xref: /freebsd/sys/net/if.c (revision 5e53a4f90f82c4345f277dd87cc9292f26e04a29)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_compat.h"
36 #include "opt_inet6.h"
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/conf.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <sys/priv.h>
64 
65 #include <machine/stdarg.h>
66 #include <vm/uma.h>
67 
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/if_media.h>
77 #include <net/if_vlan_var.h>
78 #include <net/radix.h>
79 #include <net/route.h>
80 #include <net/vnet.h>
81 
82 #if defined(INET) || defined(INET6)
83 #include <net/ethernet.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip.h>
87 #include <netinet/ip_carp.h>
88 #ifdef INET
89 #include <netinet/if_ether.h>
90 #endif /* INET */
91 #ifdef INET6
92 #include <netinet6/in6_var.h>
93 #include <netinet6/in6_ifattach.h>
94 #endif /* INET6 */
95 #endif /* INET || INET6 */
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef COMPAT_FREEBSD32
100 #include <sys/mount.h>
101 #include <compat/freebsd32/freebsd32.h>
102 #endif
103 
104 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
105 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
106 
107 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
108     &ifqmaxlen, 0, "max send queue size");
109 
110 /* Log link state change events */
111 static int log_link_state_change = 1;
112 
113 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
114 	&log_link_state_change, 0,
115 	"log interface link state change events");
116 
117 /* Log promiscuous mode change events */
118 static int log_promisc_mode_change = 1;
119 
120 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
121 	&log_promisc_mode_change, 1,
122 	"log promiscuous mode change events");
123 
124 /* Interface description */
125 static unsigned int ifdescr_maxlen = 1024;
126 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
127 	&ifdescr_maxlen, 0,
128 	"administrative maximum length for interface description");
129 
130 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
131 
132 /* global sx for non-critical path ifdescr */
133 static struct sx ifdescr_sx;
134 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
135 
136 void	(*bridge_linkstate_p)(struct ifnet *ifp);
137 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
138 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
139 /* These are external hooks for CARP. */
140 void	(*carp_linkstate_p)(struct ifnet *ifp);
141 void	(*carp_demote_adj_p)(int, char *);
142 int	(*carp_master_p)(struct ifaddr *);
143 #if defined(INET) || defined(INET6)
144 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
145 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
146     const struct sockaddr *sa);
147 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
148 int	(*carp_attach_p)(struct ifaddr *, int);
149 void	(*carp_detach_p)(struct ifaddr *, bool);
150 #endif
151 #ifdef INET
152 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
153 #endif
154 #ifdef INET6
155 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
156 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
157     const struct in6_addr *taddr);
158 #endif
159 
160 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
161 
162 /*
163  * XXX: Style; these should be sorted alphabetically, and unprototyped
164  * static functions should be prototyped. Currently they are sorted by
165  * declaration order.
166  */
167 static void	if_attachdomain(void *);
168 static void	if_attachdomain1(struct ifnet *);
169 static int	ifconf(u_long, caddr_t);
170 static void	if_freemulti(struct ifmultiaddr *);
171 static void	if_grow(void);
172 static void	if_input_default(struct ifnet *, struct mbuf *);
173 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
174 static void	if_route(struct ifnet *, int flag, int fam);
175 static int	if_setflag(struct ifnet *, int, int, int *, int);
176 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
177 static void	if_unroute(struct ifnet *, int flag, int fam);
178 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
179 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
180 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
181 static void	do_link_state_change(void *, int);
182 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
183 static int	if_getgroupmembers(struct ifgroupreq *);
184 static void	if_delgroups(struct ifnet *);
185 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
186 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
187 #ifdef VIMAGE
188 static void	if_vmove(struct ifnet *, struct vnet *);
189 #endif
190 
191 #ifdef INET6
192 /*
193  * XXX: declare here to avoid to include many inet6 related files..
194  * should be more generalized?
195  */
196 extern void	nd6_setmtu(struct ifnet *);
197 #endif
198 
199 /* ipsec helper hooks */
200 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
201 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
202 
203 VNET_DEFINE(int, if_index);
204 int	ifqmaxlen = IFQ_MAXLEN;
205 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
206 VNET_DEFINE(struct ifgrouphead, ifg_head);
207 
208 static VNET_DEFINE(int, if_indexlim) = 8;
209 
210 /* Table of ifnet by index. */
211 VNET_DEFINE(struct ifnet **, ifindex_table);
212 
213 #define	V_if_indexlim		VNET(if_indexlim)
214 #define	V_ifindex_table		VNET(ifindex_table)
215 
216 /*
217  * The global network interface list (V_ifnet) and related state (such as
218  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
219  * an rwlock.  Either may be acquired shared to stablize the list, but both
220  * must be acquired writable to modify the list.  This model allows us to
221  * both stablize the interface list during interrupt thread processing, but
222  * also to stablize it over long-running ioctls, without introducing priority
223  * inversions and deadlocks.
224  */
225 struct rwlock ifnet_rwlock;
226 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
227 struct sx ifnet_sxlock;
228 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
229 
230 /*
231  * The allocation of network interfaces is a rather non-atomic affair; we
232  * need to select an index before we are ready to expose the interface for
233  * use, so will use this pointer value to indicate reservation.
234  */
235 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
236 
237 static	if_com_alloc_t *if_com_alloc[256];
238 static	if_com_free_t *if_com_free[256];
239 
240 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
241 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
242 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
243 
244 struct ifnet *
245 ifnet_byindex_locked(u_short idx)
246 {
247 
248 	if (idx > V_if_index)
249 		return (NULL);
250 	if (V_ifindex_table[idx] == IFNET_HOLD)
251 		return (NULL);
252 	return (V_ifindex_table[idx]);
253 }
254 
255 struct ifnet *
256 ifnet_byindex(u_short idx)
257 {
258 	struct ifnet *ifp;
259 
260 	IFNET_RLOCK_NOSLEEP();
261 	ifp = ifnet_byindex_locked(idx);
262 	IFNET_RUNLOCK_NOSLEEP();
263 	return (ifp);
264 }
265 
266 struct ifnet *
267 ifnet_byindex_ref(u_short idx)
268 {
269 	struct ifnet *ifp;
270 
271 	IFNET_RLOCK_NOSLEEP();
272 	ifp = ifnet_byindex_locked(idx);
273 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
274 		IFNET_RUNLOCK_NOSLEEP();
275 		return (NULL);
276 	}
277 	if_ref(ifp);
278 	IFNET_RUNLOCK_NOSLEEP();
279 	return (ifp);
280 }
281 
282 /*
283  * Allocate an ifindex array entry; return 0 on success or an error on
284  * failure.
285  */
286 static u_short
287 ifindex_alloc(void)
288 {
289 	u_short idx;
290 
291 	IFNET_WLOCK_ASSERT();
292 retry:
293 	/*
294 	 * Try to find an empty slot below V_if_index.  If we fail, take the
295 	 * next slot.
296 	 */
297 	for (idx = 1; idx <= V_if_index; idx++) {
298 		if (V_ifindex_table[idx] == NULL)
299 			break;
300 	}
301 
302 	/* Catch if_index overflow. */
303 	if (idx >= V_if_indexlim) {
304 		if_grow();
305 		goto retry;
306 	}
307 	if (idx > V_if_index)
308 		V_if_index = idx;
309 	return (idx);
310 }
311 
312 static void
313 ifindex_free_locked(u_short idx)
314 {
315 
316 	IFNET_WLOCK_ASSERT();
317 
318 	V_ifindex_table[idx] = NULL;
319 	while (V_if_index > 0 &&
320 	    V_ifindex_table[V_if_index] == NULL)
321 		V_if_index--;
322 }
323 
324 static void
325 ifindex_free(u_short idx)
326 {
327 
328 	IFNET_WLOCK();
329 	ifindex_free_locked(idx);
330 	IFNET_WUNLOCK();
331 }
332 
333 static void
334 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
335 {
336 
337 	IFNET_WLOCK_ASSERT();
338 
339 	V_ifindex_table[idx] = ifp;
340 }
341 
342 static void
343 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
344 {
345 
346 	IFNET_WLOCK();
347 	ifnet_setbyindex_locked(idx, ifp);
348 	IFNET_WUNLOCK();
349 }
350 
351 struct ifaddr *
352 ifaddr_byindex(u_short idx)
353 {
354 	struct ifnet *ifp;
355 	struct ifaddr *ifa = NULL;
356 
357 	IFNET_RLOCK_NOSLEEP();
358 	ifp = ifnet_byindex_locked(idx);
359 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
360 		ifa_ref(ifa);
361 	IFNET_RUNLOCK_NOSLEEP();
362 	return (ifa);
363 }
364 
365 /*
366  * Network interface utility routines.
367  *
368  * Routines with ifa_ifwith* names take sockaddr *'s as
369  * parameters.
370  */
371 
372 static void
373 vnet_if_init(const void *unused __unused)
374 {
375 
376 	TAILQ_INIT(&V_ifnet);
377 	TAILQ_INIT(&V_ifg_head);
378 	IFNET_WLOCK();
379 	if_grow();				/* create initial table */
380 	IFNET_WUNLOCK();
381 	vnet_if_clone_init();
382 }
383 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
384     NULL);
385 
386 #ifdef VIMAGE
387 static void
388 vnet_if_uninit(const void *unused __unused)
389 {
390 
391 	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
392 	    "not empty", __func__, __LINE__, &V_ifnet));
393 	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
394 	    "not empty", __func__, __LINE__, &V_ifg_head));
395 
396 	free((caddr_t)V_ifindex_table, M_IFNET);
397 }
398 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
399     vnet_if_uninit, NULL);
400 
401 static void
402 vnet_if_return(const void *unused __unused)
403 {
404 	struct ifnet *ifp, *nifp;
405 
406 	/* Return all inherited interfaces to their parent vnets. */
407 	TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
408 		if (ifp->if_home_vnet != ifp->if_vnet)
409 			if_vmove(ifp, ifp->if_home_vnet);
410 	}
411 }
412 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
413     vnet_if_return, NULL);
414 #endif
415 
416 static void
417 if_grow(void)
418 {
419 	int oldlim;
420 	u_int n;
421 	struct ifnet **e;
422 
423 	IFNET_WLOCK_ASSERT();
424 	oldlim = V_if_indexlim;
425 	IFNET_WUNLOCK();
426 	n = (oldlim << 1) * sizeof(*e);
427 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
428 	IFNET_WLOCK();
429 	if (V_if_indexlim != oldlim) {
430 		free(e, M_IFNET);
431 		return;
432 	}
433 	if (V_ifindex_table != NULL) {
434 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
435 		free((caddr_t)V_ifindex_table, M_IFNET);
436 	}
437 	V_if_indexlim <<= 1;
438 	V_ifindex_table = e;
439 }
440 
441 /*
442  * Allocate a struct ifnet and an index for an interface.  A layer 2
443  * common structure will also be allocated if an allocation routine is
444  * registered for the passed type.
445  */
446 struct ifnet *
447 if_alloc(u_char type)
448 {
449 	struct ifnet *ifp;
450 	u_short idx;
451 
452 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
453 	IFNET_WLOCK();
454 	idx = ifindex_alloc();
455 	ifnet_setbyindex_locked(idx, IFNET_HOLD);
456 	IFNET_WUNLOCK();
457 	ifp->if_index = idx;
458 	ifp->if_type = type;
459 	ifp->if_alloctype = type;
460 #ifdef VIMAGE
461 	ifp->if_vnet = curvnet;
462 #endif
463 	if (if_com_alloc[type] != NULL) {
464 		ifp->if_l2com = if_com_alloc[type](type, ifp);
465 		if (ifp->if_l2com == NULL) {
466 			free(ifp, M_IFNET);
467 			ifindex_free(idx);
468 			return (NULL);
469 		}
470 	}
471 
472 	IF_ADDR_LOCK_INIT(ifp);
473 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
474 	ifp->if_afdata_initialized = 0;
475 	IF_AFDATA_LOCK_INIT(ifp);
476 	TAILQ_INIT(&ifp->if_addrhead);
477 	TAILQ_INIT(&ifp->if_multiaddrs);
478 	TAILQ_INIT(&ifp->if_groups);
479 #ifdef MAC
480 	mac_ifnet_init(ifp);
481 #endif
482 	ifq_init(&ifp->if_snd, ifp);
483 
484 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
485 	for (int i = 0; i < IFCOUNTERS; i++)
486 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
487 	ifp->if_get_counter = if_get_counter_default;
488 	ifnet_setbyindex(ifp->if_index, ifp);
489 	return (ifp);
490 }
491 
492 /*
493  * Do the actual work of freeing a struct ifnet, and layer 2 common
494  * structure.  This call is made when the last reference to an
495  * interface is released.
496  */
497 static void
498 if_free_internal(struct ifnet *ifp)
499 {
500 
501 	KASSERT((ifp->if_flags & IFF_DYING),
502 	    ("if_free_internal: interface not dying"));
503 
504 	if (if_com_free[ifp->if_alloctype] != NULL)
505 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
506 		    ifp->if_alloctype);
507 
508 #ifdef MAC
509 	mac_ifnet_destroy(ifp);
510 #endif /* MAC */
511 	if (ifp->if_description != NULL)
512 		free(ifp->if_description, M_IFDESCR);
513 	IF_AFDATA_DESTROY(ifp);
514 	IF_ADDR_LOCK_DESTROY(ifp);
515 	ifq_delete(&ifp->if_snd);
516 
517 	for (int i = 0; i < IFCOUNTERS; i++)
518 		counter_u64_free(ifp->if_counters[i]);
519 
520 	free(ifp, M_IFNET);
521 }
522 
523 /*
524  * Deregister an interface and free the associated storage.
525  */
526 void
527 if_free(struct ifnet *ifp)
528 {
529 
530 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
531 
532 	CURVNET_SET_QUIET(ifp->if_vnet);
533 	IFNET_WLOCK();
534 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
535 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
536 
537 	ifindex_free_locked(ifp->if_index);
538 	IFNET_WUNLOCK();
539 
540 	if (refcount_release(&ifp->if_refcount))
541 		if_free_internal(ifp);
542 	CURVNET_RESTORE();
543 }
544 
545 /*
546  * Interfaces to keep an ifnet type-stable despite the possibility of the
547  * driver calling if_free().  If there are additional references, we defer
548  * freeing the underlying data structure.
549  */
550 void
551 if_ref(struct ifnet *ifp)
552 {
553 
554 	/* We don't assert the ifnet list lock here, but arguably should. */
555 	refcount_acquire(&ifp->if_refcount);
556 }
557 
558 void
559 if_rele(struct ifnet *ifp)
560 {
561 
562 	if (!refcount_release(&ifp->if_refcount))
563 		return;
564 	if_free_internal(ifp);
565 }
566 
567 void
568 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
569 {
570 
571 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
572 
573 	if (ifq->ifq_maxlen == 0)
574 		ifq->ifq_maxlen = ifqmaxlen;
575 
576 	ifq->altq_type = 0;
577 	ifq->altq_disc = NULL;
578 	ifq->altq_flags &= ALTQF_CANTCHANGE;
579 	ifq->altq_tbr  = NULL;
580 	ifq->altq_ifp  = ifp;
581 }
582 
583 void
584 ifq_delete(struct ifaltq *ifq)
585 {
586 	mtx_destroy(&ifq->ifq_mtx);
587 }
588 
589 /*
590  * Perform generic interface initialization tasks and attach the interface
591  * to the list of "active" interfaces.  If vmove flag is set on entry
592  * to if_attach_internal(), perform only a limited subset of initialization
593  * tasks, given that we are moving from one vnet to another an ifnet which
594  * has already been fully initialized.
