xref: /freebsd/sys/net/if.c (revision ad6fc754f3d0dc2047171b41cb4d6e13da3deb98)
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 /* Interface description */
116 static unsigned int ifdescr_maxlen = 1024;
117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
118 	&ifdescr_maxlen, 0,
119 	"administrative maximum length for interface description");
120 
121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
122 
123 /* global sx for non-critical path ifdescr */
124 static struct sx ifdescr_sx;
125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
126 
127 void	(*bridge_linkstate_p)(struct ifnet *ifp);
128 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
129 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
130 /* These are external hooks for CARP. */
131 void	(*carp_linkstate_p)(struct ifnet *ifp);
132 void	(*carp_demote_adj_p)(int, char *);
133 int	(*carp_master_p)(struct ifaddr *);
134 #if defined(INET) || defined(INET6)
135 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
136 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
137     const struct sockaddr *sa);
138 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
139 int	(*carp_attach_p)(struct ifaddr *, int);
140 void	(*carp_detach_p)(struct ifaddr *);
141 #endif
142 #ifdef INET
143 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
144 #endif
145 #ifdef INET6
146 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
147 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
148     const struct in6_addr *taddr);
149 #endif
150 
151 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
152 
153 /*
154  * XXX: Style; these should be sorted alphabetically, and unprototyped
155  * static functions should be prototyped. Currently they are sorted by
156  * declaration order.
157  */
158 static void	if_attachdomain(void *);
159 static void	if_attachdomain1(struct ifnet *);
160 static int	ifconf(u_long, caddr_t);
161 static void	if_freemulti(struct ifmultiaddr *);
162 static void	if_grow(void);
163 static void	if_input_default(struct ifnet *, struct mbuf *);
164 static void	if_route(struct ifnet *, int flag, int fam);
165 static int	if_setflag(struct ifnet *, int, int, int *, int);
166 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
167 static void	if_unroute(struct ifnet *, int flag, int fam);
168 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
169 static int	if_rtdel(struct radix_node *, void *);
170 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
171 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
172 static void	do_link_state_change(void *, int);
173 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
174 static int	if_getgroupmembers(struct ifgroupreq *);
175 static void	if_delgroups(struct ifnet *);
176 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
177 static void	if_detach_internal(struct ifnet *, int, struct if_clone **);
178 
179 #ifdef INET6
180 /*
181  * XXX: declare here to avoid to include many inet6 related files..
182  * should be more generalized?
183  */
184 extern void	nd6_setmtu(struct ifnet *);
185 #endif
186 
187 VNET_DEFINE(int, if_index);
188 int	ifqmaxlen = IFQ_MAXLEN;
189 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
190 VNET_DEFINE(struct ifgrouphead, ifg_head);
191 
192 static VNET_DEFINE(int, if_indexlim) = 8;
193 
194 /* Table of ifnet by index. */
195 VNET_DEFINE(struct ifnet **, ifindex_table);
196 
197 #define	V_if_indexlim		VNET(if_indexlim)
198 #define	V_ifindex_table		VNET(ifindex_table)
199 
200 /*
201  * The global network interface list (V_ifnet) and related state (such as
202  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
203  * an rwlock.  Either may be acquired shared to stablize the list, but both
204  * must be acquired writable to modify the list.  This model allows us to
205  * both stablize the interface list during interrupt thread processing, but
206  * also to stablize it over long-running ioctls, without introducing priority
207  * inversions and deadlocks.
208  */
209 struct rwlock ifnet_rwlock;
210 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
211 struct sx ifnet_sxlock;
212 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
213 
214 /*
215  * The allocation of network interfaces is a rather non-atomic affair; we
216  * need to select an index before we are ready to expose the interface for
217  * use, so will use this pointer value to indicate reservation.
218  */
219 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
220 
221 static	if_com_alloc_t *if_com_alloc[256];
222 static	if_com_free_t *if_com_free[256];
223 
224 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
225 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
226 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
227 
228 struct ifnet *
229 ifnet_byindex_locked(u_short idx)
230 {
231 
232 	if (idx > V_if_index)
233 		return (NULL);
234 	if (V_ifindex_table[idx] == IFNET_HOLD)
235 		return (NULL);
236 	return (V_ifindex_table[idx]);
237 }
238 
239 struct ifnet *
240 ifnet_byindex(u_short idx)
241 {
242 	struct ifnet *ifp;
243 
244 	IFNET_RLOCK_NOSLEEP();
245 	ifp = ifnet_byindex_locked(idx);
246 	IFNET_RUNLOCK_NOSLEEP();
247 	return (ifp);
248 }
249 
250 struct ifnet *
251 ifnet_byindex_ref(u_short idx)
252 {
253 	struct ifnet *ifp;
254 
255 	IFNET_RLOCK_NOSLEEP();
256 	ifp = ifnet_byindex_locked(idx);
257 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
258 		IFNET_RUNLOCK_NOSLEEP();
259 		return (NULL);
260 	}
261 	if_ref(ifp);
262 	IFNET_RUNLOCK_NOSLEEP();
263 	return (ifp);
264 }
265 
266 /*
267  * Allocate an ifindex array entry; return 0 on success or an error on
268  * failure.
269  */
270 static u_short
271 ifindex_alloc(void)
272 {
273 	u_short idx;
274 
275 	IFNET_WLOCK_ASSERT();
276 retry:
277 	/*
278 	 * Try to find an empty slot below V_if_index.  If we fail, take the
279 	 * next slot.
280 	 */
281 	for (idx = 1; idx <= V_if_index; idx++) {
282 		if (V_ifindex_table[idx] == NULL)
283 			break;
284 	}
285 
286 	/* Catch if_index overflow. */
287 	if (idx >= V_if_indexlim) {
288 		if_grow();
289 		goto retry;
290 	}
291 	if (idx > V_if_index)
292 		V_if_index = idx;
293 	return (idx);
294 }
295 
296 static void
297 ifindex_free_locked(u_short idx)
298 {
299 
300 	IFNET_WLOCK_ASSERT();
301 
302 	V_ifindex_table[idx] = NULL;
303 	while (V_if_index > 0 &&
304 	    V_ifindex_table[V_if_index] == NULL)
305 		V_if_index--;
306 }
307 
308 static void
309 ifindex_free(u_short idx)
310 {
311 
312 	IFNET_WLOCK();
313 	ifindex_free_locked(idx);
314 	IFNET_WUNLOCK();
315 }
316 
317 static void
318 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
319 {
320 
321 	IFNET_WLOCK_ASSERT();
322 
323 	V_ifindex_table[idx] = ifp;
324 }
325 
326 static void
327 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
328 {
329 
330 	IFNET_WLOCK();
331 	ifnet_setbyindex_locked(idx, ifp);
332 	IFNET_WUNLOCK();
333 }
334 
335 struct ifaddr *
336 ifaddr_byindex(u_short idx)
337 {
338 	struct ifaddr *ifa;
339 
340 	IFNET_RLOCK_NOSLEEP();
341 	ifa = ifnet_byindex_locked(idx)->if_addr;
342 	if (ifa != NULL)
343 		ifa_ref(ifa);
344 	IFNET_RUNLOCK_NOSLEEP();
345 	return (ifa);
346 }
347 
348 /*
349  * Network interface utility routines.
350  *
351  * Routines with ifa_ifwith* names take sockaddr *'s as
352  * parameters.
353  */
354 
355 static void
356 vnet_if_init(const void *unused __unused)
357 {
358 
359 	TAILQ_INIT(&V_ifnet);
360 	TAILQ_INIT(&V_ifg_head);
361 	IFNET_WLOCK();
362 	if_grow();				/* create initial table */
363 	IFNET_WUNLOCK();
364 	vnet_if_clone_init();
365 }
366 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
367     NULL);
368 
369 #ifdef VIMAGE
370 static void
371 vnet_if_uninit(const void *unused __unused)
372 {
373 
374 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
375 	    "not empty", __func__, __LINE__, &V_ifnet));
376 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
377 	    "not empty", __func__, __LINE__, &V_ifg_head));
378 
379 	free((caddr_t)V_ifindex_table, M_IFNET);
380 }
381 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
382     vnet_if_uninit, NULL);
383 #endif
384 
385 static void
386 if_grow(void)
387 {
388 	int oldlim;
389 	u_int n;
390 	struct ifnet **e;
391 
392 	IFNET_WLOCK_ASSERT();
393 	oldlim = V_if_indexlim;
394 	IFNET_WUNLOCK();
395 	n = (oldlim << 1) * sizeof(*e);
396 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
397 	IFNET_WLOCK();
398 	if (V_if_indexlim != oldlim) {
399 		free(e, M_IFNET);
400 		return;
401 	}
402 	if (V_ifindex_table != NULL) {
403 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
404 		free((caddr_t)V_ifindex_table, M_IFNET);
405 	}
406 	V_if_indexlim <<= 1;
407 	V_ifindex_table = e;
408 }
409 
410 /*
411  * Allocate a struct ifnet and an index for an interface.  A layer 2
412  * common structure will also be allocated if an allocation routine is
413  * registered for the passed type.
414  */
415 struct ifnet *
416 if_alloc(u_char type)
417 {
418 	struct ifnet *ifp;
419 	u_short idx;
420 
421 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
422 	IFNET_WLOCK();
423 	idx = ifindex_alloc();
424 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
425 	IFNET_WUNLOCK();
426 	ifp->if_index = idx;
427 	ifp->if_type = type;
428 	ifp->if_alloctype = type;
429 	if (if_com_alloc[type] != NULL) {
430 		ifp->if_l2com = if_com_alloc[type](type, ifp);
431 		if (ifp->if_l2com == NULL) {
432 			free(ifp, M_IFNET);
433 			ifindex_free(idx);
434 			return (NULL);
435 		}
436 	}
437 
438 	IF_ADDR_LOCK_INIT(ifp);
439 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
440 	ifp->if_afdata_initialized = 0;
441 	IF_AFDATA_LOCK_INIT(ifp);
442 	TAILQ_INIT(&ifp->if_addrhead);
443 	TAILQ_INIT(&ifp->if_multiaddrs);
444 	TAILQ_INIT(&ifp->if_groups);
445 #ifdef MAC
446 	mac_ifnet_init(ifp);
447 #endif
448 	ifq_init(&ifp->if_snd, ifp);
449 
450 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
451 	for (int i = 0; i < IFCOUNTERS; i++)
452 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
453 	ifp->if_get_counter = if_get_counter_default;
454 	ifnet_setbyindex(ifp->if_index, ifp);
455 	return (ifp);
456 }
457 
458 /*
459  * Do the actual work of freeing a struct ifnet, and layer 2 common
460  * structure.  This call is made when the last reference to an
461  * interface is released.
462  */
463 static void
464 if_free_internal(struct ifnet *ifp)
465 {
466 
467 	KASSERT((ifp->if_flags & IFF_DYING),
468 	    ("if_free_internal: interface not dying"));
469 
470 	if (if_com_free[ifp->if_alloctype] != NULL)
471 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
472 		    ifp->if_alloctype);
473 
474 #ifdef MAC
475 	mac_ifnet_destroy(ifp);
476 #endif /* MAC */
477 	if (ifp->if_description != NULL)
478 		free(ifp->if_description, M_IFDESCR);
479 	IF_AFDATA_DESTROY(ifp);
480 	IF_ADDR_LOCK_DESTROY(ifp);
481 	ifq_delete(&ifp->if_snd);
482 
483 	for (int i = 0; i < IFCOUNTERS; i++)
484 		counter_u64_free(ifp->if_counters[i]);
485 
486 	free(ifp, M_IFNET);
487 }
488 
489 /*
490  * Deregister an interface and free the associated storage.
491  */
492 void
493 if_free(struct ifnet *ifp)
494 {
495 
496 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
497 
498 	CURVNET_SET_QUIET(ifp->if_vnet);
499 	IFNET_WLOCK();
500 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
501 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
502 
503 	ifindex_free_locked(ifp->if_index);
504 	IFNET_WUNLOCK();
505 
506 	if (refcount_release(&ifp->if_refcount))
507 		if_free_internal(ifp);
508 	CURVNET_RESTORE();
509 }
510 
511 /*
512  * Interfaces to keep an ifnet type-stable despite the possibility of the
513  * driver calling if_free().  If there are additional references, we defer
514  * freeing the underlying data structure.
515  */
516 void
517 if_ref(struct ifnet *ifp)
518 {
519 
520 	/* We don't assert the ifnet list lock here, but arguably should. */
521 	refcount_acquire(&ifp->if_refcount);
522 }
523 
524 void
525 if_rele(struct ifnet *ifp)
526 {
527 
528 	if (!refcount_release(&ifp->if_refcount))
529 		return;
530 	if_free_internal(ifp);
531 }
532 
533 void
534 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
535 {
536 
537 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
538 
539 	if (ifq->ifq_maxlen == 0)
540 		ifq->ifq_maxlen = ifqmaxlen;
541 
542 	ifq->altq_type = 0;
543 	ifq->altq_disc = NULL;
544 	ifq->altq_flags &= ALTQF_CANTCHANGE;
545 	ifq->altq_tbr  = NULL;
546 	ifq->altq_ifp  = ifp;
547 }
548 
549 void
550 ifq_delete(struct ifaltq *ifq)
551 {
552 	mtx_destroy(&ifq->ifq_mtx);
553 }
554 
555 /*
556  * Perform generic interface initalization tasks and attach the interface
557  * to the list of "active" interfaces.  If vmove flag is set on entry
558  * to if_attach_internal(), perform only a limited subset of initialization
559  * tasks, given that we are moving from one vnet to another an ifnet which
560  * has already been fully initialized.
561  *
562  * Note that if_detach_internal() removes group membership unconditionally
563  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
564  * Thus, when if_vmove() is applied to a cloned interface, group membership
565  * is lost while a cloned one always joins a group whose name is
566  * ifc->ifc_name.  To recover this after if_detach_internal() and
567  * if_attach_internal(), the cloner should be specified to
568  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
569  * attempts to join a group whose name is ifc->ifc_name.
570  *
571  * XXX:
572  *  - The decision to return void and thus require this function to
573  *    succeed is questionable.
574  *  - We should probably do more sanity checking.  For instance we don't
575  *    do anything to insure if_xname is unique or non-empty.
576  */
577 void
578 if_attach(struct ifnet *ifp)
579 {
580 
581 	if_attach_internal(ifp, 0, NULL);
582 }
583 
584 /*
585  * Compute the least common TSO limit.
586  */
587 void
588 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
589 {
590 	/*
591 	 * 1) If there is no limit currently, take the limit from
592 	 * the network adapter.
593 	 *
594 	 * 2) If the network adapter has a limit below the current
595 	 * limit, apply it.
