xref: /freebsd/sys/net/if.c (revision 46c1105fbb6fbff6d6ccd0a18571342eb992d637)
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_RW,
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 (IF_AFDATA_TRYLOCK(ifp) == 0)
825 		return;
826 	if (ifp->if_afdata_initialized >= domain_init_status) {
827 		IF_AFDATA_UNLOCK(ifp);
828 		log(LOG_WARNING, "%s called more than once on %s\n",
829 		    __func__, ifp->if_xname);
830 		return;
831 	}
832 	ifp->if_afdata_initialized = domain_init_status;
833 	IF_AFDATA_UNLOCK(ifp);
834 
835 	/* address family dependent data region */
836 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
837 	for (dp = domains; dp; dp = dp->dom_next) {
838 		if (dp->dom_ifattach)
839 			ifp->if_afdata[dp->dom_family] =
840 			    (*dp->dom_ifattach)(ifp);
841 	}
842 }
843 
844 /*
845  * Remove any unicast or broadcast network addresses from an interface.
846  */
847 void
848 if_purgeaddrs(struct ifnet *ifp)
849 {
850 	struct ifaddr *ifa, *next;
851 
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 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
877 		ifa_free(ifa);
878 	}
879 }
880 
881 /*
882  * Remove any multicast network addresses from an interface when an ifnet
883  * is going away.
884  */
885 static void
886 if_purgemaddrs(struct ifnet *ifp)
887 {
888 	struct ifmultiaddr *ifma;
889 	struct ifmultiaddr *next;
890 
891 	IF_ADDR_WLOCK(ifp);
892 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
893 		if_delmulti_locked(ifp, ifma, 1);
894 	IF_ADDR_WUNLOCK(ifp);
895 }
896 
897 /*
898  * Detach an interface, removing it from the list of "active" interfaces.
899  * If vmove flag is set on entry to if_detach_internal(), perform only a
900  * limited subset of cleanup tasks, given that we are moving an ifnet from
901  * one vnet to another, where it must be fully operational.
902  *
903  * XXXRW: There are some significant questions about event ordering, and
904  * how to prevent things from starting to use the interface during detach.
905  */
906 void
907 if_detach(struct ifnet *ifp)
908 {
909 
910 	CURVNET_SET_QUIET(ifp->if_vnet);
911 	if_detach_internal(ifp, 0, NULL);
912 	CURVNET_RESTORE();
913 }
914 
915 static int
916 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
917 {
918 	struct ifaddr *ifa;
919 	int i;
920 	struct domain *dp;
921  	struct ifnet *iter;
922  	int found = 0;
923 
924 	IFNET_WLOCK();
925 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
926 		if (iter == ifp) {
927 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
928 			found = 1;
929 			break;
930 		}
931 #ifdef VIMAGE
932 	if (found)
933 		curvnet->vnet_ifcnt--;
934 #endif
935 	IFNET_WUNLOCK();
936 	if (!found) {
937 		/*
938 		 * While we would want to panic here, we cannot
939 		 * guarantee that the interface is indeed still on
940 		 * the list given we don't hold locks all the way.
941 		 */
942 		return (ENOENT);
943 #if 0
944 		if (vmove)
945 			panic("%s: ifp=%p not on the ifnet tailq %p",
946 			    __func__, ifp, &V_ifnet);
947 		else
948 			return; /* XXX this should panic as well? */
949 #endif
950 	}
951 
952 	/* Check if this is a cloned interface or not. */
953 	if (vmove && ifcp != NULL)
954 		*ifcp = if_clone_findifc(ifp);
955 
956 	/*
957 	 * Remove/wait for pending events.
958 	 */
959 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
960 
961 	/*
962 	 * Remove routes and flush queues.
963 	 */
964 	if_down(ifp);
965 #ifdef ALTQ
966 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
967 		altq_disable(&ifp->if_snd);
968 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
969 		altq_detach(&ifp->if_snd);
970 #endif
971 
972 	if_purgeaddrs(ifp);
973 
974 #ifdef INET
975 	in_ifdetach(ifp);
976 #endif
977 
978 #ifdef INET6
979 	/*
980 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
981 	 * before removing routing entries below, since IPv6 interface direct
982 	 * routes are expected to be removed by the IPv6-specific kernel API.
983 	 * Otherwise, the kernel will detect some inconsistency and bark it.
984 	 */
985 	in6_ifdetach(ifp);
986 #endif
987 	if_purgemaddrs(ifp);
988 
989 	/* Announce that the interface is gone. */
990 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
991 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
992 	if (IS_DEFAULT_VNET(curvnet))
993 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
994 
995 	if (!vmove) {
996 		/*
997 		 * Prevent further calls into the device driver via ifnet.
998 		 */
999 		if_dead(ifp);
1000 
1001 		/*
1002 		 * Remove link ifaddr pointer and maybe decrement if_index.
1003 		 * Clean up all addresses.
1004 		 */
1005 		ifp->if_addr = NULL;
1006 
1007 		/* We can now free link ifaddr. */
1008 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1009 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1010 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1011 			ifa_free(ifa);
1012 		}
1013 	}
1014 
1015 	rt_flushifroutes(ifp);
1016 	if_delgroups(ifp);
1017 
1018 	/*
1019 	 * We cannot hold the lock over dom_ifdetach calls as they might
1020 	 * sleep, for example trying to drain a callout, thus open up the
1021 	 * theoretical race with re-attaching.
1022 	 */
1023 	IF_AFDATA_LOCK(ifp);
1024 	i = ifp->if_afdata_initialized;
1025 	ifp->if_afdata_initialized = 0;
1026 	IF_AFDATA_UNLOCK(ifp);
1027 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1028 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1029 			(*dp->dom_ifdetach)(ifp,
1030 			    ifp->if_afdata[dp->dom_family]);
1031 	}
1032 
1033 	return (0);
1034 }
1035 
1036 #ifdef VIMAGE
1037 /*
1038  * if_vmove() performs a limited version of if_detach() in current
1039  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1040  * An attempt is made to shrink if_index in current vnet, find an
1041  * unused if_index in target vnet and calls if_grow() if necessary,
1042  * and finally find an unused if_xname for the target vnet.
1043  */
1044 void
1045 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1046 {
1047 	struct if_clone *ifc;
1048 	u_int bif_dlt, bif_hdrlen;
1049 	int rc;
1050 
1051  	/*
1052 	 * if_detach_internal() will call the eventhandler to notify
1053 	 * interface departure.  That will detach if_bpf.  We need to
1054 	 * safe the dlt and hdrlen so we can re-attach it later.
1055 	 */
1056 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1057 
1058 	/*
1059 	 * Detach from current vnet, but preserve LLADDR info, do not
1060 	 * mark as dead etc. so that the ifnet can be reattached later.
1061 	 * If we cannot find it, we lost the race to someone else.
1062 	 */
1063 	rc = if_detach_internal(ifp, 1, &ifc);
1064 	if (rc != 0)
1065 		return;
1066 
1067 	/*
1068 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1069 	 * the if_index for that vnet if possible.
1070 	 *
1071 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1072 	 * or we'd lock on one vnet and unlock on another.
1073 	 */
1074 	IFNET_WLOCK();
1075 	ifindex_free_locked(ifp->if_index);
1076 	IFNET_WUNLOCK();
1077 
1078 	/*
1079 	 * Perform interface-specific reassignment tasks, if provided by
1080 	 * the driver.
1081 	 */
1082 	if (ifp->if_reassign != NULL)
1083 		ifp->if_reassign(ifp, new_vnet, NULL);
1084 
1085 	/*
1086 	 * Switch to the context of the target vnet.
1087 	 */
1088 	CURVNET_SET_QUIET(new_vnet);
1089 
1090 	IFNET_WLOCK();
1091 	ifp->if_index = ifindex_alloc();
1092 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1093 	IFNET_WUNLOCK();
1094 
1095 	if_attach_internal(ifp, 1, ifc);
1096 
1097 	if (ifp->if_bpf == NULL)
1098 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1099 
1100 	CURVNET_RESTORE();
1101 }
1102 
1103 /*
1104  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1105  */
1106 static int
1107 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1108 {
1109 	struct prison *pr;
1110 	struct ifnet *difp;
1111 
1112 	/* Try to find the prison within our visibility. */
1113 	sx_slock(&allprison_lock);
1114 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1115 	sx_sunlock(&allprison_lock);
1116 	if (pr == NULL)
1117 		return (ENXIO);
1118 	prison_hold_locked(pr);
1119 	mtx_unlock(&pr->pr_mtx);
1120 
1121 	/* Do not try to move the iface from and to the same prison. */
1122 	if (pr->pr_vnet == ifp->if_vnet) {
1123 		prison_free(pr);
1124 		return (EEXIST);
1125 	}
1126 
1127 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1128 	/* XXX Lock interfaces to avoid races. */
1129 	CURVNET_SET_QUIET(pr->pr_vnet);
1130 	difp = ifunit(ifname);
1131 	CURVNET_RESTORE();
1132 	if (difp != NULL) {
1133 		prison_free(pr);
1134 		return (EEXIST);
1135 	}
1136 
1137 	/* Move the interface into the child jail/vnet. */
1138 	if_vmove(ifp, pr->pr_vnet);
1139 
1140 	/* Report the new if_xname back to the userland. */
1141 	sprintf(ifname, "%s", ifp->if_xname);
1142 
1143 	prison_free(pr);
1144 	return (0);
1145 }
1146 
1147 static int
1148 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1149 {
1150 	struct prison *pr;
1151 	struct vnet *vnet_dst;
1152 	struct ifnet *ifp;
1153 
1154 	/* Try to find the prison within our visibility. */
1155 	sx_slock(&allprison_lock);
1156 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1157 	sx_sunlock(&allprison_lock);
1158 	if (pr == NULL)
1159 		return (ENXIO);
1160 	prison_hold_locked(pr);
1161 	mtx_unlock(&pr->pr_mtx);
1162 
1163 	/* Make sure the named iface exists in the source prison/vnet. */
1164 	CURVNET_SET(pr->pr_vnet);
1165 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1166 	if (ifp == NULL) {
1167 		CURVNET_RESTORE();
1168 		prison_free(pr);
1169 		return (ENXIO);
1170 	}
1171 
1172 	/* Do not try to move the iface from and to the same prison. */
1173 	vnet_dst = TD_TO_VNET(td);
1174 	if (vnet_dst == ifp->if_vnet) {
1175 		CURVNET_RESTORE();
1176 		prison_free(pr);
1177 		return (EEXIST);
1178 	}
1179 
1180 	/* Get interface back from child jail/vnet. */
1181 	if_vmove(ifp, vnet_dst);
1182 	CURVNET_RESTORE();
1183 
1184 	/* Report the new if_xname back to the userland. */
1185 	sprintf(ifname, "%s", ifp->if_xname);
1186 
1187 	prison_free(pr);
1188 	return (0);
1189 }
1190 #endif /* VIMAGE */
1191 
1192 /*
1193  * Add a group to an interface
1194  */
1195 int
1196 if_addgroup(struct ifnet *ifp, const char *groupname)
1197 {
1198 	struct ifg_list		*ifgl;
1199 	struct ifg_group	*ifg = NULL;
1200 	struct ifg_member	*ifgm;
1201 	int 			 new = 0;
1202 
1203 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1204 	    groupname[strlen(groupname) - 1] <= '9')
1205 		return (EINVAL);
1206 
1207 	IFNET_WLOCK();
1208 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1209 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1210 			IFNET_WUNLOCK();
1211 			return (EEXIST);
1212 		}
1213 
1214 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1215 	    M_NOWAIT)) == NULL) {
1216 	    	IFNET_WUNLOCK();
1217 		return (ENOMEM);
1218 	}
1219 
1220 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1221 	    M_TEMP, M_NOWAIT)) == NULL) {
1222 		free(ifgl, M_TEMP);
1223 		IFNET_WUNLOCK();
1224 		return (ENOMEM);
1225 	}
1226 
1227 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1228 		if (!strcmp(ifg->ifg_group, groupname))
1229 			break;
1230 
1231 	if (ifg == NULL) {
1232 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1233 		    M_TEMP, M_NOWAIT)) == NULL) {
1234 			free(ifgl, M_TEMP);
1235 			free(ifgm, M_TEMP);
1236 			IFNET_WUNLOCK();
1237 			return (ENOMEM);
1238 		}
1239 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1240 		ifg->ifg_refcnt = 0;
1241 		TAILQ_INIT(&ifg->ifg_members);
1242 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1243 		new = 1;
