if_free_internal(struct ifnet *ifp)
610 {
611 
612 	KASSERT((ifp->if_flags & IFF_DYING),
613 	    ("if_free_internal: interface not dying"));
614 
615 	if (if_com_free[ifp->if_alloctype] != NULL)
616 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
617 		    ifp->if_alloctype);
618 
619 #ifdef MAC
620 	mac_ifnet_destroy(ifp);
621 #endif /* MAC */
622 	IF_AFDATA_DESTROY(ifp);
623 	IF_ADDR_LOCK_DESTROY(ifp);
624 	ifq_delete(&ifp->if_snd);
625 
626 	for (int i = 0; i < IFCOUNTERS; i++)
627 		counter_u64_free(ifp->if_counters[i]);
628 
629 	free(ifp->if_description, M_IFDESCR);
630 	free(ifp->if_hw_addr, M_IFADDR);
631 	if (ifp->if_numa_domain == IF_NODOM)
632 		free(ifp, M_IFNET);
633 	else
634 		free_domain(ifp, M_IFNET);
635 }
636 
637 static void
638 if_destroy(epoch_context_t ctx)
639 {
640 	struct ifnet *ifp;
641 
642 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
643 	if_free_internal(ifp);
644 }
645 
646 /*
647  * Deregister an interface and free the associated storage.
648  */
649 void
650 if_free(struct ifnet *ifp)
651 {
652 
653 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
654 
655 	CURVNET_SET_QUIET(ifp->if_vnet);
656 	IFNET_WLOCK();
657 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
658 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
659 
660 	ifindex_free_locked(ifp->if_index);
661 	IFNET_WUNLOCK();
662 
663 	if (refcount_release(&ifp->if_refcount))
664 		NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
665 	CURVNET_RESTORE();
666 }
667 
668 /*
669  * Interfaces to keep an ifnet type-stable despite the possibility of the
670  * driver calling if_free().  If there are additional references, we defer
671  * freeing the underlying data structure.
672  */
673 void
674 if_ref(struct ifnet *ifp)
675 {
676 
677 	/* We don't assert the ifnet list lock here, but arguably should. */
678 	refcount_acquire(&ifp->if_refcount);
679 }
680 
681 void
682 if_rele(struct ifnet *ifp)
683 {
684 
685 	if (!refcount_release(&ifp->if_refcount))
686 		return;
687 	NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx);
688 }
689 
690 void
691 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
692 {
693 
694 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
695 
696 	if (ifq->ifq_maxlen == 0)
697 		ifq->ifq_maxlen = ifqmaxlen;
698 
699 	ifq->altq_type = 0;
700 	ifq->altq_disc = NULL;
701 	ifq->altq_flags &= ALTQF_CANTCHANGE;
702 	ifq->altq_tbr  = NULL;
703 	ifq->altq_ifp  = ifp;
704 }
705 
706 void
707 ifq_delete(struct ifaltq *ifq)
708 {
709 	mtx_destroy(&ifq->ifq_mtx);
710 }
711 
712 /*
713  * Perform generic interface initialization tasks and attach the interface
714  * to the list of "active" interfaces.  If vmove flag is set on entry
715  * to if_attach_internal(), perform only a limited subset of initialization
716  * tasks, given that we are moving from one vnet to another an ifnet which
717  * has already been fully initialized.
718  *
719  * Note that if_detach_internal() removes group membership unconditionally
720  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
721  * Thus, when if_vmove() is applied to a cloned interface, group membership
722  * is lost while a cloned one always joins a group whose name is
723  * ifc->ifc_name.  To recover this after if_detach_internal() and
724  * if_attach_internal(), the cloner should be specified to
725  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
726  * attempts to join a group whose name is ifc->ifc_name.
727  *
728  * XXX:
729  *  - The decision to return void and thus require this function to
730  *    succeed is questionable.
731  *  - We should probably do more sanity checking.  For instance we don't
732  *    do anything to insure if_xname is unique or non-empty.
733  */
734 void
735 if_attach(struct ifnet *ifp)
736 {
737 
738 	if_attach_internal(ifp, 0, NULL);
739 }
740 
741 /*
742  * Compute the least common TSO limit.
743  */
744 void
745 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
746 {
747 	/*
748 	 * 1) If there is no limit currently, take the limit from
749 	 * the network adapter.
750 	 *
751 	 * 2) If the network adapter has a limit below the current
752 	 * limit, apply it.
753 	 */
754 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
755 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
756 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
757 	}
758 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
759 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
760 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
761 	}
762 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
763 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
764 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
765 	}
766 }
767 
768 /*
769  * Update TSO limit of a network adapter.
770  *
771  * Returns zero if no change. Else non-zero.
772  */
773 int
774 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
775 {
776 	int retval = 0;
777 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
778 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
779 		retval++;
780 	}
781 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
782 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
783 		retval++;
784 	}
785 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
786 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
787 		retval++;
788 	}
789 	return (retval);
790 }
791 
792 static void
793 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
794 {
795 	unsigned socksize, ifasize;
796 	int namelen, masklen;
797 	struct sockaddr_dl *sdl;
798 	struct ifaddr *ifa;
799 
800 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
801 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
802 		    ifp->if_xname);
803 
804 #ifdef VIMAGE
805 	ifp->if_vnet = curvnet;
806 	if (ifp->if_home_vnet == NULL)
807 		ifp->if_home_vnet = curvnet;
808 #endif
809 
810 	if_addgroup(ifp, IFG_ALL);
811 
812 	/* Restore group membership for cloned interfaces. */
813 	if (vmove && ifc != NULL)
814 		if_clone_addgroup(ifp, ifc);
815 
816 	getmicrotime(&ifp->if_lastchange);
817 	ifp->if_epoch = time_uptime;
818 
819 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
820 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
821 	    ("transmit and qflush must both either be set or both be NULL"));
822 	if (ifp->if_transmit == NULL) {
823 		ifp->if_transmit = if_transmit;
824 		ifp->if_qflush = if_qflush;
825 	}
826 	if (ifp->if_input == NULL)
827 		ifp->if_input = if_input_default;
828 
829 	if (ifp->if_requestencap == NULL)
830 		ifp->if_requestencap = if_requestencap_default;
831 
832 	if (!vmove) {
833 #ifdef MAC
834 		mac_ifnet_create(ifp);
835 #endif
836 
837 		/*
838 		 * Create a Link Level name for this device.
839 		 */
840 		namelen = strlen(ifp->if_xname);
841 		/*
842 		 * Always save enough space for any possiable name so we
843 		 * can do a rename in place later.
844 		 */
845 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
846 		socksize = masklen + ifp->if_addrlen;
847 		if (socksize < sizeof(*sdl))
848 			socksize = sizeof(*sdl);
849 		socksize = roundup2(socksize, sizeof(long));
850 		ifasize = sizeof(*ifa) + 2 * socksize;
851 		ifa = ifa_alloc(ifasize, M_WAITOK);
852 		sdl = (struct sockaddr_dl *)(ifa + 1);
853 		sdl->sdl_len = socksize;
854 		sdl->sdl_family = AF_LINK;
855 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
856 		sdl->sdl_nlen = namelen;
857 		sdl->sdl_index = ifp->if_index;
858 		sdl->sdl_type = ifp->if_type;
859 		ifp->if_addr = ifa;
860 		ifa->ifa_ifp = ifp;
861 		ifa->ifa_addr = (struct sockaddr *)sdl;
862 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
863 		ifa->ifa_netmask = (struct sockaddr *)sdl;
864 		sdl->sdl_len = masklen;
865 		while (namelen != 0)
866 			sdl->sdl_data[--namelen] = 0xff;
867 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
868 		/* Reliably crash if used uninitialized. */
869 		ifp->if_broadcastaddr = NULL;
870 
871 		if (ifp->if_type == IFT_ETHER) {
872 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
873 			    M_WAITOK | M_ZERO);
874 		}
875 
876 #if defined(INET) || defined(INET6)
877 		/* Use defaults for TSO, if nothing is set */
878 		if (ifp->if_hw_tsomax == 0 &&
879 		    ifp->if_hw_tsomaxsegcount == 0 &&
880 		    ifp->if_hw_tsomaxsegsize == 0) {
881 			/*
882 			 * The TSO defaults needs to be such that an
883 			 * NFS mbuf list of 35 mbufs totalling just
884 			 * below 64K works and that a chain of mbufs
885 			 * can be defragged into at most 32 segments:
886 			 */
887 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
888 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
889 			ifp->if_hw_tsomaxsegcount = 35;
890 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
891 
892 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
893 			if (ifp->if_capabilities & IFCAP_TSO) {
894 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
895 				    ifp->if_hw_tsomax,
896 				    ifp->if_hw_tsomaxsegcount,
897 				    ifp->if_hw_tsomaxsegsize);
898 			}
899 		}
900 #endif
901 	}
902 #ifdef VIMAGE
903 	else {
904 		/*
905 		 * Update the interface index in the link layer address
906 		 * of the interface.
907 		 */
908 		for (ifa = ifp->if_addr; ifa != NULL;
909 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
910 			if (ifa->ifa_addr->sa_family == AF_LINK) {
911 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
912 				sdl->sdl_index = ifp->if_index;
913 			}
914 		}
915 	}
916 #endif
917 
918 	IFNET_WLOCK();
919 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
920 #ifdef VIMAGE
921 	curvnet->vnet_ifcnt++;
922 #endif
923 	IFNET_WUNLOCK();
924 
925 	if (domain_init_status >= 2)
926 		if_attachdomain1(ifp);
927 
928 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
929 	if (IS_DEFAULT_VNET(curvnet))
930 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
931 
932 	/* Announce the interface. */
933 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
934 }
935 
936 static void
937 if_epochalloc(void *dummy __unused)
938 {
939 
940 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
941 }
942 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
943 
944 static void
945 if_attachdomain(void *dummy)
946 {
947 	struct ifnet *ifp;
948 
949 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
950 		if_attachdomain1(ifp);
951 }
952 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
953     if_attachdomain, NULL);
954 
955 static void
956 if_attachdomain1(struct ifnet *ifp)
957 {
958 	struct domain *dp;
959 
960 	/*
961 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
962 	 * cannot lock ifp->if_afdata initialization, entirely.
963 	 */
964 	IF_AFDATA_LOCK(ifp);
965 	if (ifp->if_afdata_initialized >= domain_init_status) {
966 		IF_AFDATA_UNLOCK(ifp);
967 		log(LOG_WARNING, "%s called more than once on %s\n",
968 		    __func__, ifp->if_xname);
969 		return;
970 	}
971 	ifp->if_afdata_initialized = domain_init_status;
972 	IF_AFDATA_UNLOCK(ifp);
973 
974 	/* address family dependent data region */
975 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
976 	for (dp = domains; dp; dp = dp->dom_next) {
977 		if (dp->dom_ifattach)
978 			ifp->if_afdata[dp->dom_family] =
979 			    (*dp->dom_ifattach)(ifp);
980 	}
981 }
982 
983 /*
984  * Remove any unicast or broadcast network addresses from an interface.
985  */
986 void
987 if_purgeaddrs(struct ifnet *ifp)
988 {
989 	struct ifaddr *ifa;
990 
991 	while (1) {
992 		struct epoch_tracker et;
993 
994 		NET_EPOCH_ENTER(et);
995 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
996 			if (ifa->ifa_addr->sa_family != AF_LINK)
997 				break;
998 		}
999 		NET_EPOCH_EXIT(et);
1000 
1001 		if (ifa == NULL)
1002 			break;
1003 #ifdef INET
1004 		/* XXX: Ugly!! ad hoc just for INET */
1005 		if (ifa->ifa_addr->sa_family == AF_INET) {
1006 			struct ifaliasreq ifr;
1007 
1008 			bzero(&ifr, sizeof(ifr));
1009 			ifr.ifra_addr = *ifa->ifa_addr;
1010 			if (ifa->ifa_dstaddr)
1011 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1012 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1013 			    NULL) == 0)
1014 				continue;
1015 		}
1016 #endif /* INET */
1017 #ifdef INET6
1018 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1019 			in6_purgeaddr(ifa);
1020 			/* ifp_addrhead is already updated */
1021 			continue;
1022 		}
1023 #endif /* INET6 */
1024 		IF_ADDR_WLOCK(ifp);
1025 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1026 		IF_ADDR_WUNLOCK(ifp);
1027 		ifa_free(ifa);
1028 	}
1029 }
1030 
1031 /*
1032  * Remove any multicast network addresses from an interface when an ifnet
1033  * is going away.
1034  */
1035 static void
1036 if_purgemaddrs(struct ifnet *ifp)
1037 {
1038 	struct ifmultiaddr *ifma;
1039 
1040 	IF_ADDR_WLOCK(ifp);
1041 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1042 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1043 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1044 		if_delmulti_locked(ifp, ifma, 1);
1045 	}
1046 	IF_ADDR_WUNLOCK(ifp);
1047 }
1048 
1049 /*
1050  * Detach an interface, removing it from the list of "active" interfaces.
1051  * If vmove flag is set on entry to if_detach_internal(), perform only a
1052  * limited subset of cleanup tasks, given that we are moving an ifnet from
1053  * one vnet to another, where it must be fully operational.
1054  *
1055  * XXXRW: There are some significant questions about event ordering, and
1056  * how to prevent things from starting to use the interface during detach.
1057  */
1058 void
1059 if_detach(struct ifnet *ifp)
1060 {
1061 
1062 	CURVNET_SET_QUIET(ifp->if_vnet);
1063 	if_detach_internal(ifp, 0, NULL);
1064 	CURVNET_RESTORE();
1065 }
1066 
1067 /*
1068  * The vmove flag, if set, indicates that we are called from a callpath
1069  * that is moving an interface to a different vnet instance.
1070  *
1071  * The shutdown flag, if set, indicates that we are called in the
1072  * process of shutting down a vnet instance.  Currently only the
1073  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1074  * on a vnet instance shutdown without this flag being set, e.g., when
1075  * the cloned interfaces are destoyed as first thing of teardown.
1076  */
1077 static int
1078 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1079 {
1080 	struct ifaddr *ifa;
1081 	int i;
1082 	struct domain *dp;
1083  	struct ifnet *iter;
1084  	int found = 0;
1085 #ifdef VIMAGE
1086 	bool shutdown;
1087 
1088 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1089 #endif
1090 	IFNET_WLOCK();
1091 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1092 		if (iter == ifp) {
1093 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1094 			if (!vmove)
1095 				ifp->if_flags |= IFF_DYING;
1096 			found = 1;
1097 			break;
1098 		}
1099 	IFNET_WUNLOCK();
1100 	if (!found) {
1101 		/*
1102 		 * While we would want to panic here, we cannot
1103 		 * guarantee that the interface is indeed still on
1104 		 * the list given we don't hold locks all the way.
