xref: /freebsd/sys/net/if.c (revision 715d1396d6233158b94e85414b57520ee3e536cc)
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/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/refcount.h>
54 #include <sys/module.h>
55 #include <sys/rwlock.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <sys/priv.h>
64 
65 #include <machine/stdarg.h>
66 #include <vm/uma.h>
67 
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/if_media.h>
77 #include <net/if_vlan_var.h>
78 #include <net/radix.h>
79 #include <net/route.h>
80 #include <net/vnet.h>
81 
82 #if defined(INET) || defined(INET6)
83 #include <net/ethernet.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip.h>
87 #include <netinet/ip_carp.h>
88 #ifdef INET
89 #include <netinet/if_ether.h>
90 #endif /* INET */
91 #ifdef INET6
92 #include <netinet6/in6_var.h>
93 #include <netinet6/in6_ifattach.h>
94 #endif /* INET6 */
95 #endif /* INET || INET6 */
96 
97 #include <security/mac/mac_framework.h>
98 
99 /*
100  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
101  * and ifr_ifru when it is used in SIOCGIFCONF.
102  */
103 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
104     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
105 
106 #ifdef COMPAT_FREEBSD32
107 #include <sys/mount.h>
108 #include <compat/freebsd32/freebsd32.h>
109 
110 struct ifreq_buffer32 {
111 	uint32_t	length;		/* (size_t) */
112 	uint32_t	buffer;		/* (void *) */
113 };
114 
115 /*
116  * Interface request structure used for socket
117  * ioctl's.  All interface ioctl's must have parameter
118  * definitions which begin with ifr_name.  The
119  * remainder may be interface specific.
120  */
121 struct ifreq32 {
122 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
123 	union {
124 		struct sockaddr	ifru_addr;
125 		struct sockaddr	ifru_dstaddr;
126 		struct sockaddr	ifru_broadaddr;
127 		struct ifreq_buffer32 ifru_buffer;
128 		short		ifru_flags[2];
129 		short		ifru_index;
130 		int		ifru_jid;
131 		int		ifru_metric;
132 		int		ifru_mtu;
133 		int		ifru_phys;
134 		int		ifru_media;
135 		uint32_t	ifru_data;
136 		int		ifru_cap[2];
137 		u_int		ifru_fib;
138 		u_char		ifru_vlan_pcp;
139 	} ifr_ifru;
140 };
141 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
142 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
143     __offsetof(struct ifreq32, ifr_ifru));
144 
145 struct ifgroupreq32 {
146 	char	ifgr_name[IFNAMSIZ];
147 	u_int	ifgr_len;
148 	union {
149 		char		ifgru_group[IFNAMSIZ];
150 		uint32_t	ifgru_groups;
151 	} ifgr_ifgru;
152 };
153 
154 struct ifmediareq32 {
155 	char		ifm_name[IFNAMSIZ];
156 	int		ifm_current;
157 	int		ifm_mask;
158 	int		ifm_status;
159 	int		ifm_active;
160 	int		ifm_count;
161 	uint32_t	ifm_ulist;	/* (int *) */
162 };
163 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
164 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
165 
166 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
167     case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
168 #else /* !COMPAT_FREEBSD32 */
169 #define _CASE_IOC_IFGROUPREQ_32(cmd)
170 #endif /* !COMPAT_FREEBSD32 */
171 
172 #define CASE_IOC_IFGROUPREQ(cmd)	\
173     _CASE_IOC_IFGROUPREQ_32(cmd)	\
174     case (cmd)
175 
176 union ifreq_union {
177 	struct ifreq	ifr;
178 #ifdef COMPAT_FREEBSD32
179 	struct ifreq32	ifr32;
180 #endif
181 };
182 
183 union ifgroupreq_union {
184 	struct ifgroupreq ifgr;
185 #ifdef COMPAT_FREEBSD32
186 	struct ifgroupreq32 ifgr32;
187 #endif
188 };
189 
190 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
191 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
192 
193 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
194     &ifqmaxlen, 0, "max send queue size");
195 
196 /* Log link state change events */
197 static int log_link_state_change = 1;
198 
199 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
200 	&log_link_state_change, 0,
201 	"log interface link state change events");
202 
203 /* Log promiscuous mode change events */
204 static int log_promisc_mode_change = 1;
205 
206 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
207 	&log_promisc_mode_change, 1,
208 	"log promiscuous mode change events");
209 
210 /* Interface description */
211 static unsigned int ifdescr_maxlen = 1024;
212 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
213 	&ifdescr_maxlen, 0,
214 	"administrative maximum length for interface description");
215 
216 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
217 
218 /* global sx for non-critical path ifdescr */
219 static struct sx ifdescr_sx;
220 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
221 
222 void	(*bridge_linkstate_p)(struct ifnet *ifp);
223 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
224 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
225 /* These are external hooks for CARP. */
226 void	(*carp_linkstate_p)(struct ifnet *ifp);
227 void	(*carp_demote_adj_p)(int, char *);
228 int	(*carp_master_p)(struct ifaddr *);
229 #if defined(INET) || defined(INET6)
230 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
231 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
232     const struct sockaddr *sa);
233 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
234 int	(*carp_attach_p)(struct ifaddr *, int);
235 void	(*carp_detach_p)(struct ifaddr *, bool);
236 #endif
237 #ifdef INET
238 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
239 #endif
240 #ifdef INET6
241 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
242 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
243     const struct in6_addr *taddr);
244 #endif
245 
246 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
247 
248 /*
249  * XXX: Style; these should be sorted alphabetically, and unprototyped
250  * static functions should be prototyped. Currently they are sorted by
251  * declaration order.
252  */
253 static void	if_attachdomain(void *);
254 static void	if_attachdomain1(struct ifnet *);
255 static int	ifconf(u_long, caddr_t);
256 static void	if_freemulti(struct ifmultiaddr *);
257 static void	if_grow(void);
258 static void	if_input_default(struct ifnet *, struct mbuf *);
259 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
260 static void	if_route(struct ifnet *, int flag, int fam);
261 static int	if_setflag(struct ifnet *, int, int, int *, int);
262 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
263 static void	if_unroute(struct ifnet *, int flag, int fam);
264 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
265 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
266 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
267 static void	do_link_state_change(void *, int);
268 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
269 static int	if_getgroupmembers(struct ifgroupreq *);
270 static void	if_delgroups(struct ifnet *);
271 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
272 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
273 #ifdef VIMAGE
274 static void	if_vmove(struct ifnet *, struct vnet *);
275 #endif
276 
277 #ifdef INET6
278 /*
279  * XXX: declare here to avoid to include many inet6 related files..
280  * should be more generalized?
281  */
282 extern void	nd6_setmtu(struct ifnet *);
283 #endif
284 
285 /* ipsec helper hooks */
286 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
288 
289 VNET_DEFINE(int, if_index);
290 int	ifqmaxlen = IFQ_MAXLEN;
291 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
292 VNET_DEFINE(struct ifgrouphead, ifg_head);
293 
294 static VNET_DEFINE(int, if_indexlim) = 8;
295 
296 /* Table of ifnet by index. */
297 VNET_DEFINE(struct ifnet **, ifindex_table);
298 
299 #define	V_if_indexlim		VNET(if_indexlim)
300 #define	V_ifindex_table		VNET(ifindex_table)
301 
302 /*
303  * The global network interface list (V_ifnet) and related state (such as
304  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
305  * an rwlock.  Either may be acquired shared to stablize the list, but both
306  * must be acquired writable to modify the list.  This model allows us to
307  * both stablize the interface list during interrupt thread processing, but
308  * also to stablize it over long-running ioctls, without introducing priority
309  * inversions and deadlocks.
310  */
311 struct rwlock ifnet_rwlock;
312 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
313 struct sx ifnet_sxlock;
314 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
315 
316 /*
317  * The allocation of network interfaces is a rather non-atomic affair; we
318  * need to select an index before we are ready to expose the interface for
319  * use, so will use this pointer value to indicate reservation.
320  */
321 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
322 
323 static	if_com_alloc_t *if_com_alloc[256];
324 static	if_com_free_t *if_com_free[256];
325 
326 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
327 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
328 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
329 
330 struct ifnet *
331 ifnet_byindex_locked(u_short idx)
332 {
333 
334 	if (idx > V_if_index)
335 		return (NULL);
336 	if (V_ifindex_table[idx] == IFNET_HOLD)
337 		return (NULL);
338 	return (V_ifindex_table[idx]);
339 }
340 
341 struct ifnet *
342 ifnet_byindex(u_short idx)
343 {
344 	struct ifnet *ifp;
345 
346 	IFNET_RLOCK_NOSLEEP();
347 	ifp = ifnet_byindex_locked(idx);
348 	IFNET_RUNLOCK_NOSLEEP();
349 	return (ifp);
350 }
351 
352 struct ifnet *
353 ifnet_byindex_ref(u_short idx)
354 {
355 	struct ifnet *ifp;
356 
357 	IFNET_RLOCK_NOSLEEP();
358 	ifp = ifnet_byindex_locked(idx);
359 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
360 		IFNET_RUNLOCK_NOSLEEP();
361 		return (NULL);
362 	}
363 	if_ref(ifp);
364 	IFNET_RUNLOCK_NOSLEEP();
365 	return (ifp);
366 }
367 
368 /*
369  * Allocate an ifindex array entry; return 0 on success or an error on
370  * failure.
371  */
372 static u_short
373 ifindex_alloc(void)
374 {
375 	u_short idx;
376 
377 	IFNET_WLOCK_ASSERT();
378 retry:
379 	/*
380 	 * Try to find an empty slot below V_if_index.  If we fail, take the
381 	 * next slot.
382 	 */
383 	for (idx = 1; idx <= V_if_index; idx++) {
384 		if (V_ifindex_table[idx] == NULL)
385 			break;
386 	}
387 
388 	/* Catch if_index overflow. */
389 	if (idx >= V_if_indexlim) {
390 		if_grow();
391 		goto retry;
392 	}
393 	if (idx > V_if_index)
394 		V_if_index = idx;
395 	return (idx);
396 }
397 
398 static void
399 ifindex_free_locked(u_short idx)
400 {
401 
402 	IFNET_WLOCK_ASSERT();
403 
404 	V_ifindex_table[idx] = NULL;
405 	while (V_if_index > 0 &&
406 	    V_ifindex_table[V_if_index] == NULL)
407 		V_if_index--;
408 }
409 
410 static void
411 ifindex_free(u_short idx)
412 {
413 
414 	IFNET_WLOCK();
415 	ifindex_free_locked(idx);
416 	IFNET_WUNLOCK();
417 }
418 
419 static void
420 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
421 {
422 
423 	IFNET_WLOCK_ASSERT();
424 
425 	V_ifindex_table[idx] = ifp;
426 }
427 
428 static void
429 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
430 {
431 
432 	IFNET_WLOCK();
433 	ifnet_setbyindex_locked(idx, ifp);
434 	IFNET_WUNLOCK();
435 }
436 
437 struct ifaddr *
438 ifaddr_byindex(u_short idx)
439 {
440 	struct ifnet *ifp;
441 	struct ifaddr *ifa = NULL;
442 
443 	IFNET_RLOCK_NOSLEEP();
444 	ifp = ifnet_byindex_locked(idx);
445 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
446 		ifa_ref(ifa);
447 	IFNET_RUNLOCK_NOSLEEP();
448 	return (ifa);
449 }
450 
451 /*
452  * Network interface utility routines.
453  *
454  * Routines with ifa_ifwith* names take sockaddr *'s as
455  * parameters.
456  */
457 
458 static void
459 vnet_if_init(const void *unused __unused)
460 {
461 
462 	TAILQ_INIT(&V_ifnet);
463 	TAILQ_INIT(&V_ifg_head);
464 	IFNET_WLOCK();
465 	if_grow();				/* create initial table */
466 	IFNET_WUNLOCK();
467 	vnet_if_clone_init();
468 }
469 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
470     NULL);
471 
472 #ifdef VIMAGE
473 static void
474 vnet_if_uninit(const void *unused __unused)
475 {
476 
477 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
478 	    "not empty", __func__, __LINE__, &V_ifnet));
479 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
480 	    "not empty", __func__, __LINE__, &V_ifg_head));
481 
482 	free((caddr_t)V_ifindex_table, M_IFNET);
483 }
484 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
485     vnet_if_uninit, NULL);
486 
487 static void
488 vnet_if_return(const void *unused __unused)
489 {
490 	struct ifnet *ifp, *nifp;
491 
492 	/* Return all inherited interfaces to their parent vnets. */
493 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
494 		if (ifp->if_home_vnet != ifp->if_vnet)
495 			if_vmove(ifp, ifp->if_home_vnet);
496 	}
497 }
498 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
499     vnet_if_return, NULL);
500 #endif
501 
502 static void
503 if_grow(void)
504 {
505 	int oldlim;
506 	u_int n;
507 	struct ifnet **e;
508 
509 	IFNET_WLOCK_ASSERT();
510 	oldlim = V_if_indexlim;
511 	IFNET_WUNLOCK();
512 	n = (oldlim << 1) * sizeof(*e);
513 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
514 	IFNET_WLOCK();
515 	if (V_if_indexlim != oldlim) {
516 		free(e, M_IFNET);
517 		return;
518 	}
519 	if (V_ifindex_table != NULL) {
520 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
521 		free((caddr_t)V_ifindex_table, M_IFNET);
522 	}
523 	V_if_indexlim <<= 1;
524 	V_ifindex_table = e;
525 }
526 
527 /*
528  * Allocate a struct ifnet and an index for an interface.  A layer 2
529  * common structure will also be allocated if an allocation routine is
530  * registered for the passed type.
531  */
532 struct ifnet *
533 if_alloc(u_char type)
534 {
535 	struct ifnet *ifp;
536 	u_short idx;
537 
538 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
539 	IFNET_WLOCK();
540 	idx = ifindex_alloc();
541 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
542 	IFNET_WUNLOCK();
543 	ifp->if_index = idx;
544 	ifp->if_type = type;
545 	ifp->if_alloctype = type;
546 #ifdef VIMAGE
547 	ifp->if_vnet = curvnet;
548 #endif
549 	if (if_com_alloc[type] != NULL) {
550 		ifp->if_l2com = if_com_alloc[type](type, ifp);
551 		if (ifp->if_l2com == NULL) {
552 			free(ifp, M_IFNET);
553 			ifindex_free(idx);
554 			return (NULL);
555 		}
556 	}
557 
558 	IF_ADDR_LOCK_INIT(ifp);
559 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
560 	ifp->if_afdata_initialized = 0;
561 	IF_AFDATA_LOCK_INIT(ifp);
562 	TAILQ_INIT(&ifp->if_addrhead);
563 	TAILQ_INIT(&ifp->if_multiaddrs);
564 	TAILQ_INIT(&ifp->if_groups);
565 #ifdef MAC
566 	mac_ifnet_init(ifp);
567 #endif
568 	ifq_init(&ifp->if_snd, ifp);
569 
570 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
571 	for (int i = 0; i < IFCOUNTERS; i++)
572 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
573 	ifp->if_get_counter = if_get_counter_default;
574 	ifp->if_pcp = IFNET_PCP_NONE;
575 	ifnet_setbyindex(ifp->if_index, ifp);
576 	return (ifp);
577 }
578 
579 /*
580  * Do the actual work of freeing a struct ifnet, and layer 2 common
581  * structure.  This call is made when the last reference to an
582  * interface is released.
583  */
584 static void
585 if_free_internal(struct ifnet *ifp)
586 {
587 
588 	KASSERT((ifp->if_flags & IFF_DYING),
589 	    ("if_free_internal: interface not dying"));
590 
591 	if (if_com_free[ifp->if_alloctype] != NULL)
592 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
593 		    ifp->if_alloctype);
594 
595 #ifdef MAC
596 	mac_ifnet_destroy(ifp);
597 #endif /* MAC */
598 	if (ifp->if_description != NULL)
599 		free(ifp->if_description, M_IFDESCR);
600 	IF_AFDATA_DESTROY(ifp);
601 	IF_ADDR_LOCK_DESTROY(ifp);
602 	ifq_delete(&ifp->if_snd);
603 
604 	for (int i = 0; i < IFCOUNTERS; i++)
605 		counter_u64_free(ifp->if_counters[i]);
606 
607 	free(ifp, M_IFNET);
608 }
609 
610 /*
611  * Deregister an interface and free the associated storage.