595  *
596  * Note that if_detach_internal() removes group membership unconditionally
597  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
598  * Thus, when if_vmove() is applied to a cloned interface, group membership
599  * is lost while a cloned one always joins a group whose name is
600  * ifc->ifc_name.  To recover this after if_detach_internal() and
601  * if_attach_internal(), the cloner should be specified to
602  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
603  * attempts to join a group whose name is ifc->ifc_name.
604  *
605  * XXX:
606  *  - The decision to return void and thus require this function to
607  *    succeed is questionable.
608  *  - We should probably do more sanity checking.  For instance we don't
609  *    do anything to insure if_xname is unique or non-empty.
610  */
611 void
612 if_attach(struct ifnet *ifp)
613 {
614 
615 	if_attach_internal(ifp, 0, NULL);
616 }
617 
618 /*
619  * Compute the least common TSO limit.
620  */
621 void
622 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
623 {
624 	/*
625 	 * 1) If there is no limit currently, take the limit from
626 	 * the network adapter.
627 	 *
628 	 * 2) If the network adapter has a limit below the current
629 	 * limit, apply it.
630 	 */
631 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
632 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
633 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
634 	}
635 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
636 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
637 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
638 	}
639 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
640 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
641 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
642 	}
643 }
644 
645 /*
646  * Update TSO limit of a network adapter.
647  *
648  * Returns zero if no change. Else non-zero.
649  */
650 int
651 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
652 {
653 	int retval = 0;
654 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
655 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
656 		retval++;
657 	}
658 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
659 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
660 		retval++;
661 	}
662 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
663 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
664 		retval++;
665 	}
666 	return (retval);
667 }
668 
669 static void
670 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
671 {
672 	unsigned socksize, ifasize;
673 	int namelen, masklen;
674 	struct sockaddr_dl *sdl;
675 	struct ifaddr *ifa;
676 
677 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
678 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
679 		    ifp->if_xname);
680 
681 #ifdef VIMAGE
682 	ifp->if_vnet = curvnet;
683 	if (ifp->if_home_vnet == NULL)
684 		ifp->if_home_vnet = curvnet;
685 #endif
686 
687 	if_addgroup(ifp, IFG_ALL);
688 
689 	/* Restore group membership for cloned interfaces. */
690 	if (vmove && ifc != NULL)
691 		if_clone_addgroup(ifp, ifc);
692 
693 	getmicrotime(&ifp->if_lastchange);
694 	ifp->if_epoch = time_uptime;
695 
696 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
697 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
698 	    ("transmit and qflush must both either be set or both be NULL"));
699 	if (ifp->if_transmit == NULL) {
700 		ifp->if_transmit = if_transmit;
701 		ifp->if_qflush = if_qflush;
702 	}
703 	if (ifp->if_input == NULL)
704 		ifp->if_input = if_input_default;
705 
706 	if (ifp->if_requestencap == NULL)
707 		ifp->if_requestencap = if_requestencap_default;
708 
709 	if (!vmove) {
710 #ifdef MAC
711 		mac_ifnet_create(ifp);
712 #endif
713 
714 		/*
715 		 * Create a Link Level name for this device.
716 		 */
717 		namelen = strlen(ifp->if_xname);
718 		/*
719 		 * Always save enough space for any possiable name so we
720 		 * can do a rename in place later.
721 		 */
722 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
723 		socksize = masklen + ifp->if_addrlen;
724 		if (socksize < sizeof(*sdl))
725 			socksize = sizeof(*sdl);
726 		socksize = roundup2(socksize, sizeof(long));
727 		ifasize = sizeof(*ifa) + 2 * socksize;
728 		ifa = ifa_alloc(ifasize, M_WAITOK);
729 		sdl = (struct sockaddr_dl *)(ifa + 1);
730 		sdl->sdl_len = socksize;
731 		sdl->sdl_family = AF_LINK;
732 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
733 		sdl->sdl_nlen = namelen;
734 		sdl->sdl_index = ifp->if_index;
735 		sdl->sdl_type = ifp->if_type;
736 		ifp->if_addr = ifa;
737 		ifa->ifa_ifp = ifp;
738 		ifa->ifa_rtrequest = link_rtrequest;
739 		ifa->ifa_addr = (struct sockaddr *)sdl;
740 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
741 		ifa->ifa_netmask = (struct sockaddr *)sdl;
742 		sdl->sdl_len = masklen;
743 		while (namelen != 0)
744 			sdl->sdl_data[--namelen] = 0xff;
745 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
746 		/* Reliably crash if used uninitialized. */
747 		ifp->if_broadcastaddr = NULL;
748 
749 		if (ifp->if_type == IFT_ETHER) {
750 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
751 			    M_WAITOK | M_ZERO);
752 		}
753 
754 #if defined(INET) || defined(INET6)
755 		/* Use defaults for TSO, if nothing is set */
756 		if (ifp->if_hw_tsomax == 0 &&
757 		    ifp->if_hw_tsomaxsegcount == 0 &&
758 		    ifp->if_hw_tsomaxsegsize == 0) {
759 			/*
760 			 * The TSO defaults needs to be such that an
761 			 * NFS mbuf list of 35 mbufs totalling just
762 			 * below 64K works and that a chain of mbufs
763 			 * can be defragged into at most 32 segments:
764 			 */
765 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
766 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
767 			ifp->if_hw_tsomaxsegcount = 35;
768 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
769 
770 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
771 			if (ifp->if_capabilities & IFCAP_TSO) {
772 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
773 				    ifp->if_hw_tsomax,
774 				    ifp->if_hw_tsomaxsegcount,
775 				    ifp->if_hw_tsomaxsegsize);
776 			}
777 		}
778 #endif
779 	}
780 #ifdef VIMAGE
781 	else {
782 		/*
783 		 * Update the interface index in the link layer address
784 		 * of the interface.
785 		 */
786 		for (ifa = ifp->if_addr; ifa != NULL;
787 		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
788 			if (ifa->ifa_addr->sa_family == AF_LINK) {
789 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
790 				sdl->sdl_index = ifp->if_index;
791 			}
792 		}
793 	}
794 #endif
795 
796 	IFNET_WLOCK();
797 	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
798 #ifdef VIMAGE
799 	curvnet->vnet_ifcnt++;
800 #endif
801 	IFNET_WUNLOCK();
802 
803 	if (domain_init_status >= 2)
804 		if_attachdomain1(ifp);
805 
806 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
807 	if (IS_DEFAULT_VNET(curvnet))
808 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
809 
810 	/* Announce the interface. */
811 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
812 }
813 
814 static void
815 if_attachdomain(void *dummy)
816 {
817 	struct ifnet *ifp;
818 
819 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
820 		if_attachdomain1(ifp);
821 }
822 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
823     if_attachdomain, NULL);
824 
825 static void
826 if_attachdomain1(struct ifnet *ifp)
827 {
828 	struct domain *dp;
829 
830 	/*
831 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
832 	 * cannot lock ifp->if_afdata initialization, entirely.
833 	 */
834 	IF_AFDATA_LOCK(ifp);
835 	if (ifp->if_afdata_initialized >= domain_init_status) {
836 		IF_AFDATA_UNLOCK(ifp);
837 		log(LOG_WARNING, "%s called more than once on %s\n",
838 		    __func__, ifp->if_xname);
839 		return;
840 	}
841 	ifp->if_afdata_initialized = domain_init_status;
842 	IF_AFDATA_UNLOCK(ifp);
843 
844 	/* address family dependent data region */
845 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
846 	for (dp = domains; dp; dp = dp->dom_next) {
847 		if (dp->dom_ifattach)
848 			ifp->if_afdata[dp->dom_family] =
849 			    (*dp->dom_ifattach)(ifp);
850 	}
851 }
852 
853 /*
854  * Remove any unicast or broadcast network addresses from an interface.
855  */
856 void
857 if_purgeaddrs(struct ifnet *ifp)
858 {
859 	struct ifaddr *ifa, *next;
860 
861 	/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
862 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
863 		if (ifa->ifa_addr->sa_family == AF_LINK)
864 			continue;
865 #ifdef INET
866 		/* XXX: Ugly!! ad hoc just for INET */
867 		if (ifa->ifa_addr->sa_family == AF_INET) {
868 			struct ifaliasreq ifr;
869 
870 			bzero(&ifr, sizeof(ifr));
871 			ifr.ifra_addr = *ifa->ifa_addr;
872 			if (ifa->ifa_dstaddr)
873 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
874 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
875 			    NULL) == 0)
876 				continue;
877 		}
878 #endif /* INET */
879 #ifdef INET6
880 		if (ifa->ifa_addr->sa_family == AF_INET6) {
881 			in6_purgeaddr(ifa);
882 			/* ifp_addrhead is already updated */
883 			continue;
884 		}
885 #endif /* INET6 */
886 		IF_ADDR_WLOCK(ifp);
887 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
888 		IF_ADDR_WUNLOCK(ifp);
889 		ifa_free(ifa);
890 	}
891 }
892 
893 /*
894  * Remove any multicast network addresses from an interface when an ifnet
895  * is going away.
896  */
897 static void
898 if_purgemaddrs(struct ifnet *ifp)
899 {
900 	struct ifmultiaddr *ifma;
901 	struct ifmultiaddr *next;
902 
903 	IF_ADDR_WLOCK(ifp);
904 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
905 		if_delmulti_locked(ifp, ifma, 1);
906 	IF_ADDR_WUNLOCK(ifp);
907 }
908 
909 /*
910  * Detach an interface, removing it from the list of "active" interfaces.
911  * If vmove flag is set on entry to if_detach_internal(), perform only a
912  * limited subset of cleanup tasks, given that we are moving an ifnet from
913  * one vnet to another, where it must be fully operational.
914  *
915  * XXXRW: There are some significant questions about event ordering, and
916  * how to prevent things from starting to use the interface during detach.
917  */
918 void
919 if_detach(struct ifnet *ifp)
920 {
921 
922 	CURVNET_SET_QUIET(ifp->if_vnet);
923 	if_detach_internal(ifp, 0, NULL);
924 	CURVNET_RESTORE();
925 }
926 
927 /*
928  * The vmove flag, if set, indicates that we are called from a callpath
929  * that is moving an interface to a different vnet instance.
930  *
931  * The shutdown flag, if set, indicates that we are called in the
932  * process of shutting down a vnet instance.  Currently only the
933  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
934  * on a vnet instance shutdown without this flag being set, e.g., when
935  * the cloned interfaces are destoyed as first thing of teardown.
936  */
937 static int
938 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
939 {
940 	struct ifaddr *ifa;
941 	int i;
942 	struct domain *dp;
943  	struct ifnet *iter;
944  	int found = 0;
945 #ifdef VIMAGE
946 	int shutdown;
947 
948 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
949 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
950 #endif
951 	IFNET_WLOCK();
952 	TAILQ_FOREACH(iter, &V_ifnet, if_link)
953 		if (iter == ifp) {
954 			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
955 			found = 1;
956 			break;
957 		}
958 	IFNET_WUNLOCK();
959 	if (!found) {
960 		/*
961 		 * While we would want to panic here, we cannot
962 		 * guarantee that the interface is indeed still on
963 		 * the list given we don't hold locks all the way.
964 		 */
965 		return (ENOENT);
966 #if 0
967 		if (vmove)
968 			panic("%s: ifp=%p not on the ifnet tailq %p",
969 			    __func__, ifp, &V_ifnet);
970 		else
971 			return; /* XXX this should panic as well? */
972 #endif
973 	}
974 
975 	/*
976 	 * At this point we know the interface still was on the ifnet list
977 	 * and we removed it so we are in a stable state.
978 	 */
979 #ifdef VIMAGE
980 	curvnet->vnet_ifcnt--;
981 #endif
982 
983 	/*
984 	 * In any case (destroy or vmove) detach us from the groups
985 	 * and remove/wait for pending events on the taskq.
986 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
987 	 */
988 	if_delgroups(ifp);
989 
990 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
991 
992 	/*
993 	 * Check if this is a cloned interface or not. Must do even if
994 	 * shutting down as a if_vmove_reclaim() would move the ifp and
995 	 * the if_clone_addgroup() will have a corrupted string overwise
996 	 * from a gibberish pointer.
997 	 */
998 	if (vmove && ifcp != NULL)
999 		*ifcp = if_clone_findifc(ifp);
1000 
1001 	if_down(ifp);
1002 
1003 #ifdef VIMAGE
1004 	/*
1005 	 * On VNET shutdown abort here as the stack teardown will do all
1006 	 * the work top-down for us.
1007 	 */
1008 	if (shutdown) {
1009 		/*
1010 		 * In case of a vmove we are done here without error.
1011 		 * If we would signal an error it would lead to the same
1012 		 * abort as if we did not find the ifnet anymore.
1013 		 * if_detach() calls us in void context and does not care
1014 		 * about an early abort notification, so life is splendid :)
1015 		 */
1016 		goto finish_vnet_shutdown;
1017 	}
1018 #endif
1019 
1020 	/*
1021 	 * At this point we are not tearing down a VNET and are either
1022 	 * going to destroy or vmove the interface and have to cleanup
1023 	 * accordingly.
1024 	 */
1025 
1026 	/*
1027 	 * Remove routes and flush queues.
1028 	 */
1029 #ifdef ALTQ
1030 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1031 		altq_disable(&ifp->if_snd);
1032 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1033 		altq_detach(&ifp->if_snd);
1034 #endif
1035 
1036 	if_purgeaddrs(ifp);
1037 
1038 #ifdef INET
1039 	in_ifdetach(ifp);
1040 #endif
1041 
1042 #ifdef INET6
1043 	/*
1044 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1045 	 * before removing routing entries below, since IPv6 interface direct
1046 	 * routes are expected to be removed by the IPv6-specific kernel API.
1047 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1048 	 */
1049 	in6_ifdetach(ifp);
1050 #endif
1051 	if_purgemaddrs(ifp);
1052 
1053 	/* Announce that the interface is gone. */
1054 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1055 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1056 	if (IS_DEFAULT_VNET(curvnet))
1057 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1058 
1059 	if (!vmove) {
1060 		/*
1061 		 * Prevent further calls into the device driver via ifnet.
1062 		 */
1063 		if_dead(ifp);
1064 
1065 		/*
1066 		 * Remove link ifaddr pointer and maybe decrement if_index.
1067 		 * Clean up all addresses.
1068 		 */
1069 		free(ifp->if_hw_addr, M_IFADDR);
1070 		ifp->if_hw_addr = NULL;
1071 		ifp->if_addr = NULL;
1072 
1073 		/* We can now free link ifaddr. */
1074 		IF_ADDR_WLOCK(ifp);
1075 		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1076 			ifa = TAILQ_FIRST(&ifp->if_addrhead);
1077 			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1078 			IF_ADDR_WUNLOCK(ifp);
1079 			ifa_free(ifa);
1080 		} else
1081 			IF_ADDR_WUNLOCK(ifp);
1082 	}
1083 
1084 	rt_flushifroutes(ifp);
1085 
1086 #ifdef VIMAGE
1087 finish_vnet_shutdown:
1088 #endif
1089 	/*
1090 	 * We cannot hold the lock over dom_ifdetach calls as they might
1091 	 * sleep, for example trying to drain a callout, thus open up the
1092 	 * theoretical race with re-attaching.