1244 	}
1245 
1246 	ifg->ifg_refcnt++;
1247 	ifgl->ifgl_group = ifg;
1248 	ifgm->ifgm_ifp = ifp;
1249 
1250 	IF_ADDR_WLOCK(ifp);
1251 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1252 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1253 	IF_ADDR_WUNLOCK(ifp);
1254 
1255 	IFNET_WUNLOCK();
1256 
1257 	if (new)
1258 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1259 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1260 
1261 	return (0);
1262 }
1263 
1264 /*
1265  * Remove a group from an interface
1266  */
1267 int
1268 if_delgroup(struct ifnet *ifp, const char *groupname)
1269 {
1270 	struct ifg_list		*ifgl;
1271 	struct ifg_member	*ifgm;
1272 
1273 	IFNET_WLOCK();
1274 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1275 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1276 			break;
1277 	if (ifgl == NULL) {
1278 		IFNET_WUNLOCK();
1279 		return (ENOENT);
1280 	}
1281 
1282 	IF_ADDR_WLOCK(ifp);
1283 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1284 	IF_ADDR_WUNLOCK(ifp);
1285 
1286 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1287 		if (ifgm->ifgm_ifp == ifp)
1288 			break;
1289 
1290 	if (ifgm != NULL) {
1291 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1292 		free(ifgm, M_TEMP);
1293 	}
1294 
1295 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1296 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1297 		IFNET_WUNLOCK();
1298 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1299 		free(ifgl->ifgl_group, M_TEMP);
1300 	} else
1301 		IFNET_WUNLOCK();
1302 
1303 	free(ifgl, M_TEMP);
1304 
1305 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1306 
1307 	return (0);
1308 }
1309 
1310 /*
1311  * Remove an interface from all groups
1312  */
1313 static void
1314 if_delgroups(struct ifnet *ifp)
1315 {
1316 	struct ifg_list		*ifgl;
1317 	struct ifg_member	*ifgm;
1318 	char groupname[IFNAMSIZ];
1319 
1320 	IFNET_WLOCK();
1321 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1322 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1323 
1324 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1325 
1326 		IF_ADDR_WLOCK(ifp);
1327 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1328 		IF_ADDR_WUNLOCK(ifp);
1329 
1330 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1331 			if (ifgm->ifgm_ifp == ifp)
1332 				break;
1333 
1334 		if (ifgm != NULL) {
1335 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1336 			    ifgm_next);
1337 			free(ifgm, M_TEMP);
1338 		}
1339 
1340 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1341 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1342 			IFNET_WUNLOCK();
1343 			EVENTHANDLER_INVOKE(group_detach_event,
1344 			    ifgl->ifgl_group);
1345 			free(ifgl->ifgl_group, M_TEMP);
1346 		} else
1347 			IFNET_WUNLOCK();
1348 
1349 		free(ifgl, M_TEMP);
1350 
1351 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1352 
1353 		IFNET_WLOCK();
1354 	}
1355 	IFNET_WUNLOCK();
1356 }
1357 
1358 /*
1359  * Stores all groups from an interface in memory pointed
1360  * to by data
1361  */
1362 static int
1363 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1364 {
1365 	int			 len, error;
1366 	struct ifg_list		*ifgl;
1367 	struct ifg_req		 ifgrq, *ifgp;
1368 	struct ifgroupreq	*ifgr = data;
1369 
1370 	if (ifgr->ifgr_len == 0) {
1371 		IF_ADDR_RLOCK(ifp);
1372 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1373 			ifgr->ifgr_len += sizeof(struct ifg_req);
1374 		IF_ADDR_RUNLOCK(ifp);
1375 		return (0);
1376 	}
1377 
1378 	len = ifgr->ifgr_len;
1379 	ifgp = ifgr->ifgr_groups;
1380 	/* XXX: wire */
1381 	IF_ADDR_RLOCK(ifp);
1382 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1383 		if (len < sizeof(ifgrq)) {
1384 			IF_ADDR_RUNLOCK(ifp);
1385 			return (EINVAL);
1386 		}
1387 		bzero(&ifgrq, sizeof ifgrq);
1388 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1389 		    sizeof(ifgrq.ifgrq_group));
1390 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1391 		    	IF_ADDR_RUNLOCK(ifp);
1392 			return (error);
1393 		}
1394 		len -= sizeof(ifgrq);
1395 		ifgp++;
1396 	}
1397 	IF_ADDR_RUNLOCK(ifp);
1398 
1399 	return (0);
1400 }
1401 
1402 /*
1403  * Stores all members of a group in memory pointed to by data
1404  */
1405 static int
1406 if_getgroupmembers(struct ifgroupreq *data)
1407 {
1408 	struct ifgroupreq	*ifgr = data;
1409 	struct ifg_group	*ifg;
1410 	struct ifg_member	*ifgm;
1411 	struct ifg_req		 ifgrq, *ifgp;
1412 	int			 len, error;
1413 
1414 	IFNET_RLOCK();
1415 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1416 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1417 			break;
1418 	if (ifg == NULL) {
1419 		IFNET_RUNLOCK();
1420 		return (ENOENT);
1421 	}
1422 
1423 	if (ifgr->ifgr_len == 0) {
1424 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1425 			ifgr->ifgr_len += sizeof(ifgrq);
1426 		IFNET_RUNLOCK();
1427 		return (0);
1428 	}
1429 
1430 	len = ifgr->ifgr_len;
1431 	ifgp = ifgr->ifgr_groups;
1432 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1433 		if (len < sizeof(ifgrq)) {
1434 			IFNET_RUNLOCK();
1435 			return (EINVAL);
1436 		}
1437 		bzero(&ifgrq, sizeof ifgrq);
1438 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1439 		    sizeof(ifgrq.ifgrq_member));
1440 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1441 			IFNET_RUNLOCK();
1442 			return (error);
1443 		}
1444 		len -= sizeof(ifgrq);
1445 		ifgp++;
1446 	}
1447 	IFNET_RUNLOCK();
1448 
1449 	return (0);
1450 }
1451 
1452 /*
1453  * Return counter values from counter(9)s stored in ifnet.
1454  */
1455 uint64_t
1456 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1457 {
1458 
1459 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1460 
1461 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1462 }
1463 
1464 /*
1465  * Increase an ifnet counter. Usually used for counters shared
1466  * between the stack and a driver, but function supports them all.
1467  */
1468 void
1469 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1470 {
1471 
1472 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1473 
1474 	counter_u64_add(ifp->if_counters[cnt], inc);
1475 }
1476 
1477 /*
1478  * Copy data from ifnet to userland API structure if_data.
1479  */
1480 void
1481 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1482 {
1483 
1484 	ifd->ifi_type = ifp->if_type;
1485 	ifd->ifi_physical = 0;
1486 	ifd->ifi_addrlen = ifp->if_addrlen;
1487 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1488 	ifd->ifi_link_state = ifp->if_link_state;
1489 	ifd->ifi_vhid = 0;
1490 	ifd->ifi_datalen = sizeof(struct if_data);
1491 	ifd->ifi_mtu = ifp->if_mtu;
1492 	ifd->ifi_metric = ifp->if_metric;
1493 	ifd->ifi_baudrate = ifp->if_baudrate;
1494 	ifd->ifi_hwassist = ifp->if_hwassist;
1495 	ifd->ifi_epoch = ifp->if_epoch;
1496 	ifd->ifi_lastchange = ifp->if_lastchange;
1497 
1498 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1499 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1500 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1501 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1502 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1503 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1504 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1505 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1506 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1507 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1508 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1509 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1510 }
1511 
1512 /*
1513  * Wrapper functions for struct ifnet address list locking macros.  These are
1514  * used by kernel modules to avoid encoding programming interface or binary
1515  * interface assumptions that may be violated when kernel-internal locking
1516  * approaches change.
1517  */
1518 void
1519 if_addr_rlock(struct ifnet *ifp)
1520 {
1521 
1522 	IF_ADDR_RLOCK(ifp);
1523 }
1524 
1525 void
1526 if_addr_runlock(struct ifnet *ifp)
1527 {
1528 
1529 	IF_ADDR_RUNLOCK(ifp);
1530 }
1531 
1532 void
1533 if_maddr_rlock(if_t ifp)
1534 {
1535 
1536 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1537 }
1538 
1539 void
1540 if_maddr_runlock(if_t ifp)
1541 {
1542 
1543 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1544 }
1545 
1546 /*
1547  * Initialization, destruction and refcounting functions for ifaddrs.
1548  */
1549 struct ifaddr *
1550 ifa_alloc(size_t size, int flags)
1551 {
1552 	struct ifaddr *ifa;
1553 
1554 	KASSERT(size >= sizeof(struct ifaddr),
1555 	    ("%s: invalid size %zu", __func__, size));
1556 
1557 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1558 	if (ifa == NULL)
1559 		return (NULL);
1560 
1561 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1562 		goto fail;
1563 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1564 		goto fail;
1565 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1566 		goto fail;
1567 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1568 		goto fail;
1569 
1570 	refcount_init(&ifa->ifa_refcnt, 1);
1571 
1572 	return (ifa);
1573 
1574 fail:
1575 	/* free(NULL) is okay */
1576 	counter_u64_free(ifa->ifa_opackets);
1577 	counter_u64_free(ifa->ifa_ipackets);
1578 	counter_u64_free(ifa->ifa_obytes);
1579 	counter_u64_free(ifa->ifa_ibytes);
1580 	free(ifa, M_IFADDR);
1581 
1582 	return (NULL);
1583 }
1584 
1585 void
1586 ifa_ref(struct ifaddr *ifa)
1587 {
1588 
1589 	refcount_acquire(&ifa->ifa_refcnt);
1590 }
1591 
1592 void
1593 ifa_free(struct ifaddr *ifa)
1594 {
1595 
1596 	if (refcount_release(&ifa->ifa_refcnt)) {
1597 		counter_u64_free(ifa->ifa_opackets);
1598 		counter_u64_free(ifa->ifa_ipackets);
1599 		counter_u64_free(ifa->ifa_obytes);
1600 		counter_u64_free(ifa->ifa_ibytes);
1601 		free(ifa, M_IFADDR);
1602 	}
1603 }
1604 
1605 static int
1606 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1607     struct sockaddr *ia)
1608 {
1609 	int error;
1610 	struct rt_addrinfo info;
1611 	struct sockaddr_dl null_sdl;
1612 	struct ifnet *ifp;
1613 
1614 	ifp = ifa->ifa_ifp;
1615 
1616 	bzero(&info, sizeof(info));
1617 	if (cmd != RTM_DELETE)
1618 		info.rti_ifp = V_loif;
1619 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1620 	info.rti_info[RTAX_DST] = ia;
1621 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1622 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1623 
1624 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1625 
1626 	if (error != 0)
1627 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1628 		    __func__, otype, if_name(ifp), error);
1629 
1630 	return (error);
1631 }
1632 
1633 int
1634 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1635 {
1636 
1637 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1638 }
1639 
1640 int
1641 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1642 {
1643 
1644 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1645 }
1646 
1647 int
1648 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1649 {
1650 
1651 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1652 }
1653 
1654 /*
1655  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1656  * structs used to represent other address families, it is necessary
1657  * to perform a different comparison.
1658  */
1659 
1660 #define	sa_dl_equal(a1, a2)	\
1661 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1662 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1663 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1664 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1665 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1666 
1667 /*
1668  * Locate an interface based on a complete address.
