xref: /freebsd/sys/net/if.c (revision bc37ac7ea8dd10039688d7cd2cef1d079fd025ac)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/conf.h>
40 #include <sys/eventhandler.h>
41 #include <sys/malloc.h>
42 #include <sys/domainset.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/epoch.h>
46 #include <sys/mbuf.h>
47 #include <sys/systm.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/protosw.h>
53 #include <sys/kernel.h>
54 #include <sys/lock.h>
55 #include <sys/refcount.h>
56 #include <sys/module.h>
57 #include <sys/rwlock.h>
58 #include <sys/sockio.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/taskqueue.h>
63 #include <sys/domain.h>
64 #include <sys/jail.h>
65 #include <sys/priv.h>
66 
67 #include <machine/stdarg.h>
68 #include <vm/uma.h>
69 
70 #include <net/bpf.h>
71 #include <net/ethernet.h>
72 #include <net/if.h>
73 #include <net/if_arp.h>
74 #include <net/if_clone.h>
75 #include <net/if_dl.h>
76 #include <net/if_types.h>
77 #include <net/if_var.h>
78 #include <net/if_media.h>
79 #include <net/if_vlan_var.h>
80 #include <net/radix.h>
81 #include <net/route.h>
82 #include <net/vnet.h>
83 
84 #if defined(INET) || defined(INET6)
85 #include <net/ethernet.h>
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_carp.h>
90 #ifdef INET
91 #include <netinet/if_ether.h>
92 #include <netinet/netdump/netdump.h>
93 #endif /* INET */
94 #ifdef INET6
95 #include <netinet6/in6_var.h>
96 #include <netinet6/in6_ifattach.h>
97 #endif /* INET6 */
98 #endif /* INET || INET6 */
99 
100 #include <security/mac/mac_framework.h>
101 
102 /*
103  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
104  * and ifr_ifru when it is used in SIOCGIFCONF.
105  */
106 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
107     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
108 
109 __read_mostly epoch_t net_epoch_preempt;
110 __read_mostly epoch_t net_epoch;
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, 0, "Link layers");
196 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
197 
198 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
199     &ifqmaxlen, 0, "max send queue size");
200 
201 /* Log link state change events */
202 static int log_link_state_change = 1;
203 
204 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
205 	&log_link_state_change, 0,
206 	"log interface link state change events");
207 
208 /* Log promiscuous mode change events */
209 static int log_promisc_mode_change = 1;
210 
211 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
212 	&log_promisc_mode_change, 1,
213 	"log promiscuous mode change events");
214 
215 /* Interface description */
216 static unsigned int ifdescr_maxlen = 1024;
217 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
218 	&ifdescr_maxlen, 0,
219 	"administrative maximum length for interface description");
220 
221 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
222 
223 /* global sx for non-critical path ifdescr */
224 static struct sx ifdescr_sx;
225 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
226 
227 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
228 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
229 /* These are external hooks for CARP. */
230 void	(*carp_linkstate_p)(struct ifnet *ifp);
231 void	(*carp_demote_adj_p)(int, char *);
232 int	(*carp_master_p)(struct ifaddr *);
233 #if defined(INET) || defined(INET6)
234 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
235 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
236     const struct sockaddr *sa);
237 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
238 int	(*carp_attach_p)(struct ifaddr *, int);
239 void	(*carp_detach_p)(struct ifaddr *, bool);
240 #endif
241 #ifdef INET
242 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
243 #endif
244 #ifdef INET6
245 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
246 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
247     const struct in6_addr *taddr);
248 #endif
249 
250 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
251 
252 /*
253  * XXX: Style; these should be sorted alphabetically, and unprototyped
254  * static functions should be prototyped. Currently they are sorted by
255  * declaration order.
256  */
257 static void	if_attachdomain(void *);
258 static void	if_attachdomain1(struct ifnet *);
259 static int	ifconf(u_long, caddr_t);
260 static void	*if_grow(void);
261 static void	if_input_default(struct ifnet *, struct mbuf *);
262 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
263 static void	if_route(struct ifnet *, int flag, int fam);
264 static int	if_setflag(struct ifnet *, int, int, int *, int);
265 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
266 static void	if_unroute(struct ifnet *, int flag, int fam);
267 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void	do_link_state_change(void *, int);
269 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int	if_getgroupmembers(struct ifgroupreq *);
271 static void	if_delgroups(struct ifnet *);
272 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
274 #ifdef VIMAGE
275 static void	if_vmove(struct ifnet *, struct vnet *);
276 #endif
277 
278 #ifdef INET6
279 /*
280  * XXX: declare here to avoid to include many inet6 related files..
281  * should be more generalized?
282  */
283 extern void	nd6_setmtu(struct ifnet *);
284 #endif
285 
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
289 
290 VNET_DEFINE(int, if_index);
291 int	ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
294 
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
296 
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
299 
300 #define	V_if_indexlim		VNET(if_indexlim)
301 #define	V_ifindex_table		VNET(ifindex_table)
302 
303 /*
304  * The global network interface list (V_ifnet) and related state (such as
305  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306  * an rwlock.  Either may be acquired shared to stablize the list, but both
307  * must be acquired writable to modify the list.  This model allows us to
308  * both stablize the interface list during interrupt thread processing, but
309  * also to stablize it over long-running ioctls, without introducing priority
310  * inversions and deadlocks.
311  */
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
316 
317 /*
318  * The allocation of network interfaces is a rather non-atomic affair; we
319  * need to select an index before we are ready to expose the interface for
320  * use, so will use this pointer value to indicate reservation.
321  */
322 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
323 
324 static	if_com_alloc_t *if_com_alloc[256];
325 static	if_com_free_t *if_com_free[256];
326 
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
330 
331 struct ifnet *
332 ifnet_byindex(u_short idx)
333 {
334 	struct ifnet *ifp;
335 
336 	if (__predict_false(idx > V_if_index))
337 		return (NULL);
338 
339 	ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
340 	return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
341 }
342 
343 struct ifnet *
344 ifnet_byindex_ref(u_short idx)
345 {
346 	struct ifnet *ifp;
347 
348 	NET_EPOCH_ASSERT();
349 
350 	ifp = ifnet_byindex(idx);
351 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
352 		return (NULL);
353 	if_ref(ifp);
354 	return (ifp);
355 }
356 
357 /*
358  * Allocate an ifindex array entry; return 0 on success or an error on
359  * failure.
360  */
361 static u_short
362 ifindex_alloc(void **old)
363 {
364 	u_short idx;
365 
366 	IFNET_WLOCK_ASSERT();
367 	/*
368 	 * Try to find an empty slot below V_if_index.  If we fail, take the
369 	 * next slot.
370 	 */
371 	for (idx = 1; idx <= V_if_index; idx++) {
372 		if (V_ifindex_table[idx] == NULL)
373 			break;
374 	}
375 
376 	/* Catch if_index overflow. */
377 	if (idx >= V_if_indexlim) {
378 		*old = if_grow();
379 		return (USHRT_MAX);
380 	}
381 	if (idx > V_if_index)
382 		V_if_index = idx;
383 	return (idx);
384 }
385 
386 static void
387 ifindex_free_locked(u_short idx)
388 {
389 
390 	IFNET_WLOCK_ASSERT();
391 
392 	V_ifindex_table[idx] = NULL;
393 	while (V_if_index > 0 &&
394 	    V_ifindex_table[V_if_index] == NULL)
395 		V_if_index--;
396 }
397 
398 static void
399 ifindex_free(u_short idx)
400 {
401 
402 	IFNET_WLOCK();
403 	ifindex_free_locked(idx);
404 	IFNET_WUNLOCK();
405 }
406 
407 static void
408 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
409 {
410 
411 	V_ifindex_table[idx] = ifp;
412 }
413 
414 struct ifaddr *
415 ifaddr_byindex(u_short idx)
416 {
417 	struct ifnet *ifp;
418 	struct ifaddr *ifa = NULL;
419 
420 	NET_EPOCH_ASSERT();
421 
422 	ifp = ifnet_byindex(idx);
423 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
424 		ifa_ref(ifa);
425 	return (ifa);
426 }
427 
428 /*
429  * Network interface utility routines.
430  *
431  * Routines with ifa_ifwith* names take sockaddr *'s as
432  * parameters.
433  */
434 
435 static void
436 vnet_if_init(const void *unused __unused)
437 {
438 	void *old;
439 
440 	CK_STAILQ_INIT(&V_ifnet);
441 	CK_STAILQ_INIT(&V_ifg_head);
442 	IFNET_WLOCK();
443 	old = if_grow();				/* create initial table */
444 	IFNET_WUNLOCK();
445 	epoch_wait_preempt(net_epoch_preempt);
446 	free(old, M_IFNET);
447 	vnet_if_clone_init();
448 }
449 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
450     NULL);
451 
452 #ifdef VIMAGE
453 static void
454 vnet_if_uninit(const void *unused __unused)
455 {
456 
457 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
458 	    "not empty", __func__, __LINE__, &V_ifnet));
459 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
460 	    "not empty", __func__, __LINE__, &V_ifg_head));
461 
462 	free((caddr_t)V_ifindex_table, M_IFNET);
463 }
464 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
465     vnet_if_uninit, NULL);
466 
467 static void
468 vnet_if_return(const void *unused __unused)
469 {
470 	struct ifnet *ifp, *nifp;
471 
472 	/* Return all inherited interfaces to their parent vnets. */
473 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
474 		if (ifp->if_home_vnet != ifp->if_vnet)
475 			if_vmove(ifp, ifp->if_home_vnet);
476 	}
477 }
478 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
479     vnet_if_return, NULL);
480 #endif
481 
482 
483 static void *
484 if_grow(void)
485 {
486 	int oldlim;
487 	u_int n;
488 	struct ifnet **e;
489 	void *old;
490 
491 	old = NULL;
492 	IFNET_WLOCK_ASSERT();
493 	oldlim = V_if_indexlim;
494 	IFNET_WUNLOCK();
495 	n = (oldlim << 1) * sizeof(*e);
496 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
497 	IFNET_WLOCK();
498 	if (V_if_indexlim != oldlim) {
499 		free(e, M_IFNET);
500 		return (NULL);
501 	}
502 	if (V_ifindex_table != NULL) {
503 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
504 		old = V_ifindex_table;
505 	}
506 	V_if_indexlim <<= 1;
507 	V_ifindex_table = e;
508 	return (old);
509 }
510 
511 /*
512  * Allocate a struct ifnet and an index for an interface.  A layer 2
513  * common structure will also be allocated if an allocation routine is
514  * registered for the passed type.
515  */
516 struct ifnet *
517 if_alloc_domain(u_char type, int numa_domain)
518 {
519 	struct ifnet *ifp;
520 	u_short idx;
521 	void *old;
522 
523 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
524 	if (numa_domain == IF_NODOM)
525 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
526 		    M_WAITOK | M_ZERO);
527 	else
528 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
529 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
530  restart:
531 	IFNET_WLOCK();
532 	idx = ifindex_alloc(&old);
533 	if (__predict_false(idx == USHRT_MAX)) {
534 		IFNET_WUNLOCK();
535 		epoch_wait_preempt(net_epoch_preempt);
536 		free(old, M_IFNET);
537 		goto restart;
538 	}
539 	ifnet_setbyindex(idx, IFNET_HOLD);
540 	IFNET_WUNLOCK();
541 	ifp->if_index = idx;
542 	ifp->if_type = type;
543 	ifp->if_alloctype = type;
544 	ifp->if_numa_domain = numa_domain;
545 #ifdef VIMAGE
546 	ifp->if_vnet = curvnet;
547 #endif
548 	if (if_com_alloc[type] != NULL) {
549 		ifp->if_l2com = if_com_alloc[type](type, ifp);
550 		if (ifp->if_l2com == NULL) {
551 			free(ifp, M_IFNET);
552 			ifindex_free(idx);
553 			return (NULL);
554 		}
555 	}
556 
557 	IF_ADDR_LOCK_INIT(ifp);
558 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
559 	ifp->if_afdata_initialized = 0;
560 	IF_AFDATA_LOCK_INIT(ifp);
561 	CK_STAILQ_INIT(&ifp->if_addrhead);
562 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
563 	CK_STAILQ_INIT(&ifp->if_groups);
564 #ifdef MAC
565 	mac_ifnet_init(ifp);
566 #endif
567 	ifq_init(&ifp->if_snd, ifp);
568 
569 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
570 	for (int i = 0; i < IFCOUNTERS; i++)
571 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
572 	ifp->if_get_counter = if_get_counter_default;
573 	ifp->if_pcp = IFNET_PCP_NONE;
574 	ifnet_setbyindex(ifp->if_index, ifp);
575 	return (ifp);
576 }
577 
578 struct ifnet *
579 if_alloc_dev(u_char type, device_t dev)
580 {
581 	int numa_domain;
582 
583 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
584 		return (if_alloc_domain(type, IF_NODOM));
585 	return (if_alloc_domain(type, numa_domain));
586 }
587 
588 struct ifnet *
589 if_alloc(u_char type)
590 {
591 
592 	return (if_alloc_domain(type, IF_NODOM));
593 }
594 /*
595  * Do the actual work of freeing a struct ifnet, and layer 2 common
596  * structure.  This call is made when the last reference to an
597  * interface is released.
598  */
599 static void
600 if_free_internal(struct ifnet *ifp)
601 {
602 
603 	KASSERT((ifp->if_flags & IFF_DYING),
604 	    ("if_free_internal: interface not dying"));
605 
606 	if (if_com_free[ifp->if_alloctype] != NULL)
607 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
608 		    ifp->if_alloctype);
609 
610 #ifdef MAC
611 	mac_ifnet_destroy(ifp);
612 #endif /* MAC */
613 	IF_AFDATA_DESTROY(ifp);
614 	IF_ADDR_LOCK_DESTROY(ifp);
615 	ifq_delete(&ifp->if_snd);
616 
617 	for (int i = 0; i < IFCOUNTERS; i++)
618 		counter_u64_free(ifp->if_counters[i]);
619 
620 	free(ifp->if_description, M_IFDESCR);
621 	free(ifp->if_hw_addr, M_IFADDR);
622 	if (ifp->if_numa_domain == IF_NODOM)
623 		free(ifp, M_IFNET);
624 	else
625 		free_domain(ifp, M_IFNET);
626 }
627 
628 static void
629 if_destroy(epoch_context_t ctx)
630 {
631 	struct ifnet *ifp;
632 
633 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
634 	if_free_internal(ifp);
635 }
636 
637 /*
638  * Deregister an interface and free the associated storage.
639  */
640 void
641 if_free(struct ifnet *ifp)
642 {
643 
644 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
645 
646 	CURVNET_SET_QUIET(ifp->if_vnet);
647 	IFNET_WLOCK();
648 	KASSERT(ifp == ifnet_byindex(ifp->if_index),
649 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
650 
651 	ifindex_free_locked(ifp->if_index);
652 	IFNET_WUNLOCK();
653 
654 	if (refcount_release(&ifp->if_refcount))
655 		epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
656 	CURVNET_RESTORE();
657 }
658 
659 /*
660  * Interfaces to keep an ifnet type-stable despite the possibility of the
661  * driver calling if_free().  If there are additional references, we defer
662  * freeing the underlying data structure.