1093 	 */
1094 	IF_AFDATA_LOCK(ifp);
1095 	i = ifp->if_afdata_initialized;
1096 	ifp->if_afdata_initialized = 0;
1097 	IF_AFDATA_UNLOCK(ifp);
1098 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1099 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1100 			(*dp->dom_ifdetach)(ifp,
1101 			    ifp->if_afdata[dp->dom_family]);
1102 			ifp->if_afdata[dp->dom_family] = NULL;
1103 		}
1104 	}
1105 
1106 	return (0);
1107 }
1108 
1109 #ifdef VIMAGE
1110 /*
1111  * if_vmove() performs a limited version of if_detach() in current
1112  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1113  * An attempt is made to shrink if_index in current vnet, find an
1114  * unused if_index in target vnet and calls if_grow() if necessary,
1115  * and finally find an unused if_xname for the target vnet.
1116  */
1117 static void
1118 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1119 {
1120 	struct if_clone *ifc;
1121 	u_int bif_dlt, bif_hdrlen;
1122 	int rc;
1123 
1124  	/*
1125 	 * if_detach_internal() will call the eventhandler to notify
1126 	 * interface departure.  That will detach if_bpf.  We need to
1127 	 * safe the dlt and hdrlen so we can re-attach it later.
1128 	 */
1129 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1130 
1131 	/*
1132 	 * Detach from current vnet, but preserve LLADDR info, do not
1133 	 * mark as dead etc. so that the ifnet can be reattached later.
1134 	 * If we cannot find it, we lost the race to someone else.
1135 	 */
1136 	rc = if_detach_internal(ifp, 1, &ifc);
1137 	if (rc != 0)
1138 		return;
1139 
1140 	/*
1141 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1142 	 * the if_index for that vnet if possible.
1143 	 *
1144 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1145 	 * or we'd lock on one vnet and unlock on another.
1146 	 */
1147 	IFNET_WLOCK();
1148 	ifindex_free_locked(ifp->if_index);
1149 	IFNET_WUNLOCK();
1150 
1151 	/*
1152 	 * Perform interface-specific reassignment tasks, if provided by
1153 	 * the driver.
1154 	 */
1155 	if (ifp->if_reassign != NULL)
1156 		ifp->if_reassign(ifp, new_vnet, NULL);
1157 
1158 	/*
1159 	 * Switch to the context of the target vnet.
1160 	 */
1161 	CURVNET_SET_QUIET(new_vnet);
1162 
1163 	IFNET_WLOCK();
1164 	ifp->if_index = ifindex_alloc();
1165 	ifnet_setbyindex_locked(ifp->if_index, ifp);
1166 	IFNET_WUNLOCK();
1167 
1168 	if_attach_internal(ifp, 1, ifc);
1169 
1170 	if (ifp->if_bpf == NULL)
1171 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1172 
1173 	CURVNET_RESTORE();
1174 }
1175 
1176 /*
1177  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1178  */
1179 static int
1180 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1181 {
1182 	struct prison *pr;
1183 	struct ifnet *difp;
1184 	int shutdown;
1185 
1186 	/* Try to find the prison within our visibility. */
1187 	sx_slock(&allprison_lock);
1188 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1189 	sx_sunlock(&allprison_lock);
1190 	if (pr == NULL)
1191 		return (ENXIO);
1192 	prison_hold_locked(pr);
1193 	mtx_unlock(&pr->pr_mtx);
1194 
1195 	/* Do not try to move the iface from and to the same prison. */
1196 	if (pr->pr_vnet == ifp->if_vnet) {
1197 		prison_free(pr);
1198 		return (EEXIST);
1199 	}
1200 
1201 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1202 	/* XXX Lock interfaces to avoid races. */
1203 	CURVNET_SET_QUIET(pr->pr_vnet);
1204 	difp = ifunit(ifname);
1205 	if (difp != NULL) {
1206 		CURVNET_RESTORE();
1207 		prison_free(pr);
1208 		return (EEXIST);
1209 	}
1210 
1211 	/* Make sure the VNET is stable. */
1212 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1213 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1214 	if (shutdown) {
1215 		CURVNET_RESTORE();
1216 		prison_free(pr);
1217 		return (EBUSY);
1218 	}
1219 	CURVNET_RESTORE();
1220 
1221 	/* Move the interface into the child jail/vnet. */
1222 	if_vmove(ifp, pr->pr_vnet);
1223 
1224 	/* Report the new if_xname back to the userland. */
1225 	sprintf(ifname, "%s", ifp->if_xname);
1226 
1227 	prison_free(pr);
1228 	return (0);
1229 }
1230 
1231 static int
1232 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1233 {
1234 	struct prison *pr;
1235 	struct vnet *vnet_dst;
1236 	struct ifnet *ifp;
1237  	int shutdown;
1238 
1239 	/* Try to find the prison within our visibility. */
1240 	sx_slock(&allprison_lock);
1241 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1242 	sx_sunlock(&allprison_lock);
1243 	if (pr == NULL)
1244 		return (ENXIO);
1245 	prison_hold_locked(pr);
1246 	mtx_unlock(&pr->pr_mtx);
1247 
1248 	/* Make sure the named iface exists in the source prison/vnet. */
1249 	CURVNET_SET(pr->pr_vnet);
1250 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1251 	if (ifp == NULL) {
1252 		CURVNET_RESTORE();
1253 		prison_free(pr);
1254 		return (ENXIO);
1255 	}
1256 
1257 	/* Do not try to move the iface from and to the same prison. */
1258 	vnet_dst = TD_TO_VNET(td);
1259 	if (vnet_dst == ifp->if_vnet) {
1260 		CURVNET_RESTORE();
1261 		prison_free(pr);
1262 		return (EEXIST);
1263 	}
1264 
1265 	/* Make sure the VNET is stable. */
1266 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1267 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1268 	if (shutdown) {
1269 		CURVNET_RESTORE();
1270 		prison_free(pr);
1271 		return (EBUSY);
1272 	}
1273 
1274 	/* Get interface back from child jail/vnet. */
1275 	if_vmove(ifp, vnet_dst);
1276 	CURVNET_RESTORE();
1277 
1278 	/* Report the new if_xname back to the userland. */
1279 	sprintf(ifname, "%s", ifp->if_xname);
1280 
1281 	prison_free(pr);
1282 	return (0);
1283 }
1284 #endif /* VIMAGE */
1285 
1286 /*
1287  * Add a group to an interface
1288  */
1289 int
1290 if_addgroup(struct ifnet *ifp, const char *groupname)
1291 {
1292 	struct ifg_list		*ifgl;
1293 	struct ifg_group	*ifg = NULL;
1294 	struct ifg_member	*ifgm;
1295 	int 			 new = 0;
1296 
1297 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1298 	    groupname[strlen(groupname) - 1] <= '9')
1299 		return (EINVAL);
1300 
1301 	IFNET_WLOCK();
1302 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1303 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1304 			IFNET_WUNLOCK();
1305 			return (EEXIST);
1306 		}
1307 
1308 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1309 	    M_NOWAIT)) == NULL) {
1310 	    	IFNET_WUNLOCK();
1311 		return (ENOMEM);
1312 	}
1313 
1314 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1315 	    M_TEMP, M_NOWAIT)) == NULL) {
1316 		free(ifgl, M_TEMP);
1317 		IFNET_WUNLOCK();
1318 		return (ENOMEM);
1319 	}
1320 
1321 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1322 		if (!strcmp(ifg->ifg_group, groupname))
1323 			break;
1324 
1325 	if (ifg == NULL) {
1326 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1327 		    M_TEMP, M_NOWAIT)) == NULL) {
1328 			free(ifgl, M_TEMP);
1329 			free(ifgm, M_TEMP);
1330 			IFNET_WUNLOCK();
1331 			return (ENOMEM);
1332 		}
1333 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1334 		ifg->ifg_refcnt = 0;
1335 		TAILQ_INIT(&ifg->ifg_members);
1336 		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1337 		new = 1;
1338 	}
1339 
1340 	ifg->ifg_refcnt++;
1341 	ifgl->ifgl_group = ifg;
1342 	ifgm->ifgm_ifp = ifp;
1343 
1344 	IF_ADDR_WLOCK(ifp);
1345 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1346 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1347 	IF_ADDR_WUNLOCK(ifp);
1348 
1349 	IFNET_WUNLOCK();
1350 
1351 	if (new)
1352 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1353 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1354 
1355 	return (0);
1356 }
1357 
1358 /*
1359  * Remove a group from an interface
1360  */
1361 int
1362 if_delgroup(struct ifnet *ifp, const char *groupname)
1363 {
1364 	struct ifg_list		*ifgl;
1365 	struct ifg_member	*ifgm;
1366 
1367 	IFNET_WLOCK();
1368 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1369 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1370 			break;
1371 	if (ifgl == NULL) {
1372 		IFNET_WUNLOCK();
1373 		return (ENOENT);
1374 	}
1375 
1376 	IF_ADDR_WLOCK(ifp);
1377 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1378 	IF_ADDR_WUNLOCK(ifp);
1379 
1380 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1381 		if (ifgm->ifgm_ifp == ifp)
1382 			break;
1383 
1384 	if (ifgm != NULL) {
1385 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1386 		free(ifgm, M_TEMP);
1387 	}
1388 
1389 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1390 		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1391 		IFNET_WUNLOCK();
1392 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1393 		free(ifgl->ifgl_group, M_TEMP);
1394 	} else
1395 		IFNET_WUNLOCK();
1396 
1397 	free(ifgl, M_TEMP);
1398 
1399 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1400 
1401 	return (0);
1402 }
1403 
1404 /*
1405  * Remove an interface from all groups
1406  */
1407 static void
1408 if_delgroups(struct ifnet *ifp)
1409 {
1410 	struct ifg_list		*ifgl;
1411 	struct ifg_member	*ifgm;
1412 	char groupname[IFNAMSIZ];
1413 
1414 	IFNET_WLOCK();
1415 	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1416 		ifgl = TAILQ_FIRST(&ifp->if_groups);
1417 
1418 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1419 
1420 		IF_ADDR_WLOCK(ifp);
1421 		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1422 		IF_ADDR_WUNLOCK(ifp);
1423 
1424 		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1425 			if (ifgm->ifgm_ifp == ifp)
1426 				break;
1427 
1428 		if (ifgm != NULL) {
1429 			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1430 			    ifgm_next);
1431 			free(ifgm, M_TEMP);
1432 		}
1433 
1434 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1435 			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1436 			IFNET_WUNLOCK();
1437 			EVENTHANDLER_INVOKE(group_detach_event,
1438 			    ifgl->ifgl_group);
1439 			free(ifgl->ifgl_group, M_TEMP);
1440 		} else
1441 			IFNET_WUNLOCK();
1442 
1443 		free(ifgl, M_TEMP);
1444 
1445 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1446 
1447 		IFNET_WLOCK();
1448 	}
1449 	IFNET_WUNLOCK();
1450 }
1451 
1452 /*
1453  * Stores all groups from an interface in memory pointed
1454  * to by data
1455  */
1456 static int
1457 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1458 {
1459 	int			 len, error;
1460 	struct ifg_list		*ifgl;
1461 	struct ifg_req		 ifgrq, *ifgp;
1462 	struct ifgroupreq	*ifgr = data;
1463 
1464 	if (ifgr->ifgr_len == 0) {
1465 		IF_ADDR_RLOCK(ifp);
1466 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1467 			ifgr->ifgr_len += sizeof(struct ifg_req);
1468 		IF_ADDR_RUNLOCK(ifp);
1469 		return (0);
1470 	}
1471 
1472 	len = ifgr->ifgr_len;
1473 	ifgp = ifgr->ifgr_groups;
1474 	/* XXX: wire */
1475 	IF_ADDR_RLOCK(ifp);
1476 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1477 		if (len < sizeof(ifgrq)) {
1478 			IF_ADDR_RUNLOCK(ifp);
1479 			return (EINVAL);
1480 		}
1481 		bzero(&ifgrq, sizeof ifgrq);
1482 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1483 		    sizeof(ifgrq.ifgrq_group));
1484 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1485 		    	IF_ADDR_RUNLOCK(ifp);
1486 			return (error);
1487 		}
1488 		len -= sizeof(ifgrq);
1489 		ifgp++;
1490 	}
1491 	IF_ADDR_RUNLOCK(ifp);
1492 
1493 	return (0);
1494 }
1495 
1496 /*
1497  * Stores all members of a group in memory pointed to by data
1498  */
1499 static int
1500 if_getgroupmembers(struct ifgroupreq *data)
1501 {
1502 	struct ifgroupreq	*ifgr = data;
1503 	struct ifg_group	*ifg;
1504 	struct ifg_member	*ifgm;
1505 	struct ifg_req		 ifgrq, *ifgp;
1506 	int			 len, error;
1507 
1508 	IFNET_RLOCK();
1509 	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1510 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1511 			break;
1512 	if (ifg == NULL) {
1513 		IFNET_RUNLOCK();
1514 		return (ENOENT);
1515 	}
1516 
1517 	if (ifgr->ifgr_len == 0) {
1518 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1519 			ifgr->ifgr_len += sizeof(ifgrq);
1520 		IFNET_RUNLOCK();
1521 		return (0);
1522 	}
1523 
1524 	len = ifgr->ifgr_len;
1525 	ifgp = ifgr->ifgr_groups;
1526 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1527 		if (len < sizeof(ifgrq)) {
1528 			IFNET_RUNLOCK();
1529 			return (EINVAL);
1530 		}
1531 		bzero(&ifgrq, sizeof ifgrq);
1532 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1533 		    sizeof(ifgrq.ifgrq_member));
1534 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1535 			IFNET_RUNLOCK();
1536 			return (error);
1537 		}
1538 		len -= sizeof(ifgrq);
1539 		ifgp++;
1540 	}
1541 	IFNET_RUNLOCK();
1542 
1543 	return (0);
1544 }
1545 
1546 /*
1547  * Return counter values from counter(9)s stored in ifnet.
1548  */
1549 uint64_t
1550 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1551 {
1552 
1553 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1554 
1555 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1556 }
1557 
1558 /*
1559  * Increase an ifnet counter. Usually used for counters shared
1560  * between the stack and a driver, but function supports them all.
1561  */
1562 void
1563 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1564 {
1565 
1566 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1567 
1568 	counter_u64_add(ifp->if_counters[cnt], inc);
1569 }
1570 
1571 /*
1572  * Copy data from ifnet to userland API structure if_data.
1573  */
1574 void
1575 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1576 {
1577 
1578 	ifd->ifi_type = ifp->if_type;
1579 	ifd->ifi_physical = 0;
1580 	ifd->ifi_addrlen = ifp->if_addrlen;
1581 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1582 	ifd->ifi_link_state = ifp->if_link_state;
1583 	ifd->ifi_vhid = 0;
1584 	ifd->ifi_datalen = sizeof(struct if_data);
1585 	ifd->ifi_mtu = ifp->if_mtu;
1586 	ifd->ifi_metric = ifp->if_metric;
1587 	ifd->ifi_baudrate = ifp->if_baudrate;
1588 	ifd->ifi_hwassist = ifp->if_hwassist;
1589 	ifd->ifi_epoch = ifp->if_epoch;
1590 	ifd->ifi_lastchange = ifp->if_lastchange;
1591 
1592 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1593 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1594 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1595 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1596 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1597 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1598 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1599 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1600 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1601 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1602 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1603 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1604 }
1605 
1606 /*
1607  * Wrapper functions for struct ifnet address list locking macros.  These are
1608  * used by kernel modules to avoid encoding programming interface or binary
1609  * interface assumptions that may be violated when kernel-internal locking
1610  * approaches change.
