xref: /freebsd/sys/net/if.c (revision 3fc36ee018bb836bd1796067cf4ef8683f166ebc)
1 /*-
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62 
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65 
66 #include <net/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/if.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif /* INET */
89 #ifdef INET6
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
92 #endif /* INET6 */
93 #endif /* INET || INET6 */
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
100 #endif
101 
102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
104 
105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106     &ifqmaxlen, 0, "max send queue size");
107 
108 /* Log link state change events */
109 static int log_link_state_change = 1;
110 
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 	&log_link_state_change, 0,
113 	"log interface link state change events");
114 
115 /* Log promiscuous mode change events */
116 static int log_promisc_mode_change = 1;
117 
118 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
119 	&log_promisc_mode_change, 1,
120 	"log promiscuous mode change events");
121 
122 /* Interface description */
123 static unsigned int ifdescr_maxlen = 1024;
124 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
125 	&ifdescr_maxlen, 0,
126 	"administrative maximum length for interface description");
127 
128 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
129 
130 /* global sx for non-critical path ifdescr */
131 static struct sx ifdescr_sx;
132 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
133 
134 void	(*bridge_linkstate_p)(struct ifnet *ifp);
135 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
136 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
137 /* These are external hooks for CARP. */
138 void	(*carp_linkstate_p)(struct ifnet *ifp);
139 void	(*carp_demote_adj_p)(int, char *);
140 int	(*carp_master_p)(struct ifaddr *);
141 #if defined(INET) || defined(INET6)
142 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
143 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
144     const struct sockaddr *sa);
145 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
146 int	(*carp_attach_p)(struct ifaddr *, int);
147 void	(*carp_detach_p)(struct ifaddr *);
148 #endif
149 #ifdef INET
150 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
151 #endif
152 #ifdef INET6
153 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
154 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
155     const struct in6_addr *taddr);
156 #endif
157 
158 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
159 
160 /*
161  * XXX: Style; these should be sorted alphabetically, and unprototyped
162  * static functions should be prototyped. Currently they are sorted by
163  * declaration order.
164  */
165 static void	if_attachdomain(void *);
166 static void	if_attachdomain1(struct ifnet *);
167 static int	ifconf(u_long, caddr_t);
168 static void	if_freemulti(struct ifmultiaddr *);
169 static void	if_grow(void);
170 static void	if_input_default(struct ifnet *, struct mbuf *);
171 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
172 static void	if_route(struct ifnet *, int flag, int fam);
173 static int	if_setflag(struct ifnet *, int, int, int *, int);
174 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
175 static void	if_unroute(struct ifnet *, int flag, int fam);
176 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
177 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
178 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
179 static void	do_link_state_change(void *, int);
180 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
181 static int	if_getgroupmembers(struct ifgroupreq *);
182 static void	if_delgroups(struct ifnet *);
183 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
184 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
185 #ifdef VIMAGE
186 static void	if_vmove(struct ifnet *, struct vnet *);
187 #endif
188 
189 #ifdef INET6
190 /*
191  * XXX: declare here to avoid to include many inet6 related files..
192  * should be more generalized?
193  */
194 extern void	nd6_setmtu(struct ifnet *);
195 #endif
196 
197 /* ipsec helper hooks */
198 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
199 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
200 
201 VNET_DEFINE(int, if_index);
202 int	ifqmaxlen = IFQ_MAXLEN;
203 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
204 VNET_DEFINE(struct ifgrouphead, ifg_head);
205 
206 static VNET_DEFINE(int, if_indexlim) = 8;
207 
208 /* Table of ifnet by index. */
209 VNET_DEFINE(struct ifnet **, ifindex_table);
210 
211 #define	V_if_indexlim		VNET(if_indexlim)
212 #define	V_ifindex_table		VNET(ifindex_table)
213 
214 /*
215  * The global network interface list (V_ifnet) and related state (such as
216  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
217  * an rwlock.  Either may be acquired shared to stablize the list, but both
218  * must be acquired writable to modify the list.  This model allows us to
219  * both stablize the interface list during interrupt thread processing, but
220  * also to stablize it over long-running ioctls, without introducing priority
221  * inversions and deadlocks.
222  */
223 struct rwlock ifnet_rwlock;
224 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
225 struct sx ifnet_sxlock;
226 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
227 
228 /*
229  * The allocation of network interfaces is a rather non-atomic affair; we
230  * need to select an index before we are ready to expose the interface for
231  * use, so will use this pointer value to indicate reservation.
232  */
233 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
234 
235 static	if_com_alloc_t *if_com_alloc[256];
236 static	if_com_free_t *if_com_free[256];
237 
238 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
239 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
240 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
241 
242 struct ifnet *
243 ifnet_byindex_locked(u_short idx)
244 {
245 
246 	if (idx > V_if_index)
247 		return (NULL);
248 	if (V_ifindex_table[idx] == IFNET_HOLD)
249 		return (NULL);
250 	return (V_ifindex_table[idx]);
251 }
252 
253 struct ifnet *
254 ifnet_byindex(u_short idx)
255 {
256 	struct ifnet *ifp;
257 
258 	IFNET_RLOCK_NOSLEEP();
259 	ifp = ifnet_byindex_locked(idx);
260 	IFNET_RUNLOCK_NOSLEEP();
261 	return (ifp);
262 }
263 
264 struct ifnet *
265 ifnet_byindex_ref(u_short idx)
266 {
267 	struct ifnet *ifp;
268 
269 	IFNET_RLOCK_NOSLEEP();
270 	ifp = ifnet_byindex_locked(idx);
271 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
272 		IFNET_RUNLOCK_NOSLEEP();
273 		return (NULL);
274 	}
275 	if_ref(ifp);
276 	IFNET_RUNLOCK_NOSLEEP();
277 	return (ifp);
278 }
279 
280 /*
281  * Allocate an ifindex array entry; return 0 on success or an error on
282  * failure.
283  */
284 static u_short
285 ifindex_alloc(void)
286 {
287 	u_short idx;
288 
289 	IFNET_WLOCK_ASSERT();
290 retry:
291 	/*
292 	 * Try to find an empty slot below V_if_index.  If we fail, take the
293 	 * next slot.
294 	 */
295 	for (idx = 1; idx <= V_if_index; idx++) {
296 		if (V_ifindex_table[idx] == NULL)
297 			break;
298 	}
299 
300 	/* Catch if_index overflow. */
301 	if (idx >= V_if_indexlim) {
302 		if_grow();
303 		goto retry;
304 	}
305 	if (idx > V_if_index)
306 		V_if_index = idx;
307 	return (idx);
308 }
309 
310 static void
311 ifindex_free_locked(u_short idx)
312 {
313 
314 	IFNET_WLOCK_ASSERT();
315 
316 	V_ifindex_table[idx] = NULL;
317 	while (V_if_index > 0 &&
318 	    V_ifindex_table[V_if_index] == NULL)
319 		V_if_index--;
320 }
321 
322 static void
323 ifindex_free(u_short idx)
324 {
325 
326 	IFNET_WLOCK();
327 	ifindex_free_locked(idx);
328 	IFNET_WUNLOCK();
329 }
330 
331 static void
332 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
333 {
334 
335 	IFNET_WLOCK_ASSERT();
336 
337 	V_ifindex_table[idx] = ifp;
338 }
339 
340 static void
341 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
342 {
343 
344 	IFNET_WLOCK();
345 	ifnet_setbyindex_locked(idx, ifp);
346 	IFNET_WUNLOCK();
347 }
348 
349 struct ifaddr *
350 ifaddr_byindex(u_short idx)
351 {
352 	struct ifnet *ifp;
353 	struct ifaddr *ifa = NULL;
354 
355 	IFNET_RLOCK_NOSLEEP();
356 	ifp = ifnet_byindex_locked(idx);
357 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
358 		ifa_ref(ifa);
359 	IFNET_RUNLOCK_NOSLEEP();
360 	return (ifa);
361 }
362 
363 /*
364  * Network interface utility routines.
365  *
366  * Routines with ifa_ifwith* names take sockaddr *'s as
367  * parameters.
368  */
369 
370 static void
371 vnet_if_init(const void *unused __unused)
372 {
373 
374 	TAILQ_INIT(&V_ifnet);
375 	TAILQ_INIT(&V_ifg_head);
376 	IFNET_WLOCK();
377 	if_grow();				/* create initial table */
378 	IFNET_WUNLOCK();
379 	vnet_if_clone_init();
380 }
381 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
382     NULL);
383 
384 #ifdef VIMAGE
385 static void
386 vnet_if_uninit(const void *unused __unused)
387 {
388 
389 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
390 	    "not empty", __func__, __LINE__, &V_ifnet));
391 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
392 	    "not empty", __func__, __LINE__, &V_ifg_head));
393 
394 	free((caddr_t)V_ifindex_table, M_IFNET);
395 }
396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
397     vnet_if_uninit, NULL);
398 
399 static void
400 vnet_if_return(const void *unused __unused)
401 {
402 	struct ifnet *ifp, *nifp;
403 
404 	/* Return all inherited interfaces to their parent vnets. */
405 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
406 		if (ifp->if_home_vnet != ifp->if_vnet)
407 			if_vmove(ifp, ifp->if_home_vnet);
408 	}
409 }
410 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
411     vnet_if_return, NULL);
412 #endif
413 
414 static void
415 if_grow(void)
416 {
417 	int oldlim;
418 	u_int n;
419 	struct ifnet **e;
420 
421 	IFNET_WLOCK_ASSERT();
422 	oldlim = V_if_indexlim;
423 	IFNET_WUNLOCK();
424 	n = (oldlim << 1) * sizeof(*e);
425 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
426 	IFNET_WLOCK();
427 	if (V_if_indexlim != oldlim) {
428 		free(e, M_IFNET);
429 		return;
430 	}
431 	if (V_ifindex_table != NULL) {
432 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
433 		free((caddr_t)V_ifindex_table, M_IFNET);
434 	}
435 	V_if_indexlim <<= 1;
436 	V_ifindex_table = e;
437 }
438 
439 /*
440  * Allocate a struct ifnet and an index for an interface.  A layer 2
441  * common structure will also be allocated if an allocation routine is
442  * registered for the passed type.
443  */
444 struct ifnet *
445 if_alloc(u_char type)
446 {
447 	struct ifnet *ifp;
448 	u_short idx;
449 
450 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
451 	IFNET_WLOCK();
452 	idx = ifindex_alloc();
453 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
454 	IFNET_WUNLOCK();
455 	ifp->if_index = idx;
456 	ifp->if_type = type;
457 	ifp->if_alloctype = type;
458 #ifdef VIMAGE
459 	ifp->if_vnet = curvnet;
460 #endif
461 	if (if_com_alloc[type] != NULL) {
462 		ifp->if_l2com = if_com_alloc[type](type, ifp);
463 		if (ifp->if_l2com == NULL) {
464 			free(ifp, M_IFNET);
465 			ifindex_free(idx);
466 			return (NULL);
467 		}
468 	}
469 
470 	IF_ADDR_LOCK_INIT(ifp);
471 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
472 	ifp->if_afdata_initialized = 0;
473 	IF_AFDATA_LOCK_INIT(ifp);
474 	TAILQ_INIT(&ifp->if_addrhead);
475 	TAILQ_INIT(&ifp->if_multiaddrs);
476 	TAILQ_INIT(&ifp->if_groups);
477 #ifdef MAC
478 	mac_ifnet_init(ifp);
479 #endif
480 	ifq_init(&ifp->if_snd, ifp);
481 
482 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
483 	for (int i = 0; i < IFCOUNTERS; i++)
484 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
485 	ifp->if_get_counter = if_get_counter_default;
486 	ifnet_setbyindex(ifp->if_index, ifp);
487 	return (ifp);
488 }
489 
490 /*
491  * Do the actual work of freeing a struct ifnet, and layer 2 common
492  * structure.  This call is made when the last reference to an
493  * interface is released.
494  */
495 static void
496 if_free_internal(struct ifnet *ifp)
497 {
498 
499 	KASSERT((ifp->if_flags & IFF_DYING),
500 	    ("if_free_internal: interface not dying"));
501 
502 	if (if_com_free[ifp->if_alloctype] != NULL)
503 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
504 		    ifp->if_alloctype);
505 
506 #ifdef MAC
507 	mac_ifnet_destroy(ifp);
508 #endif /* MAC */
509 	if (ifp->if_description != NULL)
510 		free(ifp->if_description, M_IFDESCR);
511 	IF_AFDATA_DESTROY(ifp);
512 	IF_ADDR_LOCK_DESTROY(ifp);
513 	ifq_delete(&ifp->if_snd);
514 
515 	for (int i = 0; i < IFCOUNTERS; i++)
516 		counter_u64_free(ifp->if_counters[i]);
517 
518 	free(ifp, M_IFNET);
519 }
520 
521 /*
522  * Deregister an interface and free the associated storage.
523  */
524 void
525 if_free(struct ifnet *ifp)
526 {
527 
528 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
529 
530 	CURVNET_SET_QUIET(ifp->if_vnet);
531 	IFNET_WLOCK();
532 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
533 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
534 
535 	ifindex_free_locked(ifp->if_index);
536 	IFNET_WUNLOCK();
537 
538 	if (refcount_release(&ifp->if_refcount))
539 		if_free_internal(ifp);
540 	CURVNET_RESTORE();
541 }
542 
543 /*
544  * Interfaces to keep an ifnet type-stable despite the possibility of the
545  * driver calling if_free().  If there are additional references, we defer
546  * freeing the underlying data structure.
547  */
548 void
549 if_ref(struct ifnet *ifp)
550 {
551 
552 	/* We don't assert the ifnet list lock here, but arguably should. */
553 	refcount_acquire(&ifp->if_refcount);
554 }
555 
556 void
557 if_rele(struct ifnet *ifp)
558 {
559 
560 	if (!refcount_release(&ifp->if_refcount))
561 		return;
562 	if_free_internal(ifp);
563 }
564 
565 void
566 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
567 {
568 
569 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
570 
571 	if (ifq->ifq_maxlen == 0)
572 		ifq->ifq_maxlen = ifqmaxlen;
573 
574 	ifq->altq_type = 0;
575 	ifq->altq_disc = NULL;
576 	ifq->altq_flags &= ALTQF_CANTCHANGE;
577 	ifq->altq_tbr  = NULL;
578 	ifq->altq_ifp  = ifp;
579 }
580 
581 void
582 ifq_delete(struct ifaltq *ifq)
583 {
584 	mtx_destroy(&ifq->ifq_mtx);
585 }
586 
587 /*
588  * Perform generic interface initialization tasks and attach the interface
589  * to the list of "active" interfaces.  If vmove flag is set on entry
590  * to if_attach_internal(), perform only a limited subset of initialization
591  * tasks, given that we are moving from one vnet to another an ifnet which
592  * has already been fully initialized.
593  *
594  * Note that if_detach_internal() removes group membership unconditionally
595  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
596  * Thus, when if_vmove() is applied to a cloned interface, group membership
597  * is lost while a cloned one always joins a group whose name is
598  * ifc->ifc_name.  To recover this after if_detach_internal() and
599  * if_attach_internal(), the cloner should be specified to
600  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
601  * attempts to join a group whose name is ifc->ifc_name.