612  */
613 void
614 if_free(struct ifnet *ifp)
615 {
616 
617 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
618 
619 	CURVNET_SET_QUIET(ifp->if_vnet);
620 	IFNET_WLOCK();
621 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
622 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
623 
624 	ifindex_free_locked(ifp->if_index);
625 	IFNET_WUNLOCK();
626 
627 	if (refcount_release(&ifp->if_refcount))
628 		if_free_internal(ifp);
629 	CURVNET_RESTORE();
630 }
631 
632 /*
633  * Interfaces to keep an ifnet type-stable despite the possibility of the
634  * driver calling if_free().  If there are additional references, we defer
635  * freeing the underlying data structure.
636  */
637 void
638 if_ref(struct ifnet *ifp)
639 {
640 
641 	/* We don't assert the ifnet list lock here, but arguably should. */
642 	refcount_acquire(&ifp->if_refcount);
643 }
644 
645 void
646 if_rele(struct ifnet *ifp)
647 {
648 
649 	if (!refcount_release(&ifp->if_refcount))
650 		return;
651 	if_free_internal(ifp);
652 }
653 
654 void
655 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
656 {
657 
658 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
659 
660 	if (ifq->ifq_maxlen == 0)
661 		ifq->ifq_maxlen = ifqmaxlen;
662 
663 	ifq->altq_type = 0;
664 	ifq->altq_disc = NULL;
665 	ifq->altq_flags &= ALTQF_CANTCHANGE;
666 	ifq->altq_tbr  = NULL;
667 	ifq->altq_ifp  = ifp;
668 }
669 
670 void
671 ifq_delete(struct ifaltq *ifq)
672 {
673 	mtx_destroy(&ifq->ifq_mtx);
674 }
675 
676 /*
677  * Perform generic interface initialization tasks and attach the interface
678  * to the list of "active" interfaces.  If vmove flag is set on entry
679  * to if_attach_internal(), perform only a limited subset of initialization
680  * tasks, given that we are moving from one vnet to another an ifnet which
681  * has already been fully initialized.
682  *
683  * Note that if_detach_internal() removes group membership unconditionally
684  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
685  * Thus, when if_vmove() is applied to a cloned interface, group membership
686  * is lost while a cloned one always joins a group whose name is
687  * ifc->ifc_name.  To recover this after if_detach_internal() and
688  * if_attach_internal(), the cloner should be specified to
689  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
690  * attempts to join a group whose name is ifc->ifc_name.
691  *
692  * XXX:
693  *  - The decision to return void and thus require this function to
694  *    succeed is questionable.
695  *  - We should probably do more sanity checking.  For instance we don't
696  *    do anything to insure if_xname is unique or non-empty.
697  */
698 void
699 if_attach(struct ifnet *ifp)
700 {
701 
702 	if_attach_internal(ifp, 0, NULL);
703 }
704 
705 /*
706  * Compute the least common TSO limit.
707  */
708 void
709 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
710 {
711 	/*
712 	 * 1) If there is no limit currently, take the limit from
713 	 * the network adapter.
714 	 *
715 	 * 2) If the network adapter has a limit below the current
716 	 * limit, apply it.
717 	 */
718 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
719 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
720 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
721 	}
722 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
723 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
724 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
725 	}
726 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
727 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
728 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
729 	}
730 }
731 
732 /*
733  * Update TSO limit of a network adapter.
734  *
735  * Returns zero if no change. Else non-zero.
736  */
737 int
738 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
739 {
740 	int retval = 0;
741 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
742 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
743 		retval++;
744 	}
745 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
746 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
747 		retval++;
748 	}
749 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
750 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
751 		retval++;
752 	}
753 	return (retval);
754 }
755 
756 static void
757 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
758 {
759 	unsigned socksize, ifasize;
760 	int namelen, masklen;
761 	struct sockaddr_dl *sdl;
762 	struct ifaddr *ifa;
763 
764 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
765 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
766 		    ifp->if_xname);
767 
768 #ifdef VIMAGE
769 	ifp->if_vnet = curvnet;
770 	if (ifp->if_home_vnet == NULL)
771 		ifp->if_home_vnet = curvnet;
772 #endif
773 
774 	if_addgroup(ifp, IFG_ALL);
775 
776 	/* Restore group membership for cloned interfaces. */
777 	if (vmove && ifc != NULL)
778 		if_clone_addgroup(ifp, ifc);
779 
780 	getmicrotime(&ifp->if_lastchange);
781 	ifp->if_epoch = time_uptime;
782 
783 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
784 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
785 	    ("transmit and qflush must both either be set or both be NULL"));
786 	if (ifp->if_transmit == NULL) {
787 		ifp->if_transmit = if_transmit;
788 		ifp->if_qflush = if_qflush;
789 	}
790 	if (ifp->if_input == NULL)
791 		ifp->if_input = if_input_default;
792 
793 	if (ifp->if_requestencap == NULL)
794 		ifp->if_requestencap = if_requestencap_default;
795 
796 	if (!vmove) {
797 #ifdef MAC
798 		mac_ifnet_create(ifp);
799 #endif
800 
801 		/*
802 		 * Create a Link Level name for this device.
803 		 */
804 		namelen = strlen(ifp->if_xname);
805 		/*
806 		 * Always save enough space for any possiable name so we
807 		 * can do a rename in place later.
808 		 */
809 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
810 		socksize = masklen + ifp->if_addrlen;
811 		if (socksize < sizeof(*sdl))
812 			socksize = sizeof(*sdl);
813 		socksize = roundup2(socksize, sizeof(long));
814 		ifasize = sizeof(*ifa) + 2 * socksize;
815 		ifa = ifa_alloc(ifasize, M_WAITOK);
816 		sdl = (struct sockaddr_dl *)(ifa + 1);
817 		sdl->sdl_len = socksize;
818 		sdl->sdl_family = AF_LINK;
819 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
820 		sdl->sdl_nlen = namelen;
821 		sdl->sdl_index = ifp->if_index;
822 		sdl->sdl_type = ifp->if_type;
823 		ifp->if_addr = ifa;
824 		ifa->ifa_ifp = ifp;
825 		ifa->ifa_rtrequest = link_rtrequest;
826 		ifa->ifa_addr = (struct sockaddr *)sdl;
827 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
828 		ifa->ifa_netmask = (struct sockaddr *)sdl;
829 		sdl->sdl_len = masklen;
830 		while (namelen != 0)
831 			sdl->sdl_data[--namelen] = 0xff;
832 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
833 		/* Reliably crash if used uninitialized. */
834 		ifp->if_broadcastaddr = NULL;
835 
836 		if (ifp->if_type == IFT_ETHER) {
837 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
838 			    M_WAITOK | M_ZERO);
839 		}
840 
841 #if defined(INET) || defined(INET6)
842 		/* Use defaults for TSO, if nothing is set */
843 		if (ifp->if_hw_tsomax == 0 &&
844 		    ifp->if_hw_tsomaxsegcount == 0 &&
845 		    ifp->if_hw_tsomaxsegsize == 0) {
846 			/*
847 			 * The TSO defaults needs to be such that an
848 			 * NFS mbuf list of 35 mbufs totalling just
849 			 * below 64K works and that a chain of mbufs
850 			 * can be defragged into at most 32 segments:
851 			 */
852 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
853 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
854 			ifp->if_hw_tsomaxsegcount = 35;
855 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
856 
857 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
858 			if (ifp->if_capabilities & IFCAP_TSO) {
859 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
860 				    ifp->if_hw_tsomax,
861 				    ifp->if_hw_tsomaxsegcount,
862 				    ifp->if_hw_tsomaxsegsize);
863 			}
864 		}
865 #endif
866 	}
867 #ifdef VIMAGE
868 	else {
869 		/*
870 		 * Update the interface index in the link layer address
871 		 * of the interface.
872 		 */
873 		for (ifa = ifp->if_addr; ifa != NULL;
874 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
875 			if (ifa->ifa_addr->sa_family == AF_LINK) {
876 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
877 				sdl->sdl_index = ifp->if_index;
878 			}
879 		}
880 	}
881 #endif
882 
883 	IFNET_WLOCK();
884 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
885 #ifdef VIMAGE
886 	curvnet->vnet_ifcnt++;
887 #endif
888 	IFNET_WUNLOCK();
889 
890 	if (domain_init_status >= 2)
891 		if_attachdomain1(ifp);
892 
893 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
894 	if (IS_DEFAULT_VNET(curvnet))
895 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
896 
897 	/* Announce the interface. */
898 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
899 }
900 
901 static void
902 if_attachdomain(void *dummy)
903 {
904 	struct ifnet *ifp;
905 
906 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
907 		if_attachdomain1(ifp);
908 }
909 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
910     if_attachdomain, NULL);
911 
912 static void
913 if_attachdomain1(struct ifnet *ifp)
914 {
915 	struct domain *dp;
916 
917 	/*
918 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
919 	 * cannot lock ifp->if_afdata initialization, entirely.
920 	 */
921 	IF_AFDATA_LOCK(ifp);
922 	if (ifp->if_afdata_initialized >= domain_init_status) {
923 		IF_AFDATA_UNLOCK(ifp);
924 		log(LOG_WARNING, "%s called more than once on %s\n",
925 		    __func__, ifp->if_xname);
926 		return;
927 	}
928 	ifp->if_afdata_initialized = domain_init_status;
929 	IF_AFDATA_UNLOCK(ifp);
930 
931 	/* address family dependent data region */
932 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
933 	for (dp = domains; dp; dp = dp->dom_next) {
934 		if (dp->dom_ifattach)
935 			ifp->if_afdata[dp->dom_family] =
936 			    (*dp->dom_ifattach)(ifp);
937 	}
938 }
939 
940 /*
941  * Remove any unicast or broadcast network addresses from an interface.
942  */
943 void
944 if_purgeaddrs(struct ifnet *ifp)
945 {
946 	struct ifaddr *ifa, *next;
947 
948 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
949 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
950 		if (ifa->ifa_addr->sa_family == AF_LINK)
951 			continue;
952 #ifdef INET
953 		/* XXX: Ugly!! ad hoc just for INET */
954 		if (ifa->ifa_addr->sa_family == AF_INET) {
955 			struct ifaliasreq ifr;
956 
957 			bzero(&ifr, sizeof(ifr));
958 			ifr.ifra_addr = *ifa->ifa_addr;
959 			if (ifa->ifa_dstaddr)
960 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
961 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
962 			    NULL) == 0)
963 				continue;
964 		}
965 #endif /* INET */
966 #ifdef INET6
967 		if (ifa->ifa_addr->sa_family == AF_INET6) {
968 			in6_purgeaddr(ifa);
969 			/* ifp_addrhead is already updated */
970 			continue;
971 		}
972 #endif /* INET6 */
973 		IF_ADDR_WLOCK(ifp);
974 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
975 		IF_ADDR_WUNLOCK(ifp);
976 		ifa_free(ifa);
977 	}
978 }
979 
980 /*
981  * Remove any multicast network addresses from an interface when an ifnet
982  * is going away.
983  */
984 static void
985 if_purgemaddrs(struct ifnet *ifp)
986 {
987 	struct ifmultiaddr *ifma;
988 	struct ifmultiaddr *next;
989 
990 	IF_ADDR_WLOCK(ifp);
991 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
992 		if_delmulti_locked(ifp, ifma, 1);
993 	IF_ADDR_WUNLOCK(ifp);
994 }
995 
996 /*
997  * Detach an interface, removing it from the list of "active" interfaces.
998  * If vmove flag is set on entry to if_detach_internal(), perform only a
999  * limited subset of cleanup tasks, given that we are moving an ifnet from
1000  * one vnet to another, where it must be fully operational.
1001  *
1002  * XXXRW: There are some significant questions about event ordering, and
1003  * how to prevent things from starting to use the interface during detach.
1004  */
1005 void
1006 if_detach(struct ifnet *ifp)
1007 {
1008 
1009 	CURVNET_SET_QUIET(ifp->if_vnet);
1010 	if_detach_internal(ifp, 0, NULL);
1011 	CURVNET_RESTORE();
1012 }
1013 
1014 /*
1015  * The vmove flag, if set, indicates that we are called from a callpath
1016  * that is moving an interface to a different vnet instance.
1017  *
1018  * The shutdown flag, if set, indicates that we are called in the
1019  * process of shutting down a vnet instance.  Currently only the
1020  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1021  * on a vnet instance shutdown without this flag being set, e.g., when
1022  * the cloned interfaces are destoyed as first thing of teardown.
1023  */
1024 static int
1025 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1026 {
1027 	struct ifaddr *ifa;
1028 	int i;
1029 	struct domain *dp;
1030  	struct ifnet *iter;
1031  	int found = 0;
1032 #ifdef VIMAGE
1033 	int shutdown;
1034 
1035 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1036 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1037 #endif
1038 	IFNET_WLOCK();
1039 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
1040 		if (iter == ifp) {
1041 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
1042 			found = 1;
1043 			break;
1044 		}
1045 	IFNET_WUNLOCK();
1046 	if (!found) {
1047 		/*
1048 		 * While we would want to panic here, we cannot
1049 		 * guarantee that the interface is indeed still on
1050 		 * the list given we don't hold locks all the way.
1051 		 */
1052 		return (ENOENT);
1053 #if 0
1054 		if (vmove)
1055 			panic("%s: ifp=%p not on the ifnet tailq %p",
1056 			    __func__, ifp, &V_ifnet);
1057 		else
1058 			return; /* XXX this should panic as well? */
1059 #endif
1060 	}
1061 
1062 	/*
1063 	 * At this point we know the interface still was on the ifnet list
1064 	 * and we removed it so we are in a stable state.
1065 	 */
1066 #ifdef VIMAGE
1067 	curvnet->vnet_ifcnt--;
1068 #endif
1069 
1070 	/*
1071 	 * In any case (destroy or vmove) detach us from the groups
1072 	 * and remove/wait for pending events on the taskq.
1073 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1074 	 */
1075 	if_delgroups(ifp);
1076 
1077 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1078 
1079 	/*
1080 	 * Check if this is a cloned interface or not. Must do even if
1081 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1082 	 * the if_clone_addgroup() will have a corrupted string overwise
1083 	 * from a gibberish pointer.
1084 	 */
1085 	if (vmove && ifcp != NULL)
1086 		*ifcp = if_clone_findifc(ifp);
1087 
1088 	if_down(ifp);
1089 
1090 #ifdef VIMAGE
1091 	/*
1092 	 * On VNET shutdown abort here as the stack teardown will do all
1093 	 * the work top-down for us.
1094 	 */
1095 	if (shutdown) {
1096 		/*
1097 		 * In case of a vmove we are done here without error.
1098 		 * If we would signal an error it would lead to the same
1099 		 * abort as if we did not find the ifnet anymore.
1100 		 * if_detach() calls us in void context and does not care
1101 		 * about an early abort notification, so life is splendid :)
1102 		 */
1103 		goto finish_vnet_shutdown;
1104 	}
1105 #endif
1106 
1107 	/*
1108 	 * At this point we are not tearing down a VNET and are either
1109 	 * going to destroy or vmove the interface and have to cleanup
1110 	 * accordingly.
1111 	 */
1112 
1113 	/*
1114 	 * Remove routes and flush queues.
1115 	 */
1116 #ifdef ALTQ
1117 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1118 		altq_disable(&ifp->if_snd);
1119 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1120 		altq_detach(&ifp->if_snd);
1121 #endif
1122 
1123 	if_purgeaddrs(ifp);
1124 
1125 #ifdef INET
1126 	in_ifdetach(ifp);
1127 #endif
1128 
1129 #ifdef INET6
1130 	/*
1131 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1132 	 * before removing routing entries below, since IPv6 interface direct
1133 	 * routes are expected to be removed by the IPv6-specific kernel API.
1134 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1135 	 */
1136 	in6_ifdetach(ifp);
1137 #endif
1138 	if_purgemaddrs(ifp);
1139 
1140 	/* Announce that the interface is gone. */
1141 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1142 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1143 	if (IS_DEFAULT_VNET(curvnet))
1144 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1145 
1146 	if (!vmove) {
1147 		/*
1148 		 * Prevent further calls into the device driver via ifnet.
1149 		 */
1150 		if_dead(ifp);
1151 
1152 		/*
1153 		 * Remove link ifaddr pointer and maybe decrement if_index.
1154 		 * Clean up all addresses.