1611  */
1612 void
1613 if_addr_rlock(struct ifnet *ifp)
1614 {
1615 
1616 	IF_ADDR_RLOCK(ifp);
1617 }
1618 
1619 void
1620 if_addr_runlock(struct ifnet *ifp)
1621 {
1622 
1623 	IF_ADDR_RUNLOCK(ifp);
1624 }
1625 
1626 void
1627 if_maddr_rlock(if_t ifp)
1628 {
1629 
1630 	IF_ADDR_RLOCK((struct ifnet *)ifp);
1631 }
1632 
1633 void
1634 if_maddr_runlock(if_t ifp)
1635 {
1636 
1637 	IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1638 }
1639 
1640 /*
1641  * Initialization, destruction and refcounting functions for ifaddrs.
1642  */
1643 struct ifaddr *
1644 ifa_alloc(size_t size, int flags)
1645 {
1646 	struct ifaddr *ifa;
1647 
1648 	KASSERT(size >= sizeof(struct ifaddr),
1649 	    ("%s: invalid size %zu", __func__, size));
1650 
1651 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1652 	if (ifa == NULL)
1653 		return (NULL);
1654 
1655 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1656 		goto fail;
1657 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1658 		goto fail;
1659 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1660 		goto fail;
1661 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1662 		goto fail;
1663 
1664 	refcount_init(&ifa->ifa_refcnt, 1);
1665 
1666 	return (ifa);
1667 
1668 fail:
1669 	/* free(NULL) is okay */
1670 	counter_u64_free(ifa->ifa_opackets);
1671 	counter_u64_free(ifa->ifa_ipackets);
1672 	counter_u64_free(ifa->ifa_obytes);
1673 	counter_u64_free(ifa->ifa_ibytes);
1674 	free(ifa, M_IFADDR);
1675 
1676 	return (NULL);
1677 }
1678 
1679 void
1680 ifa_ref(struct ifaddr *ifa)
1681 {
1682 
1683 	refcount_acquire(&ifa->ifa_refcnt);
1684 }
1685 
1686 void
1687 ifa_free(struct ifaddr *ifa)
1688 {
1689 
1690 	if (refcount_release(&ifa->ifa_refcnt)) {
1691 		counter_u64_free(ifa->ifa_opackets);
1692 		counter_u64_free(ifa->ifa_ipackets);
1693 		counter_u64_free(ifa->ifa_obytes);
1694 		counter_u64_free(ifa->ifa_ibytes);
1695 		free(ifa, M_IFADDR);
1696 	}
1697 }
1698 
1699 static int
1700 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1701     struct sockaddr *ia)
1702 {
1703 	int error;
1704 	struct rt_addrinfo info;
1705 	struct sockaddr_dl null_sdl;
1706 	struct ifnet *ifp;
1707 
1708 	ifp = ifa->ifa_ifp;
1709 
1710 	bzero(&info, sizeof(info));
1711 	if (cmd != RTM_DELETE)
1712 		info.rti_ifp = V_loif;
1713 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1714 	info.rti_info[RTAX_DST] = ia;
1715 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1716 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1717 
1718 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1719 
1720 	if (error != 0)
1721 		log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1722 		    __func__, otype, if_name(ifp), error);
1723 
1724 	return (error);
1725 }
1726 
1727 int
1728 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1729 {
1730 
1731 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1732 }
1733 
1734 int
1735 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1736 {
1737 
1738 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1739 }
1740 
1741 int
1742 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1743 {
1744 
1745 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1746 }
1747 
1748 /*
1749  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1750  * structs used to represent other address families, it is necessary
1751  * to perform a different comparison.
1752  */
1753 
1754 #define	sa_dl_equal(a1, a2)	\
1755 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1756 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1757 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1758 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1759 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1760 
1761 /*
1762  * Locate an interface based on a complete address.
1763  */
1764 /*ARGSUSED*/
1765 static struct ifaddr *
1766 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1767 {
1768 	struct ifnet *ifp;
1769 	struct ifaddr *ifa;
1770 
1771 	IFNET_RLOCK_NOSLEEP();
1772 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1773 		IF_ADDR_RLOCK(ifp);
1774 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1775 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1776 				continue;
1777 			if (sa_equal(addr, ifa->ifa_addr)) {
1778 				if (getref)
1779 					ifa_ref(ifa);
1780 				IF_ADDR_RUNLOCK(ifp);
1781 				goto done;
1782 			}
1783 			/* IP6 doesn't have broadcast */
1784 			if ((ifp->if_flags & IFF_BROADCAST) &&
1785 			    ifa->ifa_broadaddr &&
1786 			    ifa->ifa_broadaddr->sa_len != 0 &&
1787 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1788 				if (getref)
1789 					ifa_ref(ifa);
1790 				IF_ADDR_RUNLOCK(ifp);
1791 				goto done;
1792 			}
1793 		}
1794 		IF_ADDR_RUNLOCK(ifp);
1795 	}
1796 	ifa = NULL;
1797 done:
1798 	IFNET_RUNLOCK_NOSLEEP();
1799 	return (ifa);
1800 }
1801 
1802 struct ifaddr *
1803 ifa_ifwithaddr(const struct sockaddr *addr)
1804 {
1805 
1806 	return (ifa_ifwithaddr_internal(addr, 1));
1807 }
1808 
1809 int
1810 ifa_ifwithaddr_check(const struct sockaddr *addr)
1811 {
1812 
1813 	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1814 }
1815 
1816 /*
1817  * Locate an interface based on the broadcast address.
1818  */
1819 /* ARGSUSED */
1820 struct ifaddr *
1821 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1822 {
1823 	struct ifnet *ifp;
1824 	struct ifaddr *ifa;
1825 
1826 	IFNET_RLOCK_NOSLEEP();
1827 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1828 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1829 			continue;
1830 		IF_ADDR_RLOCK(ifp);
1831 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1832 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1833 				continue;
1834 			if ((ifp->if_flags & IFF_BROADCAST) &&
1835 			    ifa->ifa_broadaddr &&
1836 			    ifa->ifa_broadaddr->sa_len != 0 &&
1837 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1838 				ifa_ref(ifa);
1839 				IF_ADDR_RUNLOCK(ifp);
1840 				goto done;
1841 			}
1842 		}
1843 		IF_ADDR_RUNLOCK(ifp);
1844 	}
1845 	ifa = NULL;
1846 done:
1847 	IFNET_RUNLOCK_NOSLEEP();
1848 	return (ifa);
1849 }
1850 
1851 /*
1852  * Locate the point to point interface with a given destination address.
1853  */
1854 /*ARGSUSED*/
1855 struct ifaddr *
1856 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1857 {
1858 	struct ifnet *ifp;
1859 	struct ifaddr *ifa;
1860 
1861 	IFNET_RLOCK_NOSLEEP();
1862 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1863 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1864 			continue;
1865 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1866 			continue;
1867 		IF_ADDR_RLOCK(ifp);
1868 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1869 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1870 				continue;
1871 			if (ifa->ifa_dstaddr != NULL &&
1872 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1873 				ifa_ref(ifa);
1874 				IF_ADDR_RUNLOCK(ifp);
1875 				goto done;
1876 			}
1877 		}
1878 		IF_ADDR_RUNLOCK(ifp);
1879 	}
1880 	ifa = NULL;
1881 done:
1882 	IFNET_RUNLOCK_NOSLEEP();
1883 	return (ifa);
1884 }
1885 
1886 /*
1887  * Find an interface on a specific network.  If many, choice
1888  * is most specific found.
1889  */
1890 struct ifaddr *
1891 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1892 {
1893 	struct ifnet *ifp;
1894 	struct ifaddr *ifa;
1895 	struct ifaddr *ifa_maybe = NULL;
1896 	u_int af = addr->sa_family;
1897 	const char *addr_data = addr->sa_data, *cplim;
1898 
1899 	/*
1900 	 * AF_LINK addresses can be looked up directly by their index number,
1901 	 * so do that if we can.
1902 	 */
1903 	if (af == AF_LINK) {
1904 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1905 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1906 		return (ifaddr_byindex(sdl->sdl_index));
1907 	}
1908 
1909 	/*
1910 	 * Scan though each interface, looking for ones that have addresses
1911 	 * in this address family and the requested fib.  Maintain a reference
1912 	 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1913 	 * kept it stable when we move onto the next interface.
1914 	 */
1915 	IFNET_RLOCK_NOSLEEP();
1916 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1917 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1918 			continue;
1919 		IF_ADDR_RLOCK(ifp);
1920 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1921 			const char *cp, *cp2, *cp3;
1922 
1923 			if (ifa->ifa_addr->sa_family != af)
1924 next:				continue;
1925 			if (af == AF_INET &&
1926 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1927 				/*
1928 				 * This is a bit broken as it doesn't
1929 				 * take into account that the remote end may
1930 				 * be a single node in the network we are
1931 				 * looking for.
1932 				 * The trouble is that we don't know the
1933 				 * netmask for the remote end.
1934 				 */
1935 				if (ifa->ifa_dstaddr != NULL &&
1936 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1937 					ifa_ref(ifa);
1938 					IF_ADDR_RUNLOCK(ifp);
1939 					goto done;
1940 				}
1941 			} else {
1942 				/*
1943 				 * Scan all the bits in the ifa's address.
1944 				 * If a bit dissagrees with what we are
1945 				 * looking for, mask it with the netmask
1946 				 * to see if it really matters.
1947 				 * (A byte at a time)
1948 				 */
1949 				if (ifa->ifa_netmask == 0)
1950 					continue;
1951 				cp = addr_data;
1952 				cp2 = ifa->ifa_addr->sa_data;
1953 				cp3 = ifa->ifa_netmask->sa_data;
1954 				cplim = ifa->ifa_netmask->sa_len
1955 					+ (char *)ifa->ifa_netmask;
1956 				while (cp3 < cplim)
1957 					if ((*cp++ ^ *cp2++) & *cp3++)
1958 						goto next; /* next address! */
1959 				/*
1960 				 * If the netmask of what we just found
1961 				 * is more specific than what we had before
1962 				 * (if we had one), or if the virtual status
1963 				 * of new prefix is better than of the old one,
1964 				 * then remember the new one before continuing
1965 				 * to search for an even better one.
1966 				 */
1967 				if (ifa_maybe == NULL ||
1968 				    ifa_preferred(ifa_maybe, ifa) ||
1969 				    rn_refines((caddr_t)ifa->ifa_netmask,
1970 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1971 					if (ifa_maybe != NULL)
1972 						ifa_free(ifa_maybe);
1973 					ifa_maybe = ifa;
1974 					ifa_ref(ifa_maybe);
1975 				}
1976 			}
1977 		}
1978 		IF_ADDR_RUNLOCK(ifp);
1979 	}
1980 	ifa = ifa_maybe;
1981 	ifa_maybe = NULL;
1982 done:
1983 	IFNET_RUNLOCK_NOSLEEP();
1984 	if (ifa_maybe != NULL)
1985 		ifa_free(ifa_maybe);
1986 	return (ifa);
1987 }
1988 
1989 /*
1990  * Find an interface address specific to an interface best matching
1991  * a given address.
1992  */
1993 struct ifaddr *
1994 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1995 {
1996 	struct ifaddr *ifa;
1997 	const char *cp, *cp2, *cp3;
1998 	char *cplim;
1999 	struct ifaddr *ifa_maybe = NULL;
2000 	u_int af = addr->sa_family;
2001 
2002 	if (af >= AF_MAX)
2003 		return (NULL);
2004 	IF_ADDR_RLOCK(ifp);
2005 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2006 		if (ifa->ifa_addr->sa_family != af)
2007 			continue;
2008 		if (ifa_maybe == NULL)
2009 			ifa_maybe = ifa;
2010 		if (ifa->ifa_netmask == 0) {
2011 			if (sa_equal(addr, ifa->ifa_addr) ||
2012 			    (ifa->ifa_dstaddr &&
2013 			    sa_equal(addr, ifa->ifa_dstaddr)))
2014 				goto done;
2015 			continue;
2016 		}
2017 		if (ifp->if_flags & IFF_POINTOPOINT) {
2018 			if (sa_equal(addr, ifa->ifa_dstaddr))
2019 				goto done;
2020 		} else {
2021 			cp = addr->sa_data;
2022 			cp2 = ifa->ifa_addr->sa_data;
2023 			cp3 = ifa->ifa_netmask->sa_data;
2024 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2025 			for (; cp3 < cplim; cp3++)
2026 				if ((*cp++ ^ *cp2++) & *cp3)
2027 					break;
2028 			if (cp3 == cplim)
2029 				goto done;
2030 		}
2031 	}
2032 	ifa = ifa_maybe;
2033 done:
2034 	if (ifa != NULL)
2035 		ifa_ref(ifa);
2036 	IF_ADDR_RUNLOCK(ifp);
2037 	return (ifa);
2038 }
2039 
2040 /*
2041  * See whether new ifa is better than current one:
2042  * 1) A non-virtual one is preferred over virtual.
2043  * 2) A virtual in master state preferred over any other state.
2044  *
2045  * Used in several address selecting functions.
2046  */
2047 int
2048 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2049 {
2050 
2051 	return (cur->ifa_carp && (!next->ifa_carp ||
2052 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2053 }
2054 
2055 #include <net/if_llatbl.h>
2056 
2057 /*
2058  * Default action when installing a route with a Link Level gateway.
2059  * Lookup an appropriate real ifa to point to.
2060  * This should be moved to /sys/net/link.c eventually.
2061  */
2062 static void
2063 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2064 {
2065 	struct ifaddr *ifa, *oifa;
2066 	struct sockaddr *dst;
2067 	struct ifnet *ifp;
2068 
2069 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2070 	    ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2071 		return;
2072 	ifa = ifaof_ifpforaddr(dst, ifp);
2073 	if (ifa) {
2074 		oifa = rt->rt_ifa;
2075 		rt->rt_ifa = ifa;
2076 		ifa_free(oifa);
2077 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2078 			ifa->ifa_rtrequest(cmd, rt, info);
2079 	}
2080 }
2081 
2082 struct sockaddr_dl *
2083 link_alloc_sdl(size_t size, int flags)
2084 {
2085 
2086 	return (malloc(size, M_TEMP, flags));
2087 }
2088 
2089 void
2090 link_free_sdl(struct sockaddr *sa)
2091 {
2092 	free(sa, M_TEMP);
2093 }
2094 
2095 /*
2096  * Fills in given sdl with interface basic info.
2097  * Returns pointer to filled sdl.
2098  */
2099 struct sockaddr_dl *
2100 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2101 {
2102 	struct sockaddr_dl *sdl;
2103 
2104 	sdl = (struct sockaddr_dl *)paddr;
2105 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2106 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2107 	sdl->sdl_family = AF_LINK;
2108 	sdl->sdl_index = ifp->if_index;
2109 	sdl->sdl_type = iftype;
2110 
2111 	return (sdl);
2112 }
2113 
2114 /*
2115  * Mark an interface down and notify protocols of
2116  * the transition.
2117  */
2118 static void
2119 if_unroute(struct ifnet *ifp, int flag, int fam)
2120 {
2121 	struct ifaddr *ifa;
2122 
2123 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2124 
2125 	ifp->if_flags &= ~flag;
2126 	getmicrotime(&ifp->if_lastchange);
2127 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2128 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2129 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2130 	ifp->if_qflush(ifp);
2131 
2132 	if (ifp->if_carp)
2133 		(*carp_linkstate_p)(ifp);
2134 	rt_ifmsg(ifp);
2135 }
2136 
2137 /*
2138  * Mark an interface up and notify protocols of
2139  * the transition.
