xref: /freebsd/sys/net/if.c (revision 282e23f07bf49b4e37aabdcc1c513a788db36d10)
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 	if (if_com_alloc[type] != NULL) {
459 		ifp->if_l2com = if_com_alloc[type](type, ifp);
460 		if (ifp->if_l2com == NULL) {
461 			free(ifp, M_IFNET);
462 			ifindex_free(idx);
463 			return (NULL);
464 		}
465 	}
466 
467 	IF_ADDR_LOCK_INIT(ifp);
468 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
469 	ifp->if_afdata_initialized = 0;
470 	IF_AFDATA_LOCK_INIT(ifp);
471 	TAILQ_INIT(&ifp->if_addrhead);
472 	TAILQ_INIT(&ifp->if_multiaddrs);
473 	TAILQ_INIT(&ifp->if_groups);
474 #ifdef MAC
475 	mac_ifnet_init(ifp);
476 #endif
477 	ifq_init(&ifp->if_snd, ifp);
478 
479 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
480 	for (int i = 0; i < IFCOUNTERS; i++)
481 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
482 	ifp->if_get_counter = if_get_counter_default;
483 	ifnet_setbyindex(ifp->if_index, ifp);
484 	return (ifp);
485 }
486 
487 /*
488  * Do the actual work of freeing a struct ifnet, and layer 2 common
489  * structure.  This call is made when the last reference to an
490  * interface is released.
491  */
492 static void
493 if_free_internal(struct ifnet *ifp)
494 {
495 
496 	KASSERT((ifp->if_flags & IFF_DYING),
497 	    ("if_free_internal: interface not dying"));
498 
499 	if (if_com_free[ifp->if_alloctype] != NULL)
500 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
501 		    ifp->if_alloctype);
502 
503 #ifdef MAC
504 	mac_ifnet_destroy(ifp);
505 #endif /* MAC */
506 	if (ifp->if_description != NULL)
507 		free(ifp->if_description, M_IFDESCR);
508 	IF_AFDATA_DESTROY(ifp);
509 	IF_ADDR_LOCK_DESTROY(ifp);
510 	ifq_delete(&ifp->if_snd);
511 
512 	for (int i = 0; i < IFCOUNTERS; i++)
513 		counter_u64_free(ifp->if_counters[i]);
514 
515 	free(ifp, M_IFNET);
516 }
517 
518 /*
519  * Deregister an interface and free the associated storage.
520  */
521 void
522 if_free(struct ifnet *ifp)
523 {
524 
525 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
526 
527 	CURVNET_SET_QUIET(ifp->if_vnet);
528 	IFNET_WLOCK();
529 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
530 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
531 
532 	ifindex_free_locked(ifp->if_index);
533 	IFNET_WUNLOCK();
534 
535 	if (refcount_release(&ifp->if_refcount))
536 		if_free_internal(ifp);
537 	CURVNET_RESTORE();
538 }
539 
540 /*
541  * Interfaces to keep an ifnet type-stable despite the possibility of the
542  * driver calling if_free().  If there are additional references, we defer
543  * freeing the underlying data structure.
544  */
545 void
546 if_ref(struct ifnet *ifp)
547 {
548 
549 	/* We don't assert the ifnet list lock here, but arguably should. */
550 	refcount_acquire(&ifp->if_refcount);
551 }
552 
553 void
554 if_rele(struct ifnet *ifp)
555 {
556 
557 	if (!refcount_release(&ifp->if_refcount))
558 		return;
559 	if_free_internal(ifp);
560 }
561 
562 void
563 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
564 {
565 
566 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
567 
568 	if (ifq->ifq_maxlen == 0)
569 		ifq->ifq_maxlen = ifqmaxlen;
570 
571 	ifq->altq_type = 0;
572 	ifq->altq_disc = NULL;
573 	ifq->altq_flags &= ALTQF_CANTCHANGE;
574 	ifq->altq_tbr  = NULL;
575 	ifq->altq_ifp  = ifp;
576 }
577 
578 void
579 ifq_delete(struct ifaltq *ifq)
580 {
581 	mtx_destroy(&ifq->ifq_mtx);
582 }
583 
584 /*
585  * Perform generic interface initialization tasks and attach the interface
586  * to the list of "active" interfaces.  If vmove flag is set on entry
587  * to if_attach_internal(), perform only a limited subset of initialization
588  * tasks, given that we are moving from one vnet to another an ifnet which
589  * has already been fully initialized.
590  *
591  * Note that if_detach_internal() removes group membership unconditionally
592  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
593  * Thus, when if_vmove() is applied to a cloned interface, group membership
594  * is lost while a cloned one always joins a group whose name is
595  * ifc->ifc_name.  To recover this after if_detach_internal() and
596  * if_attach_internal(), the cloner should be specified to
597  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
598  * attempts to join a group whose name is ifc->ifc_name.
599  *
600  * XXX:
601  *  - The decision to return void and thus require this function to
602  *    succeed is questionable.
603  *  - We should probably do more sanity checking.  For instance we don't
604  *    do anything to insure if_xname is unique or non-empty.
605  */
606 void
607 if_attach(struct ifnet *ifp)
608 {
609 
610 	if_attach_internal(ifp, 0, NULL);
611 }
612 
613 /*
614  * Compute the least common TSO limit.
615  */
616 void
617 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
618 {
619 	/*
620 	 * 1) If there is no limit currently, take the limit from
621 	 * the network adapter.
622 	 *
623 	 * 2) If the network adapter has a limit below the current
624 	 * limit, apply it.
625 	 */
626 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
627 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
628 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
629 	}
630 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
631 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
632 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
633 	}
634 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
635 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
636 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
637 	}
638 }
639 
640 /*
641  * Update TSO limit of a network adapter.
642  *
643  * Returns zero if no change. Else non-zero.
644  */
645 int
646 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
647 {
648 	int retval = 0;
649 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
650 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
651 		retval++;
652 	}
653 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
654 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
655 		retval++;
656 	}
657 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
658 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
659 		retval++;
660 	}
661 	return (retval);
662 }
663 
664 static void
665 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
666 {
667 	unsigned socksize, ifasize;
668 	int namelen, masklen;
669 	struct sockaddr_dl *sdl;
670 	struct ifaddr *ifa;
671 
672 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
673 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
674 		    ifp->if_xname);
675 
676 #ifdef VIMAGE
677 	ifp->if_vnet = curvnet;
678 	if (ifp->if_home_vnet == NULL)
679 		ifp->if_home_vnet = curvnet;
680 #endif
681 
682 	if_addgroup(ifp, IFG_ALL);
683 
684 	/* Restore group membership for cloned interfaces. */
685 	if (vmove && ifc != NULL)
686 		if_clone_addgroup(ifp, ifc);
687 
688 	getmicrotime(&ifp->if_lastchange);
689 	ifp->if_epoch = time_uptime;
690 
691 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
692 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
693 	    ("transmit and qflush must both either be set or both be NULL"));
694 	if (ifp->if_transmit == NULL) {
695 		ifp->if_transmit = if_transmit;
696 		ifp->if_qflush = if_qflush;
697 	}
698 	if (ifp->if_input == NULL)
699 		ifp->if_input = if_input_default;
700 
701 	if (ifp->if_requestencap == NULL)
702 		ifp->if_requestencap = if_requestencap_default;
703 
704 	if (!vmove) {
705 #ifdef MAC
706 		mac_ifnet_create(ifp);
707 #endif
708 
709 		/*
710 		 * Create a Link Level name for this device.
711 		 */
712 		namelen = strlen(ifp->if_xname);
713 		/*
714 		 * Always save enough space for any possiable name so we
715 		 * can do a rename in place later.
716 		 */
717 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
718 		socksize = masklen + ifp->if_addrlen;
719 		if (socksize < sizeof(*sdl))
720 			socksize = sizeof(*sdl);
721 		socksize = roundup2(socksize, sizeof(long));
722 		ifasize = sizeof(*ifa) + 2 * socksize;
723 		ifa = ifa_alloc(ifasize, M_WAITOK);
724 		sdl = (struct sockaddr_dl *)(ifa + 1);
725 		sdl->sdl_len = socksize;
726 		sdl->sdl_family = AF_LINK;
727 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
728 		sdl->sdl_nlen = namelen;
729 		sdl->sdl_index = ifp->if_index;
730 		sdl->sdl_type = ifp->if_type;
731 		ifp->if_addr = ifa;
732 		ifa->ifa_ifp = ifp;
733 		ifa->ifa_rtrequest = link_rtrequest;
734 		ifa->ifa_addr = (struct sockaddr *)sdl;
735 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
736 		ifa->ifa_netmask = (struct sockaddr *)sdl;
737 		sdl->sdl_len = masklen;
738 		while (namelen != 0)
739 			sdl->sdl_data[--namelen] = 0xff;
740 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
741 		/* Reliably crash if used uninitialized. */
742 		ifp->if_broadcastaddr = NULL;
743 
744 #if defined(INET) || defined(INET6)
745 		/* Use defaults for TSO, if nothing is set */
746 		if (ifp->if_hw_tsomax == 0 &&
747 		    ifp->if_hw_tsomaxsegcount == 0 &&
748 		    ifp->if_hw_tsomaxsegsize == 0) {
749 			/*
750 			 * The TSO defaults needs to be such that an
751 			 * NFS mbuf list of 35 mbufs totalling just
752 			 * below 64K works and that a chain of mbufs
753 			 * can be defragged into at most 32 segments:
754 			 */
755 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
756 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
757 			ifp->if_hw_tsomaxsegcount = 35;
758 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
759 
760 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
761 			if (ifp->if_capabilities & IFCAP_TSO) {
762 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
763 				    ifp->if_hw_tsomax,
764 				    ifp->if_hw_tsomaxsegcount,
765 				    ifp->if_hw_tsomaxsegsize);
766 			}
767 		}
768 #endif
769 	}
770 #ifdef VIMAGE
771 	else {
772 		/*
773 		 * Update the interface index in the link layer address
774 		 * of the interface.
775 		 */
776 		for (ifa = ifp->if_addr; ifa != NULL;
777 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
778 			if (ifa->ifa_addr->sa_family == AF_LINK) {
779 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
780 				sdl->sdl_index = ifp->if_index;
781 			}
782 		}
783 	}
784 #endif
785 
786 	IFNET_WLOCK();
787 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
788 #ifdef VIMAGE
789 	curvnet->vnet_ifcnt++;
790 #endif
791 	IFNET_WUNLOCK();
792 
793 	if (domain_init_status >= 2)
794 		if_attachdomain1(ifp);
795 
796 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
797 	if (IS_DEFAULT_VNET(curvnet))
798 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
799 
800 	/* Announce the interface. */
801 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
802 }
803 
804 static void
805 if_attachdomain(void *dummy)
806 {
807 	struct ifnet *ifp;
808 
809 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
810 		if_attachdomain1(ifp);
811 }
812 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
813     if_attachdomain, NULL);
814 
815 static void
816 if_attachdomain1(struct ifnet *ifp)
817 {
818 	struct domain *dp;
819 
820 	/*
821 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
822 	 * cannot lock ifp->if_afdata initialization, entirely.
823 	 */
824 	IF_AFDATA_LOCK(ifp);
825 	if (ifp->if_afdata_initialized >= domain_init_status) {
826 		IF_AFDATA_UNLOCK(ifp);
827 		log(LOG_WARNING, "%s called more than once on %s\n",
828 		    __func__, ifp->if_xname);
829 		return;
830 	}
831 	ifp->if_afdata_initialized = domain_init_status;
832 	IF_AFDATA_UNLOCK(ifp);
833 
834 	/* address family dependent data region */
835 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
836 	for (dp = domains; dp; dp = dp->dom_next) {
837 		if (dp->dom_ifattach)
838 			ifp->if_afdata[dp->dom_family] =
839 			    (*dp->dom_ifattach)(ifp);
840 	}
841 }
842 
843 /*
844  * Remove any unicast or broadcast network addresses from an interface.
845  */
846 void
847 if_purgeaddrs(struct ifnet *ifp)
848 {
849 	struct ifaddr *ifa, *next;
850 
851 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
852 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
853 		if (ifa->ifa_addr->sa_family == AF_LINK)
854 			continue;
855 #ifdef INET
856 		/* XXX: Ugly!! ad hoc just for INET */
857 		if (ifa->ifa_addr->sa_family == AF_INET) {
858 			struct ifaliasreq ifr;
859 
860 			bzero(&ifr, sizeof(ifr));
861 			ifr.ifra_addr = *ifa->ifa_addr;
862 			if (ifa->ifa_dstaddr)
863 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
864 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
865 			    NULL) == 0)
866 				continue;
867 		}
868 #endif /* INET */
869 #ifdef INET6
870 		if (ifa->ifa_addr->sa_family == AF_INET6) {
871 			in6_purgeaddr(ifa);
872 			/* ifp_addrhead is already updated */
873 			continue;
874 		}
875 #endif /* INET6 */
876 		IF_ADDR_WLOCK(ifp);
877 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
878 		IF_ADDR_WUNLOCK(ifp);
879 		ifa_free(ifa);
880 	}
881 }
882 
883 /*
884  * Remove any multicast network addresses from an interface when an ifnet
885  * is going away.
886  */
887 static void
888 if_purgemaddrs(struct ifnet *ifp)
889 {
890 	struct ifmultiaddr *ifma;
891 	struct ifmultiaddr *next;
892 
893 	IF_ADDR_WLOCK(ifp);
894 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
895 		if_delmulti_locked(ifp, ifma, 1);
896 	IF_ADDR_WUNLOCK(ifp);
897 }
898 
899 /*
900  * Detach an interface, removing it from the list of "active" interfaces.
901  * If vmove flag is set on entry to if_detach_internal(), perform only a
902  * limited subset of cleanup tasks, given that we are moving an ifnet from
903  * one vnet to another, where it must be fully operational.