1105 		 */
1106 		return (ENOENT);
1107 #if 0
1108 		if (vmove)
1109 			panic("%s: ifp=%p not on the ifnet tailq %p",
1110 			    __func__, ifp, &V_ifnet);
1111 		else
1112 			return; /* XXX this should panic as well? */
1113 #endif
1114 	}
1115 
1116 	/*
1117 	 * At this point we know the interface still was on the ifnet list
1118 	 * and we removed it so we are in a stable state.
1119 	 */
1120 #ifdef VIMAGE
1121 	curvnet->vnet_ifcnt--;
1122 #endif
1123 	epoch_wait_preempt(net_epoch_preempt);
1124 
1125 	/*
1126 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1127 	 * fixes issues about late destruction of multicast options
1128 	 * which lead to leave group calls, which in turn access the
1129 	 * belonging ifnet structure:
1130 	 */
1131 	epoch_drain_callbacks(net_epoch_preempt);
1132 
1133 	/*
1134 	 * In any case (destroy or vmove) detach us from the groups
1135 	 * and remove/wait for pending events on the taskq.
1136 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1137 	 */
1138 	if_delgroups(ifp);
1139 
1140 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1141 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1142 
1143 	/*
1144 	 * Check if this is a cloned interface or not. Must do even if
1145 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1146 	 * the if_clone_addgroup() will have a corrupted string overwise
1147 	 * from a gibberish pointer.
1148 	 */
1149 	if (vmove && ifcp != NULL)
1150 		*ifcp = if_clone_findifc(ifp);
1151 
1152 	if_down(ifp);
1153 
1154 #ifdef VIMAGE
1155 	/*
1156 	 * On VNET shutdown abort here as the stack teardown will do all
1157 	 * the work top-down for us.
1158 	 */
1159 	if (shutdown) {
1160 		/* Give interface users the chance to clean up. */
1161 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1162 
1163 		/*
1164 		 * In case of a vmove we are done here without error.
1165 		 * If we would signal an error it would lead to the same
1166 		 * abort as if we did not find the ifnet anymore.
1167 		 * if_detach() calls us in void context and does not care
1168 		 * about an early abort notification, so life is splendid :)
1169 		 */
1170 		goto finish_vnet_shutdown;
1171 	}
1172 #endif
1173 
1174 	/*
1175 	 * At this point we are not tearing down a VNET and are either
1176 	 * going to destroy or vmove the interface and have to cleanup
1177 	 * accordingly.
1178 	 */
1179 
1180 	/*
1181 	 * Remove routes and flush queues.
1182 	 */
1183 #ifdef ALTQ
1184 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1185 		altq_disable(&ifp->if_snd);
1186 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1187 		altq_detach(&ifp->if_snd);
1188 #endif
1189 
1190 	if_purgeaddrs(ifp);
1191 
1192 #ifdef INET
1193 	in_ifdetach(ifp);
1194 #endif
1195 
1196 #ifdef INET6
1197 	/*
1198 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1199 	 * before removing routing entries below, since IPv6 interface direct
1200 	 * routes are expected to be removed by the IPv6-specific kernel API.
1201 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1202 	 */
1203 	in6_ifdetach(ifp);
1204 #endif
1205 	if_purgemaddrs(ifp);
1206 
1207 	/* Announce that the interface is gone. */
1208 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1209 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1210 	if (IS_DEFAULT_VNET(curvnet))
1211 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1212 
1213 	if (!vmove) {
1214 		/*
1215 		 * Prevent further calls into the device driver via ifnet.
1216 		 */
1217 		if_dead(ifp);
1218 
1219 		/*
1220 		 * Clean up all addresses.
1221 		 */
1222 		IF_ADDR_WLOCK(ifp);
1223 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1224 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1225 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1226 			IF_ADDR_WUNLOCK(ifp);
1227 			ifa_free(ifa);
1228 		} else
1229 			IF_ADDR_WUNLOCK(ifp);
1230 	}
1231 
1232 	rt_flushifroutes(ifp);
1233 
1234 #ifdef VIMAGE
1235 finish_vnet_shutdown:
1236 #endif
1237 	/*
1238 	 * We cannot hold the lock over dom_ifdetach calls as they might
1239 	 * sleep, for example trying to drain a callout, thus open up the
1240 	 * theoretical race with re-attaching.
1241 	 */
1242 	IF_AFDATA_LOCK(ifp);
1243 	i = ifp->if_afdata_initialized;
1244 	ifp->if_afdata_initialized = 0;
1245 	IF_AFDATA_UNLOCK(ifp);
1246 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1247 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1248 			(*dp->dom_ifdetach)(ifp,
1249 			    ifp->if_afdata[dp->dom_family]);
1250 			ifp->if_afdata[dp->dom_family] = NULL;
1251 		}
1252 	}
1253 
1254 	return (0);
1255 }
1256 
1257 #ifdef VIMAGE
1258 /*
1259  * if_vmove() performs a limited version of if_detach() in current
1260  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1261  * An attempt is made to shrink if_index in current vnet, find an
1262  * unused if_index in target vnet and calls if_grow() if necessary,
1263  * and finally find an unused if_xname for the target vnet.
1264  */
1265 static int
1266 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1267 {
1268 	struct if_clone *ifc;
1269 #ifdef DEV_BPF
1270 	u_int bif_dlt, bif_hdrlen;
1271 #endif
1272 	void *old;
1273 	int rc;
1274 
1275 #ifdef DEV_BPF
1276  	/*
1277 	 * if_detach_internal() will call the eventhandler to notify
1278 	 * interface departure.  That will detach if_bpf.  We need to
1279 	 * safe the dlt and hdrlen so we can re-attach it later.
1280 	 */
1281 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1282 #endif
1283 
1284 	/*
1285 	 * Detach from current vnet, but preserve LLADDR info, do not
1286 	 * mark as dead etc. so that the ifnet can be reattached later.
1287 	 * If we cannot find it, we lost the race to someone else.
1288 	 */
1289 	rc = if_detach_internal(ifp, 1, &ifc);
1290 	if (rc != 0)
1291 		return (rc);
1292 
1293 	/*
1294 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1295 	 * the if_index for that vnet if possible.
1296 	 *
1297 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1298 	 * or we'd lock on one vnet and unlock on another.
1299 	 */
1300 	IFNET_WLOCK();
1301 	ifindex_free_locked(ifp->if_index);
1302 	IFNET_WUNLOCK();
1303 
1304 	/*
1305 	 * Perform interface-specific reassignment tasks, if provided by
1306 	 * the driver.
1307 	 */
1308 	if (ifp->if_reassign != NULL)
1309 		ifp->if_reassign(ifp, new_vnet, NULL);
1310 
1311 	/*
1312 	 * Switch to the context of the target vnet.
1313 	 */
1314 	CURVNET_SET_QUIET(new_vnet);
1315  restart:
1316 	IFNET_WLOCK();
1317 	ifp->if_index = ifindex_alloc(&old);
1318 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1319 		IFNET_WUNLOCK();
1320 		epoch_wait_preempt(net_epoch_preempt);
1321 		free(old, M_IFNET);
1322 		goto restart;
1323 	}
1324 	ifnet_setbyindex(ifp->if_index, ifp);
1325 	IFNET_WUNLOCK();
1326 
1327 	if_attach_internal(ifp, 1, ifc);
1328 
1329 #ifdef DEV_BPF
1330 	if (ifp->if_bpf == NULL)
1331 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1332 #endif
1333 
1334 	CURVNET_RESTORE();
1335 	return (0);
1336 }
1337 
1338 /*
1339  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1340  */
1341 static int
1342 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1343 {
1344 	struct prison *pr;
1345 	struct ifnet *difp;
1346 	int error;
1347 	bool shutdown;
1348 
1349 	/* Try to find the prison within our visibility. */
1350 	sx_slock(&allprison_lock);
1351 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1352 	sx_sunlock(&allprison_lock);
1353 	if (pr == NULL)
1354 		return (ENXIO);
1355 	prison_hold_locked(pr);
1356 	mtx_unlock(&pr->pr_mtx);
1357 
1358 	/* Do not try to move the iface from and to the same prison. */
1359 	if (pr->pr_vnet == ifp->if_vnet) {
1360 		prison_free(pr);
1361 		return (EEXIST);
1362 	}
1363 
1364 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1365 	/* XXX Lock interfaces to avoid races. */
1366 	CURVNET_SET_QUIET(pr->pr_vnet);
1367 	difp = ifunit(ifname);
1368 	if (difp != NULL) {
1369 		CURVNET_RESTORE();
1370 		prison_free(pr);
1371 		return (EEXIST);
1372 	}
1373 
1374 	/* Make sure the VNET is stable. */
1375 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1376 	if (shutdown) {
1377 		CURVNET_RESTORE();
1378 		prison_free(pr);
1379 		return (EBUSY);
1380 	}
1381 	CURVNET_RESTORE();
1382 
1383 	/* Move the interface into the child jail/vnet. */
1384 	error = if_vmove(ifp, pr->pr_vnet);
1385 
1386 	/* Report the new if_xname back to the userland on success. */
1387 	if (error == 0)
1388 		sprintf(ifname, "%s", ifp->if_xname);
1389 
1390 	prison_free(pr);
1391 	return (error);
1392 }
1393 
1394 static int
1395 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1396 {
1397 	struct prison *pr;
1398 	struct vnet *vnet_dst;
1399 	struct ifnet *ifp;
1400 	int error;
1401  	bool shutdown;
1402 
1403 	/* Try to find the prison within our visibility. */
1404 	sx_slock(&allprison_lock);
1405 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1406 	sx_sunlock(&allprison_lock);
1407 	if (pr == NULL)
1408 		return (ENXIO);
1409 	prison_hold_locked(pr);
1410 	mtx_unlock(&pr->pr_mtx);
1411 
1412 	/* Make sure the named iface exists in the source prison/vnet. */
1413 	CURVNET_SET(pr->pr_vnet);
1414 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1415 	if (ifp == NULL) {
1416 		CURVNET_RESTORE();
1417 		prison_free(pr);
1418 		return (ENXIO);
1419 	}
1420 
1421 	/* Do not try to move the iface from and to the same prison. */
1422 	vnet_dst = TD_TO_VNET(td);
1423 	if (vnet_dst == ifp->if_vnet) {
1424 		CURVNET_RESTORE();
1425 		prison_free(pr);
1426 		return (EEXIST);
1427 	}
1428 
1429 	/* Make sure the VNET is stable. */
1430 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1431 	if (shutdown) {
1432 		CURVNET_RESTORE();
1433 		prison_free(pr);
1434 		return (EBUSY);
1435 	}
1436 
1437 	/* Get interface back from child jail/vnet. */
1438 	error = if_vmove(ifp, vnet_dst);
1439 	CURVNET_RESTORE();
1440 
1441 	/* Report the new if_xname back to the userland on success. */
1442 	if (error == 0)
1443 		sprintf(ifname, "%s", ifp->if_xname);
1444 
1445 	prison_free(pr);
1446 	return (error);
1447 }
1448 #endif /* VIMAGE */
1449 
1450 /*
1451  * Add a group to an interface
1452  */
1453 int
1454 if_addgroup(struct ifnet *ifp, const char *groupname)
1455 {
1456 	struct ifg_list		*ifgl;
1457 	struct ifg_group	*ifg = NULL;
1458 	struct ifg_member	*ifgm;
1459 	int 			 new = 0;
1460 
1461 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1462 	    groupname[strlen(groupname) - 1] <= '9')
1463 		return (EINVAL);
1464 
1465 	IFNET_WLOCK();
1466 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1467 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1468 			IFNET_WUNLOCK();
1469 			return (EEXIST);
1470 		}
1471 
1472 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1473 	    	IFNET_WUNLOCK();
1474 		return (ENOMEM);
1475 	}
1476 
1477 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1478 		free(ifgl, M_TEMP);
1479 		IFNET_WUNLOCK();
1480 		return (ENOMEM);
1481 	}
1482 
1483 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1484 		if (!strcmp(ifg->ifg_group, groupname))
1485 			break;
1486 
1487 	if (ifg == NULL) {
1488 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1489 			free(ifgl, M_TEMP);
1490 			free(ifgm, M_TEMP);
1491 			IFNET_WUNLOCK();
1492 			return (ENOMEM);
1493 		}
1494 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1495 		ifg->ifg_refcnt = 0;
1496 		CK_STAILQ_INIT(&ifg->ifg_members);
1497 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1498 		new = 1;
1499 	}
1500 
1501 	ifg->ifg_refcnt++;
1502 	ifgl->ifgl_group = ifg;
1503 	ifgm->ifgm_ifp = ifp;
1504 
1505 	IF_ADDR_WLOCK(ifp);
1506 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1507 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1508 	IF_ADDR_WUNLOCK(ifp);
1509 
1510 	IFNET_WUNLOCK();
1511 
1512 	if (new)
1513 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1514 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1515 
1516 	return (0);
1517 }
1518 
1519 /*
1520  * Helper function to remove a group out of an interface.  Expects the global
1521  * ifnet lock to be write-locked, and drops it before returning.