2140  */
2141 static void
2142 if_route(struct ifnet *ifp, int flag, int fam)
2143 {
2144 	struct ifaddr *ifa;
2145 
2146 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2147 
2148 	ifp->if_flags |= flag;
2149 	getmicrotime(&ifp->if_lastchange);
2150 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2151 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2152 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2153 	if (ifp->if_carp)
2154 		(*carp_linkstate_p)(ifp);
2155 	rt_ifmsg(ifp);
2156 #ifdef INET6
2157 	in6_if_up(ifp);
2158 #endif
2159 }
2160 
2161 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2162 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2163 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2164 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2165 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2166 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2167 void	*(*vlan_cookie_p)(struct ifnet *);
2168 
2169 /*
2170  * Handle a change in the interface link state. To avoid LORs
2171  * between driver lock and upper layer locks, as well as possible
2172  * recursions, we post event to taskqueue, and all job
2173  * is done in static do_link_state_change().
2174  */
2175 void
2176 if_link_state_change(struct ifnet *ifp, int link_state)
2177 {
2178 	/* Return if state hasn't changed. */
2179 	if (ifp->if_link_state == link_state)
2180 		return;
2181 
2182 	ifp->if_link_state = link_state;
2183 
2184 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2185 }
2186 
2187 static void
2188 do_link_state_change(void *arg, int pending)
2189 {
2190 	struct ifnet *ifp = (struct ifnet *)arg;
2191 	int link_state = ifp->if_link_state;
2192 	CURVNET_SET(ifp->if_vnet);
2193 
2194 	/* Notify that the link state has changed. */
2195 	rt_ifmsg(ifp);
2196 	if (ifp->if_vlantrunk != NULL)
2197 		(*vlan_link_state_p)(ifp);
2198 
2199 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2200 	    ifp->if_l2com != NULL)
2201 		(*ng_ether_link_state_p)(ifp, link_state);
2202 	if (ifp->if_carp)
2203 		(*carp_linkstate_p)(ifp);
2204 	if (ifp->if_bridge)
2205 		(*bridge_linkstate_p)(ifp);
2206 	if (ifp->if_lagg)
2207 		(*lagg_linkstate_p)(ifp, link_state);
2208 
2209 	if (IS_DEFAULT_VNET(curvnet))
2210 		devctl_notify("IFNET", ifp->if_xname,
2211 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2212 		    NULL);
2213 	if (pending > 1)
2214 		if_printf(ifp, "%d link states coalesced\n", pending);
2215 	if (log_link_state_change)
2216 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2217 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2218 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2219 	CURVNET_RESTORE();
2220 }
2221 
2222 /*
2223  * Mark an interface down and notify protocols of
2224  * the transition.
2225  */
2226 void
2227 if_down(struct ifnet *ifp)
2228 {
2229 
2230 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2231 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2232 }
2233 
2234 /*
2235  * Mark an interface up and notify protocols of
2236  * the transition.
2237  */
2238 void
2239 if_up(struct ifnet *ifp)
2240 {
2241 
2242 	if_route(ifp, IFF_UP, AF_UNSPEC);
2243 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2244 }
2245 
2246 /*
2247  * Flush an interface queue.
2248  */
2249 void
2250 if_qflush(struct ifnet *ifp)
2251 {
2252 	struct mbuf *m, *n;
2253 	struct ifaltq *ifq;
2254 
2255 	ifq = &ifp->if_snd;
2256 	IFQ_LOCK(ifq);
2257 #ifdef ALTQ
2258 	if (ALTQ_IS_ENABLED(ifq))
2259 		ALTQ_PURGE(ifq);
2260 #endif
2261 	n = ifq->ifq_head;
2262 	while ((m = n) != NULL) {
2263 		n = m->m_nextpkt;
2264 		m_freem(m);
2265 	}
2266 	ifq->ifq_head = 0;
2267 	ifq->ifq_tail = 0;
2268 	ifq->ifq_len = 0;
2269 	IFQ_UNLOCK(ifq);
2270 }
2271 
2272 /*
2273  * Map interface name to interface structure pointer, with or without
2274  * returning a reference.
2275  */
2276 struct ifnet *
2277 ifunit_ref(const char *name)
2278 {
2279 	struct ifnet *ifp;
2280 
2281 	IFNET_RLOCK_NOSLEEP();
2282 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2283 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2284 		    !(ifp->if_flags & IFF_DYING))
2285 			break;
2286 	}
2287 	if (ifp != NULL)
2288 		if_ref(ifp);
2289 	IFNET_RUNLOCK_NOSLEEP();
2290 	return (ifp);
2291 }
2292 
2293 struct ifnet *
2294 ifunit(const char *name)
2295 {
2296 	struct ifnet *ifp;
2297 
2298 	IFNET_RLOCK_NOSLEEP();
2299 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2300 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2301 			break;
2302 	}
2303 	IFNET_RUNLOCK_NOSLEEP();
2304 	return (ifp);
2305 }
2306 
2307 /*
2308  * Hardware specific interface ioctls.
2309  */
2310 static int
2311 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2312 {
2313 	struct ifreq *ifr;
2314 	int error = 0, do_ifup = 0;
2315 	int new_flags, temp_flags;
2316 	size_t namelen, onamelen;
2317 	size_t descrlen;
2318 	char *descrbuf, *odescrbuf;
2319 	char new_name[IFNAMSIZ];
2320 	struct ifaddr *ifa;
2321 	struct sockaddr_dl *sdl;
2322 
2323 	ifr = (struct ifreq *)data;
2324 	switch (cmd) {
2325 	case SIOCGIFINDEX:
2326 		ifr->ifr_index = ifp->if_index;
2327 		break;
2328 
2329 	case SIOCGIFFLAGS:
2330 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2331 		ifr->ifr_flags = temp_flags & 0xffff;
2332 		ifr->ifr_flagshigh = temp_flags >> 16;
2333 		break;
2334 
2335 	case SIOCGIFCAP:
2336 		ifr->ifr_reqcap = ifp->if_capabilities;
2337 		ifr->ifr_curcap = ifp->if_capenable;
2338 		break;
2339 
2340 #ifdef MAC
2341 	case SIOCGIFMAC:
2342 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2343 		break;
2344 #endif
2345 
2346 	case SIOCGIFMETRIC:
2347 		ifr->ifr_metric = ifp->if_metric;
2348 		break;
2349 
2350 	case SIOCGIFMTU:
2351 		ifr->ifr_mtu = ifp->if_mtu;
2352 		break;
2353 
2354 	case SIOCGIFPHYS:
2355 		/* XXXGL: did this ever worked? */
2356 		ifr->ifr_phys = 0;
2357 		break;
2358 
2359 	case SIOCGIFDESCR:
2360 		error = 0;
2361 		sx_slock(&ifdescr_sx);
2362 		if (ifp->if_description == NULL)
2363 			error = ENOMSG;
2364 		else {
2365 			/* space for terminating nul */
2366 			descrlen = strlen(ifp->if_description) + 1;
2367 			if (ifr->ifr_buffer.length < descrlen)
2368 				ifr->ifr_buffer.buffer = NULL;
2369 			else
2370 				error = copyout(ifp->if_description,
2371 				    ifr->ifr_buffer.buffer, descrlen);
2372 			ifr->ifr_buffer.length = descrlen;
2373 		}
2374 		sx_sunlock(&ifdescr_sx);
2375 		break;
2376 
2377 	case SIOCSIFDESCR:
2378 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2379 		if (error)
2380 			return (error);
2381 
2382 		/*
2383 		 * Copy only (length-1) bytes to make sure that
2384 		 * if_description is always nul terminated.  The
2385 		 * length parameter is supposed to count the
2386 		 * terminating nul in.
2387 		 */
2388 		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2389 			return (ENAMETOOLONG);
2390 		else if (ifr->ifr_buffer.length == 0)
2391 			descrbuf = NULL;
2392 		else {
2393 			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2394 			    M_WAITOK | M_ZERO);
2395 			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2396 			    ifr->ifr_buffer.length - 1);
2397 			if (error) {
2398 				free(descrbuf, M_IFDESCR);
2399 				break;
2400 			}
2401 		}
2402 
2403 		sx_xlock(&ifdescr_sx);
2404 		odescrbuf = ifp->if_description;
2405 		ifp->if_description = descrbuf;
2406 		sx_xunlock(&ifdescr_sx);
2407 
2408 		getmicrotime(&ifp->if_lastchange);
2409 		free(odescrbuf, M_IFDESCR);
2410 		break;
2411 
2412 	case SIOCGIFFIB:
2413 		ifr->ifr_fib = ifp->if_fib;
2414 		break;
2415 
2416 	case SIOCSIFFIB:
2417 		error = priv_check(td, PRIV_NET_SETIFFIB);
2418 		if (error)
2419 			return (error);
2420 		if (ifr->ifr_fib >= rt_numfibs)
2421 			return (EINVAL);
2422 
2423 		ifp->if_fib = ifr->ifr_fib;
2424 		break;
2425 
2426 	case SIOCSIFFLAGS:
2427 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2428 		if (error)
2429 			return (error);
2430 		/*
2431 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2432 		 * check, so we don't need special handling here yet.
2433 		 */
2434 		new_flags = (ifr->ifr_flags & 0xffff) |
2435 		    (ifr->ifr_flagshigh << 16);
2436 		if (ifp->if_flags & IFF_UP &&
2437 		    (new_flags & IFF_UP) == 0) {
2438 			if_down(ifp);
2439 		} else if (new_flags & IFF_UP &&
2440 		    (ifp->if_flags & IFF_UP) == 0) {
2441 			do_ifup = 1;
2442 		}
2443 		/* See if permanently promiscuous mode bit is about to flip */
2444 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2445 			if (new_flags & IFF_PPROMISC)
2446 				ifp->if_flags |= IFF_PROMISC;
2447 			else if (ifp->if_pcount == 0)
2448 				ifp->if_flags &= ~IFF_PROMISC;
2449 			if (log_promisc_mode_change)
2450                                 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2451                                     ifp->if_xname,
2452                                     ((new_flags & IFF_PPROMISC) ?
2453                                      "enabled" : "disabled"));
2454 		}
2455 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2456 			(new_flags &~ IFF_CANTCHANGE);
2457 		if (ifp->if_ioctl) {
2458 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2459 		}
2460 		if (do_ifup)
2461 			if_up(ifp);
2462 		getmicrotime(&ifp->if_lastchange);
2463 		break;
2464 
2465 	case SIOCSIFCAP:
2466 		error = priv_check(td, PRIV_NET_SETIFCAP);
2467 		if (error)
2468 			return (error);
2469 		if (ifp->if_ioctl == NULL)
2470 			return (EOPNOTSUPP);
2471 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2472 			return (EINVAL);
2473 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2474 		if (error == 0)
2475 			getmicrotime(&ifp->if_lastchange);
2476 		break;
2477 
2478 #ifdef MAC
2479 	case SIOCSIFMAC:
2480 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2481 		break;
2482 #endif
2483 
2484 	case SIOCSIFNAME:
2485 		error = priv_check(td, PRIV_NET_SETIFNAME);
2486 		if (error)
2487 			return (error);
2488 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2489 		if (error != 0)
2490 			return (error);
2491 		if (new_name[0] == '\0')
2492 			return (EINVAL);
2493 		if (new_name[IFNAMSIZ-1] != '\0') {
2494 			new_name[IFNAMSIZ-1] = '\0';
2495 			if (strlen(new_name) == IFNAMSIZ-1)
2496 				return (EINVAL);
2497 		}
2498 		if (ifunit(new_name) != NULL)
2499 			return (EEXIST);
2500 
2501 		/*
2502 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2503 		 * and there are numerous other races with the
2504 		 * ifunit() checks not being atomic with namespace
2505 		 * changes (renames, vmoves, if_attach, etc).
2506 		 */
2507 		ifp->if_flags |= IFF_RENAMING;
2508 
2509 		/* Announce the departure of the interface. */
2510 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2511 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2512 
2513 		log(LOG_INFO, "%s: changing name to '%s'\n",
2514 		    ifp->if_xname, new_name);
2515 
2516 		IF_ADDR_WLOCK(ifp);
2517 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2518 		ifa = ifp->if_addr;
2519 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2520 		namelen = strlen(new_name);
2521 		onamelen = sdl->sdl_nlen;
2522 		/*
2523 		 * Move the address if needed.  This is safe because we
2524 		 * allocate space for a name of length IFNAMSIZ when we
2525 		 * create this in if_attach().
2526 		 */
2527 		if (namelen != onamelen) {
2528 			bcopy(sdl->sdl_data + onamelen,
2529 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2530 		}
2531 		bcopy(new_name, sdl->sdl_data, namelen);
2532 		sdl->sdl_nlen = namelen;
2533 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2534 		bzero(sdl->sdl_data, onamelen);
2535 		while (namelen != 0)
2536 			sdl->sdl_data[--namelen] = 0xff;
2537 		IF_ADDR_WUNLOCK(ifp);
2538 
2539 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2540 		/* Announce the return of the interface. */
2541 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2542 
2543 		ifp->if_flags &= ~IFF_RENAMING;
2544 		break;
2545 
2546 #ifdef VIMAGE
2547 	case SIOCSIFVNET:
2548 		error = priv_check(td, PRIV_NET_SETIFVNET);
2549 		if (error)
2550 			return (error);
2551 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2552 		break;
2553 #endif
2554 
2555 	case SIOCSIFMETRIC:
2556 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2557 		if (error)
2558 			return (error);
2559 		ifp->if_metric = ifr->ifr_metric;
2560 		getmicrotime(&ifp->if_lastchange);
2561 		break;
2562 
2563 	case SIOCSIFPHYS:
2564 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2565 		if (error)
2566 			return (error);
2567 		if (ifp->if_ioctl == NULL)
2568 			return (EOPNOTSUPP);
2569 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2570 		if (error == 0)
2571 			getmicrotime(&ifp->if_lastchange);
2572 		break;
2573 
2574 	case SIOCSIFMTU:
2575 	{
2576 		u_long oldmtu = ifp->if_mtu;
2577 
2578 		error = priv_check(td, PRIV_NET_SETIFMTU);
2579 		if (error)
2580 			return (error);
2581 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2582 			return (EINVAL);
2583 		if (ifp->if_ioctl == NULL)
2584 			return (EOPNOTSUPP);
2585 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2586 		if (error == 0) {
2587 			getmicrotime(&ifp->if_lastchange);
2588 			rt_ifmsg(ifp);
2589 		}
2590 		/*
2591 		 * If the link MTU changed, do network layer specific procedure.
2592 		 */
2593 		if (ifp->if_mtu != oldmtu) {
2594 #ifdef INET6
2595 			nd6_setmtu(ifp);
2596 #endif
2597 			rt_updatemtu(ifp);
2598 		}
2599 		break;
2600 	}
2601 
2602 	case SIOCADDMULTI:
2603 	case SIOCDELMULTI:
2604 		if (cmd == SIOCADDMULTI)
2605 			error = priv_check(td, PRIV_NET_ADDMULTI);
2606 		else
2607 			error = priv_check(td, PRIV_NET_DELMULTI);
2608 		if (error)
2609 			return (error);
2610 
2611 		/* Don't allow group membership on non-multicast interfaces. */
2612 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2613 			return (EOPNOTSUPP);
2614 
2615 		/* Don't let users screw up protocols' entries. */
2616 		if (ifr->ifr_addr.sa_family != AF_LINK)
2617 			return (EINVAL);
2618 
2619 		if (cmd == SIOCADDMULTI) {
2620 			struct ifmultiaddr *ifma;
2621 
2622 			/*
2623 			 * Userland is only permitted to join groups once
2624 			 * via the if_addmulti() KPI, because it cannot hold
2625 			 * struct ifmultiaddr * between calls. It may also
2626 			 * lose a race while we check if the membership
2627 			 * already exists.