904  *
905  * XXXRW: There are some significant questions about event ordering, and
906  * how to prevent things from starting to use the interface during detach.
907  */
908 void
909 if_detach(struct ifnet *ifp)
910 {
911 
912 	CURVNET_SET_QUIET(ifp->if_vnet);
913 	if_detach_internal(ifp, 0, NULL);
914 	CURVNET_RESTORE();
915 }
916 
917 /*
918  * The vmove flag, if set, indicates that we are called from a callpath
919  * that is moving an interface to a different vnet instance.
920  *
921  * The shutdown flag, if set, indicates that we are called in the
922  * process of shutting down a vnet instance.  Currently only the
923  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
924  * on a vnet instance shutdown without this flag being set, e.g., when
925  * the cloned interfaces are destoyed as first thing of teardown.
926  */
927 static int
928 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
929 {
930 	struct ifaddr *ifa;
931 	int i;
932 	struct domain *dp;
933  	struct ifnet *iter;
934  	int found = 0;
935 #ifdef VIMAGE
936 	int shutdown;
937 
938 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
939 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
940 #endif
941 	IFNET_WLOCK();
942 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
943 		if (iter == ifp) {
944 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
945 			found = 1;
946 			break;
947 		}
948 	IFNET_WUNLOCK();
949 	if (!found) {
950 		/*
951 		 * While we would want to panic here, we cannot
952 		 * guarantee that the interface is indeed still on
953 		 * the list given we don't hold locks all the way.
954 		 */
955 		return (ENOENT);
956 #if 0
957 		if (vmove)
958 			panic("%s: ifp=%p not on the ifnet tailq %p",
959 			    __func__, ifp, &V_ifnet);
960 		else
961 			return; /* XXX this should panic as well? */
962 #endif
963 	}
964 
965 	/*
966 	 * At this point we know the interface still was on the ifnet list
967 	 * and we removed it so we are in a stable state.
968 	 */
969 #ifdef VIMAGE
970 	curvnet->vnet_ifcnt--;
971 #endif
972 
973 	/*
974 	 * In any case (destroy or vmove) detach us from the groups
975 	 * and remove/wait for pending events on the taskq.
976 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
977 	 */
978 	if_delgroups(ifp);
979 
980 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
981 
982 	/*
983 	 * Check if this is a cloned interface or not. Must do even if
984 	 * shutting down as a if_vmove_reclaim() would move the ifp and
985 	 * the if_clone_addgroup() will have a corrupted string overwise
986 	 * from a gibberish pointer.
987 	 */
988 	if (vmove && ifcp != NULL)
989 		*ifcp = if_clone_findifc(ifp);
990 
991 	if_down(ifp);
992 
993 #ifdef VIMAGE
994 	/*
995 	 * On VNET shutdown abort here as the stack teardown will do all
996 	 * the work top-down for us.
997 	 */
998 	if (shutdown) {
999 		/*
1000 		 * In case of a vmove we are done here without error.
1001 		 * If we would signal an error it would lead to the same
1002 		 * abort as if we did not find the ifnet anymore.
1003 		 * if_detach() calls us in void context and does not care
1004 		 * about an early abort notification, so life is splendid :)
1005 		 */
1006 		goto finish_vnet_shutdown;
1007 	}
1008 #endif
1009 
1010 	/*
1011 	 * At this point we are not tearing down a VNET and are either
1012 	 * going to destroy or vmove the interface and have to cleanup
1013 	 * accordingly.
1014 	 */
1015 
1016 	/*
1017 	 * Remove routes and flush queues.
1018 	 */
1019 #ifdef ALTQ
1020 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1021 		altq_disable(&ifp->if_snd);
1022 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1023 		altq_detach(&ifp->if_snd);
1024 #endif
1025 
1026 	if_purgeaddrs(ifp);
1027 
1028 #ifdef INET
1029 	in_ifdetach(ifp);
1030 #endif
1031 
1032 #ifdef INET6
1033 	/*
1034 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1035 	 * before removing routing entries below, since IPv6 interface direct
1036 	 * routes are expected to be removed by the IPv6-specific kernel API.
1037 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1038 	 */
1039 	in6_ifdetach(ifp);
1040 #endif
1041 	if_purgemaddrs(ifp);
1042 
1043 	/* Announce that the interface is gone. */
1044 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1045 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1046 	if (IS_DEFAULT_VNET(curvnet))
1047 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1048 
1049 	if (!vmove) {
1050 		/*
1051 		 * Prevent further calls into the device driver via ifnet.
1052 		 */
1053 		if_dead(ifp);
1054 
1055 		/*
1056 		 * Remove link ifaddr pointer and maybe decrement if_index.
1057 		 * Clean up all addresses.
1058 		 */
1059 		ifp->if_addr = NULL;
1060 
1061 		/* We can now free link ifaddr. */
1062 		IF_ADDR_WLOCK(ifp);
1063 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1064 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1065 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1066 			IF_ADDR_WUNLOCK(ifp);
1067 			ifa_free(ifa);
1068 		} else
1069 			IF_ADDR_WUNLOCK(ifp);
1070 	}
1071 
1072 	rt_flushifroutes(ifp);
1073 
1074 #ifdef VIMAGE
1075 finish_vnet_shutdown:
1076 #endif
1077 	/*
1078 	 * We cannot hold the lock over dom_ifdetach calls as they might
1079 	 * sleep, for example trying to drain a callout, thus open up the
1080 	 * theoretical race with re-attaching.
1081 	 */
1082 	IF_AFDATA_LOCK(ifp);
1083 	i = ifp->if_afdata_initialized;
1084 	ifp->if_afdata_initialized = 0;
1085 	IF_AFDATA_UNLOCK(ifp);
1086 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1087 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1088 			(*dp->dom_ifdetach)(ifp,
1089 			    ifp->if_afdata[dp->dom_family]);
1090 			ifp->if_afdata[dp->dom_family] = NULL;
1091 		}
1092 	}
1093 
1094 	return (0);
1095 }
1096 
1097 #ifdef VIMAGE
1098 /*
1099  * if_vmove() performs a limited version of if_detach() in current
1100  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1101  * An attempt is made to shrink if_index in current vnet, find an
1102  * unused if_index in target vnet and calls if_grow() if necessary,
1103  * and finally find an unused if_xname for the target vnet.
1104  */
1105 static void
1106 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1107 {
1108 	struct if_clone *ifc;
1109 	u_int bif_dlt, bif_hdrlen;
1110 	int rc;
1111 
1112  	/*
1113 	 * if_detach_internal() will call the eventhandler to notify
1114 	 * interface departure.  That will detach if_bpf.  We need to
1115 	 * safe the dlt and hdrlen so we can re-attach it later.
1116 	 */
1117 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1118 
1119 	/*
1120 	 * Detach from current vnet, but preserve LLADDR info, do not
1121 	 * mark as dead etc. so that the ifnet can be reattached later.
1122 	 * If we cannot find it, we lost the race to someone else.
1123 	 */
1124 	rc = if_detach_internal(ifp, 1, &ifc);
1125 	if (rc != 0)
1126 		return;
1127 
1128 	/*
1129 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1130 	 * the if_index for that vnet if possible.
1131 	 *
1132 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1133 	 * or we'd lock on one vnet and unlock on another.
1134 	 */
1135 	IFNET_WLOCK();
1136 	ifindex_free_locked(ifp->if_index);
1137 	IFNET_WUNLOCK();
1138 
1139 	/*
1140 	 * Perform interface-specific reassignment tasks, if provided by
1141 	 * the driver.
1142 	 */
1143 	if (ifp->if_reassign != NULL)
1144 		ifp->if_reassign(ifp, new_vnet, NULL);
1145 
1146 	/*
1147 	 * Switch to the context of the target vnet.
1148 	 */
1149 	CURVNET_SET_QUIET(new_vnet);
1150 
1151 	IFNET_WLOCK();
1152 	ifp->if_index = ifindex_alloc();
1153 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1154 	IFNET_WUNLOCK();
1155 
1156 	if_attach_internal(ifp, 1, ifc);
1157 
1158 	if (ifp->if_bpf == NULL)
1159 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1160 
1161 	CURVNET_RESTORE();
1162 }
1163 
1164 /*
1165  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1166  */
1167 static int
1168 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1169 {
1170 	struct prison *pr;
1171 	struct ifnet *difp;
1172 	int shutdown;
1173 
1174 	/* Try to find the prison within our visibility. */
1175 	sx_slock(&allprison_lock);
1176 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1177 	sx_sunlock(&allprison_lock);
1178 	if (pr == NULL)
1179 		return (ENXIO);
1180 	prison_hold_locked(pr);
1181 	mtx_unlock(&pr->pr_mtx);
1182 
1183 	/* Do not try to move the iface from and to the same prison. */
1184 	if (pr->pr_vnet == ifp->if_vnet) {
1185 		prison_free(pr);
1186 		return (EEXIST);
1187 	}
1188 
1189 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1190 	/* XXX Lock interfaces to avoid races. */
1191 	CURVNET_SET_QUIET(pr->pr_vnet);
1192 	difp = ifunit(ifname);
1193 	if (difp != NULL) {
1194 		CURVNET_RESTORE();
1195 		prison_free(pr);
1196 		return (EEXIST);
1197 	}
1198 
1199 	/* Make sure the VNET is stable. */
1200 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1201 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1202 	if (shutdown) {
1203 		CURVNET_RESTORE();
1204 		prison_free(pr);
1205 		return (EBUSY);
1206 	}
1207 	CURVNET_RESTORE();
1208 
1209 	/* Move the interface into the child jail/vnet. */
1210 	if_vmove(ifp, pr->pr_vnet);
1211 
1212 	/* Report the new if_xname back to the userland. */
1213 	sprintf(ifname, "%s", ifp->if_xname);
1214 
1215 	prison_free(pr);
1216 	return (0);
1217 }
1218 
1219 static int
1220 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1221 {
1222 	struct prison *pr;
1223 	struct vnet *vnet_dst;
1224 	struct ifnet *ifp;
1225  	int shutdown;
1226 
1227 	/* Try to find the prison within our visibility. */
1228 	sx_slock(&allprison_lock);
1229 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1230 	sx_sunlock(&allprison_lock);
1231 	if (pr == NULL)
1232 		return (ENXIO);
1233 	prison_hold_locked(pr);
1234 	mtx_unlock(&pr->pr_mtx);
1235 
1236 	/* Make sure the named iface exists in the source prison/vnet. */
1237 	CURVNET_SET(pr->pr_vnet);
1238 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1239 	if (ifp == NULL) {
1240 		CURVNET_RESTORE();
1241 		prison_free(pr);
1242 		return (ENXIO);
1243 	}
1244 
1245 	/* Do not try to move the iface from and to the same prison. */
1246 	vnet_dst = TD_TO_VNET(td);
1247 	if (vnet_dst == ifp->if_vnet) {
1248 		CURVNET_RESTORE();
1249 		prison_free(pr);
1250 		return (EEXIST);
1251 	}
1252 
1253 	/* Make sure the VNET is stable. */
1254 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1255 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1256 	if (shutdown) {
1257 		CURVNET_RESTORE();
1258 		prison_free(pr);
1259 		return (EBUSY);
1260 	}
1261 
1262 	/* Get interface back from child jail/vnet. */
1263 	if_vmove(ifp, vnet_dst);
1264 	CURVNET_RESTORE();
1265 
1266 	/* Report the new if_xname back to the userland. */
1267 	sprintf(ifname, "%s", ifp->if_xname);
1268 
1269 	prison_free(pr);
1270 	return (0);
1271 }
1272 #endif /* VIMAGE */
1273 
1274 /*
1275  * Add a group to an interface
1276  */
1277 int
1278 if_addgroup(struct ifnet *ifp, const char *groupname)
1279 {
1280 	struct ifg_list		*ifgl;
1281 	struct ifg_group	*ifg = NULL;
1282 	struct ifg_member	*ifgm;
1283 	int 			 new = 0;
1284 
1285 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1286 	    groupname[strlen(groupname) - 1] <= '9')
1287 		return (EINVAL);
1288 
1289 	IFNET_WLOCK();
1290 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1291 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1292 			IFNET_WUNLOCK();
1293 			return (EEXIST);
1294 		}
1295 
1296 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1297 	    M_NOWAIT)) == NULL) {
1298 	    	IFNET_WUNLOCK();
1299 		return (ENOMEM);
1300 	}
1301 
1302 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1303 	    M_TEMP, M_NOWAIT)) == NULL) {
1304 		free(ifgl, M_TEMP);
1305 		IFNET_WUNLOCK();
1306 		return (ENOMEM);
1307 	}
1308 
1309 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1310 		if (!strcmp(ifg->ifg_group, groupname))
1311 			break;
1312 
1313 	if (ifg == NULL) {
1314 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1315 		    M_TEMP, M_NOWAIT)) == NULL) {
1316 			free(ifgl, M_TEMP);
1317 			free(ifgm, M_TEMP);
1318 			IFNET_WUNLOCK();
1319 			return (ENOMEM);
1320 		}
1321 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1322 		ifg->ifg_refcnt = 0;
1323 		TAILQ_INIT(&ifg->ifg_members);
1324 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1325 		new = 1;
1326 	}
1327 
1328 	ifg->ifg_refcnt++;
1329 	ifgl->ifgl_group = ifg;
1330 	ifgm->ifgm_ifp = ifp;
1331 
1332 	IF_ADDR_WLOCK(ifp);
1333 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1334 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1335 	IF_ADDR_WUNLOCK(ifp);
1336 
1337 	IFNET_WUNLOCK();
1338 
1339 	if (new)
1340 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1341 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1342 
1343 	return (0);
1344 }
1345 
1346 /*
1347  * Remove a group from an interface
1348  */
1349 int
1350 if_delgroup(struct ifnet *ifp, const char *groupname)
1351 {
1352 	struct ifg_list		*ifgl;
1353 	struct ifg_member	*ifgm;
1354 
1355 	IFNET_WLOCK();