663  */
664 void
665 if_ref(struct ifnet *ifp)
666 {
667 
668 	/* We don't assert the ifnet list lock here, but arguably should. */
669 	refcount_acquire(&ifp->if_refcount);
670 }
671 
672 void
673 if_rele(struct ifnet *ifp)
674 {
675 
676 	if (!refcount_release(&ifp->if_refcount))
677 		return;
678 	epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
679 }
680 
681 void
682 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
683 {
684 
685 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
686 
687 	if (ifq->ifq_maxlen == 0)
688 		ifq->ifq_maxlen = ifqmaxlen;
689 
690 	ifq->altq_type = 0;
691 	ifq->altq_disc = NULL;
692 	ifq->altq_flags &= ALTQF_CANTCHANGE;
693 	ifq->altq_tbr  = NULL;
694 	ifq->altq_ifp  = ifp;
695 }
696 
697 void
698 ifq_delete(struct ifaltq *ifq)
699 {
700 	mtx_destroy(&ifq->ifq_mtx);
701 }
702 
703 /*
704  * Perform generic interface initialization tasks and attach the interface
705  * to the list of "active" interfaces.  If vmove flag is set on entry
706  * to if_attach_internal(), perform only a limited subset of initialization
707  * tasks, given that we are moving from one vnet to another an ifnet which
708  * has already been fully initialized.
709  *
710  * Note that if_detach_internal() removes group membership unconditionally
711  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
712  * Thus, when if_vmove() is applied to a cloned interface, group membership
713  * is lost while a cloned one always joins a group whose name is
714  * ifc->ifc_name.  To recover this after if_detach_internal() and
715  * if_attach_internal(), the cloner should be specified to
716  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
717  * attempts to join a group whose name is ifc->ifc_name.
718  *
719  * XXX:
720  *  - The decision to return void and thus require this function to
721  *    succeed is questionable.
722  *  - We should probably do more sanity checking.  For instance we don't
723  *    do anything to insure if_xname is unique or non-empty.
724  */
725 void
726 if_attach(struct ifnet *ifp)
727 {
728 
729 	if_attach_internal(ifp, 0, NULL);
730 }
731 
732 /*
733  * Compute the least common TSO limit.
734  */
735 void
736 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
737 {
738 	/*
739 	 * 1) If there is no limit currently, take the limit from
740 	 * the network adapter.
741 	 *
742 	 * 2) If the network adapter has a limit below the current
743 	 * limit, apply it.
744 	 */
745 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
746 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
747 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
748 	}
749 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
750 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
751 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
752 	}
753 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
754 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
755 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
756 	}
757 }
758 
759 /*
760  * Update TSO limit of a network adapter.
761  *
762  * Returns zero if no change. Else non-zero.
763  */
764 int
765 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
766 {
767 	int retval = 0;
768 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
769 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
770 		retval++;
771 	}
772 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
773 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
774 		retval++;
775 	}
776 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
777 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
778 		retval++;
779 	}
780 	return (retval);
781 }
782 
783 static void
784 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
785 {
786 	unsigned socksize, ifasize;
787 	int namelen, masklen;
788 	struct sockaddr_dl *sdl;
789 	struct ifaddr *ifa;
790 
791 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
792 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
793 		    ifp->if_xname);
794 
795 #ifdef VIMAGE
796 	ifp->if_vnet = curvnet;
797 	if (ifp->if_home_vnet == NULL)
798 		ifp->if_home_vnet = curvnet;
799 #endif
800 
801 	if_addgroup(ifp, IFG_ALL);
802 
803 	/* Restore group membership for cloned interfaces. */
804 	if (vmove && ifc != NULL)
805 		if_clone_addgroup(ifp, ifc);
806 
807 	getmicrotime(&ifp->if_lastchange);
808 	ifp->if_epoch = time_uptime;
809 
810 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
811 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
812 	    ("transmit and qflush must both either be set or both be NULL"));
813 	if (ifp->if_transmit == NULL) {
814 		ifp->if_transmit = if_transmit;
815 		ifp->if_qflush = if_qflush;
816 	}
817 	if (ifp->if_input == NULL)
818 		ifp->if_input = if_input_default;
819 
820 	if (ifp->if_requestencap == NULL)
821 		ifp->if_requestencap = if_requestencap_default;
822 
823 	if (!vmove) {
824 #ifdef MAC
825 		mac_ifnet_create(ifp);
826 #endif
827 
828 		/*
829 		 * Create a Link Level name for this device.
830 		 */
831 		namelen = strlen(ifp->if_xname);
832 		/*
833 		 * Always save enough space for any possiable name so we
834 		 * can do a rename in place later.
835 		 */
836 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
837 		socksize = masklen + ifp->if_addrlen;
838 		if (socksize < sizeof(*sdl))
839 			socksize = sizeof(*sdl);
840 		socksize = roundup2(socksize, sizeof(long));
841 		ifasize = sizeof(*ifa) + 2 * socksize;
842 		ifa = ifa_alloc(ifasize, M_WAITOK);
843 		sdl = (struct sockaddr_dl *)(ifa + 1);
844 		sdl->sdl_len = socksize;
845 		sdl->sdl_family = AF_LINK;
846 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
847 		sdl->sdl_nlen = namelen;
848 		sdl->sdl_index = ifp->if_index;
849 		sdl->sdl_type = ifp->if_type;
850 		ifp->if_addr = ifa;
851 		ifa->ifa_ifp = ifp;
852 		ifa->ifa_addr = (struct sockaddr *)sdl;
853 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
854 		ifa->ifa_netmask = (struct sockaddr *)sdl;
855 		sdl->sdl_len = masklen;
856 		while (namelen != 0)
857 			sdl->sdl_data[--namelen] = 0xff;
858 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
859 		/* Reliably crash if used uninitialized. */
860 		ifp->if_broadcastaddr = NULL;
861 
862 		if (ifp->if_type == IFT_ETHER) {
863 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
864 			    M_WAITOK | M_ZERO);
865 		}
866 
867 #if defined(INET) || defined(INET6)
868 		/* Use defaults for TSO, if nothing is set */
869 		if (ifp->if_hw_tsomax == 0 &&
870 		    ifp->if_hw_tsomaxsegcount == 0 &&
871 		    ifp->if_hw_tsomaxsegsize == 0) {
872 			/*
873 			 * The TSO defaults needs to be such that an
874 			 * NFS mbuf list of 35 mbufs totalling just
875 			 * below 64K works and that a chain of mbufs
876 			 * can be defragged into at most 32 segments:
877 			 */
878 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
879 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
880 			ifp->if_hw_tsomaxsegcount = 35;
881 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
882 
883 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
884 			if (ifp->if_capabilities & IFCAP_TSO) {
885 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
886 				    ifp->if_hw_tsomax,
887 				    ifp->if_hw_tsomaxsegcount,
888 				    ifp->if_hw_tsomaxsegsize);
889 			}
890 		}
891 #endif
892 	}
893 #ifdef VIMAGE
894 	else {
895 		/*
896 		 * Update the interface index in the link layer address
897 		 * of the interface.
898 		 */
899 		for (ifa = ifp->if_addr; ifa != NULL;
900 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
901 			if (ifa->ifa_addr->sa_family == AF_LINK) {
902 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
903 				sdl->sdl_index = ifp->if_index;
904 			}
905 		}
906 	}
907 #endif
908 
909 	IFNET_WLOCK();
910 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
911 #ifdef VIMAGE
912 	curvnet->vnet_ifcnt++;
913 #endif
914 	IFNET_WUNLOCK();
915 
916 	if (domain_init_status >= 2)
917 		if_attachdomain1(ifp);
918 
919 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
920 	if (IS_DEFAULT_VNET(curvnet))
921 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
922 
923 	/* Announce the interface. */
924 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
925 }
926 
927 static void
928 if_epochalloc(void *dummy __unused)
929 {
930 
931 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
932 	net_epoch = epoch_alloc("Net", 0);
933 }
934 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
935     if_epochalloc, NULL);
936 
937 static void
938 if_attachdomain(void *dummy)
939 {
940 	struct ifnet *ifp;
941 
942 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
943 		if_attachdomain1(ifp);
944 }
945 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
946     if_attachdomain, NULL);
947 
948 static void
949 if_attachdomain1(struct ifnet *ifp)
950 {
951 	struct domain *dp;
952 
953 	/*
954 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
955 	 * cannot lock ifp->if_afdata initialization, entirely.
956 	 */
957 	IF_AFDATA_LOCK(ifp);
958 	if (ifp->if_afdata_initialized >= domain_init_status) {
959 		IF_AFDATA_UNLOCK(ifp);
960 		log(LOG_WARNING, "%s called more than once on %s\n",
961 		    __func__, ifp->if_xname);
962 		return;
963 	}
964 	ifp->if_afdata_initialized = domain_init_status;
965 	IF_AFDATA_UNLOCK(ifp);
966 
967 	/* address family dependent data region */
968 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
969 	for (dp = domains; dp; dp = dp->dom_next) {
970 		if (dp->dom_ifattach)
971 			ifp->if_afdata[dp->dom_family] =
972 			    (*dp->dom_ifattach)(ifp);
973 	}
974 }
975 
976 /*
977  * Remove any unicast or broadcast network addresses from an interface.
978  */
979 void
980 if_purgeaddrs(struct ifnet *ifp)
981 {
982 	struct ifaddr *ifa;
983 
984 	while (1) {
985 		struct epoch_tracker et;
986 
987 		NET_EPOCH_ENTER(et);
988 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
989 			if (ifa->ifa_addr->sa_family != AF_LINK)
990 				break;
991 		}
992 		NET_EPOCH_EXIT(et);
993 
994 		if (ifa == NULL)
995 			break;
996 #ifdef INET
997 		/* XXX: Ugly!! ad hoc just for INET */
998 		if (ifa->ifa_addr->sa_family == AF_INET) {
999 			struct ifaliasreq ifr;
1000 
1001 			bzero(&ifr, sizeof(ifr));
1002 			ifr.ifra_addr = *ifa->ifa_addr;
1003 			if (ifa->ifa_dstaddr)
1004 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1005 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1006 			    NULL) == 0)
1007 				continue;
1008 		}
1009 #endif /* INET */
1010 #ifdef INET6
1011 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1012 			in6_purgeaddr(ifa);
1013 			/* ifp_addrhead is already updated */
1014 			continue;
1015 		}
1016 #endif /* INET6 */
1017 		IF_ADDR_WLOCK(ifp);
1018 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1019 		IF_ADDR_WUNLOCK(ifp);
1020 		ifa_free(ifa);
1021 	}
1022 }
1023 
1024 /*
1025  * Remove any multicast network addresses from an interface when an ifnet
1026  * is going away.
1027  */
1028 static void
1029 if_purgemaddrs(struct ifnet *ifp)
1030 {
1031 	struct ifmultiaddr *ifma;
1032 
1033 	IF_ADDR_WLOCK(ifp);
1034 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1035 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1036 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1037 		if_delmulti_locked(ifp, ifma, 1);
1038 	}
1039 	IF_ADDR_WUNLOCK(ifp);
1040 }
1041 
1042 /*
1043  * Detach an interface, removing it from the list of "active" interfaces.
1044  * If vmove flag is set on entry to if_detach_internal(), perform only a
1045  * limited subset of cleanup tasks, given that we are moving an ifnet from
1046  * one vnet to another, where it must be fully operational.
1047  *
1048  * XXXRW: There are some significant questions about event ordering, and
1049  * how to prevent things from starting to use the interface during detach.
1050  */
1051 void
1052 if_detach(struct ifnet *ifp)
1053 {
1054 
1055 	CURVNET_SET_QUIET(ifp->if_vnet);
1056 	if_detach_internal(ifp, 0, NULL);
1057 	CURVNET_RESTORE();
1058 }
1059 
1060 /*
1061  * The vmove flag, if set, indicates that we are called from a callpath
1062  * that is moving an interface to a different vnet instance.
1063  *
1064  * The shutdown flag, if set, indicates that we are called in the
1065  * process of shutting down a vnet instance.  Currently only the
1066  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1067  * on a vnet instance shutdown without this flag being set, e.g., when
1068  * the cloned interfaces are destoyed as first thing of teardown.
1069  */
1070 static int
1071 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1072 {
1073 	struct ifaddr *ifa;
1074 	int i;
1075 	struct domain *dp;
1076  	struct ifnet *iter;
1077  	int found = 0;
1078 #ifdef VIMAGE
1079 	bool shutdown;
1080 
1081 	shutdown = ifp->if_vnet->vnet_shutdown;
1082 #endif
1083 	IFNET_WLOCK();
1084 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1085 		if (iter == ifp) {
1086 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1087 			if (!vmove)
1088 				ifp->if_flags |= IFF_DYING;
1089 			found = 1;
1090 			break;
1091 		}
1092 	IFNET_WUNLOCK();
1093 	if (!found) {
1094 		/*
1095 		 * While we would want to panic here, we cannot
1096 		 * guarantee that the interface is indeed still on
1097 		 * the list given we don't hold locks all the way.
1098 		 */
1099 		return (ENOENT);
1100 #if 0
1101 		if (vmove)
1102 			panic("%s: ifp=%p not on the ifnet tailq %p",
1103 			    __func__, ifp, &V_ifnet);
1104 		else
1105 			return; /* XXX this should panic as well? */
1106 #endif
1107 	}
1108 
1109 	/*
1110 	 * At this point we know the interface still was on the ifnet list
1111 	 * and we removed it so we are in a stable state.
1112 	 */
1113 #ifdef VIMAGE
1114 	curvnet->vnet_ifcnt--;
1115 #endif
1116 	epoch_wait_preempt(net_epoch_preempt);
1117 
1118 	/*
1119 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1120 	 * fixes issues about late destruction of multicast options
1121 	 * which lead to leave group calls, which in turn access the
1122 	 * belonging ifnet structure:
1123 	 */
1124 	epoch_drain_callbacks(net_epoch_preempt);
1125 
1126 	/*
1127 	 * In any case (destroy or vmove) detach us from the groups
1128 	 * and remove/wait for pending events on the taskq.
1129 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1130 	 */
1131 	if_delgroups(ifp);
1132 
1133 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1134 
1135 	/*
1136 	 * Check if this is a cloned interface or not. Must do even if
1137 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1138 	 * the if_clone_addgroup() will have a corrupted string overwise
1139 	 * from a gibberish pointer.
1140 	 */
1141 	if (vmove && ifcp != NULL)
1142 		*ifcp = if_clone_findifc(ifp);
1143 
1144 	if_down(ifp);
1145 
1146 #ifdef VIMAGE
1147 	/*
1148 	 * On VNET shutdown abort here as the stack teardown will do all
1149 	 * the work top-down for us.
1150 	 */
1151 	if (shutdown) {
1152 		/* Give interface users the chance to clean up. */
1153 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1154 
1155 		/*
1156 		 * In case of a vmove we are done here without error.
1157 		 * If we would signal an error it would lead to the same
1158 		 * abort as if we did not find the ifnet anymore.
1159 		 * if_detach() calls us in void context and does not care
1160 		 * about an early abort notification, so life is splendid :)
1161 		 */
1162 		goto finish_vnet_shutdown;
1163 	}
1164 #endif
1165 
1166 	/*
1167 	 * At this point we are not tearing down a VNET and are either
1168 	 * going to destroy or vmove the interface and have to cleanup
1169 	 * accordingly.
1170 	 */
1171 
1172 	/*
1173 	 * Remove routes and flush queues.
1174 	 */
1175 #ifdef ALTQ
1176 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1177 		altq_disable(&ifp->if_snd);
1178 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1179 		altq_detach(&ifp->if_snd);
1180 #endif
1181 
1182 	if_purgeaddrs(ifp);
1183 
1184 #ifdef INET
1185 	in_ifdetach(ifp);
1186 #endif
1187 
1188 #ifdef INET6
1189 	/*
1190 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1191 	 * before removing routing entries below, since IPv6 interface direct
1192 	 * routes are expected to be removed by the IPv6-specific kernel API.
1193 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1194 	 */
1195 	in6_ifdetach(ifp);
1196 #endif
1197 	if_purgemaddrs(ifp);
1198 
1199 	/* Announce that the interface is gone. */
1200 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1201 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1202 	if (IS_DEFAULT_VNET(curvnet))
1203 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1204 
1205 	if (!vmove) {
1206 		/*
1207 		 * Prevent further calls into the device driver via ifnet.