2628 			 */
2629 			IF_ADDR_RLOCK(ifp);
2630 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2631 			IF_ADDR_RUNLOCK(ifp);
2632 			if (ifma != NULL)
2633 				error = EADDRINUSE;
2634 			else
2635 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2636 		} else {
2637 			error = if_delmulti(ifp, &ifr->ifr_addr);
2638 		}
2639 		if (error == 0)
2640 			getmicrotime(&ifp->if_lastchange);
2641 		break;
2642 
2643 	case SIOCSIFPHYADDR:
2644 	case SIOCDIFPHYADDR:
2645 #ifdef INET6
2646 	case SIOCSIFPHYADDR_IN6:
2647 #endif
2648 	case SIOCSIFMEDIA:
2649 	case SIOCSIFGENERIC:
2650 		error = priv_check(td, PRIV_NET_HWIOCTL);
2651 		if (error)
2652 			return (error);
2653 		if (ifp->if_ioctl == NULL)
2654 			return (EOPNOTSUPP);
2655 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2656 		if (error == 0)
2657 			getmicrotime(&ifp->if_lastchange);
2658 		break;
2659 
2660 	case SIOCGIFSTATUS:
2661 	case SIOCGIFPSRCADDR:
2662 	case SIOCGIFPDSTADDR:
2663 	case SIOCGIFMEDIA:
2664 	case SIOCGIFXMEDIA:
2665 	case SIOCGIFGENERIC:
2666 	case SIOCGIFRSSKEY:
2667 	case SIOCGIFRSSHASH:
2668 		if (ifp->if_ioctl == NULL)
2669 			return (EOPNOTSUPP);
2670 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2671 		break;
2672 
2673 	case SIOCSIFLLADDR:
2674 		error = priv_check(td, PRIV_NET_SETLLADDR);
2675 		if (error)
2676 			return (error);
2677 		error = if_setlladdr(ifp,
2678 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2679 		break;
2680 
2681 	case SIOCGHWADDR:
2682 		error = if_gethwaddr(ifp, ifr);
2683 		break;
2684 
2685 	case SIOCAIFGROUP:
2686 	{
2687 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2688 
2689 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2690 		if (error)
2691 			return (error);
2692 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2693 			return (error);
2694 		break;
2695 	}
2696 
2697 	case SIOCGIFGROUP:
2698 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2699 			return (error);
2700 		break;
2701 
2702 	case SIOCDIFGROUP:
2703 	{
2704 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2705 
2706 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2707 		if (error)
2708 			return (error);
2709 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2710 			return (error);
2711 		break;
2712 	}
2713 
2714 	default:
2715 		error = ENOIOCTL;
2716 		break;
2717 	}
2718 	return (error);
2719 }
2720 
2721 #ifdef COMPAT_FREEBSD32
2722 struct ifconf32 {
2723 	int32_t	ifc_len;
2724 	union {
2725 		uint32_t	ifcu_buf;
2726 		uint32_t	ifcu_req;
2727 	} ifc_ifcu;
2728 };
2729 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2730 #endif
2731 
2732 /*
2733  * Interface ioctls.
2734  */
2735 int
2736 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2737 {
2738 	struct ifnet *ifp;
2739 	struct ifreq *ifr;
2740 	int error;
2741 	int oif_flags;
2742 #ifdef VIMAGE
2743 	int shutdown;
2744 #endif
2745 
2746 	CURVNET_SET(so->so_vnet);
2747 #ifdef VIMAGE
2748 	/* Make sure the VNET is stable. */
2749 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2750 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2751 	if (shutdown) {
2752 		CURVNET_RESTORE();
2753 		return (EBUSY);
2754 	}
2755 #endif
2756 
2757 
2758 	switch (cmd) {
2759 	case SIOCGIFCONF:
2760 		error = ifconf(cmd, data);
2761 		CURVNET_RESTORE();
2762 		return (error);
2763 
2764 #ifdef COMPAT_FREEBSD32
2765 	case SIOCGIFCONF32:
2766 		{
2767 			struct ifconf32 *ifc32;
2768 			struct ifconf ifc;
2769 
2770 			ifc32 = (struct ifconf32 *)data;
2771 			ifc.ifc_len = ifc32->ifc_len;
2772 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2773 
2774 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2775 			CURVNET_RESTORE();
2776 			if (error == 0)
2777 				ifc32->ifc_len = ifc.ifc_len;
2778 			return (error);
2779 		}
2780 #endif
2781 	}
2782 	ifr = (struct ifreq *)data;
2783 
2784 	switch (cmd) {
2785 #ifdef VIMAGE
2786 	case SIOCSIFRVNET:
2787 		error = priv_check(td, PRIV_NET_SETIFVNET);
2788 		if (error == 0)
2789 			error = if_vmove_reclaim(td, ifr->ifr_name,
2790 			    ifr->ifr_jid);
2791 		CURVNET_RESTORE();
2792 		return (error);
2793 #endif
2794 	case SIOCIFCREATE:
2795 	case SIOCIFCREATE2:
2796 		error = priv_check(td, PRIV_NET_IFCREATE);
2797 		if (error == 0)
2798 			error = if_clone_create(ifr->ifr_name,
2799 			    sizeof(ifr->ifr_name),
2800 			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2801 		CURVNET_RESTORE();
2802 		return (error);
2803 	case SIOCIFDESTROY:
2804 		error = priv_check(td, PRIV_NET_IFDESTROY);
2805 		if (error == 0)
2806 			error = if_clone_destroy(ifr->ifr_name);
2807 		CURVNET_RESTORE();
2808 		return (error);
2809 
2810 	case SIOCIFGCLONERS:
2811 		error = if_clone_list((struct if_clonereq *)data);
2812 		CURVNET_RESTORE();
2813 		return (error);
2814 	case SIOCGIFGMEMB:
2815 		error = if_getgroupmembers((struct ifgroupreq *)data);
2816 		CURVNET_RESTORE();
2817 		return (error);
2818 #if defined(INET) || defined(INET6)
2819 	case SIOCSVH:
2820 	case SIOCGVH:
2821 		if (carp_ioctl_p == NULL)
2822 			error = EPROTONOSUPPORT;
2823 		else
2824 			error = (*carp_ioctl_p)(ifr, cmd, td);
2825 		CURVNET_RESTORE();
2826 		return (error);
2827 #endif
2828 	}
2829 
2830 	ifp = ifunit_ref(ifr->ifr_name);
2831 	if (ifp == NULL) {
2832 		CURVNET_RESTORE();
2833 		return (ENXIO);
2834 	}
2835 
2836 	error = ifhwioctl(cmd, ifp, data, td);
2837 	if (error != ENOIOCTL) {
2838 		if_rele(ifp);
2839 		CURVNET_RESTORE();
2840 		return (error);
2841 	}
2842 
2843 	oif_flags = ifp->if_flags;
2844 	if (so->so_proto == NULL) {
2845 		if_rele(ifp);
2846 		CURVNET_RESTORE();
2847 		return (EOPNOTSUPP);
2848 	}
2849 
2850 	/*
2851 	 * Pass the request on to the socket control method, and if the
2852 	 * latter returns EOPNOTSUPP, directly to the interface.
2853 	 *
2854 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2855 	 * trust SIOCSIFADDR et al to come from an already privileged
2856 	 * layer, and do not perform any credentials checks or input
2857 	 * validation.
2858 	 */
2859 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2860 	    ifp, td));
2861 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2862 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2863 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2864 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2865 
2866 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2867 #ifdef INET6
2868 		if (ifp->if_flags & IFF_UP)
2869 			in6_if_up(ifp);
2870 #endif
2871 	}
2872 	if_rele(ifp);
2873 	CURVNET_RESTORE();
2874 	return (error);
2875 }
2876 
2877 /*
2878  * The code common to handling reference counted flags,
2879  * e.g., in ifpromisc() and if_allmulti().
2880  * The "pflag" argument can specify a permanent mode flag to check,
2881  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2882  *
2883  * Only to be used on stack-owned flags, not driver-owned flags.
2884  */
2885 static int
2886 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2887 {
2888 	struct ifreq ifr;
2889 	int error;
2890 	int oldflags, oldcount;
2891 
2892 	/* Sanity checks to catch programming errors */
2893 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2894 	    ("%s: setting driver-owned flag %d", __func__, flag));
2895 
2896 	if (onswitch)
2897 		KASSERT(*refcount >= 0,
2898 		    ("%s: increment negative refcount %d for flag %d",
2899 		    __func__, *refcount, flag));
2900 	else
2901 		KASSERT(*refcount > 0,
2902 		    ("%s: decrement non-positive refcount %d for flag %d",
2903 		    __func__, *refcount, flag));
2904 
2905 	/* In case this mode is permanent, just touch refcount */
2906 	if (ifp->if_flags & pflag) {
2907 		*refcount += onswitch ? 1 : -1;
2908 		return (0);
2909 	}
2910 
2911 	/* Save ifnet parameters for if_ioctl() may fail */
2912 	oldcount = *refcount;
2913 	oldflags = ifp->if_flags;
2914 
2915 	/*
2916 	 * See if we aren't the only and touching refcount is enough.
2917 	 * Actually toggle interface flag if we are the first or last.
2918 	 */
2919 	if (onswitch) {
2920 		if ((*refcount)++)
2921 			return (0);
2922 		ifp->if_flags |= flag;
2923 	} else {
2924 		if (--(*refcount))
2925 			return (0);
2926 		ifp->if_flags &= ~flag;
2927 	}
2928 
2929 	/* Call down the driver since we've changed interface flags */
2930 	if (ifp->if_ioctl == NULL) {
2931 		error = EOPNOTSUPP;
2932 		goto recover;
2933 	}
2934 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2935 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2936 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2937 	if (error)
2938 		goto recover;
2939 	/* Notify userland that interface flags have changed */
2940 	rt_ifmsg(ifp);
2941 	return (0);
2942 
2943 recover:
2944 	/* Recover after driver error */
2945 	*refcount = oldcount;
2946 	ifp->if_flags = oldflags;
2947 	return (error);
2948 }
2949 
2950 /*
2951  * Set/clear promiscuous mode on interface ifp based on the truth value
2952  * of pswitch.  The calls are reference counted so that only the first
2953  * "on" request actually has an effect, as does the final "off" request.
2954  * Results are undefined if the "off" and "on" requests are not matched.
2955  */
2956 int
2957 ifpromisc(struct ifnet *ifp, int pswitch)
2958 {
2959 	int error;
2960 	int oldflags = ifp->if_flags;
2961 
2962 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2963 			   &ifp->if_pcount, pswitch);
2964 	/* If promiscuous mode status has changed, log a message */
2965 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2966             log_promisc_mode_change)
2967 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2968 		    ifp->if_xname,
2969 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2970 	return (error);
2971 }
2972 
2973 /*
2974  * Return interface configuration
2975  * of system.  List may be used
2976  * in later ioctl's (above) to get
2977  * other information.
2978  */
2979 /*ARGSUSED*/
2980 static int
2981 ifconf(u_long cmd, caddr_t data)
2982 {
2983 	struct ifconf *ifc = (struct ifconf *)data;
2984 	struct ifnet *ifp;
2985 	struct ifaddr *ifa;
2986 	struct ifreq ifr;
2987 	struct sbuf *sb;
2988 	int error, full = 0, valid_len, max_len;
2989 
2990 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2991 	max_len = MAXPHYS - 1;
2992 
2993 	/* Prevent hostile input from being able to crash the system */
2994 	if (ifc->ifc_len <= 0)
2995 		return (EINVAL);
2996 
2997 again:
2998 	if (ifc->ifc_len <= max_len) {
2999 		max_len = ifc->ifc_len;
3000 		full = 1;
3001 	}
3002 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3003 	max_len = 0;
3004 	valid_len = 0;
3005 
3006 	IFNET_RLOCK();
3007 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3008 		int addrs;
3009 
3010 		/*
3011 		 * Zero the ifr_name buffer to make sure we don't
3012 		 * disclose the contents of the stack.
3013 		 */
3014 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
3015 
3016 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3017 		    >= sizeof(ifr.ifr_name)) {
3018 			sbuf_delete(sb);
3019 			IFNET_RUNLOCK();
3020 			return (ENAMETOOLONG);
3021 		}
3022 
3023 		addrs = 0;
3024 		IF_ADDR_RLOCK(ifp);
3025 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3026 			struct sockaddr *sa = ifa->ifa_addr;
3027 
3028 			if (prison_if(curthread->td_ucred, sa) != 0)
3029 				continue;
3030 			addrs++;
3031 			if (sa->sa_len <= sizeof(*sa)) {
3032 				ifr.ifr_addr = *sa;
3033 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3034 				max_len += sizeof(ifr);
3035 			} else {
3036 				sbuf_bcat(sb, &ifr,
3037 				    offsetof(struct ifreq, ifr_addr));
3038 				max_len += offsetof(struct ifreq, ifr_addr);
3039 				sbuf_bcat(sb, sa, sa->sa_len);
3040 				max_len += sa->sa_len;
3041 			}
3042 
3043 			if (sbuf_error(sb) == 0)
3044 				valid_len = sbuf_len(sb);
3045 		}
3046 		IF_ADDR_RUNLOCK(ifp);
3047 		if (addrs == 0) {
3048 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3049 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3050 			max_len += sizeof(ifr);
3051 
3052 			if (sbuf_error(sb) == 0)
3053 				valid_len = sbuf_len(sb);
3054 		}
3055 	}
3056 	IFNET_RUNLOCK();
3057 
3058 	/*
3059 	 * If we didn't allocate enough space (uncommon), try again.  If
3060 	 * we have already allocated as much space as we are allowed,
3061 	 * return what we've got.
3062 	 */
3063 	if (valid_len != max_len && !full) {
3064 		sbuf_delete(sb);
3065 		goto again;
3066 	}
3067 
3068 	ifc->ifc_len = valid_len;
3069 	sbuf_finish(sb);
3070 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3071 	sbuf_delete(sb);
3072 	return (error);
3073 }
3074 
3075 /*
3076  * Just like ifpromisc(), but for all-multicast-reception mode.
3077  */
3078 int
3079 if_allmulti(struct ifnet *ifp, int onswitch)
3080 {
3081 
3082 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3083 }
3084 
3085 struct ifmultiaddr *
3086 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3087 {
3088 	struct ifmultiaddr *ifma;
3089 
3090 	IF_ADDR_LOCK_ASSERT(ifp);
3091 
3092 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3093 		if (sa->sa_family == AF_LINK) {
3094 			if (sa_dl_equal(ifma->ifma_addr, sa))
3095 				break;
3096 		} else {
3097 			if (sa_equal(ifma->ifma_addr, sa))
3098 				break;
3099 		}
3100 	}
3101 
3102 	return ifma;
3103 }
3104 
3105 /*
3106  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3107  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3108  * the ifnet multicast address list here, so the caller must do that and
3109  * other setup work (such as notifying the device driver).  The reference
3110  * count is initialized to 1.