602  *
603  * XXX:
604  *  - The decision to return void and thus require this function to
605  *    succeed is questionable.
606  *  - We should probably do more sanity checking.  For instance we don't
607  *    do anything to insure if_xname is unique or non-empty.
608  */
609 void
610 if_attach(struct ifnet *ifp)
611 {
612 
613 	if_attach_internal(ifp, 0, NULL);
614 }
615 
616 /*
617  * Compute the least common TSO limit.
618  */
619 void
620 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
621 {
622 	/*
623 	 * 1) If there is no limit currently, take the limit from
624 	 * the network adapter.
625 	 *
626 	 * 2) If the network adapter has a limit below the current
627 	 * limit, apply it.
628 	 */
629 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
630 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
631 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
632 	}
633 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
634 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
635 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
636 	}
637 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
638 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
639 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
640 	}
641 }
642 
643 /*
644  * Update TSO limit of a network adapter.
645  *
646  * Returns zero if no change. Else non-zero.
647  */
648 int
649 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
650 {
651 	int retval = 0;
652 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
653 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
654 		retval++;
655 	}
656 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
657 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
658 		retval++;
659 	}
660 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
661 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
662 		retval++;
663 	}
664 	return (retval);
665 }
666 
667 static void
668 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
669 {
670 	unsigned socksize, ifasize;
671 	int namelen, masklen;
672 	struct sockaddr_dl *sdl;
673 	struct ifaddr *ifa;
674 
675 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
676 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
677 		    ifp->if_xname);
678 
679 #ifdef VIMAGE
680 	ifp->if_vnet = curvnet;
681 	if (ifp->if_home_vnet == NULL)
682 		ifp->if_home_vnet = curvnet;
683 #endif
684 
685 	if_addgroup(ifp, IFG_ALL);
686 
687 	/* Restore group membership for cloned interfaces. */
688 	if (vmove && ifc != NULL)
689 		if_clone_addgroup(ifp, ifc);
690 
691 	getmicrotime(&ifp->if_lastchange);
692 	ifp->if_epoch = time_uptime;
693 
694 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
695 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
696 	    ("transmit and qflush must both either be set or both be NULL"));
697 	if (ifp->if_transmit == NULL) {
698 		ifp->if_transmit = if_transmit;
699 		ifp->if_qflush = if_qflush;
700 	}
701 	if (ifp->if_input == NULL)
702 		ifp->if_input = if_input_default;
703 
704 	if (ifp->if_requestencap == NULL)
705 		ifp->if_requestencap = if_requestencap_default;
706 
707 	if (!vmove) {
708 #ifdef MAC
709 		mac_ifnet_create(ifp);
710 #endif
711 
712 		/*
713 		 * Create a Link Level name for this device.
714 		 */
715 		namelen = strlen(ifp->if_xname);
716 		/*
717 		 * Always save enough space for any possiable name so we
718 		 * can do a rename in place later.
719 		 */
720 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
721 		socksize = masklen + ifp->if_addrlen;
722 		if (socksize < sizeof(*sdl))
723 			socksize = sizeof(*sdl);
724 		socksize = roundup2(socksize, sizeof(long));
725 		ifasize = sizeof(*ifa) + 2 * socksize;
726 		ifa = ifa_alloc(ifasize, M_WAITOK);
727 		sdl = (struct sockaddr_dl *)(ifa + 1);
728 		sdl->sdl_len = socksize;
729 		sdl->sdl_family = AF_LINK;
730 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
731 		sdl->sdl_nlen = namelen;
732 		sdl->sdl_index = ifp->if_index;
733 		sdl->sdl_type = ifp->if_type;
734 		ifp->if_addr = ifa;
735 		ifa->ifa_ifp = ifp;
736 		ifa->ifa_rtrequest = link_rtrequest;
737 		ifa->ifa_addr = (struct sockaddr *)sdl;
738 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
739 		ifa->ifa_netmask = (struct sockaddr *)sdl;
740 		sdl->sdl_len = masklen;
741 		while (namelen != 0)
742 			sdl->sdl_data[--namelen] = 0xff;
743 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
744 		/* Reliably crash if used uninitialized. */
745 		ifp->if_broadcastaddr = NULL;
746 
747 #if defined(INET) || defined(INET6)
748 		/* Use defaults for TSO, if nothing is set */
749 		if (ifp->if_hw_tsomax == 0 &&
750 		    ifp->if_hw_tsomaxsegcount == 0 &&
751 		    ifp->if_hw_tsomaxsegsize == 0) {
752 			/*
753 			 * The TSO defaults needs to be such that an
754 			 * NFS mbuf list of 35 mbufs totalling just
755 			 * below 64K works and that a chain of mbufs
756 			 * can be defragged into at most 32 segments:
757 			 */
758 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
759 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
760 			ifp->if_hw_tsomaxsegcount = 35;
761 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
762 
763 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
764 			if (ifp->if_capabilities & IFCAP_TSO) {
765 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
766 				    ifp->if_hw_tsomax,
767 				    ifp->if_hw_tsomaxsegcount,
768 				    ifp->if_hw_tsomaxsegsize);
769 			}
770 		}
771 #endif
772 	}
773 #ifdef VIMAGE
774 	else {
775 		/*
776 		 * Update the interface index in the link layer address
777 		 * of the interface.
778 		 */
779 		for (ifa = ifp->if_addr; ifa != NULL;
780 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
781 			if (ifa->ifa_addr->sa_family == AF_LINK) {
782 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
783 				sdl->sdl_index = ifp->if_index;
784 			}
785 		}
786 	}
787 #endif
788 
789 	IFNET_WLOCK();
790 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
791 #ifdef VIMAGE
792 	curvnet->vnet_ifcnt++;
793 #endif
794 	IFNET_WUNLOCK();
795 
796 	if (domain_init_status >= 2)
797 		if_attachdomain1(ifp);
798 
799 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
800 	if (IS_DEFAULT_VNET(curvnet))
801 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
802 
803 	/* Announce the interface. */
804 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
805 }
806 
807 static void
808 if_attachdomain(void *dummy)
809 {
810 	struct ifnet *ifp;
811 
812 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
813 		if_attachdomain1(ifp);
814 }
815 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
816     if_attachdomain, NULL);
817 
818 static void
819 if_attachdomain1(struct ifnet *ifp)
820 {
821 	struct domain *dp;
822 
823 	/*
824 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
825 	 * cannot lock ifp->if_afdata initialization, entirely.
826 	 */
827 	IF_AFDATA_LOCK(ifp);
828 	if (ifp->if_afdata_initialized >= domain_init_status) {
829 		IF_AFDATA_UNLOCK(ifp);
830 		log(LOG_WARNING, "%s called more than once on %s\n",
831 		    __func__, ifp->if_xname);
832 		return;
833 	}
834 	ifp->if_afdata_initialized = domain_init_status;
835 	IF_AFDATA_UNLOCK(ifp);
836 
837 	/* address family dependent data region */
838 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
839 	for (dp = domains; dp; dp = dp->dom_next) {
840 		if (dp->dom_ifattach)
841 			ifp->if_afdata[dp->dom_family] =
842 			    (*dp->dom_ifattach)(ifp);
843 	}
844 }
845 
846 /*
847  * Remove any unicast or broadcast network addresses from an interface.
848  */
849 void
850 if_purgeaddrs(struct ifnet *ifp)
851 {
852 	struct ifaddr *ifa, *next;
853 
854 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
855 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
856 		if (ifa->ifa_addr->sa_family == AF_LINK)
857 			continue;
858 #ifdef INET
859 		/* XXX: Ugly!! ad hoc just for INET */
860 		if (ifa->ifa_addr->sa_family == AF_INET) {
861 			struct ifaliasreq ifr;
862 
863 			bzero(&ifr, sizeof(ifr));
864 			ifr.ifra_addr = *ifa->ifa_addr;
865 			if (ifa->ifa_dstaddr)
866 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
867 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
868 			    NULL) == 0)
869 				continue;
870 		}
871 #endif /* INET */
872 #ifdef INET6
873 		if (ifa->ifa_addr->sa_family == AF_INET6) {
874 			in6_purgeaddr(ifa);
875 			/* ifp_addrhead is already updated */
876 			continue;
877 		}
878 #endif /* INET6 */
879 		IF_ADDR_WLOCK(ifp);
880 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
881 		IF_ADDR_WUNLOCK(ifp);
882 		ifa_free(ifa);
883 	}
884 }
885 
886 /*
887  * Remove any multicast network addresses from an interface when an ifnet
888  * is going away.
889  */
890 static void
891 if_purgemaddrs(struct ifnet *ifp)
892 {
893 	struct ifmultiaddr *ifma;
894 	struct ifmultiaddr *next;
895 
896 	IF_ADDR_WLOCK(ifp);
897 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
898 		if_delmulti_locked(ifp, ifma, 1);
899 	IF_ADDR_WUNLOCK(ifp);
900 }
901 
902 /*
903  * Detach an interface, removing it from the list of "active" interfaces.
904  * If vmove flag is set on entry to if_detach_internal(), perform only a
905  * limited subset of cleanup tasks, given that we are moving an ifnet from
906  * one vnet to another, where it must be fully operational.
907  *
908  * XXXRW: There are some significant questions about event ordering, and
909  * how to prevent things from starting to use the interface during detach.
910  */
911 void
912 if_detach(struct ifnet *ifp)
913 {
914 
915 	CURVNET_SET_QUIET(ifp->if_vnet);
916 	if_detach_internal(ifp, 0, NULL);
917 	CURVNET_RESTORE();
918 }
919 
920 /*
921  * The vmove flag, if set, indicates that we are called from a callpath
922  * that is moving an interface to a different vnet instance.
923  *
924  * The shutdown flag, if set, indicates that we are called in the
925  * process of shutting down a vnet instance.  Currently only the
926  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
927  * on a vnet instance shutdown without this flag being set, e.g., when
928  * the cloned interfaces are destoyed as first thing of teardown.
929  */
930 static int
931 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
932 {
933 	struct ifaddr *ifa;
934 	int i;
935 	struct domain *dp;
936  	struct ifnet *iter;
937  	int found = 0;
938 #ifdef VIMAGE
939 	int shutdown;
940 
941 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
942 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
943 #endif
944 	IFNET_WLOCK();
945 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
946 		if (iter == ifp) {
947 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
948 			found = 1;
949 			break;
950 		}
951 	IFNET_WUNLOCK();
952 	if (!found) {
953 		/*
954 		 * While we would want to panic here, we cannot
955 		 * guarantee that the interface is indeed still on
956 		 * the list given we don't hold locks all the way.
957 		 */
958 		return (ENOENT);
959 #if 0
960 		if (vmove)
961 			panic("%s: ifp=%p not on the ifnet tailq %p",
962 			    __func__, ifp, &V_ifnet);
963 		else
964 			return; /* XXX this should panic as well? */
965 #endif
966 	}
967 
968 	/*
969 	 * At this point we know the interface still was on the ifnet list
970 	 * and we removed it so we are in a stable state.
971 	 */
972 #ifdef VIMAGE
973 	curvnet->vnet_ifcnt--;
974 #endif
975 
976 	/*
977 	 * In any case (destroy or vmove) detach us from the groups
978 	 * and remove/wait for pending events on the taskq.
979 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
980 	 */
981 	if_delgroups(ifp);
982 
983 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
984 
985 	/*
986 	 * Check if this is a cloned interface or not. Must do even if
987 	 * shutting down as a if_vmove_reclaim() would move the ifp and
988 	 * the if_clone_addgroup() will have a corrupted string overwise
989 	 * from a gibberish pointer.
990 	 */
991 	if (vmove && ifcp != NULL)
992 		*ifcp = if_clone_findifc(ifp);
993 
994 	if_down(ifp);
995 
996 #ifdef VIMAGE
997 	/*
998 	 * On VNET shutdown abort here as the stack teardown will do all
999 	 * the work top-down for us.
1000 	 */
1001 	if (shutdown) {
1002 		/*
1003 		 * In case of a vmove we are done here without error.
1004 		 * If we would signal an error it would lead to the same
1005 		 * abort as if we did not find the ifnet anymore.
1006 		 * if_detach() calls us in void context and does not care
1007 		 * about an early abort notification, so life is splendid :)
1008 		 */
1009 		goto finish_vnet_shutdown;
1010 	}
1011 #endif
1012 
1013 	/*
1014 	 * At this point we are not tearing down a VNET and are either
1015 	 * going to destroy or vmove the interface and have to cleanup
1016 	 * accordingly.
1017 	 */
1018 
1019 	/*
1020 	 * Remove routes and flush queues.
1021 	 */
1022 #ifdef ALTQ
1023 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1024 		altq_disable(&ifp->if_snd);
1025 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1026 		altq_detach(&ifp->if_snd);
1027 #endif
1028 
1029 	if_purgeaddrs(ifp);
1030 
1031 #ifdef INET
1032 	in_ifdetach(ifp);
1033 #endif
1034 
1035 #ifdef INET6
1036 	/*
1037 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1038 	 * before removing routing entries below, since IPv6 interface direct
1039 	 * routes are expected to be removed by the IPv6-specific kernel API.
1040 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1041 	 */
1042 	in6_ifdetach(ifp);
1043 #endif
1044 	if_purgemaddrs(ifp);
1045 
1046 	/* Announce that the interface is gone. */
1047 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1048 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1049 	if (IS_DEFAULT_VNET(curvnet))
1050 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1051 
1052 	if (!vmove) {
1053 		/*
1054 		 * Prevent further calls into the device driver via ifnet.
1055 		 */
1056 		if_dead(ifp);
1057 
1058 		/*
1059 		 * Remove link ifaddr pointer and maybe decrement if_index.
1060 		 * Clean up all addresses.
1061 		 */
1062 		ifp->if_addr = NULL;
1063 
1064 		/* We can now free link ifaddr. */
1065 		IF_ADDR_WLOCK(ifp);
1066 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1067 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1068 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1069 			IF_ADDR_WUNLOCK(ifp);
1070 			ifa_free(ifa);
1071 		} else
1072 			IF_ADDR_WUNLOCK(ifp);
1073 	}
1074 
1075 	rt_flushifroutes(ifp);
1076 
1077 #ifdef VIMAGE
1078 finish_vnet_shutdown:
1079 #endif
1080 	/*
1081 	 * We cannot hold the lock over dom_ifdetach calls as they might
1082 	 * sleep, for example trying to drain a callout, thus open up the
1083 	 * theoretical race with re-attaching.
1084 	 */
1085 	IF_AFDATA_LOCK(ifp);
1086 	i = ifp->if_afdata_initialized;
1087 	ifp->if_afdata_initialized = 0;
1088 	IF_AFDATA_UNLOCK(ifp);
1089 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1090 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1091 			(*dp->dom_ifdetach)(ifp,
1092 			    ifp->if_afdata[dp->dom_family]);
1093 			ifp->if_afdata[dp->dom_family] = NULL;
1094 		}
1095 	}
1096 
1097 	return (0);
1098 }
1099 
1100 #ifdef VIMAGE
1101 /*
1102  * if_vmove() performs a limited version of if_detach() in current
1103  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1104  * An attempt is made to shrink if_index in current vnet, find an
1105  * unused if_index in target vnet and calls if_grow() if necessary,
1106  * and finally find an unused if_xname for the target vnet.