1155 		 */
1156 		free(ifp->if_hw_addr, M_IFADDR);
1157 		ifp->if_hw_addr = NULL;
1158 		ifp->if_addr = NULL;
1159 
1160 		/* We can now free link ifaddr. */
1161 		IF_ADDR_WLOCK(ifp);
1162 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1163 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1164 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1165 			IF_ADDR_WUNLOCK(ifp);
1166 			ifa_free(ifa);
1167 		} else
1168 			IF_ADDR_WUNLOCK(ifp);
1169 	}
1170 
1171 	rt_flushifroutes(ifp);
1172 
1173 #ifdef VIMAGE
1174 finish_vnet_shutdown:
1175 #endif
1176 	/*
1177 	 * We cannot hold the lock over dom_ifdetach calls as they might
1178 	 * sleep, for example trying to drain a callout, thus open up the
1179 	 * theoretical race with re-attaching.
1180 	 */
1181 	IF_AFDATA_LOCK(ifp);
1182 	i = ifp->if_afdata_initialized;
1183 	ifp->if_afdata_initialized = 0;
1184 	IF_AFDATA_UNLOCK(ifp);
1185 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1186 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1187 			(*dp->dom_ifdetach)(ifp,
1188 			    ifp->if_afdata[dp->dom_family]);
1189 			ifp->if_afdata[dp->dom_family] = NULL;
1190 		}
1191 	}
1192 
1193 	return (0);
1194 }
1195 
1196 #ifdef VIMAGE
1197 /*
1198  * if_vmove() performs a limited version of if_detach() in current
1199  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1200  * An attempt is made to shrink if_index in current vnet, find an
1201  * unused if_index in target vnet and calls if_grow() if necessary,
1202  * and finally find an unused if_xname for the target vnet.
1203  */
1204 static void
1205 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1206 {
1207 	struct if_clone *ifc;
1208 	u_int bif_dlt, bif_hdrlen;
1209 	int rc;
1210 
1211  	/*
1212 	 * if_detach_internal() will call the eventhandler to notify
1213 	 * interface departure.  That will detach if_bpf.  We need to
1214 	 * safe the dlt and hdrlen so we can re-attach it later.
1215 	 */
1216 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1217 
1218 	/*
1219 	 * Detach from current vnet, but preserve LLADDR info, do not
1220 	 * mark as dead etc. so that the ifnet can be reattached later.
1221 	 * If we cannot find it, we lost the race to someone else.
1222 	 */
1223 	rc = if_detach_internal(ifp, 1, &ifc);
1224 	if (rc != 0)
1225 		return;
1226 
1227 	/*
1228 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1229 	 * the if_index for that vnet if possible.
1230 	 *
1231 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1232 	 * or we'd lock on one vnet and unlock on another.
1233 	 */
1234 	IFNET_WLOCK();
1235 	ifindex_free_locked(ifp->if_index);
1236 	IFNET_WUNLOCK();
1237 
1238 	/*
1239 	 * Perform interface-specific reassignment tasks, if provided by
1240 	 * the driver.
1241 	 */
1242 	if (ifp->if_reassign != NULL)
1243 		ifp->if_reassign(ifp, new_vnet, NULL);
1244 
1245 	/*
1246 	 * Switch to the context of the target vnet.
1247 	 */
1248 	CURVNET_SET_QUIET(new_vnet);
1249 
1250 	IFNET_WLOCK();
1251 	ifp->if_index = ifindex_alloc();
1252 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1253 	IFNET_WUNLOCK();
1254 
1255 	if_attach_internal(ifp, 1, ifc);
1256 
1257 	if (ifp->if_bpf == NULL)
1258 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1259 
1260 	CURVNET_RESTORE();
1261 }
1262 
1263 /*
1264  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1265  */
1266 static int
1267 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1268 {
1269 	struct prison *pr;
1270 	struct ifnet *difp;
1271 	int shutdown;
1272 
1273 	/* Try to find the prison within our visibility. */
1274 	sx_slock(&allprison_lock);
1275 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1276 	sx_sunlock(&allprison_lock);
1277 	if (pr == NULL)
1278 		return (ENXIO);
1279 	prison_hold_locked(pr);
1280 	mtx_unlock(&pr->pr_mtx);
1281 
1282 	/* Do not try to move the iface from and to the same prison. */
1283 	if (pr->pr_vnet == ifp->if_vnet) {
1284 		prison_free(pr);
1285 		return (EEXIST);
1286 	}
1287 
1288 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1289 	/* XXX Lock interfaces to avoid races. */
1290 	CURVNET_SET_QUIET(pr->pr_vnet);
1291 	difp = ifunit(ifname);
1292 	if (difp != NULL) {
1293 		CURVNET_RESTORE();
1294 		prison_free(pr);
1295 		return (EEXIST);
1296 	}
1297 
1298 	/* Make sure the VNET is stable. */
1299 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1300 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1301 	if (shutdown) {
1302 		CURVNET_RESTORE();
1303 		prison_free(pr);
1304 		return (EBUSY);
1305 	}
1306 	CURVNET_RESTORE();
1307 
1308 	/* Move the interface into the child jail/vnet. */
1309 	if_vmove(ifp, pr->pr_vnet);
1310 
1311 	/* Report the new if_xname back to the userland. */
1312 	sprintf(ifname, "%s", ifp->if_xname);
1313 
1314 	prison_free(pr);
1315 	return (0);
1316 }
1317 
1318 static int
1319 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1320 {
1321 	struct prison *pr;
1322 	struct vnet *vnet_dst;
1323 	struct ifnet *ifp;
1324  	int shutdown;
1325 
1326 	/* Try to find the prison within our visibility. */
1327 	sx_slock(&allprison_lock);
1328 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1329 	sx_sunlock(&allprison_lock);
1330 	if (pr == NULL)
1331 		return (ENXIO);
1332 	prison_hold_locked(pr);
1333 	mtx_unlock(&pr->pr_mtx);
1334 
1335 	/* Make sure the named iface exists in the source prison/vnet. */
1336 	CURVNET_SET(pr->pr_vnet);
1337 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1338 	if (ifp == NULL) {
1339 		CURVNET_RESTORE();
1340 		prison_free(pr);
1341 		return (ENXIO);
1342 	}
1343 
1344 	/* Do not try to move the iface from and to the same prison. */
1345 	vnet_dst = TD_TO_VNET(td);
1346 	if (vnet_dst == ifp->if_vnet) {
1347 		CURVNET_RESTORE();
1348 		prison_free(pr);
1349 		return (EEXIST);
1350 	}
1351 
1352 	/* Make sure the VNET is stable. */
1353 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1354 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1355 	if (shutdown) {
1356 		CURVNET_RESTORE();
1357 		prison_free(pr);
1358 		return (EBUSY);
1359 	}
1360 
1361 	/* Get interface back from child jail/vnet. */
1362 	if_vmove(ifp, vnet_dst);
1363 	CURVNET_RESTORE();
1364 
1365 	/* Report the new if_xname back to the userland. */
1366 	sprintf(ifname, "%s", ifp->if_xname);
1367 
1368 	prison_free(pr);
1369 	return (0);
1370 }
1371 #endif /* VIMAGE */
1372 
1373 /*
1374  * Add a group to an interface
1375  */
1376 int
1377 if_addgroup(struct ifnet *ifp, const char *groupname)
1378 {
1379 	struct ifg_list		*ifgl;
1380 	struct ifg_group	*ifg = NULL;
1381 	struct ifg_member	*ifgm;
1382 	int 			 new = 0;
1383 
1384 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1385 	    groupname[strlen(groupname) - 1] <= '9')
1386 		return (EINVAL);
1387 
1388 	IFNET_WLOCK();
1389 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1390 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1391 			IFNET_WUNLOCK();
1392 			return (EEXIST);
1393 		}
1394 
1395 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1396 	    M_NOWAIT)) == NULL) {
1397 	    	IFNET_WUNLOCK();
1398 		return (ENOMEM);
1399 	}
1400 
1401 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1402 	    M_TEMP, M_NOWAIT)) == NULL) {
1403 		free(ifgl, M_TEMP);
1404 		IFNET_WUNLOCK();
1405 		return (ENOMEM);
1406 	}
1407 
1408 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1409 		if (!strcmp(ifg->ifg_group, groupname))
1410 			break;
1411 
1412 	if (ifg == NULL) {
1413 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1414 		    M_TEMP, M_NOWAIT)) == NULL) {
1415 			free(ifgl, M_TEMP);
1416 			free(ifgm, M_TEMP);
1417 			IFNET_WUNLOCK();
1418 			return (ENOMEM);
1419 		}
1420 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1421 		ifg->ifg_refcnt = 0;
1422 		TAILQ_INIT(&ifg->ifg_members);
1423 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1424 		new = 1;
1425 	}
1426 
1427 	ifg->ifg_refcnt++;
1428 	ifgl->ifgl_group = ifg;
1429 	ifgm->ifgm_ifp = ifp;
1430 
1431 	IF_ADDR_WLOCK(ifp);
1432 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1433 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1434 	IF_ADDR_WUNLOCK(ifp);
1435 
1436 	IFNET_WUNLOCK();
1437 
1438 	if (new)
1439 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1440 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1441 
1442 	return (0);
1443 }
1444 
1445 /*
1446  * Remove a group from an interface
1447  */
1448 int
1449 if_delgroup(struct ifnet *ifp, const char *groupname)
1450 {
1451 	struct ifg_list		*ifgl;
1452 	struct ifg_member	*ifgm;
1453 
1454 	IFNET_WLOCK();
1455 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1456 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1457 			break;
1458 	if (ifgl == NULL) {
1459 		IFNET_WUNLOCK();
1460 		return (ENOENT);
1461 	}
1462 
1463 	IF_ADDR_WLOCK(ifp);
1464 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1465 	IF_ADDR_WUNLOCK(ifp);
1466 
1467 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1468 		if (ifgm->ifgm_ifp == ifp)
1469 			break;
1470 
1471 	if (ifgm != NULL) {
1472 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1473 		free(ifgm, M_TEMP);
1474 	}
1475 
1476 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1477 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1478 		IFNET_WUNLOCK();
1479 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1480 		free(ifgl->ifgl_group, M_TEMP);
1481 	} else
1482 		IFNET_WUNLOCK();
1483 
1484 	free(ifgl, M_TEMP);
1485 
1486 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1487 
1488 	return (0);
1489 }
1490 
1491 /*
1492  * Remove an interface from all groups
1493  */
1494 static void
1495 if_delgroups(struct ifnet *ifp)
1496 {
1497 	struct ifg_list		*ifgl;
1498 	struct ifg_member	*ifgm;
1499 	char groupname[IFNAMSIZ];
1500 
1501 	IFNET_WLOCK();
1502 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1503 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1504 
1505 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1506 
1507 		IF_ADDR_WLOCK(ifp);
1508 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1509 		IF_ADDR_WUNLOCK(ifp);
1510 
1511 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1512 			if (ifgm->ifgm_ifp == ifp)
1513 				break;
1514 
1515 		if (ifgm != NULL) {
1516 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1517 			    ifgm_next);
1518 			free(ifgm, M_TEMP);
1519 		}
1520 
1521 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1522 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1523 			IFNET_WUNLOCK();
1524 			EVENTHANDLER_INVOKE(group_detach_event,
1525 			    ifgl->ifgl_group);
1526 			free(ifgl->ifgl_group, M_TEMP);
1527 		} else
1528 			IFNET_WUNLOCK();
1529 
1530 		free(ifgl, M_TEMP);
1531 
1532 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1533 
1534 		IFNET_WLOCK();
1535 	}
1536 	IFNET_WUNLOCK();
1537 }
1538 
1539 static char *
1540 ifgr_group_get(void *ifgrp)
1541 {
1542 	union ifgroupreq_union *ifgrup;
1543 
1544 	ifgrup = ifgrp;
1545 #ifdef COMPAT_FREEBSD32
1546 	if (SV_CURPROC_FLAG(SV_ILP32))
1547 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1548 #endif
1549 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1550 }
1551 
1552 static struct ifg_req *
1553 ifgr_groups_get(void *ifgrp)
1554 {
1555 	union ifgroupreq_union *ifgrup;
1556 
1557 	ifgrup = ifgrp;
1558 #ifdef COMPAT_FREEBSD32
1559 	if (SV_CURPROC_FLAG(SV_ILP32))
1560 		return ((struct ifg_req *)(uintptr_t)
1561 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1562 #endif
1563 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1564 }
1565 
1566 /*
1567  * Stores all groups from an interface in memory pointed to by ifgr.
1568  */
1569 static int
1570 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1571 {
1572 	int			 len, error;
1573 	struct ifg_list		*ifgl;
1574 	struct ifg_req		 ifgrq, *ifgp;
1575 
1576 	if (ifgr->ifgr_len == 0) {
1577 		IF_ADDR_RLOCK(ifp);
1578 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1579 			ifgr->ifgr_len += sizeof(struct ifg_req);
1580 		IF_ADDR_RUNLOCK(ifp);
1581 		return (0);
1582 	}
1583 
1584 	len = ifgr->ifgr_len;
1585 	ifgp = ifgr_groups_get(ifgr);
1586 	/* XXX: wire */
1587 	IF_ADDR_RLOCK(ifp);
1588 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1589 		if (len < sizeof(ifgrq)) {
1590 			IF_ADDR_RUNLOCK(ifp);
1591 			return (EINVAL);
1592 		}
1593 		bzero(&ifgrq, sizeof ifgrq);
1594 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1595 		    sizeof(ifgrq.ifgrq_group));
1596 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1597 		    	IF_ADDR_RUNLOCK(ifp);
1598 			return (error);
1599 		}
1600 		len -= sizeof(ifgrq);
1601 		ifgp++;
1602 	}
1603 	IF_ADDR_RUNLOCK(ifp);
1604 
1605 	return (0);
1606 }
1607 
1608 /*
1609  * Stores all members of a group in memory pointed to by igfr
1610  */
1611 static int
1612 if_getgroupmembers(struct ifgroupreq *ifgr)
1613 {
1614 	struct ifg_group	*ifg;
1615 	struct ifg_member	*ifgm;
1616 	struct ifg_req		 ifgrq, *ifgp;
1617 	int			 len, error;
1618 
1619 	IFNET_RLOCK();
1620 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1621 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1622 			break;
1623 	if (ifg == NULL) {
1624 		IFNET_RUNLOCK();
1625 		return (ENOENT);
1626 	}
1627 
1628 	if (ifgr->ifgr_len == 0) {
1629 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1630 			ifgr->ifgr_len += sizeof(ifgrq);
1631 		IFNET_RUNLOCK();
1632 		return (0);
1633 	}
1634 
1635 	len = ifgr->ifgr_len;
1636 	ifgp = ifgr_groups_get(ifgr);
1637 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1638 		if (len < sizeof(ifgrq)) {
1639 			IFNET_RUNLOCK();
1640 			return (EINVAL);
1641 		}
1642 		bzero(&ifgrq, sizeof ifgrq);
1643 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1644 		    sizeof(ifgrq.ifgrq_member));
1645 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1646 			IFNET_RUNLOCK();
1647 			return (error);
1648 		}
1649 		len -= sizeof(ifgrq);
1650 		ifgp++;
1651 	}
1652 	IFNET_RUNLOCK();
1653 
1654 	return (0);
1655 }
1656 
1657 /*
1658  * Return counter values from counter(9)s stored in ifnet.
1659  */
1660 uint64_t
1661 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1662 {
1663 
1664 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1665 
1666 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1667 }
1668 
1669 /*
1670  * Increase an ifnet counter. Usually used for counters shared
1671  * between the stack and a driver, but function supports them all.
1672  */
1673 void
1674 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1675 {
1676 
1677 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1678 
1679 	counter_u64_add(ifp->if_counters[cnt], inc);
1680 }
1681 
1682 /*
1683  * Copy data from ifnet to userland API structure if_data.