1356 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1357 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1358 			break;
1359 	if (ifgl == NULL) {
1360 		IFNET_WUNLOCK();
1361 		return (ENOENT);
1362 	}
1363 
1364 	IF_ADDR_WLOCK(ifp);
1365 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1366 	IF_ADDR_WUNLOCK(ifp);
1367 
1368 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1369 		if (ifgm->ifgm_ifp == ifp)
1370 			break;
1371 
1372 	if (ifgm != NULL) {
1373 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1374 		free(ifgm, M_TEMP);
1375 	}
1376 
1377 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1378 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1379 		IFNET_WUNLOCK();
1380 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1381 		free(ifgl->ifgl_group, M_TEMP);
1382 	} else
1383 		IFNET_WUNLOCK();
1384 
1385 	free(ifgl, M_TEMP);
1386 
1387 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1388 
1389 	return (0);
1390 }
1391 
1392 /*
1393  * Remove an interface from all groups
1394  */
1395 static void
1396 if_delgroups(struct ifnet *ifp)
1397 {
1398 	struct ifg_list		*ifgl;
1399 	struct ifg_member	*ifgm;
1400 	char groupname[IFNAMSIZ];
1401 
1402 	IFNET_WLOCK();
1403 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1404 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1405 
1406 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1407 
1408 		IF_ADDR_WLOCK(ifp);
1409 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1410 		IF_ADDR_WUNLOCK(ifp);
1411 
1412 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1413 			if (ifgm->ifgm_ifp == ifp)
1414 				break;
1415 
1416 		if (ifgm != NULL) {
1417 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1418 			    ifgm_next);
1419 			free(ifgm, M_TEMP);
1420 		}
1421 
1422 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1423 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1424 			IFNET_WUNLOCK();
1425 			EVENTHANDLER_INVOKE(group_detach_event,
1426 			    ifgl->ifgl_group);
1427 			free(ifgl->ifgl_group, M_TEMP);
1428 		} else
1429 			IFNET_WUNLOCK();
1430 
1431 		free(ifgl, M_TEMP);
1432 
1433 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1434 
1435 		IFNET_WLOCK();
1436 	}
1437 	IFNET_WUNLOCK();
1438 }
1439 
1440 /*
1441  * Stores all groups from an interface in memory pointed
1442  * to by data
1443  */
1444 static int
1445 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1446 {
1447 	int			 len, error;
1448 	struct ifg_list		*ifgl;
1449 	struct ifg_req		 ifgrq, *ifgp;
1450 	struct ifgroupreq	*ifgr = data;
1451 
1452 	if (ifgr->ifgr_len == 0) {
1453 		IF_ADDR_RLOCK(ifp);
1454 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1455 			ifgr->ifgr_len += sizeof(struct ifg_req);
1456 		IF_ADDR_RUNLOCK(ifp);
1457 		return (0);
1458 	}
1459 
1460 	len = ifgr->ifgr_len;
1461 	ifgp = ifgr->ifgr_groups;
1462 	/* XXX: wire */
1463 	IF_ADDR_RLOCK(ifp);
1464 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1465 		if (len < sizeof(ifgrq)) {
1466 			IF_ADDR_RUNLOCK(ifp);
1467 			return (EINVAL);
1468 		}
1469 		bzero(&ifgrq, sizeof ifgrq);
1470 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1471 		    sizeof(ifgrq.ifgrq_group));
1472 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1473 		    	IF_ADDR_RUNLOCK(ifp);
1474 			return (error);
1475 		}
1476 		len -= sizeof(ifgrq);
1477 		ifgp++;
1478 	}
1479 	IF_ADDR_RUNLOCK(ifp);
1480 
1481 	return (0);
1482 }
1483 
1484 /*
1485  * Stores all members of a group in memory pointed to by data
1486  */
1487 static int
1488 if_getgroupmembers(struct ifgroupreq *data)
1489 {
1490 	struct ifgroupreq	*ifgr = data;
1491 	struct ifg_group	*ifg;
1492 	struct ifg_member	*ifgm;
1493 	struct ifg_req		 ifgrq, *ifgp;
1494 	int			 len, error;
1495 
1496 	IFNET_RLOCK();
1497 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1498 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1499 			break;
1500 	if (ifg == NULL) {
1501 		IFNET_RUNLOCK();
1502 		return (ENOENT);
1503 	}
1504 
1505 	if (ifgr->ifgr_len == 0) {
1506 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1507 			ifgr->ifgr_len += sizeof(ifgrq);
1508 		IFNET_RUNLOCK();
1509 		return (0);
1510 	}
1511 
1512 	len = ifgr->ifgr_len;
1513 	ifgp = ifgr->ifgr_groups;
1514 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1515 		if (len < sizeof(ifgrq)) {
1516 			IFNET_RUNLOCK();
1517 			return (EINVAL);
1518 		}
1519 		bzero(&ifgrq, sizeof ifgrq);
1520 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1521 		    sizeof(ifgrq.ifgrq_member));
1522 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1523 			IFNET_RUNLOCK();
1524 			return (error);
1525 		}
1526 		len -= sizeof(ifgrq);
1527 		ifgp++;
1528 	}
1529 	IFNET_RUNLOCK();
1530 
1531 	return (0);
1532 }
1533 
1534 /*
1535  * Return counter values from counter(9)s stored in ifnet.
1536  */
1537 uint64_t
1538 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1539 {
1540 
1541 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1542 
1543 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1544 }
1545 
1546 /*
1547  * Increase an ifnet counter. Usually used for counters shared
1548  * between the stack and a driver, but function supports them all.
1549  */
1550 void
1551 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1552 {
1553 
1554 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1555 
1556 	counter_u64_add(ifp->if_counters[cnt], inc);
1557 }
1558 
1559 /*
1560  * Copy data from ifnet to userland API structure if_data.
1561  */
1562 void
1563 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1564 {
1565 
1566 	ifd->ifi_type = ifp->if_type;
1567 	ifd->ifi_physical = 0;
1568 	ifd->ifi_addrlen = ifp->if_addrlen;
1569 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1570 	ifd->ifi_link_state = ifp->if_link_state;
1571 	ifd->ifi_vhid = 0;
1572 	ifd->ifi_datalen = sizeof(struct if_data);
1573 	ifd->ifi_mtu = ifp->if_mtu;
1574 	ifd->ifi_metric = ifp->if_metric;
1575 	ifd->ifi_baudrate = ifp->if_baudrate;
1576 	ifd->ifi_hwassist = ifp->if_hwassist;
1577 	ifd->ifi_epoch = ifp->if_epoch;
1578 	ifd->ifi_lastchange = ifp->if_lastchange;
1579 
1580 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1581 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1582 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1583 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1584 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1585 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1586 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1587 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1588 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1589 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1590 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1591 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1592 }
1593 
1594 /*
1595  * Wrapper functions for struct ifnet address list locking macros.  These are
1596  * used by kernel modules to avoid encoding programming interface or binary
1597  * interface assumptions that may be violated when kernel-internal locking
1598  * approaches change.
1599  */
1600 void
1601 if_addr_rlock(struct ifnet *ifp)
1602 {
1603 
1604 	IF_ADDR_RLOCK(ifp);
1605 }
1606 
1607 void
1608 if_addr_runlock(struct ifnet *ifp)
1609 {
1610 
1611 	IF_ADDR_RUNLOCK(ifp);
1612 }
1613 
1614 void
1615 if_maddr_rlock(if_t ifp)
1616 {
1617 
1618 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1619 }
1620 
1621 void
1622 if_maddr_runlock(if_t ifp)
1623 {
1624 
1625 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1626 }
1627 
1628 /*
1629  * Initialization, destruction and refcounting functions for ifaddrs.
1630  */
1631 struct ifaddr *
1632 ifa_alloc(size_t size, int flags)
1633 {
1634 	struct ifaddr *ifa;
1635 
1636 	KASSERT(size >= sizeof(struct ifaddr),
1637 	    ("%s: invalid size %zu", __func__, size));
1638 
1639 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1640 	if (ifa == NULL)
1641 		return (NULL);
1642 
1643 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1644 		goto fail;
1645 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1646 		goto fail;
1647 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1648 		goto fail;
1649 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1650 		goto fail;
1651 
1652 	refcount_init(&ifa->ifa_refcnt, 1);
1653 
1654 	return (ifa);
1655 
1656 fail:
1657 	/* free(NULL) is okay */
1658 	counter_u64_free(ifa->ifa_opackets);
1659 	counter_u64_free(ifa->ifa_ipackets);
1660 	counter_u64_free(ifa->ifa_obytes);
1661 	counter_u64_free(ifa->ifa_ibytes);
1662 	free(ifa, M_IFADDR);
1663 
1664 	return (NULL);
1665 }
1666 
1667 void
1668 ifa_ref(struct ifaddr *ifa)
1669 {
1670 
1671 	refcount_acquire(&ifa->ifa_refcnt);
1672 }
1673 
1674 void
1675 ifa_free(struct ifaddr *ifa)
1676 {
1677 
1678 	if (refcount_release(&ifa->ifa_refcnt)) {
1679 		counter_u64_free(ifa->ifa_opackets);
1680 		counter_u64_free(ifa->ifa_ipackets);
1681 		counter_u64_free(ifa->ifa_obytes);
1682 		counter_u64_free(ifa->ifa_ibytes);
1683 		free(ifa, M_IFADDR);
1684 	}
1685 }
1686 
1687 static int
1688 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1689     struct sockaddr *ia)
1690 {
1691 	int error;
1692 	struct rt_addrinfo info;
1693 	struct sockaddr_dl null_sdl;
1694 	struct ifnet *ifp;
1695 
1696 	ifp = ifa->ifa_ifp;
1697 
1698 	bzero(&info, sizeof(info));
1699 	if (cmd != RTM_DELETE)
1700 		info.rti_ifp = V_loif;
1701 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1702 	info.rti_info[RTAX_DST] = ia;
1703 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1704 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1705 
1706 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1707 
1708 	if (error != 0)
1709 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1710 		    __func__, otype, if_name(ifp), error);
1711 
1712 	return (error);
1713 }
1714 
1715 int
1716 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1717 {
1718 
1719 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1720 }
1721 
1722 int
1723 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1724 {
1725 
1726 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1727 }
1728 
1729 int
1730 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1731 {
1732 
1733 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1734 }
1735 
1736 /*
1737  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1738  * structs used to represent other address families, it is necessary
1739  * to perform a different comparison.
1740  */
1741 
1742 #define	sa_dl_equal(a1, a2)	\
1743 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1744 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1745 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1746 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1747 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1748 
1749 /*
1750  * Locate an interface based on a complete address.
1751  */
1752 /*ARGSUSED*/
1753 static struct ifaddr *
1754 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1755 {
1756 	struct ifnet *ifp;
1757 	struct ifaddr *ifa;
1758 
1759 	IFNET_RLOCK_NOSLEEP();
1760 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1761 		IF_ADDR_RLOCK(ifp);
1762 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1763 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1764 				continue;
1765 			if (sa_equal(addr, ifa->ifa_addr)) {
1766 				if (getref)
1767 					ifa_ref(ifa);
1768 				IF_ADDR_RUNLOCK(ifp);
1769 				goto done;
1770 			}
1771 			/* IP6 doesn't have broadcast */
1772 			if ((ifp->if_flags & IFF_BROADCAST) &&
1773 			    ifa->ifa_broadaddr &&
1774 			    ifa->ifa_broadaddr->sa_len != 0 &&
1775 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1776 				if (getref)
1777 					ifa_ref(ifa);
1778 				IF_ADDR_RUNLOCK(ifp);
1779 				goto done;
1780 			}
1781 		}
1782 		IF_ADDR_RUNLOCK(ifp);
1783 	}
1784 	ifa = NULL;
1785 done:
1786 	IFNET_RUNLOCK_NOSLEEP();
1787 	return (ifa);
1788 }
1789 
1790 struct ifaddr *
1791 ifa_ifwithaddr(const struct sockaddr *addr)
1792 {
1793 
1794 	return (ifa_ifwithaddr_internal(addr, 1));
1795 }
1796 
1797 int
1798 ifa_ifwithaddr_check(const struct sockaddr *addr)
1799 {
1800 
1801 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1802 }
1803 
1804 /*
1805  * Locate an interface based on the broadcast address.
1806  */
1807 /* ARGSUSED */
1808 struct ifaddr *
1809 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1810 {
1811 	struct ifnet *ifp;
1812 	struct ifaddr *ifa;
1813 
1814 	IFNET_RLOCK_NOSLEEP();
1815 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1816 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1817 			continue;
1818 		IF_ADDR_RLOCK(ifp);
1819 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1820 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1821 				continue;
1822 			if ((ifp->if_flags & IFF_BROADCAST) &&
1823 			    ifa->ifa_broadaddr &&
1824 			    ifa->ifa_broadaddr->sa_len != 0 &&
1825 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1826 				ifa_ref(ifa);
1827 				IF_ADDR_RUNLOCK(ifp);
1828 				goto done;
1829 			}
1830 		}
1831 		IF_ADDR_RUNLOCK(ifp);
1832 	}
1833 	ifa = NULL;
1834 done:
1835 	IFNET_RUNLOCK_NOSLEEP();
1836 	return (ifa);
1837 }
1838 
1839 /*
1840  * Locate the point to point interface with a given destination address.