1669  */
1670 /*ARGSUSED*/
1671 static struct ifaddr *
1672 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1673 {
1674 	struct ifnet *ifp;
1675 	struct ifaddr *ifa;
1676 
1677 	IFNET_RLOCK_NOSLEEP();
1678 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1679 		IF_ADDR_RLOCK(ifp);
1680 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1681 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1682 				continue;
1683 			if (sa_equal(addr, ifa->ifa_addr)) {
1684 				if (getref)
1685 					ifa_ref(ifa);
1686 				IF_ADDR_RUNLOCK(ifp);
1687 				goto done;
1688 			}
1689 			/* IP6 doesn't have broadcast */
1690 			if ((ifp->if_flags & IFF_BROADCAST) &&
1691 			    ifa->ifa_broadaddr &&
1692 			    ifa->ifa_broadaddr->sa_len != 0 &&
1693 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1694 				if (getref)
1695 					ifa_ref(ifa);
1696 				IF_ADDR_RUNLOCK(ifp);
1697 				goto done;
1698 			}
1699 		}
1700 		IF_ADDR_RUNLOCK(ifp);
1701 	}
1702 	ifa = NULL;
1703 done:
1704 	IFNET_RUNLOCK_NOSLEEP();
1705 	return (ifa);
1706 }
1707 
1708 struct ifaddr *
1709 ifa_ifwithaddr(const struct sockaddr *addr)
1710 {
1711 
1712 	return (ifa_ifwithaddr_internal(addr, 1));
1713 }
1714 
1715 int
1716 ifa_ifwithaddr_check(const struct sockaddr *addr)
1717 {
1718 
1719 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1720 }
1721 
1722 /*
1723  * Locate an interface based on the broadcast address.
1724  */
1725 /* ARGSUSED */
1726 struct ifaddr *
1727 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1728 {
1729 	struct ifnet *ifp;
1730 	struct ifaddr *ifa;
1731 
1732 	IFNET_RLOCK_NOSLEEP();
1733 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1734 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1735 			continue;
1736 		IF_ADDR_RLOCK(ifp);
1737 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1738 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1739 				continue;
1740 			if ((ifp->if_flags & IFF_BROADCAST) &&
1741 			    ifa->ifa_broadaddr &&
1742 			    ifa->ifa_broadaddr->sa_len != 0 &&
1743 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1744 				ifa_ref(ifa);
1745 				IF_ADDR_RUNLOCK(ifp);
1746 				goto done;
1747 			}
1748 		}
1749 		IF_ADDR_RUNLOCK(ifp);
1750 	}
1751 	ifa = NULL;
1752 done:
1753 	IFNET_RUNLOCK_NOSLEEP();
1754 	return (ifa);
1755 }
1756 
1757 /*
1758  * Locate the point to point interface with a given destination address.
1759  */
1760 /*ARGSUSED*/
1761 struct ifaddr *
1762 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1763 {
1764 	struct ifnet *ifp;
1765 	struct ifaddr *ifa;
1766 
1767 	IFNET_RLOCK_NOSLEEP();
1768 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1769 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1770 			continue;
1771 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1772 			continue;
1773 		IF_ADDR_RLOCK(ifp);
1774 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1775 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1776 				continue;
1777 			if (ifa->ifa_dstaddr != NULL &&
1778 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1779 				ifa_ref(ifa);
1780 				IF_ADDR_RUNLOCK(ifp);
1781 				goto done;
1782 			}
1783 		}
1784 		IF_ADDR_RUNLOCK(ifp);
1785 	}
1786 	ifa = NULL;
1787 done:
1788 	IFNET_RUNLOCK_NOSLEEP();
1789 	return (ifa);
1790 }
1791 
1792 /*
1793  * Find an interface on a specific network.  If many, choice
1794  * is most specific found.
1795  */
1796 struct ifaddr *
1797 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1798 {
1799 	struct ifnet *ifp;
1800 	struct ifaddr *ifa;
1801 	struct ifaddr *ifa_maybe = NULL;
1802 	u_int af = addr->sa_family;
1803 	const char *addr_data = addr->sa_data, *cplim;
1804 
1805 	/*
1806 	 * AF_LINK addresses can be looked up directly by their index number,
1807 	 * so do that if we can.
1808 	 */
1809 	if (af == AF_LINK) {
1810 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1811 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1812 		return (ifaddr_byindex(sdl->sdl_index));
1813 	}
1814 
1815 	/*
1816 	 * Scan though each interface, looking for ones that have addresses
1817 	 * in this address family and the requested fib.  Maintain a reference
1818 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1819 	 * kept it stable when we move onto the next interface.
1820 	 */
1821 	IFNET_RLOCK_NOSLEEP();
1822 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1823 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1824 			continue;
1825 		IF_ADDR_RLOCK(ifp);
1826 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1827 			const char *cp, *cp2, *cp3;
1828 
1829 			if (ifa->ifa_addr->sa_family != af)
1830 next:				continue;
1831 			if (af == AF_INET &&
1832 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1833 				/*
1834 				 * This is a bit broken as it doesn't
1835 				 * take into account that the remote end may
1836 				 * be a single node in the network we are
1837 				 * looking for.
1838 				 * The trouble is that we don't know the
1839 				 * netmask for the remote end.
1840 				 */
1841 				if (ifa->ifa_dstaddr != NULL &&
1842 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1843 					ifa_ref(ifa);
1844 					IF_ADDR_RUNLOCK(ifp);
1845 					goto done;
1846 				}
1847 			} else {
1848 				/*
1849 				 * Scan all the bits in the ifa's address.
1850 				 * If a bit dissagrees with what we are
1851 				 * looking for, mask it with the netmask
1852 				 * to see if it really matters.
1853 				 * (A byte at a time)
1854 				 */
1855 				if (ifa->ifa_netmask == 0)
1856 					continue;
1857 				cp = addr_data;
1858 				cp2 = ifa->ifa_addr->sa_data;
1859 				cp3 = ifa->ifa_netmask->sa_data;
1860 				cplim = ifa->ifa_netmask->sa_len
1861 					+ (char *)ifa->ifa_netmask;
1862 				while (cp3 < cplim)
1863 					if ((*cp++ ^ *cp2++) & *cp3++)
1864 						goto next; /* next address! */
1865 				/*
1866 				 * If the netmask of what we just found
1867 				 * is more specific than what we had before
1868 				 * (if we had one), or if the virtual status
1869 				 * of new prefix is better than of the old one,
1870 				 * then remember the new one before continuing
1871 				 * to search for an even better one.
1872 				 */
1873 				if (ifa_maybe == NULL ||
1874 				    ifa_preferred(ifa_maybe, ifa) ||
1875 				    rn_refines((caddr_t)ifa->ifa_netmask,
1876 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1877 					if (ifa_maybe != NULL)
1878 						ifa_free(ifa_maybe);
1879 					ifa_maybe = ifa;
1880 					ifa_ref(ifa_maybe);
1881 				}
1882 			}
1883 		}
1884 		IF_ADDR_RUNLOCK(ifp);
1885 	}
1886 	ifa = ifa_maybe;
1887 	ifa_maybe = NULL;
1888 done:
1889 	IFNET_RUNLOCK_NOSLEEP();
1890 	if (ifa_maybe != NULL)
1891 		ifa_free(ifa_maybe);
1892 	return (ifa);
1893 }
1894 
1895 /*
1896  * Find an interface address specific to an interface best matching
1897  * a given address.
1898  */
1899 struct ifaddr *
1900 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1901 {
1902 	struct ifaddr *ifa;
1903 	const char *cp, *cp2, *cp3;
1904 	char *cplim;
1905 	struct ifaddr *ifa_maybe = NULL;
1906 	u_int af = addr->sa_family;
1907 
1908 	if (af >= AF_MAX)
1909 		return (NULL);
1910 	IF_ADDR_RLOCK(ifp);
1911 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1912 		if (ifa->ifa_addr->sa_family != af)
1913 			continue;
1914 		if (ifa_maybe == NULL)
1915 			ifa_maybe = ifa;
1916 		if (ifa->ifa_netmask == 0) {
1917 			if (sa_equal(addr, ifa->ifa_addr) ||
1918 			    (ifa->ifa_dstaddr &&
1919 			    sa_equal(addr, ifa->ifa_dstaddr)))
1920 				goto done;
1921 			continue;
1922 		}
1923 		if (ifp->if_flags & IFF_POINTOPOINT) {
1924 			if (sa_equal(addr, ifa->ifa_dstaddr))
1925 				goto done;
1926 		} else {
1927 			cp = addr->sa_data;
1928 			cp2 = ifa->ifa_addr->sa_data;
1929 			cp3 = ifa->ifa_netmask->sa_data;
1930 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1931 			for (; cp3 < cplim; cp3++)
1932 				if ((*cp++ ^ *cp2++) & *cp3)
1933 					break;
1934 			if (cp3 == cplim)
1935 				goto done;
1936 		}
1937 	}
1938 	ifa = ifa_maybe;
1939 done:
1940 	if (ifa != NULL)
1941 		ifa_ref(ifa);
1942 	IF_ADDR_RUNLOCK(ifp);
1943 	return (ifa);
1944 }
1945 
1946 /*
1947  * See whether new ifa is better than current one:
1948  * 1) A non-virtual one is preferred over virtual.
1949  * 2) A virtual in master state preferred over any other state.
1950  *
1951  * Used in several address selecting functions.
1952  */
1953 int
1954 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1955 {
1956 
1957 	return (cur->ifa_carp && (!next->ifa_carp ||
1958 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1959 }
1960 
1961 #include <net/if_llatbl.h>
1962 
1963 /*
1964  * Default action when installing a route with a Link Level gateway.
1965  * Lookup an appropriate real ifa to point to.
1966  * This should be moved to /sys/net/link.c eventually.
1967  */
1968 static void
1969 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1970 {
1971 	struct ifaddr *ifa, *oifa;
1972 	struct sockaddr *dst;
1973 	struct ifnet *ifp;
1974 
1975 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
1976 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
1977 		return;
1978 	ifa = ifaof_ifpforaddr(dst, ifp);
1979 	if (ifa) {
1980 		oifa = rt->rt_ifa;
1981 		rt->rt_ifa = ifa;
1982 		ifa_free(oifa);
1983 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1984 			ifa->ifa_rtrequest(cmd, rt, info);
1985 	}
1986 }
1987 
1988 struct sockaddr_dl *
1989 link_alloc_sdl(size_t size, int flags)
1990 {
1991 
1992 	return (malloc(size, M_TEMP, flags));
1993 }
1994 
1995 void
1996 link_free_sdl(struct sockaddr *sa)
1997 {
1998 	free(sa, M_TEMP);
1999 }
2000 
2001 /*
2002  * Fills in given sdl with interface basic info.
2003  * Returns pointer to filled sdl.
2004  */
2005 struct sockaddr_dl *
2006 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2007 {
2008 	struct sockaddr_dl *sdl;
2009 
2010 	sdl = (struct sockaddr_dl *)paddr;
2011 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2012 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2013 	sdl->sdl_family = AF_LINK;
2014 	sdl->sdl_index = ifp->if_index;
2015 	sdl->sdl_type = iftype;
2016 
2017 	return (sdl);
2018 }
2019 
2020 /*
2021  * Mark an interface down and notify protocols of
2022  * the transition.
2023  */
2024 static void
2025 if_unroute(struct ifnet *ifp, int flag, int fam)
2026 {
2027 	struct ifaddr *ifa;
2028 
2029 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2030 
2031 	ifp->if_flags &= ~flag;
2032 	getmicrotime(&ifp->if_lastchange);
2033 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2034 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2035 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2036 	ifp->if_qflush(ifp);
2037 
2038 	if (ifp->if_carp)
2039 		(*carp_linkstate_p)(ifp);
2040 	rt_ifmsg(ifp);
2041 }
2042 
2043 /*
2044  * Mark an interface up and notify protocols of
2045  * the transition.