1684  */
1685 void
1686 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1687 {
1688 
1689 	ifd->ifi_type = ifp->if_type;
1690 	ifd->ifi_physical = 0;
1691 	ifd->ifi_addrlen = ifp->if_addrlen;
1692 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1693 	ifd->ifi_link_state = ifp->if_link_state;
1694 	ifd->ifi_vhid = 0;
1695 	ifd->ifi_datalen = sizeof(struct if_data);
1696 	ifd->ifi_mtu = ifp->if_mtu;
1697 	ifd->ifi_metric = ifp->if_metric;
1698 	ifd->ifi_baudrate = ifp->if_baudrate;
1699 	ifd->ifi_hwassist = ifp->if_hwassist;
1700 	ifd->ifi_epoch = ifp->if_epoch;
1701 	ifd->ifi_lastchange = ifp->if_lastchange;
1702 
1703 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1704 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1705 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1706 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1707 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1708 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1709 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1710 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1711 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1712 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1713 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1714 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1715 }
1716 
1717 /*
1718  * Wrapper functions for struct ifnet address list locking macros.  These are
1719  * used by kernel modules to avoid encoding programming interface or binary
1720  * interface assumptions that may be violated when kernel-internal locking
1721  * approaches change.
1722  */
1723 void
1724 if_addr_rlock(struct ifnet *ifp)
1725 {
1726 
1727 	IF_ADDR_RLOCK(ifp);
1728 }
1729 
1730 void
1731 if_addr_runlock(struct ifnet *ifp)
1732 {
1733 
1734 	IF_ADDR_RUNLOCK(ifp);
1735 }
1736 
1737 void
1738 if_maddr_rlock(if_t ifp)
1739 {
1740 
1741 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1742 }
1743 
1744 void
1745 if_maddr_runlock(if_t ifp)
1746 {
1747 
1748 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1749 }
1750 
1751 /*
1752  * Initialization, destruction and refcounting functions for ifaddrs.
1753  */
1754 struct ifaddr *
1755 ifa_alloc(size_t size, int flags)
1756 {
1757 	struct ifaddr *ifa;
1758 
1759 	KASSERT(size >= sizeof(struct ifaddr),
1760 	    ("%s: invalid size %zu", __func__, size));
1761 
1762 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1763 	if (ifa == NULL)
1764 		return (NULL);
1765 
1766 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1767 		goto fail;
1768 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1769 		goto fail;
1770 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1771 		goto fail;
1772 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1773 		goto fail;
1774 
1775 	refcount_init(&ifa->ifa_refcnt, 1);
1776 
1777 	return (ifa);
1778 
1779 fail:
1780 	/* free(NULL) is okay */
1781 	counter_u64_free(ifa->ifa_opackets);
1782 	counter_u64_free(ifa->ifa_ipackets);
1783 	counter_u64_free(ifa->ifa_obytes);
1784 	counter_u64_free(ifa->ifa_ibytes);
1785 	free(ifa, M_IFADDR);
1786 
1787 	return (NULL);
1788 }
1789 
1790 void
1791 ifa_ref(struct ifaddr *ifa)
1792 {
1793 
1794 	refcount_acquire(&ifa->ifa_refcnt);
1795 }
1796 
1797 void
1798 ifa_free(struct ifaddr *ifa)
1799 {
1800 
1801 	if (refcount_release(&ifa->ifa_refcnt)) {
1802 		counter_u64_free(ifa->ifa_opackets);
1803 		counter_u64_free(ifa->ifa_ipackets);
1804 		counter_u64_free(ifa->ifa_obytes);
1805 		counter_u64_free(ifa->ifa_ibytes);
1806 		free(ifa, M_IFADDR);
1807 	}
1808 }
1809 
1810 static int
1811 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1812     struct sockaddr *ia)
1813 {
1814 	int error;
1815 	struct rt_addrinfo info;
1816 	struct sockaddr_dl null_sdl;
1817 	struct ifnet *ifp;
1818 
1819 	ifp = ifa->ifa_ifp;
1820 
1821 	bzero(&info, sizeof(info));
1822 	if (cmd != RTM_DELETE)
1823 		info.rti_ifp = V_loif;
1824 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1825 	info.rti_info[RTAX_DST] = ia;
1826 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1827 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1828 
1829 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1830 
1831 	if (error != 0)
1832 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1833 		    __func__, otype, if_name(ifp), error);
1834 
1835 	return (error);
1836 }
1837 
1838 int
1839 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1840 {
1841 
1842 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1843 }
1844 
1845 int
1846 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1847 {
1848 
1849 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1850 }
1851 
1852 int
1853 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1854 {
1855 
1856 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1857 }
1858 
1859 /*
1860  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1861  * structs used to represent other address families, it is necessary
1862  * to perform a different comparison.
1863  */
1864 
1865 #define	sa_dl_equal(a1, a2)	\
1866 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1867 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1868 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1869 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1870 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1871 
1872 /*
1873  * Locate an interface based on a complete address.
1874  */
1875 /*ARGSUSED*/
1876 static struct ifaddr *
1877 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1878 {
1879 	struct ifnet *ifp;
1880 	struct ifaddr *ifa;
1881 
1882 	IFNET_RLOCK_NOSLEEP();
1883 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1884 		IF_ADDR_RLOCK(ifp);
1885 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1886 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1887 				continue;
1888 			if (sa_equal(addr, ifa->ifa_addr)) {
1889 				if (getref)
1890 					ifa_ref(ifa);
1891 				IF_ADDR_RUNLOCK(ifp);
1892 				goto done;
1893 			}
1894 			/* IP6 doesn't have broadcast */
1895 			if ((ifp->if_flags & IFF_BROADCAST) &&
1896 			    ifa->ifa_broadaddr &&
1897 			    ifa->ifa_broadaddr->sa_len != 0 &&
1898 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1899 				if (getref)
1900 					ifa_ref(ifa);
1901 				IF_ADDR_RUNLOCK(ifp);
1902 				goto done;
1903 			}
1904 		}
1905 		IF_ADDR_RUNLOCK(ifp);
1906 	}
1907 	ifa = NULL;
1908 done:
1909 	IFNET_RUNLOCK_NOSLEEP();
1910 	return (ifa);
1911 }
1912 
1913 struct ifaddr *
1914 ifa_ifwithaddr(const struct sockaddr *addr)
1915 {
1916 
1917 	return (ifa_ifwithaddr_internal(addr, 1));
1918 }
1919 
1920 int
1921 ifa_ifwithaddr_check(const struct sockaddr *addr)
1922 {
1923 
1924 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1925 }
1926 
1927 /*
1928  * Locate an interface based on the broadcast address.
1929  */
1930 /* ARGSUSED */
1931 struct ifaddr *
1932 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1933 {
1934 	struct ifnet *ifp;
1935 	struct ifaddr *ifa;
1936 
1937 	IFNET_RLOCK_NOSLEEP();
1938 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1939 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1940 			continue;
1941 		IF_ADDR_RLOCK(ifp);
1942 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1943 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1944 				continue;
1945 			if ((ifp->if_flags & IFF_BROADCAST) &&
1946 			    ifa->ifa_broadaddr &&
1947 			    ifa->ifa_broadaddr->sa_len != 0 &&
1948 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1949 				ifa_ref(ifa);
1950 				IF_ADDR_RUNLOCK(ifp);
1951 				goto done;
1952 			}
1953 		}
1954 		IF_ADDR_RUNLOCK(ifp);
1955 	}
1956 	ifa = NULL;
1957 done:
1958 	IFNET_RUNLOCK_NOSLEEP();
1959 	return (ifa);
1960 }
1961 
1962 /*
1963  * Locate the point to point interface with a given destination address.
1964  */
1965 /*ARGSUSED*/
1966 struct ifaddr *
1967 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1968 {
1969 	struct ifnet *ifp;
1970 	struct ifaddr *ifa;
1971 
1972 	IFNET_RLOCK_NOSLEEP();
1973 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1974 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1975 			continue;
1976 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1977 			continue;
1978 		IF_ADDR_RLOCK(ifp);
1979 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1980 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1981 				continue;
1982 			if (ifa->ifa_dstaddr != NULL &&
1983 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1984 				ifa_ref(ifa);
1985 				IF_ADDR_RUNLOCK(ifp);
1986 				goto done;
1987 			}
1988 		}
1989 		IF_ADDR_RUNLOCK(ifp);
1990 	}
1991 	ifa = NULL;
1992 done:
1993 	IFNET_RUNLOCK_NOSLEEP();
1994 	return (ifa);
1995 }
1996 
1997 /*
1998  * Find an interface on a specific network.  If many, choice
1999  * is most specific found.
2000  */
2001 struct ifaddr *
2002 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2003 {
2004 	struct ifnet *ifp;
2005 	struct ifaddr *ifa;
2006 	struct ifaddr *ifa_maybe = NULL;
2007 	u_int af = addr->sa_family;
2008 	const char *addr_data = addr->sa_data, *cplim;
2009 
2010 	/*
2011 	 * AF_LINK addresses can be looked up directly by their index number,
2012 	 * so do that if we can.
2013 	 */
2014 	if (af == AF_LINK) {
2015 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2016 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2017 		return (ifaddr_byindex(sdl->sdl_index));
2018 	}
2019 
2020 	/*
2021 	 * Scan though each interface, looking for ones that have addresses
2022 	 * in this address family and the requested fib.  Maintain a reference
2023 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2024 	 * kept it stable when we move onto the next interface.
2025 	 */
2026 	IFNET_RLOCK_NOSLEEP();
2027 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2028 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2029 			continue;
2030 		IF_ADDR_RLOCK(ifp);
2031 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2032 			const char *cp, *cp2, *cp3;
2033 
2034 			if (ifa->ifa_addr->sa_family != af)
2035 next:				continue;
2036 			if (af == AF_INET &&
2037 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2038 				/*
2039 				 * This is a bit broken as it doesn't
2040 				 * take into account that the remote end may
2041 				 * be a single node in the network we are
2042 				 * looking for.
2043 				 * The trouble is that we don't know the
2044 				 * netmask for the remote end.
2045 				 */
2046 				if (ifa->ifa_dstaddr != NULL &&
2047 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2048 					ifa_ref(ifa);
2049 					IF_ADDR_RUNLOCK(ifp);
2050 					goto done;
2051 				}
2052 			} else {
2053 				/*
2054 				 * Scan all the bits in the ifa's address.
2055 				 * If a bit dissagrees with what we are
2056 				 * looking for, mask it with the netmask
2057 				 * to see if it really matters.
2058 				 * (A byte at a time)
2059 				 */
2060 				if (ifa->ifa_netmask == 0)
2061 					continue;
2062 				cp = addr_data;
2063 				cp2 = ifa->ifa_addr->sa_data;
2064 				cp3 = ifa->ifa_netmask->sa_data;
2065 				cplim = ifa->ifa_netmask->sa_len
2066 					+ (char *)ifa->ifa_netmask;
2067 				while (cp3 < cplim)
2068 					if ((*cp++ ^ *cp2++) & *cp3++)
2069 						goto next; /* next address! */
2070 				/*
2071 				 * If the netmask of what we just found
2072 				 * is more specific than what we had before
2073 				 * (if we had one), or if the virtual status
2074 				 * of new prefix is better than of the old one,
2075 				 * then remember the new one before continuing
2076 				 * to search for an even better one.
2077 				 */
2078 				if (ifa_maybe == NULL ||
2079 				    ifa_preferred(ifa_maybe, ifa) ||
2080 				    rn_refines((caddr_t)ifa->ifa_netmask,
2081 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2082 					if (ifa_maybe != NULL)
2083 						ifa_free(ifa_maybe);
2084 					ifa_maybe = ifa;
2085 					ifa_ref(ifa_maybe);
2086 				}
2087 			}
2088 		}
2089 		IF_ADDR_RUNLOCK(ifp);
2090 	}
2091 	ifa = ifa_maybe;
2092 	ifa_maybe = NULL;
2093 done:
2094 	IFNET_RUNLOCK_NOSLEEP();
2095 	if (ifa_maybe != NULL)
2096 		ifa_free(ifa_maybe);
2097 	return (ifa);
2098 }
2099 
2100 /*
2101  * Find an interface address specific to an interface best matching
2102  * a given address.
2103  */
2104 struct ifaddr *
2105 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2106 {
2107 	struct ifaddr *ifa;
2108 	const char *cp, *cp2, *cp3;
2109 	char *cplim;
2110 	struct ifaddr *ifa_maybe = NULL;
2111 	u_int af = addr->sa_family;
2112 
2113 	if (af >= AF_MAX)
2114 		return (NULL);
2115 	IF_ADDR_RLOCK(ifp);
2116 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2117 		if (ifa->ifa_addr->sa_family != af)
2118 			continue;
2119 		if (ifa_maybe == NULL)
2120 			ifa_maybe = ifa;
2121 		if (ifa->ifa_netmask == 0) {
2122 			if (sa_equal(addr, ifa->ifa_addr) ||
2123 			    (ifa->ifa_dstaddr &&
2124 			    sa_equal(addr, ifa->ifa_dstaddr)))
2125 				goto done;
2126 			continue;
2127 		}
2128 		if (ifp->if_flags & IFF_POINTOPOINT) {
2129 			if (sa_equal(addr, ifa->ifa_dstaddr))
2130 				goto done;
2131 		} else {
2132 			cp = addr->sa_data;
2133 			cp2 = ifa->ifa_addr->sa_data;
2134 			cp3 = ifa->ifa_netmask->sa_data;
2135 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2136 			for (; cp3 < cplim; cp3++)
2137 				if ((*cp++ ^ *cp2++) & *cp3)
2138 					break;
2139 			if (cp3 == cplim)
2140 				goto done;
2141 		}
2142 	}
2143 	ifa = ifa_maybe;
2144 done:
2145 	if (ifa != NULL)
2146 		ifa_ref(ifa);
2147 	IF_ADDR_RUNLOCK(ifp);
2148 	return (ifa);
2149 }
2150 
2151 /*
2152  * See whether new ifa is better than current one:
2153  * 1) A non-virtual one is preferred over virtual.
2154  * 2) A virtual in master state preferred over any other state.
2155  *
2156  * Used in several address selecting functions.
2157  */
2158 int
2159 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2160 {
2161 
2162 	return (cur->ifa_carp && (!next->ifa_carp ||
2163 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2164 }
2165 
2166 #include <net/if_llatbl.h>
2167 
2168 /*
2169  * Default action when installing a route with a Link Level gateway.
2170  * Lookup an appropriate real ifa to point to.
2171  * This should be moved to /sys/net/link.c eventually.
2172  */
2173 static void
2174 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2175 {
2176 	struct ifaddr *ifa, *oifa;
2177 	struct sockaddr *dst;
2178 	struct ifnet *ifp;
2179 
2180 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2181 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2182 		return;
2183 	ifa = ifaof_ifpforaddr(dst, ifp);
2184 	if (ifa) {
2185 		oifa = rt->rt_ifa;
2186 		rt->rt_ifa = ifa;
2187 		ifa_free(oifa);
2188 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2189 			ifa->ifa_rtrequest(cmd, rt, info);
2190 	}
2191 }
2192 
2193 struct sockaddr_dl *
2194 link_alloc_sdl(size_t size, int flags)
2195 {
2196 
2197 	return (malloc(size, M_TEMP, flags));
2198 }
2199 
2200 void
2201 link_free_sdl(struct sockaddr *sa)
2202 {
2203 	free(sa, M_TEMP);
2204 }
2205 
2206 /*
2207  * Fills in given sdl with interface basic info.
2208  * Returns pointer to filled sdl.
2209  */
2210 struct sockaddr_dl *
2211 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2212 {
2213 	struct sockaddr_dl *sdl;
2214 
2215 	sdl = (struct sockaddr_dl *)paddr;
2216 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2217 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2218 	sdl->sdl_family = AF_LINK;
2219 	sdl->sdl_index = ifp->if_index;
2220 	sdl->sdl_type = iftype;
2221 
2222 	return (sdl);
2223 }
2224 
2225 /*
2226  * Mark an interface down and notify protocols of
2227  * the transition.
2228  */
2229 static void
2230 if_unroute(struct ifnet *ifp, int flag, int fam)
2231 {
2232 	struct ifaddr *ifa;
2233 
2234 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2235 
2236 	ifp->if_flags &= ~flag;
2237 	getmicrotime(&ifp->if_lastchange);
2238 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2239 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2240 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2241 	ifp->if_qflush(ifp);
2242 
2243 	if (ifp->if_carp)
2244 		(*carp_linkstate_p)(ifp);
2245 	rt_ifmsg(ifp);
2246 }
2247 
2248 /*
2249  * Mark an interface up and notify protocols of
2250  * the transition.