1208 		 */
1209 		if_dead(ifp);
1210 
1211 		/*
1212 		 * Clean up all addresses.
1213 		 */
1214 		IF_ADDR_WLOCK(ifp);
1215 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1216 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1217 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1218 			IF_ADDR_WUNLOCK(ifp);
1219 			ifa_free(ifa);
1220 		} else
1221 			IF_ADDR_WUNLOCK(ifp);
1222 	}
1223 
1224 	rt_flushifroutes(ifp);
1225 
1226 #ifdef VIMAGE
1227 finish_vnet_shutdown:
1228 #endif
1229 	/*
1230 	 * We cannot hold the lock over dom_ifdetach calls as they might
1231 	 * sleep, for example trying to drain a callout, thus open up the
1232 	 * theoretical race with re-attaching.
1233 	 */
1234 	IF_AFDATA_LOCK(ifp);
1235 	i = ifp->if_afdata_initialized;
1236 	ifp->if_afdata_initialized = 0;
1237 	IF_AFDATA_UNLOCK(ifp);
1238 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1239 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1240 			(*dp->dom_ifdetach)(ifp,
1241 			    ifp->if_afdata[dp->dom_family]);
1242 			ifp->if_afdata[dp->dom_family] = NULL;
1243 		}
1244 	}
1245 
1246 	return (0);
1247 }
1248 
1249 #ifdef VIMAGE
1250 /*
1251  * if_vmove() performs a limited version of if_detach() in current
1252  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1253  * An attempt is made to shrink if_index in current vnet, find an
1254  * unused if_index in target vnet and calls if_grow() if necessary,
1255  * and finally find an unused if_xname for the target vnet.
1256  */
1257 static void
1258 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1259 {
1260 	struct if_clone *ifc;
1261 	u_int bif_dlt, bif_hdrlen;
1262 	void *old;
1263 	int rc;
1264 
1265  	/*
1266 	 * if_detach_internal() will call the eventhandler to notify
1267 	 * interface departure.  That will detach if_bpf.  We need to
1268 	 * safe the dlt and hdrlen so we can re-attach it later.
1269 	 */
1270 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1271 
1272 	/*
1273 	 * Detach from current vnet, but preserve LLADDR info, do not
1274 	 * mark as dead etc. so that the ifnet can be reattached later.
1275 	 * If we cannot find it, we lost the race to someone else.
1276 	 */
1277 	rc = if_detach_internal(ifp, 1, &ifc);
1278 	if (rc != 0)
1279 		return;
1280 
1281 	/*
1282 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1283 	 * the if_index for that vnet if possible.
1284 	 *
1285 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1286 	 * or we'd lock on one vnet and unlock on another.
1287 	 */
1288 	IFNET_WLOCK();
1289 	ifindex_free_locked(ifp->if_index);
1290 	IFNET_WUNLOCK();
1291 
1292 	/*
1293 	 * Perform interface-specific reassignment tasks, if provided by
1294 	 * the driver.
1295 	 */
1296 	if (ifp->if_reassign != NULL)
1297 		ifp->if_reassign(ifp, new_vnet, NULL);
1298 
1299 	/*
1300 	 * Switch to the context of the target vnet.
1301 	 */
1302 	CURVNET_SET_QUIET(new_vnet);
1303  restart:
1304 	IFNET_WLOCK();
1305 	ifp->if_index = ifindex_alloc(&old);
1306 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1307 		IFNET_WUNLOCK();
1308 		epoch_wait_preempt(net_epoch_preempt);
1309 		free(old, M_IFNET);
1310 		goto restart;
1311 	}
1312 	ifnet_setbyindex(ifp->if_index, ifp);
1313 	IFNET_WUNLOCK();
1314 
1315 	if_attach_internal(ifp, 1, ifc);
1316 
1317 	if (ifp->if_bpf == NULL)
1318 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1319 
1320 	CURVNET_RESTORE();
1321 }
1322 
1323 /*
1324  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1325  */
1326 static int
1327 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1328 {
1329 	struct prison *pr;
1330 	struct ifnet *difp;
1331 
1332 	/* Try to find the prison within our visibility. */
1333 	sx_slock(&allprison_lock);
1334 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1335 	sx_sunlock(&allprison_lock);
1336 	if (pr == NULL)
1337 		return (ENXIO);
1338 	prison_hold_locked(pr);
1339 	mtx_unlock(&pr->pr_mtx);
1340 
1341 	/* Do not try to move the iface from and to the same prison. */
1342 	if (pr->pr_vnet == ifp->if_vnet) {
1343 		prison_free(pr);
1344 		return (EEXIST);
1345 	}
1346 
1347 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1348 	/* XXX Lock interfaces to avoid races. */
1349 	CURVNET_SET_QUIET(pr->pr_vnet);
1350 	difp = ifunit(ifname);
1351 	if (difp != NULL) {
1352 		CURVNET_RESTORE();
1353 		prison_free(pr);
1354 		return (EEXIST);
1355 	}
1356 
1357 	/* Make sure the VNET is stable. */
1358 	if (ifp->if_vnet->vnet_shutdown) {
1359 		CURVNET_RESTORE();
1360 		prison_free(pr);
1361 		return (EBUSY);
1362 	}
1363 	CURVNET_RESTORE();
1364 
1365 	/* Move the interface into the child jail/vnet. */
1366 	if_vmove(ifp, pr->pr_vnet);
1367 
1368 	/* Report the new if_xname back to the userland. */
1369 	sprintf(ifname, "%s", ifp->if_xname);
1370 
1371 	prison_free(pr);
1372 	return (0);
1373 }
1374 
1375 static int
1376 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1377 {
1378 	struct prison *pr;
1379 	struct vnet *vnet_dst;
1380 	struct ifnet *ifp;
1381 
1382 	/* Try to find the prison within our visibility. */
1383 	sx_slock(&allprison_lock);
1384 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1385 	sx_sunlock(&allprison_lock);
1386 	if (pr == NULL)
1387 		return (ENXIO);
1388 	prison_hold_locked(pr);
1389 	mtx_unlock(&pr->pr_mtx);
1390 
1391 	/* Make sure the named iface exists in the source prison/vnet. */
1392 	CURVNET_SET(pr->pr_vnet);
1393 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1394 	if (ifp == NULL) {
1395 		CURVNET_RESTORE();
1396 		prison_free(pr);
1397 		return (ENXIO);
1398 	}
1399 
1400 	/* Do not try to move the iface from and to the same prison. */
1401 	vnet_dst = TD_TO_VNET(td);
1402 	if (vnet_dst == ifp->if_vnet) {
1403 		CURVNET_RESTORE();
1404 		prison_free(pr);
1405 		return (EEXIST);
1406 	}
1407 
1408 	/* Make sure the VNET is stable. */
1409 	if (ifp->if_vnet->vnet_shutdown) {
1410 		CURVNET_RESTORE();
1411 		prison_free(pr);
1412 		return (EBUSY);
1413 	}
1414 
1415 	/* Get interface back from child jail/vnet. */
1416 	if_vmove(ifp, vnet_dst);
1417 	CURVNET_RESTORE();
1418 
1419 	/* Report the new if_xname back to the userland. */
1420 	sprintf(ifname, "%s", ifp->if_xname);
1421 
1422 	prison_free(pr);
1423 	return (0);
1424 }
1425 #endif /* VIMAGE */
1426 
1427 /*
1428  * Add a group to an interface
1429  */
1430 int
1431 if_addgroup(struct ifnet *ifp, const char *groupname)
1432 {
1433 	struct ifg_list		*ifgl;
1434 	struct ifg_group	*ifg = NULL;
1435 	struct ifg_member	*ifgm;
1436 	int 			 new = 0;
1437 
1438 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1439 	    groupname[strlen(groupname) - 1] <= '9')
1440 		return (EINVAL);
1441 
1442 	IFNET_WLOCK();
1443 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1444 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1445 			IFNET_WUNLOCK();
1446 			return (EEXIST);
1447 		}
1448 
1449 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1450 	    M_NOWAIT)) == NULL) {
1451 	    	IFNET_WUNLOCK();
1452 		return (ENOMEM);
1453 	}
1454 
1455 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1456 	    M_TEMP, M_NOWAIT)) == NULL) {
1457 		free(ifgl, M_TEMP);
1458 		IFNET_WUNLOCK();
1459 		return (ENOMEM);
1460 	}
1461 
1462 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1463 		if (!strcmp(ifg->ifg_group, groupname))
1464 			break;
1465 
1466 	if (ifg == NULL) {
1467 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1468 		    M_TEMP, M_NOWAIT)) == NULL) {
1469 			free(ifgl, M_TEMP);
1470 			free(ifgm, M_TEMP);
1471 			IFNET_WUNLOCK();
1472 			return (ENOMEM);
1473 		}
1474 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1475 		ifg->ifg_refcnt = 0;
1476 		CK_STAILQ_INIT(&ifg->ifg_members);
1477 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1478 		new = 1;
1479 	}
1480 
1481 	ifg->ifg_refcnt++;
1482 	ifgl->ifgl_group = ifg;
1483 	ifgm->ifgm_ifp = ifp;
1484 
1485 	IF_ADDR_WLOCK(ifp);
1486 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1487 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1488 	IF_ADDR_WUNLOCK(ifp);
1489 
1490 	IFNET_WUNLOCK();
1491 
1492 	if (new)
1493 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1494 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1495 
1496 	return (0);
1497 }
1498 
1499 /*
1500  * Remove a group from an interface
1501  */
1502 int
1503 if_delgroup(struct ifnet *ifp, const char *groupname)
1504 {
1505 	struct ifg_list		*ifgl;
1506 	struct ifg_member	*ifgm;
1507 	int freeifgl;
1508 
1509 	IFNET_WLOCK();
1510 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1511 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1512 			break;
1513 	if (ifgl == NULL) {
1514 		IFNET_WUNLOCK();
1515 		return (ENOENT);
1516 	}
1517 
1518 	freeifgl = 0;
1519 	IF_ADDR_WLOCK(ifp);
1520 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1521 	IF_ADDR_WUNLOCK(ifp);
1522 
1523 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1524 		if (ifgm->ifgm_ifp == ifp)
1525 			break;
1526 
1527 	if (ifgm != NULL)
1528 		CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1529 
1530 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1531 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1532 		freeifgl = 1;
1533 	}
1534 	IFNET_WUNLOCK();
1535 
1536 	epoch_wait_preempt(net_epoch_preempt);
1537 	if (freeifgl) {
1538 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1539 		free(ifgl->ifgl_group, M_TEMP);
1540 	}
1541 	free(ifgm, M_TEMP);
1542 	free(ifgl, M_TEMP);
1543 
1544 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1545 
1546 	return (0);
1547 }
1548 
1549 /*
1550  * Remove an interface from all groups
1551  */
1552 static void
1553 if_delgroups(struct ifnet *ifp)
1554 {
1555 	struct ifg_list		*ifgl;
1556 	struct ifg_member	*ifgm;
1557 	char groupname[IFNAMSIZ];
1558 	int ifglfree;
1559 
1560 	IFNET_WLOCK();
1561 	while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1562 		ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1563 
1564 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1565 
1566 		IF_ADDR_WLOCK(ifp);
1567 		CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1568 		IF_ADDR_WUNLOCK(ifp);
1569 
1570 		CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1571 			if (ifgm->ifgm_ifp == ifp)
1572 				break;
1573 
1574 		if (ifgm != NULL)
1575 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1576 			    ifgm_next);
1577 		ifglfree = 0;
1578 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1579 			CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1580 			ifglfree = 1;
1581 		}
1582 
1583 		IFNET_WUNLOCK();
1584 		epoch_wait_preempt(net_epoch_preempt);
1585 		free(ifgm, M_TEMP);
1586 		if (ifglfree) {
1587 			EVENTHANDLER_INVOKE(group_detach_event,
1588 								ifgl->ifgl_group);
1589 			free(ifgl->ifgl_group, M_TEMP);
1590 		}
1591 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1592 
1593 		IFNET_WLOCK();
1594 	}
1595 	IFNET_WUNLOCK();
1596 }
1597 
1598 static char *
1599 ifgr_group_get(void *ifgrp)
1600 {
1601 	union ifgroupreq_union *ifgrup;
1602 
1603 	ifgrup = ifgrp;
1604 #ifdef COMPAT_FREEBSD32
1605 	if (SV_CURPROC_FLAG(SV_ILP32))
1606 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1607 #endif
1608 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1609 }
1610 
1611 static struct ifg_req *
1612 ifgr_groups_get(void *ifgrp)
1613 {
1614 	union ifgroupreq_union *ifgrup;
1615 
1616 	ifgrup = ifgrp;
1617 #ifdef COMPAT_FREEBSD32
1618 	if (SV_CURPROC_FLAG(SV_ILP32))
1619 		return ((struct ifg_req *)(uintptr_t)
1620 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1621 #endif
1622 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1623 }
1624 
1625 /*
1626  * Stores all groups from an interface in memory pointed to by ifgr.
1627  */
1628 static int
1629 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1630 {
1631 	int			 len, error;
1632 	struct ifg_list		*ifgl;
1633 	struct ifg_req		 ifgrq, *ifgp;
1634 
1635 	NET_EPOCH_ASSERT();
1636 
1637 	if (ifgr->ifgr_len == 0) {
1638 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1639 			ifgr->ifgr_len += sizeof(struct ifg_req);
1640 		return (0);
1641 	}
1642 
1643 	len = ifgr->ifgr_len;
1644 	ifgp = ifgr_groups_get(ifgr);
1645 	/* XXX: wire */
1646 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1647 		if (len < sizeof(ifgrq))
1648 			return (EINVAL);
1649 		bzero(&ifgrq, sizeof ifgrq);
1650 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1651 		    sizeof(ifgrq.ifgrq_group));
1652 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1653 			return (error);
1654 		len -= sizeof(ifgrq);
1655 		ifgp++;
1656 	}
1657 
1658 	return (0);
1659 }
1660 
1661 /*
1662  * Stores all members of a group in memory pointed to by igfr
1663  */
1664 static int
1665 if_getgroupmembers(struct ifgroupreq *ifgr)
1666 {
1667 	struct ifg_group	*ifg;
1668 	struct ifg_member	*ifgm;
1669 	struct ifg_req		 ifgrq, *ifgp;
1670 	int			 len, error;
1671 
1672 	IFNET_RLOCK();
1673 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1674 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1675 			break;
1676 	if (ifg == NULL) {
1677 		IFNET_RUNLOCK();
1678 		return (ENOENT);
1679 	}
1680 
1681 	if (ifgr->ifgr_len == 0) {
1682 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1683 			ifgr->ifgr_len += sizeof(ifgrq);
1684 		IFNET_RUNLOCK();
1685 		return (0);
1686 	}
1687 
1688 	len = ifgr->ifgr_len;
1689 	ifgp = ifgr_groups_get(ifgr);
1690 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1691 		if (len < sizeof(ifgrq)) {
1692 			IFNET_RUNLOCK();
1693 			return (EINVAL);
1694 		}
1695 		bzero(&ifgrq, sizeof ifgrq);
1696 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1697 		    sizeof(ifgrq.ifgrq_member));
1698 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1699 			IFNET_RUNLOCK();
1700 			return (error);
1701 		}
1702 		len -= sizeof(ifgrq);
1703 		ifgp++;
1704 	}
1705 	IFNET_RUNLOCK();
1706 
1707 	return (0);
1708 }
1709 
1710 /*
1711  * Return counter values from counter(9)s stored in ifnet.
1712  */
1713 uint64_t
1714 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1715 {
1716 
1717 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1718 
1719 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1720 }
1721 
1722 /*
1723  * Increase an ifnet counter. Usually used for counters shared
1724  * between the stack and a driver, but function supports them all.
1725  */
1726 void
1727 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1728 {
1729 
1730 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1731 
1732 	counter_u64_add(ifp->if_counters[cnt], inc);
1733 }
1734 
1735 /*
1736  * Copy data from ifnet to userland API structure if_data.