596 	 */
597 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
598 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
599 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
600 	}
601 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
602 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
603 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
604 	}
605 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
606 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
607 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
608 	}
609 }
610 
611 /*
612  * Update TSO limit of a network adapter.
613  *
614  * Returns zero if no change. Else non-zero.
615  */
616 int
617 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
618 {
619 	int retval = 0;
620 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
621 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
622 		retval++;
623 	}
624 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
625 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
626 		retval++;
627 	}
628 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
629 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
630 		retval++;
631 	}
632 	return (retval);
633 }
634 
635 static void
636 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
637 {
638 	unsigned socksize, ifasize;
639 	int namelen, masklen;
640 	struct sockaddr_dl *sdl;
641 	struct ifaddr *ifa;
642 
643 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
644 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
645 		    ifp->if_xname);
646 
647 #ifdef VIMAGE
648 	ifp->if_vnet = curvnet;
649 	if (ifp->if_home_vnet == NULL)
650 		ifp->if_home_vnet = curvnet;
651 #endif
652 
653 	if_addgroup(ifp, IFG_ALL);
654 
655 	/* Restore group membership for cloned interfaces. */
656 	if (vmove && ifc != NULL)
657 		if_clone_addgroup(ifp, ifc);
658 
659 	getmicrotime(&ifp->if_lastchange);
660 	ifp->if_epoch = time_uptime;
661 
662 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
663 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
664 	    ("transmit and qflush must both either be set or both be NULL"));
665 	if (ifp->if_transmit == NULL) {
666 		ifp->if_transmit = if_transmit;
667 		ifp->if_qflush = if_qflush;
668 	}
669 	if (ifp->if_input == NULL)
670 		ifp->if_input = if_input_default;
671 
672 	if (!vmove) {
673 #ifdef MAC
674 		mac_ifnet_create(ifp);
675 #endif
676 
677 		/*
678 		 * Create a Link Level name for this device.
679 		 */
680 		namelen = strlen(ifp->if_xname);
681 		/*
682 		 * Always save enough space for any possiable name so we
683 		 * can do a rename in place later.
684 		 */
685 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
686 		socksize = masklen + ifp->if_addrlen;
687 		if (socksize < sizeof(*sdl))
688 			socksize = sizeof(*sdl);
689 		socksize = roundup2(socksize, sizeof(long));
690 		ifasize = sizeof(*ifa) + 2 * socksize;
691 		ifa = ifa_alloc(ifasize, M_WAITOK);
692 		sdl = (struct sockaddr_dl *)(ifa + 1);
693 		sdl->sdl_len = socksize;
694 		sdl->sdl_family = AF_LINK;
695 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
696 		sdl->sdl_nlen = namelen;
697 		sdl->sdl_index = ifp->if_index;
698 		sdl->sdl_type = ifp->if_type;
699 		ifp->if_addr = ifa;
700 		ifa->ifa_ifp = ifp;
701 		ifa->ifa_rtrequest = link_rtrequest;
702 		ifa->ifa_addr = (struct sockaddr *)sdl;
703 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
704 		ifa->ifa_netmask = (struct sockaddr *)sdl;
705 		sdl->sdl_len = masklen;
706 		while (namelen != 0)
707 			sdl->sdl_data[--namelen] = 0xff;
708 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
709 		/* Reliably crash if used uninitialized. */
710 		ifp->if_broadcastaddr = NULL;
711 
712 #if defined(INET) || defined(INET6)
713 		/* Use defaults for TSO, if nothing is set */
714 		if (ifp->if_hw_tsomax == 0 &&
715 		    ifp->if_hw_tsomaxsegcount == 0 &&
716 		    ifp->if_hw_tsomaxsegsize == 0) {
717 			/*
718 			 * The TSO defaults needs to be such that an
719 			 * NFS mbuf list of 35 mbufs totalling just
720 			 * below 64K works and that a chain of mbufs
721 			 * can be defragged into at most 32 segments:
722 			 */
723 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
724 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
725 			ifp->if_hw_tsomaxsegcount = 35;
726 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
727 
728 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
729 			if (ifp->if_capabilities & IFCAP_TSO) {
730 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
731 				    ifp->if_hw_tsomax,
732 				    ifp->if_hw_tsomaxsegcount,
733 				    ifp->if_hw_tsomaxsegsize);
734 			}
735 		}
736 #endif
737 	}
738 #ifdef VIMAGE
739 	else {
740 		/*
741 		 * Update the interface index in the link layer address
742 		 * of the interface.
743 		 */
744 		for (ifa = ifp->if_addr; ifa != NULL;
745 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
746 			if (ifa->ifa_addr->sa_family == AF_LINK) {
747 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
748 				sdl->sdl_index = ifp->if_index;
749 			}
750 		}
751 	}
752 #endif
753 
754 	IFNET_WLOCK();
755 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
756 #ifdef VIMAGE
757 	curvnet->vnet_ifcnt++;
758 #endif
759 	IFNET_WUNLOCK();
760 
761 	if (domain_init_status >= 2)
762 		if_attachdomain1(ifp);
763 
764 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
765 	if (IS_DEFAULT_VNET(curvnet))
766 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
767 
768 	/* Announce the interface. */
769 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
770 }
771 
772 static void
773 if_attachdomain(void *dummy)
774 {
775 	struct ifnet *ifp;
776 
777 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
778 		if_attachdomain1(ifp);
779 }
780 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
781     if_attachdomain, NULL);
782 
783 static void
784 if_attachdomain1(struct ifnet *ifp)
785 {
786 	struct domain *dp;
787 
788 	/*
789 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
790 	 * cannot lock ifp->if_afdata initialization, entirely.
791 	 */
792 	if (IF_AFDATA_TRYLOCK(ifp) == 0)
793 		return;
794 	if (ifp->if_afdata_initialized >= domain_init_status) {
795 		IF_AFDATA_UNLOCK(ifp);
796 		log(LOG_WARNING, "%s called more than once on %s\n",
797 		    __func__, ifp->if_xname);
798 		return;
799 	}
800 	ifp->if_afdata_initialized = domain_init_status;
801 	IF_AFDATA_UNLOCK(ifp);
802 
803 	/* address family dependent data region */
804 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
805 	for (dp = domains; dp; dp = dp->dom_next) {
806 		if (dp->dom_ifattach)
807 			ifp->if_afdata[dp->dom_family] =
808 			    (*dp->dom_ifattach)(ifp);
809 	}
810 }
811 
812 /*
813  * Remove any unicast or broadcast network addresses from an interface.
814  */
815 void
816 if_purgeaddrs(struct ifnet *ifp)
817 {
818 	struct ifaddr *ifa, *next;
819 
820 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
821 		if (ifa->ifa_addr->sa_family == AF_LINK)
822 			continue;
823 #ifdef INET
824 		/* XXX: Ugly!! ad hoc just for INET */
825 		if (ifa->ifa_addr->sa_family == AF_INET) {
826 			struct ifaliasreq ifr;
827 
828 			bzero(&ifr, sizeof(ifr));
829 			ifr.ifra_addr = *ifa->ifa_addr;
830 			if (ifa->ifa_dstaddr)
831 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
832 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
833 			    NULL) == 0)
834 				continue;
835 		}
836 #endif /* INET */
837 #ifdef INET6
838 		if (ifa->ifa_addr->sa_family == AF_INET6) {
839 			in6_purgeaddr(ifa);
840 			/* ifp_addrhead is already updated */
841 			continue;
842 		}
843 #endif /* INET6 */
844 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
845 		ifa_free(ifa);
846 	}
847 }
848 
849 /*
850  * Remove any multicast network addresses from an interface when an ifnet
851  * is going away.
852  */
853 static void
854 if_purgemaddrs(struct ifnet *ifp)
855 {
856 	struct ifmultiaddr *ifma;
857 	struct ifmultiaddr *next;
858 
859 	IF_ADDR_WLOCK(ifp);
860 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
861 		if_delmulti_locked(ifp, ifma, 1);
862 	IF_ADDR_WUNLOCK(ifp);
863 }
864 
865 /*
866  * Detach an interface, removing it from the list of "active" interfaces.
867  * If vmove flag is set on entry to if_detach_internal(), perform only a
868  * limited subset of cleanup tasks, given that we are moving an ifnet from
869  * one vnet to another, where it must be fully operational.
870  *
871  * XXXRW: There are some significant questions about event ordering, and
872  * how to prevent things from starting to use the interface during detach.
873  */
874 void
875 if_detach(struct ifnet *ifp)
876 {
877 
878 	CURVNET_SET_QUIET(ifp->if_vnet);
879 	if_detach_internal(ifp, 0, NULL);
880 	CURVNET_RESTORE();
881 }
882 
883 static void
884 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
885 {
886 	struct ifaddr *ifa;
887 	struct radix_node_head	*rnh;
888 	int i, j;
889 	struct domain *dp;
890  	struct ifnet *iter;
891  	int found = 0;
892 
893 	IFNET_WLOCK();
894 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
895 		if (iter == ifp) {
896 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
897 			found = 1;
898 			break;
899 		}
900 #ifdef VIMAGE
901 	if (found)
902 		curvnet->vnet_ifcnt--;
903 #endif
904 	IFNET_WUNLOCK();
905 	if (!found) {
906 		if (vmove)
907 			panic("%s: ifp=%p not on the ifnet tailq %p",
908 			    __func__, ifp, &V_ifnet);
909 		else
910 			return; /* XXX this should panic as well? */
911 	}
912 
913 	/* Check if this is a cloned interface or not. */
914 	if (vmove && ifcp != NULL)
915 		*ifcp = if_clone_findifc(ifp);
916 
917 	/*
918 	 * Remove/wait for pending events.
919 	 */
920 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
921 
922 	/*
923 	 * Remove routes and flush queues.
924 	 */
925 	if_down(ifp);
926 #ifdef ALTQ
927 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
928 		altq_disable(&ifp->if_snd);
929 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
930 		altq_detach(&ifp->if_snd);
931 #endif
932 
933 	if_purgeaddrs(ifp);
934 
935 #ifdef INET
936 	in_ifdetach(ifp);
937 #endif
938 
939 #ifdef INET6
940 	/*
941 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
942 	 * before removing routing entries below, since IPv6 interface direct
943 	 * routes are expected to be removed by the IPv6-specific kernel API.
944 	 * Otherwise, the kernel will detect some inconsistency and bark it.
945 	 */
946 	in6_ifdetach(ifp);
947 #endif
948 	if_purgemaddrs(ifp);
949 
950 	/* Announce that the interface is gone. */
951 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
952 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
953 	if (IS_DEFAULT_VNET(curvnet))
954 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
955 
956 	if (!vmove) {
957 		/*
958 		 * Prevent further calls into the device driver via ifnet.
959 		 */
960 		if_dead(ifp);
961 
962 		/*
963 		 * Remove link ifaddr pointer and maybe decrement if_index.
964 		 * Clean up all addresses.
965 		 */
966 		ifp->if_addr = NULL;
967 
968 		/* We can now free link ifaddr. */
969 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
970 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
971 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
972 			ifa_free(ifa);
973 		}
974 	}
975 
976 	/*
977 	 * Delete all remaining routes using this interface
978 	 * Unfortuneatly the only way to do this is to slog through
979 	 * the entire routing table looking for routes which point
980 	 * to this interface...oh well...
981 	 */
982 	for (i = 1; i <= AF_MAX; i++) {
983 		for (j = 0; j < rt_numfibs; j++) {
984 			rnh = rt_tables_get_rnh(j, i);
985 			if (rnh == NULL)
986 				continue;
987 			RADIX_NODE_HEAD_LOCK(rnh);
988 			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
989 			RADIX_NODE_HEAD_UNLOCK(rnh);
990 		}
991 	}
992 
993 	if_delgroups(ifp);
994 
995 	/*
996 	 * We cannot hold the lock over dom_ifdetach calls as they might
997 	 * sleep, for example trying to drain a callout, thus open up the
998 	 * theoretical race with re-attaching.
999 	 */
1000 	IF_AFDATA_LOCK(ifp);
1001 	i = ifp->if_afdata_initialized;
1002 	ifp->if_afdata_initialized = 0;
1003 	IF_AFDATA_UNLOCK(ifp);
1004 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1005 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1006 			(*dp->dom_ifdetach)(ifp,
1007 			    ifp->if_afdata[dp->dom_family]);
1008 	}
1009 }
1010 
1011 #ifdef VIMAGE
1012 /*
1013  * if_vmove() performs a limited version of if_detach() in current
1014  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1015  * An attempt is made to shrink if_index in current vnet, find an
1016  * unused if_index in target vnet and calls if_grow() if necessary,
1017  * and finally find an unused if_xname for the target vnet.
1018  */
1019 void
1020 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1021 {
1022 	struct if_clone *ifc;
1023 
1024 	/*
1025 	 * Detach from current vnet, but preserve LLADDR info, do not
1026 	 * mark as dead etc. so that the ifnet can be reattached later.
1027 	 */
1028 	if_detach_internal(ifp, 1, &ifc);
1029 
1030 	/*
1031 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1032 	 * the if_index for that vnet if possible.
1033 	 *
1034 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1035 	 * or we'd lock on one vnet and unlock on another.
1036 	 */
1037 	IFNET_WLOCK();
1038 	ifindex_free_locked(ifp->if_index);
1039 	IFNET_WUNLOCK();
1040 
1041 	/*
1042 	 * Perform interface-specific reassignment tasks, if provided by
1043 	 * the driver.
1044 	 */
1045 	if (ifp->if_reassign != NULL)
1046 		ifp->if_reassign(ifp, new_vnet, NULL);
1047 
1048 	/*
1049 	 * Switch to the context of the target vnet.