1107  */
1108 static void
1109 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1110 {
1111 	struct if_clone *ifc;
1112 	u_int bif_dlt, bif_hdrlen;
1113 	int rc;
1114 
1115  	/*
1116 	 * if_detach_internal() will call the eventhandler to notify
1117 	 * interface departure.  That will detach if_bpf.  We need to
1118 	 * safe the dlt and hdrlen so we can re-attach it later.
1119 	 */
1120 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1121 
1122 	/*
1123 	 * Detach from current vnet, but preserve LLADDR info, do not
1124 	 * mark as dead etc. so that the ifnet can be reattached later.
1125 	 * If we cannot find it, we lost the race to someone else.
1126 	 */
1127 	rc = if_detach_internal(ifp, 1, &ifc);
1128 	if (rc != 0)
1129 		return;
1130 
1131 	/*
1132 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1133 	 * the if_index for that vnet if possible.
1134 	 *
1135 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1136 	 * or we'd lock on one vnet and unlock on another.
1137 	 */
1138 	IFNET_WLOCK();
1139 	ifindex_free_locked(ifp->if_index);
1140 	IFNET_WUNLOCK();
1141 
1142 	/*
1143 	 * Perform interface-specific reassignment tasks, if provided by
1144 	 * the driver.
1145 	 */
1146 	if (ifp->if_reassign != NULL)
1147 		ifp->if_reassign(ifp, new_vnet, NULL);
1148 
1149 	/*
1150 	 * Switch to the context of the target vnet.
1151 	 */
1152 	CURVNET_SET_QUIET(new_vnet);
1153 
1154 	IFNET_WLOCK();
1155 	ifp->if_index = ifindex_alloc();
1156 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1157 	IFNET_WUNLOCK();
1158 
1159 	if_attach_internal(ifp, 1, ifc);
1160 
1161 	if (ifp->if_bpf == NULL)
1162 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1163 
1164 	CURVNET_RESTORE();
1165 }
1166 
1167 /*
1168  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1169  */
1170 static int
1171 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1172 {
1173 	struct prison *pr;
1174 	struct ifnet *difp;
1175 	int shutdown;
1176 
1177 	/* Try to find the prison within our visibility. */
1178 	sx_slock(&allprison_lock);
1179 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1180 	sx_sunlock(&allprison_lock);
1181 	if (pr == NULL)
1182 		return (ENXIO);
1183 	prison_hold_locked(pr);
1184 	mtx_unlock(&pr->pr_mtx);
1185 
1186 	/* Do not try to move the iface from and to the same prison. */
1187 	if (pr->pr_vnet == ifp->if_vnet) {
1188 		prison_free(pr);
1189 		return (EEXIST);
1190 	}
1191 
1192 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1193 	/* XXX Lock interfaces to avoid races. */
1194 	CURVNET_SET_QUIET(pr->pr_vnet);
1195 	difp = ifunit(ifname);
1196 	if (difp != NULL) {
1197 		CURVNET_RESTORE();
1198 		prison_free(pr);
1199 		return (EEXIST);
1200 	}
1201 
1202 	/* Make sure the VNET is stable. */
1203 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1204 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1205 	if (shutdown) {
1206 		CURVNET_RESTORE();
1207 		prison_free(pr);
1208 		return (EBUSY);
1209 	}
1210 	CURVNET_RESTORE();
1211 
1212 	/* Move the interface into the child jail/vnet. */
1213 	if_vmove(ifp, pr->pr_vnet);
1214 
1215 	/* Report the new if_xname back to the userland. */
1216 	sprintf(ifname, "%s", ifp->if_xname);
1217 
1218 	prison_free(pr);
1219 	return (0);
1220 }
1221 
1222 static int
1223 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1224 {
1225 	struct prison *pr;
1226 	struct vnet *vnet_dst;
1227 	struct ifnet *ifp;
1228  	int shutdown;
1229 
1230 	/* Try to find the prison within our visibility. */
1231 	sx_slock(&allprison_lock);
1232 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1233 	sx_sunlock(&allprison_lock);
1234 	if (pr == NULL)
1235 		return (ENXIO);
1236 	prison_hold_locked(pr);
1237 	mtx_unlock(&pr->pr_mtx);
1238 
1239 	/* Make sure the named iface exists in the source prison/vnet. */
1240 	CURVNET_SET(pr->pr_vnet);
1241 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1242 	if (ifp == NULL) {
1243 		CURVNET_RESTORE();
1244 		prison_free(pr);
1245 		return (ENXIO);
1246 	}
1247 
1248 	/* Do not try to move the iface from and to the same prison. */
1249 	vnet_dst = TD_TO_VNET(td);
1250 	if (vnet_dst == ifp->if_vnet) {
1251 		CURVNET_RESTORE();
1252 		prison_free(pr);
1253 		return (EEXIST);
1254 	}
1255 
1256 	/* Make sure the VNET is stable. */
1257 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1258 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1259 	if (shutdown) {
1260 		CURVNET_RESTORE();
1261 		prison_free(pr);
1262 		return (EBUSY);
1263 	}
1264 
1265 	/* Get interface back from child jail/vnet. */
1266 	if_vmove(ifp, vnet_dst);
1267 	CURVNET_RESTORE();
1268 
1269 	/* Report the new if_xname back to the userland. */
1270 	sprintf(ifname, "%s", ifp->if_xname);
1271 
1272 	prison_free(pr);
1273 	return (0);
1274 }
1275 #endif /* VIMAGE */
1276 
1277 /*
1278  * Add a group to an interface
1279  */
1280 int
1281 if_addgroup(struct ifnet *ifp, const char *groupname)
1282 {
1283 	struct ifg_list		*ifgl;
1284 	struct ifg_group	*ifg = NULL;
1285 	struct ifg_member	*ifgm;
1286 	int 			 new = 0;
1287 
1288 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1289 	    groupname[strlen(groupname) - 1] <= '9')
1290 		return (EINVAL);
1291 
1292 	IFNET_WLOCK();
1293 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1294 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1295 			IFNET_WUNLOCK();
1296 			return (EEXIST);
1297 		}
1298 
1299 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1300 	    M_NOWAIT)) == NULL) {
1301 	    	IFNET_WUNLOCK();
1302 		return (ENOMEM);
1303 	}
1304 
1305 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1306 	    M_TEMP, M_NOWAIT)) == NULL) {
1307 		free(ifgl, M_TEMP);
1308 		IFNET_WUNLOCK();
1309 		return (ENOMEM);
1310 	}
1311 
1312 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1313 		if (!strcmp(ifg->ifg_group, groupname))
1314 			break;
1315 
1316 	if (ifg == NULL) {
1317 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1318 		    M_TEMP, M_NOWAIT)) == NULL) {
1319 			free(ifgl, M_TEMP);
1320 			free(ifgm, M_TEMP);
1321 			IFNET_WUNLOCK();
1322 			return (ENOMEM);
1323 		}
1324 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1325 		ifg->ifg_refcnt = 0;
1326 		TAILQ_INIT(&ifg->ifg_members);
1327 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1328 		new = 1;
1329 	}
1330 
1331 	ifg->ifg_refcnt++;
1332 	ifgl->ifgl_group = ifg;
1333 	ifgm->ifgm_ifp = ifp;
1334 
1335 	IF_ADDR_WLOCK(ifp);
1336 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1337 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1338 	IF_ADDR_WUNLOCK(ifp);
1339 
1340 	IFNET_WUNLOCK();
1341 
1342 	if (new)
1343 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1344 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1345 
1346 	return (0);
1347 }
1348 
1349 /*
1350  * Remove a group from an interface
1351  */
1352 int
1353 if_delgroup(struct ifnet *ifp, const char *groupname)
1354 {
1355 	struct ifg_list		*ifgl;
1356 	struct ifg_member	*ifgm;
1357 
1358 	IFNET_WLOCK();
1359 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1360 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1361 			break;
1362 	if (ifgl == NULL) {
1363 		IFNET_WUNLOCK();
1364 		return (ENOENT);
1365 	}
1366 
1367 	IF_ADDR_WLOCK(ifp);
1368 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1369 	IF_ADDR_WUNLOCK(ifp);
1370 
1371 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1372 		if (ifgm->ifgm_ifp == ifp)
1373 			break;
1374 
1375 	if (ifgm != NULL) {
1376 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1377 		free(ifgm, M_TEMP);
1378 	}
1379 
1380 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1381 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1382 		IFNET_WUNLOCK();
1383 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1384 		free(ifgl->ifgl_group, M_TEMP);
1385 	} else
1386 		IFNET_WUNLOCK();
1387 
1388 	free(ifgl, M_TEMP);
1389 
1390 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1391 
1392 	return (0);
1393 }
1394 
1395 /*
1396  * Remove an interface from all groups
1397  */
1398 static void
1399 if_delgroups(struct ifnet *ifp)
1400 {
1401 	struct ifg_list		*ifgl;
1402 	struct ifg_member	*ifgm;
1403 	char groupname[IFNAMSIZ];
1404 
1405 	IFNET_WLOCK();
1406 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1407 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1408 
1409 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1410 
1411 		IF_ADDR_WLOCK(ifp);
1412 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1413 		IF_ADDR_WUNLOCK(ifp);
1414 
1415 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1416 			if (ifgm->ifgm_ifp == ifp)
1417 				break;
1418 
1419 		if (ifgm != NULL) {
1420 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1421 			    ifgm_next);
1422 			free(ifgm, M_TEMP);
1423 		}
1424 
1425 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1426 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1427 			IFNET_WUNLOCK();
1428 			EVENTHANDLER_INVOKE(group_detach_event,
1429 			    ifgl->ifgl_group);
1430 			free(ifgl->ifgl_group, M_TEMP);
1431 		} else
1432 			IFNET_WUNLOCK();
1433 
1434 		free(ifgl, M_TEMP);
1435 
1436 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1437 
1438 		IFNET_WLOCK();
1439 	}
1440 	IFNET_WUNLOCK();
1441 }
1442 
1443 /*
1444  * Stores all groups from an interface in memory pointed
1445  * to by data
1446  */
1447 static int
1448 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1449 {
1450 	int			 len, error;
1451 	struct ifg_list		*ifgl;
1452 	struct ifg_req		 ifgrq, *ifgp;
1453 	struct ifgroupreq	*ifgr = data;
1454 
1455 	if (ifgr->ifgr_len == 0) {
1456 		IF_ADDR_RLOCK(ifp);
1457 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1458 			ifgr->ifgr_len += sizeof(struct ifg_req);
1459 		IF_ADDR_RUNLOCK(ifp);
1460 		return (0);
1461 	}
1462 
1463 	len = ifgr->ifgr_len;
1464 	ifgp = ifgr->ifgr_groups;
1465 	/* XXX: wire */
1466 	IF_ADDR_RLOCK(ifp);
1467 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1468 		if (len < sizeof(ifgrq)) {
1469 			IF_ADDR_RUNLOCK(ifp);
1470 			return (EINVAL);
1471 		}
1472 		bzero(&ifgrq, sizeof ifgrq);
1473 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1474 		    sizeof(ifgrq.ifgrq_group));
1475 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1476 		    	IF_ADDR_RUNLOCK(ifp);
1477 			return (error);
1478 		}
1479 		len -= sizeof(ifgrq);
1480 		ifgp++;
1481 	}
1482 	IF_ADDR_RUNLOCK(ifp);
1483 
1484 	return (0);
1485 }
1486 
1487 /*
1488  * Stores all members of a group in memory pointed to by data
1489  */
1490 static int
1491 if_getgroupmembers(struct ifgroupreq *data)
1492 {
1493 	struct ifgroupreq	*ifgr = data;
1494 	struct ifg_group	*ifg;
1495 	struct ifg_member	*ifgm;
1496 	struct ifg_req		 ifgrq, *ifgp;
1497 	int			 len, error;
1498 
1499 	IFNET_RLOCK();
1500 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1501 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1502 			break;
1503 	if (ifg == NULL) {
1504 		IFNET_RUNLOCK();
1505 		return (ENOENT);
1506 	}
1507 
1508 	if (ifgr->ifgr_len == 0) {
1509 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1510 			ifgr->ifgr_len += sizeof(ifgrq);
1511 		IFNET_RUNLOCK();
1512 		return (0);
1513 	}
1514 
1515 	len = ifgr->ifgr_len;
1516 	ifgp = ifgr->ifgr_groups;
1517 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1518 		if (len < sizeof(ifgrq)) {
1519 			IFNET_RUNLOCK();
1520 			return (EINVAL);
1521 		}
1522 		bzero(&ifgrq, sizeof ifgrq);
1523 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1524 		    sizeof(ifgrq.ifgrq_member));
1525 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1526 			IFNET_RUNLOCK();
1527 			return (error);
1528 		}
1529 		len -= sizeof(ifgrq);
1530 		ifgp++;
1531 	}
1532 	IFNET_RUNLOCK();
1533 
1534 	return (0);
1535 }
1536 
1537 /*
1538  * Return counter values from counter(9)s stored in ifnet.
1539  */
1540 uint64_t
1541 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1542 {
1543 
1544 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1545 
1546 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1547 }
1548 
1549 /*
1550  * Increase an ifnet counter. Usually used for counters shared
1551  * between the stack and a driver, but function supports them all.
1552  */
1553 void
1554 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1555 {
1556 
1557 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1558 
1559 	counter_u64_add(ifp->if_counters[cnt], inc);
1560 }
1561 
1562 /*
1563  * Copy data from ifnet to userland API structure if_data.
1564  */
1565 void
1566 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1567 {
1568 
1569 	ifd->ifi_type = ifp->if_type;
1570 	ifd->ifi_physical = 0;
1571 	ifd->ifi_addrlen = ifp->if_addrlen;
1572 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1573 	ifd->ifi_link_state = ifp->if_link_state;
1574 	ifd->ifi_vhid = 0;
1575 	ifd->ifi_datalen = sizeof(struct if_data);
1576 	ifd->ifi_mtu = ifp->if_mtu;
1577 	ifd->ifi_metric = ifp->if_metric;
1578 	ifd->ifi_baudrate = ifp->if_baudrate;
1579 	ifd->ifi_hwassist = ifp->if_hwassist;
1580 	ifd->ifi_epoch = ifp->if_epoch;
1581 	ifd->ifi_lastchange = ifp->if_lastchange;
1582 
1583 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1584 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1585 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1586 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1587 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1588 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1589 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1590 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1591 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1592 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1593 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1594 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1595 }
1596 
1597 /*
1598  * Wrapper functions for struct ifnet address list locking macros.  These are
1599  * used by kernel modules to avoid encoding programming interface or binary
1600  * interface assumptions that may be violated when kernel-internal locking
1601  * approaches change.
1602  */
1603 void
1604 if_addr_rlock(struct ifnet *ifp)
1605 {
1606 
1607 	IF_ADDR_RLOCK(ifp);
1608 }
1609 
1610 void
1611 if_addr_runlock(struct ifnet *ifp)
1612 {
1613 
1614 	IF_ADDR_RUNLOCK(ifp);
1615 }
1616 
1617 void
1618 if_maddr_rlock(if_t ifp)
1619 {
1620 
1621 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1622 }
1623 
1624 void
1625 if_maddr_runlock(if_t ifp)
1626 {
1627 
1628 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1629 }
1630 
1631 /*
1632  * Initialization, destruction and refcounting functions for ifaddrs.