1737  */
1738 void
1739 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1740 {
1741 
1742 	ifd->ifi_type = ifp->if_type;
1743 	ifd->ifi_physical = 0;
1744 	ifd->ifi_addrlen = ifp->if_addrlen;
1745 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1746 	ifd->ifi_link_state = ifp->if_link_state;
1747 	ifd->ifi_vhid = 0;
1748 	ifd->ifi_datalen = sizeof(struct if_data);
1749 	ifd->ifi_mtu = ifp->if_mtu;
1750 	ifd->ifi_metric = ifp->if_metric;
1751 	ifd->ifi_baudrate = ifp->if_baudrate;
1752 	ifd->ifi_hwassist = ifp->if_hwassist;
1753 	ifd->ifi_epoch = ifp->if_epoch;
1754 	ifd->ifi_lastchange = ifp->if_lastchange;
1755 
1756 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1757 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1758 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1759 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1760 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1761 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1762 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1763 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1764 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1765 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1766 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1767 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1768 }
1769 
1770 /*
1771  * Wrapper functions for struct ifnet address list locking macros.  These are
1772  * used by kernel modules to avoid encoding programming interface or binary
1773  * interface assumptions that may be violated when kernel-internal locking
1774  * approaches change.
1775  */
1776 void
1777 if_addr_rlock(struct ifnet *ifp)
1778 {
1779 
1780 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1781 }
1782 
1783 void
1784 if_addr_runlock(struct ifnet *ifp)
1785 {
1786 
1787 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1788 }
1789 
1790 void
1791 if_maddr_rlock(if_t ifp)
1792 {
1793 
1794 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1795 }
1796 
1797 void
1798 if_maddr_runlock(if_t ifp)
1799 {
1800 
1801 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1802 }
1803 
1804 /*
1805  * Initialization, destruction and refcounting functions for ifaddrs.
1806  */
1807 struct ifaddr *
1808 ifa_alloc(size_t size, int flags)
1809 {
1810 	struct ifaddr *ifa;
1811 
1812 	KASSERT(size >= sizeof(struct ifaddr),
1813 	    ("%s: invalid size %zu", __func__, size));
1814 
1815 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1816 	if (ifa == NULL)
1817 		return (NULL);
1818 
1819 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1820 		goto fail;
1821 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1822 		goto fail;
1823 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1824 		goto fail;
1825 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1826 		goto fail;
1827 
1828 	refcount_init(&ifa->ifa_refcnt, 1);
1829 
1830 	return (ifa);
1831 
1832 fail:
1833 	/* free(NULL) is okay */
1834 	counter_u64_free(ifa->ifa_opackets);
1835 	counter_u64_free(ifa->ifa_ipackets);
1836 	counter_u64_free(ifa->ifa_obytes);
1837 	counter_u64_free(ifa->ifa_ibytes);
1838 	free(ifa, M_IFADDR);
1839 
1840 	return (NULL);
1841 }
1842 
1843 void
1844 ifa_ref(struct ifaddr *ifa)
1845 {
1846 
1847 	refcount_acquire(&ifa->ifa_refcnt);
1848 }
1849 
1850 static void
1851 ifa_destroy(epoch_context_t ctx)
1852 {
1853 	struct ifaddr *ifa;
1854 
1855 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1856 	counter_u64_free(ifa->ifa_opackets);
1857 	counter_u64_free(ifa->ifa_ipackets);
1858 	counter_u64_free(ifa->ifa_obytes);
1859 	counter_u64_free(ifa->ifa_ibytes);
1860 	free(ifa, M_IFADDR);
1861 }
1862 
1863 void
1864 ifa_free(struct ifaddr *ifa)
1865 {
1866 
1867 	if (refcount_release(&ifa->ifa_refcnt))
1868 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1869 }
1870 
1871 
1872 static int
1873 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1874     struct sockaddr *ia)
1875 {
1876 	struct epoch_tracker et;
1877 	int error;
1878 	struct rt_addrinfo info;
1879 	struct sockaddr_dl null_sdl;
1880 	struct ifnet *ifp;
1881 
1882 	ifp = ifa->ifa_ifp;
1883 
1884 	bzero(&info, sizeof(info));
1885 	if (cmd != RTM_DELETE)
1886 		info.rti_ifp = V_loif;
1887 	if (cmd == RTM_ADD) {
1888 		/* explicitly specify (loopback) ifa */
1889 		if (info.rti_ifp != NULL) {
1890 			NET_EPOCH_ENTER(et);
1891 			info.rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1892 			if (info.rti_ifa != NULL)
1893 				ifa_ref(info.rti_ifa);
1894 			NET_EPOCH_EXIT(et);
1895 		}
1896 	}
1897 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1898 	info.rti_info[RTAX_DST] = ia;
1899 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1900 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1901 
1902 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1903 
1904 	if (error != 0 &&
1905 	    !(cmd == RTM_ADD && error == EEXIST) &&
1906 	    !(cmd == RTM_DELETE && error == ENOENT))
1907 		if_printf(ifp, "%s failed: %d\n", otype, error);
1908 
1909 	return (error);
1910 }
1911 
1912 int
1913 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1914 {
1915 
1916 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1917 }
1918 
1919 int
1920 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1921 {
1922 
1923 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1924 }
1925 
1926 int
1927 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1928 {
1929 
1930 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1931 }
1932 
1933 /*
1934  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1935  * structs used to represent other address families, it is necessary
1936  * to perform a different comparison.
1937  */
1938 
1939 #define	sa_dl_equal(a1, a2)	\
1940 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1941 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1942 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1943 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1944 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1945 
1946 /*
1947  * Locate an interface based on a complete address.
1948  */
1949 /*ARGSUSED*/
1950 struct ifaddr *
1951 ifa_ifwithaddr(const struct sockaddr *addr)
1952 {
1953 	struct ifnet *ifp;
1954 	struct ifaddr *ifa;
1955 
1956 	NET_EPOCH_ASSERT();
1957 
1958 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1959 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1960 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1961 				continue;
1962 			if (sa_equal(addr, ifa->ifa_addr)) {
1963 				goto done;
1964 			}
1965 			/* IP6 doesn't have broadcast */
1966 			if ((ifp->if_flags & IFF_BROADCAST) &&
1967 			    ifa->ifa_broadaddr &&
1968 			    ifa->ifa_broadaddr->sa_len != 0 &&
1969 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1970 				goto done;
1971 			}
1972 		}
1973 	}
1974 	ifa = NULL;
1975 done:
1976 	return (ifa);
1977 }
1978 
1979 int
1980 ifa_ifwithaddr_check(const struct sockaddr *addr)
1981 {
1982 	struct epoch_tracker et;
1983 	int rc;
1984 
1985 	NET_EPOCH_ENTER(et);
1986 	rc = (ifa_ifwithaddr(addr) != NULL);
1987 	NET_EPOCH_EXIT(et);
1988 	return (rc);
1989 }
1990 
1991 /*
1992  * Locate an interface based on the broadcast address.
1993  */
1994 /* ARGSUSED */
1995 struct ifaddr *
1996 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1997 {
1998 	struct ifnet *ifp;
1999 	struct ifaddr *ifa;
2000 
2001 	MPASS(in_epoch(net_epoch_preempt));
2002 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2003 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2004 			continue;
2005 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2006 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2007 				continue;
2008 			if ((ifp->if_flags & IFF_BROADCAST) &&
2009 			    ifa->ifa_broadaddr &&
2010 			    ifa->ifa_broadaddr->sa_len != 0 &&
2011 			    sa_equal(ifa->ifa_broadaddr, addr)) {
2012 				goto done;
2013 			}
2014 		}
2015 	}
2016 	ifa = NULL;
2017 done:
2018 	return (ifa);
2019 }
2020 
2021 /*
2022  * Locate the point to point interface with a given destination address.
2023  */
2024 /*ARGSUSED*/
2025 struct ifaddr *
2026 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2027 {
2028 	struct ifnet *ifp;
2029 	struct ifaddr *ifa;
2030 
2031 	MPASS(in_epoch(net_epoch_preempt));
2032 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2033 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2034 			continue;
2035 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2036 			continue;
2037 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2038 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2039 				continue;
2040 			if (ifa->ifa_dstaddr != NULL &&
2041 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2042 				goto done;
2043 			}
2044 		}
2045 	}
2046 	ifa = NULL;
2047 done:
2048 	return (ifa);
2049 }
2050 
2051 /*
2052  * Find an interface on a specific network.  If many, choice
2053  * is most specific found.
2054  */
2055 struct ifaddr *
2056 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2057 {
2058 	struct ifnet *ifp;
2059 	struct ifaddr *ifa;
2060 	struct ifaddr *ifa_maybe = NULL;
2061 	u_int af = addr->sa_family;
2062 	const char *addr_data = addr->sa_data, *cplim;
2063 
2064 	MPASS(in_epoch(net_epoch_preempt));
2065 	/*
2066 	 * AF_LINK addresses can be looked up directly by their index number,
2067 	 * so do that if we can.
2068 	 */
2069 	if (af == AF_LINK) {
2070 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2071 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2072 		return (ifaddr_byindex(sdl->sdl_index));
2073 	}
2074 
2075 	/*
2076 	 * Scan though each interface, looking for ones that have addresses
2077 	 * in this address family and the requested fib.
2078 	 */
2079 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2080 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2081 			continue;
2082 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2083 			const char *cp, *cp2, *cp3;
2084 
2085 			if (ifa->ifa_addr->sa_family != af)
2086 next:				continue;
2087 			if (af == AF_INET &&
2088 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2089 				/*
2090 				 * This is a bit broken as it doesn't
2091 				 * take into account that the remote end may
2092 				 * be a single node in the network we are
2093 				 * looking for.
2094 				 * The trouble is that we don't know the
2095 				 * netmask for the remote end.
2096 				 */
2097 				if (ifa->ifa_dstaddr != NULL &&
2098 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2099 					goto done;
2100 				}
2101 			} else {
2102 				/*
2103 				 * Scan all the bits in the ifa's address.
2104 				 * If a bit dissagrees with what we are
2105 				 * looking for, mask it with the netmask
2106 				 * to see if it really matters.
2107 				 * (A byte at a time)
2108 				 */
2109 				if (ifa->ifa_netmask == 0)
2110 					continue;
2111 				cp = addr_data;
2112 				cp2 = ifa->ifa_addr->sa_data;
2113 				cp3 = ifa->ifa_netmask->sa_data;
2114 				cplim = ifa->ifa_netmask->sa_len
2115 					+ (char *)ifa->ifa_netmask;
2116 				while (cp3 < cplim)
2117 					if ((*cp++ ^ *cp2++) & *cp3++)
2118 						goto next; /* next address! */
2119 				/*
2120 				 * If the netmask of what we just found
2121 				 * is more specific than what we had before
2122 				 * (if we had one), or if the virtual status
2123 				 * of new prefix is better than of the old one,
2124 				 * then remember the new one before continuing
2125 				 * to search for an even better one.
2126 				 */
2127 				if (ifa_maybe == NULL ||
2128 				    ifa_preferred(ifa_maybe, ifa) ||
2129 				    rn_refines((caddr_t)ifa->ifa_netmask,
2130 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2131 					ifa_maybe = ifa;
2132 				}
2133 			}
2134 		}
2135 	}
2136 	ifa = ifa_maybe;
2137 	ifa_maybe = NULL;
2138 done:
2139 	return (ifa);
2140 }
2141 
2142 /*
2143  * Find an interface address specific to an interface best matching
2144  * a given address.
2145  */
2146 struct ifaddr *
2147 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2148 {
2149 	struct ifaddr *ifa;
2150 	const char *cp, *cp2, *cp3;
2151 	char *cplim;
2152 	struct ifaddr *ifa_maybe = NULL;
2153 	u_int af = addr->sa_family;
2154 
2155 	if (af >= AF_MAX)
2156 		return (NULL);
2157 
2158 	MPASS(in_epoch(net_epoch_preempt));
2159 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2160 		if (ifa->ifa_addr->sa_family != af)
2161 			continue;
2162 		if (ifa_maybe == NULL)
2163 			ifa_maybe = ifa;
2164 		if (ifa->ifa_netmask == 0) {
2165 			if (sa_equal(addr, ifa->ifa_addr) ||
2166 			    (ifa->ifa_dstaddr &&
2167 			    sa_equal(addr, ifa->ifa_dstaddr)))
2168 				goto done;
2169 			continue;
2170 		}
2171 		if (ifp->if_flags & IFF_POINTOPOINT) {
2172 			if (sa_equal(addr, ifa->ifa_dstaddr))
2173 				goto done;
2174 		} else {
2175 			cp = addr->sa_data;
2176 			cp2 = ifa->ifa_addr->sa_data;
2177 			cp3 = ifa->ifa_netmask->sa_data;
2178 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2179 			for (; cp3 < cplim; cp3++)
2180 				if ((*cp++ ^ *cp2++) & *cp3)
2181 					break;
2182 			if (cp3 == cplim)
2183 				goto done;
2184 		}
2185 	}
2186 	ifa = ifa_maybe;
2187 done:
2188 	return (ifa);
2189 }
2190 
2191 /*
2192  * See whether new ifa is better than current one:
2193  * 1) A non-virtual one is preferred over virtual.
2194  * 2) A virtual in master state preferred over any other state.
2195  *
2196  * Used in several address selecting functions.
2197  */
2198 int
2199 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2200 {
2201 
2202 	return (cur->ifa_carp && (!next->ifa_carp ||
2203 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2204 }
2205 
2206 struct sockaddr_dl *
2207 link_alloc_sdl(size_t size, int flags)
2208 {
2209 
2210 	return (malloc(size, M_TEMP, flags));
2211 }
2212 
2213 void
2214 link_free_sdl(struct sockaddr *sa)
2215 {
2216 	free(sa, M_TEMP);
2217 }
2218 
2219 /*
2220  * Fills in given sdl with interface basic info.
2221  * Returns pointer to filled sdl.
2222  */
2223 struct sockaddr_dl *
2224 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2225 {
2226 	struct sockaddr_dl *sdl;
2227 
2228 	sdl = (struct sockaddr_dl *)paddr;
2229 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2230 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2231 	sdl->sdl_family = AF_LINK;
2232 	sdl->sdl_index = ifp->if_index;
2233 	sdl->sdl_type = iftype;
2234 
2235 	return (sdl);
2236 }
2237 
2238 /*
2239  * Mark an interface down and notify protocols of
2240  * the transition.
2241  */
2242 static void
2243 if_unroute(struct ifnet *ifp, int flag, int fam)
2244 {
2245 	struct ifaddr *ifa;
2246 
2247 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2248 
2249 	ifp->if_flags &= ~flag;
2250 	getmicrotime(&ifp->if_lastchange);
2251 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2252 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2253 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2254 	ifp->if_qflush(ifp);
2255 
2256 	if (ifp->if_carp)
2257 		(*carp_linkstate_p)(ifp);
2258 	rt_ifmsg(ifp);
2259 }
2260 
2261 /*
2262  * Mark an interface up and notify protocols of
2263  * the transition.
2264  */
2265 static void
2266 if_route(struct ifnet *ifp, int flag, int fam)
2267 {
2268 	struct ifaddr *ifa;
2269 
2270 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2271 
2272 	ifp->if_flags |= flag;
2273 	getmicrotime(&ifp->if_lastchange);
2274 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2275 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2276 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2277 	if (ifp->if_carp)
2278 		(*carp_linkstate_p)(ifp);
2279 	rt_ifmsg(ifp);
2280 #ifdef INET6
2281 	in6_if_up(ifp);
2282 #endif
2283 }
2284 
2285 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2286 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2287 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2288 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2289 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2290 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2291 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2292 void	*(*vlan_cookie_p)(struct ifnet *);
2293 
2294 /*
2295  * Handle a change in the interface link state. To avoid LORs
2296  * between driver lock and upper layer locks, as well as possible
2297  * recursions, we post event to taskqueue, and all job
2298  * is done in static do_link_state_change().