3111  */
3112 static struct ifmultiaddr *
3113 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3114     int mflags)
3115 {
3116 	struct ifmultiaddr *ifma;
3117 	struct sockaddr *dupsa;
3118 
3119 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3120 	    M_ZERO);
3121 	if (ifma == NULL)
3122 		return (NULL);
3123 
3124 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3125 	if (dupsa == NULL) {
3126 		free(ifma, M_IFMADDR);
3127 		return (NULL);
3128 	}
3129 	bcopy(sa, dupsa, sa->sa_len);
3130 	ifma->ifma_addr = dupsa;
3131 
3132 	ifma->ifma_ifp = ifp;
3133 	ifma->ifma_refcount = 1;
3134 	ifma->ifma_protospec = NULL;
3135 
3136 	if (llsa == NULL) {
3137 		ifma->ifma_lladdr = NULL;
3138 		return (ifma);
3139 	}
3140 
3141 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3142 	if (dupsa == NULL) {
3143 		free(ifma->ifma_addr, M_IFMADDR);
3144 		free(ifma, M_IFMADDR);
3145 		return (NULL);
3146 	}
3147 	bcopy(llsa, dupsa, llsa->sa_len);
3148 	ifma->ifma_lladdr = dupsa;
3149 
3150 	return (ifma);
3151 }
3152 
3153 /*
3154  * if_freemulti: free ifmultiaddr structure and possibly attached related
3155  * addresses.  The caller is responsible for implementing reference
3156  * counting, notifying the driver, handling routing messages, and releasing
3157  * any dependent link layer state.
3158  */
3159 static void
3160 if_freemulti(struct ifmultiaddr *ifma)
3161 {
3162 
3163 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3164 	    ifma->ifma_refcount));
3165 
3166 	if (ifma->ifma_lladdr != NULL)
3167 		free(ifma->ifma_lladdr, M_IFMADDR);
3168 	free(ifma->ifma_addr, M_IFMADDR);
3169 	free(ifma, M_IFMADDR);
3170 }
3171 
3172 /*
3173  * Register an additional multicast address with a network interface.
3174  *
3175  * - If the address is already present, bump the reference count on the
3176  *   address and return.
3177  * - If the address is not link-layer, look up a link layer address.
3178  * - Allocate address structures for one or both addresses, and attach to the
3179  *   multicast address list on the interface.  If automatically adding a link
3180  *   layer address, the protocol address will own a reference to the link
3181  *   layer address, to be freed when it is freed.
3182  * - Notify the network device driver of an addition to the multicast address
3183  *   list.
3184  *
3185  * 'sa' points to caller-owned memory with the desired multicast address.
3186  *
3187  * 'retifma' will be used to return a pointer to the resulting multicast
3188  * address reference, if desired.
3189  */
3190 int
3191 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3192     struct ifmultiaddr **retifma)
3193 {
3194 	struct ifmultiaddr *ifma, *ll_ifma;
3195 	struct sockaddr *llsa;
3196 	struct sockaddr_dl sdl;
3197 	int error;
3198 
3199 	/*
3200 	 * If the address is already present, return a new reference to it;
3201 	 * otherwise, allocate storage and set up a new address.
3202 	 */
3203 	IF_ADDR_WLOCK(ifp);
3204 	ifma = if_findmulti(ifp, sa);
3205 	if (ifma != NULL) {
3206 		ifma->ifma_refcount++;
3207 		if (retifma != NULL)
3208 			*retifma = ifma;
3209 		IF_ADDR_WUNLOCK(ifp);
3210 		return (0);
3211 	}
3212 
3213 	/*
3214 	 * The address isn't already present; resolve the protocol address
3215 	 * into a link layer address, and then look that up, bump its
3216 	 * refcount or allocate an ifma for that also.
3217 	 * Most link layer resolving functions returns address data which
3218 	 * fits inside default sockaddr_dl structure. However callback
3219 	 * can allocate another sockaddr structure, in that case we need to
3220 	 * free it later.
3221 	 */
3222 	llsa = NULL;
3223 	ll_ifma = NULL;
3224 	if (ifp->if_resolvemulti != NULL) {
3225 		/* Provide called function with buffer size information */
3226 		sdl.sdl_len = sizeof(sdl);
3227 		llsa = (struct sockaddr *)&sdl;
3228 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3229 		if (error)
3230 			goto unlock_out;
3231 	}
3232 
3233 	/*
3234 	 * Allocate the new address.  Don't hook it up yet, as we may also
3235 	 * need to allocate a link layer multicast address.
3236 	 */
3237 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3238 	if (ifma == NULL) {
3239 		error = ENOMEM;
3240 		goto free_llsa_out;
3241 	}
3242 
3243 	/*
3244 	 * If a link layer address is found, we'll need to see if it's
3245 	 * already present in the address list, or allocate is as well.
3246 	 * When this block finishes, the link layer address will be on the
3247 	 * list.
3248 	 */
3249 	if (llsa != NULL) {
3250 		ll_ifma = if_findmulti(ifp, llsa);
3251 		if (ll_ifma == NULL) {
3252 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3253 			if (ll_ifma == NULL) {
3254 				--ifma->ifma_refcount;
3255 				if_freemulti(ifma);
3256 				error = ENOMEM;
3257 				goto free_llsa_out;
3258 			}
3259 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3260 			    ifma_link);
3261 		} else
3262 			ll_ifma->ifma_refcount++;
3263 		ifma->ifma_llifma = ll_ifma;
3264 	}
3265 
3266 	/*
3267 	 * We now have a new multicast address, ifma, and possibly a new or
3268 	 * referenced link layer address.  Add the primary address to the
3269 	 * ifnet address list.
3270 	 */
3271 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3272 
3273 	if (retifma != NULL)
3274 		*retifma = ifma;
3275 
3276 	/*
3277 	 * Must generate the message while holding the lock so that 'ifma'
3278 	 * pointer is still valid.
3279 	 */
3280 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3281 	IF_ADDR_WUNLOCK(ifp);
3282 
3283 	/*
3284 	 * We are certain we have added something, so call down to the
3285 	 * interface to let them know about it.
3286 	 */
3287 	if (ifp->if_ioctl != NULL) {
3288 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3289 	}
3290 
3291 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3292 		link_free_sdl(llsa);
3293 
3294 	return (0);
3295 
3296 free_llsa_out:
3297 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3298 		link_free_sdl(llsa);
3299 
3300 unlock_out:
3301 	IF_ADDR_WUNLOCK(ifp);
3302 	return (error);
3303 }
3304 
3305 /*
3306  * Delete a multicast group membership by network-layer group address.
3307  *
3308  * Returns ENOENT if the entry could not be found. If ifp no longer
3309  * exists, results are undefined. This entry point should only be used
3310  * from subsystems which do appropriate locking to hold ifp for the
3311  * duration of the call.
3312  * Network-layer protocol domains must use if_delmulti_ifma().
3313  */
3314 int
3315 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3316 {
3317 	struct ifmultiaddr *ifma;
3318 	int lastref;
3319 #ifdef INVARIANTS
3320 	struct ifnet *oifp;
3321 
3322 	IFNET_RLOCK_NOSLEEP();
3323 	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3324 		if (ifp == oifp)
3325 			break;
3326 	if (ifp != oifp)
3327 		ifp = NULL;
3328 	IFNET_RUNLOCK_NOSLEEP();
3329 
3330 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3331 #endif
3332 	if (ifp == NULL)
3333 		return (ENOENT);
3334 
3335 	IF_ADDR_WLOCK(ifp);
3336 	lastref = 0;
3337 	ifma = if_findmulti(ifp, sa);
3338 	if (ifma != NULL)
3339 		lastref = if_delmulti_locked(ifp, ifma, 0);
3340 	IF_ADDR_WUNLOCK(ifp);
3341 
3342 	if (ifma == NULL)
3343 		return (ENOENT);
3344 
3345 	if (lastref && ifp->if_ioctl != NULL) {
3346 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3347 	}
3348 
3349 	return (0);
3350 }
3351 
3352 /*
3353  * Delete all multicast group membership for an interface.
3354  * Should be used to quickly flush all multicast filters.
3355  */
3356 void
3357 if_delallmulti(struct ifnet *ifp)
3358 {
3359 	struct ifmultiaddr *ifma;
3360 	struct ifmultiaddr *next;
3361 
3362 	IF_ADDR_WLOCK(ifp);
3363 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3364 		if_delmulti_locked(ifp, ifma, 0);
3365 	IF_ADDR_WUNLOCK(ifp);
3366 }
3367 
3368 /*
3369  * Delete a multicast group membership by group membership pointer.
3370  * Network-layer protocol domains must use this routine.
3371  *
3372  * It is safe to call this routine if the ifp disappeared.
3373  */
3374 void
3375 if_delmulti_ifma(struct ifmultiaddr *ifma)
3376 {
3377 	struct ifnet *ifp;
3378 	int lastref;
3379 
3380 	ifp = ifma->ifma_ifp;
3381 #ifdef DIAGNOSTIC
3382 	if (ifp == NULL) {
3383 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3384 	} else {
3385 		struct ifnet *oifp;
3386 
3387 		IFNET_RLOCK_NOSLEEP();
3388 		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3389 			if (ifp == oifp)
3390 				break;
3391 		if (ifp != oifp) {
3392 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3393 			ifp = NULL;
3394 		}
3395 		IFNET_RUNLOCK_NOSLEEP();
3396 	}
3397 #endif
3398 	/*
3399 	 * If and only if the ifnet instance exists: Acquire the address lock.
3400 	 */
3401 	if (ifp != NULL)
3402 		IF_ADDR_WLOCK(ifp);
3403 
3404 	lastref = if_delmulti_locked(ifp, ifma, 0);
3405 
3406 	if (ifp != NULL) {
3407 		/*
3408 		 * If and only if the ifnet instance exists:
3409 		 *  Release the address lock.
3410 		 *  If the group was left: update the hardware hash filter.
3411 		 */
3412 		IF_ADDR_WUNLOCK(ifp);
3413 		if (lastref && ifp->if_ioctl != NULL) {
3414 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3415 		}
3416 	}
3417 }
3418 
3419 /*
3420  * Perform deletion of network-layer and/or link-layer multicast address.
3421  *
3422  * Return 0 if the reference count was decremented.
3423  * Return 1 if the final reference was released, indicating that the
3424  * hardware hash filter should be reprogrammed.
3425  */
3426 static int
3427 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3428 {
3429 	struct ifmultiaddr *ll_ifma;
3430 
3431 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3432 		KASSERT(ifma->ifma_ifp == ifp,
3433 		    ("%s: inconsistent ifp %p", __func__, ifp));
3434 		IF_ADDR_WLOCK_ASSERT(ifp);
3435 	}
3436 
3437 	ifp = ifma->ifma_ifp;
3438 
3439 	/*
3440 	 * If the ifnet is detaching, null out references to ifnet,
3441 	 * so that upper protocol layers will notice, and not attempt
3442 	 * to obtain locks for an ifnet which no longer exists. The
3443 	 * routing socket announcement must happen before the ifnet
3444 	 * instance is detached from the system.
3445 	 */
3446 	if (detaching) {
3447 #ifdef DIAGNOSTIC
3448 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3449 #endif
3450 		/*
3451 		 * ifp may already be nulled out if we are being reentered
3452 		 * to delete the ll_ifma.
3453 		 */
3454 		if (ifp != NULL) {
3455 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3456 			ifma->ifma_ifp = NULL;
3457 		}
3458 	}
3459 
3460 	if (--ifma->ifma_refcount > 0)
3461 		return 0;
3462 
3463 	/*
3464 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3465 	 * have been associated with it. Release it first if so.
3466 	 */
3467 	ll_ifma = ifma->ifma_llifma;
3468 	if (ll_ifma != NULL) {
3469 		KASSERT(ifma->ifma_lladdr != NULL,
3470 		    ("%s: llifma w/o lladdr", __func__));
3471 		if (detaching)
3472 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3473 		if (--ll_ifma->ifma_refcount == 0) {
3474 			if (ifp != NULL) {
3475 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3476 				    ifma_link);
3477 			}
3478 			if_freemulti(ll_ifma);
3479 		}
3480 	}
3481 
3482 	if (ifp != NULL)
3483 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3484 
3485 	if_freemulti(ifma);
3486 
3487 	/*
3488 	 * The last reference to this instance of struct ifmultiaddr
3489 	 * was released; the hardware should be notified of this change.
3490 	 */
3491 	return 1;
3492 }
3493 
3494 /*
3495  * Set the link layer address on an interface.
3496  *
3497  * At this time we only support certain types of interfaces,
3498  * and we don't allow the length of the address to change.
3499  *
3500  * Set noinline to be dtrace-friendly
3501  */
3502 __noinline int
3503 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3504 {
3505 	struct sockaddr_dl *sdl;
3506 	struct ifaddr *ifa;
3507 	struct ifreq ifr;
3508 
3509 	IF_ADDR_RLOCK(ifp);
3510 	ifa = ifp->if_addr;
3511 	if (ifa == NULL) {
3512 		IF_ADDR_RUNLOCK(ifp);
3513 		return (EINVAL);
3514 	}
3515 	ifa_ref(ifa);
3516 	IF_ADDR_RUNLOCK(ifp);
3517 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3518 	if (sdl == NULL) {
3519 		ifa_free(ifa);
3520 		return (EINVAL);
3521 	}
3522 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3523 		ifa_free(ifa);
3524 		return (EINVAL);
3525 	}
3526 	switch (ifp->if_type) {
3527 	case IFT_ETHER:
3528 	case IFT_FDDI:
3529 	case IFT_XETHER:
3530 	case IFT_ISO88025:
3531 	case IFT_L2VLAN:
3532 	case IFT_BRIDGE:
3533 	case IFT_ARCNET:
3534 	case IFT_IEEE8023ADLAG:
3535 		bcopy(lladdr, LLADDR(sdl), len);
3536 		ifa_free(ifa);
3537 		break;
3538 	default:
3539 		ifa_free(ifa);
3540 		return (ENODEV);
3541 	}
3542 
3543 	/*
3544 	 * If the interface is already up, we need
3545 	 * to re-init it in order to reprogram its
3546 	 * address filter.
3547 	 */
3548 	if ((ifp->if_flags & IFF_UP) != 0) {
3549 		if (ifp->if_ioctl) {
3550 			ifp->if_flags &= ~IFF_UP;
3551 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3552 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3553 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3554 			ifp->if_flags |= IFF_UP;
3555 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3556 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3557 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3558 		}
3559 	}
3560 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3561 	return (0);
3562 }
3563 
3564 /*
3565  * Compat function for handling basic encapsulation requests.
3566  * Not converted stacks (FDDI, IB, ..) supports traditional
3567  * output model: ARP (and other similar L2 protocols) are handled
3568  * inside output routine, arpresolve/nd6_resolve() returns MAC
3569  * address instead of full prepend.
3570  *
3571  * This function creates calculated header==MAC for IPv4/IPv6 and
3572  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3573  * address families.
3574  */
3575 static int
3576 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3577 {
3578 
3579 	if (req->rtype != IFENCAP_LL)
3580 		return (EOPNOTSUPP);
3581 
3582 	if (req->bufsize < req->lladdr_len)
3583 		return (ENOMEM);
3584 
3585 	switch (req->family) {
3586 	case AF_INET:
3587 	case AF_INET6:
3588 		break;
3589 	default:
3590 		return (EAFNOSUPPORT);
3591 	}
3592 
3593 	/* Copy lladdr to storage as is */
3594 	memmove(req->buf, req->lladdr, req->lladdr_len);
3595 	req->bufsize = req->lladdr_len;
3596 	req->lladdr_off = 0;
3597 
3598 	return (0);
3599 }
3600 
3601 /*
3602  * Get the link layer address that was read from the hardware at attach.