1841  */
1842 /*ARGSUSED*/
1843 struct ifaddr *
1844 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1845 {
1846 	struct ifnet *ifp;
1847 	struct ifaddr *ifa;
1848 
1849 	IFNET_RLOCK_NOSLEEP();
1850 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1851 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1852 			continue;
1853 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1854 			continue;
1855 		IF_ADDR_RLOCK(ifp);
1856 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1857 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1858 				continue;
1859 			if (ifa->ifa_dstaddr != NULL &&
1860 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1861 				ifa_ref(ifa);
1862 				IF_ADDR_RUNLOCK(ifp);
1863 				goto done;
1864 			}
1865 		}
1866 		IF_ADDR_RUNLOCK(ifp);
1867 	}
1868 	ifa = NULL;
1869 done:
1870 	IFNET_RUNLOCK_NOSLEEP();
1871 	return (ifa);
1872 }
1873 
1874 /*
1875  * Find an interface on a specific network.  If many, choice
1876  * is most specific found.
1877  */
1878 struct ifaddr *
1879 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1880 {
1881 	struct ifnet *ifp;
1882 	struct ifaddr *ifa;
1883 	struct ifaddr *ifa_maybe = NULL;
1884 	u_int af = addr->sa_family;
1885 	const char *addr_data = addr->sa_data, *cplim;
1886 
1887 	/*
1888 	 * AF_LINK addresses can be looked up directly by their index number,
1889 	 * so do that if we can.
1890 	 */
1891 	if (af == AF_LINK) {
1892 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1893 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1894 		return (ifaddr_byindex(sdl->sdl_index));
1895 	}
1896 
1897 	/*
1898 	 * Scan though each interface, looking for ones that have addresses
1899 	 * in this address family and the requested fib.  Maintain a reference
1900 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1901 	 * kept it stable when we move onto the next interface.
1902 	 */
1903 	IFNET_RLOCK_NOSLEEP();
1904 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1905 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1906 			continue;
1907 		IF_ADDR_RLOCK(ifp);
1908 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1909 			const char *cp, *cp2, *cp3;
1910 
1911 			if (ifa->ifa_addr->sa_family != af)
1912 next:				continue;
1913 			if (af == AF_INET &&
1914 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1915 				/*
1916 				 * This is a bit broken as it doesn't
1917 				 * take into account that the remote end may
1918 				 * be a single node in the network we are
1919 				 * looking for.
1920 				 * The trouble is that we don't know the
1921 				 * netmask for the remote end.
1922 				 */
1923 				if (ifa->ifa_dstaddr != NULL &&
1924 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1925 					ifa_ref(ifa);
1926 					IF_ADDR_RUNLOCK(ifp);
1927 					goto done;
1928 				}
1929 			} else {
1930 				/*
1931 				 * Scan all the bits in the ifa's address.
1932 				 * If a bit dissagrees with what we are
1933 				 * looking for, mask it with the netmask
1934 				 * to see if it really matters.
1935 				 * (A byte at a time)
1936 				 */
1937 				if (ifa->ifa_netmask == 0)
1938 					continue;
1939 				cp = addr_data;
1940 				cp2 = ifa->ifa_addr->sa_data;
1941 				cp3 = ifa->ifa_netmask->sa_data;
1942 				cplim = ifa->ifa_netmask->sa_len
1943 					+ (char *)ifa->ifa_netmask;
1944 				while (cp3 < cplim)
1945 					if ((*cp++ ^ *cp2++) & *cp3++)
1946 						goto next; /* next address! */
1947 				/*
1948 				 * If the netmask of what we just found
1949 				 * is more specific than what we had before
1950 				 * (if we had one), or if the virtual status
1951 				 * of new prefix is better than of the old one,
1952 				 * then remember the new one before continuing
1953 				 * to search for an even better one.
1954 				 */
1955 				if (ifa_maybe == NULL ||
1956 				    ifa_preferred(ifa_maybe, ifa) ||
1957 				    rn_refines((caddr_t)ifa->ifa_netmask,
1958 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1959 					if (ifa_maybe != NULL)
1960 						ifa_free(ifa_maybe);
1961 					ifa_maybe = ifa;
1962 					ifa_ref(ifa_maybe);
1963 				}
1964 			}
1965 		}
1966 		IF_ADDR_RUNLOCK(ifp);
1967 	}
1968 	ifa = ifa_maybe;
1969 	ifa_maybe = NULL;
1970 done:
1971 	IFNET_RUNLOCK_NOSLEEP();
1972 	if (ifa_maybe != NULL)
1973 		ifa_free(ifa_maybe);
1974 	return (ifa);
1975 }
1976 
1977 /*
1978  * Find an interface address specific to an interface best matching
1979  * a given address.
1980  */
1981 struct ifaddr *
1982 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1983 {
1984 	struct ifaddr *ifa;
1985 	const char *cp, *cp2, *cp3;
1986 	char *cplim;
1987 	struct ifaddr *ifa_maybe = NULL;
1988 	u_int af = addr->sa_family;
1989 
1990 	if (af >= AF_MAX)
1991 		return (NULL);
1992 	IF_ADDR_RLOCK(ifp);
1993 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1994 		if (ifa->ifa_addr->sa_family != af)
1995 			continue;
1996 		if (ifa_maybe == NULL)
1997 			ifa_maybe = ifa;
1998 		if (ifa->ifa_netmask == 0) {
1999 			if (sa_equal(addr, ifa->ifa_addr) ||
2000 			    (ifa->ifa_dstaddr &&
2001 			    sa_equal(addr, ifa->ifa_dstaddr)))
2002 				goto done;
2003 			continue;
2004 		}
2005 		if (ifp->if_flags & IFF_POINTOPOINT) {
2006 			if (sa_equal(addr, ifa->ifa_dstaddr))
2007 				goto done;
2008 		} else {
2009 			cp = addr->sa_data;
2010 			cp2 = ifa->ifa_addr->sa_data;
2011 			cp3 = ifa->ifa_netmask->sa_data;
2012 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2013 			for (; cp3 < cplim; cp3++)
2014 				if ((*cp++ ^ *cp2++) & *cp3)
2015 					break;
2016 			if (cp3 == cplim)
2017 				goto done;
2018 		}
2019 	}
2020 	ifa = ifa_maybe;
2021 done:
2022 	if (ifa != NULL)
2023 		ifa_ref(ifa);
2024 	IF_ADDR_RUNLOCK(ifp);
2025 	return (ifa);
2026 }
2027 
2028 /*
2029  * See whether new ifa is better than current one:
2030  * 1) A non-virtual one is preferred over virtual.
2031  * 2) A virtual in master state preferred over any other state.
2032  *
2033  * Used in several address selecting functions.
2034  */
2035 int
2036 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2037 {
2038 
2039 	return (cur->ifa_carp && (!next->ifa_carp ||
2040 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2041 }
2042 
2043 #include <net/if_llatbl.h>
2044 
2045 /*
2046  * Default action when installing a route with a Link Level gateway.
2047  * Lookup an appropriate real ifa to point to.
2048  * This should be moved to /sys/net/link.c eventually.
2049  */
2050 static void
2051 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2052 {
2053 	struct ifaddr *ifa, *oifa;
2054 	struct sockaddr *dst;
2055 	struct ifnet *ifp;
2056 
2057 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2058 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2059 		return;
2060 	ifa = ifaof_ifpforaddr(dst, ifp);
2061 	if (ifa) {
2062 		oifa = rt->rt_ifa;
2063 		rt->rt_ifa = ifa;
2064 		ifa_free(oifa);
2065 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2066 			ifa->ifa_rtrequest(cmd, rt, info);
2067 	}
2068 }
2069 
2070 struct sockaddr_dl *
2071 link_alloc_sdl(size_t size, int flags)
2072 {
2073 
2074 	return (malloc(size, M_TEMP, flags));
2075 }
2076 
2077 void
2078 link_free_sdl(struct sockaddr *sa)
2079 {
2080 	free(sa, M_TEMP);
2081 }
2082 
2083 /*
2084  * Fills in given sdl with interface basic info.
2085  * Returns pointer to filled sdl.
2086  */
2087 struct sockaddr_dl *
2088 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2089 {
2090 	struct sockaddr_dl *sdl;
2091 
2092 	sdl = (struct sockaddr_dl *)paddr;
2093 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2094 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2095 	sdl->sdl_family = AF_LINK;
2096 	sdl->sdl_index = ifp->if_index;
2097 	sdl->sdl_type = iftype;
2098 
2099 	return (sdl);
2100 }
2101 
2102 /*
2103  * Mark an interface down and notify protocols of
2104  * the transition.
2105  */
2106 static void
2107 if_unroute(struct ifnet *ifp, int flag, int fam)
2108 {
2109 	struct ifaddr *ifa;
2110 
2111 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2112 
2113 	ifp->if_flags &= ~flag;
2114 	getmicrotime(&ifp->if_lastchange);
2115 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2116 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2117 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2118 	ifp->if_qflush(ifp);
2119 
2120 	if (ifp->if_carp)
2121 		(*carp_linkstate_p)(ifp);
2122 	rt_ifmsg(ifp);
2123 }
2124 
2125 /*
2126  * Mark an interface up and notify protocols of
2127  * the transition.
2128  */
2129 static void
2130 if_route(struct ifnet *ifp, int flag, int fam)
2131 {
2132 	struct ifaddr *ifa;
2133 
2134 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2135 
2136 	ifp->if_flags |= flag;
2137 	getmicrotime(&ifp->if_lastchange);
2138 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2139 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2140 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2141 	if (ifp->if_carp)
2142 		(*carp_linkstate_p)(ifp);
2143 	rt_ifmsg(ifp);
2144 #ifdef INET6
2145 	in6_if_up(ifp);
2146 #endif
2147 }
2148 
2149 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2150 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2151 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2152 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2153 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2154 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2155 void	*(*vlan_cookie_p)(struct ifnet *);
2156 
2157 /*
2158  * Handle a change in the interface link state. To avoid LORs
2159  * between driver lock and upper layer locks, as well as possible
2160  * recursions, we post event to taskqueue, and all job
2161  * is done in static do_link_state_change().
2162  */
2163 void
2164 if_link_state_change(struct ifnet *ifp, int link_state)
2165 {
2166 	/* Return if state hasn't changed. */
2167 	if (ifp->if_link_state == link_state)
2168 		return;
2169 
2170 	ifp->if_link_state = link_state;
2171 
2172 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2173 }
2174 
2175 static void
2176 do_link_state_change(void *arg, int pending)
2177 {
2178 	struct ifnet *ifp = (struct ifnet *)arg;
2179 	int link_state = ifp->if_link_state;
2180 	CURVNET_SET(ifp->if_vnet);
2181 
2182 	/* Notify that the link state has changed. */
2183 	rt_ifmsg(ifp);
2184 	if (ifp->if_vlantrunk != NULL)
2185 		(*vlan_link_state_p)(ifp);
2186 
2187 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2188 	    ifp->if_l2com != NULL)
2189 		(*ng_ether_link_state_p)(ifp, link_state);
2190 	if (ifp->if_carp)
2191 		(*carp_linkstate_p)(ifp);
2192 	if (ifp->if_bridge)
2193 		(*bridge_linkstate_p)(ifp);
2194 	if (ifp->if_lagg)
2195 		(*lagg_linkstate_p)(ifp, link_state);
2196 
2197 	if (IS_DEFAULT_VNET(curvnet))
2198 		devctl_notify("IFNET", ifp->if_xname,
2199 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2200 		    NULL);
2201 	if (pending > 1)
2202 		if_printf(ifp, "%d link states coalesced\n", pending);
2203 	if (log_link_state_change)
2204 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2205 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2206 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2207 	CURVNET_RESTORE();
2208 }
2209 
2210 /*
2211  * Mark an interface down and notify protocols of
2212  * the transition.
2213  */
2214 void
2215 if_down(struct ifnet *ifp)
2216 {
2217 
2218 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2219 }
2220 
2221 /*
2222  * Mark an interface up and notify protocols of
2223  * the transition.
2224  */
2225 void
2226 if_up(struct ifnet *ifp)
2227 {
2228 
2229 	if_route(ifp, IFF_UP, AF_UNSPEC);
2230 }
2231 
2232 /*
2233  * Flush an interface queue.
2234  */
2235 void
2236 if_qflush(struct ifnet *ifp)
2237 {
2238 	struct mbuf *m, *n;
2239 	struct ifaltq *ifq;
2240 
2241 	ifq = &ifp->if_snd;
2242 	IFQ_LOCK(ifq);
2243 #ifdef ALTQ
2244 	if (ALTQ_IS_ENABLED(ifq))
2245 		ALTQ_PURGE(ifq);
2246 #endif
2247 	n = ifq->ifq_head;
2248 	while ((m = n) != NULL) {
2249 		n = m->m_nextpkt;
2250 		m_freem(m);
2251 	}
2252 	ifq->ifq_head = 0;
2253 	ifq->ifq_tail = 0;
2254 	ifq->ifq_len = 0;
2255 	IFQ_UNLOCK(ifq);
2256 }
2257 
2258 /*
2259  * Map interface name to interface structure pointer, with or without
2260  * returning a reference.
2261  */
2262 struct ifnet *
2263 ifunit_ref(const char *name)
2264 {
2265 	struct ifnet *ifp;
2266 
2267 	IFNET_RLOCK_NOSLEEP();
2268 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2269 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2270 		    !(ifp->if_flags & IFF_DYING))
2271 			break;
2272 	}
2273 	if (ifp != NULL)
2274 		if_ref(ifp);
2275 	IFNET_RUNLOCK_NOSLEEP();
2276 	return (ifp);
2277 }
2278 
2279 struct ifnet *
2280 ifunit(const char *name)
2281 {
2282 	struct ifnet *ifp;
2283 
2284 	IFNET_RLOCK_NOSLEEP();
2285 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2286 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2287 			break;
2288 	}
2289 	IFNET_RUNLOCK_NOSLEEP();
2290 	return (ifp);
2291 }
2292 
2293 /*
2294  * Hardware specific interface ioctls.