2046  */
2047 static void
2048 if_route(struct ifnet *ifp, int flag, int fam)
2049 {
2050 	struct ifaddr *ifa;
2051 
2052 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2053 
2054 	ifp->if_flags |= flag;
2055 	getmicrotime(&ifp->if_lastchange);
2056 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2057 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2058 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2059 	if (ifp->if_carp)
2060 		(*carp_linkstate_p)(ifp);
2061 	rt_ifmsg(ifp);
2062 #ifdef INET6
2063 	in6_if_up(ifp);
2064 #endif
2065 }
2066 
2067 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2068 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2069 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2070 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2071 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2072 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2073 void	*(*vlan_cookie_p)(struct ifnet *);
2074 
2075 /*
2076  * Handle a change in the interface link state. To avoid LORs
2077  * between driver lock and upper layer locks, as well as possible
2078  * recursions, we post event to taskqueue, and all job
2079  * is done in static do_link_state_change().
2080  */
2081 void
2082 if_link_state_change(struct ifnet *ifp, int link_state)
2083 {
2084 	/* Return if state hasn't changed. */
2085 	if (ifp->if_link_state == link_state)
2086 		return;
2087 
2088 	ifp->if_link_state = link_state;
2089 
2090 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2091 }
2092 
2093 static void
2094 do_link_state_change(void *arg, int pending)
2095 {
2096 	struct ifnet *ifp = (struct ifnet *)arg;
2097 	int link_state = ifp->if_link_state;
2098 	CURVNET_SET(ifp->if_vnet);
2099 
2100 	/* Notify that the link state has changed. */
2101 	rt_ifmsg(ifp);
2102 	if (ifp->if_vlantrunk != NULL)
2103 		(*vlan_link_state_p)(ifp);
2104 
2105 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2106 	    ifp->if_l2com != NULL)
2107 		(*ng_ether_link_state_p)(ifp, link_state);
2108 	if (ifp->if_carp)
2109 		(*carp_linkstate_p)(ifp);
2110 	if (ifp->if_bridge)
2111 		(*bridge_linkstate_p)(ifp);
2112 	if (ifp->if_lagg)
2113 		(*lagg_linkstate_p)(ifp, link_state);
2114 
2115 	if (IS_DEFAULT_VNET(curvnet))
2116 		devctl_notify("IFNET", ifp->if_xname,
2117 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2118 		    NULL);
2119 	if (pending > 1)
2120 		if_printf(ifp, "%d link states coalesced\n", pending);
2121 	if (log_link_state_change)
2122 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2123 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2124 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2125 	CURVNET_RESTORE();
2126 }
2127 
2128 /*
2129  * Mark an interface down and notify protocols of
2130  * the transition.
2131  */
2132 void
2133 if_down(struct ifnet *ifp)
2134 {
2135 
2136 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2137 }
2138 
2139 /*
2140  * Mark an interface up and notify protocols of
2141  * the transition.
2142  */
2143 void
2144 if_up(struct ifnet *ifp)
2145 {
2146 
2147 	if_route(ifp, IFF_UP, AF_UNSPEC);
2148 }
2149 
2150 /*
2151  * Flush an interface queue.
2152  */
2153 void
2154 if_qflush(struct ifnet *ifp)
2155 {
2156 	struct mbuf *m, *n;
2157 	struct ifaltq *ifq;
2158 
2159 	ifq = &ifp->if_snd;
2160 	IFQ_LOCK(ifq);
2161 #ifdef ALTQ
2162 	if (ALTQ_IS_ENABLED(ifq))
2163 		ALTQ_PURGE(ifq);
2164 #endif
2165 	n = ifq->ifq_head;
2166 	while ((m = n) != NULL) {
2167 		n = m->m_nextpkt;
2168 		m_freem(m);
2169 	}
2170 	ifq->ifq_head = 0;
2171 	ifq->ifq_tail = 0;
2172 	ifq->ifq_len = 0;
2173 	IFQ_UNLOCK(ifq);
2174 }
2175 
2176 /*
2177  * Map interface name to interface structure pointer, with or without
2178  * returning a reference.
2179  */
2180 struct ifnet *
2181 ifunit_ref(const char *name)
2182 {
2183 	struct ifnet *ifp;
2184 
2185 	IFNET_RLOCK_NOSLEEP();
2186 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2187 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2188 		    !(ifp->if_flags & IFF_DYING))
2189 			break;
2190 	}
2191 	if (ifp != NULL)
2192 		if_ref(ifp);
2193 	IFNET_RUNLOCK_NOSLEEP();
2194 	return (ifp);
2195 }
2196 
2197 struct ifnet *
2198 ifunit(const char *name)
2199 {
2200 	struct ifnet *ifp;
2201 
2202 	IFNET_RLOCK_NOSLEEP();
2203 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2204 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2205 			break;
2206 	}
2207 	IFNET_RUNLOCK_NOSLEEP();
2208 	return (ifp);
2209 }
2210 
2211 /*
2212  * Hardware specific interface ioctls.
2213  */
2214 static int
2215 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2216 {
2217 	struct ifreq *ifr;
2218 	int error = 0;
2219 	int new_flags, temp_flags;
2220 	size_t namelen, onamelen;
2221 	size_t descrlen;
2222 	char *descrbuf, *odescrbuf;
2223 	char new_name[IFNAMSIZ];
2224 	struct ifaddr *ifa;
2225 	struct sockaddr_dl *sdl;
2226 
2227 	ifr = (struct ifreq *)data;
2228 	switch (cmd) {
2229 	case SIOCGIFINDEX:
2230 		ifr->ifr_index = ifp->if_index;
2231 		break;
2232 
2233 	case SIOCGIFFLAGS:
2234 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2235 		ifr->ifr_flags = temp_flags & 0xffff;
2236 		ifr->ifr_flagshigh = temp_flags >> 16;
2237 		break;
2238 
2239 	case SIOCGIFCAP:
2240 		ifr->ifr_reqcap = ifp->if_capabilities;
2241 		ifr->ifr_curcap = ifp->if_capenable;
2242 		break;
2243 
2244 #ifdef MAC
2245 	case SIOCGIFMAC:
2246 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2247 		break;
2248 #endif
2249 
2250 	case SIOCGIFMETRIC:
2251 		ifr->ifr_metric = ifp->if_metric;
2252 		break;
2253 
2254 	case SIOCGIFMTU:
2255 		ifr->ifr_mtu = ifp->if_mtu;
2256 		break;
2257 
2258 	case SIOCGIFPHYS:
2259 		/* XXXGL: did this ever worked? */
2260 		ifr->ifr_phys = 0;
2261 		break;
2262 
2263 	case SIOCGIFDESCR:
2264 		error = 0;
2265 		sx_slock(&ifdescr_sx);
2266 		if (ifp->if_description == NULL)
2267 			error = ENOMSG;
2268 		else {
2269 			/* space for terminating nul */
2270 			descrlen = strlen(ifp->if_description) + 1;
2271 			if (ifr->ifr_buffer.length < descrlen)
2272 				ifr->ifr_buffer.buffer = NULL;
2273 			else
2274 				error = copyout(ifp->if_description,
2275 				    ifr->ifr_buffer.buffer, descrlen);
2276 			ifr->ifr_buffer.length = descrlen;
2277 		}
2278 		sx_sunlock(&ifdescr_sx);
2279 		break;
2280 
2281 	case SIOCSIFDESCR:
2282 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2283 		if (error)
2284 			return (error);
2285 
2286 		/*
2287 		 * Copy only (length-1) bytes to make sure that
2288 		 * if_description is always nul terminated.  The
2289 		 * length parameter is supposed to count the
2290 		 * terminating nul in.
2291 		 */
2292 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2293 			return (ENAMETOOLONG);
2294 		else if (ifr->ifr_buffer.length == 0)
2295 			descrbuf = NULL;
2296 		else {
2297 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2298 			    M_WAITOK | M_ZERO);
2299 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2300 			    ifr->ifr_buffer.length - 1);
2301 			if (error) {
2302 				free(descrbuf, M_IFDESCR);
2303 				break;
2304 			}
2305 		}
2306 
2307 		sx_xlock(&ifdescr_sx);
2308 		odescrbuf = ifp->if_description;
2309 		ifp->if_description = descrbuf;
2310 		sx_xunlock(&ifdescr_sx);
2311 
2312 		getmicrotime(&ifp->if_lastchange);
2313 		free(odescrbuf, M_IFDESCR);
2314 		break;
2315 
2316 	case SIOCGIFFIB:
2317 		ifr->ifr_fib = ifp->if_fib;
2318 		break;
2319 
2320 	case SIOCSIFFIB:
2321 		error = priv_check(td, PRIV_NET_SETIFFIB);
2322 		if (error)
2323 			return (error);
2324 		if (ifr->ifr_fib >= rt_numfibs)
2325 			return (EINVAL);
2326 
2327 		ifp->if_fib = ifr->ifr_fib;
2328 		break;
2329 
2330 	case SIOCSIFFLAGS:
2331 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2332 		if (error)
2333 			return (error);
2334 		/*
2335 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2336 		 * check, so we don't need special handling here yet.
2337 		 */
2338 		new_flags = (ifr->ifr_flags & 0xffff) |
2339 		    (ifr->ifr_flagshigh << 16);
2340 		if (ifp->if_flags & IFF_UP &&
2341 		    (new_flags & IFF_UP) == 0) {
2342 			if_down(ifp);
2343 		} else if (new_flags & IFF_UP &&
2344 		    (ifp->if_flags & IFF_UP) == 0) {
2345 			if_up(ifp);
2346 		}
2347 		/* See if permanently promiscuous mode bit is about to flip */
2348 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2349 			if (new_flags & IFF_PPROMISC)
2350 				ifp->if_flags |= IFF_PROMISC;
2351 			else if (ifp->if_pcount == 0)
2352 				ifp->if_flags &= ~IFF_PROMISC;
2353 			if (log_promisc_mode_change)
2354                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2355                                     ifp->if_xname,
2356                                     ((new_flags & IFF_PPROMISC) ?
2357                                      "enabled" : "disabled"));
2358 		}
2359 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2360 			(new_flags &~ IFF_CANTCHANGE);
2361 		if (ifp->if_ioctl) {
2362 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2363 		}
2364 		getmicrotime(&ifp->if_lastchange);
2365 		break;
2366 
2367 	case SIOCSIFCAP:
2368 		error = priv_check(td, PRIV_NET_SETIFCAP);
2369 		if (error)
2370 			return (error);
2371 		if (ifp->if_ioctl == NULL)
2372 			return (EOPNOTSUPP);
2373 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2374 			return (EINVAL);
2375 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2376 		if (error == 0)
2377 			getmicrotime(&ifp->if_lastchange);
2378 		break;
2379 
2380 #ifdef MAC
2381 	case SIOCSIFMAC:
2382 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2383 		break;
2384 #endif
2385 
2386 	case SIOCSIFNAME:
2387 		error = priv_check(td, PRIV_NET_SETIFNAME);
2388 		if (error)
2389 			return (error);
2390 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2391 		if (error != 0)
2392 			return (error);
2393 		if (new_name[0] == '\0')
2394 			return (EINVAL);
2395 		if (new_name[IFNAMSIZ-1] != '\0') {
2396 			new_name[IFNAMSIZ-1] = '\0';
2397 			if (strlen(new_name) == IFNAMSIZ-1)
2398 				return (EINVAL);
2399 		}
2400 		if (ifunit(new_name) != NULL)
2401 			return (EEXIST);
2402 
2403 		/*
2404 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2405 		 * and there are numerous other races with the
2406 		 * ifunit() checks not being atomic with namespace
2407 		 * changes (renames, vmoves, if_attach, etc).
2408 		 */
2409 		ifp->if_flags |= IFF_RENAMING;
2410 
2411 		/* Announce the departure of the interface. */
2412 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2413 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2414 
2415 		log(LOG_INFO, "%s: changing name to '%s'\n",
2416 		    ifp->if_xname, new_name);
2417 
2418 		IF_ADDR_WLOCK(ifp);
2419 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2420 		ifa = ifp->if_addr;
2421 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2422 		namelen = strlen(new_name);
2423 		onamelen = sdl->sdl_nlen;
2424 		/*
2425 		 * Move the address if needed.  This is safe because we
2426 		 * allocate space for a name of length IFNAMSIZ when we
2427 		 * create this in if_attach().