1050 	 */
1051 	CURVNET_SET_QUIET(new_vnet);
1052 
1053 	IFNET_WLOCK();
1054 	ifp->if_index = ifindex_alloc();
1055 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1056 	IFNET_WUNLOCK();
1057 
1058 	if_attach_internal(ifp, 1, ifc);
1059 
1060 	CURVNET_RESTORE();
1061 }
1062 
1063 /*
1064  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1065  */
1066 static int
1067 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1068 {
1069 	struct prison *pr;
1070 	struct ifnet *difp;
1071 
1072 	/* Try to find the prison within our visibility. */
1073 	sx_slock(&allprison_lock);
1074 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1075 	sx_sunlock(&allprison_lock);
1076 	if (pr == NULL)
1077 		return (ENXIO);
1078 	prison_hold_locked(pr);
1079 	mtx_unlock(&pr->pr_mtx);
1080 
1081 	/* Do not try to move the iface from and to the same prison. */
1082 	if (pr->pr_vnet == ifp->if_vnet) {
1083 		prison_free(pr);
1084 		return (EEXIST);
1085 	}
1086 
1087 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1088 	/* XXX Lock interfaces to avoid races. */
1089 	CURVNET_SET_QUIET(pr->pr_vnet);
1090 	difp = ifunit(ifname);
1091 	CURVNET_RESTORE();
1092 	if (difp != NULL) {
1093 		prison_free(pr);
1094 		return (EEXIST);
1095 	}
1096 
1097 	/* Move the interface into the child jail/vnet. */
1098 	if_vmove(ifp, pr->pr_vnet);
1099 
1100 	/* Report the new if_xname back to the userland. */
1101 	sprintf(ifname, "%s", ifp->if_xname);
1102 
1103 	prison_free(pr);
1104 	return (0);
1105 }
1106 
1107 static int
1108 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1109 {
1110 	struct prison *pr;
1111 	struct vnet *vnet_dst;
1112 	struct ifnet *ifp;
1113 
1114 	/* Try to find the prison within our visibility. */
1115 	sx_slock(&allprison_lock);
1116 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1117 	sx_sunlock(&allprison_lock);
1118 	if (pr == NULL)
1119 		return (ENXIO);
1120 	prison_hold_locked(pr);
1121 	mtx_unlock(&pr->pr_mtx);
1122 
1123 	/* Make sure the named iface exists in the source prison/vnet. */
1124 	CURVNET_SET(pr->pr_vnet);
1125 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1126 	if (ifp == NULL) {
1127 		CURVNET_RESTORE();
1128 		prison_free(pr);
1129 		return (ENXIO);
1130 	}
1131 
1132 	/* Do not try to move the iface from and to the same prison. */
1133 	vnet_dst = TD_TO_VNET(td);
1134 	if (vnet_dst == ifp->if_vnet) {
1135 		CURVNET_RESTORE();
1136 		prison_free(pr);
1137 		return (EEXIST);
1138 	}
1139 
1140 	/* Get interface back from child jail/vnet. */
1141 	if_vmove(ifp, vnet_dst);
1142 	CURVNET_RESTORE();
1143 
1144 	/* Report the new if_xname back to the userland. */
1145 	sprintf(ifname, "%s", ifp->if_xname);
1146 
1147 	prison_free(pr);
1148 	return (0);
1149 }
1150 #endif /* VIMAGE */
1151 
1152 /*
1153  * Add a group to an interface
1154  */
1155 int
1156 if_addgroup(struct ifnet *ifp, const char *groupname)
1157 {
1158 	struct ifg_list		*ifgl;
1159 	struct ifg_group	*ifg = NULL;
1160 	struct ifg_member	*ifgm;
1161 	int 			 new = 0;
1162 
1163 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1164 	    groupname[strlen(groupname) - 1] <= '9')
1165 		return (EINVAL);
1166 
1167 	IFNET_WLOCK();
1168 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1169 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1170 			IFNET_WUNLOCK();
1171 			return (EEXIST);
1172 		}
1173 
1174 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1175 	    M_NOWAIT)) == NULL) {
1176 	    	IFNET_WUNLOCK();
1177 		return (ENOMEM);
1178 	}
1179 
1180 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1181 	    M_TEMP, M_NOWAIT)) == NULL) {
1182 		free(ifgl, M_TEMP);
1183 		IFNET_WUNLOCK();
1184 		return (ENOMEM);
1185 	}
1186 
1187 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1188 		if (!strcmp(ifg->ifg_group, groupname))
1189 			break;
1190 
1191 	if (ifg == NULL) {
1192 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1193 		    M_TEMP, M_NOWAIT)) == NULL) {
1194 			free(ifgl, M_TEMP);
1195 			free(ifgm, M_TEMP);
1196 			IFNET_WUNLOCK();
1197 			return (ENOMEM);
1198 		}
1199 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1200 		ifg->ifg_refcnt = 0;
1201 		TAILQ_INIT(&ifg->ifg_members);
1202 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1203 		new = 1;
1204 	}
1205 
1206 	ifg->ifg_refcnt++;
1207 	ifgl->ifgl_group = ifg;
1208 	ifgm->ifgm_ifp = ifp;
1209 
1210 	IF_ADDR_WLOCK(ifp);
1211 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1212 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1213 	IF_ADDR_WUNLOCK(ifp);
1214 
1215 	IFNET_WUNLOCK();
1216 
1217 	if (new)
1218 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1219 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1220 
1221 	return (0);
1222 }
1223 
1224 /*
1225  * Remove a group from an interface
1226  */
1227 int
1228 if_delgroup(struct ifnet *ifp, const char *groupname)
1229 {
1230 	struct ifg_list		*ifgl;
1231 	struct ifg_member	*ifgm;
1232 
1233 	IFNET_WLOCK();
1234 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1235 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1236 			break;
1237 	if (ifgl == NULL) {
1238 		IFNET_WUNLOCK();
1239 		return (ENOENT);
1240 	}
1241 
1242 	IF_ADDR_WLOCK(ifp);
1243 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1244 	IF_ADDR_WUNLOCK(ifp);
1245 
1246 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1247 		if (ifgm->ifgm_ifp == ifp)
1248 			break;
1249 
1250 	if (ifgm != NULL) {
1251 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1252 		free(ifgm, M_TEMP);
1253 	}
1254 
1255 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1256 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1257 		IFNET_WUNLOCK();
1258 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1259 		free(ifgl->ifgl_group, M_TEMP);
1260 	} else
1261 		IFNET_WUNLOCK();
1262 
1263 	free(ifgl, M_TEMP);
1264 
1265 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1266 
1267 	return (0);
1268 }
1269 
1270 /*
1271  * Remove an interface from all groups
1272  */
1273 static void
1274 if_delgroups(struct ifnet *ifp)
1275 {
1276 	struct ifg_list		*ifgl;
1277 	struct ifg_member	*ifgm;
1278 	char groupname[IFNAMSIZ];
1279 
1280 	IFNET_WLOCK();
1281 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1282 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1283 
1284 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1285 
1286 		IF_ADDR_WLOCK(ifp);
1287 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1288 		IF_ADDR_WUNLOCK(ifp);
1289 
1290 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1291 			if (ifgm->ifgm_ifp == ifp)
1292 				break;
1293 
1294 		if (ifgm != NULL) {
1295 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1296 			    ifgm_next);
1297 			free(ifgm, M_TEMP);
1298 		}
1299 
1300 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1301 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1302 			IFNET_WUNLOCK();
1303 			EVENTHANDLER_INVOKE(group_detach_event,
1304 			    ifgl->ifgl_group);
1305 			free(ifgl->ifgl_group, M_TEMP);
1306 		} else
1307 			IFNET_WUNLOCK();
1308 
1309 		free(ifgl, M_TEMP);
1310 
1311 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1312 
1313 		IFNET_WLOCK();
1314 	}
1315 	IFNET_WUNLOCK();
1316 }
1317 
1318 /*
1319  * Stores all groups from an interface in memory pointed
1320  * to by data
1321  */
1322 static int
1323 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1324 {
1325 	int			 len, error;
1326 	struct ifg_list		*ifgl;
1327 	struct ifg_req		 ifgrq, *ifgp;
1328 	struct ifgroupreq	*ifgr = data;
1329 
1330 	if (ifgr->ifgr_len == 0) {
1331 		IF_ADDR_RLOCK(ifp);
1332 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1333 			ifgr->ifgr_len += sizeof(struct ifg_req);
1334 		IF_ADDR_RUNLOCK(ifp);
1335 		return (0);
1336 	}
1337 
1338 	len = ifgr->ifgr_len;
1339 	ifgp = ifgr->ifgr_groups;
1340 	/* XXX: wire */
1341 	IF_ADDR_RLOCK(ifp);
1342 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1343 		if (len < sizeof(ifgrq)) {
1344 			IF_ADDR_RUNLOCK(ifp);
1345 			return (EINVAL);
1346 		}
1347 		bzero(&ifgrq, sizeof ifgrq);
1348 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1349 		    sizeof(ifgrq.ifgrq_group));
1350 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1351 		    	IF_ADDR_RUNLOCK(ifp);
1352 			return (error);
1353 		}
1354 		len -= sizeof(ifgrq);
1355 		ifgp++;
1356 	}
1357 	IF_ADDR_RUNLOCK(ifp);
1358 
1359 	return (0);
1360 }
1361 
1362 /*
1363  * Stores all members of a group in memory pointed to by data
1364  */
1365 static int
1366 if_getgroupmembers(struct ifgroupreq *data)
1367 {
1368 	struct ifgroupreq	*ifgr = data;
1369 	struct ifg_group	*ifg;
1370 	struct ifg_member	*ifgm;
1371 	struct ifg_req		 ifgrq, *ifgp;
1372 	int			 len, error;
1373 
1374 	IFNET_RLOCK();
1375 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1376 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1377 			break;
1378 	if (ifg == NULL) {
1379 		IFNET_RUNLOCK();
1380 		return (ENOENT);
1381 	}
1382 
1383 	if (ifgr->ifgr_len == 0) {
1384 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1385 			ifgr->ifgr_len += sizeof(ifgrq);
1386 		IFNET_RUNLOCK();
1387 		return (0);
1388 	}
1389 
1390 	len = ifgr->ifgr_len;
1391 	ifgp = ifgr->ifgr_groups;
1392 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1393 		if (len < sizeof(ifgrq)) {
1394 			IFNET_RUNLOCK();
1395 			return (EINVAL);
1396 		}
1397 		bzero(&ifgrq, sizeof ifgrq);
1398 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1399 		    sizeof(ifgrq.ifgrq_member));
1400 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1401 			IFNET_RUNLOCK();
1402 			return (error);
1403 		}
1404 		len -= sizeof(ifgrq);
1405 		ifgp++;
1406 	}
1407 	IFNET_RUNLOCK();
1408 
1409 	return (0);
1410 }
1411 
1412 /*
1413  * Delete Routes for a Network Interface
1414  *
1415  * Called for each routing entry via the rnh->rnh_walktree() call above
1416  * to delete all route entries referencing a detaching network interface.
1417  *
1418  * Arguments:
1419  *	rn	pointer to node in the routing table
1420  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1421  *
1422  * Returns:
1423  *	0	successful
1424  *	errno	failed - reason indicated
1425  *
1426  */
1427 static int
1428 if_rtdel(struct radix_node *rn, void *arg)
1429 {
1430 	struct rtentry	*rt = (struct rtentry *)rn;
1431 	struct ifnet	*ifp = arg;
1432 	int		err;
1433 
1434 	if (rt->rt_ifp == ifp) {
1435 
1436 		/*
1437 		 * Protect (sorta) against walktree recursion problems
1438 		 * with cloned routes
1439 		 */
1440 		if ((rt->rt_flags & RTF_UP) == 0)
1441 			return (0);
1442 
1443 		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1444 				rt_mask(rt),
1445 				rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1446 				(struct rtentry **) NULL, rt->rt_fibnum);
1447 		if (err) {
1448 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1449 		}
1450 	}
1451 
1452 	return (0);
1453 }
1454 
1455 /*
1456  * Return counter values from counter(9)s stored in ifnet.
1457  */
1458 uint64_t
1459 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1460 {
1461 
1462 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1463 
1464 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1465 }
1466 
1467 /*
1468  * Increase an ifnet counter. Usually used for counters shared
1469  * between the stack and a driver, but function supports them all.
1470  */
1471 void
1472 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1473 {
1474 
1475 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1476 
1477 	counter_u64_add(ifp->if_counters[cnt], inc);
1478 }
1479 
1480 /*
1481  * Copy data from ifnet to userland API structure if_data.
1482  */
1483 void
1484 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1485 {
1486 
1487 	ifd->ifi_type = ifp->if_type;
1488 	ifd->ifi_physical = 0;
1489 	ifd->ifi_addrlen = ifp->if_addrlen;
1490 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1491 	ifd->ifi_link_state = ifp->if_link_state;
1492 	ifd->ifi_vhid = 0;
1493 	ifd->ifi_datalen = sizeof(struct if_data);
1494 	ifd->ifi_mtu = ifp->if_mtu;
1495 	ifd->ifi_metric = ifp->if_metric;
1496 	ifd->ifi_baudrate = ifp->if_baudrate;
1497 	ifd->ifi_hwassist = ifp->if_hwassist;
1498 	ifd->ifi_epoch = ifp->if_epoch;
1499 	ifd->ifi_lastchange = ifp->if_lastchange;
1500 
1501 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1502 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1503 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1504 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1505 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1506 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1507 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1508 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1509 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1510 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1511 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1512 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1513 }
1514 
1515 /*
1516  * Wrapper functions for struct ifnet address list locking macros.  These are
1517  * used by kernel modules to avoid encoding programming interface or binary
1518  * interface assumptions that may be violated when kernel-internal locking
1519  * approaches change.
1520  */
1521 void
1522 if_addr_rlock(struct ifnet *ifp)
1523 {
1524 
1525 	IF_ADDR_RLOCK(ifp);
1526 }
1527 
1528 void
1529 if_addr_runlock(struct ifnet *ifp)
1530 {
1531 
1532 	IF_ADDR_RUNLOCK(ifp);
1533 }
1534 
1535 void
1536 if_maddr_rlock(if_t ifp)
1537 {
1538 
1539 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1540 }
1541 
1542 void
1543 if_maddr_runlock(if_t ifp)
1544 {
1545 
1546 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1547 }
1548 
1549 /*
1550  * Initialization, destruction and refcounting functions for ifaddrs.