2295  */
2296 static int
2297 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2298 {
2299 	struct ifreq *ifr;
2300 	int error = 0;
2301 	int new_flags, temp_flags;
2302 	size_t namelen, onamelen;
2303 	size_t descrlen;
2304 	char *descrbuf, *odescrbuf;
2305 	char new_name[IFNAMSIZ];
2306 	struct ifaddr *ifa;
2307 	struct sockaddr_dl *sdl;
2308 
2309 	ifr = (struct ifreq *)data;
2310 	switch (cmd) {
2311 	case SIOCGIFINDEX:
2312 		ifr->ifr_index = ifp->if_index;
2313 		break;
2314 
2315 	case SIOCGIFFLAGS:
2316 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2317 		ifr->ifr_flags = temp_flags & 0xffff;
2318 		ifr->ifr_flagshigh = temp_flags >> 16;
2319 		break;
2320 
2321 	case SIOCGIFCAP:
2322 		ifr->ifr_reqcap = ifp->if_capabilities;
2323 		ifr->ifr_curcap = ifp->if_capenable;
2324 		break;
2325 
2326 #ifdef MAC
2327 	case SIOCGIFMAC:
2328 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2329 		break;
2330 #endif
2331 
2332 	case SIOCGIFMETRIC:
2333 		ifr->ifr_metric = ifp->if_metric;
2334 		break;
2335 
2336 	case SIOCGIFMTU:
2337 		ifr->ifr_mtu = ifp->if_mtu;
2338 		break;
2339 
2340 	case SIOCGIFPHYS:
2341 		/* XXXGL: did this ever worked? */
2342 		ifr->ifr_phys = 0;
2343 		break;
2344 
2345 	case SIOCGIFDESCR:
2346 		error = 0;
2347 		sx_slock(&ifdescr_sx);
2348 		if (ifp->if_description == NULL)
2349 			error = ENOMSG;
2350 		else {
2351 			/* space for terminating nul */
2352 			descrlen = strlen(ifp->if_description) + 1;
2353 			if (ifr->ifr_buffer.length < descrlen)
2354 				ifr->ifr_buffer.buffer = NULL;
2355 			else
2356 				error = copyout(ifp->if_description,
2357 				    ifr->ifr_buffer.buffer, descrlen);
2358 			ifr->ifr_buffer.length = descrlen;
2359 		}
2360 		sx_sunlock(&ifdescr_sx);
2361 		break;
2362 
2363 	case SIOCSIFDESCR:
2364 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2365 		if (error)
2366 			return (error);
2367 
2368 		/*
2369 		 * Copy only (length-1) bytes to make sure that
2370 		 * if_description is always nul terminated.  The
2371 		 * length parameter is supposed to count the
2372 		 * terminating nul in.
2373 		 */
2374 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2375 			return (ENAMETOOLONG);
2376 		else if (ifr->ifr_buffer.length == 0)
2377 			descrbuf = NULL;
2378 		else {
2379 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2380 			    M_WAITOK | M_ZERO);
2381 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2382 			    ifr->ifr_buffer.length - 1);
2383 			if (error) {
2384 				free(descrbuf, M_IFDESCR);
2385 				break;
2386 			}
2387 		}
2388 
2389 		sx_xlock(&ifdescr_sx);
2390 		odescrbuf = ifp->if_description;
2391 		ifp->if_description = descrbuf;
2392 		sx_xunlock(&ifdescr_sx);
2393 
2394 		getmicrotime(&ifp->if_lastchange);
2395 		free(odescrbuf, M_IFDESCR);
2396 		break;
2397 
2398 	case SIOCGIFFIB:
2399 		ifr->ifr_fib = ifp->if_fib;
2400 		break;
2401 
2402 	case SIOCSIFFIB:
2403 		error = priv_check(td, PRIV_NET_SETIFFIB);
2404 		if (error)
2405 			return (error);
2406 		if (ifr->ifr_fib >= rt_numfibs)
2407 			return (EINVAL);
2408 
2409 		ifp->if_fib = ifr->ifr_fib;
2410 		break;
2411 
2412 	case SIOCSIFFLAGS:
2413 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2414 		if (error)
2415 			return (error);
2416 		/*
2417 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2418 		 * check, so we don't need special handling here yet.
2419 		 */
2420 		new_flags = (ifr->ifr_flags & 0xffff) |
2421 		    (ifr->ifr_flagshigh << 16);
2422 		if (ifp->if_flags & IFF_UP &&
2423 		    (new_flags & IFF_UP) == 0) {
2424 			if_down(ifp);
2425 		} else if (new_flags & IFF_UP &&
2426 		    (ifp->if_flags & IFF_UP) == 0) {
2427 			if_up(ifp);
2428 		}
2429 		/* See if permanently promiscuous mode bit is about to flip */
2430 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2431 			if (new_flags & IFF_PPROMISC)
2432 				ifp->if_flags |= IFF_PROMISC;
2433 			else if (ifp->if_pcount == 0)
2434 				ifp->if_flags &= ~IFF_PROMISC;
2435 			if (log_promisc_mode_change)
2436                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2437                                     ifp->if_xname,
2438                                     ((new_flags & IFF_PPROMISC) ?
2439                                      "enabled" : "disabled"));
2440 		}
2441 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2442 			(new_flags &~ IFF_CANTCHANGE);
2443 		if (ifp->if_ioctl) {
2444 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2445 		}
2446 		getmicrotime(&ifp->if_lastchange);
2447 		break;
2448 
2449 	case SIOCSIFCAP:
2450 		error = priv_check(td, PRIV_NET_SETIFCAP);
2451 		if (error)
2452 			return (error);
2453 		if (ifp->if_ioctl == NULL)
2454 			return (EOPNOTSUPP);
2455 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2456 			return (EINVAL);
2457 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2458 		if (error == 0)
2459 			getmicrotime(&ifp->if_lastchange);
2460 		break;
2461 
2462 #ifdef MAC
2463 	case SIOCSIFMAC:
2464 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2465 		break;
2466 #endif
2467 
2468 	case SIOCSIFNAME:
2469 		error = priv_check(td, PRIV_NET_SETIFNAME);
2470 		if (error)
2471 			return (error);
2472 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2473 		if (error != 0)
2474 			return (error);
2475 		if (new_name[0] == '\0')
2476 			return (EINVAL);
2477 		if (new_name[IFNAMSIZ-1] != '\0') {
2478 			new_name[IFNAMSIZ-1] = '\0';
2479 			if (strlen(new_name) == IFNAMSIZ-1)
2480 				return (EINVAL);
2481 		}
2482 		if (ifunit(new_name) != NULL)
2483 			return (EEXIST);
2484 
2485 		/*
2486 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2487 		 * and there are numerous other races with the
2488 		 * ifunit() checks not being atomic with namespace
2489 		 * changes (renames, vmoves, if_attach, etc).
2490 		 */
2491 		ifp->if_flags |= IFF_RENAMING;
2492 
2493 		/* Announce the departure of the interface. */
2494 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2495 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2496 
2497 		log(LOG_INFO, "%s: changing name to '%s'\n",
2498 		    ifp->if_xname, new_name);
2499 
2500 		IF_ADDR_WLOCK(ifp);
2501 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2502 		ifa = ifp->if_addr;
2503 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2504 		namelen = strlen(new_name);
2505 		onamelen = sdl->sdl_nlen;
2506 		/*
2507 		 * Move the address if needed.  This is safe because we
2508 		 * allocate space for a name of length IFNAMSIZ when we
2509 		 * create this in if_attach().
2510 		 */
2511 		if (namelen != onamelen) {
2512 			bcopy(sdl->sdl_data + onamelen,
2513 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2514 		}
2515 		bcopy(new_name, sdl->sdl_data, namelen);
2516 		sdl->sdl_nlen = namelen;
2517 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2518 		bzero(sdl->sdl_data, onamelen);
2519 		while (namelen != 0)
2520 			sdl->sdl_data[--namelen] = 0xff;
2521 		IF_ADDR_WUNLOCK(ifp);
2522 
2523 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2524 		/* Announce the return of the interface. */
2525 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2526 
2527 		ifp->if_flags &= ~IFF_RENAMING;
2528 		break;
2529 
2530 #ifdef VIMAGE
2531 	case SIOCSIFVNET:
2532 		error = priv_check(td, PRIV_NET_SETIFVNET);
2533 		if (error)
2534 			return (error);
2535 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2536 		break;
2537 #endif
2538 
2539 	case SIOCSIFMETRIC:
2540 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2541 		if (error)
2542 			return (error);
2543 		ifp->if_metric = ifr->ifr_metric;
2544 		getmicrotime(&ifp->if_lastchange);
2545 		break;
2546 
2547 	case SIOCSIFPHYS:
2548 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2549 		if (error)
2550 			return (error);
2551 		if (ifp->if_ioctl == NULL)
2552 			return (EOPNOTSUPP);
2553 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2554 		if (error == 0)
2555 			getmicrotime(&ifp->if_lastchange);
2556 		break;
2557 
2558 	case SIOCSIFMTU:
2559 	{
2560 		u_long oldmtu = ifp->if_mtu;
2561 
2562 		error = priv_check(td, PRIV_NET_SETIFMTU);
2563 		if (error)
2564 			return (error);
2565 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2566 			return (EINVAL);
2567 		if (ifp->if_ioctl == NULL)
2568 			return (EOPNOTSUPP);
2569 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2570 		if (error == 0) {
2571 			getmicrotime(&ifp->if_lastchange);
2572 			rt_ifmsg(ifp);
2573 		}
2574 		/*
2575 		 * If the link MTU changed, do network layer specific procedure.
2576 		 */
2577 		if (ifp->if_mtu != oldmtu) {
2578 #ifdef INET6
2579 			nd6_setmtu(ifp);
2580 #endif
2581 			rt_updatemtu(ifp);
2582 		}
2583 		break;
2584 	}
2585 
2586 	case SIOCADDMULTI:
2587 	case SIOCDELMULTI:
2588 		if (cmd == SIOCADDMULTI)
2589 			error = priv_check(td, PRIV_NET_ADDMULTI);
2590 		else
2591 			error = priv_check(td, PRIV_NET_DELMULTI);
2592 		if (error)
2593 			return (error);
2594 
2595 		/* Don't allow group membership on non-multicast interfaces. */
2596 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2597 			return (EOPNOTSUPP);
2598 
2599 		/* Don't let users screw up protocols' entries. */
2600 		if (ifr->ifr_addr.sa_family != AF_LINK)
2601 			return (EINVAL);
2602 
2603 		if (cmd == SIOCADDMULTI) {
2604 			struct ifmultiaddr *ifma;
2605 
2606 			/*
2607 			 * Userland is only permitted to join groups once
2608 			 * via the if_addmulti() KPI, because it cannot hold
2609 			 * struct ifmultiaddr * between calls. It may also
2610 			 * lose a race while we check if the membership
2611 			 * already exists.
2612 			 */
2613 			IF_ADDR_RLOCK(ifp);
2614 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2615 			IF_ADDR_RUNLOCK(ifp);
2616 			if (ifma != NULL)
2617 				error = EADDRINUSE;
2618 			else
2619 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2620 		} else {
2621 			error = if_delmulti(ifp, &ifr->ifr_addr);
2622 		}
2623 		if (error == 0)
2624 			getmicrotime(&ifp->if_lastchange);
2625 		break;
2626 
2627 	case SIOCSIFPHYADDR:
2628 	case SIOCDIFPHYADDR:
2629 #ifdef INET6
2630 	case SIOCSIFPHYADDR_IN6:
2631 #endif
2632 	case SIOCSIFMEDIA:
2633 	case SIOCSIFGENERIC:
2634 		error = priv_check(td, PRIV_NET_HWIOCTL);
2635 		if (error)
2636 			return (error);
2637 		if (ifp->if_ioctl == NULL)
2638 			return (EOPNOTSUPP);
2639 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2640 		if (error == 0)
2641 			getmicrotime(&ifp->if_lastchange);
2642 		break;
2643 
2644 	case SIOCGIFSTATUS:
2645 	case SIOCGIFPSRCADDR:
2646 	case SIOCGIFPDSTADDR:
2647 	case SIOCGIFMEDIA:
2648 	case SIOCGIFXMEDIA:
2649 	case SIOCGIFGENERIC:
2650 		if (ifp->if_ioctl == NULL)
2651 			return (EOPNOTSUPP);
2652 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2653 		break;
2654 
2655 	case SIOCSIFLLADDR:
2656 		error = priv_check(td, PRIV_NET_SETLLADDR);
2657 		if (error)
2658 			return (error);
2659 		error = if_setlladdr(ifp,
2660 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2661 		break;
2662 
2663 	case SIOCAIFGROUP:
2664 	{
2665 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2666 
2667 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2668 		if (error)
2669 			return (error);
2670 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2671 			return (error);
2672 		break;
2673 	}
2674 
2675 	case SIOCGIFGROUP:
2676 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2677 			return (error);
2678 		break;
2679 
2680 	case SIOCDIFGROUP:
2681 	{
2682 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2683 
2684 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2685 		if (error)
2686 			return (error);
2687 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2688 			return (error);
2689 		break;
2690 	}
2691 
2692 	default:
2693 		error = ENOIOCTL;
2694 		break;
2695 	}
2696 	return (error);
2697 }
2698 
2699 #ifdef COMPAT_FREEBSD32
2700 struct ifconf32 {
2701 	int32_t	ifc_len;
2702 	union {
2703 		uint32_t	ifcu_buf;
2704 		uint32_t	ifcu_req;
2705 	} ifc_ifcu;
2706 };
2707 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2708 #endif
2709 
2710 /*
2711  * Interface ioctls.