2251  */
2252 static void
2253 if_route(struct ifnet *ifp, int flag, int fam)
2254 {
2255 	struct ifaddr *ifa;
2256 
2257 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2258 
2259 	ifp->if_flags |= flag;
2260 	getmicrotime(&ifp->if_lastchange);
2261 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2262 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2263 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2264 	if (ifp->if_carp)
2265 		(*carp_linkstate_p)(ifp);
2266 	rt_ifmsg(ifp);
2267 #ifdef INET6
2268 	in6_if_up(ifp);
2269 #endif
2270 }
2271 
2272 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2273 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2274 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2275 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2276 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2277 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2278 void	*(*vlan_cookie_p)(struct ifnet *);
2279 
2280 /*
2281  * Handle a change in the interface link state. To avoid LORs
2282  * between driver lock and upper layer locks, as well as possible
2283  * recursions, we post event to taskqueue, and all job
2284  * is done in static do_link_state_change().
2285  */
2286 void
2287 if_link_state_change(struct ifnet *ifp, int link_state)
2288 {
2289 	/* Return if state hasn't changed. */
2290 	if (ifp->if_link_state == link_state)
2291 		return;
2292 
2293 	ifp->if_link_state = link_state;
2294 
2295 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2296 }
2297 
2298 static void
2299 do_link_state_change(void *arg, int pending)
2300 {
2301 	struct ifnet *ifp = (struct ifnet *)arg;
2302 	int link_state = ifp->if_link_state;
2303 	CURVNET_SET(ifp->if_vnet);
2304 
2305 	/* Notify that the link state has changed. */
2306 	rt_ifmsg(ifp);
2307 	if (ifp->if_vlantrunk != NULL)
2308 		(*vlan_link_state_p)(ifp);
2309 
2310 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2311 	    ifp->if_l2com != NULL)
2312 		(*ng_ether_link_state_p)(ifp, link_state);
2313 	if (ifp->if_carp)
2314 		(*carp_linkstate_p)(ifp);
2315 	if (ifp->if_bridge)
2316 		(*bridge_linkstate_p)(ifp);
2317 	if (ifp->if_lagg)
2318 		(*lagg_linkstate_p)(ifp, link_state);
2319 
2320 	if (IS_DEFAULT_VNET(curvnet))
2321 		devctl_notify("IFNET", ifp->if_xname,
2322 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2323 		    NULL);
2324 	if (pending > 1)
2325 		if_printf(ifp, "%d link states coalesced\n", pending);
2326 	if (log_link_state_change)
2327 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2328 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2329 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2330 	CURVNET_RESTORE();
2331 }
2332 
2333 /*
2334  * Mark an interface down and notify protocols of
2335  * the transition.
2336  */
2337 void
2338 if_down(struct ifnet *ifp)
2339 {
2340 
2341 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2342 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2343 }
2344 
2345 /*
2346  * Mark an interface up and notify protocols of
2347  * the transition.
2348  */
2349 void
2350 if_up(struct ifnet *ifp)
2351 {
2352 
2353 	if_route(ifp, IFF_UP, AF_UNSPEC);
2354 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2355 }
2356 
2357 /*
2358  * Flush an interface queue.
2359  */
2360 void
2361 if_qflush(struct ifnet *ifp)
2362 {
2363 	struct mbuf *m, *n;
2364 	struct ifaltq *ifq;
2365 
2366 	ifq = &ifp->if_snd;
2367 	IFQ_LOCK(ifq);
2368 #ifdef ALTQ
2369 	if (ALTQ_IS_ENABLED(ifq))
2370 		ALTQ_PURGE(ifq);
2371 #endif
2372 	n = ifq->ifq_head;
2373 	while ((m = n) != NULL) {
2374 		n = m->m_nextpkt;
2375 		m_freem(m);
2376 	}
2377 	ifq->ifq_head = 0;
2378 	ifq->ifq_tail = 0;
2379 	ifq->ifq_len = 0;
2380 	IFQ_UNLOCK(ifq);
2381 }
2382 
2383 /*
2384  * Map interface name to interface structure pointer, with or without
2385  * returning a reference.
2386  */
2387 struct ifnet *
2388 ifunit_ref(const char *name)
2389 {
2390 	struct ifnet *ifp;
2391 
2392 	IFNET_RLOCK_NOSLEEP();
2393 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2394 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2395 		    !(ifp->if_flags & IFF_DYING))
2396 			break;
2397 	}
2398 	if (ifp != NULL)
2399 		if_ref(ifp);
2400 	IFNET_RUNLOCK_NOSLEEP();
2401 	return (ifp);
2402 }
2403 
2404 struct ifnet *
2405 ifunit(const char *name)
2406 {
2407 	struct ifnet *ifp;
2408 
2409 	IFNET_RLOCK_NOSLEEP();
2410 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2411 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2412 			break;
2413 	}
2414 	IFNET_RUNLOCK_NOSLEEP();
2415 	return (ifp);
2416 }
2417 
2418 static void *
2419 ifr_buffer_get_buffer(void *data)
2420 {
2421 	union ifreq_union *ifrup;
2422 
2423 	ifrup = data;
2424 #ifdef COMPAT_FREEBSD32
2425 	if (SV_CURPROC_FLAG(SV_ILP32))
2426 		return ((void *)(uintptr_t)
2427 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2428 #endif
2429 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2430 }
2431 
2432 static void
2433 ifr_buffer_set_buffer_null(void *data)
2434 {
2435 	union ifreq_union *ifrup;
2436 
2437 	ifrup = data;
2438 #ifdef COMPAT_FREEBSD32
2439 	if (SV_CURPROC_FLAG(SV_ILP32))
2440 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2441 	else
2442 #endif
2443 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2444 }
2445 
2446 static size_t
2447 ifr_buffer_get_length(void *data)
2448 {
2449 	union ifreq_union *ifrup;
2450 
2451 	ifrup = data;
2452 #ifdef COMPAT_FREEBSD32
2453 	if (SV_CURPROC_FLAG(SV_ILP32))
2454 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2455 #endif
2456 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2457 }
2458 
2459 static void
2460 ifr_buffer_set_length(void *data, size_t len)
2461 {
2462 	union ifreq_union *ifrup;
2463 
2464 	ifrup = data;
2465 #ifdef COMPAT_FREEBSD32
2466 	if (SV_CURPROC_FLAG(SV_ILP32))
2467 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2468 	else
2469 #endif
2470 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2471 }
2472 
2473 void *
2474 ifr_data_get_ptr(void *ifrp)
2475 {
2476 	union ifreq_union *ifrup;
2477 
2478 	ifrup = ifrp;
2479 #ifdef COMPAT_FREEBSD32
2480 	if (SV_CURPROC_FLAG(SV_ILP32))
2481 		return ((void *)(uintptr_t)
2482 		    ifrup->ifr32.ifr_ifru.ifru_data);
2483 #endif
2484 		return (ifrup->ifr.ifr_ifru.ifru_data);
2485 }
2486 
2487 /*
2488  * Hardware specific interface ioctls.
2489  */
2490 static int
2491 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2492 {
2493 	struct ifreq *ifr;
2494 	int error = 0, do_ifup = 0;
2495 	int new_flags, temp_flags;
2496 	size_t namelen, onamelen;
2497 	size_t descrlen;
2498 	char *descrbuf, *odescrbuf;
2499 	char new_name[IFNAMSIZ];
2500 	struct ifaddr *ifa;
2501 	struct sockaddr_dl *sdl;
2502 
2503 	ifr = (struct ifreq *)data;
2504 	switch (cmd) {
2505 	case SIOCGIFINDEX:
2506 		ifr->ifr_index = ifp->if_index;
2507 		break;
2508 
2509 	case SIOCGIFFLAGS:
2510 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2511 		ifr->ifr_flags = temp_flags & 0xffff;
2512 		ifr->ifr_flagshigh = temp_flags >> 16;
2513 		break;
2514 
2515 	case SIOCGIFCAP:
2516 		ifr->ifr_reqcap = ifp->if_capabilities;
2517 		ifr->ifr_curcap = ifp->if_capenable;
2518 		break;
2519 
2520 #ifdef MAC
2521 	case SIOCGIFMAC:
2522 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2523 		break;
2524 #endif
2525 
2526 	case SIOCGIFMETRIC:
2527 		ifr->ifr_metric = ifp->if_metric;
2528 		break;
2529 
2530 	case SIOCGIFMTU:
2531 		ifr->ifr_mtu = ifp->if_mtu;
2532 		break;
2533 
2534 	case SIOCGIFPHYS:
2535 		/* XXXGL: did this ever worked? */
2536 		ifr->ifr_phys = 0;
2537 		break;
2538 
2539 	case SIOCGIFDESCR:
2540 		error = 0;
2541 		sx_slock(&ifdescr_sx);
2542 		if (ifp->if_description == NULL)
2543 			error = ENOMSG;
2544 		else {
2545 			/* space for terminating nul */
2546 			descrlen = strlen(ifp->if_description) + 1;
2547 			if (ifr_buffer_get_length(ifr) < descrlen)
2548 				ifr_buffer_set_buffer_null(ifr);
2549 			else
2550 				error = copyout(ifp->if_description,
2551 				    ifr_buffer_get_buffer(ifr), descrlen);
2552 			ifr_buffer_set_length(ifr, descrlen);
2553 		}
2554 		sx_sunlock(&ifdescr_sx);
2555 		break;
2556 
2557 	case SIOCSIFDESCR:
2558 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2559 		if (error)
2560 			return (error);
2561 
2562 		/*
2563 		 * Copy only (length-1) bytes to make sure that
2564 		 * if_description is always nul terminated.  The
2565 		 * length parameter is supposed to count the
2566 		 * terminating nul in.
2567 		 */
2568 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2569 			return (ENAMETOOLONG);
2570 		else if (ifr_buffer_get_length(ifr) == 0)
2571 			descrbuf = NULL;
2572 		else {
2573 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2574 			    M_IFDESCR, M_WAITOK | M_ZERO);
2575 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2576 			    ifr_buffer_get_length(ifr) - 1);
2577 			if (error) {
2578 				free(descrbuf, M_IFDESCR);
2579 				break;
2580 			}
2581 		}
2582 
2583 		sx_xlock(&ifdescr_sx);
2584 		odescrbuf = ifp->if_description;
2585 		ifp->if_description = descrbuf;
2586 		sx_xunlock(&ifdescr_sx);
2587 
2588 		getmicrotime(&ifp->if_lastchange);
2589 		free(odescrbuf, M_IFDESCR);
2590 		break;
2591 
2592 	case SIOCGIFFIB:
2593 		ifr->ifr_fib = ifp->if_fib;
2594 		break;
2595 
2596 	case SIOCSIFFIB:
2597 		error = priv_check(td, PRIV_NET_SETIFFIB);
2598 		if (error)
2599 			return (error);
2600 		if (ifr->ifr_fib >= rt_numfibs)
2601 			return (EINVAL);
2602 
2603 		ifp->if_fib = ifr->ifr_fib;
2604 		break;
2605 
2606 	case SIOCSIFFLAGS:
2607 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2608 		if (error)
2609 			return (error);
2610 		/*
2611 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2612 		 * check, so we don't need special handling here yet.
2613 		 */
2614 		new_flags = (ifr->ifr_flags & 0xffff) |
2615 		    (ifr->ifr_flagshigh << 16);
2616 		if (ifp->if_flags & IFF_UP &&
2617 		    (new_flags & IFF_UP) == 0) {
2618 			if_down(ifp);
2619 		} else if (new_flags & IFF_UP &&
2620 		    (ifp->if_flags & IFF_UP) == 0) {
2621 			do_ifup = 1;
2622 		}
2623 		/* See if permanently promiscuous mode bit is about to flip */
2624 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2625 			if (new_flags & IFF_PPROMISC)
2626 				ifp->if_flags |= IFF_PROMISC;
2627 			else if (ifp->if_pcount == 0)
2628 				ifp->if_flags &= ~IFF_PROMISC;
2629 			if (log_promisc_mode_change)
2630                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2631                                     ifp->if_xname,
2632                                     ((new_flags & IFF_PPROMISC) ?
2633                                      "enabled" : "disabled"));
2634 		}
2635 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2636 			(new_flags &~ IFF_CANTCHANGE);
2637 		if (ifp->if_ioctl) {
2638 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2639 		}
2640 		if (do_ifup)
2641 			if_up(ifp);
2642 		getmicrotime(&ifp->if_lastchange);
2643 		break;
2644 
2645 	case SIOCSIFCAP:
2646 		error = priv_check(td, PRIV_NET_SETIFCAP);
2647 		if (error)
2648 			return (error);
2649 		if (ifp->if_ioctl == NULL)
2650 			return (EOPNOTSUPP);
2651 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2652 			return (EINVAL);
2653 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2654 		if (error == 0)
2655 			getmicrotime(&ifp->if_lastchange);
2656 		break;
2657 
2658 #ifdef MAC
2659 	case SIOCSIFMAC:
2660 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2661 		break;
2662 #endif
2663 
2664 	case SIOCSIFNAME:
2665 		error = priv_check(td, PRIV_NET_SETIFNAME);
2666 		if (error)
2667 			return (error);
2668 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2669 		    NULL);
2670 		if (error != 0)
2671 			return (error);
2672 		if (new_name[0] == '\0')
2673 			return (EINVAL);
2674 		if (new_name[IFNAMSIZ-1] != '\0') {
2675 			new_name[IFNAMSIZ-1] = '\0';
2676 			if (strlen(new_name) == IFNAMSIZ-1)
2677 				return (EINVAL);
2678 		}
2679 		if (ifunit(new_name) != NULL)
2680 			return (EEXIST);
2681 
2682 		/*
2683 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2684 		 * and there are numerous other races with the
2685 		 * ifunit() checks not being atomic with namespace
2686 		 * changes (renames, vmoves, if_attach, etc).
2687 		 */
2688 		ifp->if_flags |= IFF_RENAMING;
2689 
2690 		/* Announce the departure of the interface. */
2691 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2692 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2693 
2694 		log(LOG_INFO, "%s: changing name to '%s'\n",
2695 		    ifp->if_xname, new_name);
2696 
2697 		IF_ADDR_WLOCK(ifp);
2698 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2699 		ifa = ifp->if_addr;
2700 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2701 		namelen = strlen(new_name);
2702 		onamelen = sdl->sdl_nlen;
2703 		/*
2704 		 * Move the address if needed.  This is safe because we
2705 		 * allocate space for a name of length IFNAMSIZ when we
2706 		 * create this in if_attach().
2707 		 */
2708 		if (namelen != onamelen) {
2709 			bcopy(sdl->sdl_data + onamelen,
2710 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2711 		}
2712 		bcopy(new_name, sdl->sdl_data, namelen);
2713 		sdl->sdl_nlen = namelen;
2714 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2715 		bzero(sdl->sdl_data, onamelen);
2716 		while (namelen != 0)
2717 			sdl->sdl_data[--namelen] = 0xff;
2718 		IF_ADDR_WUNLOCK(ifp);
2719 
2720 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2721 		/* Announce the return of the interface. */
2722 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2723 
2724 		ifp->if_flags &= ~IFF_RENAMING;
2725 		break;
2726 
2727 #ifdef VIMAGE
2728 	case SIOCSIFVNET:
2729 		error = priv_check(td, PRIV_NET_SETIFVNET);
2730 		if (error)
2731 			return (error);
2732 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2733 		break;
2734 #endif
2735 
2736 	case SIOCSIFMETRIC:
2737 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2738 		if (error)
2739 			return (error);
2740 		ifp->if_metric = ifr->ifr_metric;
2741 		getmicrotime(&ifp->if_lastchange);
2742 		break;
2743 
2744 	case SIOCSIFPHYS:
2745 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2746 		if (error)
2747 			return (error);
2748 		if (ifp->if_ioctl == NULL)
2749 			return (EOPNOTSUPP);
2750 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2751 		if (error == 0)
2752 			getmicrotime(&ifp->if_lastchange);
2753 		break;
2754 
2755 	case SIOCSIFMTU:
2756 	{
2757 		u_long oldmtu = ifp->if_mtu;
2758 
2759 		error = priv_check(td, PRIV_NET_SETIFMTU);
2760 		if (error)
2761 			return (error);
2762 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2763 			return (EINVAL);
2764 		if (ifp->if_ioctl == NULL)
2765 			return (EOPNOTSUPP);
2766 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2767 		if (error == 0) {
2768 			getmicrotime(&ifp->if_lastchange);
2769 			rt_ifmsg(ifp);
2770 		}
2771 		/*
2772 		 * If the link MTU changed, do network layer specific procedure.