3603  *
3604  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3605  * their component interfaces as IFT_IEEE8023ADLAG.
3606  */
3607 int
3608 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3609 {
3610 
3611 	if (ifp->if_hw_addr == NULL)
3612 		return (ENODEV);
3613 
3614 	switch (ifp->if_type) {
3615 	case IFT_ETHER:
3616 	case IFT_IEEE8023ADLAG:
3617 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3618 		return (0);
3619 	default:
3620 		return (ENODEV);
3621 	}
3622 }
3623 
3624 /*
3625  * The name argument must be a pointer to storage which will last as
3626  * long as the interface does.  For physical devices, the result of
3627  * device_get_name(dev) is a good choice and for pseudo-devices a
3628  * static string works well.
3629  */
3630 void
3631 if_initname(struct ifnet *ifp, const char *name, int unit)
3632 {
3633 	ifp->if_dname = name;
3634 	ifp->if_dunit = unit;
3635 	if (unit != IF_DUNIT_NONE)
3636 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3637 	else
3638 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3639 }
3640 
3641 int
3642 if_printf(struct ifnet *ifp, const char * fmt, ...)
3643 {
3644 	va_list ap;
3645 	int retval;
3646 
3647 	retval = printf("%s: ", ifp->if_xname);
3648 	va_start(ap, fmt);
3649 	retval += vprintf(fmt, ap);
3650 	va_end(ap);
3651 	return (retval);
3652 }
3653 
3654 void
3655 if_start(struct ifnet *ifp)
3656 {
3657 
3658 	(*(ifp)->if_start)(ifp);
3659 }
3660 
3661 /*
3662  * Backwards compatibility interface for drivers
3663  * that have not implemented it
3664  */
3665 static int
3666 if_transmit(struct ifnet *ifp, struct mbuf *m)
3667 {
3668 	int error;
3669 
3670 	IFQ_HANDOFF(ifp, m, error);
3671 	return (error);
3672 }
3673 
3674 static void
3675 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3676 {
3677 
3678 	m_freem(m);
3679 }
3680 
3681 int
3682 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3683 {
3684 	int active = 0;
3685 
3686 	IF_LOCK(ifq);
3687 	if (_IF_QFULL(ifq)) {
3688 		IF_UNLOCK(ifq);
3689 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3690 		m_freem(m);
3691 		return (0);
3692 	}
3693 	if (ifp != NULL) {
3694 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3695 		if (m->m_flags & (M_BCAST|M_MCAST))
3696 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3697 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3698 	}
3699 	_IF_ENQUEUE(ifq, m);
3700 	IF_UNLOCK(ifq);
3701 	if (ifp != NULL && !active)
3702 		(*(ifp)->if_start)(ifp);
3703 	return (1);
3704 }
3705 
3706 void
3707 if_register_com_alloc(u_char type,
3708     if_com_alloc_t *a, if_com_free_t *f)
3709 {
3710 
3711 	KASSERT(if_com_alloc[type] == NULL,
3712 	    ("if_register_com_alloc: %d already registered", type));
3713 	KASSERT(if_com_free[type] == NULL,
3714 	    ("if_register_com_alloc: %d free already registered", type));
3715 
3716 	if_com_alloc[type] = a;
3717 	if_com_free[type] = f;
3718 }
3719 
3720 void
3721 if_deregister_com_alloc(u_char type)
3722 {
3723 
3724 	KASSERT(if_com_alloc[type] != NULL,
3725 	    ("if_deregister_com_alloc: %d not registered", type));
3726 	KASSERT(if_com_free[type] != NULL,
3727 	    ("if_deregister_com_alloc: %d free not registered", type));
3728 	if_com_alloc[type] = NULL;
3729 	if_com_free[type] = NULL;
3730 }
3731 
3732 /* API for driver access to network stack owned ifnet.*/
3733 uint64_t
3734 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3735 {
3736 	uint64_t oldbrate;
3737 
3738 	oldbrate = ifp->if_baudrate;
3739 	ifp->if_baudrate = baudrate;
3740 	return (oldbrate);
3741 }
3742 
3743 uint64_t
3744 if_getbaudrate(if_t ifp)
3745 {
3746 
3747 	return (((struct ifnet *)ifp)->if_baudrate);
3748 }
3749 
3750 int
3751 if_setcapabilities(if_t ifp, int capabilities)
3752 {
3753 	((struct ifnet *)ifp)->if_capabilities = capabilities;
3754 	return (0);
3755 }
3756 
3757 int
3758 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3759 {
3760 	((struct ifnet *)ifp)->if_capabilities |= setbit;
3761 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3762 
3763 	return (0);
3764 }
3765 
3766 int
3767 if_getcapabilities(if_t ifp)
3768 {
3769 	return ((struct ifnet *)ifp)->if_capabilities;
3770 }
3771 
3772 int
3773 if_setcapenable(if_t ifp, int capabilities)
3774 {
3775 	((struct ifnet *)ifp)->if_capenable = capabilities;
3776 	return (0);
3777 }
3778 
3779 int
3780 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3781 {
3782 	if(setcap)
3783 		((struct ifnet *)ifp)->if_capenable |= setcap;
3784 	if(clearcap)
3785 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3786 
3787 	return (0);
3788 }
3789 
3790 const char *
3791 if_getdname(if_t ifp)
3792 {
3793 	return ((struct ifnet *)ifp)->if_dname;
3794 }
3795 
3796 int
3797 if_togglecapenable(if_t ifp, int togglecap)
3798 {
3799 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
3800 	return (0);
3801 }
3802 
3803 int
3804 if_getcapenable(if_t ifp)
3805 {
3806 	return ((struct ifnet *)ifp)->if_capenable;
3807 }
3808 
3809 /*
3810  * This is largely undesirable because it ties ifnet to a device, but does
3811  * provide flexiblity for an embedded product vendor. Should be used with
3812  * the understanding that it violates the interface boundaries, and should be
3813  * a last resort only.
3814  */
3815 int
3816 if_setdev(if_t ifp, void *dev)
3817 {
3818 	return (0);
3819 }
3820 
3821 int
3822 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3823 {
3824 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3825 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3826 
3827 	return (0);
3828 }
3829 
3830 int
3831 if_getdrvflags(if_t ifp)
3832 {
3833 	return ((struct ifnet *)ifp)->if_drv_flags;
3834 }
3835 
3836 int
3837 if_setdrvflags(if_t ifp, int flags)
3838 {
3839 	((struct ifnet *)ifp)->if_drv_flags = flags;
3840 	return (0);
3841 }
3842 
3843 
3844 int
3845 if_setflags(if_t ifp, int flags)
3846 {
3847 	((struct ifnet *)ifp)->if_flags = flags;
3848 	return (0);
3849 }
3850 
3851 int
3852 if_setflagbits(if_t ifp, int set, int clear)
3853 {
3854 	((struct ifnet *)ifp)->if_flags |= set;
3855 	((struct ifnet *)ifp)->if_flags &= ~clear;
3856 
3857 	return (0);
3858 }
3859 
3860 int
3861 if_getflags(if_t ifp)
3862 {
3863 	return ((struct ifnet *)ifp)->if_flags;
3864 }
3865 
3866 int
3867 if_clearhwassist(if_t ifp)
3868 {
3869 	((struct ifnet *)ifp)->if_hwassist = 0;
3870 	return (0);
3871 }
3872 
3873 int
3874 if_sethwassistbits(if_t ifp, int toset, int toclear)
3875 {
3876 	((struct ifnet *)ifp)->if_hwassist |= toset;
3877 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3878 
3879 	return (0);
3880 }
3881 
3882 int
3883 if_sethwassist(if_t ifp, int hwassist_bit)
3884 {
3885 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3886 	return (0);
3887 }
3888 
3889 int
3890 if_gethwassist(if_t ifp)
3891 {
3892 	return ((struct ifnet *)ifp)->if_hwassist;
3893 }
3894 
3895 int
3896 if_setmtu(if_t ifp, int mtu)
3897 {
3898 	((struct ifnet *)ifp)->if_mtu = mtu;
3899 	return (0);
3900 }
3901 
3902 int
3903 if_getmtu(if_t ifp)
3904 {
3905 	return ((struct ifnet *)ifp)->if_mtu;
3906 }
3907 
3908 int
3909 if_getmtu_family(if_t ifp, int family)
3910 {
3911 	struct domain *dp;
3912 
3913 	for (dp = domains; dp; dp = dp->dom_next) {
3914 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3915 			return (dp->dom_ifmtu((struct ifnet *)ifp));
3916 	}
3917 
3918 	return (((struct ifnet *)ifp)->if_mtu);
3919 }
3920 
3921 int
3922 if_setsoftc(if_t ifp, void *softc)
3923 {
3924 	((struct ifnet *)ifp)->if_softc = softc;
3925 	return (0);
3926 }
3927 
3928 void *
3929 if_getsoftc(if_t ifp)
3930 {
3931 	return ((struct ifnet *)ifp)->if_softc;
3932 }
3933 
3934 void
3935 if_setrcvif(struct mbuf *m, if_t ifp)
3936 {
3937 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3938 }
3939 
3940 void
3941 if_setvtag(struct mbuf *m, uint16_t tag)
3942 {
3943 	m->m_pkthdr.ether_vtag = tag;
3944 }
3945 
3946 uint16_t
3947 if_getvtag(struct mbuf *m)
3948 {
3949 
3950 	return (m->m_pkthdr.ether_vtag);
3951 }
3952 
3953 int
3954 if_sendq_empty(if_t ifp)
3955 {
3956 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3957 }
3958 
3959 struct ifaddr *
3960 if_getifaddr(if_t ifp)
3961 {
3962 	return ((struct ifnet *)ifp)->if_addr;
3963 }
3964 
3965 int
3966 if_getamcount(if_t ifp)
3967 {
3968 	return ((struct ifnet *)ifp)->if_amcount;
3969 }
3970 
3971 
3972 int
3973 if_setsendqready(if_t ifp)
3974 {
3975 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3976 	return (0);
3977 }
3978 
3979 int
3980 if_setsendqlen(if_t ifp, int tx_desc_count)
3981 {
3982 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3983 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3984 
3985 	return (0);
3986 }
3987 
3988 int
3989 if_vlantrunkinuse(if_t ifp)
3990 {
3991 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3992 }
3993 
3994 int
3995 if_input(if_t ifp, struct mbuf* sendmp)
3996 {
3997 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3998 	return (0);
3999 
4000 }
4001 
4002 /* XXX */
4003 #ifndef ETH_ADDR_LEN
4004 #define ETH_ADDR_LEN 6
4005 #endif
4006 
4007 int
4008 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4009 {
4010 	struct ifmultiaddr *ifma;
4011 	uint8_t *lmta = (uint8_t *)mta;
4012 	int mcnt = 0;
4013 
4014 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4015 		if (ifma->ifma_addr->sa_family != AF_LINK)
4016 			continue;
4017 
4018 		if (mcnt == max)
4019 			break;
4020 
4021 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4022 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4023 		mcnt++;
4024 	}
4025 	*cnt = mcnt;
4026 
4027 	return (0);
4028 }
4029 
4030 int
4031 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4032 {
4033 	int error;
4034 
4035 	if_maddr_rlock(ifp);
4036 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4037 	if_maddr_runlock(ifp);
4038 	return (error);
4039 }
4040 
4041 int
4042 if_multiaddr_count(if_t ifp, int max)
4043 {
4044 	struct ifmultiaddr *ifma;
4045 	int count;
4046 
4047 	count = 0;
4048 	if_maddr_rlock(ifp);
4049 	TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4050 		if (ifma->ifma_addr->sa_family != AF_LINK)
4051 			continue;
4052 		count++;
4053 		if (count == max)
4054 			break;
4055 	}
4056 	if_maddr_runlock(ifp);
4057 	return (count);
4058 }
4059 
4060 int
4061 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4062 {
4063 	struct ifmultiaddr *ifma;
4064 	int cnt = 0;
4065 
4066 	if_maddr_rlock(ifp);
4067 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4068 		cnt += filter(arg, ifma, cnt);
4069 	if_maddr_runlock(ifp);
4070 	return (cnt);
4071 }
4072 
4073 struct mbuf *
4074 if_dequeue(if_t ifp)
4075 {
4076 	struct mbuf *m;
4077 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4078 
4079 	return (m);
4080 }
4081 
4082 int
4083 if_sendq_prepend(if_t ifp, struct mbuf *m)
4084 {
4085 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4086 	return (0);
4087 }
4088 
4089 int
4090 if_setifheaderlen(if_t ifp, int len)
4091 {
4092 	((struct ifnet *)ifp)->if_hdrlen = len;
4093 	return (0);
4094 }
4095 
4096 caddr_t
4097 if_getlladdr(if_t ifp)
4098 {
4099 	return (IF_LLADDR((struct ifnet *)ifp));
4100 }
4101 
4102 void *
4103 if_gethandle(u_char type)
4104 {
4105 	return (if_alloc(type));
4106 }
4107 
4108 void
4109 if_bpfmtap(if_t ifh, struct mbuf *m)
4110 {
4111 	struct ifnet *ifp = (struct ifnet *)ifh;
4112 
4113 	BPF_MTAP(ifp, m);
4114 }
4115 
4116 void
4117 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4118 {
4119 	struct ifnet *ifp = (struct ifnet *)ifh;
4120 
4121 	ETHER_BPF_MTAP(ifp, m);
4122 }
4123 
4124 void
4125 if_vlancap(if_t ifh)
4126 {
4127 	struct ifnet *ifp = (struct ifnet *)ifh;
4128 	VLAN_CAPABILITIES(ifp);
4129 }
4130 
4131 int
4132 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4133 {
4134 
4135 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4136         return (0);
4137 }
4138 
4139 int
4140 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4141 {
4142 
4143 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4144         return (0);
4145 }
4146 
4147 int
4148 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4149 {
4150 
4151 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4152         return (0);
4153 }
4154 
4155 u_int
4156 if_gethwtsomax(if_t ifp)
4157 {
4158 
4159 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4160 }
4161 
4162 u_int
4163 if_gethwtsomaxsegcount(if_t ifp)
4164 {
4165 
4166 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4167 }
4168 
4169 u_int
4170 if_gethwtsomaxsegsize(if_t ifp)
4171 {
4172 
4173 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4174 }
4175 
4176 void
4177 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4178 {
4179 	((struct ifnet *)ifp)->if_init = init_fn;
4180 }
4181 
4182 void
4183 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4184 {
4185 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4186 }
4187 
4188 void
4189 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4190 {
4191 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4192 }
4193 
4194 void
4195 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4196 {
4197 	((struct ifnet *)ifp)->if_transmit = start_fn;
4198 }
4199 
4200 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4201 {
4202 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4203 
4204 }
4205 
4206 void
4207 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4208 {
4209 
4210 	ifp->if_get_counter = fn;
4211 }
4212 
4213 /* Revisit these - These are inline functions originally. */
4214 int
4215 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4216 {
4217 	return drbr_inuse(ifh, br);
4218 }
4219 
4220 struct mbuf*
4221 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4222 {
4223 	return drbr_dequeue(ifh, br);
4224 }
4225 
4226 int
4227 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4228 {
4229 	return drbr_needs_enqueue(ifh, br);
4230 }
4231 
4232 int
4233 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4234 {
4235 	return drbr_enqueue(ifh, br, m);
4236 
4237 }
4238