2712  */
2713 int
2714 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2715 {
2716 	struct ifnet *ifp;
2717 	struct ifreq *ifr;
2718 	int error;
2719 	int oif_flags;
2720 #ifdef VIMAGE
2721 	int shutdown;
2722 #endif
2723 
2724 	CURVNET_SET(so->so_vnet);
2725 #ifdef VIMAGE
2726 	/* Make sure the VNET is stable. */
2727 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2728 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2729 	if (shutdown) {
2730 		CURVNET_RESTORE();
2731 		return (EBUSY);
2732 	}
2733 #endif
2734 
2735 
2736 	switch (cmd) {
2737 	case SIOCGIFCONF:
2738 		error = ifconf(cmd, data);
2739 		CURVNET_RESTORE();
2740 		return (error);
2741 
2742 #ifdef COMPAT_FREEBSD32
2743 	case SIOCGIFCONF32:
2744 		{
2745 			struct ifconf32 *ifc32;
2746 			struct ifconf ifc;
2747 
2748 			ifc32 = (struct ifconf32 *)data;
2749 			ifc.ifc_len = ifc32->ifc_len;
2750 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2751 
2752 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2753 			CURVNET_RESTORE();
2754 			if (error == 0)
2755 				ifc32->ifc_len = ifc.ifc_len;
2756 			return (error);
2757 		}
2758 #endif
2759 	}
2760 	ifr = (struct ifreq *)data;
2761 
2762 	switch (cmd) {
2763 #ifdef VIMAGE
2764 	case SIOCSIFRVNET:
2765 		error = priv_check(td, PRIV_NET_SETIFVNET);
2766 		if (error == 0)
2767 			error = if_vmove_reclaim(td, ifr->ifr_name,
2768 			    ifr->ifr_jid);
2769 		CURVNET_RESTORE();
2770 		return (error);
2771 #endif
2772 	case SIOCIFCREATE:
2773 	case SIOCIFCREATE2:
2774 		error = priv_check(td, PRIV_NET_IFCREATE);
2775 		if (error == 0)
2776 			error = if_clone_create(ifr->ifr_name,
2777 			    sizeof(ifr->ifr_name),
2778 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2779 		CURVNET_RESTORE();
2780 		return (error);
2781 	case SIOCIFDESTROY:
2782 		error = priv_check(td, PRIV_NET_IFDESTROY);
2783 		if (error == 0)
2784 			error = if_clone_destroy(ifr->ifr_name);
2785 		CURVNET_RESTORE();
2786 		return (error);
2787 
2788 	case SIOCIFGCLONERS:
2789 		error = if_clone_list((struct if_clonereq *)data);
2790 		CURVNET_RESTORE();
2791 		return (error);
2792 	case SIOCGIFGMEMB:
2793 		error = if_getgroupmembers((struct ifgroupreq *)data);
2794 		CURVNET_RESTORE();
2795 		return (error);
2796 #if defined(INET) || defined(INET6)
2797 	case SIOCSVH:
2798 	case SIOCGVH:
2799 		if (carp_ioctl_p == NULL)
2800 			error = EPROTONOSUPPORT;
2801 		else
2802 			error = (*carp_ioctl_p)(ifr, cmd, td);
2803 		CURVNET_RESTORE();
2804 		return (error);
2805 #endif
2806 	}
2807 
2808 	ifp = ifunit_ref(ifr->ifr_name);
2809 	if (ifp == NULL) {
2810 		CURVNET_RESTORE();
2811 		return (ENXIO);
2812 	}
2813 
2814 	error = ifhwioctl(cmd, ifp, data, td);
2815 	if (error != ENOIOCTL) {
2816 		if_rele(ifp);
2817 		CURVNET_RESTORE();
2818 		return (error);
2819 	}
2820 
2821 	oif_flags = ifp->if_flags;
2822 	if (so->so_proto == NULL) {
2823 		if_rele(ifp);
2824 		CURVNET_RESTORE();
2825 		return (EOPNOTSUPP);
2826 	}
2827 
2828 	/*
2829 	 * Pass the request on to the socket control method, and if the
2830 	 * latter returns EOPNOTSUPP, directly to the interface.
2831 	 *
2832 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2833 	 * trust SIOCSIFADDR et al to come from an already privileged
2834 	 * layer, and do not perform any credentials checks or input
2835 	 * validation.
2836 	 */
2837 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2838 	    ifp, td));
2839 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2840 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2841 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2842 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2843 
2844 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2845 #ifdef INET6
2846 		if (ifp->if_flags & IFF_UP)
2847 			in6_if_up(ifp);
2848 #endif
2849 	}
2850 	if_rele(ifp);
2851 	CURVNET_RESTORE();
2852 	return (error);
2853 }
2854 
2855 /*
2856  * The code common to handling reference counted flags,
2857  * e.g., in ifpromisc() and if_allmulti().
2858  * The "pflag" argument can specify a permanent mode flag to check,
2859  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2860  *
2861  * Only to be used on stack-owned flags, not driver-owned flags.
2862  */
2863 static int
2864 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2865 {
2866 	struct ifreq ifr;
2867 	int error;
2868 	int oldflags, oldcount;
2869 
2870 	/* Sanity checks to catch programming errors */
2871 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2872 	    ("%s: setting driver-owned flag %d", __func__, flag));
2873 
2874 	if (onswitch)
2875 		KASSERT(*refcount >= 0,
2876 		    ("%s: increment negative refcount %d for flag %d",
2877 		    __func__, *refcount, flag));
2878 	else
2879 		KASSERT(*refcount > 0,
2880 		    ("%s: decrement non-positive refcount %d for flag %d",
2881 		    __func__, *refcount, flag));
2882 
2883 	/* In case this mode is permanent, just touch refcount */
2884 	if (ifp->if_flags & pflag) {
2885 		*refcount += onswitch ? 1 : -1;
2886 		return (0);
2887 	}
2888 
2889 	/* Save ifnet parameters for if_ioctl() may fail */
2890 	oldcount = *refcount;
2891 	oldflags = ifp->if_flags;
2892 
2893 	/*
2894 	 * See if we aren't the only and touching refcount is enough.
2895 	 * Actually toggle interface flag if we are the first or last.
2896 	 */
2897 	if (onswitch) {
2898 		if ((*refcount)++)
2899 			return (0);
2900 		ifp->if_flags |= flag;
2901 	} else {
2902 		if (--(*refcount))
2903 			return (0);
2904 		ifp->if_flags &= ~flag;
2905 	}
2906 
2907 	/* Call down the driver since we've changed interface flags */
2908 	if (ifp->if_ioctl == NULL) {
2909 		error = EOPNOTSUPP;
2910 		goto recover;
2911 	}
2912 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2913 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2914 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2915 	if (error)
2916 		goto recover;
2917 	/* Notify userland that interface flags have changed */
2918 	rt_ifmsg(ifp);
2919 	return (0);
2920 
2921 recover:
2922 	/* Recover after driver error */
2923 	*refcount = oldcount;
2924 	ifp->if_flags = oldflags;
2925 	return (error);
2926 }
2927 
2928 /*
2929  * Set/clear promiscuous mode on interface ifp based on the truth value
2930  * of pswitch.  The calls are reference counted so that only the first
2931  * "on" request actually has an effect, as does the final "off" request.
2932  * Results are undefined if the "off" and "on" requests are not matched.
2933  */
2934 int
2935 ifpromisc(struct ifnet *ifp, int pswitch)
2936 {
2937 	int error;
2938 	int oldflags = ifp->if_flags;
2939 
2940 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2941 			   &ifp->if_pcount, pswitch);
2942 	/* If promiscuous mode status has changed, log a message */
2943 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2944             log_promisc_mode_change)
2945 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2946 		    ifp->if_xname,
2947 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2948 	return (error);
2949 }
2950 
2951 /*
2952  * Return interface configuration
2953  * of system.  List may be used
2954  * in later ioctl's (above) to get
2955  * other information.
2956  */
2957 /*ARGSUSED*/
2958 static int
2959 ifconf(u_long cmd, caddr_t data)
2960 {
2961 	struct ifconf *ifc = (struct ifconf *)data;
2962 	struct ifnet *ifp;
2963 	struct ifaddr *ifa;
2964 	struct ifreq ifr;
2965 	struct sbuf *sb;
2966 	int error, full = 0, valid_len, max_len;
2967 
2968 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2969 	max_len = MAXPHYS - 1;
2970 
2971 	/* Prevent hostile input from being able to crash the system */
2972 	if (ifc->ifc_len <= 0)
2973 		return (EINVAL);
2974 
2975 again:
2976 	if (ifc->ifc_len <= max_len) {
2977 		max_len = ifc->ifc_len;
2978 		full = 1;
2979 	}
2980 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2981 	max_len = 0;
2982 	valid_len = 0;
2983 
2984 	IFNET_RLOCK();
2985 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2986 		int addrs;
2987 
2988 		/*
2989 		 * Zero the ifr_name buffer to make sure we don't
2990 		 * disclose the contents of the stack.
2991 		 */
2992 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2993 
2994 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2995 		    >= sizeof(ifr.ifr_name)) {
2996 			sbuf_delete(sb);
2997 			IFNET_RUNLOCK();
2998 			return (ENAMETOOLONG);
2999 		}
3000 
3001 		addrs = 0;
3002 		IF_ADDR_RLOCK(ifp);
3003 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3004 			struct sockaddr *sa = ifa->ifa_addr;
3005 
3006 			if (prison_if(curthread->td_ucred, sa) != 0)
3007 				continue;
3008 			addrs++;
3009 			if (sa->sa_len <= sizeof(*sa)) {
3010 				ifr.ifr_addr = *sa;
3011 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3012 				max_len += sizeof(ifr);
3013 			} else {
3014 				sbuf_bcat(sb, &ifr,
3015 				    offsetof(struct ifreq, ifr_addr));
3016 				max_len += offsetof(struct ifreq, ifr_addr);
3017 				sbuf_bcat(sb, sa, sa->sa_len);
3018 				max_len += sa->sa_len;
3019 			}
3020 
3021 			if (sbuf_error(sb) == 0)
3022 				valid_len = sbuf_len(sb);
3023 		}
3024 		IF_ADDR_RUNLOCK(ifp);
3025 		if (addrs == 0) {
3026 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3027 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3028 			max_len += sizeof(ifr);
3029 
3030 			if (sbuf_error(sb) == 0)
3031 				valid_len = sbuf_len(sb);
3032 		}
3033 	}
3034 	IFNET_RUNLOCK();
3035 
3036 	/*
3037 	 * If we didn't allocate enough space (uncommon), try again.  If
3038 	 * we have already allocated as much space as we are allowed,
3039 	 * return what we've got.
3040 	 */
3041 	if (valid_len != max_len && !full) {
3042 		sbuf_delete(sb);
3043 		goto again;
3044 	}
3045 
3046 	ifc->ifc_len = valid_len;
3047 	sbuf_finish(sb);
3048 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3049 	sbuf_delete(sb);
3050 	return (error);
3051 }
3052 
3053 /*
3054  * Just like ifpromisc(), but for all-multicast-reception mode.
3055  */
3056 int
3057 if_allmulti(struct ifnet *ifp, int onswitch)
3058 {
3059 
3060 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3061 }
3062 
3063 struct ifmultiaddr *
3064 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3065 {
3066 	struct ifmultiaddr *ifma;
3067 
3068 	IF_ADDR_LOCK_ASSERT(ifp);
3069 
3070 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3071 		if (sa->sa_family == AF_LINK) {
3072 			if (sa_dl_equal(ifma->ifma_addr, sa))
3073 				break;
3074 		} else {
3075 			if (sa_equal(ifma->ifma_addr, sa))
3076 				break;
3077 		}
3078 	}
3079 
3080 	return ifma;
3081 }
3082 
3083 /*
3084  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3085  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3086  * the ifnet multicast address list here, so the caller must do that and
3087  * other setup work (such as notifying the device driver).  The reference
3088  * count is initialized to 1.
3089  */
3090 static struct ifmultiaddr *
3091 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3092     int mflags)
3093 {
3094 	struct ifmultiaddr *ifma;
3095 	struct sockaddr *dupsa;
3096 
3097 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3098 	    M_ZERO);
3099 	if (ifma == NULL)
3100 		return (NULL);
3101 
3102 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3103 	if (dupsa == NULL) {
3104 		free(ifma, M_IFMADDR);
3105 		return (NULL);
3106 	}
3107 	bcopy(sa, dupsa, sa->sa_len);
3108 	ifma->ifma_addr = dupsa;
3109 
3110 	ifma->ifma_ifp = ifp;
3111 	ifma->ifma_refcount = 1;
3112 	ifma->ifma_protospec = NULL;
3113 
3114 	if (llsa == NULL) {
3115 		ifma->ifma_lladdr = NULL;
3116 		return (ifma);
3117 	}
3118 
3119 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3120 	if (dupsa == NULL) {
3121 		free(ifma->ifma_addr, M_IFMADDR);
3122 		free(ifma, M_IFMADDR);
3123 		return (NULL);
3124 	}
3125 	bcopy(llsa, dupsa, llsa->sa_len);
3126 	ifma->ifma_lladdr = dupsa;
3127 
3128 	return (ifma);
3129 }
3130 
3131 /*
3132  * if_freemulti: free ifmultiaddr structure and possibly attached related
3133  * addresses.  The caller is responsible for implementing reference
3134  * counting, notifying the driver, handling routing messages, and releasing
3135  * any dependent link layer state.
3136  */
3137 static void
3138 if_freemulti(struct ifmultiaddr *ifma)
3139 {
3140 
3141 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3142 	    ifma->ifma_refcount));
3143 
3144 	if (ifma->ifma_lladdr != NULL)
3145 		free(ifma->ifma_lladdr, M_IFMADDR);
3146 	free(ifma->ifma_addr, M_IFMADDR);
3147 	free(ifma, M_IFMADDR);
3148 }
3149 
3150 /*
3151  * Register an additional multicast address with a network interface.
3152  *
3153  * - If the address is already present, bump the reference count on the
3154  *   address and return.