1551  */
1552 struct ifaddr *
1553 ifa_alloc(size_t size, int flags)
1554 {
1555 	struct ifaddr *ifa;
1556 
1557 	KASSERT(size >= sizeof(struct ifaddr),
1558 	    ("%s: invalid size %zu", __func__, size));
1559 
1560 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1561 	if (ifa == NULL)
1562 		return (NULL);
1563 
1564 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1565 		goto fail;
1566 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1567 		goto fail;
1568 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1569 		goto fail;
1570 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1571 		goto fail;
1572 
1573 	refcount_init(&ifa->ifa_refcnt, 1);
1574 
1575 	return (ifa);
1576 
1577 fail:
1578 	/* free(NULL) is okay */
1579 	counter_u64_free(ifa->ifa_opackets);
1580 	counter_u64_free(ifa->ifa_ipackets);
1581 	counter_u64_free(ifa->ifa_obytes);
1582 	counter_u64_free(ifa->ifa_ibytes);
1583 	free(ifa, M_IFADDR);
1584 
1585 	return (NULL);
1586 }
1587 
1588 void
1589 ifa_ref(struct ifaddr *ifa)
1590 {
1591 
1592 	refcount_acquire(&ifa->ifa_refcnt);
1593 }
1594 
1595 void
1596 ifa_free(struct ifaddr *ifa)
1597 {
1598 
1599 	if (refcount_release(&ifa->ifa_refcnt)) {
1600 		counter_u64_free(ifa->ifa_opackets);
1601 		counter_u64_free(ifa->ifa_ipackets);
1602 		counter_u64_free(ifa->ifa_obytes);
1603 		counter_u64_free(ifa->ifa_ibytes);
1604 		free(ifa, M_IFADDR);
1605 	}
1606 }
1607 
1608 int
1609 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1610 {
1611 	int error = 0;
1612 	struct rtentry *rt = NULL;
1613 	struct rt_addrinfo info;
1614 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1615 
1616 	bzero(&info, sizeof(info));
1617 	info.rti_ifp = V_loif;
1618 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1619 	info.rti_info[RTAX_DST] = ia;
1620 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1621 	error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1622 
1623 	if (error == 0 && rt != NULL) {
1624 		RT_LOCK(rt);
1625 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1626 			ifa->ifa_ifp->if_type;
1627 		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1628 			ifa->ifa_ifp->if_index;
1629 		RT_REMREF(rt);
1630 		RT_UNLOCK(rt);
1631 	} else if (error != 0)
1632 		log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1633 
1634 	return (error);
1635 }
1636 
1637 int
1638 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1639 {
1640 	int error = 0;
1641 	struct rt_addrinfo info;
1642 	struct sockaddr_dl null_sdl;
1643 
1644 	bzero(&null_sdl, sizeof(null_sdl));
1645 	null_sdl.sdl_len = sizeof(null_sdl);
1646 	null_sdl.sdl_family = AF_LINK;
1647 	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1648 	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1649 	bzero(&info, sizeof(info));
1650 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1651 	info.rti_info[RTAX_DST] = ia;
1652 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1653 	error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1654 
1655 	if (error != 0)
1656 		log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1657 
1658 	return (error);
1659 }
1660 
1661 int
1662 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib)
1663 {
1664 	struct rtentry *rt;
1665 
1666 	rt = rtalloc1_fib(sa, 0, 0, fib);
1667 	if (rt == NULL) {
1668 		log(LOG_DEBUG, "%s: fail", __func__);
1669 		return (EHOSTUNREACH);
1670 	}
1671 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1672 	    ifa->ifa_ifp->if_type;
1673 	((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1674 	    ifa->ifa_ifp->if_index;
1675 	RTFREE_LOCKED(rt);
1676 
1677 	return (0);
1678 }
1679 
1680 /*
1681  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1682  * structs used to represent other address families, it is necessary
1683  * to perform a different comparison.
1684  */
1685 
1686 #define	sa_dl_equal(a1, a2)	\
1687 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1688 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1689 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1690 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1691 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1692 
1693 /*
1694  * Locate an interface based on a complete address.
1695  */
1696 /*ARGSUSED*/
1697 static struct ifaddr *
1698 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1699 {
1700 	struct ifnet *ifp;
1701 	struct ifaddr *ifa;
1702 
1703 	IFNET_RLOCK_NOSLEEP();
1704 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1705 		IF_ADDR_RLOCK(ifp);
1706 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1707 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1708 				continue;
1709 			if (sa_equal(addr, ifa->ifa_addr)) {
1710 				if (getref)
1711 					ifa_ref(ifa);
1712 				IF_ADDR_RUNLOCK(ifp);
1713 				goto done;
1714 			}
1715 			/* IP6 doesn't have broadcast */
1716 			if ((ifp->if_flags & IFF_BROADCAST) &&
1717 			    ifa->ifa_broadaddr &&
1718 			    ifa->ifa_broadaddr->sa_len != 0 &&
1719 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1720 				if (getref)
1721 					ifa_ref(ifa);
1722 				IF_ADDR_RUNLOCK(ifp);
1723 				goto done;
1724 			}
1725 		}
1726 		IF_ADDR_RUNLOCK(ifp);
1727 	}
1728 	ifa = NULL;
1729 done:
1730 	IFNET_RUNLOCK_NOSLEEP();
1731 	return (ifa);
1732 }
1733 
1734 struct ifaddr *
1735 ifa_ifwithaddr(struct sockaddr *addr)
1736 {
1737 
1738 	return (ifa_ifwithaddr_internal(addr, 1));
1739 }
1740 
1741 int
1742 ifa_ifwithaddr_check(struct sockaddr *addr)
1743 {
1744 
1745 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1746 }
1747 
1748 /*
1749  * Locate an interface based on the broadcast address.
1750  */
1751 /* ARGSUSED */
1752 struct ifaddr *
1753 ifa_ifwithbroadaddr(struct sockaddr *addr, int fibnum)
1754 {
1755 	struct ifnet *ifp;
1756 	struct ifaddr *ifa;
1757 
1758 	IFNET_RLOCK_NOSLEEP();
1759 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1760 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1761 			continue;
1762 		IF_ADDR_RLOCK(ifp);
1763 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1764 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1765 				continue;
1766 			if ((ifp->if_flags & IFF_BROADCAST) &&
1767 			    ifa->ifa_broadaddr &&
1768 			    ifa->ifa_broadaddr->sa_len != 0 &&
1769 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1770 				ifa_ref(ifa);
1771 				IF_ADDR_RUNLOCK(ifp);
1772 				goto done;
1773 			}
1774 		}
1775 		IF_ADDR_RUNLOCK(ifp);
1776 	}
1777 	ifa = NULL;
1778 done:
1779 	IFNET_RUNLOCK_NOSLEEP();
1780 	return (ifa);
1781 }
1782 
1783 /*
1784  * Locate the point to point interface with a given destination address.
1785  */
1786 /*ARGSUSED*/
1787 struct ifaddr *
1788 ifa_ifwithdstaddr(struct sockaddr *addr, int fibnum)
1789 {
1790 	struct ifnet *ifp;
1791 	struct ifaddr *ifa;
1792 
1793 	IFNET_RLOCK_NOSLEEP();
1794 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1795 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1796 			continue;
1797 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1798 			continue;
1799 		IF_ADDR_RLOCK(ifp);
1800 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1801 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1802 				continue;
1803 			if (ifa->ifa_dstaddr != NULL &&
1804 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1805 				ifa_ref(ifa);
1806 				IF_ADDR_RUNLOCK(ifp);
1807 				goto done;
1808 			}
1809 		}
1810 		IF_ADDR_RUNLOCK(ifp);
1811 	}
1812 	ifa = NULL;
1813 done:
1814 	IFNET_RUNLOCK_NOSLEEP();
1815 	return (ifa);
1816 }
1817 
1818 /*
1819  * Find an interface on a specific network.  If many, choice
1820  * is most specific found.
1821  */
1822 struct ifaddr *
1823 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp, int fibnum)
1824 {
1825 	struct ifnet *ifp;
1826 	struct ifaddr *ifa;
1827 	struct ifaddr *ifa_maybe = NULL;
1828 	u_int af = addr->sa_family;
1829 	char *addr_data = addr->sa_data, *cplim;
1830 
1831 	/*
1832 	 * AF_LINK addresses can be looked up directly by their index number,
1833 	 * so do that if we can.
1834 	 */
1835 	if (af == AF_LINK) {
1836 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1837 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1838 		return (ifaddr_byindex(sdl->sdl_index));
1839 	}
1840 
1841 	/*
1842 	 * Scan though each interface, looking for ones that have addresses
1843 	 * in this address family and the requested fib.  Maintain a reference
1844 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1845 	 * kept it stable when we move onto the next interface.
1846 	 */
1847 	IFNET_RLOCK_NOSLEEP();
1848 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1849 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1850 			continue;
1851 		IF_ADDR_RLOCK(ifp);
1852 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1853 			char *cp, *cp2, *cp3;
1854 
1855 			if (ifa->ifa_addr->sa_family != af)
1856 next:				continue;
1857 			if (af == AF_INET &&
1858 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1859 				/*
1860 				 * This is a bit broken as it doesn't
1861 				 * take into account that the remote end may
1862 				 * be a single node in the network we are
1863 				 * looking for.
1864 				 * The trouble is that we don't know the
1865 				 * netmask for the remote end.
1866 				 */
1867 				if (ifa->ifa_dstaddr != NULL &&
1868 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1869 					ifa_ref(ifa);
1870 					IF_ADDR_RUNLOCK(ifp);
1871 					goto done;
1872 				}
1873 			} else {
1874 				/*
1875 				 * Scan all the bits in the ifa's address.
1876 				 * If a bit dissagrees with what we are
1877 				 * looking for, mask it with the netmask
1878 				 * to see if it really matters.
1879 				 * (A byte at a time)
1880 				 */
1881 				if (ifa->ifa_netmask == 0)
1882 					continue;
1883 				cp = addr_data;
1884 				cp2 = ifa->ifa_addr->sa_data;
1885 				cp3 = ifa->ifa_netmask->sa_data;
1886 				cplim = ifa->ifa_netmask->sa_len
1887 					+ (char *)ifa->ifa_netmask;
1888 				while (cp3 < cplim)
1889 					if ((*cp++ ^ *cp2++) & *cp3++)
1890 						goto next; /* next address! */
1891 				/*
1892 				 * If the netmask of what we just found
1893 				 * is more specific than what we had before
1894 				 * (if we had one), or if the virtual status
1895 				 * of new prefix is better than of the old one,
1896 				 * then remember the new one before continuing
1897 				 * to search for an even better one.
1898 				 */
1899 				if (ifa_maybe == NULL ||
1900 				    ifa_preferred(ifa_maybe, ifa) ||
1901 				    rn_refines((caddr_t)ifa->ifa_netmask,
1902 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1903 					if (ifa_maybe != NULL)
1904 						ifa_free(ifa_maybe);
1905 					ifa_maybe = ifa;
1906 					ifa_ref(ifa_maybe);
1907 				}
1908 			}
1909 		}
1910 		IF_ADDR_RUNLOCK(ifp);
1911 	}
1912 	ifa = ifa_maybe;
1913 	ifa_maybe = NULL;
1914 done:
1915 	IFNET_RUNLOCK_NOSLEEP();
1916 	if (ifa_maybe != NULL)
1917 		ifa_free(ifa_maybe);
1918 	return (ifa);
1919 }
1920 
1921 /*
1922  * Find an interface address specific to an interface best matching
1923  * a given address.
1924  */
1925 struct ifaddr *
1926 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1927 {
1928 	struct ifaddr *ifa;
1929 	char *cp, *cp2, *cp3;
1930 	char *cplim;
1931 	struct ifaddr *ifa_maybe = NULL;
1932 	u_int af = addr->sa_family;
1933 
1934 	if (af >= AF_MAX)
1935 		return (NULL);
1936 	IF_ADDR_RLOCK(ifp);
1937 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1938 		if (ifa->ifa_addr->sa_family != af)
1939 			continue;
1940 		if (ifa_maybe == NULL)
1941 			ifa_maybe = ifa;
1942 		if (ifa->ifa_netmask == 0) {
1943 			if (sa_equal(addr, ifa->ifa_addr) ||
1944 			    (ifa->ifa_dstaddr &&
1945 			    sa_equal(addr, ifa->ifa_dstaddr)))
1946 				goto done;
1947 			continue;
1948 		}
1949 		if (ifp->if_flags & IFF_POINTOPOINT) {
1950 			if (sa_equal(addr, ifa->ifa_dstaddr))
1951 				goto done;
1952 		} else {
1953 			cp = addr->sa_data;
1954 			cp2 = ifa->ifa_addr->sa_data;
1955 			cp3 = ifa->ifa_netmask->sa_data;
1956 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1957 			for (; cp3 < cplim; cp3++)
1958 				if ((*cp++ ^ *cp2++) & *cp3)
1959 					break;
1960 			if (cp3 == cplim)
1961 				goto done;
1962 		}
1963 	}
1964 	ifa = ifa_maybe;
1965 done:
1966 	if (ifa != NULL)
1967 		ifa_ref(ifa);
1968 	IF_ADDR_RUNLOCK(ifp);
1969 	return (ifa);
1970 }
1971 
1972 /*
1973  * See whether new ifa is better than current one:
1974  * 1) A non-virtual one is preferred over virtual.
1975  * 2) A virtual in master state preferred over any other state.
1976  *
1977  * Used in several address selecting functions.
1978  */
1979 int
1980 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1981 {
1982 
1983 	return (cur->ifa_carp && (!next->ifa_carp ||
1984 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1985 }
1986 
1987 #include <net/if_llatbl.h>
1988 
1989 /*
1990  * Default action when installing a route with a Link Level gateway.
1991  * Lookup an appropriate real ifa to point to.
1992  * This should be moved to /sys/net/link.c eventually.
1993  */
1994 static void
1995 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1996 {
1997 	struct ifaddr *ifa, *oifa;
1998 	struct sockaddr *dst;
1999 	struct ifnet *ifp;
2000 
2001 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
2002 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
2003 		return;
2004 	ifa = ifaof_ifpforaddr(dst, ifp);
2005 	if (ifa) {
2006 		oifa = rt->rt_ifa;
2007 		rt->rt_ifa = ifa;
2008 		ifa_free(oifa);
2009 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2010 			ifa->ifa_rtrequest(cmd, rt, info);
2011 	}
2012 }
2013 
2014 struct sockaddr_dl *
2015 link_alloc_sdl(size_t size, int flags)
2016 {
2017 
2018 	return (malloc(size, M_TEMP, flags));
2019 }
2020 
2021 void
2022 link_free_sdl(struct sockaddr *sa)
2023 {
2024 	free(sa, M_TEMP);
2025 }
2026 
2027 /*
2028  * Fills in given sdl with interface basic info.
2029  * Returns pointer to filled sdl.
2030  */
2031 struct sockaddr_dl *
2032 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2033 {
2034 	struct sockaddr_dl *sdl;
2035 
2036 	sdl = (struct sockaddr_dl *)paddr;
2037 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2038 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2039 	sdl->sdl_family = AF_LINK;
2040 	sdl->sdl_index = ifp->if_index;
2041 	sdl->sdl_type = iftype;
2042 
2043 	return (sdl);
2044 }
2045 
2046 /*
2047  * Mark an interface down and notify protocols of
2048  * the transition.
2049  */
2050 static void
2051 if_unroute(struct ifnet *ifp, int flag, int fam)
2052 {
2053 	struct ifaddr *ifa;
2054 
2055 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2056 
2057 	ifp->if_flags &= ~flag;
2058 	getmicrotime(&ifp->if_lastchange);
2059 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2060 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2061 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2062 	ifp->if_qflush(ifp);
2063 
2064 	if (ifp->if_carp)
2065 		(*carp_linkstate_p)(ifp);
2066 	rt_ifmsg(ifp);
2067 }
2068 
2069 /*
2070  * Mark an interface up and notify protocols of
2071  * the transition.