2299  */
2300 void
2301 if_link_state_change(struct ifnet *ifp, int link_state)
2302 {
2303 	/* Return if state hasn't changed. */
2304 	if (ifp->if_link_state == link_state)
2305 		return;
2306 
2307 	ifp->if_link_state = link_state;
2308 
2309 	/* XXXGL: reference ifp? */
2310 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2311 }
2312 
2313 static void
2314 do_link_state_change(void *arg, int pending)
2315 {
2316 	struct epoch_tracker et;
2317 	struct ifnet *ifp;
2318 	int link_state;
2319 
2320 	NET_EPOCH_ENTER(et);
2321 	ifp = arg;
2322 	link_state = ifp->if_link_state;
2323 
2324 	CURVNET_SET(ifp->if_vnet);
2325 	rt_ifmsg(ifp);
2326 	if (ifp->if_vlantrunk != NULL)
2327 		(*vlan_link_state_p)(ifp);
2328 
2329 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2330 	    ifp->if_l2com != NULL)
2331 		(*ng_ether_link_state_p)(ifp, link_state);
2332 	if (ifp->if_carp)
2333 		(*carp_linkstate_p)(ifp);
2334 	if (ifp->if_bridge)
2335 		ifp->if_bridge_linkstate(ifp);
2336 	if (ifp->if_lagg)
2337 		(*lagg_linkstate_p)(ifp, link_state);
2338 
2339 	if (IS_DEFAULT_VNET(curvnet))
2340 		devctl_notify("IFNET", ifp->if_xname,
2341 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2342 		    NULL);
2343 	if (pending > 1)
2344 		if_printf(ifp, "%d link states coalesced\n", pending);
2345 	if (log_link_state_change)
2346 		if_printf(ifp, "link state changed to %s\n",
2347 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2348 	NET_EPOCH_EXIT(et);
2349 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2350 	CURVNET_RESTORE();
2351 }
2352 
2353 /*
2354  * Mark an interface down and notify protocols of
2355  * the transition.
2356  */
2357 void
2358 if_down(struct ifnet *ifp)
2359 {
2360 
2361 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2362 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2363 }
2364 
2365 /*
2366  * Mark an interface up and notify protocols of
2367  * the transition.
2368  */
2369 void
2370 if_up(struct ifnet *ifp)
2371 {
2372 
2373 	if_route(ifp, IFF_UP, AF_UNSPEC);
2374 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2375 }
2376 
2377 /*
2378  * Flush an interface queue.
2379  */
2380 void
2381 if_qflush(struct ifnet *ifp)
2382 {
2383 	struct mbuf *m, *n;
2384 	struct ifaltq *ifq;
2385 
2386 	ifq = &ifp->if_snd;
2387 	IFQ_LOCK(ifq);
2388 #ifdef ALTQ
2389 	if (ALTQ_IS_ENABLED(ifq))
2390 		ALTQ_PURGE(ifq);
2391 #endif
2392 	n = ifq->ifq_head;
2393 	while ((m = n) != NULL) {
2394 		n = m->m_nextpkt;
2395 		m_freem(m);
2396 	}
2397 	ifq->ifq_head = 0;
2398 	ifq->ifq_tail = 0;
2399 	ifq->ifq_len = 0;
2400 	IFQ_UNLOCK(ifq);
2401 }
2402 
2403 /*
2404  * Map interface name to interface structure pointer, with or without
2405  * returning a reference.
2406  */
2407 struct ifnet *
2408 ifunit_ref(const char *name)
2409 {
2410 	struct epoch_tracker et;
2411 	struct ifnet *ifp;
2412 
2413 	NET_EPOCH_ENTER(et);
2414 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2415 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2416 		    !(ifp->if_flags & IFF_DYING))
2417 			break;
2418 	}
2419 	if (ifp != NULL)
2420 		if_ref(ifp);
2421 	NET_EPOCH_EXIT(et);
2422 	return (ifp);
2423 }
2424 
2425 struct ifnet *
2426 ifunit(const char *name)
2427 {
2428 	struct epoch_tracker et;
2429 	struct ifnet *ifp;
2430 
2431 	NET_EPOCH_ENTER(et);
2432 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2433 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2434 			break;
2435 	}
2436 	NET_EPOCH_EXIT(et);
2437 	return (ifp);
2438 }
2439 
2440 static void *
2441 ifr_buffer_get_buffer(void *data)
2442 {
2443 	union ifreq_union *ifrup;
2444 
2445 	ifrup = data;
2446 #ifdef COMPAT_FREEBSD32
2447 	if (SV_CURPROC_FLAG(SV_ILP32))
2448 		return ((void *)(uintptr_t)
2449 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2450 #endif
2451 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2452 }
2453 
2454 static void
2455 ifr_buffer_set_buffer_null(void *data)
2456 {
2457 	union ifreq_union *ifrup;
2458 
2459 	ifrup = data;
2460 #ifdef COMPAT_FREEBSD32
2461 	if (SV_CURPROC_FLAG(SV_ILP32))
2462 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2463 	else
2464 #endif
2465 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2466 }
2467 
2468 static size_t
2469 ifr_buffer_get_length(void *data)
2470 {
2471 	union ifreq_union *ifrup;
2472 
2473 	ifrup = data;
2474 #ifdef COMPAT_FREEBSD32
2475 	if (SV_CURPROC_FLAG(SV_ILP32))
2476 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2477 #endif
2478 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2479 }
2480 
2481 static void
2482 ifr_buffer_set_length(void *data, size_t len)
2483 {
2484 	union ifreq_union *ifrup;
2485 
2486 	ifrup = data;
2487 #ifdef COMPAT_FREEBSD32
2488 	if (SV_CURPROC_FLAG(SV_ILP32))
2489 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2490 	else
2491 #endif
2492 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2493 }
2494 
2495 void *
2496 ifr_data_get_ptr(void *ifrp)
2497 {
2498 	union ifreq_union *ifrup;
2499 
2500 	ifrup = ifrp;
2501 #ifdef COMPAT_FREEBSD32
2502 	if (SV_CURPROC_FLAG(SV_ILP32))
2503 		return ((void *)(uintptr_t)
2504 		    ifrup->ifr32.ifr_ifru.ifru_data);
2505 #endif
2506 		return (ifrup->ifr.ifr_ifru.ifru_data);
2507 }
2508 
2509 /*
2510  * Hardware specific interface ioctls.
2511  */
2512 int
2513 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2514 {
2515 	struct ifreq *ifr;
2516 	int error = 0, do_ifup = 0;
2517 	int new_flags, temp_flags;
2518 	size_t namelen, onamelen;
2519 	size_t descrlen;
2520 	char *descrbuf, *odescrbuf;
2521 	char new_name[IFNAMSIZ];
2522 	struct ifaddr *ifa;
2523 	struct sockaddr_dl *sdl;
2524 
2525 	ifr = (struct ifreq *)data;
2526 	switch (cmd) {
2527 	case SIOCGIFINDEX:
2528 		ifr->ifr_index = ifp->if_index;
2529 		break;
2530 
2531 	case SIOCGIFFLAGS:
2532 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2533 		ifr->ifr_flags = temp_flags & 0xffff;
2534 		ifr->ifr_flagshigh = temp_flags >> 16;
2535 		break;
2536 
2537 	case SIOCGIFCAP:
2538 		ifr->ifr_reqcap = ifp->if_capabilities;
2539 		ifr->ifr_curcap = ifp->if_capenable;
2540 		break;
2541 
2542 #ifdef MAC
2543 	case SIOCGIFMAC:
2544 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2545 		break;
2546 #endif
2547 
2548 	case SIOCGIFMETRIC:
2549 		ifr->ifr_metric = ifp->if_metric;
2550 		break;
2551 
2552 	case SIOCGIFMTU:
2553 		ifr->ifr_mtu = ifp->if_mtu;
2554 		break;
2555 
2556 	case SIOCGIFPHYS:
2557 		/* XXXGL: did this ever worked? */
2558 		ifr->ifr_phys = 0;
2559 		break;
2560 
2561 	case SIOCGIFDESCR:
2562 		error = 0;
2563 		sx_slock(&ifdescr_sx);
2564 		if (ifp->if_description == NULL)
2565 			error = ENOMSG;
2566 		else {
2567 			/* space for terminating nul */
2568 			descrlen = strlen(ifp->if_description) + 1;
2569 			if (ifr_buffer_get_length(ifr) < descrlen)
2570 				ifr_buffer_set_buffer_null(ifr);
2571 			else
2572 				error = copyout(ifp->if_description,
2573 				    ifr_buffer_get_buffer(ifr), descrlen);
2574 			ifr_buffer_set_length(ifr, descrlen);
2575 		}
2576 		sx_sunlock(&ifdescr_sx);
2577 		break;
2578 
2579 	case SIOCSIFDESCR:
2580 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2581 		if (error)
2582 			return (error);
2583 
2584 		/*
2585 		 * Copy only (length-1) bytes to make sure that
2586 		 * if_description is always nul terminated.  The
2587 		 * length parameter is supposed to count the
2588 		 * terminating nul in.
2589 		 */
2590 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2591 			return (ENAMETOOLONG);
2592 		else if (ifr_buffer_get_length(ifr) == 0)
2593 			descrbuf = NULL;
2594 		else {
2595 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2596 			    M_IFDESCR, M_WAITOK | M_ZERO);
2597 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2598 			    ifr_buffer_get_length(ifr) - 1);
2599 			if (error) {
2600 				free(descrbuf, M_IFDESCR);
2601 				break;
2602 			}
2603 		}
2604 
2605 		sx_xlock(&ifdescr_sx);
2606 		odescrbuf = ifp->if_description;
2607 		ifp->if_description = descrbuf;
2608 		sx_xunlock(&ifdescr_sx);
2609 
2610 		getmicrotime(&ifp->if_lastchange);
2611 		free(odescrbuf, M_IFDESCR);
2612 		break;
2613 
2614 	case SIOCGIFFIB:
2615 		ifr->ifr_fib = ifp->if_fib;
2616 		break;
2617 
2618 	case SIOCSIFFIB:
2619 		error = priv_check(td, PRIV_NET_SETIFFIB);
2620 		if (error)
2621 			return (error);
2622 		if (ifr->ifr_fib >= rt_numfibs)
2623 			return (EINVAL);
2624 
2625 		ifp->if_fib = ifr->ifr_fib;
2626 		break;
2627 
2628 	case SIOCSIFFLAGS:
2629 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2630 		if (error)
2631 			return (error);
2632 		/*
2633 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2634 		 * check, so we don't need special handling here yet.
2635 		 */
2636 		new_flags = (ifr->ifr_flags & 0xffff) |
2637 		    (ifr->ifr_flagshigh << 16);
2638 		if (ifp->if_flags & IFF_UP &&
2639 		    (new_flags & IFF_UP) == 0) {
2640 			if_down(ifp);
2641 		} else if (new_flags & IFF_UP &&
2642 		    (ifp->if_flags & IFF_UP) == 0) {
2643 			do_ifup = 1;
2644 		}
2645 		/* See if permanently promiscuous mode bit is about to flip */
2646 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2647 			if (new_flags & IFF_PPROMISC)
2648 				ifp->if_flags |= IFF_PROMISC;
2649 			else if (ifp->if_pcount == 0)
2650 				ifp->if_flags &= ~IFF_PROMISC;
2651 			if (log_promisc_mode_change)
2652                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2653                                     ((new_flags & IFF_PPROMISC) ?
2654                                      "enabled" : "disabled"));
2655 		}
2656 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2657 			(new_flags &~ IFF_CANTCHANGE);
2658 		if (ifp->if_ioctl) {
2659 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2660 		}
2661 		if (do_ifup)
2662 			if_up(ifp);
2663 		getmicrotime(&ifp->if_lastchange);
2664 		break;
2665 
2666 	case SIOCSIFCAP:
2667 		error = priv_check(td, PRIV_NET_SETIFCAP);
2668 		if (error)
2669 			return (error);
2670 		if (ifp->if_ioctl == NULL)
2671 			return (EOPNOTSUPP);
2672 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2673 			return (EINVAL);
2674 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2675 		if (error == 0)
2676 			getmicrotime(&ifp->if_lastchange);
2677 		break;
2678 
2679 #ifdef MAC
2680 	case SIOCSIFMAC:
2681 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2682 		break;
2683 #endif
2684 
2685 	case SIOCSIFNAME:
2686 		error = priv_check(td, PRIV_NET_SETIFNAME);
2687 		if (error)
2688 			return (error);
2689 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2690 		    NULL);
2691 		if (error != 0)
2692 			return (error);
2693 		if (new_name[0] == '\0')
2694 			return (EINVAL);
2695 		if (new_name[IFNAMSIZ-1] != '\0') {
2696 			new_name[IFNAMSIZ-1] = '\0';
2697 			if (strlen(new_name) == IFNAMSIZ-1)
2698 				return (EINVAL);
2699 		}
2700 		if (strcmp(new_name, ifp->if_xname) == 0)
2701 			break;
2702 		if (ifunit(new_name) != NULL)
2703 			return (EEXIST);
2704 
2705 		/*
2706 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2707 		 * and there are numerous other races with the
2708 		 * ifunit() checks not being atomic with namespace
2709 		 * changes (renames, vmoves, if_attach, etc).
2710 		 */
2711 		ifp->if_flags |= IFF_RENAMING;
2712 
2713 		/* Announce the departure of the interface. */
2714 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2715 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2716 
2717 		if_printf(ifp, "changing name to '%s'\n", new_name);
2718 
2719 		IF_ADDR_WLOCK(ifp);
2720 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2721 		ifa = ifp->if_addr;
2722 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2723 		namelen = strlen(new_name);
2724 		onamelen = sdl->sdl_nlen;
2725 		/*
2726 		 * Move the address if needed.  This is safe because we
2727 		 * allocate space for a name of length IFNAMSIZ when we
2728 		 * create this in if_attach().
2729 		 */
2730 		if (namelen != onamelen) {
2731 			bcopy(sdl->sdl_data + onamelen,
2732 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2733 		}
2734 		bcopy(new_name, sdl->sdl_data, namelen);
2735 		sdl->sdl_nlen = namelen;
2736 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2737 		bzero(sdl->sdl_data, onamelen);
2738 		while (namelen != 0)
2739 			sdl->sdl_data[--namelen] = 0xff;
2740 		IF_ADDR_WUNLOCK(ifp);
2741 
2742 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2743 		/* Announce the return of the interface. */
2744 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2745 
2746 		ifp->if_flags &= ~IFF_RENAMING;
2747 		break;
2748 
2749 #ifdef VIMAGE
2750 	case SIOCSIFVNET:
2751 		error = priv_check(td, PRIV_NET_SETIFVNET);
2752 		if (error)
2753 			return (error);
2754 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2755 		break;
2756 #endif
2757 
2758 	case SIOCSIFMETRIC:
2759 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2760 		if (error)
2761 			return (error);
2762 		ifp->if_metric = ifr->ifr_metric;
2763 		getmicrotime(&ifp->if_lastchange);
2764 		break;
2765 
2766 	case SIOCSIFPHYS:
2767 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2768 		if (error)
2769 			return (error);
2770 		if (ifp->if_ioctl == NULL)
2771 			return (EOPNOTSUPP);
2772 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2773 		if (error == 0)
2774 			getmicrotime(&ifp->if_lastchange);
2775 		break;
2776 
2777 	case SIOCSIFMTU:
2778 	{
2779 		u_long oldmtu = ifp->if_mtu;
2780 
2781 		error = priv_check(td, PRIV_NET_SETIFMTU);
2782 		if (error)
2783 			return (error);
2784 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2785 			return (EINVAL);
2786 		if (ifp->if_ioctl == NULL)
2787 			return (EOPNOTSUPP);
2788 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2789 		if (error == 0) {
2790 			getmicrotime(&ifp->if_lastchange);
2791 			rt_ifmsg(ifp);
2792 #ifdef INET
2793 			NETDUMP_REINIT(ifp);
2794 #endif
2795 		}
2796 		/*
2797 		 * If the link MTU changed, do network layer specific procedure.