3155  * - If the address is not link-layer, look up a link layer address.
3156  * - Allocate address structures for one or both addresses, and attach to the
3157  *   multicast address list on the interface.  If automatically adding a link
3158  *   layer address, the protocol address will own a reference to the link
3159  *   layer address, to be freed when it is freed.
3160  * - Notify the network device driver of an addition to the multicast address
3161  *   list.
3162  *
3163  * 'sa' points to caller-owned memory with the desired multicast address.
3164  *
3165  * 'retifma' will be used to return a pointer to the resulting multicast
3166  * address reference, if desired.
3167  */
3168 int
3169 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3170     struct ifmultiaddr **retifma)
3171 {
3172 	struct ifmultiaddr *ifma, *ll_ifma;
3173 	struct sockaddr *llsa;
3174 	struct sockaddr_dl sdl;
3175 	int error;
3176 
3177 	/*
3178 	 * If the address is already present, return a new reference to it;
3179 	 * otherwise, allocate storage and set up a new address.
3180 	 */
3181 	IF_ADDR_WLOCK(ifp);
3182 	ifma = if_findmulti(ifp, sa);
3183 	if (ifma != NULL) {
3184 		ifma->ifma_refcount++;
3185 		if (retifma != NULL)
3186 			*retifma = ifma;
3187 		IF_ADDR_WUNLOCK(ifp);
3188 		return (0);
3189 	}
3190 
3191 	/*
3192 	 * The address isn't already present; resolve the protocol address
3193 	 * into a link layer address, and then look that up, bump its
3194 	 * refcount or allocate an ifma for that also.
3195 	 * Most link layer resolving functions returns address data which
3196 	 * fits inside default sockaddr_dl structure. However callback
3197 	 * can allocate another sockaddr structure, in that case we need to
3198 	 * free it later.
3199 	 */
3200 	llsa = NULL;
3201 	ll_ifma = NULL;
3202 	if (ifp->if_resolvemulti != NULL) {
3203 		/* Provide called function with buffer size information */
3204 		sdl.sdl_len = sizeof(sdl);
3205 		llsa = (struct sockaddr *)&sdl;
3206 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3207 		if (error)
3208 			goto unlock_out;
3209 	}
3210 
3211 	/*
3212 	 * Allocate the new address.  Don't hook it up yet, as we may also
3213 	 * need to allocate a link layer multicast address.
3214 	 */
3215 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3216 	if (ifma == NULL) {
3217 		error = ENOMEM;
3218 		goto free_llsa_out;
3219 	}
3220 
3221 	/*
3222 	 * If a link layer address is found, we'll need to see if it's
3223 	 * already present in the address list, or allocate is as well.
3224 	 * When this block finishes, the link layer address will be on the
3225 	 * list.
3226 	 */
3227 	if (llsa != NULL) {
3228 		ll_ifma = if_findmulti(ifp, llsa);
3229 		if (ll_ifma == NULL) {
3230 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3231 			if (ll_ifma == NULL) {
3232 				--ifma->ifma_refcount;
3233 				if_freemulti(ifma);
3234 				error = ENOMEM;
3235 				goto free_llsa_out;
3236 			}
3237 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3238 			    ifma_link);
3239 		} else
3240 			ll_ifma->ifma_refcount++;
3241 		ifma->ifma_llifma = ll_ifma;
3242 	}
3243 
3244 	/*
3245 	 * We now have a new multicast address, ifma, and possibly a new or
3246 	 * referenced link layer address.  Add the primary address to the
3247 	 * ifnet address list.
3248 	 */
3249 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3250 
3251 	if (retifma != NULL)
3252 		*retifma = ifma;
3253 
3254 	/*
3255 	 * Must generate the message while holding the lock so that 'ifma'
3256 	 * pointer is still valid.
3257 	 */
3258 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3259 	IF_ADDR_WUNLOCK(ifp);
3260 
3261 	/*
3262 	 * We are certain we have added something, so call down to the
3263 	 * interface to let them know about it.
3264 	 */
3265 	if (ifp->if_ioctl != NULL) {
3266 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3267 	}
3268 
3269 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3270 		link_free_sdl(llsa);
3271 
3272 	return (0);
3273 
3274 free_llsa_out:
3275 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3276 		link_free_sdl(llsa);
3277 
3278 unlock_out:
3279 	IF_ADDR_WUNLOCK(ifp);
3280 	return (error);
3281 }
3282 
3283 /*
3284  * Delete a multicast group membership by network-layer group address.
3285  *
3286  * Returns ENOENT if the entry could not be found. If ifp no longer
3287  * exists, results are undefined. This entry point should only be used
3288  * from subsystems which do appropriate locking to hold ifp for the
3289  * duration of the call.
3290  * Network-layer protocol domains must use if_delmulti_ifma().
3291  */
3292 int
3293 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3294 {
3295 	struct ifmultiaddr *ifma;
3296 	int lastref;
3297 #ifdef INVARIANTS
3298 	struct ifnet *oifp;
3299 
3300 	IFNET_RLOCK_NOSLEEP();
3301 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3302 		if (ifp == oifp)
3303 			break;
3304 	if (ifp != oifp)
3305 		ifp = NULL;
3306 	IFNET_RUNLOCK_NOSLEEP();
3307 
3308 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3309 #endif
3310 	if (ifp == NULL)
3311 		return (ENOENT);
3312 
3313 	IF_ADDR_WLOCK(ifp);
3314 	lastref = 0;
3315 	ifma = if_findmulti(ifp, sa);
3316 	if (ifma != NULL)
3317 		lastref = if_delmulti_locked(ifp, ifma, 0);
3318 	IF_ADDR_WUNLOCK(ifp);
3319 
3320 	if (ifma == NULL)
3321 		return (ENOENT);
3322 
3323 	if (lastref && ifp->if_ioctl != NULL) {
3324 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3325 	}
3326 
3327 	return (0);
3328 }
3329 
3330 /*
3331  * Delete all multicast group membership for an interface.
3332  * Should be used to quickly flush all multicast filters.
3333  */
3334 void
3335 if_delallmulti(struct ifnet *ifp)
3336 {
3337 	struct ifmultiaddr *ifma;
3338 	struct ifmultiaddr *next;
3339 
3340 	IF_ADDR_WLOCK(ifp);
3341 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3342 		if_delmulti_locked(ifp, ifma, 0);
3343 	IF_ADDR_WUNLOCK(ifp);
3344 }
3345 
3346 /*
3347  * Delete a multicast group membership by group membership pointer.
3348  * Network-layer protocol domains must use this routine.
3349  *
3350  * It is safe to call this routine if the ifp disappeared.
3351  */
3352 void
3353 if_delmulti_ifma(struct ifmultiaddr *ifma)
3354 {
3355 	struct ifnet *ifp;
3356 	int lastref;
3357 
3358 	ifp = ifma->ifma_ifp;
3359 #ifdef DIAGNOSTIC
3360 	if (ifp == NULL) {
3361 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3362 	} else {
3363 		struct ifnet *oifp;
3364 
3365 		IFNET_RLOCK_NOSLEEP();
3366 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3367 			if (ifp == oifp)
3368 				break;
3369 		if (ifp != oifp) {
3370 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3371 			ifp = NULL;
3372 		}
3373 		IFNET_RUNLOCK_NOSLEEP();
3374 	}
3375 #endif
3376 	/*
3377 	 * If and only if the ifnet instance exists: Acquire the address lock.
3378 	 */
3379 	if (ifp != NULL)
3380 		IF_ADDR_WLOCK(ifp);
3381 
3382 	lastref = if_delmulti_locked(ifp, ifma, 0);
3383 
3384 	if (ifp != NULL) {
3385 		/*
3386 		 * If and only if the ifnet instance exists:
3387 		 *  Release the address lock.
3388 		 *  If the group was left: update the hardware hash filter.
3389 		 */
3390 		IF_ADDR_WUNLOCK(ifp);
3391 		if (lastref && ifp->if_ioctl != NULL) {
3392 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3393 		}
3394 	}
3395 }
3396 
3397 /*
3398  * Perform deletion of network-layer and/or link-layer multicast address.
3399  *
3400  * Return 0 if the reference count was decremented.
3401  * Return 1 if the final reference was released, indicating that the
3402  * hardware hash filter should be reprogrammed.
3403  */
3404 static int
3405 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3406 {
3407 	struct ifmultiaddr *ll_ifma;
3408 
3409 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3410 		KASSERT(ifma->ifma_ifp == ifp,
3411 		    ("%s: inconsistent ifp %p", __func__, ifp));
3412 		IF_ADDR_WLOCK_ASSERT(ifp);
3413 	}
3414 
3415 	ifp = ifma->ifma_ifp;
3416 
3417 	/*
3418 	 * If the ifnet is detaching, null out references to ifnet,
3419 	 * so that upper protocol layers will notice, and not attempt
3420 	 * to obtain locks for an ifnet which no longer exists. The
3421 	 * routing socket announcement must happen before the ifnet
3422 	 * instance is detached from the system.
3423 	 */
3424 	if (detaching) {
3425 #ifdef DIAGNOSTIC
3426 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3427 #endif
3428 		/*
3429 		 * ifp may already be nulled out if we are being reentered
3430 		 * to delete the ll_ifma.
3431 		 */
3432 		if (ifp != NULL) {
3433 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3434 			ifma->ifma_ifp = NULL;
3435 		}
3436 	}
3437 
3438 	if (--ifma->ifma_refcount > 0)
3439 		return 0;
3440 
3441 	/*
3442 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3443 	 * have been associated with it. Release it first if so.
3444 	 */
3445 	ll_ifma = ifma->ifma_llifma;
3446 	if (ll_ifma != NULL) {
3447 		KASSERT(ifma->ifma_lladdr != NULL,
3448 		    ("%s: llifma w/o lladdr", __func__));
3449 		if (detaching)
3450 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3451 		if (--ll_ifma->ifma_refcount == 0) {
3452 			if (ifp != NULL) {
3453 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3454 				    ifma_link);
3455 			}
3456 			if_freemulti(ll_ifma);
3457 		}
3458 	}
3459 
3460 	if (ifp != NULL)
3461 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3462 
3463 	if_freemulti(ifma);
3464 
3465 	/*
3466 	 * The last reference to this instance of struct ifmultiaddr
3467 	 * was released; the hardware should be notified of this change.
3468 	 */
3469 	return 1;
3470 }
3471 
3472 /*
3473  * Set the link layer address on an interface.
3474  *
3475  * At this time we only support certain types of interfaces,
3476  * and we don't allow the length of the address to change.
3477  *
3478  * Set noinline to be dtrace-friendly
3479  */
3480 __noinline int
3481 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3482 {
3483 	struct sockaddr_dl *sdl;
3484 	struct ifaddr *ifa;
3485 	struct ifreq ifr;
3486 
3487 	IF_ADDR_RLOCK(ifp);
3488 	ifa = ifp->if_addr;
3489 	if (ifa == NULL) {
3490 		IF_ADDR_RUNLOCK(ifp);
3491 		return (EINVAL);
3492 	}
3493 	ifa_ref(ifa);
3494 	IF_ADDR_RUNLOCK(ifp);
3495 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3496 	if (sdl == NULL) {
3497 		ifa_free(ifa);
3498 		return (EINVAL);
3499 	}
3500 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3501 		ifa_free(ifa);
3502 		return (EINVAL);
3503 	}
3504 	switch (ifp->if_type) {
3505 	case IFT_ETHER:
3506 	case IFT_FDDI:
3507 	case IFT_XETHER:
3508 	case IFT_ISO88025:
3509 	case IFT_L2VLAN:
3510 	case IFT_BRIDGE:
3511 	case IFT_ARCNET:
3512 	case IFT_IEEE8023ADLAG:
3513 	case IFT_IEEE80211:
3514 		bcopy(lladdr, LLADDR(sdl), len);
3515 		ifa_free(ifa);
3516 		break;
3517 	default:
3518 		ifa_free(ifa);
3519 		return (ENODEV);
3520 	}
3521 
3522 	/*
3523 	 * If the interface is already up, we need
3524 	 * to re-init it in order to reprogram its
3525 	 * address filter.
3526 	 */
3527 	if ((ifp->if_flags & IFF_UP) != 0) {
3528 		if (ifp->if_ioctl) {
3529 			ifp->if_flags &= ~IFF_UP;
3530 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3531 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3532 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3533 			ifp->if_flags |= IFF_UP;
3534 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3535 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3536 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3537 		}
3538 	}
3539 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3540 	return (0);
3541 }
3542 
3543 /*
3544  * Compat function for handling basic encapsulation requests.
3545  * Not converted stacks (FDDI, IB, ..) supports traditional
3546  * output model: ARP (and other similar L2 protocols) are handled
3547  * inside output routine, arpresolve/nd6_resolve() returns MAC
3548  * address instead of full prepend.
3549  *
3550  * This function creates calculated header==MAC for IPv4/IPv6 and
3551  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3552  * address families.
3553  */
3554 static int
3555 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3556 {
3557 
3558 	if (req->rtype != IFENCAP_LL)
3559 		return (EOPNOTSUPP);
3560 
3561 	if (req->bufsize < req->lladdr_len)
3562 		return (ENOMEM);
3563 
3564 	switch (req->family) {
3565 	case AF_INET:
3566 	case AF_INET6:
3567 		break;
3568 	default:
3569 		return (EAFNOSUPPORT);
3570 	}
3571 
3572 	/* Copy lladdr to storage as is */
3573 	memmove(req->buf, req->lladdr, req->lladdr_len);
3574 	req->bufsize = req->lladdr_len;
3575 	req->lladdr_off = 0;
3576 
3577 	return (0);
3578 }
3579 
3580 /*
3581  * The name argument must be a pointer to storage which will last as
3582  * long as the interface does.  For physical devices, the result of
3583  * device_get_name(dev) is a good choice and for pseudo-devices a
3584  * static string works well.