1633  */
1634 struct ifaddr *
1635 ifa_alloc(size_t size, int flags)
1636 {
1637 	struct ifaddr *ifa;
1638 
1639 	KASSERT(size >= sizeof(struct ifaddr),
1640 	    ("%s: invalid size %zu", __func__, size));
1641 
1642 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1643 	if (ifa == NULL)
1644 		return (NULL);
1645 
1646 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1647 		goto fail;
1648 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1649 		goto fail;
1650 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1651 		goto fail;
1652 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1653 		goto fail;
1654 
1655 	refcount_init(&ifa->ifa_refcnt, 1);
1656 
1657 	return (ifa);
1658 
1659 fail:
1660 	/* free(NULL) is okay */
1661 	counter_u64_free(ifa->ifa_opackets);
1662 	counter_u64_free(ifa->ifa_ipackets);
1663 	counter_u64_free(ifa->ifa_obytes);
1664 	counter_u64_free(ifa->ifa_ibytes);
1665 	free(ifa, M_IFADDR);
1666 
1667 	return (NULL);
1668 }
1669 
1670 void
1671 ifa_ref(struct ifaddr *ifa)
1672 {
1673 
1674 	refcount_acquire(&ifa->ifa_refcnt);
1675 }
1676 
1677 void
1678 ifa_free(struct ifaddr *ifa)
1679 {
1680 
1681 	if (refcount_release(&ifa->ifa_refcnt)) {
1682 		counter_u64_free(ifa->ifa_opackets);
1683 		counter_u64_free(ifa->ifa_ipackets);
1684 		counter_u64_free(ifa->ifa_obytes);
1685 		counter_u64_free(ifa->ifa_ibytes);
1686 		free(ifa, M_IFADDR);
1687 	}
1688 }
1689 
1690 static int
1691 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1692     struct sockaddr *ia)
1693 {
1694 	int error;
1695 	struct rt_addrinfo info;
1696 	struct sockaddr_dl null_sdl;
1697 	struct ifnet *ifp;
1698 
1699 	ifp = ifa->ifa_ifp;
1700 
1701 	bzero(&info, sizeof(info));
1702 	if (cmd != RTM_DELETE)
1703 		info.rti_ifp = V_loif;
1704 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1705 	info.rti_info[RTAX_DST] = ia;
1706 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1707 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1708 
1709 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1710 
1711 	if (error != 0)
1712 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1713 		    __func__, otype, if_name(ifp), error);
1714 
1715 	return (error);
1716 }
1717 
1718 int
1719 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1720 {
1721 
1722 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1723 }
1724 
1725 int
1726 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1727 {
1728 
1729 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1730 }
1731 
1732 int
1733 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1734 {
1735 
1736 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1737 }
1738 
1739 /*
1740  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1741  * structs used to represent other address families, it is necessary
1742  * to perform a different comparison.
1743  */
1744 
1745 #define	sa_dl_equal(a1, a2)	\
1746 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1747 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1748 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1749 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1750 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1751 
1752 /*
1753  * Locate an interface based on a complete address.
1754  */
1755 /*ARGSUSED*/
1756 static struct ifaddr *
1757 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1758 {
1759 	struct ifnet *ifp;
1760 	struct ifaddr *ifa;
1761 
1762 	IFNET_RLOCK_NOSLEEP();
1763 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1764 		IF_ADDR_RLOCK(ifp);
1765 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1766 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1767 				continue;
1768 			if (sa_equal(addr, ifa->ifa_addr)) {
1769 				if (getref)
1770 					ifa_ref(ifa);
1771 				IF_ADDR_RUNLOCK(ifp);
1772 				goto done;
1773 			}
1774 			/* IP6 doesn't have broadcast */
1775 			if ((ifp->if_flags & IFF_BROADCAST) &&
1776 			    ifa->ifa_broadaddr &&
1777 			    ifa->ifa_broadaddr->sa_len != 0 &&
1778 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1779 				if (getref)
1780 					ifa_ref(ifa);
1781 				IF_ADDR_RUNLOCK(ifp);
1782 				goto done;
1783 			}
1784 		}
1785 		IF_ADDR_RUNLOCK(ifp);
1786 	}
1787 	ifa = NULL;
1788 done:
1789 	IFNET_RUNLOCK_NOSLEEP();
1790 	return (ifa);
1791 }
1792 
1793 struct ifaddr *
1794 ifa_ifwithaddr(const struct sockaddr *addr)
1795 {
1796 
1797 	return (ifa_ifwithaddr_internal(addr, 1));
1798 }
1799 
1800 int
1801 ifa_ifwithaddr_check(const struct sockaddr *addr)
1802 {
1803 
1804 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1805 }
1806 
1807 /*
1808  * Locate an interface based on the broadcast address.
1809  */
1810 /* ARGSUSED */
1811 struct ifaddr *
1812 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1813 {
1814 	struct ifnet *ifp;
1815 	struct ifaddr *ifa;
1816 
1817 	IFNET_RLOCK_NOSLEEP();
1818 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1819 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1820 			continue;
1821 		IF_ADDR_RLOCK(ifp);
1822 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1823 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1824 				continue;
1825 			if ((ifp->if_flags & IFF_BROADCAST) &&
1826 			    ifa->ifa_broadaddr &&
1827 			    ifa->ifa_broadaddr->sa_len != 0 &&
1828 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1829 				ifa_ref(ifa);
1830 				IF_ADDR_RUNLOCK(ifp);
1831 				goto done;
1832 			}
1833 		}
1834 		IF_ADDR_RUNLOCK(ifp);
1835 	}
1836 	ifa = NULL;
1837 done:
1838 	IFNET_RUNLOCK_NOSLEEP();
1839 	return (ifa);
1840 }
1841 
1842 /*
1843  * Locate the point to point interface with a given destination address.
1844  */
1845 /*ARGSUSED*/
1846 struct ifaddr *
1847 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1848 {
1849 	struct ifnet *ifp;
1850 	struct ifaddr *ifa;
1851 
1852 	IFNET_RLOCK_NOSLEEP();
1853 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1854 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1855 			continue;
1856 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1857 			continue;
1858 		IF_ADDR_RLOCK(ifp);
1859 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1860 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1861 				continue;
1862 			if (ifa->ifa_dstaddr != NULL &&
1863 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1864 				ifa_ref(ifa);
1865 				IF_ADDR_RUNLOCK(ifp);
1866 				goto done;
1867 			}
1868 		}
1869 		IF_ADDR_RUNLOCK(ifp);
1870 	}
1871 	ifa = NULL;
1872 done:
1873 	IFNET_RUNLOCK_NOSLEEP();
1874 	return (ifa);
1875 }
1876 
1877 /*
1878  * Find an interface on a specific network.  If many, choice
1879  * is most specific found.
1880  */
1881 struct ifaddr *
1882 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1883 {
1884 	struct ifnet *ifp;
1885 	struct ifaddr *ifa;
1886 	struct ifaddr *ifa_maybe = NULL;
1887 	u_int af = addr->sa_family;
1888 	const char *addr_data = addr->sa_data, *cplim;
1889 
1890 	/*
1891 	 * AF_LINK addresses can be looked up directly by their index number,
1892 	 * so do that if we can.
1893 	 */
1894 	if (af == AF_LINK) {
1895 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1896 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1897 		return (ifaddr_byindex(sdl->sdl_index));
1898 	}
1899 
1900 	/*
1901 	 * Scan though each interface, looking for ones that have addresses
1902 	 * in this address family and the requested fib.  Maintain a reference
1903 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1904 	 * kept it stable when we move onto the next interface.
1905 	 */
1906 	IFNET_RLOCK_NOSLEEP();
1907 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1908 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1909 			continue;
1910 		IF_ADDR_RLOCK(ifp);
1911 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1912 			const char *cp, *cp2, *cp3;
1913 
1914 			if (ifa->ifa_addr->sa_family != af)
1915 next:				continue;
1916 			if (af == AF_INET &&
1917 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1918 				/*
1919 				 * This is a bit broken as it doesn't
1920 				 * take into account that the remote end may
1921 				 * be a single node in the network we are
1922 				 * looking for.
1923 				 * The trouble is that we don't know the
1924 				 * netmask for the remote end.
1925 				 */
1926 				if (ifa->ifa_dstaddr != NULL &&
1927 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1928 					ifa_ref(ifa);
1929 					IF_ADDR_RUNLOCK(ifp);
1930 					goto done;
1931 				}
1932 			} else {
1933 				/*
1934 				 * Scan all the bits in the ifa's address.
1935 				 * If a bit dissagrees with what we are
1936 				 * looking for, mask it with the netmask
1937 				 * to see if it really matters.
1938 				 * (A byte at a time)
1939 				 */
1940 				if (ifa->ifa_netmask == 0)
1941 					continue;
1942 				cp = addr_data;
1943 				cp2 = ifa->ifa_addr->sa_data;
1944 				cp3 = ifa->ifa_netmask->sa_data;
1945 				cplim = ifa->ifa_netmask->sa_len
1946 					+ (char *)ifa->ifa_netmask;
1947 				while (cp3 < cplim)
1948 					if ((*cp++ ^ *cp2++) & *cp3++)
1949 						goto next; /* next address! */
1950 				/*
1951 				 * If the netmask of what we just found
1952 				 * is more specific than what we had before
1953 				 * (if we had one), or if the virtual status
1954 				 * of new prefix is better than of the old one,
1955 				 * then remember the new one before continuing
1956 				 * to search for an even better one.
1957 				 */
1958 				if (ifa_maybe == NULL ||
1959 				    ifa_preferred(ifa_maybe, ifa) ||
1960 				    rn_refines((caddr_t)ifa->ifa_netmask,
1961 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1962 					if (ifa_maybe != NULL)
1963 						ifa_free(ifa_maybe);
1964 					ifa_maybe = ifa;
1965 					ifa_ref(ifa_maybe);
1966 				}
1967 			}
1968 		}
1969 		IF_ADDR_RUNLOCK(ifp);
1970 	}
1971 	ifa = ifa_maybe;
1972 	ifa_maybe = NULL;
1973 done:
1974 	IFNET_RUNLOCK_NOSLEEP();
1975 	if (ifa_maybe != NULL)
1976 		ifa_free(ifa_maybe);
1977 	return (ifa);
1978 }
1979 
1980 /*
1981  * Find an interface address specific to an interface best matching
1982  * a given address.
1983  */
1984 struct ifaddr *
1985 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1986 {
1987 	struct ifaddr *ifa;
1988 	const char *cp, *cp2, *cp3;
1989 	char *cplim;
1990 	struct ifaddr *ifa_maybe = NULL;
1991 	u_int af = addr->sa_family;
1992 
1993 	if (af >= AF_MAX)
1994 		return (NULL);
1995 	IF_ADDR_RLOCK(ifp);
1996 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1997 		if (ifa->ifa_addr->sa_family != af)
1998 			continue;
1999 		if (ifa_maybe == NULL)
2000 			ifa_maybe = ifa;
2001 		if (ifa->ifa_netmask == 0) {
2002 			if (sa_equal(addr, ifa->ifa_addr) ||
2003 			    (ifa->ifa_dstaddr &&
2004 			    sa_equal(addr, ifa->ifa_dstaddr)))
2005 				goto done;
2006 			continue;
2007 		}
2008 		if (ifp->if_flags & IFF_POINTOPOINT) {
2009 			if (sa_equal(addr, ifa->ifa_dstaddr))
2010 				goto done;
2011 		} else {
2012 			cp = addr->sa_data;
2013 			cp2 = ifa->ifa_addr->sa_data;
2014 			cp3 = ifa->ifa_netmask->sa_data;
2015 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2016 			for (; cp3 < cplim; cp3++)
2017 				if ((*cp++ ^ *cp2++) & *cp3)
2018 					break;
2019 			if (cp3 == cplim)
2020 				goto done;
2021 		}
2022 	}
2023 	ifa = ifa_maybe;
2024 done:
2025 	if (ifa != NULL)
2026 		ifa_ref(ifa);
2027 	IF_ADDR_RUNLOCK(ifp);
2028 	return (ifa);
2029 }
2030 
2031 /*
2032  * See whether new ifa is better than current one:
2033  * 1) A non-virtual one is preferred over virtual.
2034  * 2) A virtual in master state preferred over any other state.
2035  *
2036  * Used in several address selecting functions.
2037  */
2038 int
2039 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2040 {
2041 
2042 	return (cur->ifa_carp && (!next->ifa_carp ||
2043 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2044 }
2045 
2046 #include <net/if_llatbl.h>
2047 
2048 /*
2049  * Default action when installing a route with a Link Level gateway.
2050  * Lookup an appropriate real ifa to point to.
2051  * This should be moved to /sys/net/link.c eventually.
2052  */
2053 static void
2054 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2055 {
2056 	struct ifaddr *ifa, *oifa;
2057 	struct sockaddr *dst;
2058 	struct ifnet *ifp;
2059 
2060 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2061 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2062 		return;
2063 	ifa = ifaof_ifpforaddr(dst, ifp);
2064 	if (ifa) {
2065 		oifa = rt->rt_ifa;
2066 		rt->rt_ifa = ifa;
2067 		ifa_free(oifa);
2068 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2069 			ifa->ifa_rtrequest(cmd, rt, info);
2070 	}
2071 }
2072 
2073 struct sockaddr_dl *
2074 link_alloc_sdl(size_t size, int flags)
2075 {
2076 
2077 	return (malloc(size, M_TEMP, flags));
2078 }
2079 
2080 void
2081 link_free_sdl(struct sockaddr *sa)
2082 {
2083 	free(sa, M_TEMP);
2084 }
2085 
2086 /*
2087  * Fills in given sdl with interface basic info.
2088  * Returns pointer to filled sdl.
2089  */
2090 struct sockaddr_dl *
2091 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2092 {
2093 	struct sockaddr_dl *sdl;
2094 
2095 	sdl = (struct sockaddr_dl *)paddr;
2096 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2097 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2098 	sdl->sdl_family = AF_LINK;
2099 	sdl->sdl_index = ifp->if_index;
2100 	sdl->sdl_type = iftype;
2101 
2102 	return (sdl);
2103 }
2104 
2105 /*
2106  * Mark an interface down and notify protocols of
2107  * the transition.
2108  */
2109 static void
2110 if_unroute(struct ifnet *ifp, int flag, int fam)
2111 {
2112 	struct ifaddr *ifa;
2113 
2114 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2115 
2116 	ifp->if_flags &= ~flag;
2117 	getmicrotime(&ifp->if_lastchange);
2118 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2119 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2120 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2121 	ifp->if_qflush(ifp);
2122 
2123 	if (ifp->if_carp)
2124 		(*carp_linkstate_p)(ifp);
2125 	rt_ifmsg(ifp);
2126 }
2127 
2128 /*
2129  * Mark an interface up and notify protocols of
2130  * the transition.