2798 		 */
2799 		if (ifp->if_mtu != oldmtu) {
2800 #ifdef INET6
2801 			nd6_setmtu(ifp);
2802 #endif
2803 			rt_updatemtu(ifp);
2804 		}
2805 		break;
2806 	}
2807 
2808 	case SIOCADDMULTI:
2809 	case SIOCDELMULTI:
2810 		if (cmd == SIOCADDMULTI)
2811 			error = priv_check(td, PRIV_NET_ADDMULTI);
2812 		else
2813 			error = priv_check(td, PRIV_NET_DELMULTI);
2814 		if (error)
2815 			return (error);
2816 
2817 		/* Don't allow group membership on non-multicast interfaces. */
2818 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2819 			return (EOPNOTSUPP);
2820 
2821 		/* Don't let users screw up protocols' entries. */
2822 		if (ifr->ifr_addr.sa_family != AF_LINK)
2823 			return (EINVAL);
2824 
2825 		if (cmd == SIOCADDMULTI) {
2826 			struct epoch_tracker et;
2827 			struct ifmultiaddr *ifma;
2828 
2829 			/*
2830 			 * Userland is only permitted to join groups once
2831 			 * via the if_addmulti() KPI, because it cannot hold
2832 			 * struct ifmultiaddr * between calls. It may also
2833 			 * lose a race while we check if the membership
2834 			 * already exists.
2835 			 */
2836 			NET_EPOCH_ENTER(et);
2837 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2838 			NET_EPOCH_EXIT(et);
2839 			if (ifma != NULL)
2840 				error = EADDRINUSE;
2841 			else
2842 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2843 		} else {
2844 			error = if_delmulti(ifp, &ifr->ifr_addr);
2845 		}
2846 		if (error == 0)
2847 			getmicrotime(&ifp->if_lastchange);
2848 		break;
2849 
2850 	case SIOCSIFPHYADDR:
2851 	case SIOCDIFPHYADDR:
2852 #ifdef INET6
2853 	case SIOCSIFPHYADDR_IN6:
2854 #endif
2855 	case SIOCSIFMEDIA:
2856 	case SIOCSIFGENERIC:
2857 		error = priv_check(td, PRIV_NET_HWIOCTL);
2858 		if (error)
2859 			return (error);
2860 		if (ifp->if_ioctl == NULL)
2861 			return (EOPNOTSUPP);
2862 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2863 		if (error == 0)
2864 			getmicrotime(&ifp->if_lastchange);
2865 		break;
2866 
2867 	case SIOCGIFSTATUS:
2868 	case SIOCGIFPSRCADDR:
2869 	case SIOCGIFPDSTADDR:
2870 	case SIOCGIFMEDIA:
2871 	case SIOCGIFXMEDIA:
2872 	case SIOCGIFGENERIC:
2873 	case SIOCGIFRSSKEY:
2874 	case SIOCGIFRSSHASH:
2875 	case SIOCGIFDOWNREASON:
2876 		if (ifp->if_ioctl == NULL)
2877 			return (EOPNOTSUPP);
2878 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2879 		break;
2880 
2881 	case SIOCSIFLLADDR:
2882 		error = priv_check(td, PRIV_NET_SETLLADDR);
2883 		if (error)
2884 			return (error);
2885 		error = if_setlladdr(ifp,
2886 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2887 		break;
2888 
2889 	case SIOCGHWADDR:
2890 		error = if_gethwaddr(ifp, ifr);
2891 		break;
2892 
2893 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2894 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2895 		if (error)
2896 			return (error);
2897 		if ((error = if_addgroup(ifp,
2898 		    ifgr_group_get((struct ifgroupreq *)data))))
2899 			return (error);
2900 		break;
2901 
2902 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2903 	{
2904 		struct epoch_tracker et;
2905 
2906 		NET_EPOCH_ENTER(et);
2907 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2908 		NET_EPOCH_EXIT(et);
2909 		break;
2910 	}
2911 
2912 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2913 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2914 		if (error)
2915 			return (error);
2916 		if ((error = if_delgroup(ifp,
2917 		    ifgr_group_get((struct ifgroupreq *)data))))
2918 			return (error);
2919 		break;
2920 
2921 	default:
2922 		error = ENOIOCTL;
2923 		break;
2924 	}
2925 	return (error);
2926 }
2927 
2928 #ifdef COMPAT_FREEBSD32
2929 struct ifconf32 {
2930 	int32_t	ifc_len;
2931 	union {
2932 		uint32_t	ifcu_buf;
2933 		uint32_t	ifcu_req;
2934 	} ifc_ifcu;
2935 };
2936 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2937 #endif
2938 
2939 #ifdef COMPAT_FREEBSD32
2940 static void
2941 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2942 {
2943 	struct ifmediareq32 *ifmr32;
2944 
2945 	ifmr32 = (struct ifmediareq32 *)data;
2946 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2947 	    sizeof(ifmr->ifm_name));
2948 	ifmr->ifm_current = ifmr32->ifm_current;
2949 	ifmr->ifm_mask = ifmr32->ifm_mask;
2950 	ifmr->ifm_status = ifmr32->ifm_status;
2951 	ifmr->ifm_active = ifmr32->ifm_active;
2952 	ifmr->ifm_count = ifmr32->ifm_count;
2953 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2954 }
2955 
2956 static void
2957 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2958 {
2959 	struct ifmediareq32 *ifmr32;
2960 
2961 	ifmr32 = (struct ifmediareq32 *)data;
2962 	ifmr32->ifm_current = ifmr->ifm_current;
2963 	ifmr32->ifm_mask = ifmr->ifm_mask;
2964 	ifmr32->ifm_status = ifmr->ifm_status;
2965 	ifmr32->ifm_active = ifmr->ifm_active;
2966 	ifmr32->ifm_count = ifmr->ifm_count;
2967 }
2968 #endif
2969 
2970 /*
2971  * Interface ioctls.
2972  */
2973 int
2974 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2975 {
2976 #ifdef COMPAT_FREEBSD32
2977 	caddr_t saved_data = NULL;
2978 	struct ifmediareq ifmr;
2979 	struct ifmediareq *ifmrp = NULL;
2980 #endif
2981 	struct ifnet *ifp;
2982 	struct ifreq *ifr;
2983 	int error;
2984 	int oif_flags;
2985 
2986 	CURVNET_SET(so->so_vnet);
2987 #ifdef VIMAGE
2988 	/* Make sure the VNET is stable. */
2989 	if (so->so_vnet->vnet_shutdown) {
2990 		CURVNET_RESTORE();
2991 		return (EBUSY);
2992 	}
2993 #endif
2994 
2995 	switch (cmd) {
2996 	case SIOCGIFCONF:
2997 		error = ifconf(cmd, data);
2998 		goto out_noref;
2999 
3000 #ifdef COMPAT_FREEBSD32
3001 	case SIOCGIFCONF32:
3002 		{
3003 			struct ifconf32 *ifc32;
3004 			struct ifconf ifc;
3005 
3006 			ifc32 = (struct ifconf32 *)data;
3007 			ifc.ifc_len = ifc32->ifc_len;
3008 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3009 
3010 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
3011 			if (error == 0)
3012 				ifc32->ifc_len = ifc.ifc_len;
3013 			goto out_noref;
3014 		}
3015 #endif
3016 	}
3017 
3018 #ifdef COMPAT_FREEBSD32
3019 	switch (cmd) {
3020 	case SIOCGIFMEDIA32:
3021 	case SIOCGIFXMEDIA32:
3022 		ifmrp = &ifmr;
3023 		ifmr_init(ifmrp, data);
3024 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3025 		saved_data = data;
3026 		data = (caddr_t)ifmrp;
3027 	}
3028 #endif
3029 
3030 	ifr = (struct ifreq *)data;
3031 	switch (cmd) {
3032 #ifdef VIMAGE
3033 	case SIOCSIFRVNET:
3034 		error = priv_check(td, PRIV_NET_SETIFVNET);
3035 		if (error == 0)
3036 			error = if_vmove_reclaim(td, ifr->ifr_name,
3037 			    ifr->ifr_jid);
3038 		goto out_noref;
3039 #endif
3040 	case SIOCIFCREATE:
3041 	case SIOCIFCREATE2:
3042 		error = priv_check(td, PRIV_NET_IFCREATE);
3043 		if (error == 0)
3044 			error = if_clone_create(ifr->ifr_name,
3045 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3046 			    ifr_data_get_ptr(ifr) : NULL);
3047 		goto out_noref;
3048 	case SIOCIFDESTROY:
3049 		error = priv_check(td, PRIV_NET_IFDESTROY);
3050 		if (error == 0)
3051 			error = if_clone_destroy(ifr->ifr_name);
3052 		goto out_noref;
3053 
3054 	case SIOCIFGCLONERS:
3055 		error = if_clone_list((struct if_clonereq *)data);
3056 		goto out_noref;
3057 
3058 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3059 		error = if_getgroupmembers((struct ifgroupreq *)data);
3060 		goto out_noref;
3061 
3062 #if defined(INET) || defined(INET6)
3063 	case SIOCSVH:
3064 	case SIOCGVH:
3065 		if (carp_ioctl_p == NULL)
3066 			error = EPROTONOSUPPORT;
3067 		else
3068 			error = (*carp_ioctl_p)(ifr, cmd, td);
3069 		goto out_noref;
3070 #endif
3071 	}
3072 
3073 	ifp = ifunit_ref(ifr->ifr_name);
3074 	if (ifp == NULL) {
3075 		error = ENXIO;
3076 		goto out_noref;
3077 	}
3078 
3079 	error = ifhwioctl(cmd, ifp, data, td);
3080 	if (error != ENOIOCTL)
3081 		goto out_ref;
3082 
3083 	oif_flags = ifp->if_flags;
3084 	if (so->so_proto == NULL) {
3085 		error = EOPNOTSUPP;
3086 		goto out_ref;
3087 	}
3088 
3089 	/*
3090 	 * Pass the request on to the socket control method, and if the
3091 	 * latter returns EOPNOTSUPP, directly to the interface.
3092 	 *
3093 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3094 	 * trust SIOCSIFADDR et al to come from an already privileged
3095 	 * layer, and do not perform any credentials checks or input
3096 	 * validation.
3097 	 */
3098 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3099 	    ifp, td));
3100 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3101 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3102 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3103 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3104 
3105 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3106 #ifdef INET6
3107 		if (ifp->if_flags & IFF_UP)
3108 			in6_if_up(ifp);
3109 #endif
3110 	}
3111 
3112 out_ref:
3113 	if_rele(ifp);
3114 out_noref:
3115 #ifdef COMPAT_FREEBSD32
3116 	if (ifmrp != NULL) {
3117 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3118 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3119 		     cmd));
3120 		data = saved_data;
3121 		ifmr_update(ifmrp, data);
3122 	}
3123 #endif
3124 	CURVNET_RESTORE();
3125 	return (error);
3126 }
3127 
3128 /*
3129  * The code common to handling reference counted flags,
3130  * e.g., in ifpromisc() and if_allmulti().
3131  * The "pflag" argument can specify a permanent mode flag to check,
3132  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3133  *
3134  * Only to be used on stack-owned flags, not driver-owned flags.
3135  */
3136 static int
3137 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3138 {
3139 	struct ifreq ifr;
3140 	int error;
3141 	int oldflags, oldcount;
3142 
3143 	/* Sanity checks to catch programming errors */
3144 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3145 	    ("%s: setting driver-owned flag %d", __func__, flag));
3146 
3147 	if (onswitch)
3148 		KASSERT(*refcount >= 0,
3149 		    ("%s: increment negative refcount %d for flag %d",
3150 		    __func__, *refcount, flag));
3151 	else
3152 		KASSERT(*refcount > 0,
3153 		    ("%s: decrement non-positive refcount %d for flag %d",
3154 		    __func__, *refcount, flag));
3155 
3156 	/* In case this mode is permanent, just touch refcount */
3157 	if (ifp->if_flags & pflag) {
3158 		*refcount += onswitch ? 1 : -1;
3159 		return (0);
3160 	}
3161 
3162 	/* Save ifnet parameters for if_ioctl() may fail */
3163 	oldcount = *refcount;
3164 	oldflags = ifp->if_flags;
3165 
3166 	/*
3167 	 * See if we aren't the only and touching refcount is enough.
3168 	 * Actually toggle interface flag if we are the first or last.
3169 	 */
3170 	if (onswitch) {
3171 		if ((*refcount)++)
3172 			return (0);
3173 		ifp->if_flags |= flag;
3174 	} else {
3175 		if (--(*refcount))
3176 			return (0);
3177 		ifp->if_flags &= ~flag;
3178 	}
3179 
3180 	/* Call down the driver since we've changed interface flags */
3181 	if (ifp->if_ioctl == NULL) {
3182 		error = EOPNOTSUPP;
3183 		goto recover;
3184 	}
3185 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3186 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3187 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3188 	if (error)
3189 		goto recover;
3190 	/* Notify userland that interface flags have changed */
3191 	rt_ifmsg(ifp);
3192 	return (0);
3193 
3194 recover:
3195 	/* Recover after driver error */
3196 	*refcount = oldcount;
3197 	ifp->if_flags = oldflags;
3198 	return (error);
3199 }
3200 
3201 /*
3202  * Set/clear promiscuous mode on interface ifp based on the truth value
3203  * of pswitch.  The calls are reference counted so that only the first
3204  * "on" request actually has an effect, as does the final "off" request.
3205  * Results are undefined if the "off" and "on" requests are not matched.
3206  */
3207 int
3208 ifpromisc(struct ifnet *ifp, int pswitch)
3209 {
3210 	int error;
3211 	int oldflags = ifp->if_flags;
3212 
3213 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3214 			   &ifp->if_pcount, pswitch);
3215 	/* If promiscuous mode status has changed, log a message */
3216 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3217             log_promisc_mode_change)
3218 		if_printf(ifp, "promiscuous mode %s\n",
3219 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3220 	return (error);
3221 }
3222 
3223 /*
3224  * Return interface configuration
3225  * of system.  List may be used
3226  * in later ioctl's (above) to get
3227  * other information.
3228  */
3229 /*ARGSUSED*/
3230 static int
3231 ifconf(u_long cmd, caddr_t data)
3232 {
3233 	struct ifconf *ifc = (struct ifconf *)data;
3234 	struct ifnet *ifp;
3235 	struct ifaddr *ifa;
3236 	struct ifreq ifr;
3237 	struct sbuf *sb;
3238 	int error, full = 0, valid_len, max_len;
3239 
3240 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3241 	max_len = MAXPHYS - 1;
3242 
3243 	/* Prevent hostile input from being able to crash the system */
3244 	if (ifc->ifc_len <= 0)
3245 		return (EINVAL);
3246 
3247 again:
3248 	if (ifc->ifc_len <= max_len) {
3249 		max_len = ifc->ifc_len;
3250 		full = 1;
3251 	}
3252 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3253 	max_len = 0;
3254 	valid_len = 0;
3255 
3256 	IFNET_RLOCK();
3257 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3258 		struct epoch_tracker et;
3259 		int addrs;
3260 
3261 		/*
3262 		 * Zero the ifr to make sure we don't disclose the contents
3263 		 * of the stack.