2773 		 */
2774 		if (ifp->if_mtu != oldmtu) {
2775 #ifdef INET6
2776 			nd6_setmtu(ifp);
2777 #endif
2778 			rt_updatemtu(ifp);
2779 		}
2780 		break;
2781 	}
2782 
2783 	case SIOCADDMULTI:
2784 	case SIOCDELMULTI:
2785 		if (cmd == SIOCADDMULTI)
2786 			error = priv_check(td, PRIV_NET_ADDMULTI);
2787 		else
2788 			error = priv_check(td, PRIV_NET_DELMULTI);
2789 		if (error)
2790 			return (error);
2791 
2792 		/* Don't allow group membership on non-multicast interfaces. */
2793 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2794 			return (EOPNOTSUPP);
2795 
2796 		/* Don't let users screw up protocols' entries. */
2797 		if (ifr->ifr_addr.sa_family != AF_LINK)
2798 			return (EINVAL);
2799 
2800 		if (cmd == SIOCADDMULTI) {
2801 			struct ifmultiaddr *ifma;
2802 
2803 			/*
2804 			 * Userland is only permitted to join groups once
2805 			 * via the if_addmulti() KPI, because it cannot hold
2806 			 * struct ifmultiaddr * between calls. It may also
2807 			 * lose a race while we check if the membership
2808 			 * already exists.
2809 			 */
2810 			IF_ADDR_RLOCK(ifp);
2811 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2812 			IF_ADDR_RUNLOCK(ifp);
2813 			if (ifma != NULL)
2814 				error = EADDRINUSE;
2815 			else
2816 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2817 		} else {
2818 			error = if_delmulti(ifp, &ifr->ifr_addr);
2819 		}
2820 		if (error == 0)
2821 			getmicrotime(&ifp->if_lastchange);
2822 		break;
2823 
2824 	case SIOCSIFPHYADDR:
2825 	case SIOCDIFPHYADDR:
2826 #ifdef INET6
2827 	case SIOCSIFPHYADDR_IN6:
2828 #endif
2829 	case SIOCSIFMEDIA:
2830 	case SIOCSIFGENERIC:
2831 		error = priv_check(td, PRIV_NET_HWIOCTL);
2832 		if (error)
2833 			return (error);
2834 		if (ifp->if_ioctl == NULL)
2835 			return (EOPNOTSUPP);
2836 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2837 		if (error == 0)
2838 			getmicrotime(&ifp->if_lastchange);
2839 		break;
2840 
2841 	case SIOCGIFSTATUS:
2842 	case SIOCGIFPSRCADDR:
2843 	case SIOCGIFPDSTADDR:
2844 	case SIOCGIFMEDIA:
2845 	case SIOCGIFXMEDIA:
2846 	case SIOCGIFGENERIC:
2847 	case SIOCGIFRSSKEY:
2848 	case SIOCGIFRSSHASH:
2849 		if (ifp->if_ioctl == NULL)
2850 			return (EOPNOTSUPP);
2851 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2852 		break;
2853 
2854 	case SIOCSIFLLADDR:
2855 		error = priv_check(td, PRIV_NET_SETLLADDR);
2856 		if (error)
2857 			return (error);
2858 		error = if_setlladdr(ifp,
2859 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2860 		break;
2861 
2862 	case SIOCGHWADDR:
2863 		error = if_gethwaddr(ifp, ifr);
2864 		break;
2865 
2866 	CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2867 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2868 		if (error)
2869 			return (error);
2870 		if ((error = if_addgroup(ifp,
2871 		    ifgr_group_get((struct ifgroupreq *)data))))
2872 			return (error);
2873 		break;
2874 
2875 	CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2876 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2877 			return (error);
2878 		break;
2879 
2880 	CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2881 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2882 		if (error)
2883 			return (error);
2884 		if ((error = if_delgroup(ifp,
2885 		    ifgr_group_get((struct ifgroupreq *)data))))
2886 			return (error);
2887 		break;
2888 
2889 	default:
2890 		error = ENOIOCTL;
2891 		break;
2892 	}
2893 	return (error);
2894 }
2895 
2896 #ifdef COMPAT_FREEBSD32
2897 struct ifconf32 {
2898 	int32_t	ifc_len;
2899 	union {
2900 		uint32_t	ifcu_buf;
2901 		uint32_t	ifcu_req;
2902 	} ifc_ifcu;
2903 };
2904 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2905 #endif
2906 
2907 #ifdef COMPAT_FREEBSD32
2908 static void
2909 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2910 {
2911 	struct ifmediareq32 *ifmr32;
2912 
2913 	ifmr32 = (struct ifmediareq32 *)data;
2914 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2915 	    sizeof(ifmr->ifm_name));
2916 	ifmr->ifm_current = ifmr32->ifm_current;
2917 	ifmr->ifm_mask = ifmr32->ifm_mask;
2918 	ifmr->ifm_status = ifmr32->ifm_status;
2919 	ifmr->ifm_active = ifmr32->ifm_active;
2920 	ifmr->ifm_count = ifmr32->ifm_count;
2921 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2922 }
2923 
2924 static void
2925 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2926 {
2927 	struct ifmediareq32 *ifmr32;
2928 
2929 	ifmr32 = (struct ifmediareq32 *)data;
2930 	ifmr32->ifm_current = ifmr->ifm_current;
2931 	ifmr32->ifm_mask = ifmr->ifm_mask;
2932 	ifmr32->ifm_status = ifmr->ifm_status;
2933 	ifmr32->ifm_active = ifmr->ifm_active;
2934 	ifmr32->ifm_count = ifmr->ifm_count;
2935 }
2936 #endif
2937 
2938 /*
2939  * Interface ioctls.
2940  */
2941 int
2942 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2943 {
2944 #ifdef COMPAT_FREEBSD32
2945 	caddr_t saved_data;
2946 	struct ifmediareq ifmr;
2947 #endif
2948 	struct ifmediareq *ifmrp;
2949 	struct ifnet *ifp;
2950 	struct ifreq *ifr;
2951 	int error;
2952 	int oif_flags;
2953 #ifdef VIMAGE
2954 	int shutdown;
2955 #endif
2956 
2957 	CURVNET_SET(so->so_vnet);
2958 #ifdef VIMAGE
2959 	/* Make sure the VNET is stable. */
2960 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2961 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2962 	if (shutdown) {
2963 		CURVNET_RESTORE();
2964 		return (EBUSY);
2965 	}
2966 #endif
2967 
2968 
2969 	switch (cmd) {
2970 	case SIOCGIFCONF:
2971 		error = ifconf(cmd, data);
2972 		CURVNET_RESTORE();
2973 		return (error);
2974 
2975 #ifdef COMPAT_FREEBSD32
2976 	case SIOCGIFCONF32:
2977 		{
2978 			struct ifconf32 *ifc32;
2979 			struct ifconf ifc;
2980 
2981 			ifc32 = (struct ifconf32 *)data;
2982 			ifc.ifc_len = ifc32->ifc_len;
2983 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2984 
2985 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2986 			CURVNET_RESTORE();
2987 			if (error == 0)
2988 				ifc32->ifc_len = ifc.ifc_len;
2989 			return (error);
2990 		}
2991 #endif
2992 	}
2993 
2994 	ifmrp = NULL;
2995 #ifdef COMPAT_FREEBSD32
2996 	switch (cmd) {
2997 	case SIOCGIFMEDIA32:
2998 	case SIOCGIFXMEDIA32:
2999 		ifmrp = &ifmr;
3000 		ifmr_init(ifmrp, data);
3001 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3002 		saved_data = data;
3003 		data = (caddr_t)ifmrp;
3004 	}
3005 #endif
3006 
3007 	ifr = (struct ifreq *)data;
3008 	switch (cmd) {
3009 #ifdef VIMAGE
3010 	case SIOCSIFRVNET:
3011 		error = priv_check(td, PRIV_NET_SETIFVNET);
3012 		if (error == 0)
3013 			error = if_vmove_reclaim(td, ifr->ifr_name,
3014 			    ifr->ifr_jid);
3015 		goto out_noref;
3016 #endif
3017 	case SIOCIFCREATE:
3018 	case SIOCIFCREATE2:
3019 		error = priv_check(td, PRIV_NET_IFCREATE);
3020 		if (error == 0)
3021 			error = if_clone_create(ifr->ifr_name,
3022 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3023 			    ifr_data_get_ptr(ifr) : NULL);
3024 		goto out_noref;
3025 	case SIOCIFDESTROY:
3026 		error = priv_check(td, PRIV_NET_IFDESTROY);
3027 		if (error == 0)
3028 			error = if_clone_destroy(ifr->ifr_name);
3029 		goto out_noref;
3030 
3031 	case SIOCIFGCLONERS:
3032 		error = if_clone_list((struct if_clonereq *)data);
3033 		goto out_noref;
3034 
3035 	CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3036 		error = if_getgroupmembers((struct ifgroupreq *)data);
3037 		goto out_noref;
3038 
3039 #if defined(INET) || defined(INET6)
3040 	case SIOCSVH:
3041 	case SIOCGVH:
3042 		if (carp_ioctl_p == NULL)
3043 			error = EPROTONOSUPPORT;
3044 		else
3045 			error = (*carp_ioctl_p)(ifr, cmd, td);
3046 		goto out_noref;
3047 #endif
3048 	}
3049 
3050 	ifp = ifunit_ref(ifr->ifr_name);
3051 	if (ifp == NULL) {
3052 		error = ENXIO;
3053 		goto out_noref;
3054 	}
3055 
3056 	error = ifhwioctl(cmd, ifp, data, td);
3057 	if (error != ENOIOCTL)
3058 		goto out_ref;
3059 
3060 	oif_flags = ifp->if_flags;
3061 	if (so->so_proto == NULL) {
3062 		error = EOPNOTSUPP;
3063 		goto out_ref;
3064 	}
3065 
3066 	/*
3067 	 * Pass the request on to the socket control method, and if the
3068 	 * latter returns EOPNOTSUPP, directly to the interface.
3069 	 *
3070 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3071 	 * trust SIOCSIFADDR et al to come from an already privileged
3072 	 * layer, and do not perform any credentials checks or input
3073 	 * validation.
3074 	 */
3075 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3076 	    ifp, td));
3077 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3078 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3079 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3080 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3081 
3082 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3083 #ifdef INET6
3084 		if (ifp->if_flags & IFF_UP)
3085 			in6_if_up(ifp);
3086 #endif
3087 	}
3088 
3089 out_ref:
3090 	if_rele(ifp);
3091 out_noref:
3092 #ifdef COMPAT_FREEBSD32
3093 	if (ifmrp != NULL) {
3094 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3095 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3096 		     cmd));
3097 		data = saved_data;
3098 		ifmr_update(ifmrp, data);
3099 	}
3100 #endif
3101 	CURVNET_RESTORE();
3102 	return (error);
3103 }
3104 
3105 /*
3106  * The code common to handling reference counted flags,
3107  * e.g., in ifpromisc() and if_allmulti().
3108  * The "pflag" argument can specify a permanent mode flag to check,
3109  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3110  *
3111  * Only to be used on stack-owned flags, not driver-owned flags.
3112  */
3113 static int
3114 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3115 {
3116 	struct ifreq ifr;
3117 	int error;
3118 	int oldflags, oldcount;
3119 
3120 	/* Sanity checks to catch programming errors */
3121 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3122 	    ("%s: setting driver-owned flag %d", __func__, flag));
3123 
3124 	if (onswitch)
3125 		KASSERT(*refcount >= 0,
3126 		    ("%s: increment negative refcount %d for flag %d",
3127 		    __func__, *refcount, flag));
3128 	else
3129 		KASSERT(*refcount > 0,
3130 		    ("%s: decrement non-positive refcount %d for flag %d",
3131 		    __func__, *refcount, flag));
3132 
3133 	/* In case this mode is permanent, just touch refcount */
3134 	if (ifp->if_flags & pflag) {
3135 		*refcount += onswitch ? 1 : -1;
3136 		return (0);
3137 	}
3138 
3139 	/* Save ifnet parameters for if_ioctl() may fail */
3140 	oldcount = *refcount;
3141 	oldflags = ifp->if_flags;
3142 
3143 	/*
3144 	 * See if we aren't the only and touching refcount is enough.
3145 	 * Actually toggle interface flag if we are the first or last.
3146 	 */
3147 	if (onswitch) {
3148 		if ((*refcount)++)
3149 			return (0);
3150 		ifp->if_flags |= flag;
3151 	} else {
3152 		if (--(*refcount))
3153 			return (0);
3154 		ifp->if_flags &= ~flag;
3155 	}
3156 
3157 	/* Call down the driver since we've changed interface flags */
3158 	if (ifp->if_ioctl == NULL) {
3159 		error = EOPNOTSUPP;
3160 		goto recover;
3161 	}
3162 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3163 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3164 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3165 	if (error)
3166 		goto recover;
3167 	/* Notify userland that interface flags have changed */
3168 	rt_ifmsg(ifp);
3169 	return (0);
3170 
3171 recover:
3172 	/* Recover after driver error */
3173 	*refcount = oldcount;
3174 	ifp->if_flags = oldflags;
3175 	return (error);
3176 }
3177 
3178 /*
3179  * Set/clear promiscuous mode on interface ifp based on the truth value
3180  * of pswitch.  The calls are reference counted so that only the first
3181  * "on" request actually has an effect, as does the final "off" request.
3182  * Results are undefined if the "off" and "on" requests are not matched.
3183  */
3184 int
3185 ifpromisc(struct ifnet *ifp, int pswitch)
3186 {
3187 	int error;
3188 	int oldflags = ifp->if_flags;
3189 
3190 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3191 			   &ifp->if_pcount, pswitch);
3192 	/* If promiscuous mode status has changed, log a message */
3193 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3194             log_promisc_mode_change)
3195 		log(LOG_INFO, "%s: promiscuous mode %s\n",
3196 		    ifp->if_xname,
3197 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3198 	return (error);
3199 }
3200 
3201 /*
3202  * Return interface configuration
3203  * of system.  List may be used
3204  * in later ioctl's (above) to get
3205  * other information.
3206  */
3207 /*ARGSUSED*/
3208 static int
3209 ifconf(u_long cmd, caddr_t data)
3210 {
3211 	struct ifconf *ifc = (struct ifconf *)data;
3212 	struct ifnet *ifp;
3213 	struct ifaddr *ifa;
3214 	struct ifreq ifr;
3215 	struct sbuf *sb;
3216 	int error, full = 0, valid_len, max_len;
3217 
3218 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3219 	max_len = MAXPHYS - 1;
3220 
3221 	/* Prevent hostile input from being able to crash the system */
3222 	if (ifc->ifc_len <= 0)
3223 		return (EINVAL);
3224 
3225 again:
3226 	if (ifc->ifc_len <= max_len) {
3227 		max_len = ifc->ifc_len;
3228 		full = 1;
3229 	}
3230 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3231 	max_len = 0;
3232 	valid_len = 0;
3233 
3234 	IFNET_RLOCK();
3235 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3236 		int addrs;
3237 
3238 		/*
3239 		 * Zero the ifr to make sure we don't disclose the contents
3240 		 * of the stack.
3241 		 */
3242 		memset(&ifr, 0, sizeof(ifr));
3243 
3244 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3245 		    >= sizeof(ifr.ifr_name)) {
3246 			sbuf_delete(sb);
3247 			IFNET_RUNLOCK();
3248 			return (ENAMETOOLONG);
3249 		}
3250 
3251 		addrs = 0;
3252 		IF_ADDR_RLOCK(ifp);
3253 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3254 			struct sockaddr *sa = ifa->ifa_addr;
3255 
3256 			if (prison_if(curthread->td_ucred, sa) != 0)
3257 				continue;
3258 			addrs++;
3259 			if (sa->sa_len <= sizeof(*sa)) {
3260 				if (sa->sa_len < sizeof(*sa)) {
3261 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3262 					    sizeof(ifr.ifr_ifru.ifru_addr));
3263 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3264 					    sa->sa_len);
3265 				} else
3266 					ifr.ifr_ifru.ifru_addr = *sa;
3267 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3268 				max_len += sizeof(ifr);
3269 			} else {
3270 				sbuf_bcat(sb, &ifr,
3271 				    offsetof(struct ifreq, ifr_addr));
3272 				max_len += offsetof(struct ifreq, ifr_addr);
3273 				sbuf_bcat(sb, sa, sa->sa_len);
3274 				max_len += sa->sa_len;
3275 			}
3276 
3277 			if (sbuf_error(sb) == 0)
3278 				valid_len = sbuf_len(sb);
3279 		}
3280 		IF_ADDR_RUNLOCK(ifp);
3281 		if (addrs == 0) {
3282 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3283 			max_len += sizeof(ifr);
3284 
3285 			if (sbuf_error(sb) == 0)
3286 				valid_len = sbuf_len(sb);
3287 		}
3288 	}
3289 	IFNET_RUNLOCK();
3290 
3291 	/*
3292 	 * If we didn't allocate enough space (uncommon), try again.  If
3293 	 * we have already allocated as much space as we are allowed,
3294 	 * return what we've got.