2428 		 */
2429 		if (namelen != onamelen) {
2430 			bcopy(sdl->sdl_data + onamelen,
2431 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2432 		}
2433 		bcopy(new_name, sdl->sdl_data, namelen);
2434 		sdl->sdl_nlen = namelen;
2435 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2436 		bzero(sdl->sdl_data, onamelen);
2437 		while (namelen != 0)
2438 			sdl->sdl_data[--namelen] = 0xff;
2439 		IF_ADDR_WUNLOCK(ifp);
2440 
2441 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2442 		/* Announce the return of the interface. */
2443 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2444 
2445 		ifp->if_flags &= ~IFF_RENAMING;
2446 		break;
2447 
2448 #ifdef VIMAGE
2449 	case SIOCSIFVNET:
2450 		error = priv_check(td, PRIV_NET_SETIFVNET);
2451 		if (error)
2452 			return (error);
2453 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2454 		break;
2455 #endif
2456 
2457 	case SIOCSIFMETRIC:
2458 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2459 		if (error)
2460 			return (error);
2461 		ifp->if_metric = ifr->ifr_metric;
2462 		getmicrotime(&ifp->if_lastchange);
2463 		break;
2464 
2465 	case SIOCSIFPHYS:
2466 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2467 		if (error)
2468 			return (error);
2469 		if (ifp->if_ioctl == NULL)
2470 			return (EOPNOTSUPP);
2471 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2472 		if (error == 0)
2473 			getmicrotime(&ifp->if_lastchange);
2474 		break;
2475 
2476 	case SIOCSIFMTU:
2477 	{
2478 		u_long oldmtu = ifp->if_mtu;
2479 
2480 		error = priv_check(td, PRIV_NET_SETIFMTU);
2481 		if (error)
2482 			return (error);
2483 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2484 			return (EINVAL);
2485 		if (ifp->if_ioctl == NULL)
2486 			return (EOPNOTSUPP);
2487 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2488 		if (error == 0) {
2489 			getmicrotime(&ifp->if_lastchange);
2490 			rt_ifmsg(ifp);
2491 		}
2492 		/*
2493 		 * If the link MTU changed, do network layer specific procedure.
2494 		 */
2495 		if (ifp->if_mtu != oldmtu) {
2496 #ifdef INET6
2497 			nd6_setmtu(ifp);
2498 #endif
2499 			rt_updatemtu(ifp);
2500 		}
2501 		break;
2502 	}
2503 
2504 	case SIOCADDMULTI:
2505 	case SIOCDELMULTI:
2506 		if (cmd == SIOCADDMULTI)
2507 			error = priv_check(td, PRIV_NET_ADDMULTI);
2508 		else
2509 			error = priv_check(td, PRIV_NET_DELMULTI);
2510 		if (error)
2511 			return (error);
2512 
2513 		/* Don't allow group membership on non-multicast interfaces. */
2514 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2515 			return (EOPNOTSUPP);
2516 
2517 		/* Don't let users screw up protocols' entries. */
2518 		if (ifr->ifr_addr.sa_family != AF_LINK)
2519 			return (EINVAL);
2520 
2521 		if (cmd == SIOCADDMULTI) {
2522 			struct ifmultiaddr *ifma;
2523 
2524 			/*
2525 			 * Userland is only permitted to join groups once
2526 			 * via the if_addmulti() KPI, because it cannot hold
2527 			 * struct ifmultiaddr * between calls. It may also
2528 			 * lose a race while we check if the membership
2529 			 * already exists.
2530 			 */
2531 			IF_ADDR_RLOCK(ifp);
2532 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2533 			IF_ADDR_RUNLOCK(ifp);
2534 			if (ifma != NULL)
2535 				error = EADDRINUSE;
2536 			else
2537 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2538 		} else {
2539 			error = if_delmulti(ifp, &ifr->ifr_addr);
2540 		}
2541 		if (error == 0)
2542 			getmicrotime(&ifp->if_lastchange);
2543 		break;
2544 
2545 	case SIOCSIFPHYADDR:
2546 	case SIOCDIFPHYADDR:
2547 #ifdef INET6
2548 	case SIOCSIFPHYADDR_IN6:
2549 #endif
2550 	case SIOCSIFMEDIA:
2551 	case SIOCSIFGENERIC:
2552 		error = priv_check(td, PRIV_NET_HWIOCTL);
2553 		if (error)
2554 			return (error);
2555 		if (ifp->if_ioctl == NULL)
2556 			return (EOPNOTSUPP);
2557 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2558 		if (error == 0)
2559 			getmicrotime(&ifp->if_lastchange);
2560 		break;
2561 
2562 	case SIOCGIFSTATUS:
2563 	case SIOCGIFPSRCADDR:
2564 	case SIOCGIFPDSTADDR:
2565 	case SIOCGIFMEDIA:
2566 	case SIOCGIFXMEDIA:
2567 	case SIOCGIFGENERIC:
2568 		if (ifp->if_ioctl == NULL)
2569 			return (EOPNOTSUPP);
2570 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2571 		break;
2572 
2573 	case SIOCSIFLLADDR:
2574 		error = priv_check(td, PRIV_NET_SETLLADDR);
2575 		if (error)
2576 			return (error);
2577 		error = if_setlladdr(ifp,
2578 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2579 		break;
2580 
2581 	case SIOCAIFGROUP:
2582 	{
2583 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2584 
2585 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2586 		if (error)
2587 			return (error);
2588 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2589 			return (error);
2590 		break;
2591 	}
2592 
2593 	case SIOCGIFGROUP:
2594 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2595 			return (error);
2596 		break;
2597 
2598 	case SIOCDIFGROUP:
2599 	{
2600 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2601 
2602 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2603 		if (error)
2604 			return (error);
2605 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2606 			return (error);
2607 		break;
2608 	}
2609 
2610 	default:
2611 		error = ENOIOCTL;
2612 		break;
2613 	}
2614 	return (error);
2615 }
2616 
2617 #ifdef COMPAT_FREEBSD32
2618 struct ifconf32 {
2619 	int32_t	ifc_len;
2620 	union {
2621 		uint32_t	ifcu_buf;
2622 		uint32_t	ifcu_req;
2623 	} ifc_ifcu;
2624 };
2625 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2626 #endif
2627 
2628 /*
2629  * Interface ioctls.
2630  */
2631 int
2632 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2633 {
2634 	struct ifnet *ifp;
2635 	struct ifreq *ifr;
2636 	int error;
2637 	int oif_flags;
2638 
2639 	CURVNET_SET(so->so_vnet);
2640 	switch (cmd) {
2641 	case SIOCGIFCONF:
2642 		error = ifconf(cmd, data);
2643 		CURVNET_RESTORE();
2644 		return (error);
2645 
2646 #ifdef COMPAT_FREEBSD32
2647 	case SIOCGIFCONF32:
2648 		{
2649 			struct ifconf32 *ifc32;
2650 			struct ifconf ifc;
2651 
2652 			ifc32 = (struct ifconf32 *)data;
2653 			ifc.ifc_len = ifc32->ifc_len;
2654 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2655 
2656 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2657 			CURVNET_RESTORE();
2658 			if (error == 0)
2659 				ifc32->ifc_len = ifc.ifc_len;
2660 			return (error);
2661 		}
2662 #endif
2663 	}
2664 	ifr = (struct ifreq *)data;
2665 
2666 	switch (cmd) {
2667 #ifdef VIMAGE
2668 	case SIOCSIFRVNET:
2669 		error = priv_check(td, PRIV_NET_SETIFVNET);
2670 		if (error == 0)
2671 			error = if_vmove_reclaim(td, ifr->ifr_name,
2672 			    ifr->ifr_jid);
2673 		CURVNET_RESTORE();
2674 		return (error);
2675 #endif
2676 	case SIOCIFCREATE:
2677 	case SIOCIFCREATE2:
2678 		error = priv_check(td, PRIV_NET_IFCREATE);
2679 		if (error == 0)
2680 			error = if_clone_create(ifr->ifr_name,
2681 			    sizeof(ifr->ifr_name),
2682 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2683 		CURVNET_RESTORE();
2684 		return (error);
2685 	case SIOCIFDESTROY:
2686 		error = priv_check(td, PRIV_NET_IFDESTROY);
2687 		if (error == 0)
2688 			error = if_clone_destroy(ifr->ifr_name);
2689 		CURVNET_RESTORE();
2690 		return (error);
2691 
2692 	case SIOCIFGCLONERS:
2693 		error = if_clone_list((struct if_clonereq *)data);
2694 		CURVNET_RESTORE();
2695 		return (error);
2696 	case SIOCGIFGMEMB:
2697 		error = if_getgroupmembers((struct ifgroupreq *)data);
2698 		CURVNET_RESTORE();
2699 		return (error);
2700 #if defined(INET) || defined(INET6)
2701 	case SIOCSVH:
2702 	case SIOCGVH:
2703 		if (carp_ioctl_p == NULL)
2704 			error = EPROTONOSUPPORT;
2705 		else
2706 			error = (*carp_ioctl_p)(ifr, cmd, td);
2707 		CURVNET_RESTORE();
2708 		return (error);
2709 #endif
2710 	}
2711 
2712 	ifp = ifunit_ref(ifr->ifr_name);
2713 	if (ifp == NULL) {
2714 		CURVNET_RESTORE();
2715 		return (ENXIO);
2716 	}
2717 
2718 	error = ifhwioctl(cmd, ifp, data, td);
2719 	if (error != ENOIOCTL) {
2720 		if_rele(ifp);
2721 		CURVNET_RESTORE();
2722 		return (error);
2723 	}
2724 
2725 	oif_flags = ifp->if_flags;
2726 	if (so->so_proto == NULL) {
2727 		if_rele(ifp);
2728 		CURVNET_RESTORE();
2729 		return (EOPNOTSUPP);
2730 	}
2731 
2732 	/*
2733 	 * Pass the request on to the socket control method, and if the
2734 	 * latter returns EOPNOTSUPP, directly to the interface.
2735 	 *
2736 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2737 	 * trust SIOCSIFADDR et al to come from an already privileged
2738 	 * layer, and do not perform any credentials checks or input
2739 	 * validation.
2740 	 */
2741 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2742 	    ifp, td));
2743 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2744 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2745 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2746 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2747 
2748 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2749 #ifdef INET6
2750 		if (ifp->if_flags & IFF_UP)
2751 			in6_if_up(ifp);
2752 #endif
2753 	}
2754 	if_rele(ifp);
2755 	CURVNET_RESTORE();
2756 	return (error);
2757 }
2758 
2759 /*
2760  * The code common to handling reference counted flags,
2761  * e.g., in ifpromisc() and if_allmulti().
2762  * The "pflag" argument can specify a permanent mode flag to check,
2763  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2764  *
2765  * Only to be used on stack-owned flags, not driver-owned flags.
2766  */
2767 static int
2768 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2769 {
2770 	struct ifreq ifr;
2771 	int error;
2772 	int oldflags, oldcount;
2773 
2774 	/* Sanity checks to catch programming errors */
2775 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2776 	    ("%s: setting driver-owned flag %d", __func__, flag));
2777 
2778 	if (onswitch)
2779 		KASSERT(*refcount >= 0,
2780 		    ("%s: increment negative refcount %d for flag %d",
2781 		    __func__, *refcount, flag));
2782 	else
2783 		KASSERT(*refcount > 0,
2784 		    ("%s: decrement non-positive refcount %d for flag %d",
2785 		    __func__, *refcount, flag));
2786 
2787 	/* In case this mode is permanent, just touch refcount */
2788 	if (ifp->if_flags & pflag) {
2789 		*refcount += onswitch ? 1 : -1;
2790 		return (0);
2791 	}
2792 
2793 	/* Save ifnet parameters for if_ioctl() may fail */
2794 	oldcount = *refcount;
2795 	oldflags = ifp->if_flags;
2796 
2797 	/*
2798 	 * See if we aren't the only and touching refcount is enough.