1522  */
1523 static void
1524 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1525     const char *groupname)
1526 {
1527 	struct ifg_member *ifgm;
1528 	bool freeifgl;
1529 
1530 	IFNET_WLOCK_ASSERT();
1531 
1532 	IF_ADDR_WLOCK(ifp);
1533 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1534 	IF_ADDR_WUNLOCK(ifp);
1535 
1536 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1537 		if (ifgm->ifgm_ifp == ifp) {
1538 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1539 			    ifg_member, ifgm_next);
1540 			break;
1541 		}
1542 	}
1543 
1544 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1545 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1546 		    ifg_next);
1547 		freeifgl = true;
1548 	} else {
1549 		freeifgl = false;
1550 	}
1551 	IFNET_WUNLOCK();
1552 
1553 	epoch_wait_preempt(net_epoch_preempt);
1554 	if (freeifgl) {
1555 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1556 		free(ifgl->ifgl_group, M_TEMP);
1557 	}
1558 	free(ifgm, M_TEMP);
1559 	free(ifgl, M_TEMP);
1560 
1561 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1562 }
1563 
1564 /*
1565  * Remove a group from an interface
1566  */
1567 int
1568 if_delgroup(struct ifnet *ifp, const char *groupname)
1569 {
1570 	struct ifg_list *ifgl;
1571 
1572 	IFNET_WLOCK();
1573 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1574 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1575 			break;
1576 	if (ifgl == NULL) {
1577 		IFNET_WUNLOCK();
1578 		return (ENOENT);
1579 	}
1580 
1581 	_if_delgroup_locked(ifp, ifgl, groupname);
1582 
1583 	return (0);
1584 }
1585 
1586 /*
1587  * Remove an interface from all groups
1588  */
1589 static void
1590 if_delgroups(struct ifnet *ifp)
1591 {
1592 	struct ifg_list *ifgl;
1593 	char groupname[IFNAMSIZ];
1594 
1595 	IFNET_WLOCK();
1596 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1597 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1598 		_if_delgroup_locked(ifp, ifgl, groupname);
1599 		IFNET_WLOCK();
1600 	}
1601 	IFNET_WUNLOCK();
1602 }
1603 
1604 static char *
1605 ifgr_group_get(void *ifgrp)
1606 {
1607 	union ifgroupreq_union *ifgrup;
1608 
1609 	ifgrup = ifgrp;
1610 #ifdef COMPAT_FREEBSD32
1611 	if (SV_CURPROC_FLAG(SV_ILP32))
1612 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1613 #endif
1614 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1615 }
1616 
1617 static struct ifg_req *
1618 ifgr_groups_get(void *ifgrp)
1619 {
1620 	union ifgroupreq_union *ifgrup;
1621 
1622 	ifgrup = ifgrp;
1623 #ifdef COMPAT_FREEBSD32
1624 	if (SV_CURPROC_FLAG(SV_ILP32))
1625 		return ((struct ifg_req *)(uintptr_t)
1626 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1627 #endif
1628 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1629 }
1630 
1631 /*
1632  * Stores all groups from an interface in memory pointed to by ifgr.
1633  */
1634 static int
1635 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1636 {
1637 	int			 len, error;
1638 	struct ifg_list		*ifgl;
1639 	struct ifg_req		 ifgrq, *ifgp;
1640 
1641 	NET_EPOCH_ASSERT();
1642 
1643 	if (ifgr->ifgr_len == 0) {
1644 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1645 			ifgr->ifgr_len += sizeof(struct ifg_req);
1646 		return (0);
1647 	}
1648 
1649 	len = ifgr->ifgr_len;
1650 	ifgp = ifgr_groups_get(ifgr);
1651 	/* XXX: wire */
1652 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1653 		if (len < sizeof(ifgrq))
1654 			return (EINVAL);
1655 		bzero(&ifgrq, sizeof ifgrq);
1656 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1657 		    sizeof(ifgrq.ifgrq_group));
1658 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1659 			return (error);
1660 		len -= sizeof(ifgrq);
1661 		ifgp++;
1662 	}
1663 
1664 	return (0);
1665 }
1666 
1667 /*
1668  * Stores all members of a group in memory pointed to by igfr
1669  */
1670 static int
1671 if_getgroupmembers(struct ifgroupreq *ifgr)
1672 {
1673 	struct ifg_group	*ifg;
1674 	struct ifg_member	*ifgm;
1675 	struct ifg_req		 ifgrq, *ifgp;
1676 	int			 len, error;
1677 
1678 	IFNET_RLOCK();
1679 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1680 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1681 			break;
1682 	if (ifg == NULL) {
1683 		IFNET_RUNLOCK();
1684 		return (ENOENT);
1685 	}
1686 
1687 	if (ifgr->ifgr_len == 0) {
1688 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1689 			ifgr->ifgr_len += sizeof(ifgrq);
1690 		IFNET_RUNLOCK();
1691 		return (0);
1692 	}
1693 
1694 	len = ifgr->ifgr_len;
1695 	ifgp = ifgr_groups_get(ifgr);
1696 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1697 		if (len < sizeof(ifgrq)) {
1698 			IFNET_RUNLOCK();
1699 			return (EINVAL);
1700 		}
1701 		bzero(&ifgrq, sizeof ifgrq);
1702 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1703 		    sizeof(ifgrq.ifgrq_member));
1704 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1705 			IFNET_RUNLOCK();
1706 			return (error);
1707 		}
1708 		len -= sizeof(ifgrq);
1709 		ifgp++;
1710 	}
1711 	IFNET_RUNLOCK();
1712 
1713 	return (0);
1714 }
1715 
1716 /*
1717  * Return counter values from counter(9)s stored in ifnet.
1718  */
1719 uint64_t
1720 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1721 {
1722 
1723 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1724 
1725 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1726 }
1727 
1728 /*
1729  * Increase an ifnet counter. Usually used for counters shared
1730  * between the stack and a driver, but function supports them all.
1731  */
1732 void
1733 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1734 {
1735 
1736 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1737 
1738 	counter_u64_add(ifp->if_counters[cnt], inc);
1739 }
1740 
1741 /*
1742  * Copy data from ifnet to userland API structure if_data.
1743  */
1744 void
1745 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1746 {
1747 
1748 	ifd->ifi_type = ifp->if_type;
1749 	ifd->ifi_physical = 0;
1750 	ifd->ifi_addrlen = ifp->if_addrlen;
1751 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1752 	ifd->ifi_link_state = ifp->if_link_state;
1753 	ifd->ifi_vhid = 0;
1754 	ifd->ifi_datalen = sizeof(struct if_data);
1755 	ifd->ifi_mtu = ifp->if_mtu;
1756 	ifd->ifi_metric = ifp->if_metric;
1757 	ifd->ifi_baudrate = ifp->if_baudrate;
1758 	ifd->ifi_hwassist = ifp->if_hwassist;
1759 	ifd->ifi_epoch = ifp->if_epoch;
1760 	ifd->ifi_lastchange = ifp->if_lastchange;
1761 
1762 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1763 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1764 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1765 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1766 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1767 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1768 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1769 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1770 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1771 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1772 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1773 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1774 }
1775 
1776 /*
1777  * Initialization, destruction and refcounting functions for ifaddrs.
1778  */
1779 struct ifaddr *
1780 ifa_alloc(size_t size, int flags)
1781 {
1782 	struct ifaddr *ifa;
1783 
1784 	KASSERT(size >= sizeof(struct ifaddr),
1785 	    ("%s: invalid size %zu", __func__, size));
1786 
1787 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1788 	if (ifa == NULL)
1789 		return (NULL);
1790 
1791 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1792 		goto fail;
1793 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1794 		goto fail;
1795 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1796 		goto fail;
1797 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1798 		goto fail;
1799 
1800 	refcount_init(&ifa->ifa_refcnt, 1);
1801 
1802 	return (ifa);
1803 
1804 fail:
1805 	/* free(NULL) is okay */
1806 	counter_u64_free(ifa->ifa_opackets);
1807 	counter_u64_free(ifa->ifa_ipackets);
1808 	counter_u64_free(ifa->ifa_obytes);
1809 	counter_u64_free(ifa->ifa_ibytes);
1810 	free(ifa, M_IFADDR);
1811 
1812 	return (NULL);
1813 }
1814 
1815 void
1816 ifa_ref(struct ifaddr *ifa)
1817 {
1818 
1819 	refcount_acquire(&ifa->ifa_refcnt);
1820 }
1821 
1822 static void
1823 ifa_destroy(epoch_context_t ctx)
1824 {
1825 	struct ifaddr *ifa;
1826 
1827 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1828 	counter_u64_free(ifa->ifa_opackets);
1829 	counter_u64_free(ifa->ifa_ipackets);
1830 	counter_u64_free(ifa->ifa_obytes);
1831 	counter_u64_free(ifa->ifa_ibytes);
1832 	free(ifa, M_IFADDR);
1833 }
1834 
1835 void
1836 ifa_free(struct ifaddr *ifa)
1837 {
1838 
1839 	if (refcount_release(&ifa->ifa_refcnt))
1840 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1841 }
1842 
1843 
1844 static int
1845 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1846     struct sockaddr *ia)
1847 {
1848 	struct epoch_tracker et;
1849 	int error;
1850 	struct rt_addrinfo info;
1851 	struct sockaddr_dl null_sdl;
1852 	struct ifnet *ifp;
1853 	struct ifaddr *rti_ifa = NULL;
1854 
1855 	ifp = ifa->ifa_ifp;
1856 
1857 	bzero(&info, sizeof(info));
1858 	if (cmd != RTM_DELETE)
1859 		info.rti_ifp = V_loif;
1860 	if (cmd == RTM_ADD) {
1861 		/* explicitly specify (loopback) ifa */
1862 		if (info.rti_ifp != NULL) {
1863 			NET_EPOCH_ENTER(et);
1864 			rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1865 			if (rti_ifa != NULL)
1866 				ifa_ref(rti_ifa);
1867 			info.rti_ifa = rti_ifa;
1868 			NET_EPOCH_EXIT(et);
1869 		}
1870 	}
1871 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1872 	info.rti_info[RTAX_DST] = ia;
1873 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1874 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1875 
1876 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1877 
1878 	if (rti_ifa != NULL)
1879 		ifa_free(rti_ifa);
1880 
1881 	if (error == 0 ||
1882 	    (cmd == RTM_ADD && error == EEXIST) ||
1883 	    (cmd == RTM_DELETE && (error == ENOENT || error == ESRCH)))
1884 		return (error);
1885 
1886 	log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1887 		__func__, otype, if_name(ifp), error);
1888 
1889 	return (error);
1890 }
1891 
1892 int
1893 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1894 {
1895 
1896 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1897 }
1898 
1899 int
1900 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1901 {
1902 
1903 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1904 }
1905 
1906 int
1907 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1908 {
1909 
1910 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1911 }
1912 
1913 /*
1914  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1915  * structs used to represent other address families, it is necessary
1916  * to perform a different comparison.
1917  */
1918 
1919 #define	sa_dl_equal(a1, a2)	\
1920 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1921 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1922 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1923 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1924 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1925 
1926 /*
1927  * Locate an interface based on a complete address.
1928  */
1929 /*ARGSUSED*/
1930 struct ifaddr *
1931 ifa_ifwithaddr(const struct sockaddr *addr)
1932 {
1933 	struct ifnet *ifp;
1934 	struct ifaddr *ifa;
1935 
1936 	NET_EPOCH_ASSERT();
1937 
1938 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1939 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1940 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1941 				continue;
1942 			if (sa_equal(addr, ifa->ifa_addr)) {
1943 				goto done;
1944 			}
1945 			/* IP6 doesn't have broadcast */
1946 			if ((ifp->if_flags & IFF_BROADCAST) &&
1947 			    ifa->ifa_broadaddr &&
1948 			    ifa->ifa_broadaddr->sa_len != 0 &&
1949 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1950 				goto done;
1951 			}
1952 		}
1953 	}
1954 	ifa = NULL;
1955 done:
1956 	return (ifa);
1957 }
1958 
1959 int
1960 ifa_ifwithaddr_check(const struct sockaddr *addr)
1961 {
1962 	struct epoch_tracker et;
1963 	int rc;
1964 
1965 	NET_EPOCH_ENTER(et);
1966 	rc = (ifa_ifwithaddr(addr) != NULL);
1967 	NET_EPOCH_EXIT(et);
1968 	return (rc);
1969 }
1970 
1971 /*
1972  * Locate an interface based on the broadcast address.
1973  */
1974 /* ARGSUSED */
1975 struct ifaddr *
1976 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1977 {
1978 	struct ifnet *ifp;
1979 	struct ifaddr *ifa;
1980 
1981 	NET_EPOCH_ASSERT();
1982 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1983 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1984 			continue;
1985 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1986 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1987 				continue;
1988 			if ((ifp->if_flags & IFF_BROADCAST) &&
1989 			    ifa->ifa_broadaddr &&
1990 			    ifa->ifa_broadaddr->sa_len != 0 &&
1991 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1992 				goto done;
1993 			}
1994 		}
1995 	}
1996 	ifa = NULL;
1997 done:
1998 	return (ifa);
1999 }
2000 
2001 /*
2002  * Locate the point to point interface with a given destination address.
2003  */
2004 /*ARGSUSED*/
2005 struct ifaddr *
2006 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2007 {
2008 	struct ifnet *ifp;
2009 	struct ifaddr *ifa;
2010 
2011 	NET_EPOCH_ASSERT();
2012 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2013 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2014 			continue;
2015 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2016 			continue;
2017 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2018 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2019 				continue;
2020 			if (ifa->ifa_dstaddr != NULL &&
2021 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2022 				goto done;
2023 			}
2024 		}
2025 	}
2026 	ifa = NULL;
2027 done:
2028 	return (ifa);
2029 }
2030 
2031 /*
2032  * Find an interface on a specific network.  If many, choice
2033  * is most specific found.
2034  */
2035 struct ifaddr *
2036 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2037 {
2038 	struct ifnet *ifp;
2039 	struct ifaddr *ifa;
2040 	struct ifaddr *ifa_maybe = NULL;
2041 	u_int af = addr->sa_family;
2042 	const char *addr_data = addr->sa_data, *cplim;
2043 
2044 	NET_EPOCH_ASSERT();
2045 	/*
2046 	 * AF_LINK addresses can be looked up directly by their index number,
2047 	 * so do that if we can.
2048 	 */
2049 	if (af == AF_LINK) {
2050 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2051 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2052 		return (ifaddr_byindex(sdl->sdl_index));
2053 	}
2054 
2055 	/*
2056 	 * Scan though each interface, looking for ones that have addresses
2057 	 * in this address family and the requested fib.
2058 	 */
2059 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2060 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2061 			continue;
2062 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2063 			const char *cp, *cp2, *cp3;
2064 
2065 			if (ifa->ifa_addr->sa_family != af)
2066 next:				continue;
2067 			if (af == AF_INET &&
2068 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2069 				/*
2070 				 * This is a bit broken as it doesn't
2071 				 * take into account that the remote end may
2072 				 * be a single node in the network we are
2073 				 * looking for.
2074 				 * The trouble is that we don't know the
2075 				 * netmask for the remote end.
2076 				 */
2077 				if (ifa->ifa_dstaddr != NULL &&
2078 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2079 					goto done;
2080 				}
2081 			} else {
2082 				/*
2083 				 * Scan all the bits in the ifa's address.
2084 				 * If a bit dissagrees with what we are
2085 				 * looking for, mask it with the netmask
2086 				 * to see if it really matters.
2087 				 * (A byte at a time)
2088 				 */
2089 				if (ifa->ifa_netmask == 0)
2090 					continue;
2091 				cp = addr_data;
2092 				cp2 = ifa->ifa_addr->sa_data;
2093 				cp3 = ifa->ifa_netmask->sa_data;
2094 				cplim = ifa->ifa_netmask->sa_len
2095 					+ (char *)ifa->ifa_netmask;
2096 				while (cp3 < cplim)
2097 					if ((*cp++ ^ *cp2++) & *cp3++)
2098 						goto next; /* next address! */
2099 				/*
2100 				 * If the netmask of what we just found
2101 				 * is more specific than what we had before
2102 				 * (if we had one), or if the virtual status
2103 				 * of new prefix is better than of the old one,
2104 				 * then remember the new one before continuing
2105 				 * to search for an even better one.