2131  */
2132 static void
2133 if_route(struct ifnet *ifp, int flag, int fam)
2134 {
2135 	struct ifaddr *ifa;
2136 
2137 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2138 
2139 	ifp->if_flags |= flag;
2140 	getmicrotime(&ifp->if_lastchange);
2141 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2142 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2143 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2144 	if (ifp->if_carp)
2145 		(*carp_linkstate_p)(ifp);
2146 	rt_ifmsg(ifp);
2147 #ifdef INET6
2148 	in6_if_up(ifp);
2149 #endif
2150 }
2151 
2152 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2153 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2154 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2155 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2156 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2157 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2158 void	*(*vlan_cookie_p)(struct ifnet *);
2159 
2160 /*
2161  * Handle a change in the interface link state. To avoid LORs
2162  * between driver lock and upper layer locks, as well as possible
2163  * recursions, we post event to taskqueue, and all job
2164  * is done in static do_link_state_change().
2165  */
2166 void
2167 if_link_state_change(struct ifnet *ifp, int link_state)
2168 {
2169 	/* Return if state hasn't changed. */
2170 	if (ifp->if_link_state == link_state)
2171 		return;
2172 
2173 	ifp->if_link_state = link_state;
2174 
2175 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2176 }
2177 
2178 static void
2179 do_link_state_change(void *arg, int pending)
2180 {
2181 	struct ifnet *ifp = (struct ifnet *)arg;
2182 	int link_state = ifp->if_link_state;
2183 	CURVNET_SET(ifp->if_vnet);
2184 
2185 	/* Notify that the link state has changed. */
2186 	rt_ifmsg(ifp);
2187 	if (ifp->if_vlantrunk != NULL)
2188 		(*vlan_link_state_p)(ifp);
2189 
2190 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2191 	    ifp->if_l2com != NULL)
2192 		(*ng_ether_link_state_p)(ifp, link_state);
2193 	if (ifp->if_carp)
2194 		(*carp_linkstate_p)(ifp);
2195 	if (ifp->if_bridge)
2196 		(*bridge_linkstate_p)(ifp);
2197 	if (ifp->if_lagg)
2198 		(*lagg_linkstate_p)(ifp, link_state);
2199 
2200 	if (IS_DEFAULT_VNET(curvnet))
2201 		devctl_notify("IFNET", ifp->if_xname,
2202 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2203 		    NULL);
2204 	if (pending > 1)
2205 		if_printf(ifp, "%d link states coalesced\n", pending);
2206 	if (log_link_state_change)
2207 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2208 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2209 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2210 	CURVNET_RESTORE();
2211 }
2212 
2213 /*
2214  * Mark an interface down and notify protocols of
2215  * the transition.
2216  */
2217 void
2218 if_down(struct ifnet *ifp)
2219 {
2220 
2221 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2222 }
2223 
2224 /*
2225  * Mark an interface up and notify protocols of
2226  * the transition.
2227  */
2228 void
2229 if_up(struct ifnet *ifp)
2230 {
2231 
2232 	if_route(ifp, IFF_UP, AF_UNSPEC);
2233 }
2234 
2235 /*
2236  * Flush an interface queue.
2237  */
2238 void
2239 if_qflush(struct ifnet *ifp)
2240 {
2241 	struct mbuf *m, *n;
2242 	struct ifaltq *ifq;
2243 
2244 	ifq = &ifp->if_snd;
2245 	IFQ_LOCK(ifq);
2246 #ifdef ALTQ
2247 	if (ALTQ_IS_ENABLED(ifq))
2248 		ALTQ_PURGE(ifq);
2249 #endif
2250 	n = ifq->ifq_head;
2251 	while ((m = n) != NULL) {
2252 		n = m->m_nextpkt;
2253 		m_freem(m);
2254 	}
2255 	ifq->ifq_head = 0;
2256 	ifq->ifq_tail = 0;
2257 	ifq->ifq_len = 0;
2258 	IFQ_UNLOCK(ifq);
2259 }
2260 
2261 /*
2262  * Map interface name to interface structure pointer, with or without
2263  * returning a reference.
2264  */
2265 struct ifnet *
2266 ifunit_ref(const char *name)
2267 {
2268 	struct ifnet *ifp;
2269 
2270 	IFNET_RLOCK_NOSLEEP();
2271 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2272 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2273 		    !(ifp->if_flags & IFF_DYING))
2274 			break;
2275 	}
2276 	if (ifp != NULL)
2277 		if_ref(ifp);
2278 	IFNET_RUNLOCK_NOSLEEP();
2279 	return (ifp);
2280 }
2281 
2282 struct ifnet *
2283 ifunit(const char *name)
2284 {
2285 	struct ifnet *ifp;
2286 
2287 	IFNET_RLOCK_NOSLEEP();
2288 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2289 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2290 			break;
2291 	}
2292 	IFNET_RUNLOCK_NOSLEEP();
2293 	return (ifp);
2294 }
2295 
2296 /*
2297  * Hardware specific interface ioctls.
2298  */
2299 static int
2300 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2301 {
2302 	struct ifreq *ifr;
2303 	int error = 0;
2304 	int new_flags, temp_flags;
2305 	size_t namelen, onamelen;
2306 	size_t descrlen;
2307 	char *descrbuf, *odescrbuf;
2308 	char new_name[IFNAMSIZ];
2309 	struct ifaddr *ifa;
2310 	struct sockaddr_dl *sdl;
2311 
2312 	ifr = (struct ifreq *)data;
2313 	switch (cmd) {
2314 	case SIOCGIFINDEX:
2315 		ifr->ifr_index = ifp->if_index;
2316 		break;
2317 
2318 	case SIOCGIFFLAGS:
2319 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2320 		ifr->ifr_flags = temp_flags & 0xffff;
2321 		ifr->ifr_flagshigh = temp_flags >> 16;
2322 		break;
2323 
2324 	case SIOCGIFCAP:
2325 		ifr->ifr_reqcap = ifp->if_capabilities;
2326 		ifr->ifr_curcap = ifp->if_capenable;
2327 		break;
2328 
2329 #ifdef MAC
2330 	case SIOCGIFMAC:
2331 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2332 		break;
2333 #endif
2334 
2335 	case SIOCGIFMETRIC:
2336 		ifr->ifr_metric = ifp->if_metric;
2337 		break;
2338 
2339 	case SIOCGIFMTU:
2340 		ifr->ifr_mtu = ifp->if_mtu;
2341 		break;
2342 
2343 	case SIOCGIFPHYS:
2344 		/* XXXGL: did this ever worked? */
2345 		ifr->ifr_phys = 0;
2346 		break;
2347 
2348 	case SIOCGIFDESCR:
2349 		error = 0;
2350 		sx_slock(&ifdescr_sx);
2351 		if (ifp->if_description == NULL)
2352 			error = ENOMSG;
2353 		else {
2354 			/* space for terminating nul */
2355 			descrlen = strlen(ifp->if_description) + 1;
2356 			if (ifr->ifr_buffer.length < descrlen)
2357 				ifr->ifr_buffer.buffer = NULL;
2358 			else
2359 				error = copyout(ifp->if_description,
2360 				    ifr->ifr_buffer.buffer, descrlen);
2361 			ifr->ifr_buffer.length = descrlen;
2362 		}
2363 		sx_sunlock(&ifdescr_sx);
2364 		break;
2365 
2366 	case SIOCSIFDESCR:
2367 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2368 		if (error)
2369 			return (error);
2370 
2371 		/*
2372 		 * Copy only (length-1) bytes to make sure that
2373 		 * if_description is always nul terminated.  The
2374 		 * length parameter is supposed to count the
2375 		 * terminating nul in.
2376 		 */
2377 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2378 			return (ENAMETOOLONG);
2379 		else if (ifr->ifr_buffer.length == 0)
2380 			descrbuf = NULL;
2381 		else {
2382 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2383 			    M_WAITOK | M_ZERO);
2384 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2385 			    ifr->ifr_buffer.length - 1);
2386 			if (error) {
2387 				free(descrbuf, M_IFDESCR);
2388 				break;
2389 			}
2390 		}
2391 
2392 		sx_xlock(&ifdescr_sx);
2393 		odescrbuf = ifp->if_description;
2394 		ifp->if_description = descrbuf;
2395 		sx_xunlock(&ifdescr_sx);
2396 
2397 		getmicrotime(&ifp->if_lastchange);
2398 		free(odescrbuf, M_IFDESCR);
2399 		break;
2400 
2401 	case SIOCGIFFIB:
2402 		ifr->ifr_fib = ifp->if_fib;
2403 		break;
2404 
2405 	case SIOCSIFFIB:
2406 		error = priv_check(td, PRIV_NET_SETIFFIB);
2407 		if (error)
2408 			return (error);
2409 		if (ifr->ifr_fib >= rt_numfibs)
2410 			return (EINVAL);
2411 
2412 		ifp->if_fib = ifr->ifr_fib;
2413 		break;
2414 
2415 	case SIOCSIFFLAGS:
2416 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2417 		if (error)
2418 			return (error);
2419 		/*
2420 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2421 		 * check, so we don't need special handling here yet.
2422 		 */
2423 		new_flags = (ifr->ifr_flags & 0xffff) |
2424 		    (ifr->ifr_flagshigh << 16);
2425 		if (ifp->if_flags & IFF_UP &&
2426 		    (new_flags & IFF_UP) == 0) {
2427 			if_down(ifp);
2428 		} else if (new_flags & IFF_UP &&
2429 		    (ifp->if_flags & IFF_UP) == 0) {
2430 			if_up(ifp);
2431 		}
2432 		/* See if permanently promiscuous mode bit is about to flip */
2433 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2434 			if (new_flags & IFF_PPROMISC)
2435 				ifp->if_flags |= IFF_PROMISC;
2436 			else if (ifp->if_pcount == 0)
2437 				ifp->if_flags &= ~IFF_PROMISC;
2438 			if (log_promisc_mode_change)
2439                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2440                                     ifp->if_xname,
2441                                     ((new_flags & IFF_PPROMISC) ?
2442                                      "enabled" : "disabled"));
2443 		}
2444 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2445 			(new_flags &~ IFF_CANTCHANGE);
2446 		if (ifp->if_ioctl) {
2447 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2448 		}
2449 		getmicrotime(&ifp->if_lastchange);
2450 		break;
2451 
2452 	case SIOCSIFCAP:
2453 		error = priv_check(td, PRIV_NET_SETIFCAP);
2454 		if (error)
2455 			return (error);
2456 		if (ifp->if_ioctl == NULL)
2457 			return (EOPNOTSUPP);
2458 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2459 			return (EINVAL);
2460 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2461 		if (error == 0)
2462 			getmicrotime(&ifp->if_lastchange);
2463 		break;
2464 
2465 #ifdef MAC
2466 	case SIOCSIFMAC:
2467 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2468 		break;
2469 #endif
2470 
2471 	case SIOCSIFNAME:
2472 		error = priv_check(td, PRIV_NET_SETIFNAME);
2473 		if (error)
2474 			return (error);
2475 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2476 		if (error != 0)
2477 			return (error);
2478 		if (new_name[0] == '\0')
2479 			return (EINVAL);
2480 		if (new_name[IFNAMSIZ-1] != '\0') {
2481 			new_name[IFNAMSIZ-1] = '\0';
2482 			if (strlen(new_name) == IFNAMSIZ-1)
2483 				return (EINVAL);
2484 		}
2485 		if (ifunit(new_name) != NULL)
2486 			return (EEXIST);
2487 
2488 		/*
2489 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2490 		 * and there are numerous other races with the
2491 		 * ifunit() checks not being atomic with namespace
2492 		 * changes (renames, vmoves, if_attach, etc).
2493 		 */
2494 		ifp->if_flags |= IFF_RENAMING;
2495 
2496 		/* Announce the departure of the interface. */
2497 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2498 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2499 
2500 		log(LOG_INFO, "%s: changing name to '%s'\n",
2501 		    ifp->if_xname, new_name);
2502 
2503 		IF_ADDR_WLOCK(ifp);
2504 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2505 		ifa = ifp->if_addr;
2506 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2507 		namelen = strlen(new_name);
2508 		onamelen = sdl->sdl_nlen;
2509 		/*
2510 		 * Move the address if needed.  This is safe because we
2511 		 * allocate space for a name of length IFNAMSIZ when we
2512 		 * create this in if_attach().
2513 		 */
2514 		if (namelen != onamelen) {
2515 			bcopy(sdl->sdl_data + onamelen,
2516 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2517 		}
2518 		bcopy(new_name, sdl->sdl_data, namelen);
2519 		sdl->sdl_nlen = namelen;
2520 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2521 		bzero(sdl->sdl_data, onamelen);
2522 		while (namelen != 0)
2523 			sdl->sdl_data[--namelen] = 0xff;
2524 		IF_ADDR_WUNLOCK(ifp);
2525 
2526 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2527 		/* Announce the return of the interface. */
2528 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2529 
2530 		ifp->if_flags &= ~IFF_RENAMING;
2531 		break;
2532 
2533 #ifdef VIMAGE
2534 	case SIOCSIFVNET:
2535 		error = priv_check(td, PRIV_NET_SETIFVNET);
2536 		if (error)
2537 			return (error);
2538 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2539 		break;
2540 #endif
2541 
2542 	case SIOCSIFMETRIC:
2543 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2544 		if (error)
2545 			return (error);
2546 		ifp->if_metric = ifr->ifr_metric;
2547 		getmicrotime(&ifp->if_lastchange);
2548 		break;
2549 
2550 	case SIOCSIFPHYS:
2551 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2552 		if (error)
2553 			return (error);
2554 		if (ifp->if_ioctl == NULL)
2555 			return (EOPNOTSUPP);
2556 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2557 		if (error == 0)
2558 			getmicrotime(&ifp->if_lastchange);
2559 		break;
2560 
2561 	case SIOCSIFMTU:
2562 	{
2563 		u_long oldmtu = ifp->if_mtu;
2564 
2565 		error = priv_check(td, PRIV_NET_SETIFMTU);
2566 		if (error)
2567 			return (error);
2568 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2569 			return (EINVAL);
2570 		if (ifp->if_ioctl == NULL)
2571 			return (EOPNOTSUPP);
2572 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2573 		if (error == 0) {
2574 			getmicrotime(&ifp->if_lastchange);
2575 			rt_ifmsg(ifp);
2576 		}
2577 		/*
2578 		 * If the link MTU changed, do network layer specific procedure.
2579 		 */
2580 		if (ifp->if_mtu != oldmtu) {
2581 #ifdef INET6
2582 			nd6_setmtu(ifp);
2583 #endif
2584 			rt_updatemtu(ifp);
2585 		}
2586 		break;
2587 	}
2588 
2589 	case SIOCADDMULTI:
2590 	case SIOCDELMULTI:
2591 		if (cmd == SIOCADDMULTI)
2592 			error = priv_check(td, PRIV_NET_ADDMULTI);
2593 		else
2594 			error = priv_check(td, PRIV_NET_DELMULTI);
2595 		if (error)
2596 			return (error);
2597 
2598 		/* Don't allow group membership on non-multicast interfaces. */
2599 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2600 			return (EOPNOTSUPP);
2601 
2602 		/* Don't let users screw up protocols' entries. */
2603 		if (ifr->ifr_addr.sa_family != AF_LINK)
2604 			return (EINVAL);
2605 
2606 		if (cmd == SIOCADDMULTI) {
2607 			struct ifmultiaddr *ifma;
2608 
2609 			/*
2610 			 * Userland is only permitted to join groups once
2611 			 * via the if_addmulti() KPI, because it cannot hold
2612 			 * struct ifmultiaddr * between calls. It may also
2613 			 * lose a race while we check if the membership
2614 			 * already exists.