2799 	 * Actually toggle interface flag if we are the first or last.
2800 	 */
2801 	if (onswitch) {
2802 		if ((*refcount)++)
2803 			return (0);
2804 		ifp->if_flags |= flag;
2805 	} else {
2806 		if (--(*refcount))
2807 			return (0);
2808 		ifp->if_flags &= ~flag;
2809 	}
2810 
2811 	/* Call down the driver since we've changed interface flags */
2812 	if (ifp->if_ioctl == NULL) {
2813 		error = EOPNOTSUPP;
2814 		goto recover;
2815 	}
2816 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2817 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2818 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2819 	if (error)
2820 		goto recover;
2821 	/* Notify userland that interface flags have changed */
2822 	rt_ifmsg(ifp);
2823 	return (0);
2824 
2825 recover:
2826 	/* Recover after driver error */
2827 	*refcount = oldcount;
2828 	ifp->if_flags = oldflags;
2829 	return (error);
2830 }
2831 
2832 /*
2833  * Set/clear promiscuous mode on interface ifp based on the truth value
2834  * of pswitch.  The calls are reference counted so that only the first
2835  * "on" request actually has an effect, as does the final "off" request.
2836  * Results are undefined if the "off" and "on" requests are not matched.
2837  */
2838 int
2839 ifpromisc(struct ifnet *ifp, int pswitch)
2840 {
2841 	int error;
2842 	int oldflags = ifp->if_flags;
2843 
2844 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2845 			   &ifp->if_pcount, pswitch);
2846 	/* If promiscuous mode status has changed, log a message */
2847 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2848             log_promisc_mode_change)
2849 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2850 		    ifp->if_xname,
2851 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2852 	return (error);
2853 }
2854 
2855 /*
2856  * Return interface configuration
2857  * of system.  List may be used
2858  * in later ioctl's (above) to get
2859  * other information.
2860  */
2861 /*ARGSUSED*/
2862 static int
2863 ifconf(u_long cmd, caddr_t data)
2864 {
2865 	struct ifconf *ifc = (struct ifconf *)data;
2866 	struct ifnet *ifp;
2867 	struct ifaddr *ifa;
2868 	struct ifreq ifr;
2869 	struct sbuf *sb;
2870 	int error, full = 0, valid_len, max_len;
2871 
2872 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2873 	max_len = MAXPHYS - 1;
2874 
2875 	/* Prevent hostile input from being able to crash the system */
2876 	if (ifc->ifc_len <= 0)
2877 		return (EINVAL);
2878 
2879 again:
2880 	if (ifc->ifc_len <= max_len) {
2881 		max_len = ifc->ifc_len;
2882 		full = 1;
2883 	}
2884 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2885 	max_len = 0;
2886 	valid_len = 0;
2887 
2888 	IFNET_RLOCK();
2889 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2890 		int addrs;
2891 
2892 		/*
2893 		 * Zero the ifr_name buffer to make sure we don't
2894 		 * disclose the contents of the stack.
2895 		 */
2896 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2897 
2898 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2899 		    >= sizeof(ifr.ifr_name)) {
2900 			sbuf_delete(sb);
2901 			IFNET_RUNLOCK();
2902 			return (ENAMETOOLONG);
2903 		}
2904 
2905 		addrs = 0;
2906 		IF_ADDR_RLOCK(ifp);
2907 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2908 			struct sockaddr *sa = ifa->ifa_addr;
2909 
2910 			if (prison_if(curthread->td_ucred, sa) != 0)
2911 				continue;
2912 			addrs++;
2913 			if (sa->sa_len <= sizeof(*sa)) {
2914 				ifr.ifr_addr = *sa;
2915 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2916 				max_len += sizeof(ifr);
2917 			} else {
2918 				sbuf_bcat(sb, &ifr,
2919 				    offsetof(struct ifreq, ifr_addr));
2920 				max_len += offsetof(struct ifreq, ifr_addr);
2921 				sbuf_bcat(sb, sa, sa->sa_len);
2922 				max_len += sa->sa_len;
2923 			}
2924 
2925 			if (sbuf_error(sb) == 0)
2926 				valid_len = sbuf_len(sb);
2927 		}
2928 		IF_ADDR_RUNLOCK(ifp);
2929 		if (addrs == 0) {
2930 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2931 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2932 			max_len += sizeof(ifr);
2933 
2934 			if (sbuf_error(sb) == 0)
2935 				valid_len = sbuf_len(sb);
2936 		}
2937 	}
2938 	IFNET_RUNLOCK();
2939 
2940 	/*
2941 	 * If we didn't allocate enough space (uncommon), try again.  If
2942 	 * we have already allocated as much space as we are allowed,
2943 	 * return what we've got.
2944 	 */
2945 	if (valid_len != max_len && !full) {
2946 		sbuf_delete(sb);
2947 		goto again;
2948 	}
2949 
2950 	ifc->ifc_len = valid_len;
2951 	sbuf_finish(sb);
2952 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2953 	sbuf_delete(sb);
2954 	return (error);
2955 }
2956 
2957 /*
2958  * Just like ifpromisc(), but for all-multicast-reception mode.
2959  */
2960 int
2961 if_allmulti(struct ifnet *ifp, int onswitch)
2962 {
2963 
2964 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2965 }
2966 
2967 struct ifmultiaddr *
2968 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
2969 {
2970 	struct ifmultiaddr *ifma;
2971 
2972 	IF_ADDR_LOCK_ASSERT(ifp);
2973 
2974 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2975 		if (sa->sa_family == AF_LINK) {
2976 			if (sa_dl_equal(ifma->ifma_addr, sa))
2977 				break;
2978 		} else {
2979 			if (sa_equal(ifma->ifma_addr, sa))
2980 				break;
2981 		}
2982 	}
2983 
2984 	return ifma;
2985 }
2986 
2987 /*
2988  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2989  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2990  * the ifnet multicast address list here, so the caller must do that and
2991  * other setup work (such as notifying the device driver).  The reference
2992  * count is initialized to 1.
2993  */
2994 static struct ifmultiaddr *
2995 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2996     int mflags)
2997 {
2998 	struct ifmultiaddr *ifma;
2999 	struct sockaddr *dupsa;
3000 
3001 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3002 	    M_ZERO);
3003 	if (ifma == NULL)
3004 		return (NULL);
3005 
3006 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3007 	if (dupsa == NULL) {
3008 		free(ifma, M_IFMADDR);
3009 		return (NULL);
3010 	}
3011 	bcopy(sa, dupsa, sa->sa_len);
3012 	ifma->ifma_addr = dupsa;
3013 
3014 	ifma->ifma_ifp = ifp;
3015 	ifma->ifma_refcount = 1;
3016 	ifma->ifma_protospec = NULL;
3017 
3018 	if (llsa == NULL) {
3019 		ifma->ifma_lladdr = NULL;
3020 		return (ifma);
3021 	}
3022 
3023 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3024 	if (dupsa == NULL) {
3025 		free(ifma->ifma_addr, M_IFMADDR);
3026 		free(ifma, M_IFMADDR);
3027 		return (NULL);
3028 	}
3029 	bcopy(llsa, dupsa, llsa->sa_len);
3030 	ifma->ifma_lladdr = dupsa;
3031 
3032 	return (ifma);
3033 }
3034 
3035 /*
3036  * if_freemulti: free ifmultiaddr structure and possibly attached related
3037  * addresses.  The caller is responsible for implementing reference
3038  * counting, notifying the driver, handling routing messages, and releasing
3039  * any dependent link layer state.
3040  */
3041 static void
3042 if_freemulti(struct ifmultiaddr *ifma)
3043 {
3044 
3045 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3046 	    ifma->ifma_refcount));
3047 
3048 	if (ifma->ifma_lladdr != NULL)
3049 		free(ifma->ifma_lladdr, M_IFMADDR);
3050 	free(ifma->ifma_addr, M_IFMADDR);
3051 	free(ifma, M_IFMADDR);
3052 }
3053 
3054 /*
3055  * Register an additional multicast address with a network interface.
3056  *
3057  * - If the address is already present, bump the reference count on the
3058  *   address and return.
3059  * - If the address is not link-layer, look up a link layer address.
3060  * - Allocate address structures for one or both addresses, and attach to the
3061  *   multicast address list on the interface.  If automatically adding a link
3062  *   layer address, the protocol address will own a reference to the link
3063  *   layer address, to be freed when it is freed.
3064  * - Notify the network device driver of an addition to the multicast address
3065  *   list.
3066  *
3067  * 'sa' points to caller-owned memory with the desired multicast address.
3068  *
3069  * 'retifma' will be used to return a pointer to the resulting multicast
3070  * address reference, if desired.
3071  */
3072 int
3073 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3074     struct ifmultiaddr **retifma)
3075 {
3076 	struct ifmultiaddr *ifma, *ll_ifma;
3077 	struct sockaddr *llsa;
3078 	struct sockaddr_dl sdl;
3079 	int error;
3080 
3081 	/*
3082 	 * If the address is already present, return a new reference to it;
3083 	 * otherwise, allocate storage and set up a new address.
3084 	 */
3085 	IF_ADDR_WLOCK(ifp);
3086 	ifma = if_findmulti(ifp, sa);
3087 	if (ifma != NULL) {
3088 		ifma->ifma_refcount++;
3089 		if (retifma != NULL)
3090 			*retifma = ifma;
3091 		IF_ADDR_WUNLOCK(ifp);
3092 		return (0);
3093 	}
3094 
3095 	/*
3096 	 * The address isn't already present; resolve the protocol address
3097 	 * into a link layer address, and then look that up, bump its
3098 	 * refcount or allocate an ifma for that also.
3099 	 * Most link layer resolving functions returns address data which
3100 	 * fits inside default sockaddr_dl structure. However callback
3101 	 * can allocate another sockaddr structure, in that case we need to
3102 	 * free it later.
3103 	 */
3104 	llsa = NULL;
3105 	ll_ifma = NULL;
3106 	if (ifp->if_resolvemulti != NULL) {
3107 		/* Provide called function with buffer size information */
3108 		sdl.sdl_len = sizeof(sdl);
3109 		llsa = (struct sockaddr *)&sdl;
3110 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3111 		if (error)
3112 			goto unlock_out;
3113 	}
3114 
3115 	/*
3116 	 * Allocate the new address.  Don't hook it up yet, as we may also
3117 	 * need to allocate a link layer multicast address.
3118 	 */
3119 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3120 	if (ifma == NULL) {
3121 		error = ENOMEM;
3122 		goto free_llsa_out;
3123 	}
3124 
3125 	/*
3126 	 * If a link layer address is found, we'll need to see if it's
3127 	 * already present in the address list, or allocate is as well.
3128 	 * When this block finishes, the link layer address will be on the
3129 	 * list.
3130 	 */
3131 	if (llsa != NULL) {
3132 		ll_ifma = if_findmulti(ifp, llsa);
3133 		if (ll_ifma == NULL) {
3134 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3135 			if (ll_ifma == NULL) {
3136 				--ifma->ifma_refcount;
3137 				if_freemulti(ifma);
3138 				error = ENOMEM;
3139 				goto free_llsa_out;
3140 			}
3141 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3142 			    ifma_link);
3143 		} else
3144 			ll_ifma->ifma_refcount++;
3145 		ifma->ifma_llifma = ll_ifma;
3146 	}
3147 
3148 	/*
3149 	 * We now have a new multicast address, ifma, and possibly a new or
3150 	 * referenced link layer address.  Add the primary address to the
3151 	 * ifnet address list.
3152 	 */
3153 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3154 
3155 	if (retifma != NULL)
3156 		*retifma = ifma;
3157 
3158 	/*
3159 	 * Must generate the message while holding the lock so that 'ifma'
3160 	 * pointer is still valid.