2106 				 */
2107 				if (ifa_maybe == NULL ||
2108 				    ifa_preferred(ifa_maybe, ifa) ||
2109 				    rn_refines((caddr_t)ifa->ifa_netmask,
2110 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2111 					ifa_maybe = ifa;
2112 				}
2113 			}
2114 		}
2115 	}
2116 	ifa = ifa_maybe;
2117 	ifa_maybe = NULL;
2118 done:
2119 	return (ifa);
2120 }
2121 
2122 /*
2123  * Find an interface address specific to an interface best matching
2124  * a given address.
2125  */
2126 struct ifaddr *
2127 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2128 {
2129 	struct ifaddr *ifa;
2130 	const char *cp, *cp2, *cp3;
2131 	char *cplim;
2132 	struct ifaddr *ifa_maybe = NULL;
2133 	u_int af = addr->sa_family;
2134 
2135 	if (af >= AF_MAX)
2136 		return (NULL);
2137 
2138 	NET_EPOCH_ASSERT();
2139 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2140 		if (ifa->ifa_addr->sa_family != af)
2141 			continue;
2142 		if (ifa_maybe == NULL)
2143 			ifa_maybe = ifa;
2144 		if (ifa->ifa_netmask == 0) {
2145 			if (sa_equal(addr, ifa->ifa_addr) ||
2146 			    (ifa->ifa_dstaddr &&
2147 			    sa_equal(addr, ifa->ifa_dstaddr)))
2148 				goto done;
2149 			continue;
2150 		}
2151 		if (ifp->if_flags & IFF_POINTOPOINT) {
2152 			if (sa_equal(addr, ifa->ifa_dstaddr))
2153 				goto done;
2154 		} else {
2155 			cp = addr->sa_data;
2156 			cp2 = ifa->ifa_addr->sa_data;
2157 			cp3 = ifa->ifa_netmask->sa_data;
2158 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2159 			for (; cp3 < cplim; cp3++)
2160 				if ((*cp++ ^ *cp2++) & *cp3)
2161 					break;
2162 			if (cp3 == cplim)
2163 				goto done;
2164 		}
2165 	}
2166 	ifa = ifa_maybe;
2167 done:
2168 	return (ifa);
2169 }
2170 
2171 /*
2172  * See whether new ifa is better than current one:
2173  * 1) A non-virtual one is preferred over virtual.
2174  * 2) A virtual in master state preferred over any other state.
2175  *
2176  * Used in several address selecting functions.
2177  */
2178 int
2179 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2180 {
2181 
2182 	return (cur->ifa_carp && (!next->ifa_carp ||
2183 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2184 }
2185 
2186 struct sockaddr_dl *
2187 link_alloc_sdl(size_t size, int flags)
2188 {
2189 
2190 	return (malloc(size, M_TEMP, flags));
2191 }
2192 
2193 void
2194 link_free_sdl(struct sockaddr *sa)
2195 {
2196 	free(sa, M_TEMP);
2197 }
2198 
2199 /*
2200  * Fills in given sdl with interface basic info.
2201  * Returns pointer to filled sdl.
2202  */
2203 struct sockaddr_dl *
2204 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2205 {
2206 	struct sockaddr_dl *sdl;
2207 
2208 	sdl = (struct sockaddr_dl *)paddr;
2209 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2210 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2211 	sdl->sdl_family = AF_LINK;
2212 	sdl->sdl_index = ifp->if_index;
2213 	sdl->sdl_type = iftype;
2214 
2215 	return (sdl);
2216 }
2217 
2218 /*
2219  * Mark an interface down and notify protocols of
2220  * the transition.
2221  */
2222 static void
2223 if_unroute(struct ifnet *ifp, int flag, int fam)
2224 {
2225 	struct ifaddr *ifa;
2226 
2227 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2228 
2229 	ifp->if_flags &= ~flag;
2230 	getmicrotime(&ifp->if_lastchange);
2231 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2232 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2233 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2234 	ifp->if_qflush(ifp);
2235 
2236 	if (ifp->if_carp)
2237 		(*carp_linkstate_p)(ifp);
2238 	rt_ifmsg(ifp);
2239 }
2240 
2241 /*
2242  * Mark an interface up and notify protocols of
2243  * the transition.
2244  */
2245 static void
2246 if_route(struct ifnet *ifp, int flag, int fam)
2247 {
2248 	struct ifaddr *ifa;
2249 
2250 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2251 
2252 	ifp->if_flags |= flag;
2253 	getmicrotime(&ifp->if_lastchange);
2254 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2255 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2256 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2257 	if (ifp->if_carp)
2258 		(*carp_linkstate_p)(ifp);
2259 	rt_ifmsg(ifp);
2260 #ifdef INET6
2261 	in6_if_up(ifp);
2262 #endif
2263 }
2264 
2265 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2266 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2267 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2268 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2269 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2270 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2271 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2272 void	*(*vlan_cookie_p)(struct ifnet *);
2273 
2274 /*
2275  * Handle a change in the interface link state. To avoid LORs
2276  * between driver lock and upper layer locks, as well as possible
2277  * recursions, we post event to taskqueue, and all job
2278  * is done in static do_link_state_change().
2279  */
2280 void
2281 if_link_state_change(struct ifnet *ifp, int link_state)
2282 {
2283 	/* Return if state hasn't changed. */
2284 	if (ifp->if_link_state == link_state)
2285 		return;
2286 
2287 	ifp->if_link_state = link_state;
2288 
2289 	/* XXXGL: reference ifp? */
2290 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2291 }
2292 
2293 static void
2294 do_link_state_change(void *arg, int pending)
2295 {
2296 	struct ifnet *ifp;
2297 	int link_state;
2298 
2299 	ifp = arg;
2300 	link_state = ifp->if_link_state;
2301 
2302 	CURVNET_SET(ifp->if_vnet);
2303 	rt_ifmsg(ifp);
2304 	if (ifp->if_vlantrunk != NULL)
2305 		(*vlan_link_state_p)(ifp);
2306 
2307 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2308 	    ifp->if_l2com != NULL)
2309 		(*ng_ether_link_state_p)(ifp, link_state);
2310 	if (ifp->if_carp)
2311 		(*carp_linkstate_p)(ifp);
2312 	if (ifp->if_bridge)
2313 		ifp->if_bridge_linkstate(ifp);
2314 	if (ifp->if_lagg)
2315 		(*lagg_linkstate_p)(ifp, link_state);
2316 
2317 	if (IS_DEFAULT_VNET(curvnet))
2318 		devctl_notify("IFNET", ifp->if_xname,
2319 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2320 		    NULL);
2321 	if (pending > 1)
2322 		if_printf(ifp, "%d link states coalesced\n", pending);
2323 	if (log_link_state_change)
2324 		if_printf(ifp, "link state changed to %s\n",
2325 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2326 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2327 	CURVNET_RESTORE();
2328 }
2329 
2330 /*
2331  * Mark an interface down and notify protocols of
2332  * the transition.
2333  */
2334 void
2335 if_down(struct ifnet *ifp)
2336 {
2337 
2338 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2339 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2340 }
2341 
2342 /*
2343  * Mark an interface up and notify protocols of
2344  * the transition.
2345  */
2346 void
2347 if_up(struct ifnet *ifp)
2348 {
2349 
2350 	if_route(ifp, IFF_UP, AF_UNSPEC);
2351 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2352 }
2353 
2354 /*
2355  * Flush an interface queue.
2356  */
2357 void
2358 if_qflush(struct ifnet *ifp)
2359 {
2360 	struct mbuf *m, *n;
2361 	struct ifaltq *ifq;
2362 
2363 	ifq = &ifp->if_snd;
2364 	IFQ_LOCK(ifq);
2365 #ifdef ALTQ
2366 	if (ALTQ_IS_ENABLED(ifq))
2367 		ALTQ_PURGE(ifq);
2368 #endif
2369 	n = ifq->ifq_head;
2370 	while ((m = n) != NULL) {
2371 		n = m->m_nextpkt;
2372 		m_freem(m);
2373 	}
2374 	ifq->ifq_head = 0;
2375 	ifq->ifq_tail = 0;
2376 	ifq->ifq_len = 0;
2377 	IFQ_UNLOCK(ifq);
2378 }
2379 
2380 /*
2381  * Map interface name to interface structure pointer, with or without
2382  * returning a reference.
2383  */
2384 struct ifnet *
2385 ifunit_ref(const char *name)
2386 {
2387 	struct epoch_tracker et;
2388 	struct ifnet *ifp;
2389 
2390 	NET_EPOCH_ENTER(et);
2391 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2392 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2393 		    !(ifp->if_flags & IFF_DYING))
2394 			break;
2395 	}
2396 	if (ifp != NULL)
2397 		if_ref(ifp);
2398 	NET_EPOCH_EXIT(et);
2399 	return (ifp);
2400 }
2401 
2402 struct ifnet *
2403 ifunit(const char *name)
2404 {
2405 	struct epoch_tracker et;
2406 	struct ifnet *ifp;
2407 
2408 	NET_EPOCH_ENTER(et);
2409 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2410 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2411 			break;
2412 	}
2413 	NET_EPOCH_EXIT(et);
2414 	return (ifp);
2415 }
2416 
2417 void *
2418 ifr_buffer_get_buffer(void *data)
2419 {
2420 	union ifreq_union *ifrup;
2421 
2422 	ifrup = data;
2423 #ifdef COMPAT_FREEBSD32
2424 	if (SV_CURPROC_FLAG(SV_ILP32))
2425 		return ((void *)(uintptr_t)
2426 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2427 #endif
2428 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2429 }
2430 
2431 static void
2432 ifr_buffer_set_buffer_null(void *data)
2433 {
2434 	union ifreq_union *ifrup;
2435 
2436 	ifrup = data;
2437 #ifdef COMPAT_FREEBSD32
2438 	if (SV_CURPROC_FLAG(SV_ILP32))
2439 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2440 	else
2441 #endif
2442 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2443 }
2444 
2445 size_t
2446 ifr_buffer_get_length(void *data)
2447 {
2448 	union ifreq_union *ifrup;
2449 
2450 	ifrup = data;
2451 #ifdef COMPAT_FREEBSD32
2452 	if (SV_CURPROC_FLAG(SV_ILP32))
2453 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2454 #endif
2455 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2456 }
2457 
2458 static void
2459 ifr_buffer_set_length(void *data, size_t len)
2460 {
2461 	union ifreq_union *ifrup;
2462 
2463 	ifrup = data;
2464 #ifdef COMPAT_FREEBSD32
2465 	if (SV_CURPROC_FLAG(SV_ILP32))
2466 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2467 	else
2468 #endif
2469 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2470 }
2471 
2472 void *
2473 ifr_data_get_ptr(void *ifrp)
2474 {
2475 	union ifreq_union *ifrup;
2476 
2477 	ifrup = ifrp;
2478 #ifdef COMPAT_FREEBSD32
2479 	if (SV_CURPROC_FLAG(SV_ILP32))
2480 		return ((void *)(uintptr_t)
2481 		    ifrup->ifr32.ifr_ifru.ifru_data);
2482 #endif
2483 		return (ifrup->ifr.ifr_ifru.ifru_data);
2484 }
2485 
2486 /*
2487  * Hardware specific interface ioctls.
2488  */
2489 int
2490 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2491 {
2492 	struct ifreq *ifr;
2493 	int error = 0, do_ifup = 0;
2494 	int new_flags, temp_flags;
2495 	size_t namelen, onamelen;
2496 	size_t descrlen;
2497 	char *descrbuf, *odescrbuf;
2498 	char new_name[IFNAMSIZ];
2499 	struct ifaddr *ifa;
2500 	struct sockaddr_dl *sdl;
2501 
2502 	ifr = (struct ifreq *)data;
2503 	switch (cmd) {
2504 	case SIOCGIFINDEX:
2505 		ifr->ifr_index = ifp->if_index;
2506 		break;
2507 
2508 	case SIOCGIFFLAGS:
2509 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2510 		ifr->ifr_flags = temp_flags & 0xffff;
2511 		ifr->ifr_flagshigh = temp_flags >> 16;
2512 		break;
2513 
2514 	case SIOCGIFCAP:
2515 		ifr->ifr_reqcap = ifp->if_capabilities;
2516 		ifr->ifr_curcap = ifp->if_capenable;
2517 		break;
2518 
2519 #ifdef MAC
2520 	case SIOCGIFMAC:
2521 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2522 		break;
2523 #endif
2524 
2525 	case SIOCGIFMETRIC:
2526 		ifr->ifr_metric = ifp->if_metric;
2527 		break;
2528 
2529 	case SIOCGIFMTU:
2530 		ifr->ifr_mtu = ifp->if_mtu;
2531 		break;
2532 
2533 	case SIOCGIFPHYS:
2534 		/* XXXGL: did this ever worked? */
2535 		ifr->ifr_phys = 0;
2536 		break;
2537 
2538 	case SIOCGIFDESCR:
2539 		error = 0;
2540 		sx_slock(&ifdescr_sx);
2541 		if (ifp->if_description == NULL)
2542 			error = ENOMSG;
2543 		else {
2544 			/* space for terminating nul */
2545 			descrlen = strlen(ifp->if_description) + 1;
2546 			if (ifr_buffer_get_length(ifr) < descrlen)
2547 				ifr_buffer_set_buffer_null(ifr);
2548 			else
2549 				error = copyout(ifp->if_description,
2550 				    ifr_buffer_get_buffer(ifr), descrlen);
2551 			ifr_buffer_set_length(ifr, descrlen);
2552 		}
2553 		sx_sunlock(&ifdescr_sx);
2554 		break;
2555 
2556 	case SIOCSIFDESCR:
2557 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2558 		if (error)
2559 			return (error);
2560 
2561 		/*
2562 		 * Copy only (length-1) bytes to make sure that
2563 		 * if_description is always nul terminated.  The
2564 		 * length parameter is supposed to count the
2565 		 * terminating nul in.