3585  */
3586 void
3587 if_initname(struct ifnet *ifp, const char *name, int unit)
3588 {
3589 	ifp->if_dname = name;
3590 	ifp->if_dunit = unit;
3591 	if (unit != IF_DUNIT_NONE)
3592 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3593 	else
3594 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3595 }
3596 
3597 int
3598 if_printf(struct ifnet *ifp, const char * fmt, ...)
3599 {
3600 	va_list ap;
3601 	int retval;
3602 
3603 	retval = printf("%s: ", ifp->if_xname);
3604 	va_start(ap, fmt);
3605 	retval += vprintf(fmt, ap);
3606 	va_end(ap);
3607 	return (retval);
3608 }
3609 
3610 void
3611 if_start(struct ifnet *ifp)
3612 {
3613 
3614 	(*(ifp)->if_start)(ifp);
3615 }
3616 
3617 /*
3618  * Backwards compatibility interface for drivers
3619  * that have not implemented it
3620  */
3621 static int
3622 if_transmit(struct ifnet *ifp, struct mbuf *m)
3623 {
3624 	int error;
3625 
3626 	IFQ_HANDOFF(ifp, m, error);
3627 	return (error);
3628 }
3629 
3630 static void
3631 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3632 {
3633 
3634 	m_freem(m);
3635 }
3636 
3637 int
3638 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3639 {
3640 	int active = 0;
3641 
3642 	IF_LOCK(ifq);
3643 	if (_IF_QFULL(ifq)) {
3644 		IF_UNLOCK(ifq);
3645 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3646 		m_freem(m);
3647 		return (0);
3648 	}
3649 	if (ifp != NULL) {
3650 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3651 		if (m->m_flags & (M_BCAST|M_MCAST))
3652 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3653 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3654 	}
3655 	_IF_ENQUEUE(ifq, m);
3656 	IF_UNLOCK(ifq);
3657 	if (ifp != NULL && !active)
3658 		(*(ifp)->if_start)(ifp);
3659 	return (1);
3660 }
3661 
3662 void
3663 if_register_com_alloc(u_char type,
3664     if_com_alloc_t *a, if_com_free_t *f)
3665 {
3666 
3667 	KASSERT(if_com_alloc[type] == NULL,
3668 	    ("if_register_com_alloc: %d already registered", type));
3669 	KASSERT(if_com_free[type] == NULL,
3670 	    ("if_register_com_alloc: %d free already registered", type));
3671 
3672 	if_com_alloc[type] = a;
3673 	if_com_free[type] = f;
3674 }
3675 
3676 void
3677 if_deregister_com_alloc(u_char type)
3678 {
3679 
3680 	KASSERT(if_com_alloc[type] != NULL,
3681 	    ("if_deregister_com_alloc: %d not registered", type));
3682 	KASSERT(if_com_free[type] != NULL,
3683 	    ("if_deregister_com_alloc: %d free not registered", type));
3684 	if_com_alloc[type] = NULL;
3685 	if_com_free[type] = NULL;
3686 }
3687 
3688 /* API for driver access to network stack owned ifnet.*/
3689 uint64_t
3690 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3691 {
3692 	uint64_t oldbrate;
3693 
3694 	oldbrate = ifp->if_baudrate;
3695 	ifp->if_baudrate = baudrate;
3696 	return (oldbrate);
3697 }
3698 
3699 uint64_t
3700 if_getbaudrate(if_t ifp)
3701 {
3702 
3703 	return (((struct ifnet *)ifp)->if_baudrate);
3704 }
3705 
3706 int
3707 if_setcapabilities(if_t ifp, int capabilities)
3708 {
3709 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3710 	return (0);
3711 }
3712 
3713 int
3714 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3715 {
3716 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3717 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3718 
3719 	return (0);
3720 }
3721 
3722 int
3723 if_getcapabilities(if_t ifp)
3724 {
3725 	return ((struct ifnet *)ifp)->if_capabilities;
3726 }
3727 
3728 int
3729 if_setcapenable(if_t ifp, int capabilities)
3730 {
3731 	((struct ifnet *)ifp)->if_capenable = capabilities;
3732 	return (0);
3733 }
3734 
3735 int
3736 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3737 {
3738 	if(setcap)
3739 		((struct ifnet *)ifp)->if_capenable |= setcap;
3740 	if(clearcap)
3741 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3742 
3743 	return (0);
3744 }
3745 
3746 const char *
3747 if_getdname(if_t ifp)
3748 {
3749 	return ((struct ifnet *)ifp)->if_dname;
3750 }
3751 
3752 int
3753 if_togglecapenable(if_t ifp, int togglecap)
3754 {
3755 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3756 	return (0);
3757 }
3758 
3759 int
3760 if_getcapenable(if_t ifp)
3761 {
3762 	return ((struct ifnet *)ifp)->if_capenable;
3763 }
3764 
3765 /*
3766  * This is largely undesirable because it ties ifnet to a device, but does
3767  * provide flexiblity for an embedded product vendor. Should be used with
3768  * the understanding that it violates the interface boundaries, and should be
3769  * a last resort only.
3770  */
3771 int
3772 if_setdev(if_t ifp, void *dev)
3773 {
3774 	return (0);
3775 }
3776 
3777 int
3778 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3779 {
3780 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3781 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3782 
3783 	return (0);
3784 }
3785 
3786 int
3787 if_getdrvflags(if_t ifp)
3788 {
3789 	return ((struct ifnet *)ifp)->if_drv_flags;
3790 }
3791 
3792 int
3793 if_setdrvflags(if_t ifp, int flags)
3794 {
3795 	((struct ifnet *)ifp)->if_drv_flags = flags;
3796 	return (0);
3797 }
3798 
3799 
3800 int
3801 if_setflags(if_t ifp, int flags)
3802 {
3803 	((struct ifnet *)ifp)->if_flags = flags;
3804 	return (0);
3805 }
3806 
3807 int
3808 if_setflagbits(if_t ifp, int set, int clear)
3809 {
3810 	((struct ifnet *)ifp)->if_flags |= set;
3811 	((struct ifnet *)ifp)->if_flags &= ~clear;
3812 
3813 	return (0);
3814 }
3815 
3816 int
3817 if_getflags(if_t ifp)
3818 {
3819 	return ((struct ifnet *)ifp)->if_flags;
3820 }
3821 
3822 int
3823 if_clearhwassist(if_t ifp)
3824 {
3825 	((struct ifnet *)ifp)->if_hwassist = 0;
3826 	return (0);
3827 }
3828 
3829 int
3830 if_sethwassistbits(if_t ifp, int toset, int toclear)
3831 {
3832 	((struct ifnet *)ifp)->if_hwassist |= toset;
3833 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3834 
3835 	return (0);
3836 }
3837 
3838 int
3839 if_sethwassist(if_t ifp, int hwassist_bit)
3840 {
3841 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3842 	return (0);
3843 }
3844 
3845 int
3846 if_gethwassist(if_t ifp)
3847 {
3848 	return ((struct ifnet *)ifp)->if_hwassist;
3849 }
3850 
3851 int
3852 if_setmtu(if_t ifp, int mtu)
3853 {
3854 	((struct ifnet *)ifp)->if_mtu = mtu;
3855 	return (0);
3856 }
3857 
3858 int
3859 if_getmtu(if_t ifp)
3860 {
3861 	return ((struct ifnet *)ifp)->if_mtu;
3862 }
3863 
3864 int
3865 if_getmtu_family(if_t ifp, int family)
3866 {
3867 	struct domain *dp;
3868 
3869 	for (dp = domains; dp; dp = dp->dom_next) {
3870 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3871 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3872 	}
3873 
3874 	return (((struct ifnet *)ifp)->if_mtu);
3875 }
3876 
3877 int
3878 if_setsoftc(if_t ifp, void *softc)
3879 {
3880 	((struct ifnet *)ifp)->if_softc = softc;
3881 	return (0);
3882 }
3883 
3884 void *
3885 if_getsoftc(if_t ifp)
3886 {
3887 	return ((struct ifnet *)ifp)->if_softc;
3888 }
3889 
3890 void
3891 if_setrcvif(struct mbuf *m, if_t ifp)
3892 {
3893 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3894 }
3895 
3896 void
3897 if_setvtag(struct mbuf *m, uint16_t tag)
3898 {
3899 	m->m_pkthdr.ether_vtag = tag;
3900 }
3901 
3902 uint16_t
3903 if_getvtag(struct mbuf *m)
3904 {
3905 
3906 	return (m->m_pkthdr.ether_vtag);
3907 }
3908 
3909 int
3910 if_sendq_empty(if_t ifp)
3911 {
3912 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3913 }
3914 
3915 struct ifaddr *
3916 if_getifaddr(if_t ifp)
3917 {
3918 	return ((struct ifnet *)ifp)->if_addr;
3919 }
3920 
3921 int
3922 if_getamcount(if_t ifp)
3923 {
3924 	return ((struct ifnet *)ifp)->if_amcount;
3925 }
3926 
3927 
3928 int
3929 if_setsendqready(if_t ifp)
3930 {
3931 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3932 	return (0);
3933 }
3934 
3935 int
3936 if_setsendqlen(if_t ifp, int tx_desc_count)
3937 {
3938 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3939 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3940 
3941 	return (0);
3942 }
3943 
3944 int
3945 if_vlantrunkinuse(if_t ifp)
3946 {
3947 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3948 }
3949 
3950 int
3951 if_input(if_t ifp, struct mbuf* sendmp)
3952 {
3953 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3954 	return (0);
3955 
3956 }
3957 
3958 /* XXX */
3959 #ifndef ETH_ADDR_LEN
3960 #define ETH_ADDR_LEN 6
3961 #endif
3962 
3963 int
3964 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3965 {
3966 	struct ifmultiaddr *ifma;
3967 	uint8_t *lmta = (uint8_t *)mta;
3968 	int mcnt = 0;
3969 
3970 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3971 		if (ifma->ifma_addr->sa_family != AF_LINK)
3972 			continue;
3973 
3974 		if (mcnt == max)
3975 			break;
3976 
3977 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3978 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3979 		mcnt++;
3980 	}
3981 	*cnt = mcnt;
3982 
3983 	return (0);
3984 }
3985 
3986 int
3987 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3988 {
3989 	int error;
3990 
3991 	if_maddr_rlock(ifp);
3992 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3993 	if_maddr_runlock(ifp);
3994 	return (error);
3995 }
3996 
3997 int
3998 if_multiaddr_count(if_t ifp, int max)
3999 {
4000 	struct ifmultiaddr *ifma;
4001 	int count;
4002 
4003 	count = 0;
4004 	if_maddr_rlock(ifp);
4005 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4006 		if (ifma->ifma_addr->sa_family != AF_LINK)
4007 			continue;
4008 		count++;
4009 		if (count == max)
4010 			break;
4011 	}
4012 	if_maddr_runlock(ifp);
4013 	return (count);
4014 }
4015 
4016 int
4017 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4018 {
4019 	struct ifmultiaddr *ifma;
4020 	int cnt = 0;
4021 
4022 	if_maddr_rlock(ifp);
4023 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4024 		cnt += filter(arg, ifma, cnt);
4025 	if_maddr_runlock(ifp);
4026 	return (cnt);
4027 }
4028 
4029 struct mbuf *
4030 if_dequeue(if_t ifp)
4031 {
4032 	struct mbuf *m;
4033 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4034 
4035 	return (m);
4036 }
4037 
4038 int
4039 if_sendq_prepend(if_t ifp, struct mbuf *m)
4040 {
4041 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4042 	return (0);
4043 }
4044 
4045 int
4046 if_setifheaderlen(if_t ifp, int len)
4047 {
4048 	((struct ifnet *)ifp)->if_hdrlen = len;
4049 	return (0);
4050 }
4051 
4052 caddr_t
4053 if_getlladdr(if_t ifp)
4054 {
4055 	return (IF_LLADDR((struct ifnet *)ifp));
4056 }
4057 
4058 void *
4059 if_gethandle(u_char type)
4060 {
4061 	return (if_alloc(type));
4062 }
4063 
4064 void
4065 if_bpfmtap(if_t ifh, struct mbuf *m)
4066 {
4067 	struct ifnet *ifp = (struct ifnet *)ifh;
4068 
4069 	BPF_MTAP(ifp, m);
4070 }
4071 
4072 void
4073 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4074 {
4075 	struct ifnet *ifp = (struct ifnet *)ifh;
4076 
4077 	ETHER_BPF_MTAP(ifp, m);
4078 }
4079 
4080 void
4081 if_vlancap(if_t ifh)
4082 {
4083 	struct ifnet *ifp = (struct ifnet *)ifh;
4084 	VLAN_CAPABILITIES(ifp);
4085 }
4086 
4087 void
4088 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4089 {
4090 	((struct ifnet *)ifp)->if_init = init_fn;
4091 }
4092 
4093 void
4094 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4095 {
4096 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4097 }
4098 
4099 void
4100 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4101 {
4102 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4103 }
4104 
4105 void
4106 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4107 {
4108 	((struct ifnet *)ifp)->if_transmit = start_fn;
4109 }
4110 
4111 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4112 {
4113 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4114 
4115 }
4116 
4117 void
4118 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4119 {
4120 
4121 	ifp->if_get_counter = fn;
4122 }
4123 
4124 /* Revisit these - These are inline functions originally. */
4125 int
4126 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4127 {
4128 	return drbr_inuse(ifh, br);
4129 }
4130 
4131 struct mbuf*
4132 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4133 {
4134 	return drbr_dequeue(ifh, br);
4135 }
4136 
4137 int
4138 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4139 {
4140 	return drbr_needs_enqueue(ifh, br);
4141 }
4142 
4143 int
4144 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4145 {
4146 	return drbr_enqueue(ifh, br, m);
4147 
4148 }
4149