3295 	 */
3296 	if (valid_len != max_len && !full) {
3297 		sbuf_delete(sb);
3298 		goto again;
3299 	}
3300 
3301 	ifc->ifc_len = valid_len;
3302 	sbuf_finish(sb);
3303 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3304 	sbuf_delete(sb);
3305 	return (error);
3306 }
3307 
3308 /*
3309  * Just like ifpromisc(), but for all-multicast-reception mode.
3310  */
3311 int
3312 if_allmulti(struct ifnet *ifp, int onswitch)
3313 {
3314 
3315 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3316 }
3317 
3318 struct ifmultiaddr *
3319 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3320 {
3321 	struct ifmultiaddr *ifma;
3322 
3323 	IF_ADDR_LOCK_ASSERT(ifp);
3324 
3325 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3326 		if (sa->sa_family == AF_LINK) {
3327 			if (sa_dl_equal(ifma->ifma_addr, sa))
3328 				break;
3329 		} else {
3330 			if (sa_equal(ifma->ifma_addr, sa))
3331 				break;
3332 		}
3333 	}
3334 
3335 	return ifma;
3336 }
3337 
3338 /*
3339  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3340  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3341  * the ifnet multicast address list here, so the caller must do that and
3342  * other setup work (such as notifying the device driver).  The reference
3343  * count is initialized to 1.
3344  */
3345 static struct ifmultiaddr *
3346 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3347     int mflags)
3348 {
3349 	struct ifmultiaddr *ifma;
3350 	struct sockaddr *dupsa;
3351 
3352 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3353 	    M_ZERO);
3354 	if (ifma == NULL)
3355 		return (NULL);
3356 
3357 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3358 	if (dupsa == NULL) {
3359 		free(ifma, M_IFMADDR);
3360 		return (NULL);
3361 	}
3362 	bcopy(sa, dupsa, sa->sa_len);
3363 	ifma->ifma_addr = dupsa;
3364 
3365 	ifma->ifma_ifp = ifp;
3366 	ifma->ifma_refcount = 1;
3367 	ifma->ifma_protospec = NULL;
3368 
3369 	if (llsa == NULL) {
3370 		ifma->ifma_lladdr = NULL;
3371 		return (ifma);
3372 	}
3373 
3374 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3375 	if (dupsa == NULL) {
3376 		free(ifma->ifma_addr, M_IFMADDR);
3377 		free(ifma, M_IFMADDR);
3378 		return (NULL);
3379 	}
3380 	bcopy(llsa, dupsa, llsa->sa_len);
3381 	ifma->ifma_lladdr = dupsa;
3382 
3383 	return (ifma);
3384 }
3385 
3386 /*
3387  * if_freemulti: free ifmultiaddr structure and possibly attached related
3388  * addresses.  The caller is responsible for implementing reference
3389  * counting, notifying the driver, handling routing messages, and releasing
3390  * any dependent link layer state.
3391  */
3392 static void
3393 if_freemulti(struct ifmultiaddr *ifma)
3394 {
3395 
3396 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3397 	    ifma->ifma_refcount));
3398 
3399 	if (ifma->ifma_lladdr != NULL)
3400 		free(ifma->ifma_lladdr, M_IFMADDR);
3401 	free(ifma->ifma_addr, M_IFMADDR);
3402 	free(ifma, M_IFMADDR);
3403 }
3404 
3405 /*
3406  * Register an additional multicast address with a network interface.
3407  *
3408  * - If the address is already present, bump the reference count on the
3409  *   address and return.
3410  * - If the address is not link-layer, look up a link layer address.
3411  * - Allocate address structures for one or both addresses, and attach to the
3412  *   multicast address list on the interface.  If automatically adding a link
3413  *   layer address, the protocol address will own a reference to the link
3414  *   layer address, to be freed when it is freed.
3415  * - Notify the network device driver of an addition to the multicast address
3416  *   list.
3417  *
3418  * 'sa' points to caller-owned memory with the desired multicast address.
3419  *
3420  * 'retifma' will be used to return a pointer to the resulting multicast
3421  * address reference, if desired.
3422  */
3423 int
3424 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3425     struct ifmultiaddr **retifma)
3426 {
3427 	struct ifmultiaddr *ifma, *ll_ifma;
3428 	struct sockaddr *llsa;
3429 	struct sockaddr_dl sdl;
3430 	int error;
3431 
3432 	/*
3433 	 * If the address is already present, return a new reference to it;
3434 	 * otherwise, allocate storage and set up a new address.
3435 	 */
3436 	IF_ADDR_WLOCK(ifp);
3437 	ifma = if_findmulti(ifp, sa);
3438 	if (ifma != NULL) {
3439 		ifma->ifma_refcount++;
3440 		if (retifma != NULL)
3441 			*retifma = ifma;
3442 		IF_ADDR_WUNLOCK(ifp);
3443 		return (0);
3444 	}
3445 
3446 	/*
3447 	 * The address isn't already present; resolve the protocol address
3448 	 * into a link layer address, and then look that up, bump its
3449 	 * refcount or allocate an ifma for that also.
3450 	 * Most link layer resolving functions returns address data which
3451 	 * fits inside default sockaddr_dl structure. However callback
3452 	 * can allocate another sockaddr structure, in that case we need to
3453 	 * free it later.
3454 	 */
3455 	llsa = NULL;
3456 	ll_ifma = NULL;
3457 	if (ifp->if_resolvemulti != NULL) {
3458 		/* Provide called function with buffer size information */
3459 		sdl.sdl_len = sizeof(sdl);
3460 		llsa = (struct sockaddr *)&sdl;
3461 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3462 		if (error)
3463 			goto unlock_out;
3464 	}
3465 
3466 	/*
3467 	 * Allocate the new address.  Don't hook it up yet, as we may also
3468 	 * need to allocate a link layer multicast address.
3469 	 */
3470 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3471 	if (ifma == NULL) {
3472 		error = ENOMEM;
3473 		goto free_llsa_out;
3474 	}
3475 
3476 	/*
3477 	 * If a link layer address is found, we'll need to see if it's
3478 	 * already present in the address list, or allocate is as well.
3479 	 * When this block finishes, the link layer address will be on the
3480 	 * list.
3481 	 */
3482 	if (llsa != NULL) {
3483 		ll_ifma = if_findmulti(ifp, llsa);
3484 		if (ll_ifma == NULL) {
3485 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3486 			if (ll_ifma == NULL) {
3487 				--ifma->ifma_refcount;
3488 				if_freemulti(ifma);
3489 				error = ENOMEM;
3490 				goto free_llsa_out;
3491 			}
3492 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3493 			    ifma_link);
3494 		} else
3495 			ll_ifma->ifma_refcount++;
3496 		ifma->ifma_llifma = ll_ifma;
3497 	}
3498 
3499 	/*
3500 	 * We now have a new multicast address, ifma, and possibly a new or
3501 	 * referenced link layer address.  Add the primary address to the
3502 	 * ifnet address list.
3503 	 */
3504 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3505 
3506 	if (retifma != NULL)
3507 		*retifma = ifma;
3508 
3509 	/*
3510 	 * Must generate the message while holding the lock so that 'ifma'
3511 	 * pointer is still valid.
3512 	 */
3513 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3514 	IF_ADDR_WUNLOCK(ifp);
3515 
3516 	/*
3517 	 * We are certain we have added something, so call down to the
3518 	 * interface to let them know about it.
3519 	 */
3520 	if (ifp->if_ioctl != NULL) {
3521 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3522 	}
3523 
3524 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3525 		link_free_sdl(llsa);
3526 
3527 	return (0);
3528 
3529 free_llsa_out:
3530 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3531 		link_free_sdl(llsa);
3532 
3533 unlock_out:
3534 	IF_ADDR_WUNLOCK(ifp);
3535 	return (error);
3536 }
3537 
3538 /*
3539  * Delete a multicast group membership by network-layer group address.
3540  *
3541  * Returns ENOENT if the entry could not be found. If ifp no longer
3542  * exists, results are undefined. This entry point should only be used
3543  * from subsystems which do appropriate locking to hold ifp for the
3544  * duration of the call.
3545  * Network-layer protocol domains must use if_delmulti_ifma().
3546  */
3547 int
3548 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3549 {
3550 	struct ifmultiaddr *ifma;
3551 	int lastref;
3552 #ifdef INVARIANTS
3553 	struct ifnet *oifp;
3554 
3555 	IFNET_RLOCK_NOSLEEP();
3556 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3557 		if (ifp == oifp)
3558 			break;
3559 	if (ifp != oifp)
3560 		ifp = NULL;
3561 	IFNET_RUNLOCK_NOSLEEP();
3562 
3563 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3564 #endif
3565 	if (ifp == NULL)
3566 		return (ENOENT);
3567 
3568 	IF_ADDR_WLOCK(ifp);
3569 	lastref = 0;
3570 	ifma = if_findmulti(ifp, sa);
3571 	if (ifma != NULL)
3572 		lastref = if_delmulti_locked(ifp, ifma, 0);
3573 	IF_ADDR_WUNLOCK(ifp);
3574 
3575 	if (ifma == NULL)
3576 		return (ENOENT);
3577 
3578 	if (lastref && ifp->if_ioctl != NULL) {
3579 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3580 	}
3581 
3582 	return (0);
3583 }
3584 
3585 /*
3586  * Delete all multicast group membership for an interface.
3587  * Should be used to quickly flush all multicast filters.
3588  */
3589 void
3590 if_delallmulti(struct ifnet *ifp)
3591 {
3592 	struct ifmultiaddr *ifma;
3593 	struct ifmultiaddr *next;
3594 
3595 	IF_ADDR_WLOCK(ifp);
3596 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3597 		if_delmulti_locked(ifp, ifma, 0);
3598 	IF_ADDR_WUNLOCK(ifp);
3599 }
3600 
3601 /*
3602  * Delete a multicast group membership by group membership pointer.
3603  * Network-layer protocol domains must use this routine.
3604  *
3605  * It is safe to call this routine if the ifp disappeared.
3606  */
3607 void
3608 if_delmulti_ifma(struct ifmultiaddr *ifma)
3609 {
3610 	struct ifnet *ifp;
3611 	int lastref;
3612 
3613 	ifp = ifma->ifma_ifp;
3614 #ifdef DIAGNOSTIC
3615 	if (ifp == NULL) {
3616 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3617 	} else {
3618 		struct ifnet *oifp;
3619 
3620 		IFNET_RLOCK_NOSLEEP();
3621 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3622 			if (ifp == oifp)
3623 				break;
3624 		if (ifp != oifp) {
3625 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3626 			ifp = NULL;
3627 		}
3628 		IFNET_RUNLOCK_NOSLEEP();
3629 	}
3630 #endif
3631 	/*
3632 	 * If and only if the ifnet instance exists: Acquire the address lock.
3633 	 */
3634 	if (ifp != NULL)
3635 		IF_ADDR_WLOCK(ifp);
3636 
3637 	lastref = if_delmulti_locked(ifp, ifma, 0);
3638 
3639 	if (ifp != NULL) {
3640 		/*
3641 		 * If and only if the ifnet instance exists:
3642 		 *  Release the address lock.
3643 		 *  If the group was left: update the hardware hash filter.
3644 		 */
3645 		IF_ADDR_WUNLOCK(ifp);
3646 		if (lastref && ifp->if_ioctl != NULL) {
3647 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3648 		}
3649 	}
3650 }
3651 
3652 /*
3653  * Perform deletion of network-layer and/or link-layer multicast address.
3654  *
3655  * Return 0 if the reference count was decremented.
3656  * Return 1 if the final reference was released, indicating that the
3657  * hardware hash filter should be reprogrammed.
3658  */
3659 static int
3660 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3661 {
3662 	struct ifmultiaddr *ll_ifma;
3663 
3664 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3665 		KASSERT(ifma->ifma_ifp == ifp,
3666 		    ("%s: inconsistent ifp %p", __func__, ifp));
3667 		IF_ADDR_WLOCK_ASSERT(ifp);
3668 	}
3669 
3670 	ifp = ifma->ifma_ifp;
3671 
3672 	/*
3673 	 * If the ifnet is detaching, null out references to ifnet,
3674 	 * so that upper protocol layers will notice, and not attempt
3675 	 * to obtain locks for an ifnet which no longer exists. The
3676 	 * routing socket announcement must happen before the ifnet
3677 	 * instance is detached from the system.
3678 	 */
3679 	if (detaching) {
3680 #ifdef DIAGNOSTIC
3681 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3682 #endif
3683 		/*
3684 		 * ifp may already be nulled out if we are being reentered
3685 		 * to delete the ll_ifma.
3686 		 */
3687 		if (ifp != NULL) {
3688 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3689 			ifma->ifma_ifp = NULL;
3690 		}
3691 	}
3692 
3693 	if (--ifma->ifma_refcount > 0)
3694 		return 0;
3695 
3696 	/*
3697 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3698 	 * have been associated with it. Release it first if so.
3699 	 */
3700 	ll_ifma = ifma->ifma_llifma;
3701 	if (ll_ifma != NULL) {
3702 		KASSERT(ifma->ifma_lladdr != NULL,
3703 		    ("%s: llifma w/o lladdr", __func__));
3704 		if (detaching)
3705 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3706 		if (--ll_ifma->ifma_refcount == 0) {
3707 			if (ifp != NULL) {
3708 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3709 				    ifma_link);
3710 			}
3711 			if_freemulti(ll_ifma);
3712 		}
3713 	}
3714 
3715 	if (ifp != NULL)
3716 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3717 
3718 	if_freemulti(ifma);
3719 
3720 	/*
3721 	 * The last reference to this instance of struct ifmultiaddr
3722 	 * was released; the hardware should be notified of this change.
3723 	 */
3724 	return 1;
3725 }
3726 
3727 /*
3728  * Set the link layer address on an interface.
3729  *
3730  * At this time we only support certain types of interfaces,
3731  * and we don't allow the length of the address to change.
3732  *
3733  * Set noinline to be dtrace-friendly
3734  */
3735 __noinline int
3736 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3737 {
3738 	struct sockaddr_dl *sdl;
3739 	struct ifaddr *ifa;
3740 	struct ifreq ifr;
3741 
3742 	IF_ADDR_RLOCK(ifp);
3743 	ifa = ifp->if_addr;
3744 	if (ifa == NULL) {
3745 		IF_ADDR_RUNLOCK(ifp);
3746 		return (EINVAL);
3747 	}
3748 	ifa_ref(ifa);
3749 	IF_ADDR_RUNLOCK(ifp);
3750 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3751 	if (sdl == NULL) {
3752 		ifa_free(ifa);
3753 		return (EINVAL);
3754 	}
3755 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3756 		ifa_free(ifa);
3757 		return (EINVAL);
3758 	}
3759 	switch (ifp->if_type) {
3760 	case IFT_ETHER:
3761 	case IFT_XETHER:
3762 	case IFT_L2VLAN:
3763 	case IFT_BRIDGE:
3764 	case IFT_IEEE8023ADLAG:
3765 		bcopy(lladdr, LLADDR(sdl), len);
3766 		ifa_free(ifa);
3767 		break;
3768 	default:
3769 		ifa_free(ifa);
3770 		return (ENODEV);
3771 	}
3772 
3773 	/*
3774 	 * If the interface is already up, we need
3775 	 * to re-init it in order to reprogram its
3776 	 * address filter.
3777 	 */
3778 	if ((ifp->if_flags & IFF_UP) != 0) {
3779 		if (ifp->if_ioctl) {
3780 			ifp->if_flags &= ~IFF_UP;
3781 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3782 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3783 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3784 			ifp->if_flags |= IFF_UP;
3785 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3786 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3787 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3788 		}
3789 	}
3790 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3791 	return (0);
3792 }
3793 
3794 /*
3795  * Compat function for handling basic encapsulation requests.