2566 		 */
2567 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2568 			return (ENAMETOOLONG);
2569 		else if (ifr_buffer_get_length(ifr) == 0)
2570 			descrbuf = NULL;
2571 		else {
2572 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2573 			    M_IFDESCR, M_WAITOK | M_ZERO);
2574 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2575 			    ifr_buffer_get_length(ifr) - 1);
2576 			if (error) {
2577 				free(descrbuf, M_IFDESCR);
2578 				break;
2579 			}
2580 		}
2581 
2582 		sx_xlock(&ifdescr_sx);
2583 		odescrbuf = ifp->if_description;
2584 		ifp->if_description = descrbuf;
2585 		sx_xunlock(&ifdescr_sx);
2586 
2587 		getmicrotime(&ifp->if_lastchange);
2588 		free(odescrbuf, M_IFDESCR);
2589 		break;
2590 
2591 	case SIOCGIFFIB:
2592 		ifr->ifr_fib = ifp->if_fib;
2593 		break;
2594 
2595 	case SIOCSIFFIB:
2596 		error = priv_check(td, PRIV_NET_SETIFFIB);
2597 		if (error)
2598 			return (error);
2599 		if (ifr->ifr_fib >= rt_numfibs)
2600 			return (EINVAL);
2601 
2602 		ifp->if_fib = ifr->ifr_fib;
2603 		break;
2604 
2605 	case SIOCSIFFLAGS:
2606 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2607 		if (error)
2608 			return (error);
2609 		/*
2610 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2611 		 * check, so we don't need special handling here yet.
2612 		 */
2613 		new_flags = (ifr->ifr_flags & 0xffff) |
2614 		    (ifr->ifr_flagshigh << 16);
2615 		if (ifp->if_flags & IFF_UP &&
2616 		    (new_flags & IFF_UP) == 0) {
2617 			if_down(ifp);
2618 		} else if (new_flags & IFF_UP &&
2619 		    (ifp->if_flags & IFF_UP) == 0) {
2620 			do_ifup = 1;
2621 		}
2622 		/* See if permanently promiscuous mode bit is about to flip */
2623 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2624 			if (new_flags & IFF_PPROMISC)
2625 				ifp->if_flags |= IFF_PROMISC;
2626 			else if (ifp->if_pcount == 0)
2627 				ifp->if_flags &= ~IFF_PROMISC;
2628 			if (log_promisc_mode_change)
2629                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2630                                     ((new_flags & IFF_PPROMISC) ?
2631                                      "enabled" : "disabled"));
2632 		}
2633 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2634 			(new_flags &~ IFF_CANTCHANGE);
2635 		if (ifp->if_ioctl) {
2636 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2637 		}
2638 		if (do_ifup)
2639 			if_up(ifp);
2640 		getmicrotime(&ifp->if_lastchange);
2641 		break;
2642 
2643 	case SIOCSIFCAP:
2644 		error = priv_check(td, PRIV_NET_SETIFCAP);
2645 		if (error)
2646 			return (error);
2647 		if (ifp->if_ioctl == NULL)
2648 			return (EOPNOTSUPP);
2649 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2650 			return (EINVAL);
2651 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2652 		if (error == 0)
2653 			getmicrotime(&ifp->if_lastchange);
2654 		break;
2655 
2656 #ifdef MAC
2657 	case SIOCSIFMAC:
2658 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2659 		break;
2660 #endif
2661 
2662 	case SIOCSIFNAME:
2663 		error = priv_check(td, PRIV_NET_SETIFNAME);
2664 		if (error)
2665 			return (error);
2666 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2667 		    NULL);
2668 		if (error != 0)
2669 			return (error);
2670 		if (new_name[0] == '\0')
2671 			return (EINVAL);
2672 		if (new_name[IFNAMSIZ-1] != '\0') {
2673 			new_name[IFNAMSIZ-1] = '\0';
2674 			if (strlen(new_name) == IFNAMSIZ-1)
2675 				return (EINVAL);
2676 		}
2677 		if (strcmp(new_name, ifp->if_xname) == 0)
2678 			break;
2679 		if (ifunit(new_name) != NULL)
2680 			return (EEXIST);
2681 
2682 		/*
2683 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2684 		 * and there are numerous other races with the
2685 		 * ifunit() checks not being atomic with namespace
2686 		 * changes (renames, vmoves, if_attach, etc).
2687 		 */
2688 		ifp->if_flags |= IFF_RENAMING;
2689 
2690 		/* Announce the departure of the interface. */
2691 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2692 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2693 
2694 		if_printf(ifp, "changing name to '%s'\n", new_name);
2695 
2696 		IF_ADDR_WLOCK(ifp);
2697 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2698 		ifa = ifp->if_addr;
2699 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2700 		namelen = strlen(new_name);
2701 		onamelen = sdl->sdl_nlen;
2702 		/*
2703 		 * Move the address if needed.  This is safe because we
2704 		 * allocate space for a name of length IFNAMSIZ when we
2705 		 * create this in if_attach().
2706 		 */
2707 		if (namelen != onamelen) {
2708 			bcopy(sdl->sdl_data + onamelen,
2709 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2710 		}
2711 		bcopy(new_name, sdl->sdl_data, namelen);
2712 		sdl->sdl_nlen = namelen;
2713 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2714 		bzero(sdl->sdl_data, onamelen);
2715 		while (namelen != 0)
2716 			sdl->sdl_data[--namelen] = 0xff;
2717 		IF_ADDR_WUNLOCK(ifp);
2718 
2719 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2720 		/* Announce the return of the interface. */
2721 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2722 
2723 		ifp->if_flags &= ~IFF_RENAMING;
2724 		break;
2725 
2726 #ifdef VIMAGE
2727 	case SIOCSIFVNET:
2728 		error = priv_check(td, PRIV_NET_SETIFVNET);
2729 		if (error)
2730 			return (error);
2731 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2732 		break;
2733 #endif
2734 
2735 	case SIOCSIFMETRIC:
2736 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2737 		if (error)
2738 			return (error);
2739 		ifp->if_metric = ifr->ifr_metric;
2740 		getmicrotime(&ifp->if_lastchange);
2741 		break;
2742 
2743 	case SIOCSIFPHYS:
2744 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2745 		if (error)
2746 			return (error);
2747 		if (ifp->if_ioctl == NULL)
2748 			return (EOPNOTSUPP);
2749 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2750 		if (error == 0)
2751 			getmicrotime(&ifp->if_lastchange);
2752 		break;
2753 
2754 	case SIOCSIFMTU:
2755 	{
2756 		u_long oldmtu = ifp->if_mtu;
2757 
2758 		error = priv_check(td, PRIV_NET_SETIFMTU);
2759 		if (error)
2760 			return (error);
2761 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2762 			return (EINVAL);
2763 		if (ifp->if_ioctl == NULL)
2764 			return (EOPNOTSUPP);
2765 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2766 		if (error == 0) {
2767 			getmicrotime(&ifp->if_lastchange);
2768 			rt_ifmsg(ifp);
2769 #ifdef INET
2770 			DEBUGNET_NOTIFY_MTU(ifp);
2771 #endif
2772 		}
2773 		/*
2774 		 * If the link MTU changed, do network layer specific procedure.
2775 		 */
2776 		if (ifp->if_mtu != oldmtu) {
2777 #ifdef INET6
2778 			nd6_setmtu(ifp);
2779 #endif
2780 			rt_updatemtu(ifp);
2781 		}
2782 		break;
2783 	}
2784 
2785 	case SIOCADDMULTI:
2786 	case SIOCDELMULTI:
2787 		if (cmd == SIOCADDMULTI)
2788 			error = priv_check(td, PRIV_NET_ADDMULTI);
2789 		else
2790 			error = priv_check(td, PRIV_NET_DELMULTI);
2791 		if (error)
2792 			return (error);
2793 
2794 		/* Don't allow group membership on non-multicast interfaces. */
2795 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2796 			return (EOPNOTSUPP);
2797 
2798 		/* Don't let users screw up protocols' entries. */
2799 		if (ifr->ifr_addr.sa_family != AF_LINK)
2800 			return (EINVAL);
2801 
2802 		if (cmd == SIOCADDMULTI) {
2803 			struct epoch_tracker et;
2804 			struct ifmultiaddr *ifma;
2805 
2806 			/*
2807 			 * Userland is only permitted to join groups once
2808 			 * via the if_addmulti() KPI, because it cannot hold
2809 			 * struct ifmultiaddr * between calls. It may also
2810 			 * lose a race while we check if the membership
2811 			 * already exists.
2812 			 */
2813 			NET_EPOCH_ENTER(et);
2814 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2815 			NET_EPOCH_EXIT(et);
2816 			if (ifma != NULL)
2817 				error = EADDRINUSE;
2818 			else
2819 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2820 		} else {
2821 			error = if_delmulti(ifp, &ifr->ifr_addr);
2822 		}
2823 		if (error == 0)
2824 			getmicrotime(&ifp->if_lastchange);
2825 		break;
2826 
2827 	case SIOCSIFPHYADDR:
2828 	case SIOCDIFPHYADDR:
2829 #ifdef INET6
2830 	case SIOCSIFPHYADDR_IN6:
2831 #endif
2832 	case SIOCSIFMEDIA:
2833 	case SIOCSIFGENERIC:
2834 		error = priv_check(td, PRIV_NET_HWIOCTL);
2835 		if (error)
2836 			return (error);
2837 		if (ifp->if_ioctl == NULL)
2838 			return (EOPNOTSUPP);
2839 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2840 		if (error == 0)
2841 			getmicrotime(&ifp->if_lastchange);
2842 		break;
2843 
2844 	case SIOCGIFSTATUS:
2845 	case SIOCGIFPSRCADDR:
2846 	case SIOCGIFPDSTADDR:
2847 	case SIOCGIFMEDIA:
2848 	case SIOCGIFXMEDIA:
2849 	case SIOCGIFGENERIC:
2850 	case SIOCGIFRSSKEY:
2851 	case SIOCGIFRSSHASH:
2852 	case SIOCGIFDOWNREASON:
2853 		if (ifp->if_ioctl == NULL)
2854 			return (EOPNOTSUPP);
2855 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2856 		break;
2857 
2858 	case SIOCSIFLLADDR:
2859 		error = priv_check(td, PRIV_NET_SETLLADDR);
2860 		if (error)
2861 			return (error);
2862 		error = if_setlladdr(ifp,
2863 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2864 		break;
2865 
2866 	case SIOCGHWADDR:
2867 		error = if_gethwaddr(ifp, ifr);
2868 		break;
2869 
2870 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2871 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2872 		if (error)
2873 			return (error);
2874 		if ((error = if_addgroup(ifp,
2875 		    ifgr_group_get((struct ifgroupreq *)data))))
2876 			return (error);
2877 		break;
2878 
2879 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2880 	{
2881 		struct epoch_tracker et;
2882 
2883 		NET_EPOCH_ENTER(et);
2884 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2885 		NET_EPOCH_EXIT(et);
2886 		break;
2887 	}
2888 
2889 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2890 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2891 		if (error)
2892 			return (error);
2893 		if ((error = if_delgroup(ifp,
2894 		    ifgr_group_get((struct ifgroupreq *)data))))
2895 			return (error);
2896 		break;
2897 
2898 	default:
2899 		error = ENOIOCTL;
2900 		break;
2901 	}
2902 	return (error);
2903 }
2904 
2905 #ifdef COMPAT_FREEBSD32
2906 struct ifconf32 {
2907 	int32_t	ifc_len;
2908 	union {
2909 		uint32_t	ifcu_buf;
2910 		uint32_t	ifcu_req;
2911 	} ifc_ifcu;
2912 };
2913 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2914 #endif
2915 
2916 #ifdef COMPAT_FREEBSD32
2917 static void
2918 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2919 {
2920 	struct ifmediareq32 *ifmr32;
2921 
2922 	ifmr32 = (struct ifmediareq32 *)data;
2923 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2924 	    sizeof(ifmr->ifm_name));
2925 	ifmr->ifm_current = ifmr32->ifm_current;
2926 	ifmr->ifm_mask = ifmr32->ifm_mask;
2927 	ifmr->ifm_status = ifmr32->ifm_status;
2928 	ifmr->ifm_active = ifmr32->ifm_active;
2929 	ifmr->ifm_count = ifmr32->ifm_count;
2930 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2931 }
2932 
2933 static void
2934 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2935 {
2936 	struct ifmediareq32 *ifmr32;
2937 
2938 	ifmr32 = (struct ifmediareq32 *)data;
2939 	ifmr32->ifm_current = ifmr->ifm_current;
2940 	ifmr32->ifm_mask = ifmr->ifm_mask;
2941 	ifmr32->ifm_status = ifmr->ifm_status;
2942 	ifmr32->ifm_active = ifmr->ifm_active;
2943 	ifmr32->ifm_count = ifmr->ifm_count;
2944 }
2945 #endif
2946 
2947 /*
2948  * Interface ioctls.
2949  */
2950 int
2951 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2952 {
2953 #ifdef COMPAT_FREEBSD32
2954 	caddr_t saved_data = NULL;
2955 	struct ifmediareq ifmr;
2956 	struct ifmediareq *ifmrp = NULL;
2957 #endif
2958 	struct ifnet *ifp;
2959 	struct ifreq *ifr;
2960 	int error;
2961 	int oif_flags;
2962 #ifdef VIMAGE
2963 	bool shutdown;
2964 #endif
2965 
2966 	CURVNET_SET(so->so_vnet);
2967 #ifdef VIMAGE
2968 	/* Make sure the VNET is stable. */
2969 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2970 	if (shutdown) {
2971 		CURVNET_RESTORE();
2972 		return (EBUSY);
2973 	}
2974 #endif
2975 
2976 	switch (cmd) {
2977 	case SIOCGIFCONF:
2978 		error = ifconf(cmd, data);
2979 		goto out_noref;
2980 
2981 #ifdef COMPAT_FREEBSD32
2982 	case SIOCGIFCONF32:
2983 		{
2984 			struct ifconf32 *ifc32;
2985 			struct ifconf ifc;
2986 
2987 			ifc32 = (struct ifconf32 *)data;
2988 			ifc.ifc_len = ifc32->ifc_len;
2989 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2990 
2991 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2992 			if (error == 0)
2993 				ifc32->ifc_len = ifc.ifc_len;
2994 			goto out_noref;
2995 		}
2996 #endif
2997 	}
2998 
2999 #ifdef COMPAT_FREEBSD32
3000 	switch (cmd) {
3001 	case SIOCGIFMEDIA32:
3002 	case SIOCGIFXMEDIA32:
3003 		ifmrp = &ifmr;
3004 		ifmr_init(ifmrp, data);
3005 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3006 		saved_data = data;
3007 		data = (caddr_t)ifmrp;
3008 	}
3009 #endif
3010 
3011 	ifr = (struct ifreq *)data;
3012 	switch (cmd) {
3013 #ifdef VIMAGE
3014 	case SIOCSIFRVNET:
3015 		error = priv_check(td, PRIV_NET_SETIFVNET);
3016 		if (error == 0)
3017 			error = if_vmove_reclaim(td, ifr->ifr_name,
3018 			    ifr->ifr_jid);
3019 		goto out_noref;
3020 #endif
3021 	case SIOCIFCREATE:
3022 	case SIOCIFCREATE2:
3023 		error = priv_check(td, PRIV_NET_IFCREATE);
3024 		if (error == 0)
3025 			error = if_clone_create(ifr->ifr_name,
3026 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3027 			    ifr_data_get_ptr(ifr) : NULL);
3028 		goto out_noref;
3029 	case SIOCIFDESTROY:
3030 		error = priv_check(td, PRIV_NET_IFDESTROY);
3031 		if (error == 0)
3032 			error = if_clone_destroy(ifr->ifr_name);
3033 		goto out_noref;
3034 
3035 	case SIOCIFGCLONERS:
3036 		error = if_clone_list((struct if_clonereq *)data);
3037 		goto out_noref;
3038 
3039 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3040 		error = if_getgroupmembers((struct ifgroupreq *)data);
3041 		goto out_noref;
3042 
3043 #if defined(INET) || defined(INET6)
3044 	case SIOCSVH:
3045 	case SIOCGVH:
3046 		if (carp_ioctl_p == NULL)
3047 			error = EPROTONOSUPPORT;
3048 		else
3049 			error = (*carp_ioctl_p)(ifr, cmd, td);
3050 		goto out_noref;
3051 #endif
3052 	}
3053 
3054 	ifp = ifunit_ref(ifr->ifr_name);
3055 	if (ifp == NULL) {
3056 		error = ENXIO;
3057 		goto out_noref;
3058 	}
3059 
3060 	error = ifhwioctl(cmd, ifp, data, td);
3061 	if (error != ENOIOCTL)
3062 		goto out_ref;
3063 
3064 	oif_flags = ifp->if_flags;
3065 	if (so->so_proto == NULL) {
3066 		error = EOPNOTSUPP;
3067 		goto out_ref;
3068 	}
3069 
3070 	/*
3071 	 * Pass the request on to the socket control method, and if the
3072 	 * latter returns EOPNOTSUPP, directly to the interface.