2072  */
2073 static void
2074 if_route(struct ifnet *ifp, int flag, int fam)
2075 {
2076 	struct ifaddr *ifa;
2077 
2078 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2079 
2080 	ifp->if_flags |= flag;
2081 	getmicrotime(&ifp->if_lastchange);
2082 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2083 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2084 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2085 	if (ifp->if_carp)
2086 		(*carp_linkstate_p)(ifp);
2087 	rt_ifmsg(ifp);
2088 #ifdef INET6
2089 	in6_if_up(ifp);
2090 #endif
2091 }
2092 
2093 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2094 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2095 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2096 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2097 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2098 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2099 void	*(*vlan_cookie_p)(struct ifnet *);
2100 
2101 /*
2102  * Handle a change in the interface link state. To avoid LORs
2103  * between driver lock and upper layer locks, as well as possible
2104  * recursions, we post event to taskqueue, and all job
2105  * is done in static do_link_state_change().
2106  */
2107 void
2108 if_link_state_change(struct ifnet *ifp, int link_state)
2109 {
2110 	/* Return if state hasn't changed. */
2111 	if (ifp->if_link_state == link_state)
2112 		return;
2113 
2114 	ifp->if_link_state = link_state;
2115 
2116 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2117 }
2118 
2119 static void
2120 do_link_state_change(void *arg, int pending)
2121 {
2122 	struct ifnet *ifp = (struct ifnet *)arg;
2123 	int link_state = ifp->if_link_state;
2124 	CURVNET_SET(ifp->if_vnet);
2125 
2126 	/* Notify that the link state has changed. */
2127 	rt_ifmsg(ifp);
2128 	if (ifp->if_vlantrunk != NULL)
2129 		(*vlan_link_state_p)(ifp);
2130 
2131 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2132 	    ifp->if_l2com != NULL)
2133 		(*ng_ether_link_state_p)(ifp, link_state);
2134 	if (ifp->if_carp)
2135 		(*carp_linkstate_p)(ifp);
2136 	if (ifp->if_bridge)
2137 		(*bridge_linkstate_p)(ifp);
2138 	if (ifp->if_lagg)
2139 		(*lagg_linkstate_p)(ifp, link_state);
2140 
2141 	if (IS_DEFAULT_VNET(curvnet))
2142 		devctl_notify("IFNET", ifp->if_xname,
2143 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2144 		    NULL);
2145 	if (pending > 1)
2146 		if_printf(ifp, "%d link states coalesced\n", pending);
2147 	if (log_link_state_change)
2148 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2149 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2150 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2151 	CURVNET_RESTORE();
2152 }
2153 
2154 /*
2155  * Mark an interface down and notify protocols of
2156  * the transition.
2157  */
2158 void
2159 if_down(struct ifnet *ifp)
2160 {
2161 
2162 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2163 }
2164 
2165 /*
2166  * Mark an interface up and notify protocols of
2167  * the transition.
2168  */
2169 void
2170 if_up(struct ifnet *ifp)
2171 {
2172 
2173 	if_route(ifp, IFF_UP, AF_UNSPEC);
2174 }
2175 
2176 /*
2177  * Flush an interface queue.
2178  */
2179 void
2180 if_qflush(struct ifnet *ifp)
2181 {
2182 	struct mbuf *m, *n;
2183 	struct ifaltq *ifq;
2184 
2185 	ifq = &ifp->if_snd;
2186 	IFQ_LOCK(ifq);
2187 #ifdef ALTQ
2188 	if (ALTQ_IS_ENABLED(ifq))
2189 		ALTQ_PURGE(ifq);
2190 #endif
2191 	n = ifq->ifq_head;
2192 	while ((m = n) != 0) {
2193 		n = m->m_nextpkt;
2194 		m_freem(m);
2195 	}
2196 	ifq->ifq_head = 0;
2197 	ifq->ifq_tail = 0;
2198 	ifq->ifq_len = 0;
2199 	IFQ_UNLOCK(ifq);
2200 }
2201 
2202 /*
2203  * Map interface name to interface structure pointer, with or without
2204  * returning a reference.
2205  */
2206 struct ifnet *
2207 ifunit_ref(const char *name)
2208 {
2209 	struct ifnet *ifp;
2210 
2211 	IFNET_RLOCK_NOSLEEP();
2212 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2213 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2214 		    !(ifp->if_flags & IFF_DYING))
2215 			break;
2216 	}
2217 	if (ifp != NULL)
2218 		if_ref(ifp);
2219 	IFNET_RUNLOCK_NOSLEEP();
2220 	return (ifp);
2221 }
2222 
2223 struct ifnet *
2224 ifunit(const char *name)
2225 {
2226 	struct ifnet *ifp;
2227 
2228 	IFNET_RLOCK_NOSLEEP();
2229 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2230 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2231 			break;
2232 	}
2233 	IFNET_RUNLOCK_NOSLEEP();
2234 	return (ifp);
2235 }
2236 
2237 /*
2238  * Hardware specific interface ioctls.
2239  */
2240 static int
2241 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2242 {
2243 	struct ifreq *ifr;
2244 	int error = 0;
2245 	int new_flags, temp_flags;
2246 	size_t namelen, onamelen;
2247 	size_t descrlen;
2248 	char *descrbuf, *odescrbuf;
2249 	char new_name[IFNAMSIZ];
2250 	struct ifaddr *ifa;
2251 	struct sockaddr_dl *sdl;
2252 
2253 	ifr = (struct ifreq *)data;
2254 	switch (cmd) {
2255 	case SIOCGIFINDEX:
2256 		ifr->ifr_index = ifp->if_index;
2257 		break;
2258 
2259 	case SIOCGIFFLAGS:
2260 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2261 		ifr->ifr_flags = temp_flags & 0xffff;
2262 		ifr->ifr_flagshigh = temp_flags >> 16;
2263 		break;
2264 
2265 	case SIOCGIFCAP:
2266 		ifr->ifr_reqcap = ifp->if_capabilities;
2267 		ifr->ifr_curcap = ifp->if_capenable;
2268 		break;
2269 
2270 #ifdef MAC
2271 	case SIOCGIFMAC:
2272 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2273 		break;
2274 #endif
2275 
2276 	case SIOCGIFMETRIC:
2277 		ifr->ifr_metric = ifp->if_metric;
2278 		break;
2279 
2280 	case SIOCGIFMTU:
2281 		ifr->ifr_mtu = ifp->if_mtu;
2282 		break;
2283 
2284 	case SIOCGIFPHYS:
2285 		/* XXXGL: did this ever worked? */
2286 		ifr->ifr_phys = 0;
2287 		break;
2288 
2289 	case SIOCGIFDESCR:
2290 		error = 0;
2291 		sx_slock(&ifdescr_sx);
2292 		if (ifp->if_description == NULL)
2293 			error = ENOMSG;
2294 		else {
2295 			/* space for terminating nul */
2296 			descrlen = strlen(ifp->if_description) + 1;
2297 			if (ifr->ifr_buffer.length < descrlen)
2298 				ifr->ifr_buffer.buffer = NULL;
2299 			else
2300 				error = copyout(ifp->if_description,
2301 				    ifr->ifr_buffer.buffer, descrlen);
2302 			ifr->ifr_buffer.length = descrlen;
2303 		}
2304 		sx_sunlock(&ifdescr_sx);
2305 		break;
2306 
2307 	case SIOCSIFDESCR:
2308 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2309 		if (error)
2310 			return (error);
2311 
2312 		/*
2313 		 * Copy only (length-1) bytes to make sure that
2314 		 * if_description is always nul terminated.  The
2315 		 * length parameter is supposed to count the
2316 		 * terminating nul in.
2317 		 */
2318 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2319 			return (ENAMETOOLONG);
2320 		else if (ifr->ifr_buffer.length == 0)
2321 			descrbuf = NULL;
2322 		else {
2323 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2324 			    M_WAITOK | M_ZERO);
2325 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2326 			    ifr->ifr_buffer.length - 1);
2327 			if (error) {
2328 				free(descrbuf, M_IFDESCR);
2329 				break;
2330 			}
2331 		}
2332 
2333 		sx_xlock(&ifdescr_sx);
2334 		odescrbuf = ifp->if_description;
2335 		ifp->if_description = descrbuf;
2336 		sx_xunlock(&ifdescr_sx);
2337 
2338 		getmicrotime(&ifp->if_lastchange);
2339 		free(odescrbuf, M_IFDESCR);
2340 		break;
2341 
2342 	case SIOCGIFFIB:
2343 		ifr->ifr_fib = ifp->if_fib;
2344 		break;
2345 
2346 	case SIOCSIFFIB:
2347 		error = priv_check(td, PRIV_NET_SETIFFIB);
2348 		if (error)
2349 			return (error);
2350 		if (ifr->ifr_fib >= rt_numfibs)
2351 			return (EINVAL);
2352 
2353 		ifp->if_fib = ifr->ifr_fib;
2354 		break;
2355 
2356 	case SIOCSIFFLAGS:
2357 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2358 		if (error)
2359 			return (error);
2360 		/*
2361 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2362 		 * check, so we don't need special handling here yet.
2363 		 */
2364 		new_flags = (ifr->ifr_flags & 0xffff) |
2365 		    (ifr->ifr_flagshigh << 16);
2366 		if (ifp->if_flags & IFF_UP &&
2367 		    (new_flags & IFF_UP) == 0) {
2368 			if_down(ifp);
2369 		} else if (new_flags & IFF_UP &&
2370 		    (ifp->if_flags & IFF_UP) == 0) {
2371 			if_up(ifp);
2372 		}
2373 		/* See if permanently promiscuous mode bit is about to flip */
2374 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2375 			if (new_flags & IFF_PPROMISC)
2376 				ifp->if_flags |= IFF_PROMISC;
2377 			else if (ifp->if_pcount == 0)
2378 				ifp->if_flags &= ~IFF_PROMISC;
2379 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2380 			    ifp->if_xname,
2381 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2382 		}
2383 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2384 			(new_flags &~ IFF_CANTCHANGE);
2385 		if (ifp->if_ioctl) {
2386 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2387 		}
2388 		getmicrotime(&ifp->if_lastchange);
2389 		break;
2390 
2391 	case SIOCSIFCAP:
2392 		error = priv_check(td, PRIV_NET_SETIFCAP);
2393 		if (error)
2394 			return (error);
2395 		if (ifp->if_ioctl == NULL)
2396 			return (EOPNOTSUPP);
2397 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2398 			return (EINVAL);
2399 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2400 		if (error == 0)
2401 			getmicrotime(&ifp->if_lastchange);
2402 		break;
2403 
2404 #ifdef MAC
2405 	case SIOCSIFMAC:
2406 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2407 		break;
2408 #endif
2409 
2410 	case SIOCSIFNAME:
2411 		error = priv_check(td, PRIV_NET_SETIFNAME);
2412 		if (error)
2413 			return (error);
2414 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2415 		if (error != 0)
2416 			return (error);
2417 		if (new_name[0] == '\0')
2418 			return (EINVAL);
2419 		if (ifunit(new_name) != NULL)
2420 			return (EEXIST);
2421 
2422 		/*
2423 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2424 		 * and there are numerous other races with the
2425 		 * ifunit() checks not being atomic with namespace
2426 		 * changes (renames, vmoves, if_attach, etc).
2427 		 */
2428 		ifp->if_flags |= IFF_RENAMING;
2429 
2430 		/* Announce the departure of the interface. */
2431 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2432 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2433 
2434 		log(LOG_INFO, "%s: changing name to '%s'\n",
2435 		    ifp->if_xname, new_name);
2436 
2437 		IF_ADDR_WLOCK(ifp);
2438 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2439 		ifa = ifp->if_addr;
2440 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2441 		namelen = strlen(new_name);
2442 		onamelen = sdl->sdl_nlen;
2443 		/*
2444 		 * Move the address if needed.  This is safe because we
2445 		 * allocate space for a name of length IFNAMSIZ when we
2446 		 * create this in if_attach().
2447 		 */
2448 		if (namelen != onamelen) {
2449 			bcopy(sdl->sdl_data + onamelen,
2450 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2451 		}
2452 		bcopy(new_name, sdl->sdl_data, namelen);
2453 		sdl->sdl_nlen = namelen;
2454 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2455 		bzero(sdl->sdl_data, onamelen);
2456 		while (namelen != 0)
2457 			sdl->sdl_data[--namelen] = 0xff;
2458 		IF_ADDR_WUNLOCK(ifp);
2459 
2460 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2461 		/* Announce the return of the interface. */
2462 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2463 
2464 		ifp->if_flags &= ~IFF_RENAMING;
2465 		break;
2466 
2467 #ifdef VIMAGE
2468 	case SIOCSIFVNET:
2469 		error = priv_check(td, PRIV_NET_SETIFVNET);
2470 		if (error)
2471 			return (error);
2472 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2473 		break;
2474 #endif
2475 
2476 	case SIOCSIFMETRIC:
2477 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2478 		if (error)
2479 			return (error);
2480 		ifp->if_metric = ifr->ifr_metric;
2481 		getmicrotime(&ifp->if_lastchange);
2482 		break;
2483 
2484 	case SIOCSIFPHYS:
2485 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2486 		if (error)
2487 			return (error);
2488 		if (ifp->if_ioctl == NULL)
2489 			return (EOPNOTSUPP);
2490 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2491 		if (error == 0)
2492 			getmicrotime(&ifp->if_lastchange);
2493 		break;
2494 
2495 	case SIOCSIFMTU:
2496 	{
2497 		u_long oldmtu = ifp->if_mtu;
2498 
2499 		error = priv_check(td, PRIV_NET_SETIFMTU);
2500 		if (error)
2501 			return (error);
2502 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2503 			return (EINVAL);
2504 		if (ifp->if_ioctl == NULL)
2505 			return (EOPNOTSUPP);
2506 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2507 		if (error == 0) {
2508 			getmicrotime(&ifp->if_lastchange);
2509 			rt_ifmsg(ifp);
2510 		}
2511 		/*
2512 		 * If the link MTU changed, do network layer specific procedure.