2615 			 */
2616 			IF_ADDR_RLOCK(ifp);
2617 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2618 			IF_ADDR_RUNLOCK(ifp);
2619 			if (ifma != NULL)
2620 				error = EADDRINUSE;
2621 			else
2622 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2623 		} else {
2624 			error = if_delmulti(ifp, &ifr->ifr_addr);
2625 		}
2626 		if (error == 0)
2627 			getmicrotime(&ifp->if_lastchange);
2628 		break;
2629 
2630 	case SIOCSIFPHYADDR:
2631 	case SIOCDIFPHYADDR:
2632 #ifdef INET6
2633 	case SIOCSIFPHYADDR_IN6:
2634 #endif
2635 	case SIOCSIFMEDIA:
2636 	case SIOCSIFGENERIC:
2637 		error = priv_check(td, PRIV_NET_HWIOCTL);
2638 		if (error)
2639 			return (error);
2640 		if (ifp->if_ioctl == NULL)
2641 			return (EOPNOTSUPP);
2642 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2643 		if (error == 0)
2644 			getmicrotime(&ifp->if_lastchange);
2645 		break;
2646 
2647 	case SIOCGIFSTATUS:
2648 	case SIOCGIFPSRCADDR:
2649 	case SIOCGIFPDSTADDR:
2650 	case SIOCGIFMEDIA:
2651 	case SIOCGIFXMEDIA:
2652 	case SIOCGIFGENERIC:
2653 		if (ifp->if_ioctl == NULL)
2654 			return (EOPNOTSUPP);
2655 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2656 		break;
2657 
2658 	case SIOCSIFLLADDR:
2659 		error = priv_check(td, PRIV_NET_SETLLADDR);
2660 		if (error)
2661 			return (error);
2662 		error = if_setlladdr(ifp,
2663 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2664 		break;
2665 
2666 	case SIOCAIFGROUP:
2667 	{
2668 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2669 
2670 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2671 		if (error)
2672 			return (error);
2673 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2674 			return (error);
2675 		break;
2676 	}
2677 
2678 	case SIOCGIFGROUP:
2679 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2680 			return (error);
2681 		break;
2682 
2683 	case SIOCDIFGROUP:
2684 	{
2685 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2686 
2687 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2688 		if (error)
2689 			return (error);
2690 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2691 			return (error);
2692 		break;
2693 	}
2694 
2695 	default:
2696 		error = ENOIOCTL;
2697 		break;
2698 	}
2699 	return (error);
2700 }
2701 
2702 /* COMPAT_SVR4 */
2703 #define	OSIOCGIFCONF	_IOWR('i', 20, struct ifconf)
2704 
2705 #ifdef COMPAT_FREEBSD32
2706 struct ifconf32 {
2707 	int32_t	ifc_len;
2708 	union {
2709 		uint32_t	ifcu_buf;
2710 		uint32_t	ifcu_req;
2711 	} ifc_ifcu;
2712 };
2713 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2714 #endif
2715 
2716 /*
2717  * Interface ioctls.
2718  */
2719 int
2720 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2721 {
2722 	struct ifnet *ifp;
2723 	struct ifreq *ifr;
2724 	int error;
2725 	int oif_flags;
2726 #ifdef VIMAGE
2727 	int shutdown;
2728 #endif
2729 
2730 	CURVNET_SET(so->so_vnet);
2731 #ifdef VIMAGE
2732 	/* Make sure the VNET is stable. */
2733 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2734 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2735 	if (shutdown) {
2736 		CURVNET_RESTORE();
2737 		return (EBUSY);
2738 	}
2739 #endif
2740 
2741 
2742 	switch (cmd) {
2743 	case SIOCGIFCONF:
2744 	case OSIOCGIFCONF:	/* COMPAT_SVR4 */
2745 		error = ifconf(cmd, data);
2746 		CURVNET_RESTORE();
2747 		return (error);
2748 
2749 #ifdef COMPAT_FREEBSD32
2750 	case SIOCGIFCONF32:
2751 		{
2752 			struct ifconf32 *ifc32;
2753 			struct ifconf ifc;
2754 
2755 			ifc32 = (struct ifconf32 *)data;
2756 			ifc.ifc_len = ifc32->ifc_len;
2757 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2758 
2759 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2760 			CURVNET_RESTORE();
2761 			if (error == 0)
2762 				ifc32->ifc_len = ifc.ifc_len;
2763 			return (error);
2764 		}
2765 #endif
2766 	}
2767 	ifr = (struct ifreq *)data;
2768 
2769 	switch (cmd) {
2770 #ifdef VIMAGE
2771 	case SIOCSIFRVNET:
2772 		error = priv_check(td, PRIV_NET_SETIFVNET);
2773 		if (error == 0)
2774 			error = if_vmove_reclaim(td, ifr->ifr_name,
2775 			    ifr->ifr_jid);
2776 		CURVNET_RESTORE();
2777 		return (error);
2778 #endif
2779 	case SIOCIFCREATE:
2780 	case SIOCIFCREATE2:
2781 		error = priv_check(td, PRIV_NET_IFCREATE);
2782 		if (error == 0)
2783 			error = if_clone_create(ifr->ifr_name,
2784 			    sizeof(ifr->ifr_name),
2785 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2786 		CURVNET_RESTORE();
2787 		return (error);
2788 	case SIOCIFDESTROY:
2789 		error = priv_check(td, PRIV_NET_IFDESTROY);
2790 		if (error == 0)
2791 			error = if_clone_destroy(ifr->ifr_name);
2792 		CURVNET_RESTORE();
2793 		return (error);
2794 
2795 	case SIOCIFGCLONERS:
2796 		error = if_clone_list((struct if_clonereq *)data);
2797 		CURVNET_RESTORE();
2798 		return (error);
2799 	case SIOCGIFGMEMB:
2800 		error = if_getgroupmembers((struct ifgroupreq *)data);
2801 		CURVNET_RESTORE();
2802 		return (error);
2803 #if defined(INET) || defined(INET6)
2804 	case SIOCSVH:
2805 	case SIOCGVH:
2806 		if (carp_ioctl_p == NULL)
2807 			error = EPROTONOSUPPORT;
2808 		else
2809 			error = (*carp_ioctl_p)(ifr, cmd, td);
2810 		CURVNET_RESTORE();
2811 		return (error);
2812 #endif
2813 	}
2814 
2815 	ifp = ifunit_ref(ifr->ifr_name);
2816 	if (ifp == NULL) {
2817 		CURVNET_RESTORE();
2818 		return (ENXIO);
2819 	}
2820 
2821 	error = ifhwioctl(cmd, ifp, data, td);
2822 	if (error != ENOIOCTL) {
2823 		if_rele(ifp);
2824 		CURVNET_RESTORE();
2825 		return (error);
2826 	}
2827 
2828 	oif_flags = ifp->if_flags;
2829 	if (so->so_proto == NULL) {
2830 		if_rele(ifp);
2831 		CURVNET_RESTORE();
2832 		return (EOPNOTSUPP);
2833 	}
2834 
2835 	/*
2836 	 * Pass the request on to the socket control method, and if the
2837 	 * latter returns EOPNOTSUPP, directly to the interface.
2838 	 *
2839 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2840 	 * trust SIOCSIFADDR et al to come from an already privileged
2841 	 * layer, and do not perform any credentials checks or input
2842 	 * validation.
2843 	 */
2844 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2845 	    ifp, td));
2846 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2847 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2848 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2849 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2850 
2851 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2852 #ifdef INET6
2853 		if (ifp->if_flags & IFF_UP)
2854 			in6_if_up(ifp);
2855 #endif
2856 	}
2857 	if_rele(ifp);
2858 	CURVNET_RESTORE();
2859 	return (error);
2860 }
2861 
2862 /*
2863  * The code common to handling reference counted flags,
2864  * e.g., in ifpromisc() and if_allmulti().
2865  * The "pflag" argument can specify a permanent mode flag to check,
2866  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2867  *
2868  * Only to be used on stack-owned flags, not driver-owned flags.
2869  */
2870 static int
2871 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2872 {
2873 	struct ifreq ifr;
2874 	int error;
2875 	int oldflags, oldcount;
2876 
2877 	/* Sanity checks to catch programming errors */
2878 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2879 	    ("%s: setting driver-owned flag %d", __func__, flag));
2880 
2881 	if (onswitch)
2882 		KASSERT(*refcount >= 0,
2883 		    ("%s: increment negative refcount %d for flag %d",
2884 		    __func__, *refcount, flag));
2885 	else
2886 		KASSERT(*refcount > 0,
2887 		    ("%s: decrement non-positive refcount %d for flag %d",
2888 		    __func__, *refcount, flag));
2889 
2890 	/* In case this mode is permanent, just touch refcount */
2891 	if (ifp->if_flags & pflag) {
2892 		*refcount += onswitch ? 1 : -1;
2893 		return (0);
2894 	}
2895 
2896 	/* Save ifnet parameters for if_ioctl() may fail */
2897 	oldcount = *refcount;
2898 	oldflags = ifp->if_flags;
2899 
2900 	/*
2901 	 * See if we aren't the only and touching refcount is enough.
2902 	 * Actually toggle interface flag if we are the first or last.
2903 	 */
2904 	if (onswitch) {
2905 		if ((*refcount)++)
2906 			return (0);
2907 		ifp->if_flags |= flag;
2908 	} else {
2909 		if (--(*refcount))
2910 			return (0);
2911 		ifp->if_flags &= ~flag;
2912 	}
2913 
2914 	/* Call down the driver since we've changed interface flags */
2915 	if (ifp->if_ioctl == NULL) {
2916 		error = EOPNOTSUPP;
2917 		goto recover;
2918 	}
2919 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2920 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2921 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2922 	if (error)
2923 		goto recover;
2924 	/* Notify userland that interface flags have changed */
2925 	rt_ifmsg(ifp);
2926 	return (0);
2927 
2928 recover:
2929 	/* Recover after driver error */
2930 	*refcount = oldcount;
2931 	ifp->if_flags = oldflags;
2932 	return (error);
2933 }
2934 
2935 /*
2936  * Set/clear promiscuous mode on interface ifp based on the truth value
2937  * of pswitch.  The calls are reference counted so that only the first
2938  * "on" request actually has an effect, as does the final "off" request.
2939  * Results are undefined if the "off" and "on" requests are not matched.
2940  */
2941 int
2942 ifpromisc(struct ifnet *ifp, int pswitch)
2943 {
2944 	int error;
2945 	int oldflags = ifp->if_flags;
2946 
2947 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2948 			   &ifp->if_pcount, pswitch);
2949 	/* If promiscuous mode status has changed, log a message */
2950 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2951             log_promisc_mode_change)
2952 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2953 		    ifp->if_xname,
2954 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2955 	return (error);
2956 }
2957 
2958 /*
2959  * Return interface configuration
2960  * of system.  List may be used
2961  * in later ioctl's (above) to get
2962  * other information.
2963  */
2964 /*ARGSUSED*/
2965 static int
2966 ifconf(u_long cmd, caddr_t data)
2967 {
2968 	struct ifconf *ifc = (struct ifconf *)data;
2969 	struct ifnet *ifp;
2970 	struct ifaddr *ifa;
2971 	struct ifreq ifr;
2972 	struct sbuf *sb;
2973 	int error, full = 0, valid_len, max_len;
2974 
2975 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2976 	max_len = MAXPHYS - 1;
2977 
2978 	/* Prevent hostile input from being able to crash the system */
2979 	if (ifc->ifc_len <= 0)
2980 		return (EINVAL);
2981 
2982 again:
2983 	if (ifc->ifc_len <= max_len) {
2984 		max_len = ifc->ifc_len;
2985 		full = 1;
2986 	}
2987 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2988 	max_len = 0;
2989 	valid_len = 0;
2990 
2991 	IFNET_RLOCK();
2992 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2993 		int addrs;
2994 
2995 		/*
2996 		 * Zero the ifr_name buffer to make sure we don't
2997 		 * disclose the contents of the stack.
2998 		 */
2999 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
3000 
3001 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3002 		    >= sizeof(ifr.ifr_name)) {
3003 			sbuf_delete(sb);
3004 			IFNET_RUNLOCK();
3005 			return (ENAMETOOLONG);
3006 		}
3007 
3008 		addrs = 0;
3009 		IF_ADDR_RLOCK(ifp);
3010 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3011 			struct sockaddr *sa = ifa->ifa_addr;
3012 
3013 			if (prison_if(curthread->td_ucred, sa) != 0)
3014 				continue;
3015 			addrs++;
3016 			/* COMPAT_SVR4 */
3017 			if (cmd == OSIOCGIFCONF) {
3018 				struct osockaddr *osa =
3019 				    (struct osockaddr *)&ifr.ifr_addr;
3020 				ifr.ifr_addr = *sa;
3021 				osa->sa_family = sa->sa_family;
3022 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3023 				max_len += sizeof(ifr);
3024 			} else
3025 			if (sa->sa_len <= sizeof(*sa)) {
3026 				ifr.ifr_addr = *sa;
3027 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3028 				max_len += sizeof(ifr);
3029 			} else {
3030 				sbuf_bcat(sb, &ifr,
3031 				    offsetof(struct ifreq, ifr_addr));
3032 				max_len += offsetof(struct ifreq, ifr_addr);
3033 				sbuf_bcat(sb, sa, sa->sa_len);
3034 				max_len += sa->sa_len;
3035 			}
3036 
3037 			if (sbuf_error(sb) == 0)
3038 				valid_len = sbuf_len(sb);
3039 		}
3040 		IF_ADDR_RUNLOCK(ifp);
3041 		if (addrs == 0) {
3042 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3043 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3044 			max_len += sizeof(ifr);
3045 
3046 			if (sbuf_error(sb) == 0)
3047 				valid_len = sbuf_len(sb);
3048 		}
3049 	}
3050 	IFNET_RUNLOCK();
3051 
3052 	/*
3053 	 * If we didn't allocate enough space (uncommon), try again.  If
3054 	 * we have already allocated as much space as we are allowed,
3055 	 * return what we've got.
3056 	 */
3057 	if (valid_len != max_len && !full) {
3058 		sbuf_delete(sb);
3059 		goto again;
3060 	}
3061 
3062 	ifc->ifc_len = valid_len;
3063 	sbuf_finish(sb);
3064 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3065 	sbuf_delete(sb);
3066 	return (error);
3067 }
3068 
3069 /*
3070  * Just like ifpromisc(), but for all-multicast-reception mode.