3073 	 *
3074 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3075 	 * trust SIOCSIFADDR et al to come from an already privileged
3076 	 * layer, and do not perform any credentials checks or input
3077 	 * validation.
3078 	 */
3079 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3080 	    ifp, td));
3081 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3082 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3083 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3084 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3085 
3086 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3087 #ifdef INET6
3088 		if (ifp->if_flags & IFF_UP)
3089 			in6_if_up(ifp);
3090 #endif
3091 	}
3092 
3093 out_ref:
3094 	if_rele(ifp);
3095 out_noref:
3096 #ifdef COMPAT_FREEBSD32
3097 	if (ifmrp != NULL) {
3098 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3099 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3100 		     cmd));
3101 		data = saved_data;
3102 		ifmr_update(ifmrp, data);
3103 	}
3104 #endif
3105 	CURVNET_RESTORE();
3106 	return (error);
3107 }
3108 
3109 /*
3110  * The code common to handling reference counted flags,
3111  * e.g., in ifpromisc() and if_allmulti().
3112  * The "pflag" argument can specify a permanent mode flag to check,
3113  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3114  *
3115  * Only to be used on stack-owned flags, not driver-owned flags.
3116  */
3117 static int
3118 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3119 {
3120 	struct ifreq ifr;
3121 	int error;
3122 	int oldflags, oldcount;
3123 
3124 	/* Sanity checks to catch programming errors */
3125 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3126 	    ("%s: setting driver-owned flag %d", __func__, flag));
3127 
3128 	if (onswitch)
3129 		KASSERT(*refcount >= 0,
3130 		    ("%s: increment negative refcount %d for flag %d",
3131 		    __func__, *refcount, flag));
3132 	else
3133 		KASSERT(*refcount > 0,
3134 		    ("%s: decrement non-positive refcount %d for flag %d",
3135 		    __func__, *refcount, flag));
3136 
3137 	/* In case this mode is permanent, just touch refcount */
3138 	if (ifp->if_flags & pflag) {
3139 		*refcount += onswitch ? 1 : -1;
3140 		return (0);
3141 	}
3142 
3143 	/* Save ifnet parameters for if_ioctl() may fail */
3144 	oldcount = *refcount;
3145 	oldflags = ifp->if_flags;
3146 
3147 	/*
3148 	 * See if we aren't the only and touching refcount is enough.
3149 	 * Actually toggle interface flag if we are the first or last.
3150 	 */
3151 	if (onswitch) {
3152 		if ((*refcount)++)
3153 			return (0);
3154 		ifp->if_flags |= flag;
3155 	} else {
3156 		if (--(*refcount))
3157 			return (0);
3158 		ifp->if_flags &= ~flag;
3159 	}
3160 
3161 	/* Call down the driver since we've changed interface flags */
3162 	if (ifp->if_ioctl == NULL) {
3163 		error = EOPNOTSUPP;
3164 		goto recover;
3165 	}
3166 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3167 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3168 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3169 	if (error)
3170 		goto recover;
3171 	/* Notify userland that interface flags have changed */
3172 	rt_ifmsg(ifp);
3173 	return (0);
3174 
3175 recover:
3176 	/* Recover after driver error */
3177 	*refcount = oldcount;
3178 	ifp->if_flags = oldflags;
3179 	return (error);
3180 }
3181 
3182 /*
3183  * Set/clear promiscuous mode on interface ifp based on the truth value
3184  * of pswitch.  The calls are reference counted so that only the first
3185  * "on" request actually has an effect, as does the final "off" request.
3186  * Results are undefined if the "off" and "on" requests are not matched.
3187  */
3188 int
3189 ifpromisc(struct ifnet *ifp, int pswitch)
3190 {
3191 	int error;
3192 	int oldflags = ifp->if_flags;
3193 
3194 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3195 			   &ifp->if_pcount, pswitch);
3196 	/* If promiscuous mode status has changed, log a message */
3197 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3198             log_promisc_mode_change)
3199 		if_printf(ifp, "promiscuous mode %s\n",
3200 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3201 	return (error);
3202 }
3203 
3204 /*
3205  * Return interface configuration
3206  * of system.  List may be used
3207  * in later ioctl's (above) to get
3208  * other information.
3209  */
3210 /*ARGSUSED*/
3211 static int
3212 ifconf(u_long cmd, caddr_t data)
3213 {
3214 	struct ifconf *ifc = (struct ifconf *)data;
3215 	struct ifnet *ifp;
3216 	struct ifaddr *ifa;
3217 	struct ifreq ifr;
3218 	struct sbuf *sb;
3219 	int error, full = 0, valid_len, max_len;
3220 
3221 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3222 	max_len = MAXPHYS - 1;
3223 
3224 	/* Prevent hostile input from being able to crash the system */
3225 	if (ifc->ifc_len <= 0)
3226 		return (EINVAL);
3227 
3228 again:
3229 	if (ifc->ifc_len <= max_len) {
3230 		max_len = ifc->ifc_len;
3231 		full = 1;
3232 	}
3233 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3234 	max_len = 0;
3235 	valid_len = 0;
3236 
3237 	IFNET_RLOCK();
3238 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3239 		struct epoch_tracker et;
3240 		int addrs;
3241 
3242 		/*
3243 		 * Zero the ifr to make sure we don't disclose the contents
3244 		 * of the stack.
3245 		 */
3246 		memset(&ifr, 0, sizeof(ifr));
3247 
3248 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3249 		    >= sizeof(ifr.ifr_name)) {
3250 			sbuf_delete(sb);
3251 			IFNET_RUNLOCK();
3252 			return (ENAMETOOLONG);
3253 		}
3254 
3255 		addrs = 0;
3256 		NET_EPOCH_ENTER(et);
3257 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3258 			struct sockaddr *sa = ifa->ifa_addr;
3259 
3260 			if (prison_if(curthread->td_ucred, sa) != 0)
3261 				continue;
3262 			addrs++;
3263 			if (sa->sa_len <= sizeof(*sa)) {
3264 				if (sa->sa_len < sizeof(*sa)) {
3265 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3266 					    sizeof(ifr.ifr_ifru.ifru_addr));
3267 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3268 					    sa->sa_len);
3269 				} else
3270 					ifr.ifr_ifru.ifru_addr = *sa;
3271 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3272 				max_len += sizeof(ifr);
3273 			} else {
3274 				sbuf_bcat(sb, &ifr,
3275 				    offsetof(struct ifreq, ifr_addr));
3276 				max_len += offsetof(struct ifreq, ifr_addr);
3277 				sbuf_bcat(sb, sa, sa->sa_len);
3278 				max_len += sa->sa_len;
3279 			}
3280 
3281 			if (sbuf_error(sb) == 0)
3282 				valid_len = sbuf_len(sb);
3283 		}
3284 		NET_EPOCH_EXIT(et);
3285 		if (addrs == 0) {
3286 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3287 			max_len += sizeof(ifr);
3288 
3289 			if (sbuf_error(sb) == 0)
3290 				valid_len = sbuf_len(sb);
3291 		}
3292 	}
3293 	IFNET_RUNLOCK();
3294 
3295 	/*
3296 	 * If we didn't allocate enough space (uncommon), try again.  If
3297 	 * we have already allocated as much space as we are allowed,
3298 	 * return what we've got.
3299 	 */
3300 	if (valid_len != max_len && !full) {
3301 		sbuf_delete(sb);
3302 		goto again;
3303 	}
3304 
3305 	ifc->ifc_len = valid_len;
3306 	sbuf_finish(sb);
3307 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3308 	sbuf_delete(sb);
3309 	return (error);
3310 }
3311 
3312 /*
3313  * Just like ifpromisc(), but for all-multicast-reception mode.
3314  */
3315 int
3316 if_allmulti(struct ifnet *ifp, int onswitch)
3317 {
3318 
3319 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3320 }
3321 
3322 struct ifmultiaddr *
3323 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3324 {
3325 	struct ifmultiaddr *ifma;
3326 
3327 	IF_ADDR_LOCK_ASSERT(ifp);
3328 
3329 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3330 		if (sa->sa_family == AF_LINK) {
3331 			if (sa_dl_equal(ifma->ifma_addr, sa))
3332 				break;
3333 		} else {
3334 			if (sa_equal(ifma->ifma_addr, sa))
3335 				break;
3336 		}
3337 	}
3338 
3339 	return ifma;
3340 }
3341 
3342 /*
3343  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3344  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3345  * the ifnet multicast address list here, so the caller must do that and
3346  * other setup work (such as notifying the device driver).  The reference
3347  * count is initialized to 1.
3348  */
3349 static struct ifmultiaddr *
3350 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3351     int mflags)
3352 {
3353 	struct ifmultiaddr *ifma;
3354 	struct sockaddr *dupsa;
3355 
3356 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3357 	    M_ZERO);
3358 	if (ifma == NULL)
3359 		return (NULL);
3360 
3361 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3362 	if (dupsa == NULL) {
3363 		free(ifma, M_IFMADDR);
3364 		return (NULL);
3365 	}
3366 	bcopy(sa, dupsa, sa->sa_len);
3367 	ifma->ifma_addr = dupsa;
3368 
3369 	ifma->ifma_ifp = ifp;
3370 	ifma->ifma_refcount = 1;
3371 	ifma->ifma_protospec = NULL;
3372 
3373 	if (llsa == NULL) {
3374 		ifma->ifma_lladdr = NULL;
3375 		return (ifma);
3376 	}
3377 
3378 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3379 	if (dupsa == NULL) {
3380 		free(ifma->ifma_addr, M_IFMADDR);
3381 		free(ifma, M_IFMADDR);
3382 		return (NULL);
3383 	}
3384 	bcopy(llsa, dupsa, llsa->sa_len);
3385 	ifma->ifma_lladdr = dupsa;
3386 
3387 	return (ifma);
3388 }
3389 
3390 /*
3391  * if_freemulti: free ifmultiaddr structure and possibly attached related
3392  * addresses.  The caller is responsible for implementing reference
3393  * counting, notifying the driver, handling routing messages, and releasing
3394  * any dependent link layer state.
3395  */
3396 #ifdef MCAST_VERBOSE
3397 extern void kdb_backtrace(void);
3398 #endif
3399 static void
3400 if_freemulti_internal(struct ifmultiaddr *ifma)
3401 {
3402 
3403 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3404 	    ifma->ifma_refcount));
3405 
3406 	if (ifma->ifma_lladdr != NULL)
3407 		free(ifma->ifma_lladdr, M_IFMADDR);
3408 #ifdef MCAST_VERBOSE
3409 	kdb_backtrace();
3410 	printf("%s freeing ifma: %p\n", __func__, ifma);
3411 #endif
3412 	free(ifma->ifma_addr, M_IFMADDR);
3413 	free(ifma, M_IFMADDR);
3414 }
3415 
3416 static void
3417 if_destroymulti(epoch_context_t ctx)
3418 {
3419 	struct ifmultiaddr *ifma;
3420 
3421 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3422 	if_freemulti_internal(ifma);
3423 }
3424 
3425 void
3426 if_freemulti(struct ifmultiaddr *ifma)
3427 {
3428 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3429 	    ifma->ifma_refcount));
3430 
3431 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3432 }
3433 
3434 
3435 /*
3436  * Register an additional multicast address with a network interface.
3437  *
3438  * - If the address is already present, bump the reference count on the
3439  *   address and return.
3440  * - If the address is not link-layer, look up a link layer address.
3441  * - Allocate address structures for one or both addresses, and attach to the
3442  *   multicast address list on the interface.  If automatically adding a link
3443  *   layer address, the protocol address will own a reference to the link
3444  *   layer address, to be freed when it is freed.
3445  * - Notify the network device driver of an addition to the multicast address
3446  *   list.
3447  *
3448  * 'sa' points to caller-owned memory with the desired multicast address.
3449  *
3450  * 'retifma' will be used to return a pointer to the resulting multicast
3451  * address reference, if desired.
3452  */
3453 int
3454 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3455     struct ifmultiaddr **retifma)
3456 {
3457 	struct ifmultiaddr *ifma, *ll_ifma;
3458 	struct sockaddr *llsa;
3459 	struct sockaddr_dl sdl;
3460 	int error;
3461 
3462 #ifdef INET
3463 	IN_MULTI_LIST_UNLOCK_ASSERT();
3464 #endif
3465 #ifdef INET6
3466 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3467 #endif
3468 	/*
3469 	 * If the address is already present, return a new reference to it;
3470 	 * otherwise, allocate storage and set up a new address.