3796  * Not converted stacks (FDDI, IB, ..) supports traditional
3797  * output model: ARP (and other similar L2 protocols) are handled
3798  * inside output routine, arpresolve/nd6_resolve() returns MAC
3799  * address instead of full prepend.
3800  *
3801  * This function creates calculated header==MAC for IPv4/IPv6 and
3802  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3803  * address families.
3804  */
3805 static int
3806 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3807 {
3808 
3809 	if (req->rtype != IFENCAP_LL)
3810 		return (EOPNOTSUPP);
3811 
3812 	if (req->bufsize < req->lladdr_len)
3813 		return (ENOMEM);
3814 
3815 	switch (req->family) {
3816 	case AF_INET:
3817 	case AF_INET6:
3818 		break;
3819 	default:
3820 		return (EAFNOSUPPORT);
3821 	}
3822 
3823 	/* Copy lladdr to storage as is */
3824 	memmove(req->buf, req->lladdr, req->lladdr_len);
3825 	req->bufsize = req->lladdr_len;
3826 	req->lladdr_off = 0;
3827 
3828 	return (0);
3829 }
3830 
3831 /*
3832  * Get the link layer address that was read from the hardware at attach.
3833  *
3834  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3835  * their component interfaces as IFT_IEEE8023ADLAG.
3836  */
3837 int
3838 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3839 {
3840 
3841 	if (ifp->if_hw_addr == NULL)
3842 		return (ENODEV);
3843 
3844 	switch (ifp->if_type) {
3845 	case IFT_ETHER:
3846 	case IFT_IEEE8023ADLAG:
3847 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3848 		return (0);
3849 	default:
3850 		return (ENODEV);
3851 	}
3852 }
3853 
3854 /*
3855  * The name argument must be a pointer to storage which will last as
3856  * long as the interface does.  For physical devices, the result of
3857  * device_get_name(dev) is a good choice and for pseudo-devices a
3858  * static string works well.
3859  */
3860 void
3861 if_initname(struct ifnet *ifp, const char *name, int unit)
3862 {
3863 	ifp->if_dname = name;
3864 	ifp->if_dunit = unit;
3865 	if (unit != IF_DUNIT_NONE)
3866 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3867 	else
3868 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3869 }
3870 
3871 int
3872 if_printf(struct ifnet *ifp, const char * fmt, ...)
3873 {
3874 	va_list ap;
3875 	int retval;
3876 
3877 	retval = printf("%s: ", ifp->if_xname);
3878 	va_start(ap, fmt);
3879 	retval += vprintf(fmt, ap);
3880 	va_end(ap);
3881 	return (retval);
3882 }
3883 
3884 void
3885 if_start(struct ifnet *ifp)
3886 {
3887 
3888 	(*(ifp)->if_start)(ifp);
3889 }
3890 
3891 /*
3892  * Backwards compatibility interface for drivers
3893  * that have not implemented it
3894  */
3895 static int
3896 if_transmit(struct ifnet *ifp, struct mbuf *m)
3897 {
3898 	int error;
3899 
3900 	IFQ_HANDOFF(ifp, m, error);
3901 	return (error);
3902 }
3903 
3904 static void
3905 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3906 {
3907 
3908 	m_freem(m);
3909 }
3910 
3911 int
3912 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3913 {
3914 	int active = 0;
3915 
3916 	IF_LOCK(ifq);
3917 	if (_IF_QFULL(ifq)) {
3918 		IF_UNLOCK(ifq);
3919 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3920 		m_freem(m);
3921 		return (0);
3922 	}
3923 	if (ifp != NULL) {
3924 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3925 		if (m->m_flags & (M_BCAST|M_MCAST))
3926 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3927 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3928 	}
3929 	_IF_ENQUEUE(ifq, m);
3930 	IF_UNLOCK(ifq);
3931 	if (ifp != NULL && !active)
3932 		(*(ifp)->if_start)(ifp);
3933 	return (1);
3934 }
3935 
3936 void
3937 if_register_com_alloc(u_char type,
3938     if_com_alloc_t *a, if_com_free_t *f)
3939 {
3940 
3941 	KASSERT(if_com_alloc[type] == NULL,
3942 	    ("if_register_com_alloc: %d already registered", type));
3943 	KASSERT(if_com_free[type] == NULL,
3944 	    ("if_register_com_alloc: %d free already registered", type));
3945 
3946 	if_com_alloc[type] = a;
3947 	if_com_free[type] = f;
3948 }
3949 
3950 void
3951 if_deregister_com_alloc(u_char type)
3952 {
3953 
3954 	KASSERT(if_com_alloc[type] != NULL,
3955 	    ("if_deregister_com_alloc: %d not registered", type));
3956 	KASSERT(if_com_free[type] != NULL,
3957 	    ("if_deregister_com_alloc: %d free not registered", type));
3958 	if_com_alloc[type] = NULL;
3959 	if_com_free[type] = NULL;
3960 }
3961 
3962 /* API for driver access to network stack owned ifnet.*/
3963 uint64_t
3964 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3965 {
3966 	uint64_t oldbrate;
3967 
3968 	oldbrate = ifp->if_baudrate;
3969 	ifp->if_baudrate = baudrate;
3970 	return (oldbrate);
3971 }
3972 
3973 uint64_t
3974 if_getbaudrate(if_t ifp)
3975 {
3976 
3977 	return (((struct ifnet *)ifp)->if_baudrate);
3978 }
3979 
3980 int
3981 if_setcapabilities(if_t ifp, int capabilities)
3982 {
3983 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3984 	return (0);
3985 }
3986 
3987 int
3988 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3989 {
3990 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3991 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3992 
3993 	return (0);
3994 }
3995 
3996 int
3997 if_getcapabilities(if_t ifp)
3998 {
3999 	return ((struct ifnet *)ifp)->if_capabilities;
4000 }
4001 
4002 int
4003 if_setcapenable(if_t ifp, int capabilities)
4004 {
4005 	((struct ifnet *)ifp)->if_capenable = capabilities;
4006 	return (0);
4007 }
4008 
4009 int
4010 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4011 {
4012 	if(setcap)
4013 		((struct ifnet *)ifp)->if_capenable |= setcap;
4014 	if(clearcap)
4015 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4016 
4017 	return (0);
4018 }
4019 
4020 const char *
4021 if_getdname(if_t ifp)
4022 {
4023 	return ((struct ifnet *)ifp)->if_dname;
4024 }
4025 
4026 int
4027 if_togglecapenable(if_t ifp, int togglecap)
4028 {
4029 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4030 	return (0);
4031 }
4032 
4033 int
4034 if_getcapenable(if_t ifp)
4035 {
4036 	return ((struct ifnet *)ifp)->if_capenable;
4037 }
4038 
4039 /*
4040  * This is largely undesirable because it ties ifnet to a device, but does
4041  * provide flexiblity for an embedded product vendor. Should be used with
4042  * the understanding that it violates the interface boundaries, and should be
4043  * a last resort only.
4044  */
4045 int
4046 if_setdev(if_t ifp, void *dev)
4047 {
4048 	return (0);
4049 }
4050 
4051 int
4052 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4053 {
4054 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4055 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4056 
4057 	return (0);
4058 }
4059 
4060 int
4061 if_getdrvflags(if_t ifp)
4062 {
4063 	return ((struct ifnet *)ifp)->if_drv_flags;
4064 }
4065 
4066 int
4067 if_setdrvflags(if_t ifp, int flags)
4068 {
4069 	((struct ifnet *)ifp)->if_drv_flags = flags;
4070 	return (0);
4071 }
4072 
4073 
4074 int
4075 if_setflags(if_t ifp, int flags)
4076 {
4077 	((struct ifnet *)ifp)->if_flags = flags;
4078 	return (0);
4079 }
4080 
4081 int
4082 if_setflagbits(if_t ifp, int set, int clear)
4083 {
4084 	((struct ifnet *)ifp)->if_flags |= set;
4085 	((struct ifnet *)ifp)->if_flags &= ~clear;
4086 
4087 	return (0);
4088 }
4089 
4090 int
4091 if_getflags(if_t ifp)
4092 {
4093 	return ((struct ifnet *)ifp)->if_flags;
4094 }
4095 
4096 int
4097 if_clearhwassist(if_t ifp)
4098 {
4099 	((struct ifnet *)ifp)->if_hwassist = 0;
4100 	return (0);
4101 }
4102 
4103 int
4104 if_sethwassistbits(if_t ifp, int toset, int toclear)
4105 {
4106 	((struct ifnet *)ifp)->if_hwassist |= toset;
4107 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4108 
4109 	return (0);
4110 }
4111 
4112 int
4113 if_sethwassist(if_t ifp, int hwassist_bit)
4114 {
4115 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4116 	return (0);
4117 }
4118 
4119 int
4120 if_gethwassist(if_t ifp)
4121 {
4122 	return ((struct ifnet *)ifp)->if_hwassist;
4123 }
4124 
4125 int
4126 if_setmtu(if_t ifp, int mtu)
4127 {
4128 	((struct ifnet *)ifp)->if_mtu = mtu;
4129 	return (0);
4130 }
4131 
4132 int
4133 if_getmtu(if_t ifp)
4134 {
4135 	return ((struct ifnet *)ifp)->if_mtu;
4136 }
4137 
4138 int
4139 if_getmtu_family(if_t ifp, int family)
4140 {
4141 	struct domain *dp;
4142 
4143 	for (dp = domains; dp; dp = dp->dom_next) {
4144 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4145 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4146 	}
4147 
4148 	return (((struct ifnet *)ifp)->if_mtu);
4149 }
4150 
4151 int
4152 if_setsoftc(if_t ifp, void *softc)
4153 {
4154 	((struct ifnet *)ifp)->if_softc = softc;
4155 	return (0);
4156 }
4157 
4158 void *
4159 if_getsoftc(if_t ifp)
4160 {
4161 	return ((struct ifnet *)ifp)->if_softc;
4162 }
4163 
4164 void
4165 if_setrcvif(struct mbuf *m, if_t ifp)
4166 {
4167 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4168 }
4169 
4170 void
4171 if_setvtag(struct mbuf *m, uint16_t tag)
4172 {
4173 	m->m_pkthdr.ether_vtag = tag;
4174 }
4175 
4176 uint16_t
4177 if_getvtag(struct mbuf *m)
4178 {
4179 
4180 	return (m->m_pkthdr.ether_vtag);
4181 }
4182 
4183 int
4184 if_sendq_empty(if_t ifp)
4185 {
4186 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4187 }
4188 
4189 struct ifaddr *
4190 if_getifaddr(if_t ifp)
4191 {
4192 	return ((struct ifnet *)ifp)->if_addr;
4193 }
4194 
4195 int
4196 if_getamcount(if_t ifp)
4197 {
4198 	return ((struct ifnet *)ifp)->if_amcount;
4199 }
4200 
4201 
4202 int
4203 if_setsendqready(if_t ifp)
4204 {
4205 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4206 	return (0);
4207 }
4208 
4209 int
4210 if_setsendqlen(if_t ifp, int tx_desc_count)
4211 {
4212 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4213 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4214 
4215 	return (0);
4216 }
4217 
4218 int
4219 if_vlantrunkinuse(if_t ifp)
4220 {
4221 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4222 }
4223 
4224 int
4225 if_input(if_t ifp, struct mbuf* sendmp)
4226 {
4227 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4228 	return (0);
4229 
4230 }
4231 
4232 /* XXX */
4233 #ifndef ETH_ADDR_LEN
4234 #define ETH_ADDR_LEN 6
4235 #endif
4236 
4237 int
4238 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4239 {
4240 	struct ifmultiaddr *ifma;
4241 	uint8_t *lmta = (uint8_t *)mta;
4242 	int mcnt = 0;
4243 
4244 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4245 		if (ifma->ifma_addr->sa_family != AF_LINK)
4246 			continue;
4247 
4248 		if (mcnt == max)
4249 			break;
4250 
4251 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4252 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4253 		mcnt++;
4254 	}
4255 	*cnt = mcnt;
4256 
4257 	return (0);
4258 }
4259 
4260 int
4261 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4262 {
4263 	int error;
4264 
4265 	if_maddr_rlock(ifp);
4266 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4267 	if_maddr_runlock(ifp);
4268 	return (error);
4269 }
4270 
4271 int
4272 if_multiaddr_count(if_t ifp, int max)
4273 {
4274 	struct ifmultiaddr *ifma;
4275 	int count;
4276 
4277 	count = 0;
4278 	if_maddr_rlock(ifp);
4279 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4280 		if (ifma->ifma_addr->sa_family != AF_LINK)
4281 			continue;
4282 		count++;
4283 		if (count == max)
4284 			break;
4285 	}
4286 	if_maddr_runlock(ifp);
4287 	return (count);
4288 }
4289 
4290 int
4291 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4292 {
4293 	struct ifmultiaddr *ifma;
4294 	int cnt = 0;
4295 
4296 	if_maddr_rlock(ifp);
4297 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4298 		cnt += filter(arg, ifma, cnt);
4299 	if_maddr_runlock(ifp);
4300 	return (cnt);
4301 }
4302 
4303 struct mbuf *
4304 if_dequeue(if_t ifp)
4305 {
4306 	struct mbuf *m;
4307 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4308 
4309 	return (m);
4310 }
4311 
4312 int
4313 if_sendq_prepend(if_t ifp, struct mbuf *m)
4314 {
4315 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4316 	return (0);
4317 }
4318 
4319 int
4320 if_setifheaderlen(if_t ifp, int len)
4321 {
4322 	((struct ifnet *)ifp)->if_hdrlen = len;
4323 	return (0);
4324 }
4325 
4326 caddr_t
4327 if_getlladdr(if_t ifp)
4328 {
4329 	return (IF_LLADDR((struct ifnet *)ifp));
4330 }
4331 
4332 void *
4333 if_gethandle(u_char type)
4334 {
4335 	return (if_alloc(type));
4336 }
4337 
4338 void
4339 if_bpfmtap(if_t ifh, struct mbuf *m)
4340 {
4341 	struct ifnet *ifp = (struct ifnet *)ifh;
4342 
4343 	BPF_MTAP(ifp, m);
4344 }
4345 
4346 void
4347 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4348 {
4349 	struct ifnet *ifp = (struct ifnet *)ifh;
4350 
4351 	ETHER_BPF_MTAP(ifp, m);
4352 }
4353 
4354 void
4355 if_vlancap(if_t ifh)
4356 {
4357 	struct ifnet *ifp = (struct ifnet *)ifh;
4358 	VLAN_CAPABILITIES(ifp);
4359 }
4360 
4361 int
4362 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4363 {
4364 
4365 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4366         return (0);
4367 }
4368 
4369 int
4370 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4371 {
4372 
4373 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4374         return (0);
4375 }
4376 
4377 int
4378 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4379 {
4380 
4381 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4382         return (0);
4383 }
4384 
4385 u_int
4386 if_gethwtsomax(if_t ifp)
4387 {
4388 
4389 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4390 }
4391 
4392 u_int
4393 if_gethwtsomaxsegcount(if_t ifp)
4394 {
4395 
4396 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4397 }
4398 
4399 u_int
4400 if_gethwtsomaxsegsize(if_t ifp)
4401 {
4402 
4403 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4404 }
4405 
4406 void
4407 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4408 {
4409 	((struct ifnet *)ifp)->if_init = init_fn;
4410 }
4411 
4412 void
4413 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4414 {
4415 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4416 }
4417 
4418 void
4419 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4420 {
4421 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4422 }
4423 
4424 void
4425 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4426 {
4427 	((struct ifnet *)ifp)->if_transmit = start_fn;
4428 }
4429 
4430 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4431 {
4432 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4433 
4434 }
4435 
4436 void
4437 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4438 {
4439 
4440 	ifp->if_get_counter = fn;
4441 }
4442 
4443 /* Revisit these - These are inline functions originally. */
4444 int
4445 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4446 {
4447 	return drbr_inuse(ifh, br);
4448 }
4449 
4450 struct mbuf*
4451 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4452 {
4453 	return drbr_dequeue(ifh, br);
4454 }
4455 
4456 int
4457 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4458 {
4459 	return drbr_needs_enqueue(ifh, br);
4460 }
4461 
4462 int
4463 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4464 {
4465 	return drbr_enqueue(ifh, br, m);
4466 
4467 }
4468