3071  */
3072 int
3073 if_allmulti(struct ifnet *ifp, int onswitch)
3074 {
3075 
3076 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3077 }
3078 
3079 struct ifmultiaddr *
3080 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3081 {
3082 	struct ifmultiaddr *ifma;
3083 
3084 	IF_ADDR_LOCK_ASSERT(ifp);
3085 
3086 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3087 		if (sa->sa_family == AF_LINK) {
3088 			if (sa_dl_equal(ifma->ifma_addr, sa))
3089 				break;
3090 		} else {
3091 			if (sa_equal(ifma->ifma_addr, sa))
3092 				break;
3093 		}
3094 	}
3095 
3096 	return ifma;
3097 }
3098 
3099 /*
3100  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3101  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3102  * the ifnet multicast address list here, so the caller must do that and
3103  * other setup work (such as notifying the device driver).  The reference
3104  * count is initialized to 1.
3105  */
3106 static struct ifmultiaddr *
3107 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3108     int mflags)
3109 {
3110 	struct ifmultiaddr *ifma;
3111 	struct sockaddr *dupsa;
3112 
3113 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3114 	    M_ZERO);
3115 	if (ifma == NULL)
3116 		return (NULL);
3117 
3118 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3119 	if (dupsa == NULL) {
3120 		free(ifma, M_IFMADDR);
3121 		return (NULL);
3122 	}
3123 	bcopy(sa, dupsa, sa->sa_len);
3124 	ifma->ifma_addr = dupsa;
3125 
3126 	ifma->ifma_ifp = ifp;
3127 	ifma->ifma_refcount = 1;
3128 	ifma->ifma_protospec = NULL;
3129 
3130 	if (llsa == NULL) {
3131 		ifma->ifma_lladdr = NULL;
3132 		return (ifma);
3133 	}
3134 
3135 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3136 	if (dupsa == NULL) {
3137 		free(ifma->ifma_addr, M_IFMADDR);
3138 		free(ifma, M_IFMADDR);
3139 		return (NULL);
3140 	}
3141 	bcopy(llsa, dupsa, llsa->sa_len);
3142 	ifma->ifma_lladdr = dupsa;
3143 
3144 	return (ifma);
3145 }
3146 
3147 /*
3148  * if_freemulti: free ifmultiaddr structure and possibly attached related
3149  * addresses.  The caller is responsible for implementing reference
3150  * counting, notifying the driver, handling routing messages, and releasing
3151  * any dependent link layer state.
3152  */
3153 static void
3154 if_freemulti(struct ifmultiaddr *ifma)
3155 {
3156 
3157 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3158 	    ifma->ifma_refcount));
3159 
3160 	if (ifma->ifma_lladdr != NULL)
3161 		free(ifma->ifma_lladdr, M_IFMADDR);
3162 	free(ifma->ifma_addr, M_IFMADDR);
3163 	free(ifma, M_IFMADDR);
3164 }
3165 
3166 /*
3167  * Register an additional multicast address with a network interface.
3168  *
3169  * - If the address is already present, bump the reference count on the
3170  *   address and return.
3171  * - If the address is not link-layer, look up a link layer address.
3172  * - Allocate address structures for one or both addresses, and attach to the
3173  *   multicast address list on the interface.  If automatically adding a link
3174  *   layer address, the protocol address will own a reference to the link
3175  *   layer address, to be freed when it is freed.
3176  * - Notify the network device driver of an addition to the multicast address
3177  *   list.
3178  *
3179  * 'sa' points to caller-owned memory with the desired multicast address.
3180  *
3181  * 'retifma' will be used to return a pointer to the resulting multicast
3182  * address reference, if desired.
3183  */
3184 int
3185 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3186     struct ifmultiaddr **retifma)
3187 {
3188 	struct ifmultiaddr *ifma, *ll_ifma;
3189 	struct sockaddr *llsa;
3190 	struct sockaddr_dl sdl;
3191 	int error;
3192 
3193 	/*
3194 	 * If the address is already present, return a new reference to it;
3195 	 * otherwise, allocate storage and set up a new address.
3196 	 */
3197 	IF_ADDR_WLOCK(ifp);
3198 	ifma = if_findmulti(ifp, sa);
3199 	if (ifma != NULL) {
3200 		ifma->ifma_refcount++;
3201 		if (retifma != NULL)
3202 			*retifma = ifma;
3203 		IF_ADDR_WUNLOCK(ifp);
3204 		return (0);
3205 	}
3206 
3207 	/*
3208 	 * The address isn't already present; resolve the protocol address
3209 	 * into a link layer address, and then look that up, bump its
3210 	 * refcount or allocate an ifma for that also.
3211 	 * Most link layer resolving functions returns address data which
3212 	 * fits inside default sockaddr_dl structure. However callback
3213 	 * can allocate another sockaddr structure, in that case we need to
3214 	 * free it later.
3215 	 */
3216 	llsa = NULL;
3217 	ll_ifma = NULL;
3218 	if (ifp->if_resolvemulti != NULL) {
3219 		/* Provide called function with buffer size information */
3220 		sdl.sdl_len = sizeof(sdl);
3221 		llsa = (struct sockaddr *)&sdl;
3222 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3223 		if (error)
3224 			goto unlock_out;
3225 	}
3226 
3227 	/*
3228 	 * Allocate the new address.  Don't hook it up yet, as we may also
3229 	 * need to allocate a link layer multicast address.
3230 	 */
3231 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3232 	if (ifma == NULL) {
3233 		error = ENOMEM;
3234 		goto free_llsa_out;
3235 	}
3236 
3237 	/*
3238 	 * If a link layer address is found, we'll need to see if it's
3239 	 * already present in the address list, or allocate is as well.
3240 	 * When this block finishes, the link layer address will be on the
3241 	 * list.
3242 	 */
3243 	if (llsa != NULL) {
3244 		ll_ifma = if_findmulti(ifp, llsa);
3245 		if (ll_ifma == NULL) {
3246 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3247 			if (ll_ifma == NULL) {
3248 				--ifma->ifma_refcount;
3249 				if_freemulti(ifma);
3250 				error = ENOMEM;
3251 				goto free_llsa_out;
3252 			}
3253 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3254 			    ifma_link);
3255 		} else
3256 			ll_ifma->ifma_refcount++;
3257 		ifma->ifma_llifma = ll_ifma;
3258 	}
3259 
3260 	/*
3261 	 * We now have a new multicast address, ifma, and possibly a new or
3262 	 * referenced link layer address.  Add the primary address to the
3263 	 * ifnet address list.
3264 	 */
3265 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3266 
3267 	if (retifma != NULL)
3268 		*retifma = ifma;
3269 
3270 	/*
3271 	 * Must generate the message while holding the lock so that 'ifma'
3272 	 * pointer is still valid.
3273 	 */
3274 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3275 	IF_ADDR_WUNLOCK(ifp);
3276 
3277 	/*
3278 	 * We are certain we have added something, so call down to the
3279 	 * interface to let them know about it.
3280 	 */
3281 	if (ifp->if_ioctl != NULL) {
3282 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3283 	}
3284 
3285 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3286 		link_free_sdl(llsa);
3287 
3288 	return (0);
3289 
3290 free_llsa_out:
3291 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3292 		link_free_sdl(llsa);
3293 
3294 unlock_out:
3295 	IF_ADDR_WUNLOCK(ifp);
3296 	return (error);
3297 }
3298 
3299 /*
3300  * Delete a multicast group membership by network-layer group address.
3301  *
3302  * Returns ENOENT if the entry could not be found. If ifp no longer
3303  * exists, results are undefined. This entry point should only be used
3304  * from subsystems which do appropriate locking to hold ifp for the
3305  * duration of the call.
3306  * Network-layer protocol domains must use if_delmulti_ifma().
3307  */
3308 int
3309 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3310 {
3311 	struct ifmultiaddr *ifma;
3312 	int lastref;
3313 #ifdef INVARIANTS
3314 	struct ifnet *oifp;
3315 
3316 	IFNET_RLOCK_NOSLEEP();
3317 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3318 		if (ifp == oifp)
3319 			break;
3320 	if (ifp != oifp)
3321 		ifp = NULL;
3322 	IFNET_RUNLOCK_NOSLEEP();
3323 
3324 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3325 #endif
3326 	if (ifp == NULL)
3327 		return (ENOENT);
3328 
3329 	IF_ADDR_WLOCK(ifp);
3330 	lastref = 0;
3331 	ifma = if_findmulti(ifp, sa);
3332 	if (ifma != NULL)
3333 		lastref = if_delmulti_locked(ifp, ifma, 0);
3334 	IF_ADDR_WUNLOCK(ifp);
3335 
3336 	if (ifma == NULL)
3337 		return (ENOENT);
3338 
3339 	if (lastref && ifp->if_ioctl != NULL) {
3340 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3341 	}
3342 
3343 	return (0);
3344 }
3345 
3346 /*
3347  * Delete all multicast group membership for an interface.
3348  * Should be used to quickly flush all multicast filters.
3349  */
3350 void
3351 if_delallmulti(struct ifnet *ifp)
3352 {
3353 	struct ifmultiaddr *ifma;
3354 	struct ifmultiaddr *next;
3355 
3356 	IF_ADDR_WLOCK(ifp);
3357 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3358 		if_delmulti_locked(ifp, ifma, 0);
3359 	IF_ADDR_WUNLOCK(ifp);
3360 }
3361 
3362 /*
3363  * Delete a multicast group membership by group membership pointer.
3364  * Network-layer protocol domains must use this routine.
3365  *
3366  * It is safe to call this routine if the ifp disappeared.
3367  */
3368 void
3369 if_delmulti_ifma(struct ifmultiaddr *ifma)
3370 {
3371 	struct ifnet *ifp;
3372 	int lastref;
3373 
3374 	ifp = ifma->ifma_ifp;
3375 #ifdef DIAGNOSTIC
3376 	if (ifp == NULL) {
3377 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3378 	} else {
3379 		struct ifnet *oifp;
3380 
3381 		IFNET_RLOCK_NOSLEEP();
3382 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3383 			if (ifp == oifp)
3384 				break;
3385 		if (ifp != oifp) {
3386 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3387 			ifp = NULL;
3388 		}
3389 		IFNET_RUNLOCK_NOSLEEP();
3390 	}
3391 #endif
3392 	/*
3393 	 * If and only if the ifnet instance exists: Acquire the address lock.
3394 	 */
3395 	if (ifp != NULL)
3396 		IF_ADDR_WLOCK(ifp);
3397 
3398 	lastref = if_delmulti_locked(ifp, ifma, 0);
3399 
3400 	if (ifp != NULL) {
3401 		/*
3402 		 * If and only if the ifnet instance exists:
3403 		 *  Release the address lock.
3404 		 *  If the group was left: update the hardware hash filter.
3405 		 */
3406 		IF_ADDR_WUNLOCK(ifp);
3407 		if (lastref && ifp->if_ioctl != NULL) {
3408 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3409 		}
3410 	}
3411 }
3412 
3413 /*
3414  * Perform deletion of network-layer and/or link-layer multicast address.
3415  *
3416  * Return 0 if the reference count was decremented.
3417  * Return 1 if the final reference was released, indicating that the
3418  * hardware hash filter should be reprogrammed.
3419  */
3420 static int
3421 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3422 {
3423 	struct ifmultiaddr *ll_ifma;
3424 
3425 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3426 		KASSERT(ifma->ifma_ifp == ifp,
3427 		    ("%s: inconsistent ifp %p", __func__, ifp));
3428 		IF_ADDR_WLOCK_ASSERT(ifp);
3429 	}
3430 
3431 	ifp = ifma->ifma_ifp;
3432 
3433 	/*
3434 	 * If the ifnet is detaching, null out references to ifnet,
3435 	 * so that upper protocol layers will notice, and not attempt
3436 	 * to obtain locks for an ifnet which no longer exists. The
3437 	 * routing socket announcement must happen before the ifnet
3438 	 * instance is detached from the system.
3439 	 */
3440 	if (detaching) {
3441 #ifdef DIAGNOSTIC
3442 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3443 #endif
3444 		/*
3445 		 * ifp may already be nulled out if we are being reentered
3446 		 * to delete the ll_ifma.
3447 		 */
3448 		if (ifp != NULL) {
3449 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3450 			ifma->ifma_ifp = NULL;
3451 		}
3452 	}
3453 
3454 	if (--ifma->ifma_refcount > 0)
3455 		return 0;
3456 
3457 	/*
3458 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3459 	 * have been associated with it. Release it first if so.
3460 	 */
3461 	ll_ifma = ifma->ifma_llifma;
3462 	if (ll_ifma != NULL) {
3463 		KASSERT(ifma->ifma_lladdr != NULL,
3464 		    ("%s: llifma w/o lladdr", __func__));
3465 		if (detaching)
3466 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3467 		if (--ll_ifma->ifma_refcount == 0) {
3468 			if (ifp != NULL) {
3469 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3470 				    ifma_link);
3471 			}
3472 			if_freemulti(ll_ifma);
3473 		}
3474 	}
3475 
3476 	if (ifp != NULL)
3477 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3478 
3479 	if_freemulti(ifma);
3480 
3481 	/*
3482 	 * The last reference to this instance of struct ifmultiaddr
3483 	 * was released; the hardware should be notified of this change.
3484 	 */
3485 	return 1;
3486 }
3487 
3488 /*
3489  * Set the link layer address on an interface.
3490  *
3491  * At this time we only support certain types of interfaces,
3492  * and we don't allow the length of the address to change.
3493  *
3494  * Set noinline to be dtrace-friendly
3495  */
3496 __noinline int
3497 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3498 {
3499 	struct sockaddr_dl *sdl;
3500 	struct ifaddr *ifa;
3501 	struct ifreq ifr;
3502 
3503 	IF_ADDR_RLOCK(ifp);
3504 	ifa = ifp->if_addr;
3505 	if (ifa == NULL) {
3506 		IF_ADDR_RUNLOCK(ifp);
3507 		return (EINVAL);
3508 	}
3509 	ifa_ref(ifa);
3510 	IF_ADDR_RUNLOCK(ifp);
3511 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3512 	if (sdl == NULL) {
3513 		ifa_free(ifa);
3514 		return (EINVAL);
3515 	}
3516 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3517 		ifa_free(ifa);
3518 		return (EINVAL);
3519 	}
3520 	switch (ifp->if_type) {
3521 	case IFT_ETHER:
3522 	case IFT_FDDI:
3523 	case IFT_XETHER:
3524 	case IFT_ISO88025:
3525 	case IFT_L2VLAN:
3526 	case IFT_BRIDGE:
3527 	case IFT_ARCNET:
3528 	case IFT_IEEE8023ADLAG:
3529 	case IFT_IEEE80211:
3530 		bcopy(lladdr, LLADDR(sdl), len);
3531 		ifa_free(ifa);
3532 		break;
3533 	default:
3534 		ifa_free(ifa);
3535 		return (ENODEV);
3536 	}
3537 
3538 	/*
3539 	 * If the interface is already up, we need
3540 	 * to re-init it in order to reprogram its
3541 	 * address filter.
3542 	 */
3543 	if ((ifp->if_flags & IFF_UP) != 0) {
3544 		if (ifp->if_ioctl) {
3545 			ifp->if_flags &= ~IFF_UP;
3546 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3547 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3548 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3549 			ifp->if_flags |= IFF_UP;
3550 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3551 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3552 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3553 		}
3554 	}
3555 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3556 	return (0);
3557 }
3558 
3559 /*
3560  * Compat function for handling basic encapsulation requests.