3471 	 */
3472 	IF_ADDR_WLOCK(ifp);
3473 	ifma = if_findmulti(ifp, sa);
3474 	if (ifma != NULL) {
3475 		ifma->ifma_refcount++;
3476 		if (retifma != NULL)
3477 			*retifma = ifma;
3478 		IF_ADDR_WUNLOCK(ifp);
3479 		return (0);
3480 	}
3481 
3482 	/*
3483 	 * The address isn't already present; resolve the protocol address
3484 	 * into a link layer address, and then look that up, bump its
3485 	 * refcount or allocate an ifma for that also.
3486 	 * Most link layer resolving functions returns address data which
3487 	 * fits inside default sockaddr_dl structure. However callback
3488 	 * can allocate another sockaddr structure, in that case we need to
3489 	 * free it later.
3490 	 */
3491 	llsa = NULL;
3492 	ll_ifma = NULL;
3493 	if (ifp->if_resolvemulti != NULL) {
3494 		/* Provide called function with buffer size information */
3495 		sdl.sdl_len = sizeof(sdl);
3496 		llsa = (struct sockaddr *)&sdl;
3497 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3498 		if (error)
3499 			goto unlock_out;
3500 	}
3501 
3502 	/*
3503 	 * Allocate the new address.  Don't hook it up yet, as we may also
3504 	 * need to allocate a link layer multicast address.
3505 	 */
3506 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3507 	if (ifma == NULL) {
3508 		error = ENOMEM;
3509 		goto free_llsa_out;
3510 	}
3511 
3512 	/*
3513 	 * If a link layer address is found, we'll need to see if it's
3514 	 * already present in the address list, or allocate is as well.
3515 	 * When this block finishes, the link layer address will be on the
3516 	 * list.
3517 	 */
3518 	if (llsa != NULL) {
3519 		ll_ifma = if_findmulti(ifp, llsa);
3520 		if (ll_ifma == NULL) {
3521 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3522 			if (ll_ifma == NULL) {
3523 				--ifma->ifma_refcount;
3524 				if_freemulti(ifma);
3525 				error = ENOMEM;
3526 				goto free_llsa_out;
3527 			}
3528 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3529 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3530 			    ifma_link);
3531 		} else
3532 			ll_ifma->ifma_refcount++;
3533 		ifma->ifma_llifma = ll_ifma;
3534 	}
3535 
3536 	/*
3537 	 * We now have a new multicast address, ifma, and possibly a new or
3538 	 * referenced link layer address.  Add the primary address to the
3539 	 * ifnet address list.
3540 	 */
3541 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3542 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3543 
3544 	if (retifma != NULL)
3545 		*retifma = ifma;
3546 
3547 	/*
3548 	 * Must generate the message while holding the lock so that 'ifma'
3549 	 * pointer is still valid.
3550 	 */
3551 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3552 	IF_ADDR_WUNLOCK(ifp);
3553 
3554 	/*
3555 	 * We are certain we have added something, so call down to the
3556 	 * interface to let them know about it.
3557 	 */
3558 	if (ifp->if_ioctl != NULL) {
3559 		if (THREAD_CAN_SLEEP())
3560 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3561 		else
3562 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3563 	}
3564 
3565 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3566 		link_free_sdl(llsa);
3567 
3568 	return (0);
3569 
3570 free_llsa_out:
3571 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3572 		link_free_sdl(llsa);
3573 
3574 unlock_out:
3575 	IF_ADDR_WUNLOCK(ifp);
3576 	return (error);
3577 }
3578 
3579 static void
3580 if_siocaddmulti(void *arg, int pending)
3581 {
3582 	struct ifnet *ifp;
3583 
3584 	ifp = arg;
3585 #ifdef DIAGNOSTIC
3586 	if (pending > 1)
3587 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3588 #endif
3589 	CURVNET_SET(ifp->if_vnet);
3590 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3591 	CURVNET_RESTORE();
3592 }
3593 
3594 /*
3595  * Delete a multicast group membership by network-layer group address.
3596  *
3597  * Returns ENOENT if the entry could not be found. If ifp no longer
3598  * exists, results are undefined. This entry point should only be used
3599  * from subsystems which do appropriate locking to hold ifp for the
3600  * duration of the call.
3601  * Network-layer protocol domains must use if_delmulti_ifma().
3602  */
3603 int
3604 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3605 {
3606 	struct ifmultiaddr *ifma;
3607 	int lastref;
3608 
3609 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3610 
3611 	IF_ADDR_WLOCK(ifp);
3612 	lastref = 0;
3613 	ifma = if_findmulti(ifp, sa);
3614 	if (ifma != NULL)
3615 		lastref = if_delmulti_locked(ifp, ifma, 0);
3616 	IF_ADDR_WUNLOCK(ifp);
3617 
3618 	if (ifma == NULL)
3619 		return (ENOENT);
3620 
3621 	if (lastref && ifp->if_ioctl != NULL) {
3622 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3623 	}
3624 
3625 	return (0);
3626 }
3627 
3628 /*
3629  * Delete all multicast group membership for an interface.
3630  * Should be used to quickly flush all multicast filters.
3631  */
3632 void
3633 if_delallmulti(struct ifnet *ifp)
3634 {
3635 	struct ifmultiaddr *ifma;
3636 	struct ifmultiaddr *next;
3637 
3638 	IF_ADDR_WLOCK(ifp);
3639 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3640 		if_delmulti_locked(ifp, ifma, 0);
3641 	IF_ADDR_WUNLOCK(ifp);
3642 }
3643 
3644 void
3645 if_delmulti_ifma(struct ifmultiaddr *ifma)
3646 {
3647 	if_delmulti_ifma_flags(ifma, 0);
3648 }
3649 
3650 /*
3651  * Delete a multicast group membership by group membership pointer.
3652  * Network-layer protocol domains must use this routine.
3653  *
3654  * It is safe to call this routine if the ifp disappeared.
3655  */
3656 void
3657 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3658 {
3659 	struct ifnet *ifp;
3660 	int lastref;
3661 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3662 #ifdef INET
3663 	IN_MULTI_LIST_UNLOCK_ASSERT();
3664 #endif
3665 	ifp = ifma->ifma_ifp;
3666 #ifdef DIAGNOSTIC
3667 	if (ifp == NULL) {
3668 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3669 	} else {
3670 		struct epoch_tracker et;
3671 		struct ifnet *oifp;
3672 
3673 		NET_EPOCH_ENTER(et);
3674 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3675 			if (ifp == oifp)
3676 				break;
3677 		NET_EPOCH_EXIT(et);
3678 		if (ifp != oifp)
3679 			ifp = NULL;
3680 	}
3681 #endif
3682 	/*
3683 	 * If and only if the ifnet instance exists: Acquire the address lock.
3684 	 */
3685 	if (ifp != NULL)
3686 		IF_ADDR_WLOCK(ifp);
3687 
3688 	lastref = if_delmulti_locked(ifp, ifma, flags);
3689 
3690 	if (ifp != NULL) {
3691 		/*
3692 		 * If and only if the ifnet instance exists:
3693 		 *  Release the address lock.
3694 		 *  If the group was left: update the hardware hash filter.
3695 		 */
3696 		IF_ADDR_WUNLOCK(ifp);
3697 		if (lastref && ifp->if_ioctl != NULL) {
3698 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3699 		}
3700 	}
3701 }
3702 
3703 /*
3704  * Perform deletion of network-layer and/or link-layer multicast address.
3705  *
3706  * Return 0 if the reference count was decremented.
3707  * Return 1 if the final reference was released, indicating that the
3708  * hardware hash filter should be reprogrammed.
3709  */
3710 static int
3711 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3712 {
3713 	struct ifmultiaddr *ll_ifma;
3714 
3715 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3716 		KASSERT(ifma->ifma_ifp == ifp,
3717 		    ("%s: inconsistent ifp %p", __func__, ifp));
3718 		IF_ADDR_WLOCK_ASSERT(ifp);
3719 	}
3720 
3721 	ifp = ifma->ifma_ifp;
3722 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3723 
3724 	/*
3725 	 * If the ifnet is detaching, null out references to ifnet,
3726 	 * so that upper protocol layers will notice, and not attempt
3727 	 * to obtain locks for an ifnet which no longer exists. The
3728 	 * routing socket announcement must happen before the ifnet
3729 	 * instance is detached from the system.
3730 	 */
3731 	if (detaching) {
3732 #ifdef DIAGNOSTIC
3733 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3734 #endif
3735 		/*
3736 		 * ifp may already be nulled out if we are being reentered
3737 		 * to delete the ll_ifma.
3738 		 */
3739 		if (ifp != NULL) {
3740 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3741 			ifma->ifma_ifp = NULL;
3742 		}
3743 	}
3744 
3745 	if (--ifma->ifma_refcount > 0)
3746 		return 0;
3747 
3748 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3749 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3750 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3751 	}
3752 	/*
3753 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3754 	 * have been associated with it. Release it first if so.
3755 	 */
3756 	ll_ifma = ifma->ifma_llifma;
3757 	if (ll_ifma != NULL) {
3758 		KASSERT(ifma->ifma_lladdr != NULL,
3759 		    ("%s: llifma w/o lladdr", __func__));
3760 		if (detaching)
3761 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3762 		if (--ll_ifma->ifma_refcount == 0) {
3763 			if (ifp != NULL) {
3764 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3765 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3766 						ifma_link);
3767 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3768 				}
3769 			}
3770 			if_freemulti(ll_ifma);
3771 		}
3772 	}
3773 #ifdef INVARIANTS
3774 	if (ifp) {
3775 		struct ifmultiaddr *ifmatmp;
3776 
3777 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3778 			MPASS(ifma != ifmatmp);
3779 	}
3780 #endif
3781 	if_freemulti(ifma);
3782 	/*
3783 	 * The last reference to this instance of struct ifmultiaddr
3784 	 * was released; the hardware should be notified of this change.
3785 	 */
3786 	return 1;
3787 }
3788 
3789 /*
3790  * Set the link layer address on an interface.
3791  *
3792  * At this time we only support certain types of interfaces,
3793  * and we don't allow the length of the address to change.
3794  *
3795  * Set noinline to be dtrace-friendly
3796  */
3797 __noinline int
3798 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3799 {
3800 	struct sockaddr_dl *sdl;
3801 	struct ifaddr *ifa;
3802 	struct ifreq ifr;
3803 
3804 	ifa = ifp->if_addr;
3805 	if (ifa == NULL)
3806 		return (EINVAL);
3807 
3808 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3809 	if (sdl == NULL)
3810 		return (EINVAL);
3811 
3812 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3813 		return (EINVAL);
3814 
3815 	switch (ifp->if_type) {
3816 	case IFT_ETHER:
3817 	case IFT_XETHER:
3818 	case IFT_L2VLAN:
3819 	case IFT_BRIDGE:
3820 	case IFT_IEEE8023ADLAG:
3821 		bcopy(lladdr, LLADDR(sdl), len);
3822 		break;
3823 	default:
3824 		return (ENODEV);
3825 	}
3826 
3827 	/*
3828 	 * If the interface is already up, we need
3829 	 * to re-init it in order to reprogram its
3830 	 * address filter.
3831 	 */
3832 	if ((ifp->if_flags & IFF_UP) != 0) {
3833 		if (ifp->if_ioctl) {
3834 			ifp->if_flags &= ~IFF_UP;
3835 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3836 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3837 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3838 			ifp->if_flags |= IFF_UP;
3839 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3840 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3841 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3842 		}
3843 	}
3844 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3845 
3846 	return (0);
3847 }
3848 
3849 /*
3850  * Compat function for handling basic encapsulation requests.
3851  * Not converted stacks (FDDI, IB, ..) supports traditional
3852  * output model: ARP (and other similar L2 protocols) are handled
3853  * inside output routine, arpresolve/nd6_resolve() returns MAC
3854  * address instead of full prepend.
3855  *
3856  * This function creates calculated header==MAC for IPv4/IPv6 and
3857  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3858  * address families.
3859  */
3860 static int
3861 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3862 {
3863 
3864 	if (req->rtype != IFENCAP_LL)
3865 		return (EOPNOTSUPP);
3866 
3867 	if (req->bufsize < req->lladdr_len)
3868 		return (ENOMEM);
3869 
3870 	switch (req->family) {
3871 	case AF_INET:
3872 	case AF_INET6:
3873 		break;
3874 	default:
3875 		return (EAFNOSUPPORT);
3876 	}
3877 
3878 	/* Copy lladdr to storage as is */
3879 	memmove(req->buf, req->lladdr, req->lladdr_len);
3880 	req->bufsize = req->lladdr_len;
3881 	req->lladdr_off = 0;
3882 
3883 	return (0);
3884 }
3885 
3886 /*
3887  * Tunnel interfaces can nest, also they may cause infinite recursion
3888  * calls when misconfigured. We'll prevent this by detecting loops.
3889  * High nesting level may cause stack exhaustion. We'll prevent this
3890  * by introducing upper limit.
3891  *
3892  * Return 0, if tunnel nesting count is equal or less than limit.
3893  */
3894 int
3895 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3896     int limit)
3897 {
3898 	struct m_tag *mtag;
3899 	int count;
3900 
3901 	count = 1;
3902 	mtag = NULL;
3903 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3904 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3905 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3906 			return (EIO);
3907 		}
3908 		count++;
3909 	}
3910 	if (count > limit) {
3911 		log(LOG_NOTICE,
3912 		    "%s: if_output recursively called too many times(%d)\n",
3913 		    if_name(ifp), count);
3914 		return (EIO);
3915 	}
3916 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3917 	if (mtag == NULL)
3918 		return (ENOMEM);
3919 	*(struct ifnet **)(mtag + 1) = ifp;
3920 	m_tag_prepend(m, mtag);
3921 	return (0);
3922 }
3923 
3924 /*
3925  * Get the link layer address that was read from the hardware at attach.
3926  *
3927  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3928  * their component interfaces as IFT_IEEE8023ADLAG.
3929  */
3930 int
3931 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3932 {
3933 
3934 	if (ifp->if_hw_addr == NULL)
3935 		return (ENODEV);
3936 
3937 	switch (ifp->if_type) {
3938 	case IFT_ETHER:
3939 	case IFT_IEEE8023ADLAG:
3940 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3941 		return (0);
3942 	default:
3943 		return (ENODEV);
3944 	}
3945 }
3946 
3947 /*
3948  * The name argument must be a pointer to storage which will last as
3949  * long as the interface does.  For physical devices, the result of
3950  * device_get_name(dev) is a good choice and for pseudo-devices a
3951  * static string works well.
3952  */
3953 void
3954 if_initname(struct ifnet *ifp, const char *name, int unit)
3955 {
3956 	ifp->if_dname = name;
3957 	ifp->if_dunit = unit;
3958 	if (unit != IF_DUNIT_NONE)
3959 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3960 	else
3961 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3962 }
3963 
3964 int
3965 if_printf(struct ifnet *ifp, const char *fmt, ...)