2513 		 */
2514 		if (ifp->if_mtu != oldmtu) {
2515 #ifdef INET6
2516 			nd6_setmtu(ifp);
2517 #endif
2518 			rt_updatemtu(ifp);
2519 		}
2520 		break;
2521 	}
2522 
2523 	case SIOCADDMULTI:
2524 	case SIOCDELMULTI:
2525 		if (cmd == SIOCADDMULTI)
2526 			error = priv_check(td, PRIV_NET_ADDMULTI);
2527 		else
2528 			error = priv_check(td, PRIV_NET_DELMULTI);
2529 		if (error)
2530 			return (error);
2531 
2532 		/* Don't allow group membership on non-multicast interfaces. */
2533 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2534 			return (EOPNOTSUPP);
2535 
2536 		/* Don't let users screw up protocols' entries. */
2537 		if (ifr->ifr_addr.sa_family != AF_LINK)
2538 			return (EINVAL);
2539 
2540 		if (cmd == SIOCADDMULTI) {
2541 			struct ifmultiaddr *ifma;
2542 
2543 			/*
2544 			 * Userland is only permitted to join groups once
2545 			 * via the if_addmulti() KPI, because it cannot hold
2546 			 * struct ifmultiaddr * between calls. It may also
2547 			 * lose a race while we check if the membership
2548 			 * already exists.
2549 			 */
2550 			IF_ADDR_RLOCK(ifp);
2551 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2552 			IF_ADDR_RUNLOCK(ifp);
2553 			if (ifma != NULL)
2554 				error = EADDRINUSE;
2555 			else
2556 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2557 		} else {
2558 			error = if_delmulti(ifp, &ifr->ifr_addr);
2559 		}
2560 		if (error == 0)
2561 			getmicrotime(&ifp->if_lastchange);
2562 		break;
2563 
2564 	case SIOCSIFPHYADDR:
2565 	case SIOCDIFPHYADDR:
2566 #ifdef INET6
2567 	case SIOCSIFPHYADDR_IN6:
2568 #endif
2569 	case SIOCSIFMEDIA:
2570 	case SIOCSIFGENERIC:
2571 		error = priv_check(td, PRIV_NET_HWIOCTL);
2572 		if (error)
2573 			return (error);
2574 		if (ifp->if_ioctl == NULL)
2575 			return (EOPNOTSUPP);
2576 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2577 		if (error == 0)
2578 			getmicrotime(&ifp->if_lastchange);
2579 		break;
2580 
2581 	case SIOCGIFSTATUS:
2582 	case SIOCGIFPSRCADDR:
2583 	case SIOCGIFPDSTADDR:
2584 	case SIOCGIFMEDIA:
2585 	case SIOCGIFGENERIC:
2586 		if (ifp->if_ioctl == NULL)
2587 			return (EOPNOTSUPP);
2588 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2589 		break;
2590 
2591 	case SIOCSIFLLADDR:
2592 		error = priv_check(td, PRIV_NET_SETLLADDR);
2593 		if (error)
2594 			return (error);
2595 		error = if_setlladdr(ifp,
2596 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2597 		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2598 		break;
2599 
2600 	case SIOCAIFGROUP:
2601 	{
2602 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2603 
2604 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2605 		if (error)
2606 			return (error);
2607 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2608 			return (error);
2609 		break;
2610 	}
2611 
2612 	case SIOCGIFGROUP:
2613 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2614 			return (error);
2615 		break;
2616 
2617 	case SIOCDIFGROUP:
2618 	{
2619 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2620 
2621 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2622 		if (error)
2623 			return (error);
2624 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2625 			return (error);
2626 		break;
2627 	}
2628 
2629 	default:
2630 		error = ENOIOCTL;
2631 		break;
2632 	}
2633 	return (error);
2634 }
2635 
2636 #ifdef COMPAT_FREEBSD32
2637 struct ifconf32 {
2638 	int32_t	ifc_len;
2639 	union {
2640 		uint32_t	ifcu_buf;
2641 		uint32_t	ifcu_req;
2642 	} ifc_ifcu;
2643 };
2644 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2645 #endif
2646 
2647 /*
2648  * Interface ioctls.
2649  */
2650 int
2651 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2652 {
2653 	struct ifnet *ifp;
2654 	struct ifreq *ifr;
2655 	int error;
2656 	int oif_flags;
2657 
2658 	CURVNET_SET(so->so_vnet);
2659 	switch (cmd) {
2660 	case SIOCGIFCONF:
2661 		error = ifconf(cmd, data);
2662 		CURVNET_RESTORE();
2663 		return (error);
2664 
2665 #ifdef COMPAT_FREEBSD32
2666 	case SIOCGIFCONF32:
2667 		{
2668 			struct ifconf32 *ifc32;
2669 			struct ifconf ifc;
2670 
2671 			ifc32 = (struct ifconf32 *)data;
2672 			ifc.ifc_len = ifc32->ifc_len;
2673 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2674 
2675 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2676 			CURVNET_RESTORE();
2677 			if (error == 0)
2678 				ifc32->ifc_len = ifc.ifc_len;
2679 			return (error);
2680 		}
2681 #endif
2682 	}
2683 	ifr = (struct ifreq *)data;
2684 
2685 	switch (cmd) {
2686 #ifdef VIMAGE
2687 	case SIOCSIFRVNET:
2688 		error = priv_check(td, PRIV_NET_SETIFVNET);
2689 		if (error == 0)
2690 			error = if_vmove_reclaim(td, ifr->ifr_name,
2691 			    ifr->ifr_jid);
2692 		CURVNET_RESTORE();
2693 		return (error);
2694 #endif
2695 	case SIOCIFCREATE:
2696 	case SIOCIFCREATE2:
2697 		error = priv_check(td, PRIV_NET_IFCREATE);
2698 		if (error == 0)
2699 			error = if_clone_create(ifr->ifr_name,
2700 			    sizeof(ifr->ifr_name),
2701 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2702 		CURVNET_RESTORE();
2703 		return (error);
2704 	case SIOCIFDESTROY:
2705 		error = priv_check(td, PRIV_NET_IFDESTROY);
2706 		if (error == 0)
2707 			error = if_clone_destroy(ifr->ifr_name);
2708 		CURVNET_RESTORE();
2709 		return (error);
2710 
2711 	case SIOCIFGCLONERS:
2712 		error = if_clone_list((struct if_clonereq *)data);
2713 		CURVNET_RESTORE();
2714 		return (error);
2715 	case SIOCGIFGMEMB:
2716 		error = if_getgroupmembers((struct ifgroupreq *)data);
2717 		CURVNET_RESTORE();
2718 		return (error);
2719 #if defined(INET) || defined(INET6)
2720 	case SIOCSVH:
2721 	case SIOCGVH:
2722 		if (carp_ioctl_p == NULL)
2723 			error = EPROTONOSUPPORT;
2724 		else
2725 			error = (*carp_ioctl_p)(ifr, cmd, td);
2726 		CURVNET_RESTORE();
2727 		return (error);
2728 #endif
2729 	}
2730 
2731 	ifp = ifunit_ref(ifr->ifr_name);
2732 	if (ifp == NULL) {
2733 		CURVNET_RESTORE();
2734 		return (ENXIO);
2735 	}
2736 
2737 	error = ifhwioctl(cmd, ifp, data, td);
2738 	if (error != ENOIOCTL) {
2739 		if_rele(ifp);
2740 		CURVNET_RESTORE();
2741 		return (error);
2742 	}
2743 
2744 	oif_flags = ifp->if_flags;
2745 	if (so->so_proto == NULL) {
2746 		if_rele(ifp);
2747 		CURVNET_RESTORE();
2748 		return (EOPNOTSUPP);
2749 	}
2750 
2751 	/*
2752 	 * Pass the request on to the socket control method, and if the
2753 	 * latter returns EOPNOTSUPP, directly to the interface.
2754 	 *
2755 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2756 	 * trust SIOCSIFADDR et al to come from an already privileged
2757 	 * layer, and do not perform any credentials checks or input
2758 	 * validation.
2759 	 */
2760 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2761 	    ifp, td));
2762 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2763 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2764 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2765 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2766 
2767 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2768 #ifdef INET6
2769 		if (ifp->if_flags & IFF_UP)
2770 			in6_if_up(ifp);
2771 #endif
2772 	}
2773 	if_rele(ifp);
2774 	CURVNET_RESTORE();
2775 	return (error);
2776 }
2777 
2778 /*
2779  * The code common to handling reference counted flags,
2780  * e.g., in ifpromisc() and if_allmulti().
2781  * The "pflag" argument can specify a permanent mode flag to check,
2782  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2783  *
2784  * Only to be used on stack-owned flags, not driver-owned flags.
2785  */
2786 static int
2787 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2788 {
2789 	struct ifreq ifr;
2790 	int error;
2791 	int oldflags, oldcount;
2792 
2793 	/* Sanity checks to catch programming errors */
2794 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2795 	    ("%s: setting driver-owned flag %d", __func__, flag));
2796 
2797 	if (onswitch)
2798 		KASSERT(*refcount >= 0,
2799 		    ("%s: increment negative refcount %d for flag %d",
2800 		    __func__, *refcount, flag));
2801 	else
2802 		KASSERT(*refcount > 0,
2803 		    ("%s: decrement non-positive refcount %d for flag %d",
2804 		    __func__, *refcount, flag));
2805 
2806 	/* In case this mode is permanent, just touch refcount */
2807 	if (ifp->if_flags & pflag) {
2808 		*refcount += onswitch ? 1 : -1;
2809 		return (0);
2810 	}
2811 
2812 	/* Save ifnet parameters for if_ioctl() may fail */
2813 	oldcount = *refcount;
2814 	oldflags = ifp->if_flags;
2815 
2816 	/*
2817 	 * See if we aren't the only and touching refcount is enough.
2818 	 * Actually toggle interface flag if we are the first or last.
2819 	 */
2820 	if (onswitch) {
2821 		if ((*refcount)++)
2822 			return (0);
2823 		ifp->if_flags |= flag;
2824 	} else {
2825 		if (--(*refcount))
2826 			return (0);
2827 		ifp->if_flags &= ~flag;
2828 	}
2829 
2830 	/* Call down the driver since we've changed interface flags */
2831 	if (ifp->if_ioctl == NULL) {
2832 		error = EOPNOTSUPP;
2833 		goto recover;
2834 	}
2835 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2836 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2837 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2838 	if (error)
2839 		goto recover;
2840 	/* Notify userland that interface flags have changed */
2841 	rt_ifmsg(ifp);
2842 	return (0);
2843 
2844 recover:
2845 	/* Recover after driver error */
2846 	*refcount = oldcount;
2847 	ifp->if_flags = oldflags;
2848 	return (error);
2849 }
2850 
2851 /*
2852  * Set/clear promiscuous mode on interface ifp based on the truth value
2853  * of pswitch.  The calls are reference counted so that only the first
2854  * "on" request actually has an effect, as does the final "off" request.
2855  * Results are undefined if the "off" and "on" requests are not matched.
2856  */
2857 int
2858 ifpromisc(struct ifnet *ifp, int pswitch)
2859 {
2860 	int error;
2861 	int oldflags = ifp->if_flags;
2862 
2863 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2864 			   &ifp->if_pcount, pswitch);
2865 	/* If promiscuous mode status has changed, log a message */
2866 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2867 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2868 		    ifp->if_xname,
2869 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2870 	return (error);
2871 }
2872 
2873 /*
2874  * Return interface configuration
2875  * of system.  List may be used
2876  * in later ioctl's (above) to get
2877  * other information.
2878  */
2879 /*ARGSUSED*/
2880 static int
2881 ifconf(u_long cmd, caddr_t data)
2882 {
2883 	struct ifconf *ifc = (struct ifconf *)data;
2884 	struct ifnet *ifp;
2885 	struct ifaddr *ifa;
2886 	struct ifreq ifr;
2887 	struct sbuf *sb;
2888 	int error, full = 0, valid_len, max_len;
2889 
2890 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2891 	max_len = MAXPHYS - 1;
2892 
2893 	/* Prevent hostile input from being able to crash the system */
2894 	if (ifc->ifc_len <= 0)
2895 		return (EINVAL);
2896 
2897 again:
2898 	if (ifc->ifc_len <= max_len) {
2899 		max_len = ifc->ifc_len;
2900 		full = 1;
2901 	}
2902 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2903 	max_len = 0;
2904 	valid_len = 0;
2905 
2906 	IFNET_RLOCK();
2907 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2908 		int addrs;
2909 
2910 		/*
2911 		 * Zero the ifr_name buffer to make sure we don't
2912 		 * disclose the contents of the stack.
2913 		 */
2914 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2915 
2916 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2917 		    >= sizeof(ifr.ifr_name)) {
2918 			sbuf_delete(sb);
2919 			IFNET_RUNLOCK();
2920 			return (ENAMETOOLONG);
2921 		}
2922 
2923 		addrs = 0;
2924 		IF_ADDR_RLOCK(ifp);
2925 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2926 			struct sockaddr *sa = ifa->ifa_addr;
2927 
2928 			if (prison_if(curthread->td_ucred, sa) != 0)
2929 				continue;
2930 			addrs++;
2931 			if (sa->sa_len <= sizeof(*sa)) {
2932 				ifr.ifr_addr = *sa;
2933 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2934 				max_len += sizeof(ifr);
2935 			} else {
2936 				sbuf_bcat(sb, &ifr,
2937 				    offsetof(struct ifreq, ifr_addr));
2938 				max_len += offsetof(struct ifreq, ifr_addr);
2939 				sbuf_bcat(sb, sa, sa->sa_len);
2940 				max_len += sa->sa_len;
2941 			}
2942 
2943 			if (sbuf_error(sb) == 0)
2944 				valid_len = sbuf_len(sb);
2945 		}
2946 		IF_ADDR_RUNLOCK(ifp);
2947 		if (addrs == 0) {
2948 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2949 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2950 			max_len += sizeof(ifr);
2951 
2952 			if (sbuf_error(sb) == 0)
2953 				valid_len = sbuf_len(sb);
2954 		}
2955 	}
2956 	IFNET_RUNLOCK();
2957 
2958 	/*
2959 	 * If we didn't allocate enough space (uncommon), try again.  If
2960 	 * we have already allocated as much space as we are allowed,
2961 	 * return what we've got.
2962 	 */
2963 	if (valid_len != max_len && !full) {
2964 		sbuf_delete(sb);
2965 		goto again;
2966 	}
2967 
2968 	ifc->ifc_len = valid_len;
2969 	sbuf_finish(sb);
2970 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2971 	sbuf_delete(sb);
2972 	return (error);
2973 }
2974 
2975 /*
2976  * Just like ifpromisc(), but for all-multicast-reception mode.