3264 		 */
3265 		memset(&ifr, 0, sizeof(ifr));
3266 
3267 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3268 		    >= sizeof(ifr.ifr_name)) {
3269 			sbuf_delete(sb);
3270 			IFNET_RUNLOCK();
3271 			return (ENAMETOOLONG);
3272 		}
3273 
3274 		addrs = 0;
3275 		NET_EPOCH_ENTER(et);
3276 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3277 			struct sockaddr *sa = ifa->ifa_addr;
3278 
3279 			if (prison_if(curthread->td_ucred, sa) != 0)
3280 				continue;
3281 			addrs++;
3282 			if (sa->sa_len <= sizeof(*sa)) {
3283 				if (sa->sa_len < sizeof(*sa)) {
3284 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3285 					    sizeof(ifr.ifr_ifru.ifru_addr));
3286 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3287 					    sa->sa_len);
3288 				} else
3289 					ifr.ifr_ifru.ifru_addr = *sa;
3290 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3291 				max_len += sizeof(ifr);
3292 			} else {
3293 				sbuf_bcat(sb, &ifr,
3294 				    offsetof(struct ifreq, ifr_addr));
3295 				max_len += offsetof(struct ifreq, ifr_addr);
3296 				sbuf_bcat(sb, sa, sa->sa_len);
3297 				max_len += sa->sa_len;
3298 			}
3299 
3300 			if (sbuf_error(sb) == 0)
3301 				valid_len = sbuf_len(sb);
3302 		}
3303 		NET_EPOCH_EXIT(et);
3304 		if (addrs == 0) {
3305 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3306 			max_len += sizeof(ifr);
3307 
3308 			if (sbuf_error(sb) == 0)
3309 				valid_len = sbuf_len(sb);
3310 		}
3311 	}
3312 	IFNET_RUNLOCK();
3313 
3314 	/*
3315 	 * If we didn't allocate enough space (uncommon), try again.  If
3316 	 * we have already allocated as much space as we are allowed,
3317 	 * return what we've got.
3318 	 */
3319 	if (valid_len != max_len && !full) {
3320 		sbuf_delete(sb);
3321 		goto again;
3322 	}
3323 
3324 	ifc->ifc_len = valid_len;
3325 	sbuf_finish(sb);
3326 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3327 	sbuf_delete(sb);
3328 	return (error);
3329 }
3330 
3331 /*
3332  * Just like ifpromisc(), but for all-multicast-reception mode.
3333  */
3334 int
3335 if_allmulti(struct ifnet *ifp, int onswitch)
3336 {
3337 
3338 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3339 }
3340 
3341 struct ifmultiaddr *
3342 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3343 {
3344 	struct ifmultiaddr *ifma;
3345 
3346 	IF_ADDR_LOCK_ASSERT(ifp);
3347 
3348 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3349 		if (sa->sa_family == AF_LINK) {
3350 			if (sa_dl_equal(ifma->ifma_addr, sa))
3351 				break;
3352 		} else {
3353 			if (sa_equal(ifma->ifma_addr, sa))
3354 				break;
3355 		}
3356 	}
3357 
3358 	return ifma;
3359 }
3360 
3361 /*
3362  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3363  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3364  * the ifnet multicast address list here, so the caller must do that and
3365  * other setup work (such as notifying the device driver).  The reference
3366  * count is initialized to 1.
3367  */
3368 static struct ifmultiaddr *
3369 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3370     int mflags)
3371 {
3372 	struct ifmultiaddr *ifma;
3373 	struct sockaddr *dupsa;
3374 
3375 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3376 	    M_ZERO);
3377 	if (ifma == NULL)
3378 		return (NULL);
3379 
3380 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3381 	if (dupsa == NULL) {
3382 		free(ifma, M_IFMADDR);
3383 		return (NULL);
3384 	}
3385 	bcopy(sa, dupsa, sa->sa_len);
3386 	ifma->ifma_addr = dupsa;
3387 
3388 	ifma->ifma_ifp = ifp;
3389 	ifma->ifma_refcount = 1;
3390 	ifma->ifma_protospec = NULL;
3391 
3392 	if (llsa == NULL) {
3393 		ifma->ifma_lladdr = NULL;
3394 		return (ifma);
3395 	}
3396 
3397 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3398 	if (dupsa == NULL) {
3399 		free(ifma->ifma_addr, M_IFMADDR);
3400 		free(ifma, M_IFMADDR);
3401 		return (NULL);
3402 	}
3403 	bcopy(llsa, dupsa, llsa->sa_len);
3404 	ifma->ifma_lladdr = dupsa;
3405 
3406 	return (ifma);
3407 }
3408 
3409 /*
3410  * if_freemulti: free ifmultiaddr structure and possibly attached related
3411  * addresses.  The caller is responsible for implementing reference
3412  * counting, notifying the driver, handling routing messages, and releasing
3413  * any dependent link layer state.
3414  */
3415 #ifdef MCAST_VERBOSE
3416 extern void kdb_backtrace(void);
3417 #endif
3418 static void
3419 if_freemulti_internal(struct ifmultiaddr *ifma)
3420 {
3421 
3422 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3423 	    ifma->ifma_refcount));
3424 
3425 	if (ifma->ifma_lladdr != NULL)
3426 		free(ifma->ifma_lladdr, M_IFMADDR);
3427 #ifdef MCAST_VERBOSE
3428 	kdb_backtrace();
3429 	printf("%s freeing ifma: %p\n", __func__, ifma);
3430 #endif
3431 	free(ifma->ifma_addr, M_IFMADDR);
3432 	free(ifma, M_IFMADDR);
3433 }
3434 
3435 static void
3436 if_destroymulti(epoch_context_t ctx)
3437 {
3438 	struct ifmultiaddr *ifma;
3439 
3440 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3441 	if_freemulti_internal(ifma);
3442 }
3443 
3444 void
3445 if_freemulti(struct ifmultiaddr *ifma)
3446 {
3447 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3448 	    ifma->ifma_refcount));
3449 
3450 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3451 }
3452 
3453 
3454 /*
3455  * Register an additional multicast address with a network interface.
3456  *
3457  * - If the address is already present, bump the reference count on the
3458  *   address and return.
3459  * - If the address is not link-layer, look up a link layer address.
3460  * - Allocate address structures for one or both addresses, and attach to the
3461  *   multicast address list on the interface.  If automatically adding a link
3462  *   layer address, the protocol address will own a reference to the link
3463  *   layer address, to be freed when it is freed.
3464  * - Notify the network device driver of an addition to the multicast address
3465  *   list.
3466  *
3467  * 'sa' points to caller-owned memory with the desired multicast address.
3468  *
3469  * 'retifma' will be used to return a pointer to the resulting multicast
3470  * address reference, if desired.
3471  */
3472 int
3473 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3474     struct ifmultiaddr **retifma)
3475 {
3476 	struct ifmultiaddr *ifma, *ll_ifma;
3477 	struct sockaddr *llsa;
3478 	struct sockaddr_dl sdl;
3479 	int error;
3480 
3481 #ifdef INET
3482 	IN_MULTI_LIST_UNLOCK_ASSERT();
3483 #endif
3484 #ifdef INET6
3485 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3486 #endif
3487 	/*
3488 	 * If the address is already present, return a new reference to it;
3489 	 * otherwise, allocate storage and set up a new address.
3490 	 */
3491 	IF_ADDR_WLOCK(ifp);
3492 	ifma = if_findmulti(ifp, sa);
3493 	if (ifma != NULL) {
3494 		ifma->ifma_refcount++;
3495 		if (retifma != NULL)
3496 			*retifma = ifma;
3497 		IF_ADDR_WUNLOCK(ifp);
3498 		return (0);
3499 	}
3500 
3501 	/*
3502 	 * The address isn't already present; resolve the protocol address
3503 	 * into a link layer address, and then look that up, bump its
3504 	 * refcount or allocate an ifma for that also.
3505 	 * Most link layer resolving functions returns address data which
3506 	 * fits inside default sockaddr_dl structure. However callback
3507 	 * can allocate another sockaddr structure, in that case we need to
3508 	 * free it later.
3509 	 */
3510 	llsa = NULL;
3511 	ll_ifma = NULL;
3512 	if (ifp->if_resolvemulti != NULL) {
3513 		/* Provide called function with buffer size information */
3514 		sdl.sdl_len = sizeof(sdl);
3515 		llsa = (struct sockaddr *)&sdl;
3516 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3517 		if (error)
3518 			goto unlock_out;
3519 	}
3520 
3521 	/*
3522 	 * Allocate the new address.  Don't hook it up yet, as we may also
3523 	 * need to allocate a link layer multicast address.
3524 	 */
3525 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3526 	if (ifma == NULL) {
3527 		error = ENOMEM;
3528 		goto free_llsa_out;
3529 	}
3530 
3531 	/*
3532 	 * If a link layer address is found, we'll need to see if it's
3533 	 * already present in the address list, or allocate is as well.
3534 	 * When this block finishes, the link layer address will be on the
3535 	 * list.
3536 	 */
3537 	if (llsa != NULL) {
3538 		ll_ifma = if_findmulti(ifp, llsa);
3539 		if (ll_ifma == NULL) {
3540 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3541 			if (ll_ifma == NULL) {
3542 				--ifma->ifma_refcount;
3543 				if_freemulti(ifma);
3544 				error = ENOMEM;
3545 				goto free_llsa_out;
3546 			}
3547 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3548 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3549 			    ifma_link);
3550 		} else
3551 			ll_ifma->ifma_refcount++;
3552 		ifma->ifma_llifma = ll_ifma;
3553 	}
3554 
3555 	/*
3556 	 * We now have a new multicast address, ifma, and possibly a new or
3557 	 * referenced link layer address.  Add the primary address to the
3558 	 * ifnet address list.
3559 	 */
3560 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3561 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3562 
3563 	if (retifma != NULL)
3564 		*retifma = ifma;
3565 
3566 	/*
3567 	 * Must generate the message while holding the lock so that 'ifma'
3568 	 * pointer is still valid.
3569 	 */
3570 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3571 	IF_ADDR_WUNLOCK(ifp);
3572 
3573 	/*
3574 	 * We are certain we have added something, so call down to the
3575 	 * interface to let them know about it.
3576 	 */
3577 	if (ifp->if_ioctl != NULL) {
3578 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3579 	}
3580 
3581 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3582 		link_free_sdl(llsa);
3583 
3584 	return (0);
3585 
3586 free_llsa_out:
3587 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3588 		link_free_sdl(llsa);
3589 
3590 unlock_out:
3591 	IF_ADDR_WUNLOCK(ifp);
3592 	return (error);
3593 }
3594 
3595 /*
3596  * Delete a multicast group membership by network-layer group address.
3597  *
3598  * Returns ENOENT if the entry could not be found. If ifp no longer
3599  * exists, results are undefined. This entry point should only be used
3600  * from subsystems which do appropriate locking to hold ifp for the
3601  * duration of the call.
3602  * Network-layer protocol domains must use if_delmulti_ifma().
3603  */
3604 int
3605 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3606 {
3607 	struct ifmultiaddr *ifma;
3608 	int lastref;
3609 
3610 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3611 
3612 	IF_ADDR_WLOCK(ifp);
3613 	lastref = 0;
3614 	ifma = if_findmulti(ifp, sa);
3615 	if (ifma != NULL)
3616 		lastref = if_delmulti_locked(ifp, ifma, 0);
3617 	IF_ADDR_WUNLOCK(ifp);
3618 
3619 	if (ifma == NULL)
3620 		return (ENOENT);
3621 
3622 	if (lastref && ifp->if_ioctl != NULL) {
3623 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3624 	}
3625 
3626 	return (0);
3627 }
3628 
3629 /*
3630  * Delete all multicast group membership for an interface.
3631  * Should be used to quickly flush all multicast filters.
3632  */
3633 void
3634 if_delallmulti(struct ifnet *ifp)
3635 {
3636 	struct ifmultiaddr *ifma;
3637 	struct ifmultiaddr *next;
3638 
3639 	IF_ADDR_WLOCK(ifp);
3640 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3641 		if_delmulti_locked(ifp, ifma, 0);
3642 	IF_ADDR_WUNLOCK(ifp);
3643 }
3644 
3645 void
3646 if_delmulti_ifma(struct ifmultiaddr *ifma)
3647 {
3648 	if_delmulti_ifma_flags(ifma, 0);
3649 }
3650 
3651 /*
3652  * Delete a multicast group membership by group membership pointer.
3653  * Network-layer protocol domains must use this routine.
3654  *
3655  * It is safe to call this routine if the ifp disappeared.
3656  */
3657 void
3658 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3659 {
3660 	struct ifnet *ifp;
3661 	int lastref;
3662 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3663 #ifdef INET
3664 	IN_MULTI_LIST_UNLOCK_ASSERT();
3665 #endif
3666 	ifp = ifma->ifma_ifp;
3667 #ifdef DIAGNOSTIC
3668 	if (ifp == NULL) {
3669 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3670 	} else {
3671 		struct epoch_tracker et;
3672 		struct ifnet *oifp;
3673 
3674 		NET_EPOCH_ENTER(et);
3675 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3676 			if (ifp == oifp)
3677 				break;
3678 		NET_EPOCH_EXIT(et);
3679 		if (ifp != oifp)
3680 			ifp = NULL;
3681 	}
3682 #endif
3683 	/*
3684 	 * If and only if the ifnet instance exists: Acquire the address lock.
3685 	 */
3686 	if (ifp != NULL)
3687 		IF_ADDR_WLOCK(ifp);
3688 
3689 	lastref = if_delmulti_locked(ifp, ifma, flags);
3690 
3691 	if (ifp != NULL) {
3692 		/*
3693 		 * If and only if the ifnet instance exists:
3694 		 *  Release the address lock.
3695 		 *  If the group was left: update the hardware hash filter.
3696 		 */
3697 		IF_ADDR_WUNLOCK(ifp);
3698 		if (lastref && ifp->if_ioctl != NULL) {
3699 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3700 		}
3701 	}
3702 }
3703 
3704 /*
3705  * Perform deletion of network-layer and/or link-layer multicast address.
3706  *
3707  * Return 0 if the reference count was decremented.
3708  * Return 1 if the final reference was released, indicating that the
3709  * hardware hash filter should be reprogrammed.
3710  */
3711 static int
3712 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3713 {
3714 	struct ifmultiaddr *ll_ifma;
3715 
3716 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3717 		KASSERT(ifma->ifma_ifp == ifp,
3718 		    ("%s: inconsistent ifp %p", __func__, ifp));
3719 		IF_ADDR_WLOCK_ASSERT(ifp);
3720 	}
3721 
3722 	ifp = ifma->ifma_ifp;
3723 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3724 
3725 	/*
3726 	 * If the ifnet is detaching, null out references to ifnet,
3727 	 * so that upper protocol layers will notice, and not attempt
3728 	 * to obtain locks for an ifnet which no longer exists. The
3729 	 * routing socket announcement must happen before the ifnet
3730 	 * instance is detached from the system.
3731 	 */
3732 	if (detaching) {
3733 #ifdef DIAGNOSTIC
3734 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3735 #endif
3736 		/*
3737 		 * ifp may already be nulled out if we are being reentered
3738 		 * to delete the ll_ifma.
3739 		 */
3740 		if (ifp != NULL) {
3741 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3742 			ifma->ifma_ifp = NULL;
3743 		}
3744 	}
3745 
3746 	if (--ifma->ifma_refcount > 0)
3747 		return 0;
3748 
3749 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3750 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3751 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3752 	}
3753 	/*
3754 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3755 	 * have been associated with it. Release it first if so.
3756 	 */
3757 	ll_ifma = ifma->ifma_llifma;
3758 	if (ll_ifma != NULL) {
3759 		KASSERT(ifma->ifma_lladdr != NULL,
3760 		    ("%s: llifma w/o lladdr", __func__));
3761 		if (detaching)
3762 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3763 		if (--ll_ifma->ifma_refcount == 0) {
3764 			if (ifp != NULL) {
3765 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3766 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3767 						ifma_link);
3768 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3769 				}
3770 			}
3771 			if_freemulti(ll_ifma);
3772 		}
3773 	}
3774 #ifdef INVARIANTS
3775 	if (ifp) {
3776 		struct ifmultiaddr *ifmatmp;
3777 
3778 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3779 			MPASS(ifma != ifmatmp);
3780 	}
3781 #endif
3782 	if_freemulti(ifma);
3783 	/*
3784 	 * The last reference to this instance of struct ifmultiaddr
3785 	 * was released; the hardware should be notified of this change.