3161 	 */
3162 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3163 	IF_ADDR_WUNLOCK(ifp);
3164 
3165 	/*
3166 	 * We are certain we have added something, so call down to the
3167 	 * interface to let them know about it.
3168 	 */
3169 	if (ifp->if_ioctl != NULL) {
3170 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3171 	}
3172 
3173 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3174 		link_free_sdl(llsa);
3175 
3176 	return (0);
3177 
3178 free_llsa_out:
3179 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3180 		link_free_sdl(llsa);
3181 
3182 unlock_out:
3183 	IF_ADDR_WUNLOCK(ifp);
3184 	return (error);
3185 }
3186 
3187 /*
3188  * Delete a multicast group membership by network-layer group address.
3189  *
3190  * Returns ENOENT if the entry could not be found. If ifp no longer
3191  * exists, results are undefined. This entry point should only be used
3192  * from subsystems which do appropriate locking to hold ifp for the
3193  * duration of the call.
3194  * Network-layer protocol domains must use if_delmulti_ifma().
3195  */
3196 int
3197 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3198 {
3199 	struct ifmultiaddr *ifma;
3200 	int lastref;
3201 #ifdef INVARIANTS
3202 	struct ifnet *oifp;
3203 
3204 	IFNET_RLOCK_NOSLEEP();
3205 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3206 		if (ifp == oifp)
3207 			break;
3208 	if (ifp != oifp)
3209 		ifp = NULL;
3210 	IFNET_RUNLOCK_NOSLEEP();
3211 
3212 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3213 #endif
3214 	if (ifp == NULL)
3215 		return (ENOENT);
3216 
3217 	IF_ADDR_WLOCK(ifp);
3218 	lastref = 0;
3219 	ifma = if_findmulti(ifp, sa);
3220 	if (ifma != NULL)
3221 		lastref = if_delmulti_locked(ifp, ifma, 0);
3222 	IF_ADDR_WUNLOCK(ifp);
3223 
3224 	if (ifma == NULL)
3225 		return (ENOENT);
3226 
3227 	if (lastref && ifp->if_ioctl != NULL) {
3228 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3229 	}
3230 
3231 	return (0);
3232 }
3233 
3234 /*
3235  * Delete all multicast group membership for an interface.
3236  * Should be used to quickly flush all multicast filters.
3237  */
3238 void
3239 if_delallmulti(struct ifnet *ifp)
3240 {
3241 	struct ifmultiaddr *ifma;
3242 	struct ifmultiaddr *next;
3243 
3244 	IF_ADDR_WLOCK(ifp);
3245 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3246 		if_delmulti_locked(ifp, ifma, 0);
3247 	IF_ADDR_WUNLOCK(ifp);
3248 }
3249 
3250 /*
3251  * Delete a multicast group membership by group membership pointer.
3252  * Network-layer protocol domains must use this routine.
3253  *
3254  * It is safe to call this routine if the ifp disappeared.
3255  */
3256 void
3257 if_delmulti_ifma(struct ifmultiaddr *ifma)
3258 {
3259 	struct ifnet *ifp;
3260 	int lastref;
3261 
3262 	ifp = ifma->ifma_ifp;
3263 #ifdef DIAGNOSTIC
3264 	if (ifp == NULL) {
3265 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3266 	} else {
3267 		struct ifnet *oifp;
3268 
3269 		IFNET_RLOCK_NOSLEEP();
3270 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3271 			if (ifp == oifp)
3272 				break;
3273 		if (ifp != oifp) {
3274 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3275 			ifp = NULL;
3276 		}
3277 		IFNET_RUNLOCK_NOSLEEP();
3278 	}
3279 #endif
3280 	/*
3281 	 * If and only if the ifnet instance exists: Acquire the address lock.
3282 	 */
3283 	if (ifp != NULL)
3284 		IF_ADDR_WLOCK(ifp);
3285 
3286 	lastref = if_delmulti_locked(ifp, ifma, 0);
3287 
3288 	if (ifp != NULL) {
3289 		/*
3290 		 * If and only if the ifnet instance exists:
3291 		 *  Release the address lock.
3292 		 *  If the group was left: update the hardware hash filter.
3293 		 */
3294 		IF_ADDR_WUNLOCK(ifp);
3295 		if (lastref && ifp->if_ioctl != NULL) {
3296 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3297 		}
3298 	}
3299 }
3300 
3301 /*
3302  * Perform deletion of network-layer and/or link-layer multicast address.
3303  *
3304  * Return 0 if the reference count was decremented.
3305  * Return 1 if the final reference was released, indicating that the
3306  * hardware hash filter should be reprogrammed.
3307  */
3308 static int
3309 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3310 {
3311 	struct ifmultiaddr *ll_ifma;
3312 
3313 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3314 		KASSERT(ifma->ifma_ifp == ifp,
3315 		    ("%s: inconsistent ifp %p", __func__, ifp));
3316 		IF_ADDR_WLOCK_ASSERT(ifp);
3317 	}
3318 
3319 	ifp = ifma->ifma_ifp;
3320 
3321 	/*
3322 	 * If the ifnet is detaching, null out references to ifnet,
3323 	 * so that upper protocol layers will notice, and not attempt
3324 	 * to obtain locks for an ifnet which no longer exists. The
3325 	 * routing socket announcement must happen before the ifnet
3326 	 * instance is detached from the system.
3327 	 */
3328 	if (detaching) {
3329 #ifdef DIAGNOSTIC
3330 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3331 #endif
3332 		/*
3333 		 * ifp may already be nulled out if we are being reentered
3334 		 * to delete the ll_ifma.
3335 		 */
3336 		if (ifp != NULL) {
3337 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3338 			ifma->ifma_ifp = NULL;
3339 		}
3340 	}
3341 
3342 	if (--ifma->ifma_refcount > 0)
3343 		return 0;
3344 
3345 	/*
3346 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3347 	 * have been associated with it. Release it first if so.
3348 	 */
3349 	ll_ifma = ifma->ifma_llifma;
3350 	if (ll_ifma != NULL) {
3351 		KASSERT(ifma->ifma_lladdr != NULL,
3352 		    ("%s: llifma w/o lladdr", __func__));
3353 		if (detaching)
3354 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3355 		if (--ll_ifma->ifma_refcount == 0) {
3356 			if (ifp != NULL) {
3357 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3358 				    ifma_link);
3359 			}
3360 			if_freemulti(ll_ifma);
3361 		}
3362 	}
3363 
3364 	if (ifp != NULL)
3365 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3366 
3367 	if_freemulti(ifma);
3368 
3369 	/*
3370 	 * The last reference to this instance of struct ifmultiaddr
3371 	 * was released; the hardware should be notified of this change.
3372 	 */
3373 	return 1;
3374 }
3375 
3376 /*
3377  * Set the link layer address on an interface.
3378  *
3379  * At this time we only support certain types of interfaces,
3380  * and we don't allow the length of the address to change.
3381  *
3382  * Set noinline to be dtrace-friendly
3383  */
3384 __noinline int
3385 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3386 {
3387 	struct sockaddr_dl *sdl;
3388 	struct ifaddr *ifa;
3389 	struct ifreq ifr;
3390 
3391 	IF_ADDR_RLOCK(ifp);
3392 	ifa = ifp->if_addr;
3393 	if (ifa == NULL) {
3394 		IF_ADDR_RUNLOCK(ifp);
3395 		return (EINVAL);
3396 	}
3397 	ifa_ref(ifa);
3398 	IF_ADDR_RUNLOCK(ifp);
3399 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3400 	if (sdl == NULL) {
3401 		ifa_free(ifa);
3402 		return (EINVAL);
3403 	}
3404 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3405 		ifa_free(ifa);
3406 		return (EINVAL);
3407 	}
3408 	switch (ifp->if_type) {
3409 	case IFT_ETHER:
3410 	case IFT_FDDI:
3411 	case IFT_XETHER:
3412 	case IFT_ISO88025:
3413 	case IFT_L2VLAN:
3414 	case IFT_BRIDGE:
3415 	case IFT_ARCNET:
3416 	case IFT_IEEE8023ADLAG:
3417 	case IFT_IEEE80211:
3418 		bcopy(lladdr, LLADDR(sdl), len);
3419 		ifa_free(ifa);
3420 		break;
3421 	default:
3422 		ifa_free(ifa);
3423 		return (ENODEV);
3424 	}
3425 
3426 	/*
3427 	 * If the interface is already up, we need
3428 	 * to re-init it in order to reprogram its
3429 	 * address filter.
3430 	 */
3431 	if ((ifp->if_flags & IFF_UP) != 0) {
3432 		if (ifp->if_ioctl) {
3433 			ifp->if_flags &= ~IFF_UP;
3434 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3435 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3436 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3437 			ifp->if_flags |= IFF_UP;
3438 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3439 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3440 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3441 		}
3442 	}
3443 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3444 	return (0);
3445 }
3446 
3447 /*
3448  * Compat function for handling basic encapsulation requests.
3449  * Not converted stacks (FDDI, IB, ..) supports traditional
3450  * output model: ARP (and other similar L2 protocols) are handled
3451  * inside output routine, arpresolve/nd6_resolve() returns MAC
3452  * address instead of full prepend.
3453  *
3454  * This function creates calculated header==MAC for IPv4/IPv6 and
3455  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3456  * address families.
3457  */
3458 static int
3459 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3460 {
3461 
3462 	if (req->rtype != IFENCAP_LL)
3463 		return (EOPNOTSUPP);
3464 
3465 	if (req->bufsize < req->lladdr_len)
3466 		return (ENOMEM);
3467 
3468 	switch (req->family) {
3469 	case AF_INET:
3470 	case AF_INET6:
3471 		break;
3472 	default:
3473 		return (EAFNOSUPPORT);
3474 	}
3475 
3476 	/* Copy lladdr to storage as is */
3477 	memmove(req->buf, req->lladdr, req->lladdr_len);
3478 	req->bufsize = req->lladdr_len;
3479 	req->lladdr_off = 0;
3480 
3481 	return (0);
3482 }
3483 
3484 /*
3485  * The name argument must be a pointer to storage which will last as
3486  * long as the interface does.  For physical devices, the result of
3487  * device_get_name(dev) is a good choice and for pseudo-devices a
3488  * static string works well.
3489  */
3490 void
3491 if_initname(struct ifnet *ifp, const char *name, int unit)
3492 {
3493 	ifp->if_dname = name;
3494 	ifp->if_dunit = unit;
3495 	if (unit != IF_DUNIT_NONE)
3496 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3497 	else
3498 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3499 }
3500 
3501 int
3502 if_printf(struct ifnet *ifp, const char * fmt, ...)