2977  */
2978 int
2979 if_allmulti(struct ifnet *ifp, int onswitch)
2980 {
2981 
2982 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2983 }
2984 
2985 struct ifmultiaddr *
2986 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2987 {
2988 	struct ifmultiaddr *ifma;
2989 
2990 	IF_ADDR_LOCK_ASSERT(ifp);
2991 
2992 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2993 		if (sa->sa_family == AF_LINK) {
2994 			if (sa_dl_equal(ifma->ifma_addr, sa))
2995 				break;
2996 		} else {
2997 			if (sa_equal(ifma->ifma_addr, sa))
2998 				break;
2999 		}
3000 	}
3001 
3002 	return ifma;
3003 }
3004 
3005 /*
3006  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3007  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3008  * the ifnet multicast address list here, so the caller must do that and
3009  * other setup work (such as notifying the device driver).  The reference
3010  * count is initialized to 1.
3011  */
3012 static struct ifmultiaddr *
3013 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3014     int mflags)
3015 {
3016 	struct ifmultiaddr *ifma;
3017 	struct sockaddr *dupsa;
3018 
3019 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3020 	    M_ZERO);
3021 	if (ifma == NULL)
3022 		return (NULL);
3023 
3024 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3025 	if (dupsa == NULL) {
3026 		free(ifma, M_IFMADDR);
3027 		return (NULL);
3028 	}
3029 	bcopy(sa, dupsa, sa->sa_len);
3030 	ifma->ifma_addr = dupsa;
3031 
3032 	ifma->ifma_ifp = ifp;
3033 	ifma->ifma_refcount = 1;
3034 	ifma->ifma_protospec = NULL;
3035 
3036 	if (llsa == NULL) {
3037 		ifma->ifma_lladdr = NULL;
3038 		return (ifma);
3039 	}
3040 
3041 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3042 	if (dupsa == NULL) {
3043 		free(ifma->ifma_addr, M_IFMADDR);
3044 		free(ifma, M_IFMADDR);
3045 		return (NULL);
3046 	}
3047 	bcopy(llsa, dupsa, llsa->sa_len);
3048 	ifma->ifma_lladdr = dupsa;
3049 
3050 	return (ifma);
3051 }
3052 
3053 /*
3054  * if_freemulti: free ifmultiaddr structure and possibly attached related
3055  * addresses.  The caller is responsible for implementing reference
3056  * counting, notifying the driver, handling routing messages, and releasing
3057  * any dependent link layer state.
3058  */
3059 static void
3060 if_freemulti(struct ifmultiaddr *ifma)
3061 {
3062 
3063 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3064 	    ifma->ifma_refcount));
3065 
3066 	if (ifma->ifma_lladdr != NULL)
3067 		free(ifma->ifma_lladdr, M_IFMADDR);
3068 	free(ifma->ifma_addr, M_IFMADDR);
3069 	free(ifma, M_IFMADDR);
3070 }
3071 
3072 /*
3073  * Register an additional multicast address with a network interface.
3074  *
3075  * - If the address is already present, bump the reference count on the
3076  *   address and return.
3077  * - If the address is not link-layer, look up a link layer address.
3078  * - Allocate address structures for one or both addresses, and attach to the
3079  *   multicast address list on the interface.  If automatically adding a link
3080  *   layer address, the protocol address will own a reference to the link
3081  *   layer address, to be freed when it is freed.
3082  * - Notify the network device driver of an addition to the multicast address
3083  *   list.
3084  *
3085  * 'sa' points to caller-owned memory with the desired multicast address.
3086  *
3087  * 'retifma' will be used to return a pointer to the resulting multicast
3088  * address reference, if desired.
3089  */
3090 int
3091 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3092     struct ifmultiaddr **retifma)
3093 {
3094 	struct ifmultiaddr *ifma, *ll_ifma;
3095 	struct sockaddr *llsa;
3096 	struct sockaddr_dl sdl;
3097 	int error;
3098 
3099 	/*
3100 	 * If the address is already present, return a new reference to it;
3101 	 * otherwise, allocate storage and set up a new address.
3102 	 */
3103 	IF_ADDR_WLOCK(ifp);
3104 	ifma = if_findmulti(ifp, sa);
3105 	if (ifma != NULL) {
3106 		ifma->ifma_refcount++;
3107 		if (retifma != NULL)
3108 			*retifma = ifma;
3109 		IF_ADDR_WUNLOCK(ifp);
3110 		return (0);
3111 	}
3112 
3113 	/*
3114 	 * The address isn't already present; resolve the protocol address
3115 	 * into a link layer address, and then look that up, bump its
3116 	 * refcount or allocate an ifma for that also.
3117 	 * Most link layer resolving functions returns address data which
3118 	 * fits inside default sockaddr_dl structure. However callback
3119 	 * can allocate another sockaddr structure, in that case we need to
3120 	 * free it later.
3121 	 */
3122 	llsa = NULL;
3123 	ll_ifma = NULL;
3124 	if (ifp->if_resolvemulti != NULL) {
3125 		/* Provide called function with buffer size information */
3126 		sdl.sdl_len = sizeof(sdl);
3127 		llsa = (struct sockaddr *)&sdl;
3128 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3129 		if (error)
3130 			goto unlock_out;
3131 	}
3132 
3133 	/*
3134 	 * Allocate the new address.  Don't hook it up yet, as we may also
3135 	 * need to allocate a link layer multicast address.
3136 	 */
3137 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3138 	if (ifma == NULL) {
3139 		error = ENOMEM;
3140 		goto free_llsa_out;
3141 	}
3142 
3143 	/*
3144 	 * If a link layer address is found, we'll need to see if it's
3145 	 * already present in the address list, or allocate is as well.
3146 	 * When this block finishes, the link layer address will be on the
3147 	 * list.
3148 	 */
3149 	if (llsa != NULL) {
3150 		ll_ifma = if_findmulti(ifp, llsa);
3151 		if (ll_ifma == NULL) {
3152 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3153 			if (ll_ifma == NULL) {
3154 				--ifma->ifma_refcount;
3155 				if_freemulti(ifma);
3156 				error = ENOMEM;
3157 				goto free_llsa_out;
3158 			}
3159 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3160 			    ifma_link);
3161 		} else
3162 			ll_ifma->ifma_refcount++;
3163 		ifma->ifma_llifma = ll_ifma;
3164 	}
3165 
3166 	/*
3167 	 * We now have a new multicast address, ifma, and possibly a new or
3168 	 * referenced link layer address.  Add the primary address to the
3169 	 * ifnet address list.
3170 	 */
3171 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3172 
3173 	if (retifma != NULL)
3174 		*retifma = ifma;
3175 
3176 	/*
3177 	 * Must generate the message while holding the lock so that 'ifma'
3178 	 * pointer is still valid.
3179 	 */
3180 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3181 	IF_ADDR_WUNLOCK(ifp);
3182 
3183 	/*
3184 	 * We are certain we have added something, so call down to the
3185 	 * interface to let them know about it.
3186 	 */
3187 	if (ifp->if_ioctl != NULL) {
3188 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3189 	}
3190 
3191 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3192 		link_free_sdl(llsa);
3193 
3194 	return (0);
3195 
3196 free_llsa_out:
3197 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3198 		link_free_sdl(llsa);
3199 
3200 unlock_out:
3201 	IF_ADDR_WUNLOCK(ifp);
3202 	return (error);
3203 }
3204 
3205 /*
3206  * Delete a multicast group membership by network-layer group address.
3207  *
3208  * Returns ENOENT if the entry could not be found. If ifp no longer
3209  * exists, results are undefined. This entry point should only be used
3210  * from subsystems which do appropriate locking to hold ifp for the
3211  * duration of the call.
3212  * Network-layer protocol domains must use if_delmulti_ifma().
3213  */
3214 int
3215 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3216 {
3217 	struct ifmultiaddr *ifma;
3218 	int lastref;
3219 #ifdef INVARIANTS
3220 	struct ifnet *oifp;
3221 
3222 	IFNET_RLOCK_NOSLEEP();
3223 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3224 		if (ifp == oifp)
3225 			break;
3226 	if (ifp != oifp)
3227 		ifp = NULL;
3228 	IFNET_RUNLOCK_NOSLEEP();
3229 
3230 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3231 #endif
3232 	if (ifp == NULL)
3233 		return (ENOENT);
3234 
3235 	IF_ADDR_WLOCK(ifp);
3236 	lastref = 0;
3237 	ifma = if_findmulti(ifp, sa);
3238 	if (ifma != NULL)
3239 		lastref = if_delmulti_locked(ifp, ifma, 0);
3240 	IF_ADDR_WUNLOCK(ifp);
3241 
3242 	if (ifma == NULL)
3243 		return (ENOENT);
3244 
3245 	if (lastref && ifp->if_ioctl != NULL) {
3246 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3247 	}
3248 
3249 	return (0);
3250 }
3251 
3252 /*
3253  * Delete all multicast group membership for an interface.
3254  * Should be used to quickly flush all multicast filters.
3255  */
3256 void
3257 if_delallmulti(struct ifnet *ifp)
3258 {
3259 	struct ifmultiaddr *ifma;
3260 	struct ifmultiaddr *next;
3261 
3262 	IF_ADDR_WLOCK(ifp);
3263 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3264 		if_delmulti_locked(ifp, ifma, 0);
3265 	IF_ADDR_WUNLOCK(ifp);
3266 }
3267 
3268 /*
3269  * Delete a multicast group membership by group membership pointer.
3270  * Network-layer protocol domains must use this routine.
3271  *
3272  * It is safe to call this routine if the ifp disappeared.
3273  */
3274 void
3275 if_delmulti_ifma(struct ifmultiaddr *ifma)
3276 {
3277 	struct ifnet *ifp;
3278 	int lastref;
3279 
3280 	ifp = ifma->ifma_ifp;
3281 #ifdef DIAGNOSTIC
3282 	if (ifp == NULL) {
3283 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3284 	} else {
3285 		struct ifnet *oifp;
3286 
3287 		IFNET_RLOCK_NOSLEEP();
3288 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3289 			if (ifp == oifp)
3290 				break;
3291 		if (ifp != oifp) {
3292 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3293 			ifp = NULL;
3294 		}
3295 		IFNET_RUNLOCK_NOSLEEP();
3296 	}
3297 #endif
3298 	/*
3299 	 * If and only if the ifnet instance exists: Acquire the address lock.
3300 	 */
3301 	if (ifp != NULL)
3302 		IF_ADDR_WLOCK(ifp);
3303 
3304 	lastref = if_delmulti_locked(ifp, ifma, 0);
3305 
3306 	if (ifp != NULL) {
3307 		/*
3308 		 * If and only if the ifnet instance exists:
3309 		 *  Release the address lock.
3310 		 *  If the group was left: update the hardware hash filter.
3311 		 */
3312 		IF_ADDR_WUNLOCK(ifp);
3313 		if (lastref && ifp->if_ioctl != NULL) {
3314 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3315 		}
3316 	}
3317 }
3318 
3319 /*
3320  * Perform deletion of network-layer and/or link-layer multicast address.
3321  *
3322  * Return 0 if the reference count was decremented.
3323  * Return 1 if the final reference was released, indicating that the
3324  * hardware hash filter should be reprogrammed.
3325  */
3326 static int
3327 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3328 {
3329 	struct ifmultiaddr *ll_ifma;
3330 
3331 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3332 		KASSERT(ifma->ifma_ifp == ifp,
3333 		    ("%s: inconsistent ifp %p", __func__, ifp));
3334 		IF_ADDR_WLOCK_ASSERT(ifp);
3335 	}
3336 
3337 	ifp = ifma->ifma_ifp;
3338 
3339 	/*
3340 	 * If the ifnet is detaching, null out references to ifnet,
3341 	 * so that upper protocol layers will notice, and not attempt
3342 	 * to obtain locks for an ifnet which no longer exists. The
3343 	 * routing socket announcement must happen before the ifnet
3344 	 * instance is detached from the system.
3345 	 */
3346 	if (detaching) {
3347 #ifdef DIAGNOSTIC
3348 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3349 #endif
3350 		/*
3351 		 * ifp may already be nulled out if we are being reentered
3352 		 * to delete the ll_ifma.
3353 		 */
3354 		if (ifp != NULL) {
3355 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3356 			ifma->ifma_ifp = NULL;
3357 		}
3358 	}
3359 
3360 	if (--ifma->ifma_refcount > 0)
3361 		return 0;
3362 
3363 	/*
3364 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3365 	 * have been associated with it. Release it first if so.
3366 	 */
3367 	ll_ifma = ifma->ifma_llifma;
3368 	if (ll_ifma != NULL) {
3369 		KASSERT(ifma->ifma_lladdr != NULL,
3370 		    ("%s: llifma w/o lladdr", __func__));
3371 		if (detaching)
3372 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3373 		if (--ll_ifma->ifma_refcount == 0) {
3374 			if (ifp != NULL) {
3375 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3376 				    ifma_link);
3377 			}
3378 			if_freemulti(ll_ifma);
3379 		}
3380 	}
3381 
3382 	if (ifp != NULL)
3383 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3384 
3385 	if_freemulti(ifma);
3386 
3387 	/*
3388 	 * The last reference to this instance of struct ifmultiaddr
3389 	 * was released; the hardware should be notified of this change.
3390 	 */
3391 	return 1;
3392 }
3393 
3394 /*
3395  * Set the link layer address on an interface.
3396  *
3397  * At this time we only support certain types of interfaces,
3398  * and we don't allow the length of the address to change.
3399  */
3400 int
3401 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3402 {
3403 	struct sockaddr_dl *sdl;
3404 	struct ifaddr *ifa;
3405 	struct ifreq ifr;
3406 
3407 	IF_ADDR_RLOCK(ifp);
3408 	ifa = ifp->if_addr;
3409 	if (ifa == NULL) {
3410 		IF_ADDR_RUNLOCK(ifp);
3411 		return (EINVAL);
3412 	}
3413 	ifa_ref(ifa);
3414 	IF_ADDR_RUNLOCK(ifp);
3415 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3416 	if (sdl == NULL) {
3417 		ifa_free(ifa);
3418 		return (EINVAL);
3419 	}
3420 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3421 		ifa_free(ifa);
3422 		return (EINVAL);
3423 	}
3424 	switch (ifp->if_type) {
3425 	case IFT_ETHER:
3426 	case IFT_FDDI:
3427 	case IFT_XETHER:
3428 	case IFT_ISO88025:
3429 	case IFT_L2VLAN:
3430 	case IFT_BRIDGE:
3431 	case IFT_ARCNET:
3432 	case IFT_IEEE8023ADLAG:
3433 	case IFT_IEEE80211:
3434 		bcopy(lladdr, LLADDR(sdl), len);
3435 		ifa_free(ifa);
3436 		break;
3437 	default:
3438 		ifa_free(ifa);
3439 		return (ENODEV);
3440 	}
3441 
3442 	/*
3443 	 * If the interface is already up, we need
3444 	 * to re-init it in order to reprogram its
3445 	 * address filter.