3786 	 */
3787 	return 1;
3788 }
3789 
3790 /*
3791  * Set the link layer address on an interface.
3792  *
3793  * At this time we only support certain types of interfaces,
3794  * and we don't allow the length of the address to change.
3795  *
3796  * Set noinline to be dtrace-friendly
3797  */
3798 __noinline int
3799 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3800 {
3801 	struct sockaddr_dl *sdl;
3802 	struct ifaddr *ifa;
3803 	struct ifreq ifr;
3804 
3805 	ifa = ifp->if_addr;
3806 	if (ifa == NULL)
3807 		return (EINVAL);
3808 
3809 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3810 	if (sdl == NULL)
3811 		return (EINVAL);
3812 
3813 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3814 		return (EINVAL);
3815 
3816 	switch (ifp->if_type) {
3817 	case IFT_ETHER:
3818 	case IFT_XETHER:
3819 	case IFT_L2VLAN:
3820 	case IFT_BRIDGE:
3821 	case IFT_IEEE8023ADLAG:
3822 		bcopy(lladdr, LLADDR(sdl), len);
3823 		break;
3824 	default:
3825 		return (ENODEV);
3826 	}
3827 
3828 	/*
3829 	 * If the interface is already up, we need
3830 	 * to re-init it in order to reprogram its
3831 	 * address filter.
3832 	 */
3833 	if ((ifp->if_flags & IFF_UP) != 0) {
3834 		if (ifp->if_ioctl) {
3835 			ifp->if_flags &= ~IFF_UP;
3836 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3837 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3838 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3839 			ifp->if_flags |= IFF_UP;
3840 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3841 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3842 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3843 		}
3844 	}
3845 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3846 
3847 	return (0);
3848 }
3849 
3850 /*
3851  * Compat function for handling basic encapsulation requests.
3852  * Not converted stacks (FDDI, IB, ..) supports traditional
3853  * output model: ARP (and other similar L2 protocols) are handled
3854  * inside output routine, arpresolve/nd6_resolve() returns MAC
3855  * address instead of full prepend.
3856  *
3857  * This function creates calculated header==MAC for IPv4/IPv6 and
3858  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3859  * address families.
3860  */
3861 static int
3862 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3863 {
3864 
3865 	if (req->rtype != IFENCAP_LL)
3866 		return (EOPNOTSUPP);
3867 
3868 	if (req->bufsize < req->lladdr_len)
3869 		return (ENOMEM);
3870 
3871 	switch (req->family) {
3872 	case AF_INET:
3873 	case AF_INET6:
3874 		break;
3875 	default:
3876 		return (EAFNOSUPPORT);
3877 	}
3878 
3879 	/* Copy lladdr to storage as is */
3880 	memmove(req->buf, req->lladdr, req->lladdr_len);
3881 	req->bufsize = req->lladdr_len;
3882 	req->lladdr_off = 0;
3883 
3884 	return (0);
3885 }
3886 
3887 /*
3888  * Tunnel interfaces can nest, also they may cause infinite recursion
3889  * calls when misconfigured. We'll prevent this by detecting loops.
3890  * High nesting level may cause stack exhaustion. We'll prevent this
3891  * by introducing upper limit.
3892  *
3893  * Return 0, if tunnel nesting count is equal or less than limit.
3894  */
3895 int
3896 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3897     int limit)
3898 {
3899 	struct m_tag *mtag;
3900 	int count;
3901 
3902 	count = 1;
3903 	mtag = NULL;
3904 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3905 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3906 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3907 			return (EIO);
3908 		}
3909 		count++;
3910 	}
3911 	if (count > limit) {
3912 		log(LOG_NOTICE,
3913 		    "%s: if_output recursively called too many times(%d)\n",
3914 		    if_name(ifp), count);
3915 		return (EIO);
3916 	}
3917 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3918 	if (mtag == NULL)
3919 		return (ENOMEM);
3920 	*(struct ifnet **)(mtag + 1) = ifp;
3921 	m_tag_prepend(m, mtag);
3922 	return (0);
3923 }
3924 
3925 /*
3926  * Get the link layer address that was read from the hardware at attach.
3927  *
3928  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3929  * their component interfaces as IFT_IEEE8023ADLAG.
3930  */
3931 int
3932 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3933 {
3934 
3935 	if (ifp->if_hw_addr == NULL)
3936 		return (ENODEV);
3937 
3938 	switch (ifp->if_type) {
3939 	case IFT_ETHER:
3940 	case IFT_IEEE8023ADLAG:
3941 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3942 		return (0);
3943 	default:
3944 		return (ENODEV);
3945 	}
3946 }
3947 
3948 /*
3949  * The name argument must be a pointer to storage which will last as
3950  * long as the interface does.  For physical devices, the result of
3951  * device_get_name(dev) is a good choice and for pseudo-devices a
3952  * static string works well.
3953  */
3954 void
3955 if_initname(struct ifnet *ifp, const char *name, int unit)
3956 {
3957 	ifp->if_dname = name;
3958 	ifp->if_dunit = unit;
3959 	if (unit != IF_DUNIT_NONE)
3960 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3961 	else
3962 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3963 }
3964 
3965 int
3966 if_printf(struct ifnet *ifp, const char *fmt, ...)
3967 {
3968 	char if_fmt[256];
3969 	va_list ap;
3970 
3971 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3972 	va_start(ap, fmt);
3973 	vlog(LOG_INFO, if_fmt, ap);
3974 	va_end(ap);
3975 	return (0);
3976 }
3977 
3978 void
3979 if_start(struct ifnet *ifp)
3980 {
3981 
3982 	(*(ifp)->if_start)(ifp);
3983 }
3984 
3985 /*
3986  * Backwards compatibility interface for drivers
3987  * that have not implemented it
3988  */
3989 static int
3990 if_transmit(struct ifnet *ifp, struct mbuf *m)
3991 {
3992 	int error;
3993 
3994 	IFQ_HANDOFF(ifp, m, error);
3995 	return (error);
3996 }
3997 
3998 static void
3999 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4000 {
4001 
4002 	m_freem(m);
4003 }
4004 
4005 int
4006 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4007 {
4008 	int active = 0;
4009 
4010 	IF_LOCK(ifq);
4011 	if (_IF_QFULL(ifq)) {
4012 		IF_UNLOCK(ifq);
4013 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4014 		m_freem(m);
4015 		return (0);
4016 	}
4017 	if (ifp != NULL) {
4018 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4019 		if (m->m_flags & (M_BCAST|M_MCAST))
4020 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4021 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4022 	}
4023 	_IF_ENQUEUE(ifq, m);
4024 	IF_UNLOCK(ifq);
4025 	if (ifp != NULL && !active)
4026 		(*(ifp)->if_start)(ifp);
4027 	return (1);
4028 }
4029 
4030 void
4031 if_register_com_alloc(u_char type,
4032     if_com_alloc_t *a, if_com_free_t *f)
4033 {
4034 
4035 	KASSERT(if_com_alloc[type] == NULL,
4036 	    ("if_register_com_alloc: %d already registered", type));
4037 	KASSERT(if_com_free[type] == NULL,
4038 	    ("if_register_com_alloc: %d free already registered", type));
4039 
4040 	if_com_alloc[type] = a;
4041 	if_com_free[type] = f;
4042 }
4043 
4044 void
4045 if_deregister_com_alloc(u_char type)
4046 {
4047 
4048 	KASSERT(if_com_alloc[type] != NULL,
4049 	    ("if_deregister_com_alloc: %d not registered", type));
4050 	KASSERT(if_com_free[type] != NULL,
4051 	    ("if_deregister_com_alloc: %d free not registered", type));
4052 	if_com_alloc[type] = NULL;
4053 	if_com_free[type] = NULL;
4054 }
4055 
4056 /* API for driver access to network stack owned ifnet.*/
4057 uint64_t
4058 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4059 {
4060 	uint64_t oldbrate;
4061 
4062 	oldbrate = ifp->if_baudrate;
4063 	ifp->if_baudrate = baudrate;
4064 	return (oldbrate);
4065 }
4066 
4067 uint64_t
4068 if_getbaudrate(if_t ifp)
4069 {
4070 
4071 	return (((struct ifnet *)ifp)->if_baudrate);
4072 }
4073 
4074 int
4075 if_setcapabilities(if_t ifp, int capabilities)
4076 {
4077 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4078 	return (0);
4079 }
4080 
4081 int
4082 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4083 {
4084 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4085 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4086 
4087 	return (0);
4088 }
4089 
4090 int
4091 if_getcapabilities(if_t ifp)
4092 {
4093 	return ((struct ifnet *)ifp)->if_capabilities;
4094 }
4095 
4096 int
4097 if_setcapenable(if_t ifp, int capabilities)
4098 {
4099 	((struct ifnet *)ifp)->if_capenable = capabilities;
4100 	return (0);
4101 }
4102 
4103 int
4104 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4105 {
4106 	if(setcap)
4107 		((struct ifnet *)ifp)->if_capenable |= setcap;
4108 	if(clearcap)
4109 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4110 
4111 	return (0);
4112 }
4113 
4114 const char *
4115 if_getdname(if_t ifp)
4116 {
4117 	return ((struct ifnet *)ifp)->if_dname;
4118 }
4119 
4120 int
4121 if_togglecapenable(if_t ifp, int togglecap)
4122 {
4123 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4124 	return (0);
4125 }
4126 
4127 int
4128 if_getcapenable(if_t ifp)
4129 {
4130 	return ((struct ifnet *)ifp)->if_capenable;
4131 }
4132 
4133 /*
4134  * This is largely undesirable because it ties ifnet to a device, but does
4135  * provide flexiblity for an embedded product vendor. Should be used with
4136  * the understanding that it violates the interface boundaries, and should be
4137  * a last resort only.
4138  */
4139 int
4140 if_setdev(if_t ifp, void *dev)
4141 {
4142 	return (0);
4143 }
4144 
4145 int
4146 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4147 {
4148 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4149 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4150 
4151 	return (0);
4152 }
4153 
4154 int
4155 if_getdrvflags(if_t ifp)
4156 {
4157 	return ((struct ifnet *)ifp)->if_drv_flags;
4158 }
4159 
4160 int
4161 if_setdrvflags(if_t ifp, int flags)
4162 {
4163 	((struct ifnet *)ifp)->if_drv_flags = flags;
4164 	return (0);
4165 }
4166 
4167 
4168 int
4169 if_setflags(if_t ifp, int flags)
4170 {
4171 	((struct ifnet *)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 /* XXX */
4412 #ifndef ETH_ADDR_LEN
4413 #define ETH_ADDR_LEN 6
4414 #endif
4415 
4416 int
4417 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4418 {
4419 	struct ifmultiaddr *ifma;
4420 	uint8_t *lmta = (uint8_t *)mta;
4421 	int mcnt = 0;
4422 
4423 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4424 		if (ifma->ifma_addr->sa_family != AF_LINK)
4425 			continue;
4426 
4427 		if (mcnt == max)
4428 			break;
4429 
4430 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4431 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4432 		mcnt++;
4433 	}
4434 	*cnt = mcnt;
4435 
4436 	return (0);
4437 }
4438 
4439 int
4440 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4441 {
4442 	int error;
4443 
4444 	if_maddr_rlock(ifp);
4445 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4446 	if_maddr_runlock(ifp);
4447 	return (error);
4448 }
4449 
4450 int
4451 if_multiaddr_count(if_t ifp, int max)
4452 {
4453 	struct ifmultiaddr *ifma;
4454 	int count;
4455 
4456 	count = 0;
4457 	if_maddr_rlock(ifp);
4458 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4459 		if (ifma->ifma_addr->sa_family != AF_LINK)
4460 			continue;
4461 		count++;
4462 		if (count == max)
4463 			break;
4464 	}
4465 	if_maddr_runlock(ifp);
4466 	return (count);
4467 }
4468 
4469 int
4470 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4471 {
4472 	struct ifmultiaddr *ifma;
4473 	int cnt = 0;
4474 
4475 	if_maddr_rlock(ifp);
4476 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4477 		cnt += filter(arg, ifma, cnt);
4478 	if_maddr_runlock(ifp);
4479 	return (cnt);
4480 }
4481 
4482 struct mbuf *
4483 if_dequeue(if_t ifp)
4484 {
4485 	struct mbuf *m;
4486 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4487 
4488 	return (m);
4489 }
4490 
4491 int
4492 if_sendq_prepend(if_t ifp, struct mbuf *m)
4493 {
4494 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4495 	return (0);
4496 }
4497 
4498 int
4499 if_setifheaderlen(if_t ifp, int len)
4500 {
4501 	((struct ifnet *)ifp)->if_hdrlen = len;
4502 	return (0);
4503 }
4504 
4505 caddr_t
4506 if_getlladdr(if_t ifp)
4507 {
4508 	return (IF_LLADDR((struct ifnet *)ifp));
4509 }
4510 
4511 void *
4512 if_gethandle(u_char type)
4513 {
4514 	return (if_alloc(type));
4515 }
4516 
4517 void
4518 if_bpfmtap(if_t ifh, struct mbuf *m)
4519 {
4520 	struct ifnet *ifp = (struct ifnet *)ifh;
4521 
4522 	BPF_MTAP(ifp, m);
4523 }
4524 
4525 void
4526 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4527 {
4528 	struct ifnet *ifp = (struct ifnet *)ifh;
4529 
4530 	ETHER_BPF_MTAP(ifp, m);
4531 }
4532 
4533 void
4534 if_vlancap(if_t ifh)
4535 {
4536 	struct ifnet *ifp = (struct ifnet *)ifh;
4537 	VLAN_CAPABILITIES(ifp);
4538 }
4539 
4540 int
4541 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4542 {
4543 
4544 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4545         return (0);
4546 }
4547 
4548 int
4549 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4550 {
4551 
4552 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4553         return (0);
4554 }
4555 
4556 int
4557 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4558 {
4559 
4560 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4561         return (0);
4562 }
4563 
4564 u_int
4565 if_gethwtsomax(if_t ifp)
4566 {
4567 
4568 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4569 }
4570 
4571 u_int
4572 if_gethwtsomaxsegcount(if_t ifp)
4573 {
4574 
4575 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4576 }
4577 
4578 u_int
4579 if_gethwtsomaxsegsize(if_t ifp)
4580 {
4581 
4582 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4583 }
4584 
4585 void
4586 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4587 {
4588 	((struct ifnet *)ifp)->if_init = init_fn;
4589 }
4590 
4591 void
4592 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4593 {
4594 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4595 }
4596 
4597 void
4598 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4599 {
4600 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4601 }
4602 
4603 void
4604 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4605 {
4606 	((struct ifnet *)ifp)->if_transmit = start_fn;
4607 }
4608 
4609 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4610 {
4611 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4612 
4613 }
4614 
4615 void
4616 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4617 {
4618 
4619 	ifp->if_get_counter = fn;
4620 }
4621 
4622 /* Revisit these - These are inline functions originally. */
4623 int
4624 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4625 {
4626 	return drbr_inuse(ifh, br);
4627 }
4628 
4629 struct mbuf*
4630 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4631 {
4632 	return drbr_dequeue(ifh, br);
4633 }
4634 
4635 int
4636 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4637 {
4638 	return drbr_needs_enqueue(ifh, br);
4639 }
4640 
4641 int
4642 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4643 {
4644 	return drbr_enqueue(ifh, br, m);
4645 
4646 }
4647