3503 {
3504 	va_list ap;
3505 	int retval;
3506 
3507 	retval = printf("%s: ", ifp->if_xname);
3508 	va_start(ap, fmt);
3509 	retval += vprintf(fmt, ap);
3510 	va_end(ap);
3511 	return (retval);
3512 }
3513 
3514 void
3515 if_start(struct ifnet *ifp)
3516 {
3517 
3518 	(*(ifp)->if_start)(ifp);
3519 }
3520 
3521 /*
3522  * Backwards compatibility interface for drivers
3523  * that have not implemented it
3524  */
3525 static int
3526 if_transmit(struct ifnet *ifp, struct mbuf *m)
3527 {
3528 	int error;
3529 
3530 	IFQ_HANDOFF(ifp, m, error);
3531 	return (error);
3532 }
3533 
3534 static void
3535 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3536 {
3537 
3538 	m_freem(m);
3539 }
3540 
3541 int
3542 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3543 {
3544 	int active = 0;
3545 
3546 	IF_LOCK(ifq);
3547 	if (_IF_QFULL(ifq)) {
3548 		IF_UNLOCK(ifq);
3549 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3550 		m_freem(m);
3551 		return (0);
3552 	}
3553 	if (ifp != NULL) {
3554 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3555 		if (m->m_flags & (M_BCAST|M_MCAST))
3556 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3557 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3558 	}
3559 	_IF_ENQUEUE(ifq, m);
3560 	IF_UNLOCK(ifq);
3561 	if (ifp != NULL && !active)
3562 		(*(ifp)->if_start)(ifp);
3563 	return (1);
3564 }
3565 
3566 void
3567 if_register_com_alloc(u_char type,
3568     if_com_alloc_t *a, if_com_free_t *f)
3569 {
3570 
3571 	KASSERT(if_com_alloc[type] == NULL,
3572 	    ("if_register_com_alloc: %d already registered", type));
3573 	KASSERT(if_com_free[type] == NULL,
3574 	    ("if_register_com_alloc: %d free already registered", type));
3575 
3576 	if_com_alloc[type] = a;
3577 	if_com_free[type] = f;
3578 }
3579 
3580 void
3581 if_deregister_com_alloc(u_char type)
3582 {
3583 
3584 	KASSERT(if_com_alloc[type] != NULL,
3585 	    ("if_deregister_com_alloc: %d not registered", type));
3586 	KASSERT(if_com_free[type] != NULL,
3587 	    ("if_deregister_com_alloc: %d free not registered", type));
3588 	if_com_alloc[type] = NULL;
3589 	if_com_free[type] = NULL;
3590 }
3591 
3592 /* API for driver access to network stack owned ifnet.*/
3593 uint64_t
3594 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3595 {
3596 	uint64_t oldbrate;
3597 
3598 	oldbrate = ifp->if_baudrate;
3599 	ifp->if_baudrate = baudrate;
3600 	return (oldbrate);
3601 }
3602 
3603 uint64_t
3604 if_getbaudrate(if_t ifp)
3605 {
3606 
3607 	return (((struct ifnet *)ifp)->if_baudrate);
3608 }
3609 
3610 int
3611 if_setcapabilities(if_t ifp, int capabilities)
3612 {
3613 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3614 	return (0);
3615 }
3616 
3617 int
3618 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3619 {
3620 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3621 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3622 
3623 	return (0);
3624 }
3625 
3626 int
3627 if_getcapabilities(if_t ifp)
3628 {
3629 	return ((struct ifnet *)ifp)->if_capabilities;
3630 }
3631 
3632 int
3633 if_setcapenable(if_t ifp, int capabilities)
3634 {
3635 	((struct ifnet *)ifp)->if_capenable = capabilities;
3636 	return (0);
3637 }
3638 
3639 int
3640 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3641 {
3642 	if(setcap)
3643 		((struct ifnet *)ifp)->if_capenable |= setcap;
3644 	if(clearcap)
3645 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3646 
3647 	return (0);
3648 }
3649 
3650 const char *
3651 if_getdname(if_t ifp)
3652 {
3653 	return ((struct ifnet *)ifp)->if_dname;
3654 }
3655 
3656 int
3657 if_togglecapenable(if_t ifp, int togglecap)
3658 {
3659 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3660 	return (0);
3661 }
3662 
3663 int
3664 if_getcapenable(if_t ifp)
3665 {
3666 	return ((struct ifnet *)ifp)->if_capenable;
3667 }
3668 
3669 /*
3670  * This is largely undesirable because it ties ifnet to a device, but does
3671  * provide flexiblity for an embedded product vendor. Should be used with
3672  * the understanding that it violates the interface boundaries, and should be
3673  * a last resort only.
3674  */
3675 int
3676 if_setdev(if_t ifp, void *dev)
3677 {
3678 	return (0);
3679 }
3680 
3681 int
3682 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3683 {
3684 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3685 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3686 
3687 	return (0);
3688 }
3689 
3690 int
3691 if_getdrvflags(if_t ifp)
3692 {
3693 	return ((struct ifnet *)ifp)->if_drv_flags;
3694 }
3695 
3696 int
3697 if_setdrvflags(if_t ifp, int flags)
3698 {
3699 	((struct ifnet *)ifp)->if_drv_flags = flags;
3700 	return (0);
3701 }
3702 
3703 
3704 int
3705 if_setflags(if_t ifp, int flags)
3706 {
3707 	((struct ifnet *)ifp)->if_flags = flags;
3708 	return (0);
3709 }
3710 
3711 int
3712 if_setflagbits(if_t ifp, int set, int clear)
3713 {
3714 	((struct ifnet *)ifp)->if_flags |= set;
3715 	((struct ifnet *)ifp)->if_flags &= ~clear;
3716 
3717 	return (0);
3718 }
3719 
3720 int
3721 if_getflags(if_t ifp)
3722 {
3723 	return ((struct ifnet *)ifp)->if_flags;
3724 }
3725 
3726 int
3727 if_clearhwassist(if_t ifp)
3728 {
3729 	((struct ifnet *)ifp)->if_hwassist = 0;
3730 	return (0);
3731 }
3732 
3733 int
3734 if_sethwassistbits(if_t ifp, int toset, int toclear)
3735 {
3736 	((struct ifnet *)ifp)->if_hwassist |= toset;
3737 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3738 
3739 	return (0);
3740 }
3741 
3742 int
3743 if_sethwassist(if_t ifp, int hwassist_bit)
3744 {
3745 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3746 	return (0);
3747 }
3748 
3749 int
3750 if_gethwassist(if_t ifp)
3751 {
3752 	return ((struct ifnet *)ifp)->if_hwassist;
3753 }
3754 
3755 int
3756 if_setmtu(if_t ifp, int mtu)
3757 {
3758 	((struct ifnet *)ifp)->if_mtu = mtu;
3759 	return (0);
3760 }
3761 
3762 int
3763 if_getmtu(if_t ifp)
3764 {
3765 	return ((struct ifnet *)ifp)->if_mtu;
3766 }
3767 
3768 int
3769 if_getmtu_family(if_t ifp, int family)
3770 {
3771 	struct domain *dp;
3772 
3773 	for (dp = domains; dp; dp = dp->dom_next) {
3774 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3775 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3776 	}
3777 
3778 	return (((struct ifnet *)ifp)->if_mtu);
3779 }
3780 
3781 int
3782 if_setsoftc(if_t ifp, void *softc)
3783 {
3784 	((struct ifnet *)ifp)->if_softc = softc;
3785 	return (0);
3786 }
3787 
3788 void *
3789 if_getsoftc(if_t ifp)
3790 {
3791 	return ((struct ifnet *)ifp)->if_softc;
3792 }
3793 
3794 void
3795 if_setrcvif(struct mbuf *m, if_t ifp)
3796 {
3797 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3798 }
3799 
3800 void
3801 if_setvtag(struct mbuf *m, uint16_t tag)
3802 {
3803 	m->m_pkthdr.ether_vtag = tag;
3804 }
3805 
3806 uint16_t
3807 if_getvtag(struct mbuf *m)
3808 {
3809 
3810 	return (m->m_pkthdr.ether_vtag);
3811 }
3812 
3813 int
3814 if_sendq_empty(if_t ifp)
3815 {
3816 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3817 }
3818 
3819 struct ifaddr *
3820 if_getifaddr(if_t ifp)
3821 {
3822 	return ((struct ifnet *)ifp)->if_addr;
3823 }
3824 
3825 int
3826 if_getamcount(if_t ifp)
3827 {
3828 	return ((struct ifnet *)ifp)->if_amcount;
3829 }
3830 
3831 
3832 int
3833 if_setsendqready(if_t ifp)
3834 {
3835 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3836 	return (0);
3837 }
3838 
3839 int
3840 if_setsendqlen(if_t ifp, int tx_desc_count)
3841 {
3842 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3843 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3844 
3845 	return (0);
3846 }
3847 
3848 int
3849 if_vlantrunkinuse(if_t ifp)
3850 {
3851 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3852 }
3853 
3854 int
3855 if_input(if_t ifp, struct mbuf* sendmp)
3856 {
3857 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3858 	return (0);
3859 
3860 }
3861 
3862 /* XXX */
3863 #ifndef ETH_ADDR_LEN
3864 #define ETH_ADDR_LEN 6
3865 #endif
3866 
3867 int
3868 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3869 {
3870 	struct ifmultiaddr *ifma;
3871 	uint8_t *lmta = (uint8_t *)mta;
3872 	int mcnt = 0;
3873 
3874 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3875 		if (ifma->ifma_addr->sa_family != AF_LINK)
3876 			continue;
3877 
3878 		if (mcnt == max)
3879 			break;
3880 
3881 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3882 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3883 		mcnt++;
3884 	}
3885 	*cnt = mcnt;
3886 
3887 	return (0);
3888 }
3889 
3890 int
3891 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3892 {
3893 	int error;
3894 
3895 	if_maddr_rlock(ifp);
3896 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3897 	if_maddr_runlock(ifp);
3898 	return (error);
3899 }
3900 
3901 int
3902 if_multiaddr_count(if_t ifp, int max)
3903 {
3904 	struct ifmultiaddr *ifma;
3905 	int count;
3906 
3907 	count = 0;
3908 	if_maddr_rlock(ifp);
3909 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3910 		if (ifma->ifma_addr->sa_family != AF_LINK)
3911 			continue;
3912 		count++;
3913 		if (count == max)
3914 			break;
3915 	}
3916 	if_maddr_runlock(ifp);
3917 	return (count);
3918 }
3919 
3920 int
3921 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
3922 {
3923 	struct ifmultiaddr *ifma;
3924 	int cnt = 0;
3925 
3926 	if_maddr_rlock(ifp);
3927 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
3928 		cnt += filter(arg, ifma, cnt);
3929 	if_maddr_runlock(ifp);
3930 	return (cnt);
3931 }
3932 
3933 struct mbuf *
3934 if_dequeue(if_t ifp)
3935 {
3936 	struct mbuf *m;
3937 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3938 
3939 	return (m);
3940 }
3941 
3942 int
3943 if_sendq_prepend(if_t ifp, struct mbuf *m)
3944 {
3945 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3946 	return (0);
3947 }
3948 
3949 int
3950 if_setifheaderlen(if_t ifp, int len)
3951 {
3952 	((struct ifnet *)ifp)->if_hdrlen = len;
3953 	return (0);
3954 }
3955 
3956 caddr_t
3957 if_getlladdr(if_t ifp)
3958 {
3959 	return (IF_LLADDR((struct ifnet *)ifp));
3960 }
3961 
3962 void *
3963 if_gethandle(u_char type)
3964 {
3965 	return (if_alloc(type));
3966 }
3967 
3968 void
3969 if_bpfmtap(if_t ifh, struct mbuf *m)
3970 {
3971 	struct ifnet *ifp = (struct ifnet *)ifh;
3972 
3973 	BPF_MTAP(ifp, m);
3974 }
3975 
3976 void
3977 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3978 {
3979 	struct ifnet *ifp = (struct ifnet *)ifh;
3980 
3981 	ETHER_BPF_MTAP(ifp, m);
3982 }
3983 
3984 void
3985 if_vlancap(if_t ifh)
3986 {
3987 	struct ifnet *ifp = (struct ifnet *)ifh;
3988 	VLAN_CAPABILITIES(ifp);
3989 }
3990 
3991 void
3992 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3993 {
3994 	((struct ifnet *)ifp)->if_init = init_fn;
3995 }
3996 
3997 void
3998 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3999 {
4000 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4001 }
4002 
4003 void
4004 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4005 {
4006 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4007 }
4008 
4009 void
4010 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4011 {
4012 	((struct ifnet *)ifp)->if_transmit = start_fn;
4013 }
4014 
4015 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4016 {
4017 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4018 
4019 }
4020 
4021 void
4022 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4023 {
4024 
4025 	ifp->if_get_counter = fn;
4026 }
4027 
4028 /* Revisit these - These are inline functions originally. */
4029 int
4030 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4031 {
4032 	return drbr_inuse(ifh, br);
4033 }
4034 
4035 struct mbuf*
4036 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4037 {
4038 	return drbr_dequeue(ifh, br);
4039 }
4040 
4041 int
4042 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4043 {
4044 	return drbr_needs_enqueue(ifh, br);
4045 }
4046 
4047 int
4048 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4049 {
4050 	return drbr_enqueue(ifh, br, m);
4051 
4052 }
4053