3561  * Not converted stacks (FDDI, IB, ..) supports traditional
3562  * output model: ARP (and other similar L2 protocols) are handled
3563  * inside output routine, arpresolve/nd6_resolve() returns MAC
3564  * address instead of full prepend.
3565  *
3566  * This function creates calculated header==MAC for IPv4/IPv6 and
3567  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3568  * address families.
3569  */
3570 static int
3571 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3572 {
3573 
3574 	if (req->rtype != IFENCAP_LL)
3575 		return (EOPNOTSUPP);
3576 
3577 	if (req->bufsize < req->lladdr_len)
3578 		return (ENOMEM);
3579 
3580 	switch (req->family) {
3581 	case AF_INET:
3582 	case AF_INET6:
3583 		break;
3584 	default:
3585 		return (EAFNOSUPPORT);
3586 	}
3587 
3588 	/* Copy lladdr to storage as is */
3589 	memmove(req->buf, req->lladdr, req->lladdr_len);
3590 	req->bufsize = req->lladdr_len;
3591 	req->lladdr_off = 0;
3592 
3593 	return (0);
3594 }
3595 
3596 /*
3597  * The name argument must be a pointer to storage which will last as
3598  * long as the interface does.  For physical devices, the result of
3599  * device_get_name(dev) is a good choice and for pseudo-devices a
3600  * static string works well.
3601  */
3602 void
3603 if_initname(struct ifnet *ifp, const char *name, int unit)
3604 {
3605 	ifp->if_dname = name;
3606 	ifp->if_dunit = unit;
3607 	if (unit != IF_DUNIT_NONE)
3608 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3609 	else
3610 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3611 }
3612 
3613 int
3614 if_printf(struct ifnet *ifp, const char * fmt, ...)
3615 {
3616 	va_list ap;
3617 	int retval;
3618 
3619 	retval = printf("%s: ", ifp->if_xname);
3620 	va_start(ap, fmt);
3621 	retval += vprintf(fmt, ap);
3622 	va_end(ap);
3623 	return (retval);
3624 }
3625 
3626 void
3627 if_start(struct ifnet *ifp)
3628 {
3629 
3630 	(*(ifp)->if_start)(ifp);
3631 }
3632 
3633 /*
3634  * Backwards compatibility interface for drivers
3635  * that have not implemented it
3636  */
3637 static int
3638 if_transmit(struct ifnet *ifp, struct mbuf *m)
3639 {
3640 	int error;
3641 
3642 	IFQ_HANDOFF(ifp, m, error);
3643 	return (error);
3644 }
3645 
3646 static void
3647 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3648 {
3649 
3650 	m_freem(m);
3651 }
3652 
3653 int
3654 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3655 {
3656 	int active = 0;
3657 
3658 	IF_LOCK(ifq);
3659 	if (_IF_QFULL(ifq)) {
3660 		IF_UNLOCK(ifq);
3661 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3662 		m_freem(m);
3663 		return (0);
3664 	}
3665 	if (ifp != NULL) {
3666 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3667 		if (m->m_flags & (M_BCAST|M_MCAST))
3668 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3669 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3670 	}
3671 	_IF_ENQUEUE(ifq, m);
3672 	IF_UNLOCK(ifq);
3673 	if (ifp != NULL && !active)
3674 		(*(ifp)->if_start)(ifp);
3675 	return (1);
3676 }
3677 
3678 void
3679 if_register_com_alloc(u_char type,
3680     if_com_alloc_t *a, if_com_free_t *f)
3681 {
3682 
3683 	KASSERT(if_com_alloc[type] == NULL,
3684 	    ("if_register_com_alloc: %d already registered", type));
3685 	KASSERT(if_com_free[type] == NULL,
3686 	    ("if_register_com_alloc: %d free already registered", type));
3687 
3688 	if_com_alloc[type] = a;
3689 	if_com_free[type] = f;
3690 }
3691 
3692 void
3693 if_deregister_com_alloc(u_char type)
3694 {
3695 
3696 	KASSERT(if_com_alloc[type] != NULL,
3697 	    ("if_deregister_com_alloc: %d not registered", type));
3698 	KASSERT(if_com_free[type] != NULL,
3699 	    ("if_deregister_com_alloc: %d free not registered", type));
3700 	if_com_alloc[type] = NULL;
3701 	if_com_free[type] = NULL;
3702 }
3703 
3704 /* API for driver access to network stack owned ifnet.*/
3705 uint64_t
3706 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3707 {
3708 	uint64_t oldbrate;
3709 
3710 	oldbrate = ifp->if_baudrate;
3711 	ifp->if_baudrate = baudrate;
3712 	return (oldbrate);
3713 }
3714 
3715 uint64_t
3716 if_getbaudrate(if_t ifp)
3717 {
3718 
3719 	return (((struct ifnet *)ifp)->if_baudrate);
3720 }
3721 
3722 int
3723 if_setcapabilities(if_t ifp, int capabilities)
3724 {
3725 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3726 	return (0);
3727 }
3728 
3729 int
3730 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3731 {
3732 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3733 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3734 
3735 	return (0);
3736 }
3737 
3738 int
3739 if_getcapabilities(if_t ifp)
3740 {
3741 	return ((struct ifnet *)ifp)->if_capabilities;
3742 }
3743 
3744 int
3745 if_setcapenable(if_t ifp, int capabilities)
3746 {
3747 	((struct ifnet *)ifp)->if_capenable = capabilities;
3748 	return (0);
3749 }
3750 
3751 int
3752 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3753 {
3754 	if(setcap)
3755 		((struct ifnet *)ifp)->if_capenable |= setcap;
3756 	if(clearcap)
3757 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3758 
3759 	return (0);
3760 }
3761 
3762 const char *
3763 if_getdname(if_t ifp)
3764 {
3765 	return ((struct ifnet *)ifp)->if_dname;
3766 }
3767 
3768 int
3769 if_togglecapenable(if_t ifp, int togglecap)
3770 {
3771 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3772 	return (0);
3773 }
3774 
3775 int
3776 if_getcapenable(if_t ifp)
3777 {
3778 	return ((struct ifnet *)ifp)->if_capenable;
3779 }
3780 
3781 /*
3782  * This is largely undesirable because it ties ifnet to a device, but does
3783  * provide flexiblity for an embedded product vendor. Should be used with
3784  * the understanding that it violates the interface boundaries, and should be
3785  * a last resort only.
3786  */
3787 int
3788 if_setdev(if_t ifp, void *dev)
3789 {
3790 	return (0);
3791 }
3792 
3793 int
3794 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3795 {
3796 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3797 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3798 
3799 	return (0);
3800 }
3801 
3802 int
3803 if_getdrvflags(if_t ifp)
3804 {
3805 	return ((struct ifnet *)ifp)->if_drv_flags;
3806 }
3807 
3808 int
3809 if_setdrvflags(if_t ifp, int flags)
3810 {
3811 	((struct ifnet *)ifp)->if_drv_flags = flags;
3812 	return (0);
3813 }
3814 
3815 
3816 int
3817 if_setflags(if_t ifp, int flags)
3818 {
3819 	((struct ifnet *)ifp)->if_flags = flags;
3820 	return (0);
3821 }
3822 
3823 int
3824 if_setflagbits(if_t ifp, int set, int clear)
3825 {
3826 	((struct ifnet *)ifp)->if_flags |= set;
3827 	((struct ifnet *)ifp)->if_flags &= ~clear;
3828 
3829 	return (0);
3830 }
3831 
3832 int
3833 if_getflags(if_t ifp)
3834 {
3835 	return ((struct ifnet *)ifp)->if_flags;
3836 }
3837 
3838 int
3839 if_clearhwassist(if_t ifp)
3840 {
3841 	((struct ifnet *)ifp)->if_hwassist = 0;
3842 	return (0);
3843 }
3844 
3845 int
3846 if_sethwassistbits(if_t ifp, int toset, int toclear)
3847 {
3848 	((struct ifnet *)ifp)->if_hwassist |= toset;
3849 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3850 
3851 	return (0);
3852 }
3853 
3854 int
3855 if_sethwassist(if_t ifp, int hwassist_bit)
3856 {
3857 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3858 	return (0);
3859 }
3860 
3861 int
3862 if_gethwassist(if_t ifp)
3863 {
3864 	return ((struct ifnet *)ifp)->if_hwassist;
3865 }
3866 
3867 int
3868 if_setmtu(if_t ifp, int mtu)
3869 {
3870 	((struct ifnet *)ifp)->if_mtu = mtu;
3871 	return (0);
3872 }
3873 
3874 int
3875 if_getmtu(if_t ifp)
3876 {
3877 	return ((struct ifnet *)ifp)->if_mtu;
3878 }
3879 
3880 int
3881 if_getmtu_family(if_t ifp, int family)
3882 {
3883 	struct domain *dp;
3884 
3885 	for (dp = domains; dp; dp = dp->dom_next) {
3886 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3887 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3888 	}
3889 
3890 	return (((struct ifnet *)ifp)->if_mtu);
3891 }
3892 
3893 int
3894 if_setsoftc(if_t ifp, void *softc)
3895 {
3896 	((struct ifnet *)ifp)->if_softc = softc;
3897 	return (0);
3898 }
3899 
3900 void *
3901 if_getsoftc(if_t ifp)
3902 {
3903 	return ((struct ifnet *)ifp)->if_softc;
3904 }
3905 
3906 void
3907 if_setrcvif(struct mbuf *m, if_t ifp)
3908 {
3909 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3910 }
3911 
3912 void
3913 if_setvtag(struct mbuf *m, uint16_t tag)
3914 {
3915 	m->m_pkthdr.ether_vtag = tag;
3916 }
3917 
3918 uint16_t
3919 if_getvtag(struct mbuf *m)
3920 {
3921 
3922 	return (m->m_pkthdr.ether_vtag);
3923 }
3924 
3925 int
3926 if_sendq_empty(if_t ifp)
3927 {
3928 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3929 }
3930 
3931 struct ifaddr *
3932 if_getifaddr(if_t ifp)
3933 {
3934 	return ((struct ifnet *)ifp)->if_addr;
3935 }
3936 
3937 int
3938 if_getamcount(if_t ifp)
3939 {
3940 	return ((struct ifnet *)ifp)->if_amcount;
3941 }
3942 
3943 
3944 int
3945 if_setsendqready(if_t ifp)
3946 {
3947 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3948 	return (0);
3949 }
3950 
3951 int
3952 if_setsendqlen(if_t ifp, int tx_desc_count)
3953 {
3954 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3955 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3956 
3957 	return (0);
3958 }
3959 
3960 int
3961 if_vlantrunkinuse(if_t ifp)
3962 {
3963 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3964 }
3965 
3966 int
3967 if_input(if_t ifp, struct mbuf* sendmp)
3968 {
3969 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3970 	return (0);
3971 
3972 }
3973 
3974 /* XXX */
3975 #ifndef ETH_ADDR_LEN
3976 #define ETH_ADDR_LEN 6
3977 #endif
3978 
3979 int
3980 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3981 {
3982 	struct ifmultiaddr *ifma;
3983 	uint8_t *lmta = (uint8_t *)mta;
3984 	int mcnt = 0;
3985 
3986 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3987 		if (ifma->ifma_addr->sa_family != AF_LINK)
3988 			continue;
3989 
3990 		if (mcnt == max)
3991 			break;
3992 
3993 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3994 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3995 		mcnt++;
3996 	}
3997 	*cnt = mcnt;
3998 
3999 	return (0);
4000 }
4001 
4002 int
4003 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4004 {
4005 	int error;
4006 
4007 	if_maddr_rlock(ifp);
4008 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4009 	if_maddr_runlock(ifp);
4010 	return (error);
4011 }
4012 
4013 int
4014 if_multiaddr_count(if_t ifp, int max)
4015 {
4016 	struct ifmultiaddr *ifma;
4017 	int count;
4018 
4019 	count = 0;
4020 	if_maddr_rlock(ifp);
4021 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4022 		if (ifma->ifma_addr->sa_family != AF_LINK)
4023 			continue;
4024 		count++;
4025 		if (count == max)
4026 			break;
4027 	}
4028 	if_maddr_runlock(ifp);
4029 	return (count);
4030 }
4031 
4032 int
4033 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4034 {
4035 	struct ifmultiaddr *ifma;
4036 	int cnt = 0;
4037 
4038 	if_maddr_rlock(ifp);
4039 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4040 		cnt += filter(arg, ifma, cnt);
4041 	if_maddr_runlock(ifp);
4042 	return (cnt);
4043 }
4044 
4045 struct mbuf *
4046 if_dequeue(if_t ifp)
4047 {
4048 	struct mbuf *m;
4049 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4050 
4051 	return (m);
4052 }
4053 
4054 int
4055 if_sendq_prepend(if_t ifp, struct mbuf *m)
4056 {
4057 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4058 	return (0);
4059 }
4060 
4061 int
4062 if_setifheaderlen(if_t ifp, int len)
4063 {
4064 	((struct ifnet *)ifp)->if_hdrlen = len;
4065 	return (0);
4066 }
4067 
4068 caddr_t
4069 if_getlladdr(if_t ifp)
4070 {
4071 	return (IF_LLADDR((struct ifnet *)ifp));
4072 }
4073 
4074 void *
4075 if_gethandle(u_char type)
4076 {
4077 	return (if_alloc(type));
4078 }
4079 
4080 void
4081 if_bpfmtap(if_t ifh, struct mbuf *m)
4082 {
4083 	struct ifnet *ifp = (struct ifnet *)ifh;
4084 
4085 	BPF_MTAP(ifp, m);
4086 }
4087 
4088 void
4089 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4090 {
4091 	struct ifnet *ifp = (struct ifnet *)ifh;
4092 
4093 	ETHER_BPF_MTAP(ifp, m);
4094 }
4095 
4096 void
4097 if_vlancap(if_t ifh)
4098 {
4099 	struct ifnet *ifp = (struct ifnet *)ifh;
4100 	VLAN_CAPABILITIES(ifp);
4101 }
4102 
4103 void
4104 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4105 {
4106 	((struct ifnet *)ifp)->if_init = init_fn;
4107 }
4108 
4109 void
4110 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4111 {
4112 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4113 }
4114 
4115 void
4116 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4117 {
4118 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4119 }
4120 
4121 void
4122 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4123 {
4124 	((struct ifnet *)ifp)->if_transmit = start_fn;
4125 }
4126 
4127 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4128 {
4129 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4130 
4131 }
4132 
4133 void
4134 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4135 {
4136 
4137 	ifp->if_get_counter = fn;
4138 }
4139 
4140 /* Revisit these - These are inline functions originally. */
4141 int
4142 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4143 {
4144 	return drbr_inuse(ifh, br);
4145 }
4146 
4147 struct mbuf*
4148 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4149 {
4150 	return drbr_dequeue(ifh, br);
4151 }
4152 
4153 int
4154 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4155 {
4156 	return drbr_needs_enqueue(ifh, br);
4157 }
4158 
4159 int
4160 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4161 {
4162 	return drbr_enqueue(ifh, br, m);
4163 
4164 }
4165