3446 	 */
3447 	if ((ifp->if_flags & IFF_UP) != 0) {
3448 		if (ifp->if_ioctl) {
3449 			ifp->if_flags &= ~IFF_UP;
3450 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3451 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3452 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3453 			ifp->if_flags |= IFF_UP;
3454 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3455 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3456 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3457 		}
3458 #ifdef INET
3459 		/*
3460 		 * Also send gratuitous ARPs to notify other nodes about
3461 		 * the address change.
3462 		 */
3463 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3464 			if (ifa->ifa_addr->sa_family == AF_INET)
3465 				arp_ifinit(ifp, ifa);
3466 		}
3467 #endif
3468 	}
3469 	return (0);
3470 }
3471 
3472 /*
3473  * The name argument must be a pointer to storage which will last as
3474  * long as the interface does.  For physical devices, the result of
3475  * device_get_name(dev) is a good choice and for pseudo-devices a
3476  * static string works well.
3477  */
3478 void
3479 if_initname(struct ifnet *ifp, const char *name, int unit)
3480 {
3481 	ifp->if_dname = name;
3482 	ifp->if_dunit = unit;
3483 	if (unit != IF_DUNIT_NONE)
3484 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3485 	else
3486 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3487 }
3488 
3489 int
3490 if_printf(struct ifnet *ifp, const char * fmt, ...)
3491 {
3492 	va_list ap;
3493 	int retval;
3494 
3495 	retval = printf("%s: ", ifp->if_xname);
3496 	va_start(ap, fmt);
3497 	retval += vprintf(fmt, ap);
3498 	va_end(ap);
3499 	return (retval);
3500 }
3501 
3502 void
3503 if_start(struct ifnet *ifp)
3504 {
3505 
3506 	(*(ifp)->if_start)(ifp);
3507 }
3508 
3509 /*
3510  * Backwards compatibility interface for drivers
3511  * that have not implemented it
3512  */
3513 static int
3514 if_transmit(struct ifnet *ifp, struct mbuf *m)
3515 {
3516 	int error;
3517 
3518 	IFQ_HANDOFF(ifp, m, error);
3519 	return (error);
3520 }
3521 
3522 static void
3523 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3524 {
3525 
3526 	m_freem(m);
3527 }
3528 
3529 int
3530 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3531 {
3532 	int active = 0;
3533 
3534 	IF_LOCK(ifq);
3535 	if (_IF_QFULL(ifq)) {
3536 		IF_UNLOCK(ifq);
3537 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3538 		m_freem(m);
3539 		return (0);
3540 	}
3541 	if (ifp != NULL) {
3542 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3543 		if (m->m_flags & (M_BCAST|M_MCAST))
3544 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3545 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3546 	}
3547 	_IF_ENQUEUE(ifq, m);
3548 	IF_UNLOCK(ifq);
3549 	if (ifp != NULL && !active)
3550 		(*(ifp)->if_start)(ifp);
3551 	return (1);
3552 }
3553 
3554 void
3555 if_register_com_alloc(u_char type,
3556     if_com_alloc_t *a, if_com_free_t *f)
3557 {
3558 
3559 	KASSERT(if_com_alloc[type] == NULL,
3560 	    ("if_register_com_alloc: %d already registered", type));
3561 	KASSERT(if_com_free[type] == NULL,
3562 	    ("if_register_com_alloc: %d free already registered", type));
3563 
3564 	if_com_alloc[type] = a;
3565 	if_com_free[type] = f;
3566 }
3567 
3568 void
3569 if_deregister_com_alloc(u_char type)
3570 {
3571 
3572 	KASSERT(if_com_alloc[type] != NULL,
3573 	    ("if_deregister_com_alloc: %d not registered", type));
3574 	KASSERT(if_com_free[type] != NULL,
3575 	    ("if_deregister_com_alloc: %d free not registered", type));
3576 	if_com_alloc[type] = NULL;
3577 	if_com_free[type] = NULL;
3578 }
3579 
3580 /* API for driver access to network stack owned ifnet.*/
3581 uint64_t
3582 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3583 {
3584 	uint64_t oldbrate;
3585 
3586 	oldbrate = ifp->if_baudrate;
3587 	ifp->if_baudrate = baudrate;
3588 	return (oldbrate);
3589 }
3590 
3591 uint64_t
3592 if_getbaudrate(if_t ifp)
3593 {
3594 
3595 	return (((struct ifnet *)ifp)->if_baudrate);
3596 }
3597 
3598 int
3599 if_setcapabilities(if_t ifp, int capabilities)
3600 {
3601 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3602 	return (0);
3603 }
3604 
3605 int
3606 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3607 {
3608 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3609 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3610 
3611 	return (0);
3612 }
3613 
3614 int
3615 if_getcapabilities(if_t ifp)
3616 {
3617 	return ((struct ifnet *)ifp)->if_capabilities;
3618 }
3619 
3620 int
3621 if_setcapenable(if_t ifp, int capabilities)
3622 {
3623 	((struct ifnet *)ifp)->if_capenable = capabilities;
3624 	return (0);
3625 }
3626 
3627 int
3628 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3629 {
3630 	if(setcap)
3631 		((struct ifnet *)ifp)->if_capenable |= setcap;
3632 	if(clearcap)
3633 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3634 
3635 	return (0);
3636 }
3637 
3638 const char *
3639 if_getdname(if_t ifp)
3640 {
3641 	return ((struct ifnet *)ifp)->if_dname;
3642 }
3643 
3644 int
3645 if_togglecapenable(if_t ifp, int togglecap)
3646 {
3647 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3648 	return (0);
3649 }
3650 
3651 int
3652 if_getcapenable(if_t ifp)
3653 {
3654 	return ((struct ifnet *)ifp)->if_capenable;
3655 }
3656 
3657 /*
3658  * This is largely undesirable because it ties ifnet to a device, but does
3659  * provide flexiblity for an embedded product vendor. Should be used with
3660  * the understanding that it violates the interface boundaries, and should be
3661  * a last resort only.
3662  */
3663 int
3664 if_setdev(if_t ifp, void *dev)
3665 {
3666 	return (0);
3667 }
3668 
3669 int
3670 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3671 {
3672 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3673 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3674 
3675 	return (0);
3676 }
3677 
3678 int
3679 if_getdrvflags(if_t ifp)
3680 {
3681 	return ((struct ifnet *)ifp)->if_drv_flags;
3682 }
3683 
3684 int
3685 if_setdrvflags(if_t ifp, int flags)
3686 {
3687 	((struct ifnet *)ifp)->if_drv_flags = flags;
3688 	return (0);
3689 }
3690 
3691 
3692 int
3693 if_setflags(if_t ifp, int flags)
3694 {
3695 	((struct ifnet *)ifp)->if_flags = flags;
3696 	return (0);
3697 }
3698 
3699 int
3700 if_setflagbits(if_t ifp, int set, int clear)
3701 {
3702 	((struct ifnet *)ifp)->if_flags |= set;
3703 	((struct ifnet *)ifp)->if_flags &= ~clear;
3704 
3705 	return (0);
3706 }
3707 
3708 int
3709 if_getflags(if_t ifp)
3710 {
3711 	return ((struct ifnet *)ifp)->if_flags;
3712 }
3713 
3714 int
3715 if_clearhwassist(if_t ifp)
3716 {
3717 	((struct ifnet *)ifp)->if_hwassist = 0;
3718 	return (0);
3719 }
3720 
3721 int
3722 if_sethwassistbits(if_t ifp, int toset, int toclear)
3723 {
3724 	((struct ifnet *)ifp)->if_hwassist |= toset;
3725 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3726 
3727 	return (0);
3728 }
3729 
3730 int
3731 if_sethwassist(if_t ifp, int hwassist_bit)
3732 {
3733 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3734 	return (0);
3735 }
3736 
3737 int
3738 if_gethwassist(if_t ifp)
3739 {
3740 	return ((struct ifnet *)ifp)->if_hwassist;
3741 }
3742 
3743 int
3744 if_setmtu(if_t ifp, int mtu)
3745 {
3746 	((struct ifnet *)ifp)->if_mtu = mtu;
3747 	return (0);
3748 }
3749 
3750 int
3751 if_getmtu(if_t ifp)
3752 {
3753 	return ((struct ifnet *)ifp)->if_mtu;
3754 }
3755 
3756 int
3757 if_getmtu_family(if_t ifp, int family)
3758 {
3759 	struct domain *dp;
3760 
3761 	for (dp = domains; dp; dp = dp->dom_next) {
3762 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3763 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3764 	}
3765 
3766 	return (((struct ifnet *)ifp)->if_mtu);
3767 }
3768 
3769 int
3770 if_setsoftc(if_t ifp, void *softc)
3771 {
3772 	((struct ifnet *)ifp)->if_softc = softc;
3773 	return (0);
3774 }
3775 
3776 void *
3777 if_getsoftc(if_t ifp)
3778 {
3779 	return ((struct ifnet *)ifp)->if_softc;
3780 }
3781 
3782 void
3783 if_setrcvif(struct mbuf *m, if_t ifp)
3784 {
3785 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3786 }
3787 
3788 void
3789 if_setvtag(struct mbuf *m, uint16_t tag)
3790 {
3791 	m->m_pkthdr.ether_vtag = tag;
3792 }
3793 
3794 uint16_t
3795 if_getvtag(struct mbuf *m)
3796 {
3797 
3798 	return (m->m_pkthdr.ether_vtag);
3799 }
3800 
3801 int
3802 if_sendq_empty(if_t ifp)
3803 {
3804 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3805 }
3806 
3807 struct ifaddr *
3808 if_getifaddr(if_t ifp)
3809 {
3810 	return ((struct ifnet *)ifp)->if_addr;
3811 }
3812 
3813 int
3814 if_getamcount(if_t ifp)
3815 {
3816 	return ((struct ifnet *)ifp)->if_amcount;
3817 }
3818 
3819 
3820 int
3821 if_setsendqready(if_t ifp)
3822 {
3823 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3824 	return (0);
3825 }
3826 
3827 int
3828 if_setsendqlen(if_t ifp, int tx_desc_count)
3829 {
3830 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3831 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3832 
3833 	return (0);
3834 }
3835 
3836 int
3837 if_vlantrunkinuse(if_t ifp)
3838 {
3839 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3840 }
3841 
3842 int
3843 if_input(if_t ifp, struct mbuf* sendmp)
3844 {
3845 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3846 	return (0);
3847 
3848 }
3849 
3850 /* XXX */
3851 #ifndef ETH_ADDR_LEN
3852 #define ETH_ADDR_LEN 6
3853 #endif
3854 
3855 int
3856 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3857 {
3858 	struct ifmultiaddr *ifma;
3859 	uint8_t *lmta = (uint8_t *)mta;
3860 	int mcnt = 0;
3861 
3862 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3863 		if (ifma->ifma_addr->sa_family != AF_LINK)
3864 			continue;
3865 
3866 		if (mcnt == max)
3867 			break;
3868 
3869 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3870 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3871 		mcnt++;
3872 	}
3873 	*cnt = mcnt;
3874 
3875 	return (0);
3876 }
3877 
3878 int
3879 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3880 {
3881 	int error;
3882 
3883 	if_maddr_rlock(ifp);
3884 	error = if_setupmultiaddr(ifp, mta, cnt, max);
3885 	if_maddr_runlock(ifp);
3886 	return (error);
3887 }
3888 
3889 int
3890 if_multiaddr_count(if_t ifp, int max)
3891 {
3892 	struct ifmultiaddr *ifma;
3893 	int count;
3894 
3895 	count = 0;
3896 	if_maddr_rlock(ifp);
3897 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3898 		if (ifma->ifma_addr->sa_family != AF_LINK)
3899 			continue;
3900 		count++;
3901 		if (count == max)
3902 			break;
3903 	}
3904 	if_maddr_runlock(ifp);
3905 	return (count);
3906 }
3907 
3908 struct mbuf *
3909 if_dequeue(if_t ifp)
3910 {
3911 	struct mbuf *m;
3912 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3913 
3914 	return (m);
3915 }
3916 
3917 int
3918 if_sendq_prepend(if_t ifp, struct mbuf *m)
3919 {
3920 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3921 	return (0);
3922 }
3923 
3924 int
3925 if_setifheaderlen(if_t ifp, int len)
3926 {
3927 	((struct ifnet *)ifp)->if_hdrlen = len;
3928 	return (0);
3929 }
3930 
3931 caddr_t
3932 if_getlladdr(if_t ifp)
3933 {
3934 	return (IF_LLADDR((struct ifnet *)ifp));
3935 }
3936 
3937 void *
3938 if_gethandle(u_char type)
3939 {
3940 	return (if_alloc(type));
3941 }
3942 
3943 void
3944 if_bpfmtap(if_t ifh, struct mbuf *m)
3945 {
3946 	struct ifnet *ifp = (struct ifnet *)ifh;
3947 
3948 	BPF_MTAP(ifp, m);
3949 }
3950 
3951 void
3952 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3953 {
3954 	struct ifnet *ifp = (struct ifnet *)ifh;
3955 
3956 	ETHER_BPF_MTAP(ifp, m);
3957 }
3958 
3959 void
3960 if_vlancap(if_t ifh)
3961 {
3962 	struct ifnet *ifp = (struct ifnet *)ifh;
3963 	VLAN_CAPABILITIES(ifp);
3964 }
3965 
3966 void
3967 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3968 {
3969 	((struct ifnet *)ifp)->if_init = init_fn;
3970 }
3971 
3972 void
3973 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3974 {
3975 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3976 }
3977 
3978 void
3979 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3980 {
3981 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
3982 }
3983 
3984 void
3985 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3986 {
3987 	((struct ifnet *)ifp)->if_transmit = start_fn;
3988 }
3989 
3990 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3991 {
3992 	((struct ifnet *)ifp)->if_qflush = flush_fn;
3993 
3994 }
3995 
3996 void
3997 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
3998 {
3999 
4000 	ifp->if_get_counter = fn;
4001 }
4002 
4003 /* Revisit these - These are inline functions originally. */
4004 int
4005 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4006 {
4007 	return drbr_inuse_drv(ifh, br);
4008 }
4009 
4010 struct mbuf*
4011 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4012 {
4013 	return drbr_dequeue(ifh, br);
4014 }
4015 
4016 int
4017 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4018 {
4019 	return drbr_needs_enqueue(ifh, br);
4020 }
4021 
4022 int
4023 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4024 {
4025 	return drbr_enqueue(ifh, br, m);
4026 
4027 }
4028