3966 {
3967 	char if_fmt[256];
3968 	va_list ap;
3969 
3970 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3971 	va_start(ap, fmt);
3972 	vlog(LOG_INFO, if_fmt, ap);
3973 	va_end(ap);
3974 	return (0);
3975 }
3976 
3977 void
3978 if_start(struct ifnet *ifp)
3979 {
3980 
3981 	(*(ifp)->if_start)(ifp);
3982 }
3983 
3984 /*
3985  * Backwards compatibility interface for drivers
3986  * that have not implemented it
3987  */
3988 static int
3989 if_transmit(struct ifnet *ifp, struct mbuf *m)
3990 {
3991 	int error;
3992 
3993 	IFQ_HANDOFF(ifp, m, error);
3994 	return (error);
3995 }
3996 
3997 static void
3998 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3999 {
4000 
4001 	m_freem(m);
4002 }
4003 
4004 int
4005 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4006 {
4007 	int active = 0;
4008 
4009 	IF_LOCK(ifq);
4010 	if (_IF_QFULL(ifq)) {
4011 		IF_UNLOCK(ifq);
4012 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4013 		m_freem(m);
4014 		return (0);
4015 	}
4016 	if (ifp != NULL) {
4017 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4018 		if (m->m_flags & (M_BCAST|M_MCAST))
4019 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4020 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4021 	}
4022 	_IF_ENQUEUE(ifq, m);
4023 	IF_UNLOCK(ifq);
4024 	if (ifp != NULL && !active)
4025 		(*(ifp)->if_start)(ifp);
4026 	return (1);
4027 }
4028 
4029 void
4030 if_register_com_alloc(u_char type,
4031     if_com_alloc_t *a, if_com_free_t *f)
4032 {
4033 
4034 	KASSERT(if_com_alloc[type] == NULL,
4035 	    ("if_register_com_alloc: %d already registered", type));
4036 	KASSERT(if_com_free[type] == NULL,
4037 	    ("if_register_com_alloc: %d free already registered", type));
4038 
4039 	if_com_alloc[type] = a;
4040 	if_com_free[type] = f;
4041 }
4042 
4043 void
4044 if_deregister_com_alloc(u_char type)
4045 {
4046 
4047 	KASSERT(if_com_alloc[type] != NULL,
4048 	    ("if_deregister_com_alloc: %d not registered", type));
4049 	KASSERT(if_com_free[type] != NULL,
4050 	    ("if_deregister_com_alloc: %d free not registered", type));
4051 	if_com_alloc[type] = NULL;
4052 	if_com_free[type] = NULL;
4053 }
4054 
4055 /* API for driver access to network stack owned ifnet.*/
4056 uint64_t
4057 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4058 {
4059 	uint64_t oldbrate;
4060 
4061 	oldbrate = ifp->if_baudrate;
4062 	ifp->if_baudrate = baudrate;
4063 	return (oldbrate);
4064 }
4065 
4066 uint64_t
4067 if_getbaudrate(if_t ifp)
4068 {
4069 
4070 	return (((struct ifnet *)ifp)->if_baudrate);
4071 }
4072 
4073 int
4074 if_setcapabilities(if_t ifp, int capabilities)
4075 {
4076 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4077 	return (0);
4078 }
4079 
4080 int
4081 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4082 {
4083 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4084 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4085 
4086 	return (0);
4087 }
4088 
4089 int
4090 if_getcapabilities(if_t ifp)
4091 {
4092 	return ((struct ifnet *)ifp)->if_capabilities;
4093 }
4094 
4095 int
4096 if_setcapenable(if_t ifp, int capabilities)
4097 {
4098 	((struct ifnet *)ifp)->if_capenable = capabilities;
4099 	return (0);
4100 }
4101 
4102 int
4103 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4104 {
4105 	if(setcap)
4106 		((struct ifnet *)ifp)->if_capenable |= setcap;
4107 	if(clearcap)
4108 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4109 
4110 	return (0);
4111 }
4112 
4113 const char *
4114 if_getdname(if_t ifp)
4115 {
4116 	return ((struct ifnet *)ifp)->if_dname;
4117 }
4118 
4119 int
4120 if_togglecapenable(if_t ifp, int togglecap)
4121 {
4122 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4123 	return (0);
4124 }
4125 
4126 int
4127 if_getcapenable(if_t ifp)
4128 {
4129 	return ((struct ifnet *)ifp)->if_capenable;
4130 }
4131 
4132 /*
4133  * This is largely undesirable because it ties ifnet to a device, but does
4134  * provide flexiblity for an embedded product vendor. Should be used with
4135  * the understanding that it violates the interface boundaries, and should be
4136  * a last resort only.
4137  */
4138 int
4139 if_setdev(if_t ifp, void *dev)
4140 {
4141 	return (0);
4142 }
4143 
4144 int
4145 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4146 {
4147 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4148 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4149 
4150 	return (0);
4151 }
4152 
4153 int
4154 if_getdrvflags(if_t ifp)
4155 {
4156 	return ((struct ifnet *)ifp)->if_drv_flags;
4157 }
4158 
4159 int
4160 if_setdrvflags(if_t ifp, int flags)
4161 {
4162 	((struct ifnet *)ifp)->if_drv_flags = flags;
4163 	return (0);
4164 }
4165 
4166 
4167 int
4168 if_setflags(if_t ifp, int flags)
4169 {
4170 
4171 	ifp->if_flags = flags;
4172 	return (0);
4173 }
4174 
4175 int
4176 if_setflagbits(if_t ifp, int set, int clear)
4177 {
4178 	((struct ifnet *)ifp)->if_flags |= set;
4179 	((struct ifnet *)ifp)->if_flags &= ~clear;
4180 
4181 	return (0);
4182 }
4183 
4184 int
4185 if_getflags(if_t ifp)
4186 {
4187 	return ((struct ifnet *)ifp)->if_flags;
4188 }
4189 
4190 int
4191 if_clearhwassist(if_t ifp)
4192 {
4193 	((struct ifnet *)ifp)->if_hwassist = 0;
4194 	return (0);
4195 }
4196 
4197 int
4198 if_sethwassistbits(if_t ifp, int toset, int toclear)
4199 {
4200 	((struct ifnet *)ifp)->if_hwassist |= toset;
4201 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4202 
4203 	return (0);
4204 }
4205 
4206 int
4207 if_sethwassist(if_t ifp, int hwassist_bit)
4208 {
4209 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4210 	return (0);
4211 }
4212 
4213 int
4214 if_gethwassist(if_t ifp)
4215 {
4216 	return ((struct ifnet *)ifp)->if_hwassist;
4217 }
4218 
4219 int
4220 if_setmtu(if_t ifp, int mtu)
4221 {
4222 	((struct ifnet *)ifp)->if_mtu = mtu;
4223 	return (0);
4224 }
4225 
4226 int
4227 if_getmtu(if_t ifp)
4228 {
4229 	return ((struct ifnet *)ifp)->if_mtu;
4230 }
4231 
4232 int
4233 if_getmtu_family(if_t ifp, int family)
4234 {
4235 	struct domain *dp;
4236 
4237 	for (dp = domains; dp; dp = dp->dom_next) {
4238 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4239 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4240 	}
4241 
4242 	return (((struct ifnet *)ifp)->if_mtu);
4243 }
4244 
4245 /*
4246  * Methods for drivers to access interface unicast and multicast
4247  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4248  * 'struct ifmultiaddr'.
4249  */
4250 u_int
4251 if_lladdr_count(if_t ifp)
4252 {
4253 	struct epoch_tracker et;
4254 	struct ifaddr *ifa;
4255 	u_int count;
4256 
4257 	count = 0;
4258 	NET_EPOCH_ENTER(et);
4259 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4260 		if (ifa->ifa_addr->sa_family == AF_LINK)
4261 			count++;
4262 	NET_EPOCH_EXIT(et);
4263 
4264 	return (count);
4265 }
4266 
4267 u_int
4268 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4269 {
4270 	struct epoch_tracker et;
4271 	struct ifaddr *ifa;
4272 	u_int count;
4273 
4274 	MPASS(cb);
4275 
4276 	count = 0;
4277 	NET_EPOCH_ENTER(et);
4278 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4279 		if (ifa->ifa_addr->sa_family != AF_LINK)
4280 			continue;
4281 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4282 		    count);
4283 	}
4284 	NET_EPOCH_EXIT(et);
4285 
4286 	return (count);
4287 }
4288 
4289 u_int
4290 if_llmaddr_count(if_t ifp)
4291 {
4292 	struct epoch_tracker et;
4293 	struct ifmultiaddr *ifma;
4294 	int count;
4295 
4296 	count = 0;
4297 	NET_EPOCH_ENTER(et);
4298 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4299 		if (ifma->ifma_addr->sa_family == AF_LINK)
4300 			count++;
4301 	NET_EPOCH_EXIT(et);
4302 
4303 	return (count);
4304 }
4305 
4306 u_int
4307 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4308 {
4309 	struct epoch_tracker et;
4310 	struct ifmultiaddr *ifma;
4311 	u_int count;
4312 
4313 	MPASS(cb);
4314 
4315 	count = 0;
4316 	NET_EPOCH_ENTER(et);
4317 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4318 		if (ifma->ifma_addr->sa_family != AF_LINK)
4319 			continue;
4320 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4321 		    count);
4322 	}
4323 	NET_EPOCH_EXIT(et);
4324 
4325 	return (count);
4326 }
4327 
4328 int
4329 if_setsoftc(if_t ifp, void *softc)
4330 {
4331 	((struct ifnet *)ifp)->if_softc = softc;
4332 	return (0);
4333 }
4334 
4335 void *
4336 if_getsoftc(if_t ifp)
4337 {
4338 	return ((struct ifnet *)ifp)->if_softc;
4339 }
4340 
4341 void
4342 if_setrcvif(struct mbuf *m, if_t ifp)
4343 {
4344 
4345 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4346 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4347 }
4348 
4349 void
4350 if_setvtag(struct mbuf *m, uint16_t tag)
4351 {
4352 	m->m_pkthdr.ether_vtag = tag;
4353 }
4354 
4355 uint16_t
4356 if_getvtag(struct mbuf *m)
4357 {
4358 
4359 	return (m->m_pkthdr.ether_vtag);
4360 }
4361 
4362 int
4363 if_sendq_empty(if_t ifp)
4364 {
4365 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4366 }
4367 
4368 struct ifaddr *
4369 if_getifaddr(if_t ifp)
4370 {
4371 	return ((struct ifnet *)ifp)->if_addr;
4372 }
4373 
4374 int
4375 if_getamcount(if_t ifp)
4376 {
4377 	return ((struct ifnet *)ifp)->if_amcount;
4378 }
4379 
4380 
4381 int
4382 if_setsendqready(if_t ifp)
4383 {
4384 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4385 	return (0);
4386 }
4387 
4388 int
4389 if_setsendqlen(if_t ifp, int tx_desc_count)
4390 {
4391 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4392 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4393 
4394 	return (0);
4395 }
4396 
4397 int
4398 if_vlantrunkinuse(if_t ifp)
4399 {
4400 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4401 }
4402 
4403 int
4404 if_input(if_t ifp, struct mbuf* sendmp)
4405 {
4406 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4407 	return (0);
4408 
4409 }
4410 
4411 struct mbuf *
4412 if_dequeue(if_t ifp)
4413 {
4414 	struct mbuf *m;
4415 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4416 
4417 	return (m);
4418 }
4419 
4420 int
4421 if_sendq_prepend(if_t ifp, struct mbuf *m)
4422 {
4423 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4424 	return (0);
4425 }
4426 
4427 int
4428 if_setifheaderlen(if_t ifp, int len)
4429 {
4430 	((struct ifnet *)ifp)->if_hdrlen = len;
4431 	return (0);
4432 }
4433 
4434 caddr_t
4435 if_getlladdr(if_t ifp)
4436 {
4437 	return (IF_LLADDR((struct ifnet *)ifp));
4438 }
4439 
4440 void *
4441 if_gethandle(u_char type)
4442 {
4443 	return (if_alloc(type));
4444 }
4445 
4446 void
4447 if_bpfmtap(if_t ifh, struct mbuf *m)
4448 {
4449 	struct ifnet *ifp = (struct ifnet *)ifh;
4450 
4451 	BPF_MTAP(ifp, m);
4452 }
4453 
4454 void
4455 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4456 {
4457 	struct ifnet *ifp = (struct ifnet *)ifh;
4458 
4459 	ETHER_BPF_MTAP(ifp, m);
4460 }
4461 
4462 void
4463 if_vlancap(if_t ifh)
4464 {
4465 	struct ifnet *ifp = (struct ifnet *)ifh;
4466 	VLAN_CAPABILITIES(ifp);
4467 }
4468 
4469 int
4470 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4471 {
4472 
4473 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4474         return (0);
4475 }
4476 
4477 int
4478 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4479 {
4480 
4481 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4482         return (0);
4483 }
4484 
4485 int
4486 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4487 {
4488 
4489 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4490         return (0);
4491 }
4492 
4493 u_int
4494 if_gethwtsomax(if_t ifp)
4495 {
4496 
4497 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4498 }
4499 
4500 u_int
4501 if_gethwtsomaxsegcount(if_t ifp)
4502 {
4503 
4504 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4505 }
4506 
4507 u_int
4508 if_gethwtsomaxsegsize(if_t ifp)
4509 {
4510 
4511 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4512 }
4513 
4514 void
4515 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4516 {
4517 	((struct ifnet *)ifp)->if_init = init_fn;
4518 }
4519 
4520 void
4521 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4522 {
4523 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4524 }
4525 
4526 void
4527 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4528 {
4529 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4530 }
4531 
4532 void
4533 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4534 {
4535 	((struct ifnet *)ifp)->if_transmit = start_fn;
4536 }
4537 
4538 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4539 {
4540 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4541 
4542 }
4543 
4544 void
4545 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4546 {
4547 
4548 	ifp->if_get_counter = fn;
4549 }
4550 
4551 /* Revisit these - These are inline functions originally. */
4552 int
4553 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4554 {
4555 	return drbr_inuse(ifh, br);
4556 }
4557 
4558 struct mbuf*
4559 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4560 {
4561 	return drbr_dequeue(ifh, br);
4562 }
4563 
4564 int
4565 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4566 {
4567 	return drbr_needs_enqueue(ifh, br);
4568 }
4569 
4570 int
4571 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4572 {
4573 	return drbr_enqueue(ifh, br, m);
4574 
4575 }
4576